From d3a23e1678a5827c38ed8a465ad91d65e59fa911 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Wed, 5 Jan 2011 20:01:21 +1100
Subject: Revert "fs: use RCU read side protection in d_validate"

This reverts commit 3825bdb7ed920845961f32f364454bee5f469abb.

You cannot dget() a dentry without having a reference, or holding
a lock that guarantees it remains valid.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 23702a9..cc2b938 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1491,26 +1491,33 @@ out:
  * This is used by ncpfs in its readdir implementation.
  * Zero is returned in the dentry is invalid.
  */
-int d_validate(struct dentry *dentry, struct dentry *parent)
+ 
+int d_validate(struct dentry *dentry, struct dentry *dparent)
 {
-	struct hlist_head *head = d_hash(parent, dentry->d_name.hash);
-	struct hlist_node *node;
-	struct dentry *d;
+	struct hlist_head *base;
+	struct hlist_node *lhp;
 
 	/* Check whether the ptr might be valid at all.. */
 	if (!kmem_ptr_validate(dentry_cache, dentry))
-		return 0;
-	if (dentry->d_parent != parent)
-		return 0;
+		goto out;
 
-	rcu_read_lock();
-	hlist_for_each_entry_rcu(d, node, head, d_hash) {
-		if (d == dentry) {
-			dget(dentry);
+	if (dentry->d_parent != dparent)
+		goto out;
+
+	spin_lock(&dcache_lock);
+	base = d_hash(dparent, dentry->d_name.hash);
+	hlist_for_each(lhp,base) { 
+		/* hlist_for_each_entry_rcu() not required for d_hash list
+		 * as it is parsed under dcache_lock
+		 */
+		if (dentry == hlist_entry(lhp, struct dentry, d_hash)) {
+			__dget_locked(dentry);
+			spin_unlock(&dcache_lock);
 			return 1;
 		}
 	}
-	rcu_read_unlock();
+	spin_unlock(&dcache_lock);
+out:
 	return 0;
 }
 EXPORT_SYMBOL(d_validate);
-- 
cgit v0.10.2


From 786a5e15b613a9cee4fc9139fc3113a5ab0fde79 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:16 +1100
Subject: fs: d_validate fixes

d_validate has been broken for a long time.

kmem_ptr_validate does not guarantee that a pointer can be dereferenced
if it can go away at any time. Even rcu_read_lock doesn't help, because
the pointer might be queued in RCU callbacks but not executed yet.

So the parent cannot be checked, nor the name hashed. The dentry pointer
can not be touched until it can be verified under lock. Hashing simply
cannot be used.

Instead, verify the parent/child relationship by traversing parent's
d_child list. It's slow, but only ncpfs and the destaged smbfs care
about it, at this point.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index cc2b938..9d1a59df 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1483,41 +1483,30 @@ out:
 }
 
 /**
- * d_validate - verify dentry provided from insecure source
+ * d_validate - verify dentry provided from insecure source (deprecated)
  * @dentry: The dentry alleged to be valid child of @dparent
  * @dparent: The parent dentry (known to be valid)
  *
  * An insecure source has sent us a dentry, here we verify it and dget() it.
  * This is used by ncpfs in its readdir implementation.
  * Zero is returned in the dentry is invalid.
+ *
+ * This function is slow for big directories, and deprecated, do not use it.
  */
- 
 int d_validate(struct dentry *dentry, struct dentry *dparent)
 {
-	struct hlist_head *base;
-	struct hlist_node *lhp;
-
-	/* Check whether the ptr might be valid at all.. */
-	if (!kmem_ptr_validate(dentry_cache, dentry))
-		goto out;
-
-	if (dentry->d_parent != dparent)
-		goto out;
+	struct dentry *child;
 
 	spin_lock(&dcache_lock);
-	base = d_hash(dparent, dentry->d_name.hash);
-	hlist_for_each(lhp,base) { 
-		/* hlist_for_each_entry_rcu() not required for d_hash list
-		 * as it is parsed under dcache_lock
-		 */
-		if (dentry == hlist_entry(lhp, struct dentry, d_hash)) {
+	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
+		if (dentry == child) {
 			__dget_locked(dentry);
 			spin_unlock(&dcache_lock);
 			return 1;
 		}
 	}
 	spin_unlock(&dcache_lock);
-out:
+
 	return 0;
 }
 EXPORT_SYMBOL(d_validate);
-- 
cgit v0.10.2


From ccd35fb9f4da856b105ea0f1e0cab3702e8ae6ba Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:17 +1100
Subject: kernel: kmem_ptr_validate considered harmful

This is a nasty and error prone API. It is no longer used, remove it.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/include/linux/slab.h b/include/linux/slab.h
index 59260e2..fa90866 100644
--- a/include/linux/slab.h
+++ b/include/linux/slab.h
@@ -106,8 +106,6 @@ int kmem_cache_shrink(struct kmem_cache *);
 void kmem_cache_free(struct kmem_cache *, void *);
 unsigned int kmem_cache_size(struct kmem_cache *);
 const char *kmem_cache_name(struct kmem_cache *);
-int kern_ptr_validate(const void *ptr, unsigned long size);
-int kmem_ptr_validate(struct kmem_cache *cachep, const void *ptr);
 
 /*
  * Please use this macro to create slab caches. Simply specify the
diff --git a/mm/slab.c b/mm/slab.c
index b1e40da..6107f23 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -2781,7 +2781,7 @@ static void slab_put_obj(struct kmem_cache *cachep, struct slab *slabp,
 /*
  * Map pages beginning at addr to the given cache and slab. This is required
  * for the slab allocator to be able to lookup the cache and slab of a
- * virtual address for kfree, ksize, kmem_ptr_validate, and slab debugging.
+ * virtual address for kfree, ksize, and slab debugging.
  */
 static void slab_map_pages(struct kmem_cache *cache, struct slab *slab,
 			   void *addr)
@@ -3660,36 +3660,6 @@ void *kmem_cache_alloc_notrace(struct kmem_cache *cachep, gfp_t flags)
 EXPORT_SYMBOL(kmem_cache_alloc_notrace);
 #endif
 
-/**
- * kmem_ptr_validate - check if an untrusted pointer might be a slab entry.
- * @cachep: the cache we're checking against
- * @ptr: pointer to validate
- *
- * This verifies that the untrusted pointer looks sane;
- * it is _not_ a guarantee that the pointer is actually
- * part of the slab cache in question, but it at least
- * validates that the pointer can be dereferenced and
- * looks half-way sane.
- *
- * Currently only used for dentry validation.
- */
-int kmem_ptr_validate(struct kmem_cache *cachep, const void *ptr)
-{
-	unsigned long size = cachep->buffer_size;
-	struct page *page;
-
-	if (unlikely(!kern_ptr_validate(ptr, size)))
-		goto out;
-	page = virt_to_page(ptr);
-	if (unlikely(!PageSlab(page)))
-		goto out;
-	if (unlikely(page_get_cache(page) != cachep))
-		goto out;
-	return 1;
-out:
-	return 0;
-}
-
 #ifdef CONFIG_NUMA
 void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
 {
diff --git a/mm/slob.c b/mm/slob.c
index 617b6d6..3588eaa 100644
--- a/mm/slob.c
+++ b/mm/slob.c
@@ -678,11 +678,6 @@ int kmem_cache_shrink(struct kmem_cache *d)
 }
 EXPORT_SYMBOL(kmem_cache_shrink);
 
-int kmem_ptr_validate(struct kmem_cache *a, const void *b)
-{
-	return 0;
-}
-
 static unsigned int slob_ready __read_mostly;
 
 int slab_is_available(void)
diff --git a/mm/slub.c b/mm/slub.c
index bec0e35..a2fe172 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -1917,17 +1917,6 @@ void kmem_cache_free(struct kmem_cache *s, void *x)
 }
 EXPORT_SYMBOL(kmem_cache_free);
 
-/* Figure out on which slab page the object resides */
-static struct page *get_object_page(const void *x)
-{
-	struct page *page = virt_to_head_page(x);
-
-	if (!PageSlab(page))
-		return NULL;
-
-	return page;
-}
-
 /*
  * Object placement in a slab is made very easy because we always start at
  * offset 0. If we tune the size of the object to the alignment then we can
@@ -2386,35 +2375,6 @@ error:
 }
 
 /*
- * Check if a given pointer is valid
- */
-int kmem_ptr_validate(struct kmem_cache *s, const void *object)
-{
-	struct page *page;
-
-	if (!kern_ptr_validate(object, s->size))
-		return 0;
-
-	page = get_object_page(object);
-
-	if (!page || s != page->slab)
-		/* No slab or wrong slab */
-		return 0;
-
-	if (!check_valid_pointer(s, page, object))
-		return 0;
-
-	/*
-	 * We could also check if the object is on the slabs freelist.
-	 * But this would be too expensive and it seems that the main
-	 * purpose of kmem_ptr_valid() is to check if the object belongs
-	 * to a certain slab.
-	 */
-	return 1;
-}
-EXPORT_SYMBOL(kmem_ptr_validate);
-
-/*
  * Determine the size of a slab object
  */
 unsigned int kmem_cache_size(struct kmem_cache *s)
diff --git a/mm/util.c b/mm/util.c
index 73dac81..f126975 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -186,27 +186,6 @@ void kzfree(const void *p)
 }
 EXPORT_SYMBOL(kzfree);
 
-int kern_ptr_validate(const void *ptr, unsigned long size)
-{
-	unsigned long addr = (unsigned long)ptr;
-	unsigned long min_addr = PAGE_OFFSET;
-	unsigned long align_mask = sizeof(void *) - 1;
-
-	if (unlikely(addr < min_addr))
-		goto out;
-	if (unlikely(addr > (unsigned long)high_memory - size))
-		goto out;
-	if (unlikely(addr & align_mask))
-		goto out;
-	if (unlikely(!kern_addr_valid(addr)))
-		goto out;
-	if (unlikely(!kern_addr_valid(addr + size - 1)))
-		goto out;
-	return 1;
-out:
-	return 0;
-}
-
 /*
  * strndup_user - duplicate an existing string from user space
  * @s: The string to duplicate
-- 
cgit v0.10.2


From 86c8749ede0c59e590de9267066932a26f1ce796 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:18 +1100
Subject: vfs: revert per-cpu nr_unused counters for dentry and inodes

The nr_unused counters count the number of objects on an LRU, and as such they
are synchronized with LRU object insertion and removal and scanning, and
protected under the LRU lock.

Making it per-cpu does not actually get any concurrency improvements because of
this lock, and summing the counter is much slower, and
incrementing/decrementing it costs more code size and is slower too.

These counters should stay per-LRU, which currently means global.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 9d1a59df..f62ba90 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -68,14 +68,12 @@ struct dentry_stat_t dentry_stat = {
 };
 
 static struct percpu_counter nr_dentry __cacheline_aligned_in_smp;
-static struct percpu_counter nr_dentry_unused __cacheline_aligned_in_smp;
 
 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
 int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
 		   size_t *lenp, loff_t *ppos)
 {
 	dentry_stat.nr_dentry = percpu_counter_sum_positive(&nr_dentry);
-	dentry_stat.nr_unused = percpu_counter_sum_positive(&nr_dentry_unused);
 	return proc_dointvec(table, write, buffer, lenp, ppos);
 }
 #endif
@@ -140,7 +138,7 @@ static void dentry_lru_add(struct dentry *dentry)
 	if (list_empty(&dentry->d_lru)) {
 		list_add(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
 		dentry->d_sb->s_nr_dentry_unused++;
-		percpu_counter_inc(&nr_dentry_unused);
+		dentry_stat.nr_unused++;
 	}
 }
 
@@ -149,7 +147,7 @@ static void dentry_lru_del(struct dentry *dentry)
 	if (!list_empty(&dentry->d_lru)) {
 		list_del_init(&dentry->d_lru);
 		dentry->d_sb->s_nr_dentry_unused--;
-		percpu_counter_dec(&nr_dentry_unused);
+		dentry_stat.nr_unused--;
 	}
 }
 
@@ -158,7 +156,7 @@ static void dentry_lru_move_tail(struct dentry *dentry)
 	if (list_empty(&dentry->d_lru)) {
 		list_add_tail(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
 		dentry->d_sb->s_nr_dentry_unused++;
-		percpu_counter_inc(&nr_dentry_unused);
+		dentry_stat.nr_unused++;
 	} else {
 		list_move_tail(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
 	}
@@ -546,7 +544,7 @@ static void prune_dcache(int count)
 {
 	struct super_block *sb, *p = NULL;
 	int w_count;
-	int unused = percpu_counter_sum_positive(&nr_dentry_unused);
+	int unused = dentry_stat.nr_unused;
 	int prune_ratio;
 	int pruned;
 
@@ -908,16 +906,13 @@ EXPORT_SYMBOL(shrink_dcache_parent);
  */
 static int shrink_dcache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask)
 {
-	int nr_unused;
-
 	if (nr) {
 		if (!(gfp_mask & __GFP_FS))
 			return -1;
 		prune_dcache(nr);
 	}
 
-	nr_unused = percpu_counter_sum_positive(&nr_dentry_unused);
-	return (nr_unused / 100) * sysctl_vfs_cache_pressure;
+	return (dentry_stat.nr_unused / 100) * sysctl_vfs_cache_pressure;
 }
 
 static struct shrinker dcache_shrinker = {
@@ -2424,7 +2419,6 @@ static void __init dcache_init(void)
 	int loop;
 
 	percpu_counter_init(&nr_dentry, 0);
-	percpu_counter_init(&nr_dentry_unused, 0);
 
 	/* 
 	 * A constructor could be added for stable state like the lists,
diff --git a/fs/inode.c b/fs/inode.c
index ae2727a..efc4397 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -103,7 +103,6 @@ static DECLARE_RWSEM(iprune_sem);
 struct inodes_stat_t inodes_stat;
 
 static struct percpu_counter nr_inodes __cacheline_aligned_in_smp;
-static struct percpu_counter nr_inodes_unused __cacheline_aligned_in_smp;
 
 static struct kmem_cache *inode_cachep __read_mostly;
 
@@ -114,7 +113,7 @@ static inline int get_nr_inodes(void)
 
 static inline int get_nr_inodes_unused(void)
 {
-	return percpu_counter_sum_positive(&nr_inodes_unused);
+	return inodes_stat.nr_unused;
 }
 
 int get_nr_dirty_inodes(void)
@@ -132,7 +131,6 @@ int proc_nr_inodes(ctl_table *table, int write,
 		   void __user *buffer, size_t *lenp, loff_t *ppos)
 {
 	inodes_stat.nr_inodes = get_nr_inodes();
-	inodes_stat.nr_unused = get_nr_inodes_unused();
 	return proc_dointvec(table, write, buffer, lenp, ppos);
 }
 #endif
@@ -335,7 +333,7 @@ static void inode_lru_list_add(struct inode *inode)
 {
 	if (list_empty(&inode->i_lru)) {
 		list_add(&inode->i_lru, &inode_lru);
-		percpu_counter_inc(&nr_inodes_unused);
+		inodes_stat.nr_unused++;
 	}
 }
 
@@ -343,7 +341,7 @@ static void inode_lru_list_del(struct inode *inode)
 {
 	if (!list_empty(&inode->i_lru)) {
 		list_del_init(&inode->i_lru);
-		percpu_counter_dec(&nr_inodes_unused);
+		inodes_stat.nr_unused--;
 	}
 }
 
@@ -513,7 +511,7 @@ void evict_inodes(struct super_block *sb)
 		list_move(&inode->i_lru, &dispose);
 		list_del_init(&inode->i_wb_list);
 		if (!(inode->i_state & (I_DIRTY | I_SYNC)))
-			percpu_counter_dec(&nr_inodes_unused);
+			inodes_stat.nr_unused--;
 	}
 	spin_unlock(&inode_lock);
 
@@ -554,7 +552,7 @@ int invalidate_inodes(struct super_block *sb)
 		list_move(&inode->i_lru, &dispose);
 		list_del_init(&inode->i_wb_list);
 		if (!(inode->i_state & (I_DIRTY | I_SYNC)))
-			percpu_counter_dec(&nr_inodes_unused);
+			inodes_stat.nr_unused--;
 	}
 	spin_unlock(&inode_lock);
 
@@ -616,7 +614,7 @@ static void prune_icache(int nr_to_scan)
 		if (atomic_read(&inode->i_count) ||
 		    (inode->i_state & ~I_REFERENCED)) {
 			list_del_init(&inode->i_lru);
-			percpu_counter_dec(&nr_inodes_unused);
+			inodes_stat.nr_unused--;
 			continue;
 		}
 
@@ -650,7 +648,7 @@ static void prune_icache(int nr_to_scan)
 		 */
 		list_move(&inode->i_lru, &freeable);
 		list_del_init(&inode->i_wb_list);
-		percpu_counter_dec(&nr_inodes_unused);
+		inodes_stat.nr_unused--;
 	}
 	if (current_is_kswapd())
 		__count_vm_events(KSWAPD_INODESTEAL, reap);
@@ -1649,7 +1647,6 @@ void __init inode_init(void)
 					 init_once);
 	register_shrinker(&icache_shrinker);
 	percpu_counter_init(&nr_inodes, 0);
-	percpu_counter_init(&nr_inodes_unused, 0);
 
 	/* Hash may have been set up in inode_init_early */
 	if (!hashdist)
-- 
cgit v0.10.2


From 3e880fb5e4bb6a012035e3edd0586ee2817c2e24 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:19 +1100
Subject: fs: use fast counters for vfs caches

percpu_counter library generates quite nasty code, so unless you need
to dynamically allocate counters or take fast approximate value, a
simple per cpu set of counters is much better.

The percpu_counter can never be made to work as well, because it has an
indirection from pointer to percpu memory, and it can't use direct
this_cpu_inc interfaces because it doesn't use static PER_CPU data, so
code will always be worse.

In the fastpath, it is the difference between this:

        incl %gs:nr_dentry      # nr_dentry

and this:

        movl    percpu_counter_batch(%rip), %edx        # percpu_counter_batch,
        movl    $1, %esi        #,
        movq    $nr_dentry, %rdi        #,
        call    __percpu_counter_add    # (plus I clobber registers)

__percpu_counter_add:
        pushq   %rbp    #
        movq    %rsp, %rbp      #,
        subq    $32, %rsp       #,
        movq    %rbx, -24(%rbp) #,
        movq    %r12, -16(%rbp) #,
        movq    %r13, -8(%rbp)  #,
        movq    %rdi, %rbx      # fbc, fbc
#APP
# 216 "/home/npiggin/usr/src/linux-2.6/arch/x86/include/asm/thread_info.h" 1
        movq %gs:kernel_stack,%rax      #, pfo_ret__
# 0 "" 2
#NO_APP
        incl    -8124(%rax)     # <variable>.preempt_count
        movq    32(%rdi), %r12  # <variable>.counters, tcp_ptr__
#APP
# 78 "lib/percpu_counter.c" 1
        add %gs:this_cpu_off, %r12      # this_cpu_off, tcp_ptr__
# 0 "" 2
#NO_APP
        movslq  (%r12),%r13     #* tcp_ptr__, tmp73
        movslq  %edx,%rax       # batch, batch
        addq    %rsi, %r13      # amount, count
        cmpq    %rax, %r13      # batch, count
        jge     .L27    #,
        negl    %edx    # tmp76
        movslq  %edx,%rdx       # tmp76, tmp77
        cmpq    %rdx, %r13      # tmp77, count
        jg      .L28    #,
.L27:
        movq    %rbx, %rdi      # fbc,
        call    _raw_spin_lock  #
        addq    %r13, 8(%rbx)   # count, <variable>.count
        movq    %rbx, %rdi      # fbc,
        movl    $0, (%r12)      #,* tcp_ptr__
        call    _raw_spin_unlock        #
.L29:
#APP
# 216 "/home/npiggin/usr/src/linux-2.6/arch/x86/include/asm/thread_info.h" 1
        movq %gs:kernel_stack,%rax      #, pfo_ret__
# 0 "" 2
#NO_APP
        decl    -8124(%rax)     # <variable>.preempt_count
        movq    -8136(%rax), %rax       #, D.14625
        testb   $8, %al #, D.14625
        jne     .L32    #,
.L31:
        movq    -24(%rbp), %rbx #,
        movq    -16(%rbp), %r12 #,
        movq    -8(%rbp), %r13  #,
        leave
        ret
        .p2align 4,,10
        .p2align 3
.L28:
        movl    %r13d, (%r12)   # count,*
        jmp     .L29    #
.L32:
        call    preempt_schedule        #
        .p2align 4,,6
        jmp     .L31    #
        .size   __percpu_counter_add, .-__percpu_counter_add
        .p2align 4,,15

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index f62ba90..b2cb266 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -67,13 +67,22 @@ struct dentry_stat_t dentry_stat = {
 	.age_limit = 45,
 };
 
-static struct percpu_counter nr_dentry __cacheline_aligned_in_smp;
+static DEFINE_PER_CPU(unsigned int, nr_dentry);
 
 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
+static int get_nr_dentry(void)
+{
+	int i;
+	int sum = 0;
+	for_each_possible_cpu(i)
+		sum += per_cpu(nr_dentry, i);
+	return sum < 0 ? 0 : sum;
+}
+
 int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
 		   size_t *lenp, loff_t *ppos)
 {
-	dentry_stat.nr_dentry = percpu_counter_sum_positive(&nr_dentry);
+	dentry_stat.nr_dentry = get_nr_dentry();
 	return proc_dointvec(table, write, buffer, lenp, ppos);
 }
 #endif
@@ -93,7 +102,7 @@ static void __d_free(struct rcu_head *head)
  */
 static void d_free(struct dentry *dentry)
 {
-	percpu_counter_dec(&nr_dentry);
+	this_cpu_dec(nr_dentry);
 	if (dentry->d_op && dentry->d_op->d_release)
 		dentry->d_op->d_release(dentry);
 
@@ -981,7 +990,7 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 		list_add(&dentry->d_u.d_child, &parent->d_subdirs);
 	spin_unlock(&dcache_lock);
 
-	percpu_counter_inc(&nr_dentry);
+	this_cpu_inc(nr_dentry);
 
 	return dentry;
 }
@@ -2418,8 +2427,6 @@ static void __init dcache_init(void)
 {
 	int loop;
 
-	percpu_counter_init(&nr_dentry, 0);
-
 	/* 
 	 * A constructor could be added for stable state like the lists,
 	 * but it is probably not worth it because of the cache nature
diff --git a/fs/inode.c b/fs/inode.c
index efc4397..5a0a898 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -102,13 +102,17 @@ static DECLARE_RWSEM(iprune_sem);
  */
 struct inodes_stat_t inodes_stat;
 
-static struct percpu_counter nr_inodes __cacheline_aligned_in_smp;
+static DEFINE_PER_CPU(unsigned int, nr_inodes);
 
 static struct kmem_cache *inode_cachep __read_mostly;
 
-static inline int get_nr_inodes(void)
+static int get_nr_inodes(void)
 {
-	return percpu_counter_sum_positive(&nr_inodes);
+	int i;
+	int sum = 0;
+	for_each_possible_cpu(i)
+		sum += per_cpu(nr_inodes, i);
+	return sum < 0 ? 0 : sum;
 }
 
 static inline int get_nr_inodes_unused(void)
@@ -118,9 +122,9 @@ static inline int get_nr_inodes_unused(void)
 
 int get_nr_dirty_inodes(void)
 {
+	/* not actually dirty inodes, but a wild approximation */
 	int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
 	return nr_dirty > 0 ? nr_dirty : 0;
-
 }
 
 /*
@@ -222,7 +226,7 @@ int inode_init_always(struct super_block *sb, struct inode *inode)
 	inode->i_fsnotify_mask = 0;
 #endif
 
-	percpu_counter_inc(&nr_inodes);
+	this_cpu_inc(nr_inodes);
 
 	return 0;
 out:
@@ -264,7 +268,7 @@ void __destroy_inode(struct inode *inode)
 	if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
 		posix_acl_release(inode->i_default_acl);
 #endif
-	percpu_counter_dec(&nr_inodes);
+	this_cpu_dec(nr_inodes);
 }
 EXPORT_SYMBOL(__destroy_inode);
 
@@ -1646,7 +1650,6 @@ void __init inode_init(void)
 					 SLAB_MEM_SPREAD),
 					 init_once);
 	register_shrinker(&icache_shrinker);
-	percpu_counter_init(&nr_inodes, 0);
 
 	/* Hash may have been set up in inode_init_early */
 	if (!hashdist)
-- 
cgit v0.10.2


From 5adcee1d8d32d7f305f6f5aaefdbf8f35adca177 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:20 +1100
Subject: cgroup fs: avoid switching ->d_op on live dentry

Switching d_op on a live dentry is racy in general, so avoid it. In this case
it is a negative dentry, which is safer, but there are still concurrent ops
which may be called on d_op in that case (eg. d_revalidate). So in general
a filesystem may not do this. Fix cgroupfs so as not to do this.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 66a416b..163c890 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -763,6 +763,8 @@ EXPORT_SYMBOL_GPL(cgroup_unlock);
  * -> cgroup_mkdir.
  */
 
+static struct dentry *cgroup_lookup(struct inode *dir,
+			struct dentry *dentry, struct nameidata *nd);
 static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, int mode);
 static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
 static int cgroup_populate_dir(struct cgroup *cgrp);
@@ -2180,7 +2182,7 @@ static const struct file_operations cgroup_file_operations = {
 };
 
 static const struct inode_operations cgroup_dir_inode_operations = {
-	.lookup = simple_lookup,
+	.lookup = cgroup_lookup,
 	.mkdir = cgroup_mkdir,
 	.rmdir = cgroup_rmdir,
 	.rename = cgroup_rename,
@@ -2196,13 +2198,29 @@ static inline struct cftype *__file_cft(struct file *file)
 	return __d_cft(file->f_dentry);
 }
 
-static int cgroup_create_file(struct dentry *dentry, mode_t mode,
-				struct super_block *sb)
+static int cgroup_delete_dentry(struct dentry *dentry)
+{
+	return 1;
+}
+
+static struct dentry *cgroup_lookup(struct inode *dir,
+			struct dentry *dentry, struct nameidata *nd)
 {
-	static const struct dentry_operations cgroup_dops = {
+	static const struct dentry_operations cgroup_dentry_operations = {
+		.d_delete = cgroup_delete_dentry,
 		.d_iput = cgroup_diput,
 	};
 
+	if (dentry->d_name.len > NAME_MAX)
+		return ERR_PTR(-ENAMETOOLONG);
+	dentry->d_op = &cgroup_dentry_operations;
+	d_add(dentry, NULL);
+	return NULL;
+}
+
+static int cgroup_create_file(struct dentry *dentry, mode_t mode,
+				struct super_block *sb)
+{
 	struct inode *inode;
 
 	if (!dentry)
@@ -2228,7 +2246,6 @@ static int cgroup_create_file(struct dentry *dentry, mode_t mode,
 		inode->i_size = 0;
 		inode->i_fop = &cgroup_file_operations;
 	}
-	dentry->d_op = &cgroup_dops;
 	d_instantiate(dentry, inode);
 	dget(dentry);	/* Extra count - pin the dentry in core */
 	return 0;
-- 
cgit v0.10.2


From fbc8d4c04626e015b18cc61199f505920abb48f0 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:21 +1100
Subject: config fs: avoid switching ->d_op on live dentry

Switching d_op on a live dentry is racy in general, so avoid it. In this case
it is a negative dentry, which is safer, but there are still concurrent ops
which may be called on d_op in that case (eg. d_revalidate). So in general
a filesystem may not do this. Fix configfs so as not to do this.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index 0b502f8..5787069 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -232,10 +232,8 @@ int configfs_make_dirent(struct configfs_dirent * parent_sd,
 
 	sd->s_mode = mode;
 	sd->s_dentry = dentry;
-	if (dentry) {
+	if (dentry)
 		dentry->d_fsdata = configfs_get(sd);
-		dentry->d_op = &configfs_dentry_ops;
-	}
 
 	return 0;
 }
@@ -278,7 +276,6 @@ static int create_dir(struct config_item * k, struct dentry * p,
 		error = configfs_create(d, mode, init_dir);
 		if (!error) {
 			inc_nlink(p->d_inode);
-			(d)->d_op = &configfs_dentry_ops;
 		} else {
 			struct configfs_dirent *sd = d->d_fsdata;
 			if (sd) {
@@ -371,9 +368,7 @@ int configfs_create_link(struct configfs_symlink *sl,
 				   CONFIGFS_ITEM_LINK);
 	if (!err) {
 		err = configfs_create(dentry, mode, init_symlink);
-		if (!err)
-			dentry->d_op = &configfs_dentry_ops;
-		else {
+		if (err) {
 			struct configfs_dirent *sd = dentry->d_fsdata;
 			if (sd) {
 				spin_lock(&configfs_dirent_lock);
@@ -493,7 +488,11 @@ static struct dentry * configfs_lookup(struct inode *dir,
 		 * If it doesn't exist and it isn't a NOT_PINNED item,
 		 * it must be negative.
 		 */
-		return simple_lookup(dir, dentry, nd);
+		if (dentry->d_name.len > NAME_MAX)
+			return ERR_PTR(-ENAMETOOLONG);
+		dentry->d_op = &configfs_dentry_ops;
+		d_add(dentry, NULL);
+		return NULL;
 	}
 
 out:
@@ -685,6 +684,7 @@ static int create_default_group(struct config_group *parent_group,
 	ret = -ENOMEM;
 	child = d_alloc(parent, &name);
 	if (child) {
+		child->d_op = &configfs_dentry_ops;
 		d_add(child, NULL);
 
 		ret = configfs_attach_group(&parent_group->cg_item,
@@ -1682,6 +1682,7 @@ int configfs_register_subsystem(struct configfs_subsystem *subsys)
 	err = -ENOMEM;
 	dentry = d_alloc(configfs_sb->s_root, &name);
 	if (dentry) {
+		dentry->d_op = &configfs_dentry_ops;
 		d_add(dentry, NULL);
 
 		err = configfs_attach_group(sd->s_element, &group->cg_item,
-- 
cgit v0.10.2


From 5eef7fa905c814826f518aca2d414ca77508ce30 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:22 +1100
Subject: fs: dcache documentation cleanup

Remove redundant (and incorrect, since dcache RCU lookup) dentry locking
documentation and point to the canonical document.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 6a4aea3..fff9755 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -141,22 +141,16 @@ struct dentry_operations {
 	char *(*d_dname)(struct dentry *, char *, int);
 };
 
-/* the dentry parameter passed to d_hash and d_compare is the parent
+/*
+ * Locking rules for dentry_operations callbacks are to be found in
+ * Documentation/filesystems/Locking. Keep it updated!
+ *
+ * the dentry parameter passed to d_hash and d_compare is the parent
  * directory of the entries to be compared. It is used in case these
  * functions need any directory specific information for determining
  * equivalency classes.  Using the dentry itself might not work, as it
  * might be a negative dentry which has no information associated with
- * it */
-
-/*
-locking rules:
-		big lock	dcache_lock	d_lock   may block
-d_revalidate:	no		no		no       yes
-d_hash		no		no		no       yes
-d_compare:	no		yes		yes      no
-d_delete:	no		yes		no       no
-d_release:	no		no		no       yes
-d_iput:		no		no		no       yes
+ * it.
  */
 
 /* d_flags entries */
-- 
cgit v0.10.2


From fe15ce446beb3a33583af81ffe6c9d01a75314ed Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:23 +1100
Subject: fs: change d_delete semantics

Change d_delete from a dentry deletion notification to a dentry caching
advise, more like ->drop_inode. Require it to be constant and idempotent,
and not take d_lock. This is how all existing filesystems use the callback
anyway.

This makes fine grained dentry locking of dput and dentry lru scanning
much simpler.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index b12c895..9e71c9a 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -318,3 +318,11 @@ if it's zero is not *and* *never* *had* *been* enough.  Final unlink() and iput(
 may happen while the inode is in the middle of ->write_inode(); e.g. if you blindly
 free the on-disk inode, you may end up doing that while ->write_inode() is writing
 to it.
+
+---
+[mandatory]
+
+	.d_delete() now only advises the dcache as to whether or not to cache
+unreferenced dentries, and is now only called when the dentry refcount goes to
+0. Even on 0 refcount transition, it must be able to tolerate being called 0,
+1, or more times (eg. constant, idempotent).
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index 20899e0..95c0a93f 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -847,9 +847,9 @@ defined:
 
 struct dentry_operations {
 	int (*d_revalidate)(struct dentry *, struct nameidata *);
-	int (*d_hash) (struct dentry *, struct qstr *);
-	int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
-	int (*d_delete)(struct dentry *);
+	int (*d_hash)(struct dentry *, struct qstr *);
+	int (*d_compare)(struct dentry *, struct qstr *, struct qstr *);
+	int (*d_delete)(const struct dentry *);
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
 	char *(*d_dname)(struct dentry *, char *, int);
@@ -864,9 +864,11 @@ struct dentry_operations {
 
   d_compare: called when a dentry should be compared with another
 
-  d_delete: called when the last reference to a dentry is
-	deleted. This means no-one is using the dentry, however it is
-	still valid and in the dcache
+  d_delete: called when the last reference to a dentry is dropped and the
+	dcache is deciding whether or not to cache it. Return 1 to delete
+	immediately, or 0 to cache the dentry. Default is NULL which means to
+	always cache a reachable dentry. d_delete must be constant and
+	idempotent.
 
   d_release: called when a dentry is really deallocated
 
@@ -910,14 +912,11 @@ manipulate dentries:
 	the usage count)
 
   dput: close a handle for a dentry (decrements the usage count). If
-	the usage count drops to 0, the "d_delete" method is called
-	and the dentry is placed on the unused list if the dentry is
-	still in its parents hash list. Putting the dentry on the
-	unused list just means that if the system needs some RAM, it
-	goes through the unused list of dentries and deallocates them.
-	If the dentry has already been unhashed and the usage count
-	drops to 0, in this case the dentry is deallocated after the
-	"d_delete" method is called
+	the usage count drops to 0, and the dentry is still in its
+	parent's hash, the "d_delete" method is called to check whether
+	it should be cached. If it should not be cached, or if the dentry
+	is not hashed, it is deleted. Otherwise cached dentries are put
+	into an LRU list to be reclaimed on memory shortage.
 
   d_drop: this unhashes a dentry from its parents hash list. A
 	subsequent call to dput() will deallocate the dentry if its
diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index 39e534f..d39d8a5 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -2185,7 +2185,7 @@ static const struct file_operations pfm_file_ops = {
 };
 
 static int
-pfmfs_delete_dentry(struct dentry *dentry)
+pfmfs_delete_dentry(const struct dentry *dentry)
 {
 	return 1;
 }
diff --git a/drivers/staging/smbfs/dir.c b/drivers/staging/smbfs/dir.c
index f088ea2..2270d48 100644
--- a/drivers/staging/smbfs/dir.c
+++ b/drivers/staging/smbfs/dir.c
@@ -276,7 +276,7 @@ smb_dir_open(struct inode *dir, struct file *file)
 static int smb_lookup_validate(struct dentry *, struct nameidata *);
 static int smb_hash_dentry(struct dentry *, struct qstr *);
 static int smb_compare_dentry(struct dentry *, struct qstr *, struct qstr *);
-static int smb_delete_dentry(struct dentry *);
+static int smb_delete_dentry(const struct dentry *);
 
 static const struct dentry_operations smbfs_dentry_operations =
 {
@@ -367,7 +367,7 @@ out:
  * We use this to unhash dentries with bad inodes.
  */
 static int
-smb_delete_dentry(struct dentry * dentry)
+smb_delete_dentry(const struct dentry *dentry)
 {
 	if (dentry->d_inode) {
 		if (is_bad_inode(dentry->d_inode)) {
diff --git a/fs/9p/vfs_dentry.c b/fs/9p/vfs_dentry.c
index cbf4e50..466d2a4 100644
--- a/fs/9p/vfs_dentry.c
+++ b/fs/9p/vfs_dentry.c
@@ -51,7 +51,7 @@
  *
  */
 
-static int v9fs_dentry_delete(struct dentry *dentry)
+static int v9fs_dentry_delete(const struct dentry *dentry)
 {
 	P9_DPRINTK(P9_DEBUG_VFS, " dentry: %s (%p)\n", dentry->d_name.name,
 									dentry);
@@ -68,7 +68,7 @@ static int v9fs_dentry_delete(struct dentry *dentry)
  *
  */
 
-static int v9fs_cached_dentry_delete(struct dentry *dentry)
+static int v9fs_cached_dentry_delete(const struct dentry *dentry)
 {
 	struct inode *inode = dentry->d_inode;
 	P9_DPRINTK(P9_DEBUG_VFS, " dentry: %s (%p)\n", dentry->d_name.name,
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index 5439e1b..2c18cde 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -23,7 +23,7 @@ static struct dentry *afs_lookup(struct inode *dir, struct dentry *dentry,
 static int afs_dir_open(struct inode *inode, struct file *file);
 static int afs_readdir(struct file *file, void *dirent, filldir_t filldir);
 static int afs_d_revalidate(struct dentry *dentry, struct nameidata *nd);
-static int afs_d_delete(struct dentry *dentry);
+static int afs_d_delete(const struct dentry *dentry);
 static void afs_d_release(struct dentry *dentry);
 static int afs_lookup_filldir(void *_cookie, const char *name, int nlen,
 				  loff_t fpos, u64 ino, unsigned dtype);
@@ -730,7 +730,7 @@ out_bad:
  * - called from dput() when d_count is going to 0.
  * - return 1 to request dentry be unhashed, 0 otherwise
  */
-static int afs_d_delete(struct dentry *dentry)
+static int afs_d_delete(const struct dentry *dentry)
 {
 	_enter("%s", dentry->d_name.name);
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 72f31ec..7ce9f89 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4127,7 +4127,7 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 	return inode;
 }
 
-static int btrfs_dentry_delete(struct dentry *dentry)
+static int btrfs_dentry_delete(const struct dentry *dentry)
 {
 	struct btrfs_root *root;
 
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index 5d8b355..4cce3b07 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -47,7 +47,7 @@ static int coda_readdir(struct file *file, void *buf, filldir_t filldir);
 
 /* dentry ops */
 static int coda_dentry_revalidate(struct dentry *de, struct nameidata *nd);
-static int coda_dentry_delete(struct dentry *);
+static int coda_dentry_delete(const struct dentry *);
 
 /* support routines */
 static int coda_venus_readdir(struct file *coda_file, void *buf,
@@ -577,7 +577,7 @@ out:
  * This is the callback from dput() when d_count is going to 0.
  * We use this to unhash dentries with bad inodes.
  */
-static int coda_dentry_delete(struct dentry * dentry)
+static int coda_dentry_delete(const struct dentry * dentry)
 {
 	int flags;
 
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index 5787069..20024a9 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -67,7 +67,7 @@ static void configfs_d_iput(struct dentry * dentry,
  * We _must_ delete our dentries on last dput, as the chain-to-parent
  * behavior is required to clear the parents of default_groups.
  */
-static int configfs_d_delete(struct dentry *dentry)
+static int configfs_d_delete(const struct dentry *dentry)
 {
 	return 1;
 }
diff --git a/fs/dcache.c b/fs/dcache.c
index b2cb266..6ee6bc4 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -453,8 +453,6 @@ static void prune_one_dentry(struct dentry * dentry)
 		if (!atomic_dec_and_lock(&dentry->d_count, &dentry->d_lock))
 			return;
 
-		if (dentry->d_op && dentry->d_op->d_delete)
-			dentry->d_op->d_delete(dentry);
 		dentry_lru_del(dentry);
 		__d_drop(dentry);
 		dentry = d_kill(dentry);
diff --git a/fs/gfs2/dentry.c b/fs/gfs2/dentry.c
index 6798755..e80fea2 100644
--- a/fs/gfs2/dentry.c
+++ b/fs/gfs2/dentry.c
@@ -106,7 +106,7 @@ static int gfs2_dhash(struct dentry *dentry, struct qstr *str)
 	return 0;
 }
 
-static int gfs2_dentry_delete(struct dentry *dentry)
+static int gfs2_dentry_delete(const struct dentry *dentry)
 {
 	struct gfs2_inode *ginode;
 
diff --git a/fs/hostfs/hostfs_kern.c b/fs/hostfs/hostfs_kern.c
index 2c0f148..cfe8bc7 100644
--- a/fs/hostfs/hostfs_kern.c
+++ b/fs/hostfs/hostfs_kern.c
@@ -32,7 +32,7 @@ static inline struct hostfs_inode_info *HOSTFS_I(struct inode *inode)
 
 #define FILE_HOSTFS_I(file) HOSTFS_I((file)->f_path.dentry->d_inode)
 
-static int hostfs_d_delete(struct dentry *dentry)
+static int hostfs_d_delete(const struct dentry *dentry)
 {
 	return 1;
 }
diff --git a/fs/libfs.c b/fs/libfs.c
index a3accdf..b9d25d8 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -37,7 +37,7 @@ int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
  * Retaining negative dentries for an in-memory filesystem just wastes
  * memory and lookup time: arrange for them to be deleted immediately.
  */
-static int simple_delete_dentry(struct dentry *dentry)
+static int simple_delete_dentry(const struct dentry *dentry)
 {
 	return 1;
 }
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index f22b12e..d6e6453 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -76,7 +76,7 @@ const struct inode_operations ncp_dir_inode_operations =
 static int ncp_lookup_validate(struct dentry *, struct nameidata *);
 static int ncp_hash_dentry(struct dentry *, struct qstr *);
 static int ncp_compare_dentry (struct dentry *, struct qstr *, struct qstr *);
-static int ncp_delete_dentry(struct dentry *);
+static int ncp_delete_dentry(const struct dentry *);
 
 static const struct dentry_operations ncp_dentry_operations =
 {
@@ -162,7 +162,7 @@ ncp_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b)
  * Closing files can be safely postponed until iput() - it's done there anyway.
  */
 static int
-ncp_delete_dentry(struct dentry * dentry)
+ncp_delete_dentry(const struct dentry * dentry)
 {
 	struct inode *inode = dentry->d_inode;
 
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 996dd89..9184c7c 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1117,7 +1117,7 @@ out_error:
 /*
  * This is called from dput() when d_count is going to 0.
  */
-static int nfs_dentry_delete(struct dentry *dentry)
+static int nfs_dentry_delete(const struct dentry *dentry)
 {
 	dfprintk(VFS, "NFS: dentry_delete(%s/%s, %x)\n",
 		dentry->d_parent->d_name.name, dentry->d_name.name,
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 1828451..d932fdb 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1744,7 +1744,7 @@ static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
 	return 0;
 }
 
-static int pid_delete_dentry(struct dentry * dentry)
+static int pid_delete_dentry(const struct dentry * dentry)
 {
 	/* Is the task we represent dead?
 	 * If so, then don't put the dentry on the lru list,
diff --git a/fs/proc/generic.c b/fs/proc/generic.c
index dd29f03..1d607be 100644
--- a/fs/proc/generic.c
+++ b/fs/proc/generic.c
@@ -400,7 +400,7 @@ static const struct inode_operations proc_link_inode_operations = {
  * smarter: we could keep a "volatile" flag in the 
  * inode to indicate which ones to keep.
  */
-static int proc_delete_dentry(struct dentry * dentry)
+static int proc_delete_dentry(const struct dentry * dentry)
 {
 	return 1;
 }
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index b652cb0..a256d77 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -392,7 +392,7 @@ static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
 	return !PROC_I(dentry->d_inode)->sysctl->unregistering;
 }
 
-static int proc_sys_delete(struct dentry *dentry)
+static int proc_sys_delete(const struct dentry *dentry)
 {
 	return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
 }
diff --git a/fs/sysfs/dir.c b/fs/sysfs/dir.c
index 7e54bac..27e1102 100644
--- a/fs/sysfs/dir.c
+++ b/fs/sysfs/dir.c
@@ -231,7 +231,7 @@ void release_sysfs_dirent(struct sysfs_dirent * sd)
 		goto repeat;
 }
 
-static int sysfs_dentry_delete(struct dentry *dentry)
+static int sysfs_dentry_delete(const struct dentry *dentry)
 {
 	struct sysfs_dirent *sd = dentry->d_fsdata;
 	return !!(sd->s_flags & SYSFS_FLAG_REMOVED);
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index fff9755..cbfc956 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -133,9 +133,9 @@ enum dentry_d_lock_class
 
 struct dentry_operations {
 	int (*d_revalidate)(struct dentry *, struct nameidata *);
-	int (*d_hash) (struct dentry *, struct qstr *);
-	int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
-	int (*d_delete)(struct dentry *);
+	int (*d_hash)(struct dentry *, struct qstr *);
+	int (*d_compare)(struct dentry *, struct qstr *, struct qstr *);
+	int (*d_delete)(const struct dentry *);
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
 	char *(*d_dname)(struct dentry *, char *, int);
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 163c890..746055b 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -2198,7 +2198,7 @@ static inline struct cftype *__file_cft(struct file *file)
 	return __d_cft(file->f_dentry);
 }
 
-static int cgroup_delete_dentry(struct dentry *dentry)
+static int cgroup_delete_dentry(const struct dentry *dentry)
 {
 	return 1;
 }
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index 10a17a3..a0dc1a8 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -430,7 +430,7 @@ void rpc_put_mount(void)
 }
 EXPORT_SYMBOL_GPL(rpc_put_mount);
 
-static int rpc_delete_dentry(struct dentry *dentry)
+static int rpc_delete_dentry(const struct dentry *dentry)
 {
 	return 1;
 }
-- 
cgit v0.10.2


From 79eb4dde742fe2e9c9e301432b894a7410261ce7 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:24 +1100
Subject: cifs: dont overwrite dentry name in d_revalidate

Use vfat's method for dealing with negative dentries to preserve case,
rather than overwrite dentry name in d_revalidate, which is a bit ugly
and also gets in the way of doing lock-free path walking.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index 3840edd..521d841 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -656,22 +656,34 @@ lookup_out:
 static int
 cifs_d_revalidate(struct dentry *direntry, struct nameidata *nd)
 {
-	int isValid = 1;
-
 	if (direntry->d_inode) {
 		if (cifs_revalidate_dentry(direntry))
 			return 0;
-	} else {
-		cFYI(1, "neg dentry 0x%p name = %s",
-			 direntry, direntry->d_name.name);
-		if (time_after(jiffies, direntry->d_time + HZ) ||
-			!lookupCacheEnabled) {
-			d_drop(direntry);
-			isValid = 0;
-		}
+		else
+			return 1;
 	}
 
-	return isValid;
+	/*
+	 * This may be nfsd (or something), anyway, we can't see the
+	 * intent of this. So, since this can be for creation, drop it.
+	 */
+	if (!nd)
+		return 0;
+
+	/*
+	 * Drop the negative dentry, in order to make sure to use the
+	 * case sensitive name which is specified by user if this is
+	 * for creation.
+	 */
+	if (!(nd->flags & (LOOKUP_CONTINUE | LOOKUP_PARENT))) {
+		if (nd->flags & (LOOKUP_CREATE | LOOKUP_RENAME_TARGET))
+			return 0;
+	}
+
+	if (time_after(jiffies, direntry->d_time + HZ) || !lookupCacheEnabled)
+		return 0;
+
+	return 1;
 }
 
 /* static int cifs_d_delete(struct dentry *direntry)
@@ -709,15 +721,8 @@ static int cifs_ci_compare(struct dentry *dentry, struct qstr *a,
 	struct nls_table *codepage = CIFS_SB(dentry->d_inode->i_sb)->local_nls;
 
 	if ((a->len == b->len) &&
-	    (nls_strnicmp(codepage, a->name, b->name, a->len) == 0)) {
-		/*
-		 * To preserve case, don't let an existing negative dentry's
-		 * case take precedence.  If a is not a negative dentry, this
-		 * should have no side effects
-		 */
-		memcpy((void *)a->name, b->name, a->len);
+	    (nls_strnicmp(codepage, a->name, b->name, a->len) == 0))
 		return 0;
-	}
 	return 1;
 }
 
-- 
cgit v0.10.2


From 2bc334dcc7c77be3700dd443d92a78603f76976b Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:25 +1100
Subject: jfs: dont overwrite dentry name in d_revalidate

Use vfat's method for dealing with negative dentries to preserve case,
rather than overwrite dentry name in d_revalidate, which is a bit ugly
and also gets in the way of doing lock-free path walking.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/jfs/namei.c b/fs/jfs/namei.c
index 231ca4a..2da1546 100644
--- a/fs/jfs/namei.c
+++ b/fs/jfs/namei.c
@@ -18,6 +18,7 @@
  */
 
 #include <linux/fs.h>
+#include <linux/namei.h>
 #include <linux/ctype.h>
 #include <linux/quotaops.h>
 #include <linux/exportfs.h>
@@ -1597,21 +1598,47 @@ static int jfs_ci_compare(struct dentry *dir, struct qstr *a, struct qstr *b)
 			goto out;
 	}
 	result = 0;
+out:
+	return result;
+}
 
+static int jfs_ci_revalidate(struct dentry *dentry, struct nameidata *nd)
+{
 	/*
-	 * We want creates to preserve case.  A negative dentry, a, that
-	 * has a different case than b may cause a new entry to be created
-	 * with the wrong case.  Since we can't tell if a comes from a negative
-	 * dentry, we blindly replace it with b.  This should be harmless if
-	 * a is not a negative dentry.
+	 * This is not negative dentry. Always valid.
+	 *
+	 * Note, rename() to existing directory entry will have ->d_inode,
+	 * and will use existing name which isn't specified name by user.
+	 *
+	 * We may be able to drop this positive dentry here. But dropping
+	 * positive dentry isn't good idea. So it's unsupported like
+	 * rename("filename", "FILENAME") for now.
 	 */
-	memcpy((unsigned char *)a->name, b->name, a->len);
-out:
-	return result;
+	if (dentry->d_inode)
+		return 1;
+
+	/*
+	 * This may be nfsd (or something), anyway, we can't see the
+	 * intent of this. So, since this can be for creation, drop it.
+	 */
+	if (!nd)
+		return 0;
+
+	/*
+	 * Drop the negative dentry, in order to make sure to use the
+	 * case sensitive name which is specified by user if this is
+	 * for creation.
+	 */
+	if (!(nd->flags & (LOOKUP_CONTINUE | LOOKUP_PARENT))) {
+		if (nd->flags & (LOOKUP_CREATE | LOOKUP_RENAME_TARGET))
+			return 0;
+	}
+	return 1;
 }
 
 const struct dentry_operations jfs_ci_dentry_operations =
 {
 	.d_hash = jfs_ci_hash,
 	.d_compare = jfs_ci_compare,
+	.d_revalidate = jfs_ci_revalidate,
 };
-- 
cgit v0.10.2


From fb2d5b86aff355a27ebfc132d3c99f4a940cc3fe Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:26 +1100
Subject: fs: name case update method

smpfs and ncpfs want to update a live dentry name in-place. Rather than
have them open code the locking, provide a documented dcache API.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/drivers/staging/smbfs/cache.c b/drivers/staging/smbfs/cache.c
index dbb9865..dbd2e1d 100644
--- a/drivers/staging/smbfs/cache.c
+++ b/drivers/staging/smbfs/cache.c
@@ -145,8 +145,8 @@ smb_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
 			goto end_advance;
 	} else {
 		hashed = 1;
-		memcpy((char *) newdent->d_name.name, qname->name,
-		       newdent->d_name.len);
+		/* dir i_mutex is locked because we're in readdir */
+		dentry_update_name_case(newdent, qname);
 	}
 
 	if (!newdent->d_inode) {
diff --git a/fs/dcache.c b/fs/dcache.c
index 6ee6bc4..814e5f4 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1589,6 +1589,33 @@ void d_rehash(struct dentry * entry)
 }
 EXPORT_SYMBOL(d_rehash);
 
+/**
+ * dentry_update_name_case - update case insensitive dentry with a new name
+ * @dentry: dentry to be updated
+ * @name: new name
+ *
+ * Update a case insensitive dentry with new case of name.
+ *
+ * dentry must have been returned by d_lookup with name @name. Old and new
+ * name lengths must match (ie. no d_compare which allows mismatched name
+ * lengths).
+ *
+ * Parent inode i_mutex must be held over d_lookup and into this call (to
+ * keep renames and concurrent inserts, and readdir(2) away).
+ */
+void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
+{
+	BUG_ON(!mutex_is_locked(&dentry->d_inode->i_mutex));
+	BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */
+
+	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
+	memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
+	spin_unlock(&dentry->d_lock);
+	spin_unlock(&dcache_lock);
+}
+EXPORT_SYMBOL(dentry_update_name_case);
+
 /*
  * When switching names, the actual string doesn't strictly have to
  * be preserved in the target - because we're dropping the target
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index d6e6453..e80ea4e 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -611,35 +611,12 @@ ncp_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
 			shrink_dcache_parent(newdent);
 
 		/*
-		 * It is not as dangerous as it looks.  NetWare's OS2 namespace is
-		 * case preserving yet case insensitive.  So we update dentry's name
-		 * as received from server.  We found dentry via d_lookup with our
-		 * hash, so we know that hash does not change, and so replacing name
-		 * should be reasonably safe.
+		 * NetWare's OS2 namespace is case preserving yet case
+		 * insensitive.  So we update dentry's name as received from
+		 * server. Parent dir's i_mutex is locked because we're in
+		 * readdir.
 		 */
-		if (qname.len == newdent->d_name.len &&
-		    memcmp(newdent->d_name.name, qname.name, newdent->d_name.len)) {
-			struct inode *inode = newdent->d_inode;
-
-			/*
-			 * Inside ncpfs all uses of d_name are either for debugging,
-			 * or on functions which acquire inode mutex (mknod, creat,
-			 * lookup).  So grab i_mutex here, to be sure.  d_path
-			 * uses dcache_lock when generating path, so we should too.
-			 * And finally d_compare is protected by dentry's d_lock, so
-			 * here we go.
-			 */
-			if (inode)
-				mutex_lock(&inode->i_mutex);
-			spin_lock(&dcache_lock);
-			spin_lock(&newdent->d_lock);
-			memcpy((char *) newdent->d_name.name, qname.name,
-								newdent->d_name.len);
-			spin_unlock(&newdent->d_lock);
-			spin_unlock(&dcache_lock);
-			if (inode)
-				mutex_unlock(&inode->i_mutex);
-		}
+		dentry_update_name_case(newdent, &qname);
 	}
 
 	if (!newdent->d_inode) {
@@ -657,7 +634,7 @@ ncp_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
 	} else {
 		struct inode *inode = newdent->d_inode;
 
-		mutex_lock(&inode->i_mutex);
+		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
 		ncp_update_inode2(inode, entry);
 		mutex_unlock(&inode->i_mutex);
 	}
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index cbfc956..6cdf499 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -290,6 +290,8 @@ static inline struct dentry *d_add_unique(struct dentry *entry, struct inode *in
 	return res;
 }
 
+extern void dentry_update_name_case(struct dentry *, struct qstr *);
+
 /* used for rename() and baskets */
 extern void d_move(struct dentry *, struct dentry *);
 extern struct dentry *d_ancestor(struct dentry *, struct dentry *);
-- 
cgit v0.10.2


From 621e155a3591962420eacdd39f6f0aa29ceb221e Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:27 +1100
Subject: fs: change d_compare for rcu-walk

Change d_compare so it may be called from lock-free RCU lookups. This
does put significant restrictions on what may be done from the callback,
however there don't seem to have been any problems with in-tree fses.
If some strange use case pops up that _really_ cannot cope with the
rcu-walk rules, we can just add new rcu-unaware callbacks, which would
cause name lookup to drop out of rcu-walk mode.

For in-tree filesystems, this is just a mechanical change.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index 33fa3e5..9a76f8d 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -11,7 +11,9 @@ be able to use diff(1).
 prototypes:
 	int (*d_revalidate)(struct dentry *, int);
 	int (*d_hash) (struct dentry *, struct qstr *);
-	int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
+	int (*d_compare)(const struct dentry *, const struct inode *,
+			const struct dentry *, const struct inode *,
+			unsigned int, const char *, const struct qstr *);
 	int (*d_delete)(struct dentry *);
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index 9e71c9a..d44511e 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -326,3 +326,10 @@ to it.
 unreferenced dentries, and is now only called when the dentry refcount goes to
 0. Even on 0 refcount transition, it must be able to tolerate being called 0,
 1, or more times (eg. constant, idempotent).
+
+---
+[mandatory]
+
+	.d_compare() calling convention and locking rules are significantly
+changed. Read updated documentation in Documentation/filesystems/vfs.txt (and
+look at examples of other filesystems) for guidance.
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index 95c0a93f..250681b 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -848,7 +848,9 @@ defined:
 struct dentry_operations {
 	int (*d_revalidate)(struct dentry *, struct nameidata *);
 	int (*d_hash)(struct dentry *, struct qstr *);
-	int (*d_compare)(struct dentry *, struct qstr *, struct qstr *);
+	int (*d_compare)(const struct dentry *, const struct inode *,
+			const struct dentry *, const struct inode *,
+			unsigned int, const char *, const struct qstr *);
 	int (*d_delete)(const struct dentry *);
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
@@ -860,9 +862,27 @@ struct dentry_operations {
 	dcache. Most filesystems leave this as NULL, because all their
 	dentries in the dcache are valid
 
-  d_hash: called when the VFS adds a dentry to the hash table
+  d_hash: called when the VFS adds a dentry to the hash table. The first
+	dentry passed to d_hash is the parent directory that the name is
+	to be hashed into.
 
-  d_compare: called when a dentry should be compared with another
+  d_compare: called to compare a dentry name with a given name. The first
+	dentry is the parent of the dentry to be compared, the second is
+	the parent's inode, then the dentry and inode (may be NULL) of the
+	child dentry. len and name string are properties of the dentry to be
+	compared. qstr is the name to compare it with.
+
+	Must be constant and idempotent, and should not take locks if
+	possible, and should not or store into the dentry or inodes.
+	Should not dereference pointers outside the dentry or inodes without
+	lots of care (eg.  d_parent, d_inode, d_name should not be used).
+
+	However, our vfsmount is pinned, and RCU held, so the dentries and
+	inodes won't disappear, neither will our sb or filesystem module.
+	->i_sb and ->d_sb may be used.
+
+	It is a tricky calling convention because it needs to be called under
+	"rcu-walk", ie. without any locks or references on things.
 
   d_delete: called when the last reference to a dentry is dropped and the
 	dcache is deciding whether or not to cache it. Return 1 to delete
diff --git a/drivers/staging/smbfs/dir.c b/drivers/staging/smbfs/dir.c
index 2270d48..bd63229 100644
--- a/drivers/staging/smbfs/dir.c
+++ b/drivers/staging/smbfs/dir.c
@@ -275,7 +275,10 @@ smb_dir_open(struct inode *dir, struct file *file)
  */
 static int smb_lookup_validate(struct dentry *, struct nameidata *);
 static int smb_hash_dentry(struct dentry *, struct qstr *);
-static int smb_compare_dentry(struct dentry *, struct qstr *, struct qstr *);
+static int smb_compare_dentry(const struct dentry *,
+		const struct inode *,
+		const struct dentry *, const struct inode *,
+		unsigned int, const char *, const struct qstr *);
 static int smb_delete_dentry(const struct dentry *);
 
 static const struct dentry_operations smbfs_dentry_operations =
@@ -347,14 +350,17 @@ smb_hash_dentry(struct dentry *dir, struct qstr *this)
 }
 
 static int
-smb_compare_dentry(struct dentry *dir, struct qstr *a, struct qstr *b)
+smb_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
 	int i, result = 1;
 
-	if (a->len != b->len)
+	if (len != name->len)
 		goto out;
-	for (i=0; i < a->len; i++) {
-		if (tolower(a->name[i]) != tolower(b->name[i]))
+	for (i=0; i < len; i++) {
+		if (tolower(str[i]) != tolower(name->name[i]))
 			goto out;
 	}
 	result = 0;
diff --git a/fs/adfs/dir.c b/fs/adfs/dir.c
index f4287e4..c8ed661 100644
--- a/fs/adfs/dir.c
+++ b/fs/adfs/dir.c
@@ -237,17 +237,19 @@ adfs_hash(struct dentry *parent, struct qstr *qstr)
  * requirements of the underlying filesystem.
  */
 static int
-adfs_compare(struct dentry *parent, struct qstr *entry, struct qstr *name)
+adfs_compare(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
 	int i;
 
-	if (entry->len != name->len)
+	if (len != name->len)
 		return 1;
 
 	for (i = 0; i < name->len; i++) {
 		char a, b;
 
-		a = entry->name[i];
+		a = str[i];
 		b = name->name[i];
 
 		if (a >= 'A' && a <= 'Z')
diff --git a/fs/affs/namei.c b/fs/affs/namei.c
index 914d1c0..547d5de 100644
--- a/fs/affs/namei.c
+++ b/fs/affs/namei.c
@@ -14,10 +14,16 @@ typedef int (*toupper_t)(int);
 
 static int	 affs_toupper(int ch);
 static int	 affs_hash_dentry(struct dentry *, struct qstr *);
-static int       affs_compare_dentry(struct dentry *, struct qstr *, struct qstr *);
+static int       affs_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
 static int	 affs_intl_toupper(int ch);
 static int	 affs_intl_hash_dentry(struct dentry *, struct qstr *);
-static int       affs_intl_compare_dentry(struct dentry *, struct qstr *, struct qstr *);
+static int       affs_intl_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
 
 const struct dentry_operations affs_dentry_operations = {
 	.d_hash		= affs_hash_dentry,
@@ -88,29 +94,29 @@ affs_intl_hash_dentry(struct dentry *dentry, struct qstr *qstr)
 	return __affs_hash_dentry(dentry, qstr, affs_intl_toupper);
 }
 
-static inline int
-__affs_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b, toupper_t toupper)
+static inline int __affs_compare_dentry(unsigned int len,
+		const char *str, const struct qstr *name, toupper_t toupper)
 {
-	const u8 *aname = a->name;
-	const u8 *bname = b->name;
-	int len;
+	const u8 *aname = str;
+	const u8 *bname = name->name;
 
-	/* 'a' is the qstr of an already existing dentry, so the name
-	 * must be valid. 'b' must be validated first.
+	/*
+	 * 'str' is the name of an already existing dentry, so the name
+	 * must be valid. 'name' must be validated first.
 	 */
 
-	if (affs_check_name(b->name,b->len))
+	if (affs_check_name(name->name, name->len))
 		return 1;
 
-	/* If the names are longer than the allowed 30 chars,
+	/*
+	 * If the names are longer than the allowed 30 chars,
 	 * the excess is ignored, so their length may differ.
 	 */
-	len = a->len;
 	if (len >= 30) {
-		if (b->len < 30)
+		if (name->len < 30)
 			return 1;
 		len = 30;
-	} else if (len != b->len)
+	} else if (len != name->len)
 		return 1;
 
 	for (; len > 0; len--)
@@ -121,14 +127,18 @@ __affs_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b, tou
 }
 
 static int
-affs_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b)
+affs_compare_dentry(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	return __affs_compare_dentry(dentry, a, b, affs_toupper);
+	return __affs_compare_dentry(len, str, name, affs_toupper);
 }
 static int
-affs_intl_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b)
+affs_intl_compare_dentry(const struct dentry *parent,const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	return __affs_compare_dentry(dentry, a, b, affs_intl_toupper);
+	return __affs_compare_dentry(len, str, name, affs_intl_toupper);
 }
 
 /*
diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index 521d841..c60133f 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -715,13 +715,15 @@ static int cifs_ci_hash(struct dentry *dentry, struct qstr *q)
 	return 0;
 }
 
-static int cifs_ci_compare(struct dentry *dentry, struct qstr *a,
-			   struct qstr *b)
+static int cifs_ci_compare(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	struct nls_table *codepage = CIFS_SB(dentry->d_inode->i_sb)->local_nls;
+	struct nls_table *codepage = CIFS_SB(pinode->i_sb)->local_nls;
 
-	if ((a->len == b->len) &&
-	    (nls_strnicmp(codepage, a->name, b->name, a->len) == 0))
+	if ((name->len == len) &&
+	    (nls_strnicmp(codepage, name->name, str, len) == 0))
 		return 0;
 	return 1;
 }
diff --git a/fs/dcache.c b/fs/dcache.c
index 814e5f4..7075555 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1437,7 +1437,9 @@ struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
 		 */
 		qstr = &dentry->d_name;
 		if (parent->d_op && parent->d_op->d_compare) {
-			if (parent->d_op->d_compare(parent, qstr, name))
+			if (parent->d_op->d_compare(parent, parent->d_inode,
+						dentry, dentry->d_inode,
+						qstr->len, qstr->name, name))
 				goto next;
 		} else {
 			if (qstr->len != len)
diff --git a/fs/fat/namei_msdos.c b/fs/fat/namei_msdos.c
index 3345aab..99d3c7a 100644
--- a/fs/fat/namei_msdos.c
+++ b/fs/fat/namei_msdos.c
@@ -164,16 +164,18 @@ static int msdos_hash(struct dentry *dentry, struct qstr *qstr)
  * Compare two msdos names. If either of the names are invalid,
  * we fall back to doing the standard name comparison.
  */
-static int msdos_cmp(struct dentry *dentry, struct qstr *a, struct qstr *b)
+static int msdos_cmp(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	struct fat_mount_options *options = &MSDOS_SB(dentry->d_sb)->options;
+	struct fat_mount_options *options = &MSDOS_SB(parent->d_sb)->options;
 	unsigned char a_msdos_name[MSDOS_NAME], b_msdos_name[MSDOS_NAME];
 	int error;
 
-	error = msdos_format_name(a->name, a->len, a_msdos_name, options);
+	error = msdos_format_name(name->name, name->len, a_msdos_name, options);
 	if (error)
 		goto old_compare;
-	error = msdos_format_name(b->name, b->len, b_msdos_name, options);
+	error = msdos_format_name(str, len, b_msdos_name, options);
 	if (error)
 		goto old_compare;
 	error = memcmp(a_msdos_name, b_msdos_name, MSDOS_NAME);
@@ -182,8 +184,8 @@ out:
 
 old_compare:
 	error = 1;
-	if (a->len == b->len)
-		error = memcmp(a->name, b->name, a->len);
+	if (name->len == len)
+		error = memcmp(name->name, str, len);
 	goto out;
 }
 
diff --git a/fs/fat/namei_vfat.c b/fs/fat/namei_vfat.c
index b936703..95e00ab 100644
--- a/fs/fat/namei_vfat.c
+++ b/fs/fat/namei_vfat.c
@@ -85,15 +85,18 @@ static int vfat_revalidate_ci(struct dentry *dentry, struct nameidata *nd)
 }
 
 /* returns the length of a struct qstr, ignoring trailing dots */
-static unsigned int vfat_striptail_len(struct qstr *qstr)
+static unsigned int __vfat_striptail_len(unsigned int len, const char *name)
 {
-	unsigned int len = qstr->len;
-
-	while (len && qstr->name[len - 1] == '.')
+	while (len && name[len - 1] == '.')
 		len--;
 	return len;
 }
 
+static unsigned int vfat_striptail_len(const struct qstr *qstr)
+{
+	return __vfat_striptail_len(qstr->len, qstr->name);
+}
+
 /*
  * Compute the hash for the vfat name corresponding to the dentry.
  * Note: if the name is invalid, we leave the hash code unchanged so
@@ -133,16 +136,18 @@ static int vfat_hashi(struct dentry *dentry, struct qstr *qstr)
 /*
  * Case insensitive compare of two vfat names.
  */
-static int vfat_cmpi(struct dentry *dentry, struct qstr *a, struct qstr *b)
+static int vfat_cmpi(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	struct nls_table *t = MSDOS_SB(dentry->d_inode->i_sb)->nls_io;
+	struct nls_table *t = MSDOS_SB(parent->d_sb)->nls_io;
 	unsigned int alen, blen;
 
 	/* A filename cannot end in '.' or we treat it like it has none */
-	alen = vfat_striptail_len(a);
-	blen = vfat_striptail_len(b);
+	alen = vfat_striptail_len(name);
+	blen = __vfat_striptail_len(len, str);
 	if (alen == blen) {
-		if (nls_strnicmp(t, a->name, b->name, alen) == 0)
+		if (nls_strnicmp(t, name->name, str, alen) == 0)
 			return 0;
 	}
 	return 1;
@@ -151,15 +156,17 @@ static int vfat_cmpi(struct dentry *dentry, struct qstr *a, struct qstr *b)
 /*
  * Case sensitive compare of two vfat names.
  */
-static int vfat_cmp(struct dentry *dentry, struct qstr *a, struct qstr *b)
+static int vfat_cmp(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
 	unsigned int alen, blen;
 
 	/* A filename cannot end in '.' or we treat it like it has none */
-	alen = vfat_striptail_len(a);
-	blen = vfat_striptail_len(b);
+	alen = vfat_striptail_len(name);
+	blen = __vfat_striptail_len(len, str);
 	if (alen == blen) {
-		if (strncmp(a->name, b->name, alen) == 0)
+		if (strncmp(name->name, str, alen) == 0)
 			return 0;
 	}
 	return 1;
diff --git a/fs/hfs/hfs_fs.h b/fs/hfs/hfs_fs.h
index c8cffb8..8cd876f 100644
--- a/fs/hfs/hfs_fs.h
+++ b/fs/hfs/hfs_fs.h
@@ -216,7 +216,10 @@ extern const struct dentry_operations hfs_dentry_operations;
 extern int hfs_hash_dentry(struct dentry *, struct qstr *);
 extern int hfs_strcmp(const unsigned char *, unsigned int,
 		      const unsigned char *, unsigned int);
-extern int hfs_compare_dentry(struct dentry *, struct qstr *, struct qstr *);
+extern int hfs_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
 
 /* trans.c */
 extern void hfs_asc2mac(struct super_block *, struct hfs_name *, struct qstr *);
diff --git a/fs/hfs/string.c b/fs/hfs/string.c
index 927a5af..aaf90d0 100644
--- a/fs/hfs/string.c
+++ b/fs/hfs/string.c
@@ -92,21 +92,21 @@ int hfs_strcmp(const unsigned char *s1, unsigned int len1,
  * Test for equality of two strings in the HFS filename character ordering.
  * return 1 on failure and 0 on success
  */
-int hfs_compare_dentry(struct dentry *dentry, struct qstr *s1, struct qstr *s2)
+int hfs_compare_dentry(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
 	const unsigned char *n1, *n2;
-	int len;
 
-	len = s1->len;
 	if (len >= HFS_NAMELEN) {
-		if (s2->len < HFS_NAMELEN)
+		if (name->len < HFS_NAMELEN)
 			return 1;
 		len = HFS_NAMELEN;
-	} else if (len != s2->len)
+	} else if (len != name->len)
 		return 1;
 
-	n1 = s1->name;
-	n2 = s2->name;
+	n1 = str;
+	n2 = name->name;
 	while (len--) {
 		if (caseorder[*n1++] != caseorder[*n2++])
 			return 1;
diff --git a/fs/hfsplus/hfsplus_fs.h b/fs/hfsplus/hfsplus_fs.h
index cb3653e..7aa96ee 100644
--- a/fs/hfsplus/hfsplus_fs.h
+++ b/fs/hfsplus/hfsplus_fs.h
@@ -380,7 +380,10 @@ int hfsplus_strcmp(const struct hfsplus_unistr *, const struct hfsplus_unistr *)
 int hfsplus_uni2asc(struct super_block *, const struct hfsplus_unistr *, char *, int *);
 int hfsplus_asc2uni(struct super_block *, struct hfsplus_unistr *, const char *, int);
 int hfsplus_hash_dentry(struct dentry *dentry, struct qstr *str);
-int hfsplus_compare_dentry(struct dentry *dentry, struct qstr *s1, struct qstr *s2);
+int hfsplus_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
 
 /* wrapper.c */
 int hfsplus_read_wrapper(struct super_block *);
diff --git a/fs/hfsplus/unicode.c b/fs/hfsplus/unicode.c
index b66d67d..b178c99 100644
--- a/fs/hfsplus/unicode.c
+++ b/fs/hfsplus/unicode.c
@@ -363,9 +363,12 @@ int hfsplus_hash_dentry(struct dentry *dentry, struct qstr *str)
  * Composed unicode characters are decomposed and case-folding is performed
  * if the appropriate bits are (un)set on the superblock.
  */
-int hfsplus_compare_dentry(struct dentry *dentry, struct qstr *s1, struct qstr *s2)
+int hfsplus_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	struct super_block *sb = dentry->d_sb;
+	struct super_block *sb = parent->d_sb;
 	int casefold, decompose, size;
 	int dsize1, dsize2, len1, len2;
 	const u16 *dstr1, *dstr2;
@@ -375,10 +378,10 @@ int hfsplus_compare_dentry(struct dentry *dentry, struct qstr *s1, struct qstr *
 
 	casefold = test_bit(HFSPLUS_SB_CASEFOLD, &HFSPLUS_SB(sb)->flags);
 	decompose = !test_bit(HFSPLUS_SB_NODECOMPOSE, &HFSPLUS_SB(sb)->flags);
-	astr1 = s1->name;
-	len1 = s1->len;
-	astr2 = s2->name;
-	len2 = s2->len;
+	astr1 = str;
+	len1 = len;
+	astr2 = name->name;
+	len2 = name->len;
 	dsize1 = dsize2 = 0;
 	dstr1 = dstr2 = NULL;
 
diff --git a/fs/hpfs/dentry.c b/fs/hpfs/dentry.c
index 67d9d36..dd9b1e7 100644
--- a/fs/hpfs/dentry.c
+++ b/fs/hpfs/dentry.c
@@ -34,19 +34,25 @@ static int hpfs_hash_dentry(struct dentry *dentry, struct qstr *qstr)
 	return 0;
 }
 
-static int hpfs_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b)
+static int hpfs_compare_dentry(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	unsigned al=a->len;
-	unsigned bl=b->len;
-	hpfs_adjust_length(a->name, &al);
+	unsigned al = len;
+	unsigned bl = name->len;
+
+	hpfs_adjust_length(str, &al);
 	/*hpfs_adjust_length(b->name, &bl);*/
-	/* 'a' is the qstr of an already existing dentry, so the name
-	 * must be valid. 'b' must be validated first.
+
+	/*
+	 * 'str' is the nane of an already existing dentry, so the name
+	 * must be valid. 'name' must be validated first.
 	 */
 
-	if (hpfs_chk_name(b->name, &bl))
+	if (hpfs_chk_name(name->name, &bl))
 		return 1;
-	if (hpfs_compare_names(dentry->d_sb, a->name, al, b->name, bl, 0))
+	if (hpfs_compare_names(parent->d_sb, str, al, name->name, bl, 0))
 		return 1;
 	return 0;
 }
diff --git a/fs/isofs/inode.c b/fs/isofs/inode.c
index bfdeb82..7b0fbc6 100644
--- a/fs/isofs/inode.c
+++ b/fs/isofs/inode.c
@@ -28,14 +28,26 @@
 
 static int isofs_hashi(struct dentry *parent, struct qstr *qstr);
 static int isofs_hash(struct dentry *parent, struct qstr *qstr);
-static int isofs_dentry_cmpi(struct dentry *dentry, struct qstr *a, struct qstr *b);
-static int isofs_dentry_cmp(struct dentry *dentry, struct qstr *a, struct qstr *b);
+static int isofs_dentry_cmpi(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
+static int isofs_dentry_cmp(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
 
 #ifdef CONFIG_JOLIET
 static int isofs_hashi_ms(struct dentry *parent, struct qstr *qstr);
 static int isofs_hash_ms(struct dentry *parent, struct qstr *qstr);
-static int isofs_dentry_cmpi_ms(struct dentry *dentry, struct qstr *a, struct qstr *b);
-static int isofs_dentry_cmp_ms(struct dentry *dentry, struct qstr *a, struct qstr *b);
+static int isofs_dentry_cmpi_ms(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
+static int isofs_dentry_cmp_ms(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name);
 #endif
 
 static void isofs_put_super(struct super_block *sb)
@@ -206,49 +218,31 @@ isofs_hashi_common(struct dentry *dentry, struct qstr *qstr, int ms)
 }
 
 /*
- * Case insensitive compare of two isofs names.
+ * Compare of two isofs names.
  */
-static int isofs_dentry_cmpi_common(struct dentry *dentry, struct qstr *a,
-				struct qstr *b, int ms)
+static int isofs_dentry_cmp_common(
+		unsigned int len, const char *str,
+		const struct qstr *name, int ms, int ci)
 {
 	int alen, blen;
 
 	/* A filename cannot end in '.' or we treat it like it has none */
-	alen = a->len;
-	blen = b->len;
+	alen = name->len;
+	blen = len;
 	if (ms) {
-		while (alen && a->name[alen-1] == '.')
+		while (alen && name->name[alen-1] == '.')
 			alen--;
-		while (blen && b->name[blen-1] == '.')
+		while (blen && str[blen-1] == '.')
 			blen--;
 	}
 	if (alen == blen) {
-		if (strnicmp(a->name, b->name, alen) == 0)
-			return 0;
-	}
-	return 1;
-}
-
-/*
- * Case sensitive compare of two isofs names.
- */
-static int isofs_dentry_cmp_common(struct dentry *dentry, struct qstr *a,
-					struct qstr *b, int ms)
-{
-	int alen, blen;
-
-	/* A filename cannot end in '.' or we treat it like it has none */
-	alen = a->len;
-	blen = b->len;
-	if (ms) {
-		while (alen && a->name[alen-1] == '.')
-			alen--;
-		while (blen && b->name[blen-1] == '.')
-			blen--;
-	}
-	if (alen == blen) {
-		if (strncmp(a->name, b->name, alen) == 0)
-			return 0;
+		if (ci) {
+			if (strnicmp(name->name, str, alen) == 0)
+				return 0;
+		} else {
+			if (strncmp(name->name, str, alen) == 0)
+				return 0;
+		}
 	}
 	return 1;
 }
@@ -266,15 +260,19 @@ isofs_hashi(struct dentry *dentry, struct qstr *qstr)
 }
 
 static int
-isofs_dentry_cmp(struct dentry *dentry,struct qstr *a,struct qstr *b)
+isofs_dentry_cmp(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	return isofs_dentry_cmp_common(dentry, a, b, 0);
+	return isofs_dentry_cmp_common(len, str, name, 0, 0);
 }
 
 static int
-isofs_dentry_cmpi(struct dentry *dentry,struct qstr *a,struct qstr *b)
+isofs_dentry_cmpi(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	return isofs_dentry_cmpi_common(dentry, a, b, 0);
+	return isofs_dentry_cmp_common(len, str, name, 0, 1);
 }
 
 #ifdef CONFIG_JOLIET
@@ -291,15 +289,19 @@ isofs_hashi_ms(struct dentry *dentry, struct qstr *qstr)
 }
 
 static int
-isofs_dentry_cmp_ms(struct dentry *dentry,struct qstr *a,struct qstr *b)
+isofs_dentry_cmp_ms(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	return isofs_dentry_cmp_common(dentry, a, b, 1);
+	return isofs_dentry_cmp_common(len, str, name, 1, 0);
 }
 
 static int
-isofs_dentry_cmpi_ms(struct dentry *dentry,struct qstr *a,struct qstr *b)
+isofs_dentry_cmpi_ms(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	return isofs_dentry_cmpi_common(dentry, a, b, 1);
+	return isofs_dentry_cmp_common(len, str, name, 1, 1);
 }
 #endif
 
diff --git a/fs/isofs/namei.c b/fs/isofs/namei.c
index 0d23abf..715f7d3 100644
--- a/fs/isofs/namei.c
+++ b/fs/isofs/namei.c
@@ -37,7 +37,8 @@ isofs_cmp(struct dentry *dentry, const char *compare, int dlen)
 
 	qstr.name = compare;
 	qstr.len = dlen;
-	return dentry->d_op->d_compare(dentry, &dentry->d_name, &qstr);
+	return dentry->d_op->d_compare(NULL, NULL, NULL, NULL,
+			dentry->d_name.len, dentry->d_name.name, &qstr);
 }
 
 /*
diff --git a/fs/jfs/namei.c b/fs/jfs/namei.c
index 2da1546..9212901 100644
--- a/fs/jfs/namei.c
+++ b/fs/jfs/namei.c
@@ -1587,14 +1587,17 @@ static int jfs_ci_hash(struct dentry *dir, struct qstr *this)
 	return 0;
 }
 
-static int jfs_ci_compare(struct dentry *dir, struct qstr *a, struct qstr *b)
+static int jfs_ci_compare(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
 	int i, result = 1;
 
-	if (a->len != b->len)
+	if (len != name->len)
 		goto out;
-	for (i=0; i < a->len; i++) {
-		if (tolower(a->name[i]) != tolower(b->name[i]))
+	for (i=0; i < len; i++) {
+		if (tolower(str[i]) != tolower(name->name[i]))
 			goto out;
 	}
 	result = 0;
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index e80ea4e..3bcc68a 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -75,7 +75,9 @@ const struct inode_operations ncp_dir_inode_operations =
  */
 static int ncp_lookup_validate(struct dentry *, struct nameidata *);
 static int ncp_hash_dentry(struct dentry *, struct qstr *);
-static int ncp_compare_dentry (struct dentry *, struct qstr *, struct qstr *);
+static int ncp_compare_dentry(const struct dentry *, const struct inode *,
+		const struct dentry *, const struct inode *,
+		unsigned int, const char *, const struct qstr *);
 static int ncp_delete_dentry(const struct dentry *);
 
 static const struct dentry_operations ncp_dentry_operations =
@@ -113,10 +115,10 @@ static inline int ncp_preserve_entry_case(struct inode *i, __u32 nscreator)
 
 #define ncp_preserve_case(i)	(ncp_namespace(i) != NW_NS_DOS)
 
-static inline int ncp_case_sensitive(struct dentry *dentry)
+static inline int ncp_case_sensitive(const struct inode *i)
 {
 #ifdef CONFIG_NCPFS_NFS_NS
-	return ncp_namespace(dentry->d_inode) == NW_NS_NFS;
+	return ncp_namespace(i) == NW_NS_NFS;
 #else
 	return 0;
 #endif /* CONFIG_NCPFS_NFS_NS */
@@ -129,12 +131,13 @@ static inline int ncp_case_sensitive(struct dentry *dentry)
 static int 
 ncp_hash_dentry(struct dentry *dentry, struct qstr *this)
 {
-	if (!ncp_case_sensitive(dentry)) {
+	if (!ncp_case_sensitive(dentry->d_inode)) {
+		struct super_block *sb = dentry->d_sb;
 		struct nls_table *t;
 		unsigned long hash;
 		int i;
 
-		t = NCP_IO_TABLE(dentry);
+		t = NCP_IO_TABLE(sb);
 		hash = init_name_hash();
 		for (i=0; i<this->len ; i++)
 			hash = partial_name_hash(ncp_tolower(t, this->name[i]),
@@ -145,15 +148,17 @@ ncp_hash_dentry(struct dentry *dentry, struct qstr *this)
 }
 
 static int
-ncp_compare_dentry(struct dentry *dentry, struct qstr *a, struct qstr *b)
+ncp_compare_dentry(const struct dentry *parent, const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	if (a->len != b->len)
+	if (len != name->len)
 		return 1;
 
-	if (ncp_case_sensitive(dentry))
-		return strncmp(a->name, b->name, a->len);
+	if (ncp_case_sensitive(pinode))
+		return strncmp(str, name->name, len);
 
-	return ncp_strnicmp(NCP_IO_TABLE(dentry), a->name, b->name, a->len);
+	return ncp_strnicmp(NCP_IO_TABLE(pinode->i_sb), str, name->name, len);
 }
 
 /*
diff --git a/fs/ncpfs/ncplib_kernel.h b/fs/ncpfs/ncplib_kernel.h
index 3c57eca..244d1b7 100644
--- a/fs/ncpfs/ncplib_kernel.h
+++ b/fs/ncpfs/ncplib_kernel.h
@@ -135,7 +135,7 @@ int ncp__vol2io(struct ncp_server *, unsigned char *, unsigned int *,
 				const unsigned char *, unsigned int, int);
 
 #define NCP_ESC			':'
-#define NCP_IO_TABLE(dentry)	(NCP_SERVER((dentry)->d_inode)->nls_io)
+#define NCP_IO_TABLE(sb)	(NCP_SBP(sb)->nls_io)
 #define ncp_tolower(t, c)	nls_tolower(t, c)
 #define ncp_toupper(t, c)	nls_toupper(t, c)
 #define ncp_strnicmp(t, s1, s2, len) \
@@ -150,15 +150,15 @@ int ncp__io2vol(unsigned char *, unsigned int *,
 int ncp__vol2io(unsigned char *, unsigned int *,
 				const unsigned char *, unsigned int, int);
 
-#define NCP_IO_TABLE(dentry)	NULL
+#define NCP_IO_TABLE(sb)	NULL
 #define ncp_tolower(t, c)	tolower(c)
 #define ncp_toupper(t, c)	toupper(c)
 #define ncp_io2vol(S,m,i,n,k,U)	ncp__io2vol(m,i,n,k,U)
 #define ncp_vol2io(S,m,i,n,k,U)	ncp__vol2io(m,i,n,k,U)
 
 
-static inline int ncp_strnicmp(struct nls_table *t, const unsigned char *s1,
-		const unsigned char *s2, int len)
+static inline int ncp_strnicmp(const struct nls_table *t,
+		const unsigned char *s1, const unsigned char *s2, int len)
 {
 	while (len--) {
 		if (tolower(*s1++) != tolower(*s2++))
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a256d77..ae4b0fd 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -397,15 +397,16 @@ static int proc_sys_delete(const struct dentry *dentry)
 	return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
 }
 
-static int proc_sys_compare(struct dentry *dir, struct qstr *qstr,
-			    struct qstr *name)
+static int proc_sys_compare(const struct dentry *parent,
+		const struct inode *pinode,
+		const struct dentry *dentry, const struct inode *inode,
+		unsigned int len, const char *str, const struct qstr *name)
 {
-	struct dentry *dentry = container_of(qstr, struct dentry, d_name);
-	if (qstr->len != name->len)
+	if (name->len != len)
 		return 1;
-	if (memcmp(qstr->name, name->name, name->len))
+	if (memcmp(name->name, str, len))
 		return 1;
-	return !sysctl_is_seen(PROC_I(dentry->d_inode)->sysctl);
+	return !sysctl_is_seen(PROC_I(inode)->sysctl);
 }
 
 static const struct dentry_operations proc_sys_dentry_operations = {
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 6cdf499..75a072b 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -134,7 +134,9 @@ enum dentry_d_lock_class
 struct dentry_operations {
 	int (*d_revalidate)(struct dentry *, struct nameidata *);
 	int (*d_hash)(struct dentry *, struct qstr *);
-	int (*d_compare)(struct dentry *, struct qstr *, struct qstr *);
+	int (*d_compare)(const struct dentry *, const struct inode *,
+			const struct dentry *, const struct inode *,
+			unsigned int, const char *, const struct qstr *);
 	int (*d_delete)(const struct dentry *);
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
@@ -145,12 +147,8 @@ struct dentry_operations {
  * Locking rules for dentry_operations callbacks are to be found in
  * Documentation/filesystems/Locking. Keep it updated!
  *
- * the dentry parameter passed to d_hash and d_compare is the parent
- * directory of the entries to be compared. It is used in case these
- * functions need any directory specific information for determining
- * equivalency classes.  Using the dentry itself might not work, as it
- * might be a negative dentry which has no information associated with
- * it.
+ * FUrther descriptions are found in Documentation/filesystems/vfs.txt.
+ * Keep it updated too!
  */
 
 /* d_flags entries */
diff --git a/include/linux/ncp_fs.h b/include/linux/ncp_fs.h
index ef66306..1c27f20 100644
--- a/include/linux/ncp_fs.h
+++ b/include/linux/ncp_fs.h
@@ -184,13 +184,13 @@ struct ncp_entry_info {
 	__u8			file_handle[6];
 };
 
-static inline struct ncp_server *NCP_SBP(struct super_block *sb)
+static inline struct ncp_server *NCP_SBP(const struct super_block *sb)
 {
 	return sb->s_fs_info;
 }
 
 #define NCP_SERVER(inode)	NCP_SBP((inode)->i_sb)
-static inline struct ncp_inode_info *NCP_FINFO(struct inode *inode)
+static inline struct ncp_inode_info *NCP_FINFO(const struct inode *inode)
 {
 	return container_of(inode, struct ncp_inode_info, vfs_inode);
 }
-- 
cgit v0.10.2


From b1e6a015a580ad145689ad1d6b4aa0e03e6c868b Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:28 +1100
Subject: fs: change d_hash for rcu-walk

Change d_hash so it may be called from lock-free RCU lookups. See similar
patch for d_compare for details.

For in-tree filesystems, this is just a mechanical change.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index 9a76f8d..a15ee20 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -10,7 +10,8 @@ be able to use diff(1).
 --------------------------- dentry_operations --------------------------
 prototypes:
 	int (*d_revalidate)(struct dentry *, int);
-	int (*d_hash) (struct dentry *, struct qstr *);
+	int (*d_hash)(const struct dentry *, const struct inode *,
+			struct qstr *);
 	int (*d_compare)(const struct dentry *, const struct inode *,
 			const struct dentry *, const struct inode *,
 			unsigned int, const char *, const struct qstr *);
@@ -22,7 +23,7 @@ prototypes:
 locking rules:
 		dcache_lock	rename_lock	->d_lock	may block
 d_revalidate:	no		no		no		yes
-d_hash		no		no		no		yes
+d_hash		no		no		no		no
 d_compare:	no		yes		no		no 
 d_delete:	yes		no		yes		no
 d_release:	no		no		no		yes
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index d44511e..9fd3194 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -333,3 +333,10 @@ unreferenced dentries, and is now only called when the dentry refcount goes to
 	.d_compare() calling convention and locking rules are significantly
 changed. Read updated documentation in Documentation/filesystems/vfs.txt (and
 look at examples of other filesystems) for guidance.
+
+---
+[mandatory]
+
+	.d_hash() calling convention and locking rules are significantly
+changed. Read updated documentation in Documentation/filesystems/vfs.txt (and
+look at examples of other filesystems) for guidance.
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index 250681b..69b10ff 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -847,7 +847,8 @@ defined:
 
 struct dentry_operations {
 	int (*d_revalidate)(struct dentry *, struct nameidata *);
-	int (*d_hash)(struct dentry *, struct qstr *);
+	int (*d_hash)(const struct dentry *, const struct inode *,
+			struct qstr *);
 	int (*d_compare)(const struct dentry *, const struct inode *,
 			const struct dentry *, const struct inode *,
 			unsigned int, const char *, const struct qstr *);
@@ -864,7 +865,10 @@ struct dentry_operations {
 
   d_hash: called when the VFS adds a dentry to the hash table. The first
 	dentry passed to d_hash is the parent directory that the name is
-	to be hashed into.
+	to be hashed into. The inode is the dentry's inode.
+
+	Same locking and synchronisation rules as d_compare regarding
+	what is safe to dereference etc.
 
   d_compare: called to compare a dentry name with a given name. The first
 	dentry is the parent of the dentry to be compared, the second is
diff --git a/drivers/staging/smbfs/cache.c b/drivers/staging/smbfs/cache.c
index dbd2e1d..0beded2 100644
--- a/drivers/staging/smbfs/cache.c
+++ b/drivers/staging/smbfs/cache.c
@@ -134,7 +134,7 @@ smb_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
 	qname->hash = full_name_hash(qname->name, qname->len);
 
 	if (dentry->d_op && dentry->d_op->d_hash)
-		if (dentry->d_op->d_hash(dentry, qname) != 0)
+		if (dentry->d_op->d_hash(dentry, inode, qname) != 0)
 			goto end_advance;
 
 	newdent = d_lookup(dentry, qname);
diff --git a/drivers/staging/smbfs/dir.c b/drivers/staging/smbfs/dir.c
index bd63229..5f79799 100644
--- a/drivers/staging/smbfs/dir.c
+++ b/drivers/staging/smbfs/dir.c
@@ -274,7 +274,8 @@ smb_dir_open(struct inode *dir, struct file *file)
  * Dentry operations routines
  */
 static int smb_lookup_validate(struct dentry *, struct nameidata *);
-static int smb_hash_dentry(struct dentry *, struct qstr *);
+static int smb_hash_dentry(const struct dentry *, const struct inode *,
+		struct qstr *);
 static int smb_compare_dentry(const struct dentry *,
 		const struct inode *,
 		const struct dentry *, const struct inode *,
@@ -336,7 +337,8 @@ smb_lookup_validate(struct dentry * dentry, struct nameidata *nd)
 }
 
 static int 
-smb_hash_dentry(struct dentry *dir, struct qstr *this)
+smb_hash_dentry(const struct dentry *dir, const struct inode *inode,
+		struct qstr *this)
 {
 	unsigned long hash;
 	int i;
diff --git a/fs/adfs/dir.c b/fs/adfs/dir.c
index c8ed661..a11e5e1 100644
--- a/fs/adfs/dir.c
+++ b/fs/adfs/dir.c
@@ -201,7 +201,8 @@ const struct file_operations adfs_dir_operations = {
 };
 
 static int
-adfs_hash(struct dentry *parent, struct qstr *qstr)
+adfs_hash(const struct dentry *parent, const struct inode *inode,
+		struct qstr *qstr)
 {
 	const unsigned int name_len = ADFS_SB(parent->d_sb)->s_namelen;
 	const unsigned char *name;
diff --git a/fs/affs/namei.c b/fs/affs/namei.c
index 547d5de..5aca08c 100644
--- a/fs/affs/namei.c
+++ b/fs/affs/namei.c
@@ -13,13 +13,15 @@
 typedef int (*toupper_t)(int);
 
 static int	 affs_toupper(int ch);
-static int	 affs_hash_dentry(struct dentry *, struct qstr *);
+static int	 affs_hash_dentry(const struct dentry *,
+		const struct inode *, struct qstr *);
 static int       affs_compare_dentry(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
 		unsigned int len, const char *str, const struct qstr *name);
 static int	 affs_intl_toupper(int ch);
-static int	 affs_intl_hash_dentry(struct dentry *, struct qstr *);
+static int	 affs_intl_hash_dentry(const struct dentry *,
+		const struct inode *, struct qstr *);
 static int       affs_intl_compare_dentry(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
@@ -64,13 +66,13 @@ affs_get_toupper(struct super_block *sb)
  * Note: the dentry argument is the parent dentry.
  */
 static inline int
-__affs_hash_dentry(struct dentry *dentry, struct qstr *qstr, toupper_t toupper)
+__affs_hash_dentry(struct qstr *qstr, toupper_t toupper)
 {
 	const u8 *name = qstr->name;
 	unsigned long hash;
 	int i;
 
-	i = affs_check_name(qstr->name,qstr->len);
+	i = affs_check_name(qstr->name, qstr->len);
 	if (i)
 		return i;
 
@@ -84,14 +86,16 @@ __affs_hash_dentry(struct dentry *dentry, struct qstr *qstr, toupper_t toupper)
 }
 
 static int
-affs_hash_dentry(struct dentry *dentry, struct qstr *qstr)
+affs_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
-	return __affs_hash_dentry(dentry, qstr, affs_toupper);
+	return __affs_hash_dentry(qstr, affs_toupper);
 }
 static int
-affs_intl_hash_dentry(struct dentry *dentry, struct qstr *qstr)
+affs_intl_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
-	return __affs_hash_dentry(dentry, qstr, affs_intl_toupper);
+	return __affs_hash_dentry(qstr, affs_intl_toupper);
 }
 
 static inline int __affs_compare_dentry(unsigned int len,
diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index c60133f..88bfe68 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -700,9 +700,10 @@ const struct dentry_operations cifs_dentry_ops = {
 /* d_delete:       cifs_d_delete,      */ /* not needed except for debugging */
 };
 
-static int cifs_ci_hash(struct dentry *dentry, struct qstr *q)
+static int cifs_ci_hash(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *q)
 {
-	struct nls_table *codepage = CIFS_SB(dentry->d_inode->i_sb)->local_nls;
+	struct nls_table *codepage = CIFS_SB(dentry->d_sb)->local_nls;
 	unsigned long hash;
 	int i;
 
diff --git a/fs/cifs/readdir.c b/fs/cifs/readdir.c
index a73eb9f..ee463ae 100644
--- a/fs/cifs/readdir.c
+++ b/fs/cifs/readdir.c
@@ -79,7 +79,7 @@ cifs_readdir_lookup(struct dentry *parent, struct qstr *name,
 	cFYI(1, "For %s", name->name);
 
 	if (parent->d_op && parent->d_op->d_hash)
-		parent->d_op->d_hash(parent, name);
+		parent->d_op->d_hash(parent, parent->d_inode, name);
 	else
 		name->hash = full_name_hash(name->name, name->len);
 
diff --git a/fs/dcache.c b/fs/dcache.c
index 7075555..6f59f37 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1478,7 +1478,7 @@ struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
 	 */
 	name->hash = full_name_hash(name->name, name->len);
 	if (dir->d_op && dir->d_op->d_hash) {
-		if (dir->d_op->d_hash(dir, name) < 0)
+		if (dir->d_op->d_hash(dir, dir->d_inode, name) < 0)
 			goto out;
 	}
 	dentry = d_lookup(dir, name);
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index 9d1a22d..a1ed7a7 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -454,7 +454,7 @@ static struct dentry *ecryptfs_lookup(struct inode *ecryptfs_dir_inode,
 	lower_name.hash = ecryptfs_dentry->d_name.hash;
 	if (lower_dir_dentry->d_op && lower_dir_dentry->d_op->d_hash) {
 		rc = lower_dir_dentry->d_op->d_hash(lower_dir_dentry,
-						    &lower_name);
+				lower_dir_dentry->d_inode, &lower_name);
 		if (rc < 0)
 			goto out_d_drop;
 	}
@@ -489,7 +489,7 @@ static struct dentry *ecryptfs_lookup(struct inode *ecryptfs_dir_inode,
 	lower_name.hash = full_name_hash(lower_name.name, lower_name.len);
 	if (lower_dir_dentry->d_op && lower_dir_dentry->d_op->d_hash) {
 		rc = lower_dir_dentry->d_op->d_hash(lower_dir_dentry,
-						    &lower_name);
+				lower_dir_dentry->d_inode, &lower_name);
 		if (rc < 0)
 			goto out_d_drop;
 	}
diff --git a/fs/fat/namei_msdos.c b/fs/fat/namei_msdos.c
index 99d3c7a..3b3e072 100644
--- a/fs/fat/namei_msdos.c
+++ b/fs/fat/namei_msdos.c
@@ -148,7 +148,8 @@ static int msdos_find(struct inode *dir, const unsigned char *name, int len,
  * that the existing dentry can be used. The msdos fs routines will
  * return ENOENT or EINVAL as appropriate.
  */
-static int msdos_hash(struct dentry *dentry, struct qstr *qstr)
+static int msdos_hash(const struct dentry *dentry, const struct inode *inode,
+	       struct qstr *qstr)
 {
 	struct fat_mount_options *options = &MSDOS_SB(dentry->d_sb)->options;
 	unsigned char msdos_name[MSDOS_NAME];
diff --git a/fs/fat/namei_vfat.c b/fs/fat/namei_vfat.c
index 95e00ab..4fc0627 100644
--- a/fs/fat/namei_vfat.c
+++ b/fs/fat/namei_vfat.c
@@ -103,7 +103,8 @@ static unsigned int vfat_striptail_len(const struct qstr *qstr)
  * that the existing dentry can be used. The vfat fs routines will
  * return ENOENT or EINVAL as appropriate.
  */
-static int vfat_hash(struct dentry *dentry, struct qstr *qstr)
+static int vfat_hash(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	qstr->hash = full_name_hash(qstr->name, vfat_striptail_len(qstr));
 	return 0;
@@ -115,9 +116,10 @@ static int vfat_hash(struct dentry *dentry, struct qstr *qstr)
  * that the existing dentry can be used. The vfat fs routines will
  * return ENOENT or EINVAL as appropriate.
  */
-static int vfat_hashi(struct dentry *dentry, struct qstr *qstr)
+static int vfat_hashi(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
-	struct nls_table *t = MSDOS_SB(dentry->d_inode->i_sb)->nls_io;
+	struct nls_table *t = MSDOS_SB(dentry->d_sb)->nls_io;
 	const unsigned char *name;
 	unsigned int len;
 	unsigned long hash;
diff --git a/fs/gfs2/dentry.c b/fs/gfs2/dentry.c
index e80fea2..50497f6 100644
--- a/fs/gfs2/dentry.c
+++ b/fs/gfs2/dentry.c
@@ -100,7 +100,8 @@ fail:
 	return 0;
 }
 
-static int gfs2_dhash(struct dentry *dentry, struct qstr *str)
+static int gfs2_dhash(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *str)
 {
 	str->hash = gfs2_disk_hash(str->name, str->len);
 	return 0;
diff --git a/fs/hfs/hfs_fs.h b/fs/hfs/hfs_fs.h
index 8cd876f..ad97c2d 100644
--- a/fs/hfs/hfs_fs.h
+++ b/fs/hfs/hfs_fs.h
@@ -213,7 +213,8 @@ extern int hfs_part_find(struct super_block *, sector_t *, sector_t *);
 /* string.c */
 extern const struct dentry_operations hfs_dentry_operations;
 
-extern int hfs_hash_dentry(struct dentry *, struct qstr *);
+extern int hfs_hash_dentry(const struct dentry *, const struct inode *,
+		struct qstr *);
 extern int hfs_strcmp(const unsigned char *, unsigned int,
 		      const unsigned char *, unsigned int);
 extern int hfs_compare_dentry(const struct dentry *parent,
diff --git a/fs/hfs/string.c b/fs/hfs/string.c
index aaf90d0..495a976 100644
--- a/fs/hfs/string.c
+++ b/fs/hfs/string.c
@@ -51,7 +51,8 @@ static unsigned char caseorder[256] = {
 /*
  * Hash a string to an integer in a case-independent way
  */
-int hfs_hash_dentry(struct dentry *dentry, struct qstr *this)
+int hfs_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *this)
 {
 	const unsigned char *name = this->name;
 	unsigned int hash, len = this->len;
diff --git a/fs/hfsplus/hfsplus_fs.h b/fs/hfsplus/hfsplus_fs.h
index 7aa96ee..a5308f4 100644
--- a/fs/hfsplus/hfsplus_fs.h
+++ b/fs/hfsplus/hfsplus_fs.h
@@ -379,7 +379,8 @@ int hfsplus_strcasecmp(const struct hfsplus_unistr *, const struct hfsplus_unist
 int hfsplus_strcmp(const struct hfsplus_unistr *, const struct hfsplus_unistr *);
 int hfsplus_uni2asc(struct super_block *, const struct hfsplus_unistr *, char *, int *);
 int hfsplus_asc2uni(struct super_block *, struct hfsplus_unistr *, const char *, int);
-int hfsplus_hash_dentry(struct dentry *dentry, struct qstr *str);
+int hfsplus_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *str);
 int hfsplus_compare_dentry(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
diff --git a/fs/hfsplus/unicode.c b/fs/hfsplus/unicode.c
index b178c99..d800aa0 100644
--- a/fs/hfsplus/unicode.c
+++ b/fs/hfsplus/unicode.c
@@ -320,7 +320,8 @@ int hfsplus_asc2uni(struct super_block *sb, struct hfsplus_unistr *ustr,
  * Composed unicode characters are decomposed and case-folding is performed
  * if the appropriate bits are (un)set on the superblock.
  */
-int hfsplus_hash_dentry(struct dentry *dentry, struct qstr *str)
+int hfsplus_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *str)
 {
 	struct super_block *sb = dentry->d_sb;
 	const char *astr;
diff --git a/fs/hpfs/dentry.c b/fs/hpfs/dentry.c
index dd9b1e7..35526df 100644
--- a/fs/hpfs/dentry.c
+++ b/fs/hpfs/dentry.c
@@ -12,7 +12,8 @@
  * Note: the dentry argument is the parent dentry.
  */
 
-static int hpfs_hash_dentry(struct dentry *dentry, struct qstr *qstr)
+static int hpfs_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	unsigned long	 hash;
 	int		 i;
diff --git a/fs/isofs/inode.c b/fs/isofs/inode.c
index 7b0fbc6..d204ee4 100644
--- a/fs/isofs/inode.c
+++ b/fs/isofs/inode.c
@@ -26,8 +26,10 @@
 
 #define BEQUIET
 
-static int isofs_hashi(struct dentry *parent, struct qstr *qstr);
-static int isofs_hash(struct dentry *parent, struct qstr *qstr);
+static int isofs_hashi(const struct dentry *parent, const struct inode *inode,
+		struct qstr *qstr);
+static int isofs_hash(const struct dentry *parent, const struct inode *inode,
+		struct qstr *qstr);
 static int isofs_dentry_cmpi(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
@@ -38,8 +40,10 @@ static int isofs_dentry_cmp(const struct dentry *parent,
 		unsigned int len, const char *str, const struct qstr *name);
 
 #ifdef CONFIG_JOLIET
-static int isofs_hashi_ms(struct dentry *parent, struct qstr *qstr);
-static int isofs_hash_ms(struct dentry *parent, struct qstr *qstr);
+static int isofs_hashi_ms(const struct dentry *parent, const struct inode *inode,
+		struct qstr *qstr);
+static int isofs_hash_ms(const struct dentry *parent, const struct inode *inode,
+		struct qstr *qstr);
 static int isofs_dentry_cmpi_ms(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
@@ -172,7 +176,7 @@ struct iso9660_options{
  * Compute the hash for the isofs name corresponding to the dentry.
  */
 static int
-isofs_hash_common(struct dentry *dentry, struct qstr *qstr, int ms)
+isofs_hash_common(const struct dentry *dentry, struct qstr *qstr, int ms)
 {
 	const char *name;
 	int len;
@@ -193,7 +197,7 @@ isofs_hash_common(struct dentry *dentry, struct qstr *qstr, int ms)
  * Compute the hash for the isofs name corresponding to the dentry.
  */
 static int
-isofs_hashi_common(struct dentry *dentry, struct qstr *qstr, int ms)
+isofs_hashi_common(const struct dentry *dentry, struct qstr *qstr, int ms)
 {
 	const char *name;
 	int len;
@@ -248,13 +252,15 @@ static int isofs_dentry_cmp_common(
 }
 
 static int
-isofs_hash(struct dentry *dentry, struct qstr *qstr)
+isofs_hash(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	return isofs_hash_common(dentry, qstr, 0);
 }
 
 static int
-isofs_hashi(struct dentry *dentry, struct qstr *qstr)
+isofs_hashi(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	return isofs_hashi_common(dentry, qstr, 0);
 }
@@ -277,13 +283,15 @@ isofs_dentry_cmpi(const struct dentry *parent, const struct inode *pinode,
 
 #ifdef CONFIG_JOLIET
 static int
-isofs_hash_ms(struct dentry *dentry, struct qstr *qstr)
+isofs_hash_ms(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	return isofs_hash_common(dentry, qstr, 1);
 }
 
 static int
-isofs_hashi_ms(struct dentry *dentry, struct qstr *qstr)
+isofs_hashi_ms(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	return isofs_hashi_common(dentry, qstr, 1);
 }
diff --git a/fs/jfs/namei.c b/fs/jfs/namei.c
index 9212901..57f90da 100644
--- a/fs/jfs/namei.c
+++ b/fs/jfs/namei.c
@@ -1574,7 +1574,8 @@ const struct file_operations jfs_dir_operations = {
 	.llseek		= generic_file_llseek,
 };
 
-static int jfs_ci_hash(struct dentry *dir, struct qstr *this)
+static int jfs_ci_hash(const struct dentry *dir, const struct inode *inode,
+		struct qstr *this)
 {
 	unsigned long hash;
 	int i;
diff --git a/fs/namei.c b/fs/namei.c
index 4ff7ca5..f3b5ca4 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -731,7 +731,8 @@ static int do_lookup(struct nameidata *nd, struct qstr *name,
 	 * to use its own hash..
 	 */
 	if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) {
-		int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, name);
+		int err = nd->path.dentry->d_op->d_hash(nd->path.dentry,
+				nd->path.dentry->d_inode, name);
 		if (err < 0)
 			return err;
 	}
@@ -1134,7 +1135,7 @@ static struct dentry *__lookup_hash(struct qstr *name,
 	 * to use its own hash..
 	 */
 	if (base->d_op && base->d_op->d_hash) {
-		err = base->d_op->d_hash(base, name);
+		err = base->d_op->d_hash(base, inode, name);
 		dentry = ERR_PTR(err);
 		if (err < 0)
 			goto out;
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index 3bcc68a..bbbf7922 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -74,7 +74,8 @@ const struct inode_operations ncp_dir_inode_operations =
  * Dentry operations routines
  */
 static int ncp_lookup_validate(struct dentry *, struct nameidata *);
-static int ncp_hash_dentry(struct dentry *, struct qstr *);
+static int ncp_hash_dentry(const struct dentry *, const struct inode *,
+		struct qstr *);
 static int ncp_compare_dentry(const struct dentry *, const struct inode *,
 		const struct dentry *, const struct inode *,
 		unsigned int, const char *, const struct qstr *);
@@ -129,9 +130,10 @@ static inline int ncp_case_sensitive(const struct inode *i)
  * is case-sensitive.
  */
 static int 
-ncp_hash_dentry(struct dentry *dentry, struct qstr *this)
+ncp_hash_dentry(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *this)
 {
-	if (!ncp_case_sensitive(dentry->d_inode)) {
+	if (!ncp_case_sensitive(inode)) {
 		struct super_block *sb = dentry->d_sb;
 		struct nls_table *t;
 		unsigned long hash;
@@ -597,7 +599,7 @@ ncp_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
 	qname.hash = full_name_hash(qname.name, qname.len);
 
 	if (dentry->d_op && dentry->d_op->d_hash)
-		if (dentry->d_op->d_hash(dentry, &qname) != 0)
+		if (dentry->d_op->d_hash(dentry, dentry->d_inode, &qname) != 0)
 			goto end_advance;
 
 	newdent = d_lookup(dentry, &qname);
diff --git a/fs/sysv/namei.c b/fs/sysv/namei.c
index 11e7f7d..7507aeb 100644
--- a/fs/sysv/namei.c
+++ b/fs/sysv/namei.c
@@ -27,7 +27,8 @@ static int add_nondir(struct dentry *dentry, struct inode *inode)
 	return err;
 }
 
-static int sysv_hash(struct dentry *dentry, struct qstr *qstr)
+static int sysv_hash(const struct dentry *dentry, const struct inode *inode,
+		struct qstr *qstr)
 {
 	/* Truncate the name in place, avoids having to define a compare
 	   function. */
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 75a072b..1149e70 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -133,7 +133,8 @@ enum dentry_d_lock_class
 
 struct dentry_operations {
 	int (*d_revalidate)(struct dentry *, struct nameidata *);
-	int (*d_hash)(struct dentry *, struct qstr *);
+	int (*d_hash)(const struct dentry *, const struct inode *,
+			struct qstr *);
 	int (*d_compare)(const struct dentry *, const struct inode *,
 			const struct dentry *, const struct inode *,
 			unsigned int, const char *, const struct qstr *);
-- 
cgit v0.10.2


From ec2447c278ee973d35f38e53ca16ba7f965ae33d Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:29 +1100
Subject: hostfs: simplify locking

Remove dcache_lock locking from hostfs filesystem, and move it into dcache
helpers. All that is required is a coherent path name. Protection from
concurrent modification of the namespace after path name generation is not
provided in current code, because dcache_lock is dropped before the path is
used.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 6f59f37..61beb40 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -2171,7 +2171,7 @@ char *dynamic_dname(struct dentry *dentry, char *buffer, int buflen,
 /*
  * Write full pathname from the root of the filesystem into the buffer.
  */
-char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
+static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
 {
 	char *end = buf + buflen;
 	char *retval;
@@ -2198,7 +2198,18 @@ char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
 Elong:
 	return ERR_PTR(-ENAMETOOLONG);
 }
-EXPORT_SYMBOL(__dentry_path);
+
+char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
+{
+	char *retval;
+
+	spin_lock(&dcache_lock);
+	retval = __dentry_path(dentry, buf, buflen);
+	spin_unlock(&dcache_lock);
+
+	return retval;
+}
+EXPORT_SYMBOL(dentry_path_raw);
 
 char *dentry_path(struct dentry *dentry, char *buf, int buflen)
 {
diff --git a/fs/hostfs/hostfs_kern.c b/fs/hostfs/hostfs_kern.c
index cfe8bc7..39dc505 100644
--- a/fs/hostfs/hostfs_kern.c
+++ b/fs/hostfs/hostfs_kern.c
@@ -92,12 +92,10 @@ __uml_setup("hostfs=", hostfs_args,
 
 static char *__dentry_name(struct dentry *dentry, char *name)
 {
-	char *p = __dentry_path(dentry, name, PATH_MAX);
+	char *p = dentry_path_raw(dentry, name, PATH_MAX);
 	char *root;
 	size_t len;
 
-	spin_unlock(&dcache_lock);
-
 	root = dentry->d_sb->s_fs_info;
 	len = strlen(root);
 	if (IS_ERR(p)) {
@@ -123,25 +121,23 @@ static char *dentry_name(struct dentry *dentry)
 	if (!name)
 		return NULL;
 
-	spin_lock(&dcache_lock);
 	return __dentry_name(dentry, name); /* will unlock */
 }
 
 static char *inode_name(struct inode *ino)
 {
 	struct dentry *dentry;
-	char *name = __getname();
-	if (!name)
-		return NULL;
+	char *name;
 
-	spin_lock(&dcache_lock);
-	if (list_empty(&ino->i_dentry)) {
-		spin_unlock(&dcache_lock);
-		__putname(name);
+	dentry = d_find_alias(ino);
+	if (!dentry)
 		return NULL;
-	}
-	dentry = list_first_entry(&ino->i_dentry, struct dentry, d_alias);
-	return __dentry_name(dentry, name); /* will unlock */
+
+	name = dentry_name(dentry);
+
+	dput(dentry);
+
+	return name;
 }
 
 static char *follow_link(char *link)
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 1149e70..cea27df 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -311,7 +311,7 @@ extern char *dynamic_dname(struct dentry *, char *, int, const char *, ...);
 extern char *__d_path(const struct path *path, struct path *root, char *, int);
 extern char *d_path(const struct path *, char *, int);
 extern char *d_path_with_unreachable(const struct path *, char *, int);
-extern char *__dentry_path(struct dentry *, char *, int);
+extern char *dentry_path_raw(struct dentry *, char *, int);
 extern char *dentry_path(struct dentry *, char *, int);
 
 /* Allocation counts.. */
-- 
cgit v0.10.2


From 789680d1ee9311cdf095241dc02bd9784d799cd1 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:30 +1100
Subject: fs: dcache scale hash

Add a new lock, dcache_hash_lock, to protect the dcache hash table from
concurrent modification. d_hash is also protected by d_lock.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 61beb40..1e124d4 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -35,10 +35,24 @@
 #include <linux/hardirq.h>
 #include "internal.h"
 
+/*
+ * Usage:
+ * dcache_hash_lock protects dcache hash table, s_anon lists
+ *
+ * Ordering:
+ * dcache_lock
+ *   dentry->d_lock
+ *     dcache_hash_lock
+ *
+ * if (dentry1 < dentry2)
+ *   dentry1->d_lock
+ *     dentry2->d_lock
+ */
 int sysctl_vfs_cache_pressure __read_mostly = 100;
 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
 
- __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
+static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_hash_lock);
+__cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
 EXPORT_SYMBOL(dcache_lock);
@@ -196,6 +210,42 @@ static struct dentry *d_kill(struct dentry *dentry)
 	return parent;
 }
 
+/**
+ * d_drop - drop a dentry
+ * @dentry: dentry to drop
+ *
+ * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
+ * be found through a VFS lookup any more. Note that this is different from
+ * deleting the dentry - d_delete will try to mark the dentry negative if
+ * possible, giving a successful _negative_ lookup, while d_drop will
+ * just make the cache lookup fail.
+ *
+ * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
+ * reason (NFS timeouts or autofs deletes).
+ *
+ * __d_drop requires dentry->d_lock.
+ */
+void __d_drop(struct dentry *dentry)
+{
+	if (!(dentry->d_flags & DCACHE_UNHASHED)) {
+		dentry->d_flags |= DCACHE_UNHASHED;
+		spin_lock(&dcache_hash_lock);
+		hlist_del_rcu(&dentry->d_hash);
+		spin_unlock(&dcache_hash_lock);
+	}
+}
+EXPORT_SYMBOL(__d_drop);
+
+void d_drop(struct dentry *dentry)
+{
+	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
+	__d_drop(dentry);
+	spin_unlock(&dentry->d_lock);
+	spin_unlock(&dcache_lock);
+}
+EXPORT_SYMBOL(d_drop);
+
 /* 
  * This is dput
  *
@@ -1199,7 +1249,9 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	tmp->d_flags |= DCACHE_DISCONNECTED;
 	tmp->d_flags &= ~DCACHE_UNHASHED;
 	list_add(&tmp->d_alias, &inode->i_dentry);
+	spin_lock(&dcache_hash_lock);
 	hlist_add_head(&tmp->d_hash, &inode->i_sb->s_anon);
+	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&tmp->d_lock);
 
 	spin_unlock(&dcache_lock);
@@ -1585,7 +1637,9 @@ void d_rehash(struct dentry * entry)
 {
 	spin_lock(&dcache_lock);
 	spin_lock(&entry->d_lock);
+	spin_lock(&dcache_hash_lock);
 	_d_rehash(entry);
+	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&entry->d_lock);
 	spin_unlock(&dcache_lock);
 }
@@ -1692,8 +1746,6 @@ static void switch_names(struct dentry *dentry, struct dentry *target)
  */
 static void d_move_locked(struct dentry * dentry, struct dentry * target)
 {
-	struct hlist_head *list;
-
 	if (!dentry->d_inode)
 		printk(KERN_WARNING "VFS: moving negative dcache entry\n");
 
@@ -1710,14 +1762,11 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	}
 
 	/* Move the dentry to the target hash queue, if on different bucket */
-	if (d_unhashed(dentry))
-		goto already_unhashed;
-
-	hlist_del_rcu(&dentry->d_hash);
-
-already_unhashed:
-	list = d_hash(target->d_parent, target->d_name.hash);
-	__d_rehash(dentry, list);
+	spin_lock(&dcache_hash_lock);
+	if (!d_unhashed(dentry))
+		hlist_del_rcu(&dentry->d_hash);
+	__d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
+	spin_unlock(&dcache_hash_lock);
 
 	/* Unhash the target: dput() will then get rid of it */
 	__d_drop(target);
@@ -1914,7 +1963,9 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 found_lock:
 	spin_lock(&actual->d_lock);
 found:
+	spin_lock(&dcache_hash_lock);
 	_d_rehash(actual);
+	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&actual->d_lock);
 	spin_unlock(&dcache_lock);
 out_nolock:
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index cea27df..2feb624 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -184,39 +184,6 @@ struct dentry_operations {
 extern spinlock_t dcache_lock;
 extern seqlock_t rename_lock;
 
-/**
- * d_drop - drop a dentry
- * @dentry: dentry to drop
- *
- * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
- * be found through a VFS lookup any more. Note that this is different from
- * deleting the dentry - d_delete will try to mark the dentry negative if
- * possible, giving a successful _negative_ lookup, while d_drop will
- * just make the cache lookup fail.
- *
- * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
- * reason (NFS timeouts or autofs deletes).
- *
- * __d_drop requires dentry->d_lock.
- */
-
-static inline void __d_drop(struct dentry *dentry)
-{
-	if (!(dentry->d_flags & DCACHE_UNHASHED)) {
-		dentry->d_flags |= DCACHE_UNHASHED;
-		hlist_del_rcu(&dentry->d_hash);
-	}
-}
-
-static inline void d_drop(struct dentry *dentry)
-{
-	spin_lock(&dcache_lock);
-	spin_lock(&dentry->d_lock);
- 	__d_drop(dentry);
-	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
-}
-
 static inline int dname_external(struct dentry *dentry)
 {
 	return dentry->d_name.name != dentry->d_iname;
@@ -228,6 +195,8 @@ static inline int dname_external(struct dentry *dentry)
 extern void d_instantiate(struct dentry *, struct inode *);
 extern struct dentry * d_instantiate_unique(struct dentry *, struct inode *);
 extern struct dentry * d_materialise_unique(struct dentry *, struct inode *);
+extern void __d_drop(struct dentry *dentry);
+extern void d_drop(struct dentry *dentry);
 extern void d_delete(struct dentry *);
 
 /* allocate/de-allocate */
-- 
cgit v0.10.2


From 2304450783dfde7b0b94ae234edd0dbffa865073 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:31 +1100
Subject: fs: dcache scale lru

Add a new lock, dcache_lru_lock, to protect the dcache LRU list from concurrent
modification. d_lru is also protected by d_lock, which allows LRU lists to be
accessed without the lru lock, using RCU in future patches.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 1e124d4..3d3c843 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -37,11 +37,19 @@
 
 /*
  * Usage:
- * dcache_hash_lock protects dcache hash table, s_anon lists
+ * dcache_hash_lock protects:
+ *   - the dcache hash table, s_anon lists
+ * dcache_lru_lock protects:
+ *   - the dcache lru lists and counters
+ * d_lock protects:
+ *   - d_flags
+ *   - d_name
+ *   - d_lru
  *
  * Ordering:
  * dcache_lock
  *   dentry->d_lock
+ *     dcache_lru_lock
  *     dcache_hash_lock
  *
  * if (dentry1 < dentry2)
@@ -52,6 +60,7 @@ int sysctl_vfs_cache_pressure __read_mostly = 100;
 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
 
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_hash_lock);
+static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
@@ -154,28 +163,38 @@ static void dentry_iput(struct dentry * dentry)
 }
 
 /*
- * dentry_lru_(add|del|move_tail) must be called with dcache_lock held.
+ * dentry_lru_(add|del|move_tail) must be called with d_lock held.
  */
 static void dentry_lru_add(struct dentry *dentry)
 {
 	if (list_empty(&dentry->d_lru)) {
+		spin_lock(&dcache_lru_lock);
 		list_add(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
 		dentry->d_sb->s_nr_dentry_unused++;
 		dentry_stat.nr_unused++;
+		spin_unlock(&dcache_lru_lock);
 	}
 }
 
+static void __dentry_lru_del(struct dentry *dentry)
+{
+	list_del_init(&dentry->d_lru);
+	dentry->d_sb->s_nr_dentry_unused--;
+	dentry_stat.nr_unused--;
+}
+
 static void dentry_lru_del(struct dentry *dentry)
 {
 	if (!list_empty(&dentry->d_lru)) {
-		list_del_init(&dentry->d_lru);
-		dentry->d_sb->s_nr_dentry_unused--;
-		dentry_stat.nr_unused--;
+		spin_lock(&dcache_lru_lock);
+		__dentry_lru_del(dentry);
+		spin_unlock(&dcache_lru_lock);
 	}
 }
 
 static void dentry_lru_move_tail(struct dentry *dentry)
 {
+	spin_lock(&dcache_lru_lock);
 	if (list_empty(&dentry->d_lru)) {
 		list_add_tail(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
 		dentry->d_sb->s_nr_dentry_unused++;
@@ -183,6 +202,7 @@ static void dentry_lru_move_tail(struct dentry *dentry)
 	} else {
 		list_move_tail(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
 	}
+	spin_unlock(&dcache_lru_lock);
 }
 
 /**
@@ -192,6 +212,8 @@ static void dentry_lru_move_tail(struct dentry *dentry)
  * The dentry must already be unhashed and removed from the LRU.
  *
  * If this is the root of the dentry tree, return NULL.
+ *
+ * dcache_lock and d_lock must be held by caller, are dropped by d_kill.
  */
 static struct dentry *d_kill(struct dentry *dentry)
 	__releases(dentry->d_lock)
@@ -383,10 +405,19 @@ int d_invalidate(struct dentry * dentry)
 EXPORT_SYMBOL(d_invalidate);
 
 /* This should be called _only_ with dcache_lock held */
+static inline struct dentry * __dget_locked_dlock(struct dentry *dentry)
+{
+	atomic_inc(&dentry->d_count);
+	dentry_lru_del(dentry);
+	return dentry;
+}
+
 static inline struct dentry * __dget_locked(struct dentry *dentry)
 {
 	atomic_inc(&dentry->d_count);
+	spin_lock(&dentry->d_lock);
 	dentry_lru_del(dentry);
+	spin_unlock(&dentry->d_lock);
 	return dentry;
 }
 
@@ -465,7 +496,7 @@ restart:
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		spin_lock(&dentry->d_lock);
 		if (!atomic_read(&dentry->d_count)) {
-			__dget_locked(dentry);
+			__dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
@@ -489,7 +520,6 @@ EXPORT_SYMBOL(d_prune_aliases);
 static void prune_one_dentry(struct dentry * dentry)
 	__releases(dentry->d_lock)
 	__releases(dcache_lock)
-	__acquires(dcache_lock)
 {
 	__d_drop(dentry);
 	dentry = d_kill(dentry);
@@ -498,15 +528,16 @@ static void prune_one_dentry(struct dentry * dentry)
 	 * Prune ancestors.  Locking is simpler than in dput(),
 	 * because dcache_lock needs to be taken anyway.
 	 */
-	spin_lock(&dcache_lock);
 	while (dentry) {
-		if (!atomic_dec_and_lock(&dentry->d_count, &dentry->d_lock))
+		spin_lock(&dcache_lock);
+		if (!atomic_dec_and_lock(&dentry->d_count, &dentry->d_lock)) {
+			spin_unlock(&dcache_lock);
 			return;
+		}
 
 		dentry_lru_del(dentry);
 		__d_drop(dentry);
 		dentry = d_kill(dentry);
-		spin_lock(&dcache_lock);
 	}
 }
 
@@ -516,21 +547,31 @@ static void shrink_dentry_list(struct list_head *list)
 
 	while (!list_empty(list)) {
 		dentry = list_entry(list->prev, struct dentry, d_lru);
-		dentry_lru_del(dentry);
+
+		if (!spin_trylock(&dentry->d_lock)) {
+			spin_unlock(&dcache_lru_lock);
+			cpu_relax();
+			spin_lock(&dcache_lru_lock);
+			continue;
+		}
+
+		__dentry_lru_del(dentry);
 
 		/*
 		 * We found an inuse dentry which was not removed from
 		 * the LRU because of laziness during lookup.  Do not free
 		 * it - just keep it off the LRU list.
 		 */
-		spin_lock(&dentry->d_lock);
 		if (atomic_read(&dentry->d_count)) {
 			spin_unlock(&dentry->d_lock);
 			continue;
 		}
+		spin_unlock(&dcache_lru_lock);
+
 		prune_one_dentry(dentry);
-		/* dentry->d_lock was dropped in prune_one_dentry() */
-		cond_resched_lock(&dcache_lock);
+		/* dcache_lock and dentry->d_lock dropped */
+		spin_lock(&dcache_lock);
+		spin_lock(&dcache_lru_lock);
 	}
 }
 
@@ -551,32 +592,36 @@ static void __shrink_dcache_sb(struct super_block *sb, int *count, int flags)
 	int cnt = *count;
 
 	spin_lock(&dcache_lock);
+relock:
+	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
 		dentry = list_entry(sb->s_dentry_lru.prev,
 				struct dentry, d_lru);
 		BUG_ON(dentry->d_sb != sb);
 
+		if (!spin_trylock(&dentry->d_lock)) {
+			spin_unlock(&dcache_lru_lock);
+			cpu_relax();
+			goto relock;
+		}
+
 		/*
 		 * If we are honouring the DCACHE_REFERENCED flag and the
 		 * dentry has this flag set, don't free it.  Clear the flag
 		 * and put it back on the LRU.
 		 */
-		if (flags & DCACHE_REFERENCED) {
-			spin_lock(&dentry->d_lock);
-			if (dentry->d_flags & DCACHE_REFERENCED) {
-				dentry->d_flags &= ~DCACHE_REFERENCED;
-				list_move(&dentry->d_lru, &referenced);
-				spin_unlock(&dentry->d_lock);
-				cond_resched_lock(&dcache_lock);
-				continue;
-			}
+		if (flags & DCACHE_REFERENCED &&
+				dentry->d_flags & DCACHE_REFERENCED) {
+			dentry->d_flags &= ~DCACHE_REFERENCED;
+			list_move(&dentry->d_lru, &referenced);
+			spin_unlock(&dentry->d_lock);
+		} else {
+			list_move_tail(&dentry->d_lru, &tmp);
 			spin_unlock(&dentry->d_lock);
+			if (!--cnt)
+				break;
 		}
-
-		list_move_tail(&dentry->d_lru, &tmp);
-		if (!--cnt)
-			break;
-		cond_resched_lock(&dcache_lock);
+		/* XXX: re-add cond_resched_lock when dcache_lock goes away */
 	}
 
 	*count = cnt;
@@ -584,6 +629,7 @@ static void __shrink_dcache_sb(struct super_block *sb, int *count, int flags)
 
 	if (!list_empty(&referenced))
 		list_splice(&referenced, &sb->s_dentry_lru);
+	spin_unlock(&dcache_lru_lock);
 	spin_unlock(&dcache_lock);
 
 }
@@ -679,10 +725,12 @@ void shrink_dcache_sb(struct super_block *sb)
 	LIST_HEAD(tmp);
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
 		list_splice_init(&sb->s_dentry_lru, &tmp);
 		shrink_dentry_list(&tmp);
 	}
+	spin_unlock(&dcache_lru_lock);
 	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(shrink_dcache_sb);
@@ -701,7 +749,9 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 
 	/* detach this root from the system */
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	dentry_lru_del(dentry);
+	spin_unlock(&dentry->d_lock);
 	__d_drop(dentry);
 	spin_unlock(&dcache_lock);
 
@@ -715,7 +765,9 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 			spin_lock(&dcache_lock);
 			list_for_each_entry(loop, &dentry->d_subdirs,
 					    d_u.d_child) {
+				spin_lock(&loop->d_lock);
 				dentry_lru_del(loop);
+				spin_unlock(&loop->d_lock);
 				__d_drop(loop);
 				cond_resched_lock(&dcache_lock);
 			}
@@ -892,6 +944,8 @@ resume:
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
 
+		spin_lock(&dentry->d_lock);
+
 		/* 
 		 * move only zero ref count dentries to the end 
 		 * of the unused list for prune_dcache
@@ -903,6 +957,8 @@ resume:
 			dentry_lru_del(dentry);
 		}
 
+		spin_unlock(&dentry->d_lock);
+
 		/*
 		 * We can return to the caller if we have found some (this
 		 * ensures forward progress). We'll be coming back to find
-- 
cgit v0.10.2


From b7ab39f631f505edc2bbdb86620d5493f995c9da Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:32 +1100
Subject: fs: dcache scale dentry refcount

Make d_count non-atomic and protect it with d_lock. This allows us to ensure a
0 refcount dentry remains 0 without dcache_lock. It is also fairly natural when
we start protecting many other dentry members with d_lock.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/arch/powerpc/platforms/cell/spufs/inode.c b/arch/powerpc/platforms/cell/spufs/inode.c
index 3532b92..29a406a 100644
--- a/arch/powerpc/platforms/cell/spufs/inode.c
+++ b/arch/powerpc/platforms/cell/spufs/inode.c
@@ -162,7 +162,7 @@ static void spufs_prune_dir(struct dentry *dir)
 		spin_lock(&dcache_lock);
 		spin_lock(&dentry->d_lock);
 		if (!(d_unhashed(dentry)) && dentry->d_inode) {
-			dget_locked(dentry);
+			dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			simple_unlink(dir->d_inode, dentry);
diff --git a/drivers/infiniband/hw/ipath/ipath_fs.c b/drivers/infiniband/hw/ipath/ipath_fs.c
index 8c8afc7..18aee04 100644
--- a/drivers/infiniband/hw/ipath/ipath_fs.c
+++ b/drivers/infiniband/hw/ipath/ipath_fs.c
@@ -280,7 +280,7 @@ static int remove_file(struct dentry *parent, char *name)
 	spin_lock(&dcache_lock);
 	spin_lock(&tmp->d_lock);
 	if (!(d_unhashed(tmp) && tmp->d_inode)) {
-		dget_locked(tmp);
+		dget_locked_dlock(tmp);
 		__d_drop(tmp);
 		spin_unlock(&tmp->d_lock);
 		spin_unlock(&dcache_lock);
diff --git a/drivers/infiniband/hw/qib/qib_fs.c b/drivers/infiniband/hw/qib/qib_fs.c
index f99bddc..fe4b242 100644
--- a/drivers/infiniband/hw/qib/qib_fs.c
+++ b/drivers/infiniband/hw/qib/qib_fs.c
@@ -456,7 +456,7 @@ static int remove_file(struct dentry *parent, char *name)
 	spin_lock(&dcache_lock);
 	spin_lock(&tmp->d_lock);
 	if (!(d_unhashed(tmp) && tmp->d_inode)) {
-		dget_locked(tmp);
+		dget_locked_dlock(tmp);
 		__d_drop(tmp);
 		spin_unlock(&tmp->d_lock);
 		spin_unlock(&dcache_lock);
diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index a796c94..413b564 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -198,7 +198,7 @@ static int autofs4_tree_busy(struct vfsmount *mnt,
 			else
 				ino_count++;
 
-			if (atomic_read(&p->d_count) > ino_count) {
+			if (p->d_count > ino_count) {
 				top_ino->last_used = jiffies;
 				dput(p);
 				return 1;
@@ -347,7 +347,7 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 
 			/* Path walk currently on this dentry? */
 			ino_count = atomic_read(&ino->count) + 2;
-			if (atomic_read(&dentry->d_count) > ino_count)
+			if (dentry->d_count > ino_count)
 				goto next;
 
 			/* Can we umount this guy */
@@ -369,7 +369,7 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 		if (!exp_leaves) {
 			/* Path walk currently on this dentry? */
 			ino_count = atomic_read(&ino->count) + 1;
-			if (atomic_read(&dentry->d_count) > ino_count)
+			if (dentry->d_count > ino_count)
 				goto next;
 
 			if (!autofs4_tree_busy(mnt, dentry, timeout, do_now)) {
@@ -383,7 +383,7 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 		} else {
 			/* Path walk currently on this dentry? */
 			ino_count = atomic_read(&ino->count) + 1;
-			if (atomic_read(&dentry->d_count) > ino_count)
+			if (dentry->d_count > ino_count)
 				goto next;
 
 			expired = autofs4_check_leaves(mnt, dentry, timeout, do_now);
diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
index d34896c..7922509 100644
--- a/fs/autofs4/root.c
+++ b/fs/autofs4/root.c
@@ -436,7 +436,7 @@ static struct dentry *autofs4_lookup_active(struct dentry *dentry)
 		spin_lock(&active->d_lock);
 
 		/* Already gone? */
-		if (atomic_read(&active->d_count) == 0)
+		if (active->d_count == 0)
 			goto next;
 
 		qstr = &active->d_name;
@@ -452,7 +452,7 @@ static struct dentry *autofs4_lookup_active(struct dentry *dentry)
 			goto next;
 
 		if (d_unhashed(active)) {
-			dget(active);
+			dget_dlock(active);
 			spin_unlock(&active->d_lock);
 			spin_unlock(&sbi->lookup_lock);
 			spin_unlock(&dcache_lock);
@@ -507,7 +507,7 @@ static struct dentry *autofs4_lookup_expiring(struct dentry *dentry)
 			goto next;
 
 		if (d_unhashed(expiring)) {
-			dget(expiring);
+			dget_dlock(expiring);
 			spin_unlock(&expiring->d_lock);
 			spin_unlock(&sbi->lookup_lock);
 			spin_unlock(&dcache_lock);
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index d902948..3ecf915 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -150,7 +150,9 @@ more:
 		di = ceph_dentry(dentry);
 	}
 
-	atomic_inc(&dentry->d_count);
+	spin_lock(&dentry->d_lock);
+	dentry->d_count++;
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 
 	dout(" %llu (%llu) dentry %p %.*s %p\n", di->offset, filp->f_pos,
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index bf12865..bb68c79 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -879,8 +879,8 @@ static struct dentry *splice_dentry(struct dentry *dn, struct inode *in,
 	} else if (realdn) {
 		dout("dn %p (%d) spliced with %p (%d) "
 		     "inode %p ino %llx.%llx\n",
-		     dn, atomic_read(&dn->d_count),
-		     realdn, atomic_read(&realdn->d_count),
+		     dn, dn->d_count,
+		     realdn, realdn->d_count,
 		     realdn->d_inode, ceph_vinop(realdn->d_inode));
 		dput(dn);
 		dn = realdn;
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 38800ea..a50fca1 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -1486,7 +1486,7 @@ retry:
 	*base = ceph_ino(temp->d_inode);
 	*plen = len;
 	dout("build_path on %p %d built %llx '%.*s'\n",
-	     dentry, atomic_read(&dentry->d_count), *base, len, path);
+	     dentry, dentry->d_count, *base, len, path);
 	return path;
 }
 
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index 4cce3b07..9e37e8b 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -559,7 +559,7 @@ static int coda_dentry_revalidate(struct dentry *de, struct nameidata *nd)
 	if (cii->c_flags & C_FLUSH) 
 		coda_flag_inode_children(inode, C_FLUSH);
 
-	if (atomic_read(&de->d_count) > 1)
+	if (de->d_count > 1)
 		/* pretend it's valid, but don't change the flags */
 		goto out;
 
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index 20024a9..e9acea4 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -394,8 +394,7 @@ static void remove_dir(struct dentry * d)
 	if (d->d_inode)
 		simple_rmdir(parent->d_inode,d);
 
-	pr_debug(" o %s removing done (%d)\n",d->d_name.name,
-		 atomic_read(&d->d_count));
+	pr_debug(" o %s removing done (%d)\n",d->d_name.name, d->d_count);
 
 	dput(parent);
 }
diff --git a/fs/configfs/inode.c b/fs/configfs/inode.c
index 253476d..79b3776 100644
--- a/fs/configfs/inode.c
+++ b/fs/configfs/inode.c
@@ -253,7 +253,7 @@ void configfs_drop_dentry(struct configfs_dirent * sd, struct dentry * parent)
 		spin_lock(&dcache_lock);
 		spin_lock(&dentry->d_lock);
 		if (!(d_unhashed(dentry) && dentry->d_inode)) {
-			dget_locked(dentry);
+			dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
diff --git a/fs/dcache.c b/fs/dcache.c
index 3d3c843..81e9150 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -45,6 +45,7 @@
  *   - d_flags
  *   - d_name
  *   - d_lru
+ *   - d_count
  *
  * Ordering:
  * dcache_lock
@@ -125,6 +126,7 @@ static void __d_free(struct rcu_head *head)
  */
 static void d_free(struct dentry *dentry)
 {
+	BUG_ON(dentry->d_count);
 	this_cpu_dec(nr_dentry);
 	if (dentry->d_op && dentry->d_op->d_release)
 		dentry->d_op->d_release(dentry);
@@ -222,8 +224,11 @@ static struct dentry *d_kill(struct dentry *dentry)
 	struct dentry *parent;
 
 	list_del(&dentry->d_u.d_child);
-	/*drops the locks, at that point nobody can reach this dentry */
 	dentry_iput(dentry);
+	/*
+	 * dentry_iput drops the locks, at which point nobody (except
+	 * transient RCU lookups) can reach this dentry.
+	 */
 	if (IS_ROOT(dentry))
 		parent = NULL;
 	else
@@ -303,13 +308,23 @@ void dput(struct dentry *dentry)
 		return;
 
 repeat:
-	if (atomic_read(&dentry->d_count) == 1)
+	if (dentry->d_count == 1)
 		might_sleep();
-	if (!atomic_dec_and_lock(&dentry->d_count, &dcache_lock))
-		return;
-
 	spin_lock(&dentry->d_lock);
-	if (atomic_read(&dentry->d_count)) {
+	if (dentry->d_count == 1) {
+		if (!spin_trylock(&dcache_lock)) {
+			/*
+			 * Something of a livelock possibility we could avoid
+			 * by taking dcache_lock and trying again, but we
+			 * want to reduce dcache_lock anyway so this will
+			 * get improved.
+			 */
+			spin_unlock(&dentry->d_lock);
+			goto repeat;
+		}
+	}
+	dentry->d_count--;
+	if (dentry->d_count) {
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		return;
@@ -389,7 +404,7 @@ int d_invalidate(struct dentry * dentry)
 	 * working directory or similar).
 	 */
 	spin_lock(&dentry->d_lock);
-	if (atomic_read(&dentry->d_count) > 1) {
+	if (dentry->d_count > 1) {
 		if (dentry->d_inode && S_ISDIR(dentry->d_inode->i_mode)) {
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
@@ -404,29 +419,61 @@ int d_invalidate(struct dentry * dentry)
 }
 EXPORT_SYMBOL(d_invalidate);
 
-/* This should be called _only_ with dcache_lock held */
+/* This must be called with dcache_lock and d_lock held */
 static inline struct dentry * __dget_locked_dlock(struct dentry *dentry)
 {
-	atomic_inc(&dentry->d_count);
+	dentry->d_count++;
 	dentry_lru_del(dentry);
 	return dentry;
 }
 
+/* This should be called _only_ with dcache_lock held */
 static inline struct dentry * __dget_locked(struct dentry *dentry)
 {
-	atomic_inc(&dentry->d_count);
 	spin_lock(&dentry->d_lock);
-	dentry_lru_del(dentry);
+	__dget_locked_dlock(dentry);
 	spin_unlock(&dentry->d_lock);
 	return dentry;
 }
 
+struct dentry * dget_locked_dlock(struct dentry *dentry)
+{
+	return __dget_locked_dlock(dentry);
+}
+
 struct dentry * dget_locked(struct dentry *dentry)
 {
 	return __dget_locked(dentry);
 }
 EXPORT_SYMBOL(dget_locked);
 
+struct dentry *dget_parent(struct dentry *dentry)
+{
+	struct dentry *ret;
+
+repeat:
+	spin_lock(&dentry->d_lock);
+	ret = dentry->d_parent;
+	if (!ret)
+		goto out;
+	if (dentry == ret) {
+		ret->d_count++;
+		goto out;
+	}
+	if (!spin_trylock(&ret->d_lock)) {
+		spin_unlock(&dentry->d_lock);
+		cpu_relax();
+		goto repeat;
+	}
+	BUG_ON(!ret->d_count);
+	ret->d_count++;
+	spin_unlock(&ret->d_lock);
+out:
+	spin_unlock(&dentry->d_lock);
+	return ret;
+}
+EXPORT_SYMBOL(dget_parent);
+
 /**
  * d_find_alias - grab a hashed alias of inode
  * @inode: inode in question
@@ -495,7 +542,7 @@ restart:
 	spin_lock(&dcache_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		spin_lock(&dentry->d_lock);
-		if (!atomic_read(&dentry->d_count)) {
+		if (!dentry->d_count) {
 			__dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
@@ -530,7 +577,10 @@ static void prune_one_dentry(struct dentry * dentry)
 	 */
 	while (dentry) {
 		spin_lock(&dcache_lock);
-		if (!atomic_dec_and_lock(&dentry->d_count, &dentry->d_lock)) {
+		spin_lock(&dentry->d_lock);
+		dentry->d_count--;
+		if (dentry->d_count) {
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 			return;
 		}
@@ -562,7 +612,7 @@ static void shrink_dentry_list(struct list_head *list)
 		 * the LRU because of laziness during lookup.  Do not free
 		 * it - just keep it off the LRU list.
 		 */
-		if (atomic_read(&dentry->d_count)) {
+		if (dentry->d_count) {
 			spin_unlock(&dentry->d_lock);
 			continue;
 		}
@@ -783,7 +833,7 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 		do {
 			struct inode *inode;
 
-			if (atomic_read(&dentry->d_count) != 0) {
+			if (dentry->d_count != 0) {
 				printk(KERN_ERR
 				       "BUG: Dentry %p{i=%lx,n=%s}"
 				       " still in use (%d)"
@@ -792,7 +842,7 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 				       dentry->d_inode ?
 				       dentry->d_inode->i_ino : 0UL,
 				       dentry->d_name.name,
-				       atomic_read(&dentry->d_count),
+				       dentry->d_count,
 				       dentry->d_sb->s_type->name,
 				       dentry->d_sb->s_id);
 				BUG();
@@ -802,7 +852,9 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 				parent = NULL;
 			else {
 				parent = dentry->d_parent;
-				atomic_dec(&parent->d_count);
+				spin_lock(&parent->d_lock);
+				parent->d_count--;
+				spin_unlock(&parent->d_lock);
 			}
 
 			list_del(&dentry->d_u.d_child);
@@ -853,7 +905,9 @@ void shrink_dcache_for_umount(struct super_block *sb)
 
 	dentry = sb->s_root;
 	sb->s_root = NULL;
-	atomic_dec(&dentry->d_count);
+	spin_lock(&dentry->d_lock);
+	dentry->d_count--;
+	spin_unlock(&dentry->d_lock);
 	shrink_dcache_for_umount_subtree(dentry);
 
 	while (!hlist_empty(&sb->s_anon)) {
@@ -950,7 +1004,7 @@ resume:
 		 * move only zero ref count dentries to the end 
 		 * of the unused list for prune_dcache
 		 */
-		if (!atomic_read(&dentry->d_count)) {
+		if (!dentry->d_count) {
 			dentry_lru_move_tail(dentry);
 			found++;
 		} else {
@@ -1068,7 +1122,7 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	memcpy(dname, name->name, name->len);
 	dname[name->len] = 0;
 
-	atomic_set(&dentry->d_count, 1);
+	dentry->d_count = 1;
 	dentry->d_flags = DCACHE_UNHASHED;
 	spin_lock_init(&dentry->d_lock);
 	dentry->d_inode = NULL;
@@ -1556,7 +1610,7 @@ struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
 				goto next;
 		}
 
-		atomic_inc(&dentry->d_count);
+		dentry->d_count++;
 		found = dentry;
 		spin_unlock(&dentry->d_lock);
 		break;
@@ -1653,7 +1707,7 @@ void d_delete(struct dentry * dentry)
 	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	isdir = S_ISDIR(dentry->d_inode->i_mode);
-	if (atomic_read(&dentry->d_count) == 1) {
+	if (dentry->d_count == 1) {
 		dentry->d_flags &= ~DCACHE_CANT_MOUNT;
 		dentry_iput(dentry);
 		fsnotify_nameremove(dentry, isdir);
@@ -2494,11 +2548,15 @@ resume:
 			this_parent = dentry;
 			goto repeat;
 		}
-		atomic_dec(&dentry->d_count);
+		spin_lock(&dentry->d_lock);
+		dentry->d_count--;
+		spin_unlock(&dentry->d_lock);
 	}
 	if (this_parent != root) {
 		next = this_parent->d_u.d_child.next;
-		atomic_dec(&this_parent->d_count);
+		spin_lock(&this_parent->d_lock);
+		this_parent->d_count--;
+		spin_unlock(&this_parent->d_lock);
 		this_parent = this_parent->d_parent;
 		goto resume;
 	}
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index a1ed7a7..5e5c7ec 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -260,7 +260,7 @@ int ecryptfs_lookup_and_interpose_lower(struct dentry *ecryptfs_dentry,
 				   ecryptfs_dentry->d_parent));
 	lower_inode = lower_dentry->d_inode;
 	fsstack_copy_attr_atime(ecryptfs_dir_inode, lower_dir_dentry->d_inode);
-	BUG_ON(!atomic_read(&lower_dentry->d_count));
+	BUG_ON(!lower_dentry->d_count);
 	ecryptfs_set_dentry_private(ecryptfs_dentry,
 				    kmem_cache_alloc(ecryptfs_dentry_info_cache,
 						     GFP_KERNEL));
diff --git a/fs/locks.c b/fs/locks.c
index 8729347..08415b2 100644
--- a/fs/locks.c
+++ b/fs/locks.c
@@ -1389,7 +1389,7 @@ int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
 		if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
 			goto out;
 		if ((arg == F_WRLCK)
-		    && ((atomic_read(&dentry->d_count) > 1)
+		    && ((dentry->d_count > 1)
 			|| (atomic_read(&inode->i_count) > 1)))
 			goto out;
 	}
diff --git a/fs/namei.c b/fs/namei.c
index f3b5ca4..cbfa5fb 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2133,7 +2133,7 @@ void dentry_unhash(struct dentry *dentry)
 	shrink_dcache_parent(dentry);
 	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
-	if (atomic_read(&dentry->d_count) == 2)
+	if (dentry->d_count == 2)
 		__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 9184c7c..12de824 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1720,7 +1720,7 @@ static int nfs_unlink(struct inode *dir, struct dentry *dentry)
 
 	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
-	if (atomic_read(&dentry->d_count) > 1) {
+	if (dentry->d_count > 1) {
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		/* Start asynchronous writeout of the inode */
@@ -1868,7 +1868,7 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	dfprintk(VFS, "NFS: rename(%s/%s -> %s/%s, ct=%d)\n",
 		 old_dentry->d_parent->d_name.name, old_dentry->d_name.name,
 		 new_dentry->d_parent->d_name.name, new_dentry->d_name.name,
-		 atomic_read(&new_dentry->d_count));
+		 new_dentry->d_count);
 
 	/*
 	 * For non-directories, check whether the target is busy and if so,
@@ -1886,7 +1886,7 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			rehash = new_dentry;
 		}
 
-		if (atomic_read(&new_dentry->d_count) > 2) {
+		if (new_dentry->d_count > 2) {
 			int err;
 
 			/* copy the target dentry's name */
diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c
index 7bdec85..8fe9eb4 100644
--- a/fs/nfs/unlink.c
+++ b/fs/nfs/unlink.c
@@ -496,7 +496,7 @@ nfs_sillyrename(struct inode *dir, struct dentry *dentry)
 
 	dfprintk(VFS, "NFS: silly-rename(%s/%s, ct=%d)\n",
 		dentry->d_parent->d_name.name, dentry->d_name.name,
-		atomic_read(&dentry->d_count));
+		dentry->d_count);
 	nfs_inc_stats(dir, NFSIOS_SILLYRENAME);
 
 	/*
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 184938f..3a35902 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -1756,8 +1756,7 @@ nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
 		goto out_dput_new;
 
 	if (svc_msnfs(ffhp) &&
-		((atomic_read(&odentry->d_count) > 1)
-		 || (atomic_read(&ndentry->d_count) > 1))) {
+		((odentry->d_count > 1) || (ndentry->d_count > 1))) {
 			host_err = -EPERM;
 			goto out_dput_new;
 	}
@@ -1843,7 +1842,7 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 	if (type != S_IFDIR) { /* It's UNLINK */
 #ifdef MSNFS
 		if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
-			(atomic_read(&rdentry->d_count) > 1)) {
+			(rdentry->d_count > 1)) {
 			host_err = -EPERM;
 		} else
 #endif
diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
index f804d41..d36fc7e 100644
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@@ -838,7 +838,7 @@ static int nilfs_attach_snapshot(struct super_block *s, __u64 cno,
 
 static int nilfs_tree_was_touched(struct dentry *root_dentry)
 {
-	return atomic_read(&root_dentry->d_count) > 1;
+	return root_dentry->d_count > 1;
 }
 
 /**
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 2feb624..b0ade2d 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -87,7 +87,7 @@ full_name_hash(const unsigned char *name, unsigned int len)
 #endif
 
 struct dentry {
-	atomic_t d_count;
+	unsigned int d_count;		/* protected by d_lock */
 	unsigned int d_flags;		/* protected by d_lock */
 	spinlock_t d_lock;		/* per dentry lock */
 	int d_mounted;
@@ -297,17 +297,28 @@ extern char *dentry_path(struct dentry *, char *, int);
  *	needs and they take necessary precautions) you should hold dcache_lock
  *	and call dget_locked() instead of dget().
  */
- 
+static inline struct dentry *dget_dlock(struct dentry *dentry)
+{
+	if (dentry) {
+		BUG_ON(!dentry->d_count);
+		dentry->d_count++;
+	}
+	return dentry;
+}
 static inline struct dentry *dget(struct dentry *dentry)
 {
 	if (dentry) {
-		BUG_ON(!atomic_read(&dentry->d_count));
-		atomic_inc(&dentry->d_count);
+		spin_lock(&dentry->d_lock);
+		dget_dlock(dentry);
+		spin_unlock(&dentry->d_lock);
 	}
 	return dentry;
 }
 
 extern struct dentry * dget_locked(struct dentry *);
+extern struct dentry * dget_locked_dlock(struct dentry *);
+
+extern struct dentry *dget_parent(struct dentry *dentry);
 
 /**
  *	d_unhashed -	is dentry hashed
@@ -338,16 +349,6 @@ static inline void dont_mount(struct dentry *dentry)
 	spin_unlock(&dentry->d_lock);
 }
 
-static inline struct dentry *dget_parent(struct dentry *dentry)
-{
-	struct dentry *ret;
-
-	spin_lock(&dentry->d_lock);
-	ret = dget(dentry->d_parent);
-	spin_unlock(&dentry->d_lock);
-	return ret;
-}
-
 extern void dput(struct dentry *);
 
 static inline int d_mountpoint(struct dentry *dentry)
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 746055b..eb7af39 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -3655,9 +3655,7 @@ again:
 	list_del(&cgrp->sibling);
 	cgroup_unlock_hierarchy(cgrp->root);
 
-	spin_lock(&cgrp->dentry->d_lock);
 	d = dget(cgrp->dentry);
-	spin_unlock(&d->d_lock);
 
 	cgroup_d_remove_dir(d);
 	dput(d);
-- 
cgit v0.10.2


From da5029563a0a026c64821b09e8e7b4fd81d3fe1b Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:33 +1100
Subject: fs: dcache scale d_unhashed

Protect d_unhashed(dentry) condition with d_lock. This means keeping
DCACHE_UNHASHED bit in synch with hash manipulations.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/arch/powerpc/platforms/cell/spufs/inode.c b/arch/powerpc/platforms/cell/spufs/inode.c
index 29a406a..5aef1a7 100644
--- a/arch/powerpc/platforms/cell/spufs/inode.c
+++ b/arch/powerpc/platforms/cell/spufs/inode.c
@@ -166,6 +166,9 @@ static void spufs_prune_dir(struct dentry *dir)
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			simple_unlink(dir->d_inode, dentry);
+			/* XXX: what is dcache_lock protecting here? Other
+			 * filesystems (IB, configfs) release dcache_lock
+			 * before unlink */
 			spin_unlock(&dcache_lock);
 			dput(dentry);
 		} else {
diff --git a/drivers/usb/core/inode.c b/drivers/usb/core/inode.c
index b690aa3..e3ab443 100644
--- a/drivers/usb/core/inode.c
+++ b/drivers/usb/core/inode.c
@@ -347,10 +347,13 @@ static int usbfs_empty (struct dentry *dentry)
 
 	list_for_each(list, &dentry->d_subdirs) {
 		struct dentry *de = list_entry(list, struct dentry, d_u.d_child);
+		spin_lock(&de->d_lock);
 		if (usbfs_positive(de)) {
+			spin_unlock(&de->d_lock);
 			spin_unlock(&dcache_lock);
 			return 0;
 		}
+		spin_unlock(&de->d_lock);
 	}
 
 	spin_unlock(&dcache_lock);
diff --git a/fs/autofs4/autofs_i.h b/fs/autofs4/autofs_i.h
index 3d283ab..3912dcf 100644
--- a/fs/autofs4/autofs_i.h
+++ b/fs/autofs4/autofs_i.h
@@ -254,19 +254,6 @@ static inline int simple_positive(struct dentry *dentry)
 	return dentry->d_inode && !d_unhashed(dentry);
 }
 
-static inline int __simple_empty(struct dentry *dentry)
-{
-	struct dentry *child;
-	int ret = 0;
-
-	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
-		if (simple_positive(child))
-			goto out;
-	ret = 1;
-out:
-	return ret;
-}
-
 static inline void autofs4_add_expiring(struct dentry *dentry)
 {
 	struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index 413b564..ee64020 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -160,14 +160,18 @@ static int autofs4_tree_busy(struct vfsmount *mnt,
 
 	spin_lock(&dcache_lock);
 	for (p = top; p; p = next_dentry(p, top)) {
+		spin_lock(&p->d_lock);
 		/* Negative dentry - give up */
-		if (!simple_positive(p))
+		if (!simple_positive(p)) {
+			spin_unlock(&p->d_lock);
 			continue;
+		}
 
 		DPRINTK("dentry %p %.*s",
 			p, (int) p->d_name.len, p->d_name.name);
 
-		p = dget(p);
+		p = dget_dlock(p);
+		spin_unlock(&p->d_lock);
 		spin_unlock(&dcache_lock);
 
 		/*
@@ -228,14 +232,18 @@ static struct dentry *autofs4_check_leaves(struct vfsmount *mnt,
 
 	spin_lock(&dcache_lock);
 	for (p = parent; p; p = next_dentry(p, parent)) {
+		spin_lock(&p->d_lock);
 		/* Negative dentry - give up */
-		if (!simple_positive(p))
+		if (!simple_positive(p)) {
+			spin_unlock(&p->d_lock);
 			continue;
+		}
 
 		DPRINTK("dentry %p %.*s",
 			p, (int) p->d_name.len, p->d_name.name);
 
-		p = dget(p);
+		p = dget_dlock(p);
+		spin_unlock(&p->d_lock);
 		spin_unlock(&dcache_lock);
 
 		if (d_mountpoint(p)) {
@@ -324,12 +332,15 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 		struct dentry *dentry = list_entry(next, struct dentry, d_u.d_child);
 
 		/* Negative dentry - give up */
+		spin_lock(&dentry->d_lock);
 		if (!simple_positive(dentry)) {
 			next = next->next;
+			spin_unlock(&dentry->d_lock);
 			continue;
 		}
 
-		dentry = dget(dentry);
+		dentry = dget_dlock(dentry);
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 
 		spin_lock(&sbi->fs_lock);
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 3ecf915..571f270 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -136,6 +136,7 @@ more:
 			fi->at_end = 1;
 			goto out_unlock;
 		}
+		spin_lock(&dentry->d_lock);
 		if (!d_unhashed(dentry) && dentry->d_inode &&
 		    ceph_snap(dentry->d_inode) != CEPH_SNAPDIR &&
 		    ceph_ino(dentry->d_inode) != CEPH_INO_CEPH &&
@@ -145,13 +146,13 @@ more:
 		     dentry->d_name.len, dentry->d_name.name, di->offset,
 		     filp->f_pos, d_unhashed(dentry) ? " unhashed" : "",
 		     !dentry->d_inode ? " null" : "");
+		spin_unlock(&dentry->d_lock);
 		p = p->prev;
 		dentry = list_entry(p, struct dentry, d_u.d_child);
 		di = ceph_dentry(dentry);
 	}
 
-	spin_lock(&dentry->d_lock);
-	dentry->d_count++;
+	dget_dlock(dentry);
 	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 
diff --git a/fs/configfs/configfs_internal.h b/fs/configfs/configfs_internal.h
index da6061a..e58b4c3 100644
--- a/fs/configfs/configfs_internal.h
+++ b/fs/configfs/configfs_internal.h
@@ -121,6 +121,7 @@ static inline struct config_item *configfs_get_config_item(struct dentry *dentry
 	struct config_item * item = NULL;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	if (!d_unhashed(dentry)) {
 		struct configfs_dirent * sd = dentry->d_fsdata;
 		if (sd->s_type & CONFIGFS_ITEM_LINK) {
@@ -129,6 +130,7 @@ static inline struct config_item *configfs_get_config_item(struct dentry *dentry
 		} else
 			item = config_item_get(sd->s_element);
 	}
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 
 	return item;
diff --git a/fs/dcache.c b/fs/dcache.c
index 81e9150..ee127f9 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -46,6 +46,7 @@
  *   - d_name
  *   - d_lru
  *   - d_count
+ *   - d_unhashed()
  *
  * Ordering:
  * dcache_lock
@@ -53,6 +54,13 @@
  *     dcache_lru_lock
  *     dcache_hash_lock
  *
+ * If there is an ancestor relationship:
+ * dentry->d_parent->...->d_parent->d_lock
+ *   ...
+ *     dentry->d_parent->d_lock
+ *       dentry->d_lock
+ *
+ * If no ancestor relationship:
  * if (dentry1 < dentry2)
  *   dentry1->d_lock
  *     dentry2->d_lock
@@ -379,7 +387,9 @@ int d_invalidate(struct dentry * dentry)
 	 * If it's already been dropped, return OK.
 	 */
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	if (d_unhashed(dentry)) {
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		return 0;
 	}
@@ -388,9 +398,11 @@ int d_invalidate(struct dentry * dentry)
 	 * to get rid of unused child entries.
 	 */
 	if (!list_empty(&dentry->d_subdirs)) {
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		shrink_dcache_parent(dentry);
 		spin_lock(&dcache_lock);
+		spin_lock(&dentry->d_lock);
 	}
 
 	/*
@@ -403,7 +415,6 @@ int d_invalidate(struct dentry * dentry)
 	 * we might still populate it if it was a
 	 * working directory or similar).
 	 */
-	spin_lock(&dentry->d_lock);
 	if (dentry->d_count > 1) {
 		if (dentry->d_inode && S_ISDIR(dentry->d_inode->i_mode)) {
 			spin_unlock(&dentry->d_lock);
@@ -490,35 +501,44 @@ EXPORT_SYMBOL(dget_parent);
  * any other hashed alias over that one unless @want_discon is set,
  * in which case only return an IS_ROOT, DCACHE_DISCONNECTED alias.
  */
-
-static struct dentry * __d_find_alias(struct inode *inode, int want_discon)
+static struct dentry *__d_find_alias(struct inode *inode, int want_discon)
 {
-	struct list_head *head, *next, *tmp;
-	struct dentry *alias, *discon_alias=NULL;
+	struct dentry *alias, *discon_alias;
 
-	head = &inode->i_dentry;
-	next = inode->i_dentry.next;
-	while (next != head) {
-		tmp = next;
-		next = tmp->next;
-		prefetch(next);
-		alias = list_entry(tmp, struct dentry, d_alias);
+again:
+	discon_alias = NULL;
+	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
+		spin_lock(&alias->d_lock);
  		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
 			if (IS_ROOT(alias) &&
-			    (alias->d_flags & DCACHE_DISCONNECTED))
+			    (alias->d_flags & DCACHE_DISCONNECTED)) {
 				discon_alias = alias;
-			else if (!want_discon) {
-				__dget_locked(alias);
+			} else if (!want_discon) {
+				__dget_locked_dlock(alias);
+				spin_unlock(&alias->d_lock);
+				return alias;
+			}
+		}
+		spin_unlock(&alias->d_lock);
+	}
+	if (discon_alias) {
+		alias = discon_alias;
+		spin_lock(&alias->d_lock);
+		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
+			if (IS_ROOT(alias) &&
+			    (alias->d_flags & DCACHE_DISCONNECTED)) {
+				__dget_locked_dlock(alias);
+				spin_unlock(&alias->d_lock);
 				return alias;
 			}
 		}
+		spin_unlock(&alias->d_lock);
+		goto again;
 	}
-	if (discon_alias)
-		__dget_locked(discon_alias);
-	return discon_alias;
+	return NULL;
 }
 
-struct dentry * d_find_alias(struct inode *inode)
+struct dentry *d_find_alias(struct inode *inode)
 {
 	struct dentry *de = NULL;
 
@@ -801,8 +821,8 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	dentry_lru_del(dentry);
-	spin_unlock(&dentry->d_lock);
 	__d_drop(dentry);
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 
 	for (;;) {
@@ -817,8 +837,8 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 					    d_u.d_child) {
 				spin_lock(&loop->d_lock);
 				dentry_lru_del(loop);
-				spin_unlock(&loop->d_lock);
 				__d_drop(loop);
+				spin_unlock(&loop->d_lock);
 				cond_resched_lock(&dcache_lock);
 			}
 			spin_unlock(&dcache_lock);
@@ -1863,7 +1883,10 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	/*
 	 * XXXX: do we really need to take target->d_lock?
 	 */
-	if (target < dentry) {
+	if (d_ancestor(dentry, target)) {
+		spin_lock(&dentry->d_lock);
+		spin_lock_nested(&target->d_lock, DENTRY_D_LOCK_NESTED);
+	} else if (d_ancestor(target, dentry) || target < dentry) {
 		spin_lock(&target->d_lock);
 		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 	} else {
@@ -2542,13 +2565,16 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
-		if (d_unhashed(dentry)||!dentry->d_inode)
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+		if (d_unhashed(dentry) || !dentry->d_inode) {
+			spin_unlock(&dentry->d_lock);
 			continue;
+		}
 		if (!list_empty(&dentry->d_subdirs)) {
+			spin_unlock(&dentry->d_lock);
 			this_parent = dentry;
 			goto repeat;
 		}
-		spin_lock(&dentry->d_lock);
 		dentry->d_count--;
 		spin_unlock(&dentry->d_lock);
 	}
diff --git a/fs/libfs.c b/fs/libfs.c
index b9d25d8..433e713 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -16,6 +16,11 @@
 
 #include <asm/uaccess.h>
 
+static inline int simple_positive(struct dentry *dentry)
+{
+	return dentry->d_inode && !d_unhashed(dentry);
+}
+
 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
 		   struct kstat *stat)
 {
@@ -100,8 +105,10 @@ loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
 			while (n && p != &file->f_path.dentry->d_subdirs) {
 				struct dentry *next;
 				next = list_entry(p, struct dentry, d_u.d_child);
-				if (!d_unhashed(next) && next->d_inode)
+				spin_lock(&next->d_lock);
+				if (simple_positive(next))
 					n--;
+				spin_unlock(&next->d_lock);
 				p = p->next;
 			}
 			list_add_tail(&cursor->d_u.d_child, p);
@@ -155,9 +162,13 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 			for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
 				struct dentry *next;
 				next = list_entry(p, struct dentry, d_u.d_child);
-				if (d_unhashed(next) || !next->d_inode)
+				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
+				if (!simple_positive(next)) {
+					spin_unlock(&next->d_lock);
 					continue;
+				}
 
+				spin_unlock(&next->d_lock);
 				spin_unlock(&dcache_lock);
 				if (filldir(dirent, next->d_name.name, 
 					    next->d_name.len, filp->f_pos, 
@@ -259,20 +270,20 @@ int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *den
 	return 0;
 }
 
-static inline int simple_positive(struct dentry *dentry)
-{
-	return dentry->d_inode && !d_unhashed(dentry);
-}
-
 int simple_empty(struct dentry *dentry)
 {
 	struct dentry *child;
 	int ret = 0;
 
 	spin_lock(&dcache_lock);
-	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
-		if (simple_positive(child))
+	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
+		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
+		if (simple_positive(child)) {
+			spin_unlock(&child->d_lock);
 			goto out;
+		}
+		spin_unlock(&child->d_lock);
+	}
 	ret = 1;
 out:
 	spin_unlock(&dcache_lock);
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index 895532ac..35e5f5a 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -174,13 +174,16 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 	list_for_each(p, &inode->i_dentry) {
 		dentry = list_entry(p, struct dentry, d_alias);
 
+		spin_lock(&dentry->d_lock);
 		if (ocfs2_match_dentry(dentry, parent_blkno, skip_unhashed)) {
 			mlog(0, "dentry found: %.*s\n",
 			     dentry->d_name.len, dentry->d_name.name);
 
-			dget_locked(dentry);
+			dget_locked_dlock(dentry);
+			spin_unlock(&dentry->d_lock);
 			break;
 		}
+		spin_unlock(&dentry->d_lock);
 
 		dentry = NULL;
 	}
diff --git a/security/tomoyo/realpath.c b/security/tomoyo/realpath.c
index 1d0bf8f..d1e05b0 100644
--- a/security/tomoyo/realpath.c
+++ b/security/tomoyo/realpath.c
@@ -14,6 +14,7 @@
 #include <linux/slab.h>
 #include <net/sock.h>
 #include "common.h"
+#include "../../fs/internal.h"
 
 /**
  * tomoyo_encode: Convert binary string to ascii string.
-- 
cgit v0.10.2


From 2fd6b7f50797f2e993eea59e0a0b8c6399c811dc Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:34 +1100
Subject: fs: dcache scale subdirs

Protect d_subdirs and d_child with d_lock, except in filesystems that aren't
using dcache_lock for these anyway (eg. using i_mutex).

Note: if we change the locking rule in future so that ->d_child protection is
provided only with ->d_parent->d_lock, it may allow us to reduce some locking.
But it would be an exception to an otherwise regular locking scheme, so we'd
have to see some good results. Probably not worthwhile.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/drivers/staging/smbfs/cache.c b/drivers/staging/smbfs/cache.c
index 0beded2..920434b 100644
--- a/drivers/staging/smbfs/cache.c
+++ b/drivers/staging/smbfs/cache.c
@@ -63,6 +63,7 @@ smb_invalidate_dircache_entries(struct dentry *parent)
 	struct dentry *dentry;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
 		dentry = list_entry(next, struct dentry, d_u.d_child);
@@ -70,6 +71,7 @@ smb_invalidate_dircache_entries(struct dentry *parent)
 		smb_age_dentry(server, dentry);
 		next = next->next;
 	}
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
@@ -97,6 +99,7 @@ smb_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 
 	/* If a pointer is invalid, we search the dentry. */
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
 		dent = list_entry(next, struct dentry, d_u.d_child);
@@ -111,6 +114,7 @@ smb_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 	}
 	dent = NULL;
 out_unlock:
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return dent;
 }
diff --git a/drivers/usb/core/inode.c b/drivers/usb/core/inode.c
index e3ab443..89a0e83 100644
--- a/drivers/usb/core/inode.c
+++ b/drivers/usb/core/inode.c
@@ -344,18 +344,20 @@ static int usbfs_empty (struct dentry *dentry)
 	struct list_head *list;
 
 	spin_lock(&dcache_lock);
-
+	spin_lock(&dentry->d_lock);
 	list_for_each(list, &dentry->d_subdirs) {
 		struct dentry *de = list_entry(list, struct dentry, d_u.d_child);
-		spin_lock(&de->d_lock);
+
+		spin_lock_nested(&de->d_lock, DENTRY_D_LOCK_NESTED);
 		if (usbfs_positive(de)) {
 			spin_unlock(&de->d_lock);
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 			return 0;
 		}
 		spin_unlock(&de->d_lock);
 	}
-
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 	return 1;
 }
diff --git a/fs/autofs4/autofs_i.h b/fs/autofs4/autofs_i.h
index 3912dcf..9d2ae9b 100644
--- a/fs/autofs4/autofs_i.h
+++ b/fs/autofs4/autofs_i.h
@@ -254,6 +254,17 @@ static inline int simple_positive(struct dentry *dentry)
 	return dentry->d_inode && !d_unhashed(dentry);
 }
 
+static inline void __autofs4_add_expiring(struct dentry *dentry)
+{
+	struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
+	struct autofs_info *ino = autofs4_dentry_ino(dentry);
+	if (ino) {
+		if (list_empty(&ino->expiring))
+			list_add(&ino->expiring, &sbi->expiring_list);
+	}
+	return;
+}
+
 static inline void autofs4_add_expiring(struct dentry *dentry)
 {
 	struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index ee64020..968c143 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -91,24 +91,64 @@ done:
 }
 
 /*
- * Calculate next entry in top down tree traversal.
- * From next_mnt in namespace.c - elegant.
+ * Calculate and dget next entry in top down tree traversal.
  */
-static struct dentry *next_dentry(struct dentry *p, struct dentry *root)
+static struct dentry *get_next_positive_dentry(struct dentry *prev,
+						struct dentry *root)
 {
-	struct list_head *next = p->d_subdirs.next;
+	struct list_head *next;
+	struct dentry *p, *ret;
+
+	if (prev == NULL)
+		return dget(prev);
 
+	spin_lock(&dcache_lock);
+relock:
+	p = prev;
+	spin_lock(&p->d_lock);
+again:
+	next = p->d_subdirs.next;
 	if (next == &p->d_subdirs) {
 		while (1) {
-			if (p == root)
+			struct dentry *parent;
+
+			if (p == root) {
+				spin_unlock(&p->d_lock);
+				spin_unlock(&dcache_lock);
+				dput(prev);
 				return NULL;
+			}
+
+			parent = p->d_parent;
+			if (!spin_trylock(&parent->d_lock)) {
+				spin_unlock(&p->d_lock);
+				cpu_relax();
+				goto relock;
+			}
+			spin_unlock(&p->d_lock);
 			next = p->d_u.d_child.next;
-			if (next != &p->d_parent->d_subdirs)
+			p = parent;
+			if (next != &parent->d_subdirs)
 				break;
-			p = p->d_parent;
 		}
 	}
-	return list_entry(next, struct dentry, d_u.d_child);
+	ret = list_entry(next, struct dentry, d_u.d_child);
+
+	spin_lock_nested(&ret->d_lock, DENTRY_D_LOCK_NESTED);
+	/* Negative dentry - try next */
+	if (!simple_positive(ret)) {
+		spin_unlock(&ret->d_lock);
+		p = ret;
+		goto again;
+	}
+	dget_dlock(ret);
+	spin_unlock(&ret->d_lock);
+	spin_unlock(&p->d_lock);
+	spin_unlock(&dcache_lock);
+
+	dput(prev);
+
+	return ret;
 }
 
 /*
@@ -158,22 +198,11 @@ static int autofs4_tree_busy(struct vfsmount *mnt,
 	if (!simple_positive(top))
 		return 1;
 
-	spin_lock(&dcache_lock);
-	for (p = top; p; p = next_dentry(p, top)) {
-		spin_lock(&p->d_lock);
-		/* Negative dentry - give up */
-		if (!simple_positive(p)) {
-			spin_unlock(&p->d_lock);
-			continue;
-		}
-
+	p = NULL;
+	while ((p = get_next_positive_dentry(p, top))) {
 		DPRINTK("dentry %p %.*s",
 			p, (int) p->d_name.len, p->d_name.name);
 
-		p = dget_dlock(p);
-		spin_unlock(&p->d_lock);
-		spin_unlock(&dcache_lock);
-
 		/*
 		 * Is someone visiting anywhere in the subtree ?
 		 * If there's no mount we need to check the usage
@@ -208,10 +237,7 @@ static int autofs4_tree_busy(struct vfsmount *mnt,
 				return 1;
 			}
 		}
-		dput(p);
-		spin_lock(&dcache_lock);
 	}
-	spin_unlock(&dcache_lock);
 
 	/* Timeout of a tree mount is ultimately determined by its top dentry */
 	if (!autofs4_can_expire(top, timeout, do_now))
@@ -230,36 +256,21 @@ static struct dentry *autofs4_check_leaves(struct vfsmount *mnt,
 	DPRINTK("parent %p %.*s",
 		parent, (int)parent->d_name.len, parent->d_name.name);
 
-	spin_lock(&dcache_lock);
-	for (p = parent; p; p = next_dentry(p, parent)) {
-		spin_lock(&p->d_lock);
-		/* Negative dentry - give up */
-		if (!simple_positive(p)) {
-			spin_unlock(&p->d_lock);
-			continue;
-		}
-
+	p = NULL;
+	while ((p = get_next_positive_dentry(p, parent))) {
 		DPRINTK("dentry %p %.*s",
 			p, (int) p->d_name.len, p->d_name.name);
 
-		p = dget_dlock(p);
-		spin_unlock(&p->d_lock);
-		spin_unlock(&dcache_lock);
-
 		if (d_mountpoint(p)) {
 			/* Can we umount this guy */
 			if (autofs4_mount_busy(mnt, p))
-				goto cont;
+				continue;
 
 			/* Can we expire this guy */
 			if (autofs4_can_expire(p, timeout, do_now))
 				return p;
 		}
-cont:
-		dput(p);
-		spin_lock(&dcache_lock);
 	}
-	spin_unlock(&dcache_lock);
 	return NULL;
 }
 
@@ -310,8 +321,8 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 {
 	unsigned long timeout;
 	struct dentry *root = sb->s_root;
+	struct dentry *dentry;
 	struct dentry *expired = NULL;
-	struct list_head *next;
 	int do_now = how & AUTOFS_EXP_IMMEDIATE;
 	int exp_leaves = how & AUTOFS_EXP_LEAVES;
 	struct autofs_info *ino;
@@ -323,26 +334,8 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 	now = jiffies;
 	timeout = sbi->exp_timeout;
 
-	spin_lock(&dcache_lock);
-	next = root->d_subdirs.next;
-
-	/* On exit from the loop expire is set to a dgot dentry
-	 * to expire or it's NULL */
-	while ( next != &root->d_subdirs ) {
-		struct dentry *dentry = list_entry(next, struct dentry, d_u.d_child);
-
-		/* Negative dentry - give up */
-		spin_lock(&dentry->d_lock);
-		if (!simple_positive(dentry)) {
-			next = next->next;
-			spin_unlock(&dentry->d_lock);
-			continue;
-		}
-
-		dentry = dget_dlock(dentry);
-		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
-
+	dentry = NULL;
+	while ((dentry = get_next_positive_dentry(dentry, root))) {
 		spin_lock(&sbi->fs_lock);
 		ino = autofs4_dentry_ino(dentry);
 
@@ -405,11 +398,7 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 		}
 next:
 		spin_unlock(&sbi->fs_lock);
-		dput(dentry);
-		spin_lock(&dcache_lock);
-		next = next->next;
 	}
-	spin_unlock(&dcache_lock);
 	return NULL;
 
 found:
@@ -420,7 +409,11 @@ found:
 	init_completion(&ino->expire_complete);
 	spin_unlock(&sbi->fs_lock);
 	spin_lock(&dcache_lock);
+	spin_lock(&expired->d_parent->d_lock);
+	spin_lock_nested(&expired->d_lock, DENTRY_D_LOCK_NESTED);
 	list_move(&expired->d_parent->d_subdirs, &expired->d_u.d_child);
+	spin_unlock(&expired->d_lock);
+	spin_unlock(&expired->d_parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return expired;
 }
diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
index 7922509..7a9ed6b 100644
--- a/fs/autofs4/root.c
+++ b/fs/autofs4/root.c
@@ -143,10 +143,13 @@ static int autofs4_dir_open(struct inode *inode, struct file *file)
 	 * it.
 	 */
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	if (!d_mountpoint(dentry) && list_empty(&dentry->d_subdirs)) {
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		return -ENOENT;
 	}
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 
 out:
@@ -253,7 +256,9 @@ static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
 	lookup_type = autofs4_need_mount(nd->flags);
 	spin_lock(&sbi->fs_lock);
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	if (!(lookup_type || ino->flags & AUTOFS_INF_PENDING)) {
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		spin_unlock(&sbi->fs_lock);
 		goto follow;
@@ -266,6 +271,7 @@ static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
 	 */
 	if (ino->flags & AUTOFS_INF_PENDING ||
 	    (!d_mountpoint(dentry) && list_empty(&dentry->d_subdirs))) {
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 		spin_unlock(&sbi->fs_lock);
 
@@ -275,6 +281,7 @@ static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
 
 		goto follow;
 	}
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 	spin_unlock(&sbi->fs_lock);
 follow:
@@ -347,10 +354,12 @@ static int autofs4_revalidate(struct dentry *dentry, struct nameidata *nd)
 
 	/* Check for a non-mountpoint directory with no contents */
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	if (S_ISDIR(dentry->d_inode->i_mode) &&
 	    !d_mountpoint(dentry) && list_empty(&dentry->d_subdirs)) {
 		DPRINTK("dentry=%p %.*s, emptydir",
 			 dentry, dentry->d_name.len, dentry->d_name.name);
+		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_lock);
 
 		/* The daemon never causes a mount to trigger */
@@ -367,6 +376,7 @@ static int autofs4_revalidate(struct dentry *dentry, struct nameidata *nd)
 
 		return status;
 	}
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 
 	return 1;
@@ -776,12 +786,16 @@ static int autofs4_dir_rmdir(struct inode *dir, struct dentry *dentry)
 		return -EACCES;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&sbi->lookup_lock);
+	spin_lock(&dentry->d_lock);
 	if (!list_empty(&dentry->d_subdirs)) {
+		spin_unlock(&dentry->d_lock);
+		spin_unlock(&sbi->lookup_lock);
 		spin_unlock(&dcache_lock);
 		return -ENOTEMPTY;
 	}
-	autofs4_add_expiring(dentry);
-	spin_lock(&dentry->d_lock);
+	__autofs4_add_expiring(dentry);
+	spin_unlock(&sbi->lookup_lock);
 	__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 571f270..2c924e8 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -113,6 +113,7 @@ static int __dcache_readdir(struct file *filp,
 	     last);
 
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 
 	/* start at beginning? */
 	if (filp->f_pos == 2 || last == NULL ||
@@ -136,7 +137,7 @@ more:
 			fi->at_end = 1;
 			goto out_unlock;
 		}
-		spin_lock(&dentry->d_lock);
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 		if (!d_unhashed(dentry) && dentry->d_inode &&
 		    ceph_snap(dentry->d_inode) != CEPH_SNAPDIR &&
 		    ceph_ino(dentry->d_inode) != CEPH_INO_CEPH &&
@@ -154,6 +155,7 @@ more:
 
 	dget_dlock(dentry);
 	spin_unlock(&dentry->d_lock);
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 
 	dout(" %llu (%llu) dentry %p %.*s %p\n", di->offset, filp->f_pos,
@@ -188,10 +190,12 @@ more:
 	}
 
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	p = p->prev;	/* advance to next dentry */
 	goto more;
 
 out_unlock:
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 out:
 	if (last)
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index bb68c79..2c69444 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -842,11 +842,13 @@ static void ceph_set_dentry_offset(struct dentry *dn)
 	spin_unlock(&inode->i_lock);
 
 	spin_lock(&dcache_lock);
-	spin_lock(&dn->d_lock);
+	spin_lock(&dir->d_lock);
+	spin_lock_nested(&dn->d_lock, DENTRY_D_LOCK_NESTED);
 	list_move(&dn->d_u.d_child, &dir->d_subdirs);
 	dout("set_dentry_offset %p %lld (%p %p)\n", dn, di->offset,
 	     dn->d_u.d_child.prev, dn->d_u.d_child.next);
 	spin_unlock(&dn->d_lock);
+	spin_unlock(&dir->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
@@ -1232,9 +1234,11 @@ retry_lookup:
 		} else {
 			/* reorder parent's d_subdirs */
 			spin_lock(&dcache_lock);
-			spin_lock(&dn->d_lock);
+			spin_lock(&parent->d_lock);
+			spin_lock_nested(&dn->d_lock, DENTRY_D_LOCK_NESTED);
 			list_move(&dn->d_u.d_child, &parent->d_subdirs);
 			spin_unlock(&dn->d_lock);
+			spin_unlock(&parent->d_lock);
 			spin_unlock(&dcache_lock);
 		}
 
diff --git a/fs/coda/cache.c b/fs/coda/cache.c
index 9060f08..859393f 100644
--- a/fs/coda/cache.c
+++ b/fs/coda/cache.c
@@ -94,6 +94,7 @@ static void coda_flag_children(struct dentry *parent, int flag)
 	struct dentry *de;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	list_for_each(child, &parent->d_subdirs)
 	{
 		de = list_entry(child, struct dentry, d_u.d_child);
@@ -102,6 +103,7 @@ static void coda_flag_children(struct dentry *parent, int flag)
 			continue;
 		coda_flag_inode(de->d_inode, flag);
 	}
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return; 
 }
diff --git a/fs/dcache.c b/fs/dcache.c
index ee127f9..a661247 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -47,6 +47,8 @@
  *   - d_lru
  *   - d_count
  *   - d_unhashed()
+ *   - d_parent and d_subdirs
+ *   - childrens' d_child and d_parent
  *
  * Ordering:
  * dcache_lock
@@ -223,24 +225,22 @@ static void dentry_lru_move_tail(struct dentry *dentry)
  *
  * If this is the root of the dentry tree, return NULL.
  *
- * dcache_lock and d_lock must be held by caller, are dropped by d_kill.
+ * dcache_lock and d_lock and d_parent->d_lock must be held by caller, and
+ * are dropped by d_kill.
  */
-static struct dentry *d_kill(struct dentry *dentry)
+static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
+	__releases(parent->d_lock)
 	__releases(dcache_lock)
 {
-	struct dentry *parent;
-
 	list_del(&dentry->d_u.d_child);
+	if (parent)
+		spin_unlock(&parent->d_lock);
 	dentry_iput(dentry);
 	/*
 	 * dentry_iput drops the locks, at which point nobody (except
 	 * transient RCU lookups) can reach this dentry.
 	 */
-	if (IS_ROOT(dentry))
-		parent = NULL;
-	else
-		parent = dentry->d_parent;
 	d_free(dentry);
 	return parent;
 }
@@ -312,6 +312,7 @@ EXPORT_SYMBOL(d_drop);
 
 void dput(struct dentry *dentry)
 {
+	struct dentry *parent;
 	if (!dentry)
 		return;
 
@@ -319,6 +320,10 @@ repeat:
 	if (dentry->d_count == 1)
 		might_sleep();
 	spin_lock(&dentry->d_lock);
+	if (IS_ROOT(dentry))
+		parent = NULL;
+	else
+		parent = dentry->d_parent;
 	if (dentry->d_count == 1) {
 		if (!spin_trylock(&dcache_lock)) {
 			/*
@@ -330,10 +335,17 @@ repeat:
 			spin_unlock(&dentry->d_lock);
 			goto repeat;
 		}
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&dcache_lock);
+			goto repeat;
+		}
 	}
 	dentry->d_count--;
 	if (dentry->d_count) {
 		spin_unlock(&dentry->d_lock);
+		if (parent)
+			spin_unlock(&parent->d_lock);
 		spin_unlock(&dcache_lock);
 		return;
 	}
@@ -355,6 +367,8 @@ repeat:
 	dentry_lru_add(dentry);
 
  	spin_unlock(&dentry->d_lock);
+	if (parent)
+		spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return;
 
@@ -363,7 +377,7 @@ unhash_it:
 kill_it:
 	/* if dentry was on the d_lru list delete it from there */
 	dentry_lru_del(dentry);
-	dentry = d_kill(dentry);
+	dentry = d_kill(dentry, parent);
 	if (dentry)
 		goto repeat;
 }
@@ -584,12 +598,13 @@ EXPORT_SYMBOL(d_prune_aliases);
  * quadratic behavior of shrink_dcache_parent(), but is also expected
  * to be beneficial in reducing dentry cache fragmentation.
  */
-static void prune_one_dentry(struct dentry * dentry)
+static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
+	__releases(parent->d_lock)
 	__releases(dcache_lock)
 {
 	__d_drop(dentry);
-	dentry = d_kill(dentry);
+	dentry = d_kill(dentry, parent);
 
 	/*
 	 * Prune ancestors.  Locking is simpler than in dput(),
@@ -597,9 +612,20 @@ static void prune_one_dentry(struct dentry * dentry)
 	 */
 	while (dentry) {
 		spin_lock(&dcache_lock);
+again:
 		spin_lock(&dentry->d_lock);
+		if (IS_ROOT(dentry))
+			parent = NULL;
+		else
+			parent = dentry->d_parent;
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dentry->d_lock);
+			goto again;
+		}
 		dentry->d_count--;
 		if (dentry->d_count) {
+			if (parent)
+				spin_unlock(&parent->d_lock);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 			return;
@@ -607,7 +633,7 @@ static void prune_one_dentry(struct dentry * dentry)
 
 		dentry_lru_del(dentry);
 		__d_drop(dentry);
-		dentry = d_kill(dentry);
+		dentry = d_kill(dentry, parent);
 	}
 }
 
@@ -616,29 +642,40 @@ static void shrink_dentry_list(struct list_head *list)
 	struct dentry *dentry;
 
 	while (!list_empty(list)) {
+		struct dentry *parent;
+
 		dentry = list_entry(list->prev, struct dentry, d_lru);
 
 		if (!spin_trylock(&dentry->d_lock)) {
+relock:
 			spin_unlock(&dcache_lru_lock);
 			cpu_relax();
 			spin_lock(&dcache_lru_lock);
 			continue;
 		}
 
-		__dentry_lru_del(dentry);
-
 		/*
 		 * We found an inuse dentry which was not removed from
 		 * the LRU because of laziness during lookup.  Do not free
 		 * it - just keep it off the LRU list.
 		 */
 		if (dentry->d_count) {
+			__dentry_lru_del(dentry);
 			spin_unlock(&dentry->d_lock);
 			continue;
 		}
+		if (IS_ROOT(dentry))
+			parent = NULL;
+		else
+			parent = dentry->d_parent;
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dentry->d_lock);
+			goto relock;
+		}
+		__dentry_lru_del(dentry);
 		spin_unlock(&dcache_lru_lock);
 
-		prune_one_dentry(dentry);
+		prune_one_dentry(dentry, parent);
 		/* dcache_lock and dentry->d_lock dropped */
 		spin_lock(&dcache_lock);
 		spin_lock(&dcache_lru_lock);
@@ -833,14 +870,16 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 			/* this is a branch with children - detach all of them
 			 * from the system in one go */
 			spin_lock(&dcache_lock);
+			spin_lock(&dentry->d_lock);
 			list_for_each_entry(loop, &dentry->d_subdirs,
 					    d_u.d_child) {
-				spin_lock(&loop->d_lock);
+				spin_lock_nested(&loop->d_lock,
+						DENTRY_D_LOCK_NESTED);
 				dentry_lru_del(loop);
 				__d_drop(loop);
 				spin_unlock(&loop->d_lock);
-				cond_resched_lock(&dcache_lock);
 			}
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 
 			/* move to the first child */
@@ -868,16 +907,17 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 				BUG();
 			}
 
-			if (IS_ROOT(dentry))
+			if (IS_ROOT(dentry)) {
 				parent = NULL;
-			else {
+				list_del(&dentry->d_u.d_child);
+			} else {
 				parent = dentry->d_parent;
 				spin_lock(&parent->d_lock);
 				parent->d_count--;
+				list_del(&dentry->d_u.d_child);
 				spin_unlock(&parent->d_lock);
 			}
 
-			list_del(&dentry->d_u.d_child);
 			detached++;
 
 			inode = dentry->d_inode;
@@ -958,6 +998,7 @@ int have_submounts(struct dentry *parent)
 	spin_lock(&dcache_lock);
 	if (d_mountpoint(parent))
 		goto positive;
+	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
 resume:
@@ -965,22 +1006,34 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
+
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 		/* Have we found a mount point ? */
-		if (d_mountpoint(dentry))
+		if (d_mountpoint(dentry)) {
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&this_parent->d_lock);
 			goto positive;
+		}
 		if (!list_empty(&dentry->d_subdirs)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
 			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
+		spin_unlock(&dentry->d_lock);
 	}
 	/*
 	 * All done at this level ... ascend and resume the search.
 	 */
 	if (this_parent != parent) {
 		next = this_parent->d_u.d_child.next;
+		spin_unlock(&this_parent->d_lock);
 		this_parent = this_parent->d_parent;
+		spin_lock(&this_parent->d_lock);
 		goto resume;
 	}
+	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return 0; /* No mount points found in tree */
 positive:
@@ -1010,6 +1063,7 @@ static int select_parent(struct dentry * parent)
 	int found = 0;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
 resume:
@@ -1017,8 +1071,9 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
+		BUG_ON(this_parent == dentry);
 
-		spin_lock(&dentry->d_lock);
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 
 		/* 
 		 * move only zero ref count dentries to the end 
@@ -1031,33 +1086,44 @@ resume:
 			dentry_lru_del(dentry);
 		}
 
-		spin_unlock(&dentry->d_lock);
-
 		/*
 		 * We can return to the caller if we have found some (this
 		 * ensures forward progress). We'll be coming back to find
 		 * the rest.
 		 */
-		if (found && need_resched())
+		if (found && need_resched()) {
+			spin_unlock(&dentry->d_lock);
 			goto out;
+		}
 
 		/*
 		 * Descend a level if the d_subdirs list is non-empty.
 		 */
 		if (!list_empty(&dentry->d_subdirs)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
 			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
+
+		spin_unlock(&dentry->d_lock);
 	}
 	/*
 	 * All done at this level ... ascend and resume the search.
 	 */
 	if (this_parent != parent) {
+		struct dentry *tmp;
 		next = this_parent->d_u.d_child.next;
-		this_parent = this_parent->d_parent;
+		tmp = this_parent->d_parent;
+		spin_unlock(&this_parent->d_lock);
+		BUG_ON(tmp == this_parent);
+		this_parent = tmp;
+		spin_lock(&this_parent->d_lock);
 		goto resume;
 	}
 out:
+	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return found;
 }
@@ -1155,18 +1221,19 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	INIT_LIST_HEAD(&dentry->d_lru);
 	INIT_LIST_HEAD(&dentry->d_subdirs);
 	INIT_LIST_HEAD(&dentry->d_alias);
+	INIT_LIST_HEAD(&dentry->d_u.d_child);
 
 	if (parent) {
-		dentry->d_parent = dget(parent);
+		spin_lock(&dcache_lock);
+		spin_lock(&parent->d_lock);
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+		dentry->d_parent = dget_dlock(parent);
 		dentry->d_sb = parent->d_sb;
-	} else {
-		INIT_LIST_HEAD(&dentry->d_u.d_child);
-	}
-
-	spin_lock(&dcache_lock);
-	if (parent)
 		list_add(&dentry->d_u.d_child, &parent->d_subdirs);
-	spin_unlock(&dcache_lock);
+		spin_unlock(&dentry->d_lock);
+		spin_unlock(&parent->d_lock);
+		spin_unlock(&dcache_lock);
+	}
 
 	this_cpu_inc(nr_dentry);
 
@@ -1684,13 +1751,18 @@ int d_validate(struct dentry *dentry, struct dentry *dparent)
 	struct dentry *child;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dparent->d_lock);
 	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
 		if (dentry == child) {
-			__dget_locked(dentry);
+			spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+			__dget_locked_dlock(dentry);
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&dparent->d_lock);
 			spin_unlock(&dcache_lock);
 			return 1;
 		}
 	}
+	spin_unlock(&dparent->d_lock);
 	spin_unlock(&dcache_lock);
 
 	return 0;
@@ -1802,17 +1874,6 @@ void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
 }
 EXPORT_SYMBOL(dentry_update_name_case);
 
-/*
- * When switching names, the actual string doesn't strictly have to
- * be preserved in the target - because we're dropping the target
- * anyway. As such, we can just do a simple memcpy() to copy over
- * the new name before we switch.
- *
- * Note that we have to be a lot more careful about getting the hash
- * switched - we have to switch the hash value properly even if it
- * then no longer matches the actual (corrupted) string of the target.
- * The hash value has to match the hash queue that the dentry is on..
- */
 static void switch_names(struct dentry *dentry, struct dentry *target)
 {
 	if (dname_external(target)) {
@@ -1854,18 +1915,53 @@ static void switch_names(struct dentry *dentry, struct dentry *target)
 	swap(dentry->d_name.len, target->d_name.len);
 }
 
+static void dentry_lock_for_move(struct dentry *dentry, struct dentry *target)
+{
+	/*
+	 * XXXX: do we really need to take target->d_lock?
+	 */
+	if (IS_ROOT(dentry) || dentry->d_parent == target->d_parent)
+		spin_lock(&target->d_parent->d_lock);
+	else {
+		if (d_ancestor(dentry->d_parent, target->d_parent)) {
+			spin_lock(&dentry->d_parent->d_lock);
+			spin_lock_nested(&target->d_parent->d_lock,
+						DENTRY_D_LOCK_NESTED);
+		} else {
+			spin_lock(&target->d_parent->d_lock);
+			spin_lock_nested(&dentry->d_parent->d_lock,
+						DENTRY_D_LOCK_NESTED);
+		}
+	}
+	if (target < dentry) {
+		spin_lock_nested(&target->d_lock, 2);
+		spin_lock_nested(&dentry->d_lock, 3);
+	} else {
+		spin_lock_nested(&dentry->d_lock, 2);
+		spin_lock_nested(&target->d_lock, 3);
+	}
+}
+
+static void dentry_unlock_parents_for_move(struct dentry *dentry,
+					struct dentry *target)
+{
+	if (target->d_parent != dentry->d_parent)
+		spin_unlock(&dentry->d_parent->d_lock);
+	if (target->d_parent != target)
+		spin_unlock(&target->d_parent->d_lock);
+}
+
 /*
- * We cannibalize "target" when moving dentry on top of it,
- * because it's going to be thrown away anyway. We could be more
- * polite about it, though.
- *
- * This forceful removal will result in ugly /proc output if
- * somebody holds a file open that got deleted due to a rename.
- * We could be nicer about the deleted file, and let it show
- * up under the name it had before it was deleted rather than
- * under the original name of the file that was moved on top of it.
+ * When switching names, the actual string doesn't strictly have to
+ * be preserved in the target - because we're dropping the target
+ * anyway. As such, we can just do a simple memcpy() to copy over
+ * the new name before we switch.
+ *
+ * Note that we have to be a lot more careful about getting the hash
+ * switched - we have to switch the hash value properly even if it
+ * then no longer matches the actual (corrupted) string of the target.
+ * The hash value has to match the hash queue that the dentry is on..
  */
- 
 /*
  * d_move_locked - move a dentry
  * @dentry: entry to move
@@ -1879,20 +1975,12 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	if (!dentry->d_inode)
 		printk(KERN_WARNING "VFS: moving negative dcache entry\n");
 
+	BUG_ON(d_ancestor(dentry, target));
+	BUG_ON(d_ancestor(target, dentry));
+
 	write_seqlock(&rename_lock);
-	/*
-	 * XXXX: do we really need to take target->d_lock?
-	 */
-	if (d_ancestor(dentry, target)) {
-		spin_lock(&dentry->d_lock);
-		spin_lock_nested(&target->d_lock, DENTRY_D_LOCK_NESTED);
-	} else if (d_ancestor(target, dentry) || target < dentry) {
-		spin_lock(&target->d_lock);
-		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
-	} else {
-		spin_lock(&dentry->d_lock);
-		spin_lock_nested(&target->d_lock, DENTRY_D_LOCK_NESTED);
-	}
+
+	dentry_lock_for_move(dentry, target);
 
 	/* Move the dentry to the target hash queue, if on different bucket */
 	spin_lock(&dcache_hash_lock);
@@ -1924,6 +2012,8 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	}
 
 	list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
+
+	dentry_unlock_parents_for_move(dentry, target);
 	spin_unlock(&target->d_lock);
 	fsnotify_d_move(dentry);
 	spin_unlock(&dentry->d_lock);
@@ -2013,17 +2103,20 @@ out_err:
 /*
  * Prepare an anonymous dentry for life in the superblock's dentry tree as a
  * named dentry in place of the dentry to be replaced.
+ * returns with anon->d_lock held!
  */
 static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
 {
 	struct dentry *dparent, *aparent;
 
-	switch_names(dentry, anon);
-	swap(dentry->d_name.hash, anon->d_name.hash);
+	dentry_lock_for_move(anon, dentry);
 
 	dparent = dentry->d_parent;
 	aparent = anon->d_parent;
 
+	switch_names(dentry, anon);
+	swap(dentry->d_name.hash, anon->d_name.hash);
+
 	dentry->d_parent = (aparent == anon) ? dentry : aparent;
 	list_del(&dentry->d_u.d_child);
 	if (!IS_ROOT(dentry))
@@ -2038,6 +2131,10 @@ static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
 	else
 		INIT_LIST_HEAD(&anon->d_u.d_child);
 
+	dentry_unlock_parents_for_move(anon, dentry);
+	spin_unlock(&dentry->d_lock);
+
+	/* anon->d_lock still locked, returns locked */
 	anon->d_flags &= ~DCACHE_DISCONNECTED;
 }
 
@@ -2073,7 +2170,6 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 			/* Is this an anonymous mountpoint that we could splice
 			 * into our tree? */
 			if (IS_ROOT(alias)) {
-				spin_lock(&alias->d_lock);
 				__d_materialise_dentry(dentry, alias);
 				__d_drop(alias);
 				goto found;
@@ -2558,6 +2654,7 @@ void d_genocide(struct dentry *root)
 	struct list_head *next;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
 resume:
@@ -2571,8 +2668,10 @@ resume:
 			continue;
 		}
 		if (!list_empty(&dentry->d_subdirs)) {
-			spin_unlock(&dentry->d_lock);
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
 			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
 		dentry->d_count--;
@@ -2580,12 +2679,13 @@ resume:
 	}
 	if (this_parent != root) {
 		next = this_parent->d_u.d_child.next;
-		spin_lock(&this_parent->d_lock);
 		this_parent->d_count--;
 		spin_unlock(&this_parent->d_lock);
 		this_parent = this_parent->d_parent;
+		spin_lock(&this_parent->d_lock);
 		goto resume;
 	}
+	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
diff --git a/fs/libfs.c b/fs/libfs.c
index 433e713..cc47949 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -81,7 +81,8 @@ int dcache_dir_close(struct inode *inode, struct file *file)
 
 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
 {
-	mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
+	struct dentry *dentry = file->f_path.dentry;
+	mutex_lock(&dentry->d_inode->i_mutex);
 	switch (origin) {
 		case 1:
 			offset += file->f_pos;
@@ -89,7 +90,7 @@ loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
 			if (offset >= 0)
 				break;
 		default:
-			mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
+			mutex_unlock(&dentry->d_inode->i_mutex);
 			return -EINVAL;
 	}
 	if (offset != file->f_pos) {
@@ -100,22 +101,25 @@ loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
 			loff_t n = file->f_pos - 2;
 
 			spin_lock(&dcache_lock);
+			spin_lock(&dentry->d_lock);
+			/* d_lock not required for cursor */
 			list_del(&cursor->d_u.d_child);
-			p = file->f_path.dentry->d_subdirs.next;
-			while (n && p != &file->f_path.dentry->d_subdirs) {
+			p = dentry->d_subdirs.next;
+			while (n && p != &dentry->d_subdirs) {
 				struct dentry *next;
 				next = list_entry(p, struct dentry, d_u.d_child);
-				spin_lock(&next->d_lock);
+				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
 				if (simple_positive(next))
 					n--;
 				spin_unlock(&next->d_lock);
 				p = p->next;
 			}
 			list_add_tail(&cursor->d_u.d_child, p);
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 		}
 	}
-	mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
+	mutex_unlock(&dentry->d_inode->i_mutex);
 	return offset;
 }
 
@@ -156,6 +160,7 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 			/* fallthrough */
 		default:
 			spin_lock(&dcache_lock);
+			spin_lock(&dentry->d_lock);
 			if (filp->f_pos == 2)
 				list_move(q, &dentry->d_subdirs);
 
@@ -169,6 +174,7 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 				}
 
 				spin_unlock(&next->d_lock);
+				spin_unlock(&dentry->d_lock);
 				spin_unlock(&dcache_lock);
 				if (filldir(dirent, next->d_name.name, 
 					    next->d_name.len, filp->f_pos, 
@@ -176,11 +182,15 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 					    dt_type(next->d_inode)) < 0)
 					return 0;
 				spin_lock(&dcache_lock);
+				spin_lock(&dentry->d_lock);
+				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
 				/* next is still alive */
 				list_move(q, p);
+				spin_unlock(&next->d_lock);
 				p = q;
 				filp->f_pos++;
 			}
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 	}
 	return 0;
@@ -276,6 +286,7 @@ int simple_empty(struct dentry *dentry)
 	int ret = 0;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
 		if (simple_positive(child)) {
@@ -286,6 +297,7 @@ int simple_empty(struct dentry *dentry)
 	}
 	ret = 1;
 out:
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 	return ret;
 }
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index bbbf7922..102278e 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -392,6 +392,7 @@ ncp_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 
 	/* If a pointer is invalid, we search the dentry. */
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
 		dent = list_entry(next, struct dentry, d_u.d_child);
@@ -400,11 +401,13 @@ ncp_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 				dget_locked(dent);
 			else
 				dent = NULL;
+			spin_unlock(&parent->d_lock);
 			spin_unlock(&dcache_lock);
 			goto out;
 		}
 		next = next->next;
 	}
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return NULL;
 
diff --git a/fs/ncpfs/ncplib_kernel.h b/fs/ncpfs/ncplib_kernel.h
index 244d1b7..c4b718f 100644
--- a/fs/ncpfs/ncplib_kernel.h
+++ b/fs/ncpfs/ncplib_kernel.h
@@ -194,6 +194,7 @@ ncp_renew_dentries(struct dentry *parent)
 	struct dentry *dentry;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
 		dentry = list_entry(next, struct dentry, d_u.d_child);
@@ -205,6 +206,7 @@ ncp_renew_dentries(struct dentry *parent)
 
 		next = next->next;
 	}
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
@@ -216,6 +218,7 @@ ncp_invalidate_dircache_entries(struct dentry *parent)
 	struct dentry *dentry;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
 		dentry = list_entry(next, struct dentry, d_u.d_child);
@@ -223,6 +226,7 @@ ncp_invalidate_dircache_entries(struct dentry *parent)
 		ncp_age_dentry(server, dentry);
 		next = next->next;
 	}
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index 20dc218..aa4f25e 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -68,17 +68,19 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 		/* run all of the children of the original inode and fix their
 		 * d_flags to indicate parental interest (their parent is the
 		 * original inode) */
+		spin_lock(&alias->d_lock);
 		list_for_each_entry(child, &alias->d_subdirs, d_u.d_child) {
 			if (!child->d_inode)
 				continue;
 
-			spin_lock(&child->d_lock);
+			spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
 			if (watched)
 				child->d_flags |= DCACHE_FSNOTIFY_PARENT_WATCHED;
 			else
 				child->d_flags &= ~DCACHE_FSNOTIFY_PARENT_WATCHED;
 			spin_unlock(&child->d_lock);
 		}
+		spin_unlock(&alias->d_lock);
 	}
 	spin_unlock(&dcache_lock);
 }
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index b0ade2d..ddf4f55 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -305,6 +305,7 @@ static inline struct dentry *dget_dlock(struct dentry *dentry)
 	}
 	return dentry;
 }
+
 static inline struct dentry *dget(struct dentry *dentry)
 {
 	if (dentry) {
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index eb7af39..7b4705b 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -877,23 +877,31 @@ static void cgroup_clear_directory(struct dentry *dentry)
 
 	BUG_ON(!mutex_is_locked(&dentry->d_inode->i_mutex));
 	spin_lock(&dcache_lock);
+	spin_lock(&dentry->d_lock);
 	node = dentry->d_subdirs.next;
 	while (node != &dentry->d_subdirs) {
 		struct dentry *d = list_entry(node, struct dentry, d_u.d_child);
+
+		spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
 		list_del_init(node);
 		if (d->d_inode) {
 			/* This should never be called on a cgroup
 			 * directory with child cgroups */
 			BUG_ON(d->d_inode->i_mode & S_IFDIR);
-			d = dget_locked(d);
+			dget_locked_dlock(d);
+			spin_unlock(&d->d_lock);
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 			d_delete(d);
 			simple_unlink(dentry->d_inode, d);
 			dput(d);
 			spin_lock(&dcache_lock);
-		}
+			spin_lock(&dentry->d_lock);
+		} else
+			spin_unlock(&d->d_lock);
 		node = dentry->d_subdirs.next;
 	}
+	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
@@ -902,10 +910,17 @@ static void cgroup_clear_directory(struct dentry *dentry)
  */
 static void cgroup_d_remove_dir(struct dentry *dentry)
 {
+	struct dentry *parent;
+
 	cgroup_clear_directory(dentry);
 
 	spin_lock(&dcache_lock);
+	parent = dentry->d_parent;
+	spin_lock(&parent->d_lock);
+	spin_lock(&dentry->d_lock);
 	list_del_init(&dentry->d_u.d_child);
+	spin_unlock(&dentry->d_lock);
+	spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 	remove_dir(dentry);
 }
diff --git a/security/selinux/selinuxfs.c b/security/selinux/selinuxfs.c
index 073fd5b..017ec09 100644
--- a/security/selinux/selinuxfs.c
+++ b/security/selinux/selinuxfs.c
@@ -1146,22 +1146,30 @@ static void sel_remove_entries(struct dentry *de)
 	struct list_head *node;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&de->d_lock);
 	node = de->d_subdirs.next;
 	while (node != &de->d_subdirs) {
 		struct dentry *d = list_entry(node, struct dentry, d_u.d_child);
+
+		spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
 		list_del_init(node);
 
 		if (d->d_inode) {
-			d = dget_locked(d);
+			dget_locked_dlock(d);
+			spin_unlock(&de->d_lock);
+			spin_unlock(&d->d_lock);
 			spin_unlock(&dcache_lock);
 			d_delete(d);
 			simple_unlink(de->d_inode, d);
 			dput(d);
 			spin_lock(&dcache_lock);
-		}
+			spin_lock(&de->d_lock);
+		} else
+			spin_unlock(&d->d_lock);
 		node = de->d_subdirs.next;
 	}
 
+	spin_unlock(&de->d_lock);
 	spin_unlock(&dcache_lock);
 }
 
-- 
cgit v0.10.2


From b23fb0a60379a95e10c671f646b259ea2558421e Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:35 +1100
Subject: fs: scale inode alias list

Add a new lock, dcache_inode_lock, to protect the inode's i_dentry list
from concurrent modification. d_alias is also protected by d_lock.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 34bf71b..47dfd5d 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -271,9 +271,11 @@ static struct dentry *v9fs_dentry_from_dir_inode(struct inode *inode)
 	struct dentry *dentry;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	/* Directory should have only one entry. */
 	BUG_ON(S_ISDIR(inode->i_mode) && !list_is_singular(&inode->i_dentry));
 	dentry = list_entry(inode->i_dentry.next, struct dentry, d_alias);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	return dentry;
 }
diff --git a/fs/affs/amigaffs.c b/fs/affs/amigaffs.c
index 7d0f0a3..2321cc9 100644
--- a/fs/affs/amigaffs.c
+++ b/fs/affs/amigaffs.c
@@ -129,6 +129,7 @@ affs_fix_dcache(struct dentry *dentry, u32 entry_ino)
 	struct list_head *head, *next;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	head = &inode->i_dentry;
 	next = head->next;
 	while (next != head) {
@@ -139,6 +140,7 @@ affs_fix_dcache(struct dentry *dentry, u32 entry_ino)
 		}
 		next = next->next;
 	}
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 }
 
diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index 589f3e3..0036983 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -810,12 +810,15 @@ inode_has_hashed_dentries(struct inode *inode)
 	struct dentry *dentry;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		if (!d_unhashed(dentry) || IS_ROOT(dentry)) {
+			spin_unlock(&dcache_inode_lock);
 			spin_unlock(&dcache_lock);
 			return true;
 		}
 	}
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	return false;
 }
diff --git a/fs/dcache.c b/fs/dcache.c
index a661247..de38680 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -37,6 +37,8 @@
 
 /*
  * Usage:
+ * dcache_inode_lock protects:
+ *   - i_dentry, d_alias, d_inode
  * dcache_hash_lock protects:
  *   - the dcache hash table, s_anon lists
  * dcache_lru_lock protects:
@@ -49,12 +51,14 @@
  *   - d_unhashed()
  *   - d_parent and d_subdirs
  *   - childrens' d_child and d_parent
+ *   - d_alias, d_inode
  *
  * Ordering:
  * dcache_lock
- *   dentry->d_lock
- *     dcache_lru_lock
- *     dcache_hash_lock
+ *   dcache_inode_lock
+ *     dentry->d_lock
+ *       dcache_lru_lock
+ *       dcache_hash_lock
  *
  * If there is an ancestor relationship:
  * dentry->d_parent->...->d_parent->d_lock
@@ -70,11 +74,13 @@
 int sysctl_vfs_cache_pressure __read_mostly = 100;
 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
 
+__cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_inode_lock);
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_hash_lock);
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
+EXPORT_SYMBOL(dcache_inode_lock);
 EXPORT_SYMBOL(dcache_lock);
 
 static struct kmem_cache *dentry_cache __read_mostly;
@@ -154,6 +160,7 @@ static void d_free(struct dentry *dentry)
  */
 static void dentry_iput(struct dentry * dentry)
 	__releases(dentry->d_lock)
+	__releases(dcache_inode_lock)
 	__releases(dcache_lock)
 {
 	struct inode *inode = dentry->d_inode;
@@ -161,6 +168,7 @@ static void dentry_iput(struct dentry * dentry)
 		dentry->d_inode = NULL;
 		list_del_init(&dentry->d_alias);
 		spin_unlock(&dentry->d_lock);
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 		if (!inode->i_nlink)
 			fsnotify_inoderemove(inode);
@@ -170,6 +178,7 @@ static void dentry_iput(struct dentry * dentry)
 			iput(inode);
 	} else {
 		spin_unlock(&dentry->d_lock);
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 	}
 }
@@ -231,6 +240,7 @@ static void dentry_lru_move_tail(struct dentry *dentry)
 static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
 	__releases(parent->d_lock)
+	__releases(dcache_inode_lock)
 	__releases(dcache_lock)
 {
 	list_del(&dentry->d_u.d_child);
@@ -332,13 +342,18 @@ repeat:
 			 * want to reduce dcache_lock anyway so this will
 			 * get improved.
 			 */
+drop1:
 			spin_unlock(&dentry->d_lock);
 			goto repeat;
 		}
-		if (parent && !spin_trylock(&parent->d_lock)) {
-			spin_unlock(&dentry->d_lock);
+		if (!spin_trylock(&dcache_inode_lock)) {
+drop2:
 			spin_unlock(&dcache_lock);
-			goto repeat;
+			goto drop1;
+		}
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dcache_inode_lock);
+			goto drop2;
 		}
 	}
 	dentry->d_count--;
@@ -369,6 +384,7 @@ repeat:
  	spin_unlock(&dentry->d_lock);
 	if (parent)
 		spin_unlock(&parent->d_lock);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	return;
 
@@ -558,7 +574,9 @@ struct dentry *d_find_alias(struct inode *inode)
 
 	if (!list_empty(&inode->i_dentry)) {
 		spin_lock(&dcache_lock);
+		spin_lock(&dcache_inode_lock);
 		de = __d_find_alias(inode, 0);
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 	}
 	return de;
@@ -574,18 +592,21 @@ void d_prune_aliases(struct inode *inode)
 	struct dentry *dentry;
 restart:
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		spin_lock(&dentry->d_lock);
 		if (!dentry->d_count) {
 			__dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
+			spin_unlock(&dcache_inode_lock);
 			spin_unlock(&dcache_lock);
 			dput(dentry);
 			goto restart;
 		}
 		spin_unlock(&dentry->d_lock);
 	}
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(d_prune_aliases);
@@ -601,6 +622,7 @@ EXPORT_SYMBOL(d_prune_aliases);
 static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
 	__releases(parent->d_lock)
+	__releases(dcache_inode_lock)
 	__releases(dcache_lock)
 {
 	__d_drop(dentry);
@@ -612,6 +634,7 @@ static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
 	 */
 	while (dentry) {
 		spin_lock(&dcache_lock);
+		spin_lock(&dcache_inode_lock);
 again:
 		spin_lock(&dentry->d_lock);
 		if (IS_ROOT(dentry))
@@ -627,6 +650,7 @@ again:
 			if (parent)
 				spin_unlock(&parent->d_lock);
 			spin_unlock(&dentry->d_lock);
+			spin_unlock(&dcache_inode_lock);
 			spin_unlock(&dcache_lock);
 			return;
 		}
@@ -676,8 +700,9 @@ relock:
 		spin_unlock(&dcache_lru_lock);
 
 		prune_one_dentry(dentry, parent);
-		/* dcache_lock and dentry->d_lock dropped */
+		/* dcache_lock, dcache_inode_lock and dentry->d_lock dropped */
 		spin_lock(&dcache_lock);
+		spin_lock(&dcache_inode_lock);
 		spin_lock(&dcache_lru_lock);
 	}
 }
@@ -699,6 +724,7 @@ static void __shrink_dcache_sb(struct super_block *sb, int *count, int flags)
 	int cnt = *count;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 relock:
 	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
@@ -737,8 +763,8 @@ relock:
 	if (!list_empty(&referenced))
 		list_splice(&referenced, &sb->s_dentry_lru);
 	spin_unlock(&dcache_lru_lock);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
-
 }
 
 /**
@@ -832,12 +858,14 @@ void shrink_dcache_sb(struct super_block *sb)
 	LIST_HEAD(tmp);
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
 		list_splice_init(&sb->s_dentry_lru, &tmp);
 		shrink_dentry_list(&tmp);
 	}
 	spin_unlock(&dcache_lru_lock);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(shrink_dcache_sb);
@@ -1255,9 +1283,11 @@ EXPORT_SYMBOL(d_alloc_name);
 /* the caller must hold dcache_lock */
 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 {
+	spin_lock(&dentry->d_lock);
 	if (inode)
 		list_add(&dentry->d_alias, &inode->i_dentry);
 	dentry->d_inode = inode;
+	spin_unlock(&dentry->d_lock);
 	fsnotify_d_instantiate(dentry, inode);
 }
 
@@ -1280,7 +1310,9 @@ void d_instantiate(struct dentry *entry, struct inode * inode)
 {
 	BUG_ON(!list_empty(&entry->d_alias));
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	__d_instantiate(entry, inode);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	security_d_instantiate(entry, inode);
 }
@@ -1341,7 +1373,9 @@ struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
 	BUG_ON(!list_empty(&entry->d_alias));
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	result = __d_instantiate_unique(entry, inode);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 
 	if (!result) {
@@ -1432,8 +1466,10 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	tmp->d_parent = tmp; /* make sure dput doesn't croak */
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	res = __d_find_alias(inode, 0);
 	if (res) {
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 		dput(tmp);
 		goto out_iput;
@@ -1450,6 +1486,7 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	hlist_add_head(&tmp->d_hash, &inode->i_sb->s_anon);
 	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&tmp->d_lock);
+	spin_unlock(&dcache_inode_lock);
 
 	spin_unlock(&dcache_lock);
 	return tmp;
@@ -1482,9 +1519,11 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 
 	if (inode && S_ISDIR(inode->i_mode)) {
 		spin_lock(&dcache_lock);
+		spin_lock(&dcache_inode_lock);
 		new = __d_find_alias(inode, 1);
 		if (new) {
 			BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
+			spin_unlock(&dcache_inode_lock);
 			spin_unlock(&dcache_lock);
 			security_d_instantiate(new, inode);
 			d_move(new, dentry);
@@ -1492,6 +1531,7 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 		} else {
 			/* already taking dcache_lock, so d_add() by hand */
 			__d_instantiate(dentry, inode);
+			spin_unlock(&dcache_inode_lock);
 			spin_unlock(&dcache_lock);
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
@@ -1566,8 +1606,10 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	 * already has a dentry.
 	 */
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	if (!S_ISDIR(inode->i_mode) || list_empty(&inode->i_dentry)) {
 		__d_instantiate(found, inode);
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 		security_d_instantiate(found, inode);
 		return found;
@@ -1579,6 +1621,7 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	 */
 	new = list_entry(inode->i_dentry.next, struct dentry, d_alias);
 	dget_locked(new);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	security_d_instantiate(found, inode);
 	d_move(new, found);
@@ -1797,6 +1840,7 @@ void d_delete(struct dentry * dentry)
 	 * Are we the only user?
 	 */
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	spin_lock(&dentry->d_lock);
 	isdir = S_ISDIR(dentry->d_inode->i_mode);
 	if (dentry->d_count == 1) {
@@ -1810,6 +1854,7 @@ void d_delete(struct dentry * dentry)
 		__d_drop(dentry);
 
 	spin_unlock(&dentry->d_lock);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 
 	fsnotify_nameremove(dentry, isdir);
@@ -2067,6 +2112,7 @@ struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
  */
 static struct dentry *__d_unalias(struct dentry *dentry, struct dentry *alias)
 	__releases(dcache_lock)
+	__releases(dcache_inode_lock)
 {
 	struct mutex *m1 = NULL, *m2 = NULL;
 	struct dentry *ret;
@@ -2092,6 +2138,7 @@ out_unalias:
 	d_move_locked(alias, dentry);
 	ret = alias;
 out_err:
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	if (m2)
 		mutex_unlock(m2);
@@ -2153,6 +2200,7 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 	BUG_ON(!d_unhashed(dentry));
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 
 	if (!inode) {
 		actual = dentry;
@@ -2196,6 +2244,7 @@ found:
 	_d_rehash(actual);
 	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&actual->d_lock);
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 out_nolock:
 	if (actual == dentry) {
@@ -2207,6 +2256,7 @@ out_nolock:
 	return actual;
 
 shouldnt_be_hashed:
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 	BUG();
 }
diff --git a/fs/exportfs/expfs.c b/fs/exportfs/expfs.c
index 51b3040..84b8c46 100644
--- a/fs/exportfs/expfs.c
+++ b/fs/exportfs/expfs.c
@@ -48,8 +48,10 @@ find_acceptable_alias(struct dentry *result,
 		return result;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &result->d_inode->i_dentry, d_alias) {
 		dget_locked(dentry);
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 		if (toput)
 			dput(toput);
@@ -58,8 +60,10 @@ find_acceptable_alias(struct dentry *result,
 			return dentry;
 		}
 		spin_lock(&dcache_lock);
+		spin_lock(&dcache_inode_lock);
 		toput = dentry;
 	}
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 
 	if (toput)
diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c
index ac7b814..850f67d 100644
--- a/fs/nfs/getroot.c
+++ b/fs/nfs/getroot.c
@@ -64,7 +64,11 @@ static int nfs_superblock_set_dummy_root(struct super_block *sb, struct inode *i
 		 * Oops, since the test for IS_ROOT() will fail.
 		 */
 		spin_lock(&dcache_lock);
+		spin_lock(&dcache_inode_lock);
+		spin_lock(&sb->s_root->d_lock);
 		list_del_init(&sb->s_root->d_alias);
+		spin_unlock(&sb->s_root->d_lock);
+		spin_unlock(&dcache_inode_lock);
 		spin_unlock(&dcache_lock);
 	}
 	return 0;
diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index aa4f25e..ae769fc 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -60,6 +60,7 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 	watched = fsnotify_inode_watches_children(inode);
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dcache_inode_lock);
 	/* run all of the dentries associated with this inode.  Since this is a
 	 * directory, there damn well better only be one item on this list */
 	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
@@ -82,6 +83,7 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 		}
 		spin_unlock(&alias->d_lock);
 	}
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 }
 
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index 35e5f5a..c31b5c6 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -170,7 +170,7 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 	struct dentry *dentry = NULL;
 
 	spin_lock(&dcache_lock);
-
+	spin_lock(&dcache_inode_lock);
 	list_for_each(p, &inode->i_dentry) {
 		dentry = list_entry(p, struct dentry, d_alias);
 
@@ -188,6 +188,7 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 		dentry = NULL;
 	}
 
+	spin_unlock(&dcache_inode_lock);
 	spin_unlock(&dcache_lock);
 
 	return dentry;
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index ddf4f55..bda5ec0 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -181,6 +181,7 @@ struct dentry_operations {
 
 #define DCACHE_CANT_MOUNT	0x0100
 
+extern spinlock_t dcache_inode_lock;
 extern spinlock_t dcache_lock;
 extern seqlock_t rename_lock;
 
-- 
cgit v0.10.2


From 9abca36087288fe28de4749c71ca003d4b9e3ed0 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:36 +1100
Subject: fs: increase d_name lock coverage

Cover d_name with d_lock in more cases, where there may be concurrent
modification to it.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index de38680..a09f077 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1350,6 +1350,11 @@ static struct dentry *__d_instantiate_unique(struct dentry *entry,
 	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
 		struct qstr *qstr = &alias->d_name;
 
+		/*
+		 * Don't need alias->d_lock here, because aliases with
+		 * d_parent == entry->d_parent are not subject to name or
+		 * parent changes, because the parent inode i_mutex is held.
+		 */
 		if (qstr->hash != hash)
 			continue;
 		if (alias->d_parent != entry->d_parent)
@@ -2313,7 +2318,9 @@ static int prepend_path(const struct path *path, struct path *root,
 		}
 		parent = dentry->d_parent;
 		prefetch(parent);
+		spin_lock(&dentry->d_lock);
 		error = prepend_name(buffer, buflen, &dentry->d_name);
+		spin_unlock(&dentry->d_lock);
 		if (!error)
 			error = prepend(buffer, buflen, "/", 1);
 		if (error)
@@ -2515,10 +2522,13 @@ static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
 
 	while (!IS_ROOT(dentry)) {
 		struct dentry *parent = dentry->d_parent;
+		int error;
 
 		prefetch(parent);
-		if ((prepend_name(&end, &buflen, &dentry->d_name) != 0) ||
-		    (prepend(&end, &buflen, "/", 1) != 0))
+		spin_lock(&dentry->d_lock);
+		error = prepend_name(&end, &buflen, &dentry->d_name);
+		spin_unlock(&dentry->d_lock);
+		if (error != 0 || prepend(&end, &buflen, "/", 1) != 0)
 			goto Elong;
 
 		retval = end;
-- 
cgit v0.10.2


From 949854d02455080d20cd3e1db28a3a18daf7599d Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:37 +1100
Subject: fs: Use rename lock and RCU for multi-step operations

The remaining usages for dcache_lock is to allow atomic, multi-step read-side
operations over the directory tree by excluding modifications to the tree.
Also, to walk in the leaf->root direction in the tree where we don't have
a natural d_lock ordering.

This could be accomplished by taking every d_lock, but this would mean a
huge number of locks and actually gets very tricky.

Solve this instead by using the rename seqlock for multi-step read-side
operations, retry in case of a rename so we don't walk up the wrong parent.
Concurrent dentry insertions are not serialised against.  Concurrent deletes
are tricky when walking up the directory: our parent might have been deleted
when dropping locks so also need to check and retry for that.

We can also use the rename lock in cases where livelock is a worry (and it
is introduced in subsequent patch).

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/drivers/staging/pohmelfs/path_entry.c b/drivers/staging/pohmelfs/path_entry.c
index 8ec83d2..bbe42f4 100644
--- a/drivers/staging/pohmelfs/path_entry.c
+++ b/drivers/staging/pohmelfs/path_entry.c
@@ -83,10 +83,11 @@ out:
 int pohmelfs_path_length(struct pohmelfs_inode *pi)
 {
 	struct dentry *d, *root, *first;
-	int len = 1; /* Root slash */
+	int len;
+	unsigned seq;
 
-	first = d = d_find_alias(&pi->vfs_inode);
-	if (!d) {
+	first = d_find_alias(&pi->vfs_inode);
+	if (!first) {
 		dprintk("%s: ino: %llu, mode: %o.\n", __func__, pi->ino, pi->vfs_inode.i_mode);
 		return -ENOENT;
 	}
@@ -95,6 +96,11 @@ int pohmelfs_path_length(struct pohmelfs_inode *pi)
 	root = dget(current->fs->root.dentry);
 	spin_unlock(&current->fs->lock);
 
+rename_retry:
+	len = 1; /* Root slash */
+	d = first;
+	seq = read_seqbegin(&rename_lock);
+	rcu_read_lock();
 	spin_lock(&dcache_lock);
 
 	if (!IS_ROOT(d) && d_unhashed(d))
@@ -105,6 +111,9 @@ int pohmelfs_path_length(struct pohmelfs_inode *pi)
 		d = d->d_parent;
 	}
 	spin_unlock(&dcache_lock);
+	rcu_read_unlock();
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 
 	dput(root);
 	dput(first);
diff --git a/fs/autofs4/waitq.c b/fs/autofs4/waitq.c
index 2341375..4be8f77 100644
--- a/fs/autofs4/waitq.c
+++ b/fs/autofs4/waitq.c
@@ -186,16 +186,25 @@ static int autofs4_getpath(struct autofs_sb_info *sbi,
 {
 	struct dentry *root = sbi->sb->s_root;
 	struct dentry *tmp;
-	char *buf = *name;
+	char *buf;
 	char *p;
-	int len = 0;
-
+	int len;
+	unsigned seq;
+
+rename_retry:
+	buf = *name;
+	len = 0;
+	seq = read_seqbegin(&rename_lock);
+	rcu_read_lock();
 	spin_lock(&dcache_lock);
 	for (tmp = dentry ; tmp != root ; tmp = tmp->d_parent)
 		len += tmp->d_name.len + 1;
 
 	if (!len || --len > NAME_MAX) {
 		spin_unlock(&dcache_lock);
+		rcu_read_unlock();
+		if (read_seqretry(&rename_lock, seq))
+			goto rename_retry;
 		return 0;
 	}
 
@@ -209,6 +218,9 @@ static int autofs4_getpath(struct autofs_sb_info *sbi,
 		strncpy(p, tmp->d_name.name, tmp->d_name.len);
 	}
 	spin_unlock(&dcache_lock);
+	rcu_read_unlock();
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 
 	return len;
 }
diff --git a/fs/dcache.c b/fs/dcache.c
index a09f077..a9bc4ec 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -80,6 +80,7 @@ static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
+EXPORT_SYMBOL(rename_lock);
 EXPORT_SYMBOL(dcache_inode_lock);
 EXPORT_SYMBOL(dcache_lock);
 
@@ -243,6 +244,7 @@ static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 	__releases(dcache_inode_lock)
 	__releases(dcache_lock)
 {
+	dentry->d_parent = NULL;
 	list_del(&dentry->d_u.d_child);
 	if (parent)
 		spin_unlock(&parent->d_lock);
@@ -1017,11 +1019,15 @@ void shrink_dcache_for_umount(struct super_block *sb)
  * Return true if the parent or its subdirectories contain
  * a mount point
  */
- 
 int have_submounts(struct dentry *parent)
 {
-	struct dentry *this_parent = parent;
+	struct dentry *this_parent;
 	struct list_head *next;
+	unsigned seq;
+
+rename_retry:
+	this_parent = parent;
+	seq = read_seqbegin(&rename_lock);
 
 	spin_lock(&dcache_lock);
 	if (d_mountpoint(parent))
@@ -1055,17 +1061,37 @@ resume:
 	 * All done at this level ... ascend and resume the search.
 	 */
 	if (this_parent != parent) {
-		next = this_parent->d_u.d_child.next;
+		struct dentry *tmp;
+		struct dentry *child;
+
+		tmp = this_parent->d_parent;
+		rcu_read_lock();
 		spin_unlock(&this_parent->d_lock);
-		this_parent = this_parent->d_parent;
+		child = this_parent;
+		this_parent = tmp;
 		spin_lock(&this_parent->d_lock);
+		/* might go back up the wrong parent if we have had a rename
+		 * or deletion */
+		if (this_parent != child->d_parent ||
+				read_seqretry(&rename_lock, seq)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_unlock(&dcache_lock);
+			rcu_read_unlock();
+			goto rename_retry;
+		}
+		rcu_read_unlock();
+		next = child->d_u.d_child.next;
 		goto resume;
 	}
 	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 	return 0; /* No mount points found in tree */
 positive:
 	spin_unlock(&dcache_lock);
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 	return 1;
 }
 EXPORT_SYMBOL(have_submounts);
@@ -1086,10 +1112,15 @@ EXPORT_SYMBOL(have_submounts);
  */
 static int select_parent(struct dentry * parent)
 {
-	struct dentry *this_parent = parent;
+	struct dentry *this_parent;
 	struct list_head *next;
+	unsigned seq;
 	int found = 0;
 
+rename_retry:
+	this_parent = parent;
+	seq = read_seqbegin(&rename_lock);
+
 	spin_lock(&dcache_lock);
 	spin_lock(&this_parent->d_lock);
 repeat:
@@ -1099,7 +1130,6 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
-		BUG_ON(this_parent == dentry);
 
 		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 
@@ -1142,17 +1172,32 @@ resume:
 	 */
 	if (this_parent != parent) {
 		struct dentry *tmp;
-		next = this_parent->d_u.d_child.next;
+		struct dentry *child;
+
 		tmp = this_parent->d_parent;
+		rcu_read_lock();
 		spin_unlock(&this_parent->d_lock);
-		BUG_ON(tmp == this_parent);
+		child = this_parent;
 		this_parent = tmp;
 		spin_lock(&this_parent->d_lock);
+		/* might go back up the wrong parent if we have had a rename
+		 * or deletion */
+		if (this_parent != child->d_parent ||
+				read_seqretry(&rename_lock, seq)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_unlock(&dcache_lock);
+			rcu_read_unlock();
+			goto rename_retry;
+		}
+		rcu_read_unlock();
+		next = child->d_u.d_child.next;
 		goto resume;
 	}
 out:
 	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 	return found;
 }
 
@@ -1654,7 +1699,7 @@ EXPORT_SYMBOL(d_add_ci);
 struct dentry * d_lookup(struct dentry * parent, struct qstr * name)
 {
 	struct dentry * dentry = NULL;
-	unsigned long seq;
+	unsigned seq;
 
         do {
                 seq = read_seqbegin(&rename_lock);
@@ -2290,7 +2335,7 @@ static int prepend_name(char **buffer, int *buflen, struct qstr *name)
  * @buffer: pointer to the end of the buffer
  * @buflen: pointer to buffer length
  *
- * Caller holds the dcache_lock.
+ * Caller holds the rename_lock.
  *
  * If path is not reachable from the supplied root, then the value of
  * root is changed (without modifying refcounts).
@@ -2377,7 +2422,9 @@ char *__d_path(const struct path *path, struct path *root,
 
 	prepend(&res, &buflen, "\0", 1);
 	spin_lock(&dcache_lock);
+	write_seqlock(&rename_lock);
 	error = prepend_path(path, root, &res, &buflen);
+	write_sequnlock(&rename_lock);
 	spin_unlock(&dcache_lock);
 
 	if (error)
@@ -2441,10 +2488,12 @@ char *d_path(const struct path *path, char *buf, int buflen)
 
 	get_fs_root(current->fs, &root);
 	spin_lock(&dcache_lock);
+	write_seqlock(&rename_lock);
 	tmp = root;
 	error = path_with_deleted(path, &tmp, &res, &buflen);
 	if (error)
 		res = ERR_PTR(error);
+	write_sequnlock(&rename_lock);
 	spin_unlock(&dcache_lock);
 	path_put(&root);
 	return res;
@@ -2472,10 +2521,12 @@ char *d_path_with_unreachable(const struct path *path, char *buf, int buflen)
 
 	get_fs_root(current->fs, &root);
 	spin_lock(&dcache_lock);
+	write_seqlock(&rename_lock);
 	tmp = root;
 	error = path_with_deleted(path, &tmp, &res, &buflen);
 	if (!error && !path_equal(&tmp, &root))
 		error = prepend_unreachable(&res, &buflen);
+	write_sequnlock(&rename_lock);
 	spin_unlock(&dcache_lock);
 	path_put(&root);
 	if (error)
@@ -2544,7 +2595,9 @@ char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
 	char *retval;
 
 	spin_lock(&dcache_lock);
+	write_seqlock(&rename_lock);
 	retval = __dentry_path(dentry, buf, buflen);
+	write_sequnlock(&rename_lock);
 	spin_unlock(&dcache_lock);
 
 	return retval;
@@ -2557,6 +2610,7 @@ char *dentry_path(struct dentry *dentry, char *buf, int buflen)
 	char *retval;
 
 	spin_lock(&dcache_lock);
+	write_seqlock(&rename_lock);
 	if (d_unlinked(dentry)) {
 		p = buf + buflen;
 		if (prepend(&p, &buflen, "//deleted", 10) != 0)
@@ -2564,6 +2618,7 @@ char *dentry_path(struct dentry *dentry, char *buf, int buflen)
 		buflen++;
 	}
 	retval = __dentry_path(dentry, buf, buflen);
+	write_sequnlock(&rename_lock);
 	spin_unlock(&dcache_lock);
 	if (!IS_ERR(retval) && p)
 		*p = '/';	/* restore '/' overriden with '\0' */
@@ -2604,6 +2659,7 @@ SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
 
 	error = -ENOENT;
 	spin_lock(&dcache_lock);
+	write_seqlock(&rename_lock);
 	if (!d_unlinked(pwd.dentry)) {
 		unsigned long len;
 		struct path tmp = root;
@@ -2612,6 +2668,7 @@ SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
 
 		prepend(&cwd, &buflen, "\0", 1);
 		error = prepend_path(&pwd, &tmp, &cwd, &buflen);
+		write_sequnlock(&rename_lock);
 		spin_unlock(&dcache_lock);
 
 		if (error)
@@ -2631,8 +2688,10 @@ SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
 			if (copy_to_user(buf, cwd, len))
 				error = -EFAULT;
 		}
-	} else
+	} else {
+		write_sequnlock(&rename_lock);
 		spin_unlock(&dcache_lock);
+	}
 
 out:
 	path_put(&pwd);
@@ -2660,25 +2719,25 @@ out:
 int is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
 {
 	int result;
-	unsigned long seq;
+	unsigned seq;
 
 	if (new_dentry == old_dentry)
 		return 1;
 
-	/*
-	 * Need rcu_readlock to protect against the d_parent trashing
-	 * due to d_move
-	 */
-	rcu_read_lock();
 	do {
 		/* for restarting inner loop in case of seq retry */
 		seq = read_seqbegin(&rename_lock);
+		/*
+		 * Need rcu_readlock to protect against the d_parent trashing
+		 * due to d_move
+		 */
+		rcu_read_lock();
 		if (d_ancestor(old_dentry, new_dentry))
 			result = 1;
 		else
 			result = 0;
+		rcu_read_unlock();
 	} while (read_seqretry(&rename_lock, seq));
-	rcu_read_unlock();
 
 	return result;
 }
@@ -2710,9 +2769,13 @@ EXPORT_SYMBOL(path_is_under);
 
 void d_genocide(struct dentry *root)
 {
-	struct dentry *this_parent = root;
+	struct dentry *this_parent;
 	struct list_head *next;
+	unsigned seq;
 
+rename_retry:
+	this_parent = root;
+	seq = read_seqbegin(&rename_lock);
 	spin_lock(&dcache_lock);
 	spin_lock(&this_parent->d_lock);
 repeat:
@@ -2722,6 +2785,7 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
+
 		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 		if (d_unhashed(dentry) || !dentry->d_inode) {
 			spin_unlock(&dentry->d_lock);
@@ -2734,19 +2798,43 @@ resume:
 			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
-		dentry->d_count--;
+		if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
+			dentry->d_flags |= DCACHE_GENOCIDE;
+			dentry->d_count--;
+		}
 		spin_unlock(&dentry->d_lock);
 	}
 	if (this_parent != root) {
-		next = this_parent->d_u.d_child.next;
-		this_parent->d_count--;
+		struct dentry *tmp;
+		struct dentry *child;
+
+		tmp = this_parent->d_parent;
+		if (!(this_parent->d_flags & DCACHE_GENOCIDE)) {
+			this_parent->d_flags |= DCACHE_GENOCIDE;
+			this_parent->d_count--;
+		}
+		rcu_read_lock();
 		spin_unlock(&this_parent->d_lock);
-		this_parent = this_parent->d_parent;
+		child = this_parent;
+		this_parent = tmp;
 		spin_lock(&this_parent->d_lock);
+		/* might go back up the wrong parent if we have had a rename
+		 * or deletion */
+		if (this_parent != child->d_parent ||
+				read_seqretry(&rename_lock, seq)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_unlock(&dcache_lock);
+			rcu_read_unlock();
+			goto rename_retry;
+		}
+		rcu_read_unlock();
+		next = child->d_u.d_child.next;
 		goto resume;
 	}
 	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 }
 
 /**
diff --git a/fs/nfs/namespace.c b/fs/nfs/namespace.c
index db6aa36..78c0ebb 100644
--- a/fs/nfs/namespace.c
+++ b/fs/nfs/namespace.c
@@ -49,11 +49,17 @@ char *nfs_path(const char *base,
 	       const struct dentry *dentry,
 	       char *buffer, ssize_t buflen)
 {
-	char *end = buffer+buflen;
+	char *end;
 	int namelen;
+	unsigned seq;
 
+rename_retry:
+	end = buffer+buflen;
 	*--end = '\0';
 	buflen--;
+
+	seq = read_seqbegin(&rename_lock);
+	rcu_read_lock();
 	spin_lock(&dcache_lock);
 	while (!IS_ROOT(dentry) && dentry != droot) {
 		namelen = dentry->d_name.len;
@@ -66,6 +72,9 @@ char *nfs_path(const char *base,
 		dentry = dentry->d_parent;
 	}
 	spin_unlock(&dcache_lock);
+	rcu_read_unlock();
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 	if (*end != '/') {
 		if (--buflen < 0)
 			goto Elong;
@@ -83,6 +92,9 @@ char *nfs_path(const char *base,
 	return end;
 Elong_unlock:
 	spin_unlock(&dcache_lock);
+	rcu_read_unlock();
+	if (read_seqretry(&rename_lock, seq))
+		goto rename_retry;
 Elong:
 	return ERR_PTR(-ENAMETOOLONG);
 }
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index bda5ec0..c963eba 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -180,6 +180,7 @@ struct dentry_operations {
 #define DCACHE_FSNOTIFY_PARENT_WATCHED	0x0080 /* Parent inode is watched by some fsnotify listener */
 
 #define DCACHE_CANT_MOUNT	0x0100
+#define DCACHE_GENOCIDE		0x0200
 
 extern spinlock_t dcache_inode_lock;
 extern spinlock_t dcache_lock;
-- 
cgit v0.10.2


From b5c84bf6f6fa3a7dfdcb556023a62953574b60ee Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:38 +1100
Subject: fs: dcache remove dcache_lock

dcache_lock no longer protects anything. remove it.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index a15ee20..bdad641 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -21,14 +21,14 @@ prototypes:
 	char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
 
 locking rules:
-		dcache_lock	rename_lock	->d_lock	may block
-d_revalidate:	no		no		no		yes
-d_hash		no		no		no		no
-d_compare:	no		yes		no		no 
-d_delete:	yes		no		yes		no
-d_release:	no		no		no		yes
-d_iput:		no		no		no		yes
-d_dname:	no		no		no		no
+		rename_lock	->d_lock	may block
+d_revalidate:	no		no		yes
+d_hash		no		no		no
+d_compare:	yes		no		no
+d_delete:	no		yes		no
+d_release:	no		no		yes
+d_iput:		no		no		yes
+d_dname:	no		no		no
 
 --------------------------- inode_operations --------------------------- 
 prototypes:
diff --git a/Documentation/filesystems/dentry-locking.txt b/Documentation/filesystems/dentry-locking.txt
index 79334ed..30b6a40 100644
--- a/Documentation/filesystems/dentry-locking.txt
+++ b/Documentation/filesystems/dentry-locking.txt
@@ -31,6 +31,7 @@ significant change is the way d_lookup traverses the hash chain, it
 doesn't acquire the dcache_lock for this and rely on RCU to ensure
 that the dentry has not been *freed*.
 
+dcache_lock no longer exists, dentry locking is explained in fs/dcache.c
 
 Dcache locking details
 ======================
@@ -50,14 +51,12 @@ Safe lock-free look-up of dcache hash table
 
 Dcache is a complex data structure with the hash table entries also
 linked together in other lists. In 2.4 kernel, dcache_lock protected
-all the lists. We applied RCU only on hash chain walking. The rest of
-the lists are still protected by dcache_lock.  Some of the important
-changes are :
+all the lists. RCU dentry hash walking works like this:
 
 1. The deletion from hash chain is done using hlist_del_rcu() macro
    which doesn't initialize next pointer of the deleted dentry and
    this allows us to walk safely lock-free while a deletion is
-   happening.
+   happening. This is a standard hlist_rcu iteration.
 
 2. Insertion of a dentry into the hash table is done using
    hlist_add_head_rcu() which take care of ordering the writes - the
@@ -66,19 +65,18 @@ changes are :
    which has since been replaced by hlist_for_each_entry_rcu(), while
    walking the hash chain. The only requirement is that all
    initialization to the dentry must be done before
-   hlist_add_head_rcu() since we don't have dcache_lock protection
-   while traversing the hash chain. This isn't different from the
-   existing code.
-
-3. The dentry looked up without holding dcache_lock by cannot be
-   returned for walking if it is unhashed. It then may have a NULL
-   d_inode or other bogosity since RCU doesn't protect the other
-   fields in the dentry. We therefore use a flag DCACHE_UNHASHED to
-   indicate unhashed dentries and use this in conjunction with a
-   per-dentry lock (d_lock). Once looked up without the dcache_lock,
-   we acquire the per-dentry lock (d_lock) and check if the dentry is
-   unhashed. If so, the look-up is failed. If not, the reference count
-   of the dentry is increased and the dentry is returned.
+   hlist_add_head_rcu() since we don't have lock protection
+   while traversing the hash chain.
+
+3. The dentry looked up without holding locks cannot be returned for
+   walking if it is unhashed. It then may have a NULL d_inode or other
+   bogosity since RCU doesn't protect the other fields in the dentry. We
+   therefore use a flag DCACHE_UNHASHED to indicate unhashed dentries
+   and use this in conjunction with a per-dentry lock (d_lock). Once
+   looked up without locks, we acquire the per-dentry lock (d_lock) and
+   check if the dentry is unhashed. If so, the look-up is failed. If not,
+   the reference count of the dentry is increased and the dentry is
+   returned.
 
 4. Once a dentry is looked up, it must be ensured during the path walk
    for that component it doesn't go away. In pre-2.5.10 code, this was
@@ -86,10 +84,10 @@ changes are :
    In some sense, dcache_rcu path walking looks like the pre-2.5.10
    version.
 
-5. All dentry hash chain updates must take the dcache_lock as well as
-   the per-dentry lock in that order. dput() does this to ensure that
-   a dentry that has just been looked up in another CPU doesn't get
-   deleted before dget() can be done on it.
+5. All dentry hash chain updates must take the per-dentry lock (see
+   fs/dcache.c). This excludes dput() to ensure that a dentry that has
+   been looked up concurrently does not get deleted before dget() can
+   take a ref.
 
 6. There are several ways to do reference counting of RCU protected
    objects. One such example is in ipv4 route cache where deferred
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index 9fd3194..1eb7695 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -216,7 +216,6 @@ had ->revalidate()) add calls in ->follow_link()/->readlink().
 ->d_parent changes are not protected by BKL anymore.  Read access is safe
 if at least one of the following is true:
 	* filesystem has no cross-directory rename()
-	* dcache_lock is held
 	* we know that parent had been locked (e.g. we are looking at
 ->d_parent of ->lookup() argument).
 	* we are called from ->rename().
@@ -340,3 +339,10 @@ look at examples of other filesystems) for guidance.
 	.d_hash() calling convention and locking rules are significantly
 changed. Read updated documentation in Documentation/filesystems/vfs.txt (and
 look at examples of other filesystems) for guidance.
+
+---
+[mandatory]
+	dcache_lock is gone, replaced by fine grained locks. See fs/dcache.c
+for details of what locks to replace dcache_lock with in order to protect
+particular things. Most of the time, a filesystem only needs ->d_lock, which
+protects *all* the dcache state of a given dentry.
diff --git a/arch/powerpc/platforms/cell/spufs/inode.c b/arch/powerpc/platforms/cell/spufs/inode.c
index 5aef1a7..2662b50 100644
--- a/arch/powerpc/platforms/cell/spufs/inode.c
+++ b/arch/powerpc/platforms/cell/spufs/inode.c
@@ -159,21 +159,18 @@ static void spufs_prune_dir(struct dentry *dir)
 
 	mutex_lock(&dir->d_inode->i_mutex);
 	list_for_each_entry_safe(dentry, tmp, &dir->d_subdirs, d_u.d_child) {
-		spin_lock(&dcache_lock);
 		spin_lock(&dentry->d_lock);
 		if (!(d_unhashed(dentry)) && dentry->d_inode) {
 			dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			simple_unlink(dir->d_inode, dentry);
-			/* XXX: what is dcache_lock protecting here? Other
+			/* XXX: what was dcache_lock protecting here? Other
 			 * filesystems (IB, configfs) release dcache_lock
 			 * before unlink */
-			spin_unlock(&dcache_lock);
 			dput(dentry);
 		} else {
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 		}
 	}
 	shrink_dcache_parent(dir);
diff --git a/drivers/infiniband/hw/ipath/ipath_fs.c b/drivers/infiniband/hw/ipath/ipath_fs.c
index 18aee04..925e882 100644
--- a/drivers/infiniband/hw/ipath/ipath_fs.c
+++ b/drivers/infiniband/hw/ipath/ipath_fs.c
@@ -277,18 +277,14 @@ static int remove_file(struct dentry *parent, char *name)
 		goto bail;
 	}
 
-	spin_lock(&dcache_lock);
 	spin_lock(&tmp->d_lock);
 	if (!(d_unhashed(tmp) && tmp->d_inode)) {
 		dget_locked_dlock(tmp);
 		__d_drop(tmp);
 		spin_unlock(&tmp->d_lock);
-		spin_unlock(&dcache_lock);
 		simple_unlink(parent->d_inode, tmp);
-	} else {
+	} else
 		spin_unlock(&tmp->d_lock);
-		spin_unlock(&dcache_lock);
-	}
 
 	ret = 0;
 bail:
diff --git a/drivers/infiniband/hw/qib/qib_fs.c b/drivers/infiniband/hw/qib/qib_fs.c
index fe4b242..49af4a6 100644
--- a/drivers/infiniband/hw/qib/qib_fs.c
+++ b/drivers/infiniband/hw/qib/qib_fs.c
@@ -453,17 +453,14 @@ static int remove_file(struct dentry *parent, char *name)
 		goto bail;
 	}
 
-	spin_lock(&dcache_lock);
 	spin_lock(&tmp->d_lock);
 	if (!(d_unhashed(tmp) && tmp->d_inode)) {
 		dget_locked_dlock(tmp);
 		__d_drop(tmp);
 		spin_unlock(&tmp->d_lock);
-		spin_unlock(&dcache_lock);
 		simple_unlink(parent->d_inode, tmp);
 	} else {
 		spin_unlock(&tmp->d_lock);
-		spin_unlock(&dcache_lock);
 	}
 
 	ret = 0;
diff --git a/drivers/staging/pohmelfs/path_entry.c b/drivers/staging/pohmelfs/path_entry.c
index bbe42f4..400a9fc 100644
--- a/drivers/staging/pohmelfs/path_entry.c
+++ b/drivers/staging/pohmelfs/path_entry.c
@@ -101,7 +101,6 @@ rename_retry:
 	d = first;
 	seq = read_seqbegin(&rename_lock);
 	rcu_read_lock();
-	spin_lock(&dcache_lock);
 
 	if (!IS_ROOT(d) && d_unhashed(d))
 		len += UNHASHED_OBSCURE_STRING_SIZE; /* Obscure " (deleted)" string */
@@ -110,7 +109,6 @@ rename_retry:
 		len += d->d_name.len + 1; /* Plus slash */
 		d = d->d_parent;
 	}
-	spin_unlock(&dcache_lock);
 	rcu_read_unlock();
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
diff --git a/drivers/staging/smbfs/cache.c b/drivers/staging/smbfs/cache.c
index 920434b..75dfd40 100644
--- a/drivers/staging/smbfs/cache.c
+++ b/drivers/staging/smbfs/cache.c
@@ -62,7 +62,6 @@ smb_invalidate_dircache_entries(struct dentry *parent)
 	struct list_head *next;
 	struct dentry *dentry;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
@@ -72,7 +71,6 @@ smb_invalidate_dircache_entries(struct dentry *parent)
 		next = next->next;
 	}
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 /*
@@ -98,7 +96,6 @@ smb_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 	}
 
 	/* If a pointer is invalid, we search the dentry. */
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
@@ -115,7 +112,6 @@ smb_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 	dent = NULL;
 out_unlock:
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 	return dent;
 }
 
diff --git a/drivers/usb/core/inode.c b/drivers/usb/core/inode.c
index 89a0e83..1b125c2 100644
--- a/drivers/usb/core/inode.c
+++ b/drivers/usb/core/inode.c
@@ -343,7 +343,6 @@ static int usbfs_empty (struct dentry *dentry)
 {
 	struct list_head *list;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	list_for_each(list, &dentry->d_subdirs) {
 		struct dentry *de = list_entry(list, struct dentry, d_u.d_child);
@@ -352,13 +351,11 @@ static int usbfs_empty (struct dentry *dentry)
 		if (usbfs_positive(de)) {
 			spin_unlock(&de->d_lock);
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 			return 0;
 		}
 		spin_unlock(&de->d_lock);
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 	return 1;
 }
 
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 47dfd5d..1073bca 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -270,13 +270,11 @@ static struct dentry *v9fs_dentry_from_dir_inode(struct inode *inode)
 {
 	struct dentry *dentry;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	/* Directory should have only one entry. */
 	BUG_ON(S_ISDIR(inode->i_mode) && !list_is_singular(&inode->i_dentry));
 	dentry = list_entry(inode->i_dentry.next, struct dentry, d_alias);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	return dentry;
 }
 
diff --git a/fs/affs/amigaffs.c b/fs/affs/amigaffs.c
index 2321cc9..600101a 100644
--- a/fs/affs/amigaffs.c
+++ b/fs/affs/amigaffs.c
@@ -128,7 +128,6 @@ affs_fix_dcache(struct dentry *dentry, u32 entry_ino)
 	void *data = dentry->d_fsdata;
 	struct list_head *head, *next;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	head = &inode->i_dentry;
 	next = head->next;
@@ -141,7 +140,6 @@ affs_fix_dcache(struct dentry *dentry, u32 entry_ino)
 		next = next->next;
 	}
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 }
 
 
diff --git a/fs/autofs4/autofs_i.h b/fs/autofs4/autofs_i.h
index 9d2ae9b..0fffe1c 100644
--- a/fs/autofs4/autofs_i.h
+++ b/fs/autofs4/autofs_i.h
@@ -16,6 +16,7 @@
 #include <linux/auto_fs4.h>
 #include <linux/auto_dev-ioctl.h>
 #include <linux/mutex.h>
+#include <linux/spinlock.h>
 #include <linux/list.h>
 
 /* This is the range of ioctl() numbers we claim as ours */
@@ -60,6 +61,8 @@ do {							\
 		current->pid, __func__, ##args);	\
 } while (0)
 
+extern spinlock_t autofs4_lock;
+
 /* Unified info structure.  This is pointed to by both the dentry and
    inode structures.  Each file in the filesystem has an instance of this
    structure.  It holds a reference to the dentry, so dentries are never
diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index 968c143..2f7951d 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -102,7 +102,7 @@ static struct dentry *get_next_positive_dentry(struct dentry *prev,
 	if (prev == NULL)
 		return dget(prev);
 
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 relock:
 	p = prev;
 	spin_lock(&p->d_lock);
@@ -114,7 +114,7 @@ again:
 
 			if (p == root) {
 				spin_unlock(&p->d_lock);
-				spin_unlock(&dcache_lock);
+				spin_unlock(&autofs4_lock);
 				dput(prev);
 				return NULL;
 			}
@@ -144,7 +144,7 @@ again:
 	dget_dlock(ret);
 	spin_unlock(&ret->d_lock);
 	spin_unlock(&p->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 	dput(prev);
 
@@ -408,13 +408,13 @@ found:
 	ino->flags |= AUTOFS_INF_EXPIRING;
 	init_completion(&ino->expire_complete);
 	spin_unlock(&sbi->fs_lock);
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&expired->d_parent->d_lock);
 	spin_lock_nested(&expired->d_lock, DENTRY_D_LOCK_NESTED);
 	list_move(&expired->d_parent->d_subdirs, &expired->d_u.d_child);
 	spin_unlock(&expired->d_lock);
 	spin_unlock(&expired->d_parent->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 	return expired;
 }
 
diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
index 7a9ed6b..10ca68a 100644
--- a/fs/autofs4/root.c
+++ b/fs/autofs4/root.c
@@ -23,6 +23,8 @@
 
 #include "autofs_i.h"
 
+DEFINE_SPINLOCK(autofs4_lock);
+
 static int autofs4_dir_symlink(struct inode *,struct dentry *,const char *);
 static int autofs4_dir_unlink(struct inode *,struct dentry *);
 static int autofs4_dir_rmdir(struct inode *,struct dentry *);
@@ -142,15 +144,15 @@ static int autofs4_dir_open(struct inode *inode, struct file *file)
 	 * autofs file system so just let the libfs routines handle
 	 * it.
 	 */
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&dentry->d_lock);
 	if (!d_mountpoint(dentry) && list_empty(&dentry->d_subdirs)) {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
+		spin_unlock(&autofs4_lock);
 		return -ENOENT;
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 out:
 	return dcache_dir_open(inode, file);
@@ -255,11 +257,11 @@ static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
 	/* We trigger a mount for almost all flags */
 	lookup_type = autofs4_need_mount(nd->flags);
 	spin_lock(&sbi->fs_lock);
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&dentry->d_lock);
 	if (!(lookup_type || ino->flags & AUTOFS_INF_PENDING)) {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
+		spin_unlock(&autofs4_lock);
 		spin_unlock(&sbi->fs_lock);
 		goto follow;
 	}
@@ -272,7 +274,7 @@ static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
 	if (ino->flags & AUTOFS_INF_PENDING ||
 	    (!d_mountpoint(dentry) && list_empty(&dentry->d_subdirs))) {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
+		spin_unlock(&autofs4_lock);
 		spin_unlock(&sbi->fs_lock);
 
 		status = try_to_fill_dentry(dentry, nd->flags);
@@ -282,7 +284,7 @@ static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
 		goto follow;
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 	spin_unlock(&sbi->fs_lock);
 follow:
 	/*
@@ -353,14 +355,14 @@ static int autofs4_revalidate(struct dentry *dentry, struct nameidata *nd)
 		return 0;
 
 	/* Check for a non-mountpoint directory with no contents */
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&dentry->d_lock);
 	if (S_ISDIR(dentry->d_inode->i_mode) &&
 	    !d_mountpoint(dentry) && list_empty(&dentry->d_subdirs)) {
 		DPRINTK("dentry=%p %.*s, emptydir",
 			 dentry, dentry->d_name.len, dentry->d_name.name);
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
+		spin_unlock(&autofs4_lock);
 
 		/* The daemon never causes a mount to trigger */
 		if (oz_mode)
@@ -377,7 +379,7 @@ static int autofs4_revalidate(struct dentry *dentry, struct nameidata *nd)
 		return status;
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 	return 1;
 }
@@ -432,7 +434,7 @@ static struct dentry *autofs4_lookup_active(struct dentry *dentry)
 	const unsigned char *str = name->name;
 	struct list_head *p, *head;
 
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&sbi->lookup_lock);
 	head = &sbi->active_list;
 	list_for_each(p, head) {
@@ -465,14 +467,14 @@ static struct dentry *autofs4_lookup_active(struct dentry *dentry)
 			dget_dlock(active);
 			spin_unlock(&active->d_lock);
 			spin_unlock(&sbi->lookup_lock);
-			spin_unlock(&dcache_lock);
+			spin_unlock(&autofs4_lock);
 			return active;
 		}
 next:
 		spin_unlock(&active->d_lock);
 	}
 	spin_unlock(&sbi->lookup_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 	return NULL;
 }
@@ -487,7 +489,7 @@ static struct dentry *autofs4_lookup_expiring(struct dentry *dentry)
 	const unsigned char *str = name->name;
 	struct list_head *p, *head;
 
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&sbi->lookup_lock);
 	head = &sbi->expiring_list;
 	list_for_each(p, head) {
@@ -520,14 +522,14 @@ static struct dentry *autofs4_lookup_expiring(struct dentry *dentry)
 			dget_dlock(expiring);
 			spin_unlock(&expiring->d_lock);
 			spin_unlock(&sbi->lookup_lock);
-			spin_unlock(&dcache_lock);
+			spin_unlock(&autofs4_lock);
 			return expiring;
 		}
 next:
 		spin_unlock(&expiring->d_lock);
 	}
 	spin_unlock(&sbi->lookup_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 	return NULL;
 }
@@ -763,12 +765,12 @@ static int autofs4_dir_unlink(struct inode *dir, struct dentry *dentry)
 
 	dir->i_mtime = CURRENT_TIME;
 
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	autofs4_add_expiring(dentry);
 	spin_lock(&dentry->d_lock);
 	__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 	return 0;
 }
@@ -785,20 +787,20 @@ static int autofs4_dir_rmdir(struct inode *dir, struct dentry *dentry)
 	if (!autofs4_oz_mode(sbi))
 		return -EACCES;
 
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	spin_lock(&sbi->lookup_lock);
 	spin_lock(&dentry->d_lock);
 	if (!list_empty(&dentry->d_subdirs)) {
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&sbi->lookup_lock);
-		spin_unlock(&dcache_lock);
+		spin_unlock(&autofs4_lock);
 		return -ENOTEMPTY;
 	}
 	__autofs4_add_expiring(dentry);
 	spin_unlock(&sbi->lookup_lock);
 	__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 
 	if (atomic_dec_and_test(&ino->count)) {
 		p_ino = autofs4_dentry_ino(dentry->d_parent);
diff --git a/fs/autofs4/waitq.c b/fs/autofs4/waitq.c
index 4be8f77..c5f8459 100644
--- a/fs/autofs4/waitq.c
+++ b/fs/autofs4/waitq.c
@@ -194,14 +194,15 @@ static int autofs4_getpath(struct autofs_sb_info *sbi,
 rename_retry:
 	buf = *name;
 	len = 0;
+
 	seq = read_seqbegin(&rename_lock);
 	rcu_read_lock();
-	spin_lock(&dcache_lock);
+	spin_lock(&autofs4_lock);
 	for (tmp = dentry ; tmp != root ; tmp = tmp->d_parent)
 		len += tmp->d_name.len + 1;
 
 	if (!len || --len > NAME_MAX) {
-		spin_unlock(&dcache_lock);
+		spin_unlock(&autofs4_lock);
 		rcu_read_unlock();
 		if (read_seqretry(&rename_lock, seq))
 			goto rename_retry;
@@ -217,7 +218,7 @@ rename_retry:
 		p -= tmp->d_name.len;
 		strncpy(p, tmp->d_name.name, tmp->d_name.len);
 	}
-	spin_unlock(&dcache_lock);
+	spin_unlock(&autofs4_lock);
 	rcu_read_unlock();
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 2c924e8..58abc3d 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -112,7 +112,6 @@ static int __dcache_readdir(struct file *filp,
 	dout("__dcache_readdir %p at %llu (last %p)\n", dir, filp->f_pos,
 	     last);
 
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 
 	/* start at beginning? */
@@ -156,7 +155,6 @@ more:
 	dget_dlock(dentry);
 	spin_unlock(&dentry->d_lock);
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 
 	dout(" %llu (%llu) dentry %p %.*s %p\n", di->offset, filp->f_pos,
 	     dentry, dentry->d_name.len, dentry->d_name.name, dentry->d_inode);
@@ -182,21 +180,19 @@ more:
 
 	filp->f_pos++;
 
-	/* make sure a dentry wasn't dropped while we didn't have dcache_lock */
+	/* make sure a dentry wasn't dropped while we didn't have parent lock */
 	if (!ceph_i_test(dir, CEPH_I_COMPLETE)) {
 		dout(" lost I_COMPLETE on %p; falling back to mds\n", dir);
 		err = -EAGAIN;
 		goto out;
 	}
 
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	p = p->prev;	/* advance to next dentry */
 	goto more;
 
 out_unlock:
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 out:
 	if (last)
 		dput(last);
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 2c69444..2a48caf 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -841,7 +841,6 @@ static void ceph_set_dentry_offset(struct dentry *dn)
 	di->offset = ceph_inode(inode)->i_max_offset++;
 	spin_unlock(&inode->i_lock);
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dir->d_lock);
 	spin_lock_nested(&dn->d_lock, DENTRY_D_LOCK_NESTED);
 	list_move(&dn->d_u.d_child, &dir->d_subdirs);
@@ -849,7 +848,6 @@ static void ceph_set_dentry_offset(struct dentry *dn)
 	     dn->d_u.d_child.prev, dn->d_u.d_child.next);
 	spin_unlock(&dn->d_lock);
 	spin_unlock(&dir->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 /*
@@ -1233,13 +1231,11 @@ retry_lookup:
 			goto retry_lookup;
 		} else {
 			/* reorder parent's d_subdirs */
-			spin_lock(&dcache_lock);
 			spin_lock(&parent->d_lock);
 			spin_lock_nested(&dn->d_lock, DENTRY_D_LOCK_NESTED);
 			list_move(&dn->d_u.d_child, &parent->d_subdirs);
 			spin_unlock(&dn->d_lock);
 			spin_unlock(&parent->d_lock);
-			spin_unlock(&dcache_lock);
 		}
 
 		di = dn->d_fsdata;
diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index 0036983..99b9a2c 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -809,17 +809,14 @@ inode_has_hashed_dentries(struct inode *inode)
 {
 	struct dentry *dentry;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		if (!d_unhashed(dentry) || IS_ROOT(dentry)) {
 			spin_unlock(&dcache_inode_lock);
-			spin_unlock(&dcache_lock);
 			return true;
 		}
 	}
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	return false;
 }
 
diff --git a/fs/coda/cache.c b/fs/coda/cache.c
index 859393f..5525e1c 100644
--- a/fs/coda/cache.c
+++ b/fs/coda/cache.c
@@ -93,7 +93,6 @@ static void coda_flag_children(struct dentry *parent, int flag)
 	struct list_head *child;
 	struct dentry *de;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	list_for_each(child, &parent->d_subdirs)
 	{
@@ -104,7 +103,6 @@ static void coda_flag_children(struct dentry *parent, int flag)
 		coda_flag_inode(de->d_inode, flag);
 	}
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 	return; 
 }
 
diff --git a/fs/configfs/configfs_internal.h b/fs/configfs/configfs_internal.h
index e58b4c3..026cf68 100644
--- a/fs/configfs/configfs_internal.h
+++ b/fs/configfs/configfs_internal.h
@@ -120,7 +120,6 @@ static inline struct config_item *configfs_get_config_item(struct dentry *dentry
 {
 	struct config_item * item = NULL;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	if (!d_unhashed(dentry)) {
 		struct configfs_dirent * sd = dentry->d_fsdata;
@@ -131,7 +130,6 @@ static inline struct config_item *configfs_get_config_item(struct dentry *dentry
 			item = config_item_get(sd->s_element);
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 
 	return item;
 }
diff --git a/fs/configfs/inode.c b/fs/configfs/inode.c
index 79b3776..fb3a55f 100644
--- a/fs/configfs/inode.c
+++ b/fs/configfs/inode.c
@@ -250,18 +250,14 @@ void configfs_drop_dentry(struct configfs_dirent * sd, struct dentry * parent)
 	struct dentry * dentry = sd->s_dentry;
 
 	if (dentry) {
-		spin_lock(&dcache_lock);
 		spin_lock(&dentry->d_lock);
 		if (!(d_unhashed(dentry) && dentry->d_inode)) {
 			dget_locked_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 			simple_unlink(parent->d_inode, dentry);
-		} else {
+		} else
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
-		}
 	}
 }
 
diff --git a/fs/dcache.c b/fs/dcache.c
index a9bc4ec..0dbae053b 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -54,11 +54,10 @@
  *   - d_alias, d_inode
  *
  * Ordering:
- * dcache_lock
- *   dcache_inode_lock
- *     dentry->d_lock
- *       dcache_lru_lock
- *       dcache_hash_lock
+ * dcache_inode_lock
+ *   dentry->d_lock
+ *     dcache_lru_lock
+ *     dcache_hash_lock
  *
  * If there is an ancestor relationship:
  * dentry->d_parent->...->d_parent->d_lock
@@ -77,12 +76,10 @@ EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_inode_lock);
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_hash_lock);
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
-__cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
 EXPORT_SYMBOL(rename_lock);
 EXPORT_SYMBOL(dcache_inode_lock);
-EXPORT_SYMBOL(dcache_lock);
 
 static struct kmem_cache *dentry_cache __read_mostly;
 
@@ -139,7 +136,7 @@ static void __d_free(struct rcu_head *head)
 }
 
 /*
- * no dcache_lock, please.
+ * no locks, please.
  */
 static void d_free(struct dentry *dentry)
 {
@@ -162,7 +159,6 @@ static void d_free(struct dentry *dentry)
 static void dentry_iput(struct dentry * dentry)
 	__releases(dentry->d_lock)
 	__releases(dcache_inode_lock)
-	__releases(dcache_lock)
 {
 	struct inode *inode = dentry->d_inode;
 	if (inode) {
@@ -170,7 +166,6 @@ static void dentry_iput(struct dentry * dentry)
 		list_del_init(&dentry->d_alias);
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 		if (!inode->i_nlink)
 			fsnotify_inoderemove(inode);
 		if (dentry->d_op && dentry->d_op->d_iput)
@@ -180,7 +175,6 @@ static void dentry_iput(struct dentry * dentry)
 	} else {
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 	}
 }
 
@@ -235,14 +229,13 @@ static void dentry_lru_move_tail(struct dentry *dentry)
  *
  * If this is the root of the dentry tree, return NULL.
  *
- * dcache_lock and d_lock and d_parent->d_lock must be held by caller, and
- * are dropped by d_kill.
+ * dentry->d_lock and parent->d_lock must be held by caller, and are dropped by
+ * d_kill.
  */
 static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
 	__releases(parent->d_lock)
 	__releases(dcache_inode_lock)
-	__releases(dcache_lock)
 {
 	dentry->d_parent = NULL;
 	list_del(&dentry->d_u.d_child);
@@ -285,11 +278,9 @@ EXPORT_SYMBOL(__d_drop);
 
 void d_drop(struct dentry *dentry)
 {
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(d_drop);
 
@@ -337,21 +328,10 @@ repeat:
 	else
 		parent = dentry->d_parent;
 	if (dentry->d_count == 1) {
-		if (!spin_trylock(&dcache_lock)) {
-			/*
-			 * Something of a livelock possibility we could avoid
-			 * by taking dcache_lock and trying again, but we
-			 * want to reduce dcache_lock anyway so this will
-			 * get improved.
-			 */
-drop1:
-			spin_unlock(&dentry->d_lock);
-			goto repeat;
-		}
 		if (!spin_trylock(&dcache_inode_lock)) {
 drop2:
-			spin_unlock(&dcache_lock);
-			goto drop1;
+			spin_unlock(&dentry->d_lock);
+			goto repeat;
 		}
 		if (parent && !spin_trylock(&parent->d_lock)) {
 			spin_unlock(&dcache_inode_lock);
@@ -363,7 +343,6 @@ drop2:
 		spin_unlock(&dentry->d_lock);
 		if (parent)
 			spin_unlock(&parent->d_lock);
-		spin_unlock(&dcache_lock);
 		return;
 	}
 
@@ -387,7 +366,6 @@ drop2:
 	if (parent)
 		spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	return;
 
 unhash_it:
@@ -418,11 +396,9 @@ int d_invalidate(struct dentry * dentry)
 	/*
 	 * If it's already been dropped, return OK.
 	 */
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	if (d_unhashed(dentry)) {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
 		return 0;
 	}
 	/*
@@ -431,9 +407,7 @@ int d_invalidate(struct dentry * dentry)
 	 */
 	if (!list_empty(&dentry->d_subdirs)) {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
 		shrink_dcache_parent(dentry);
-		spin_lock(&dcache_lock);
 		spin_lock(&dentry->d_lock);
 	}
 
@@ -450,19 +424,17 @@ int d_invalidate(struct dentry * dentry)
 	if (dentry->d_count > 1) {
 		if (dentry->d_inode && S_ISDIR(dentry->d_inode->i_mode)) {
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 			return -EBUSY;
 		}
 	}
 
 	__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 	return 0;
 }
 EXPORT_SYMBOL(d_invalidate);
 
-/* This must be called with dcache_lock and d_lock held */
+/* This must be called with d_lock held */
 static inline struct dentry * __dget_locked_dlock(struct dentry *dentry)
 {
 	dentry->d_count++;
@@ -470,7 +442,7 @@ static inline struct dentry * __dget_locked_dlock(struct dentry *dentry)
 	return dentry;
 }
 
-/* This should be called _only_ with dcache_lock held */
+/* This must be called with d_lock held */
 static inline struct dentry * __dget_locked(struct dentry *dentry)
 {
 	spin_lock(&dentry->d_lock);
@@ -575,11 +547,9 @@ struct dentry *d_find_alias(struct inode *inode)
 	struct dentry *de = NULL;
 
 	if (!list_empty(&inode->i_dentry)) {
-		spin_lock(&dcache_lock);
 		spin_lock(&dcache_inode_lock);
 		de = __d_find_alias(inode, 0);
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 	}
 	return de;
 }
@@ -593,7 +563,6 @@ void d_prune_aliases(struct inode *inode)
 {
 	struct dentry *dentry;
 restart:
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		spin_lock(&dentry->d_lock);
@@ -602,14 +571,12 @@ restart:
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_inode_lock);
-			spin_unlock(&dcache_lock);
 			dput(dentry);
 			goto restart;
 		}
 		spin_unlock(&dentry->d_lock);
 	}
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(d_prune_aliases);
 
@@ -625,17 +592,14 @@ static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
 	__releases(parent->d_lock)
 	__releases(dcache_inode_lock)
-	__releases(dcache_lock)
 {
 	__d_drop(dentry);
 	dentry = d_kill(dentry, parent);
 
 	/*
-	 * Prune ancestors.  Locking is simpler than in dput(),
-	 * because dcache_lock needs to be taken anyway.
+	 * Prune ancestors.
 	 */
 	while (dentry) {
-		spin_lock(&dcache_lock);
 		spin_lock(&dcache_inode_lock);
 again:
 		spin_lock(&dentry->d_lock);
@@ -653,7 +617,6 @@ again:
 				spin_unlock(&parent->d_lock);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_inode_lock);
-			spin_unlock(&dcache_lock);
 			return;
 		}
 
@@ -702,8 +665,7 @@ relock:
 		spin_unlock(&dcache_lru_lock);
 
 		prune_one_dentry(dentry, parent);
-		/* dcache_lock, dcache_inode_lock and dentry->d_lock dropped */
-		spin_lock(&dcache_lock);
+		/* dcache_inode_lock and dentry->d_lock dropped */
 		spin_lock(&dcache_inode_lock);
 		spin_lock(&dcache_lru_lock);
 	}
@@ -725,7 +687,6 @@ static void __shrink_dcache_sb(struct super_block *sb, int *count, int flags)
 	LIST_HEAD(tmp);
 	int cnt = *count;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 relock:
 	spin_lock(&dcache_lru_lock);
@@ -766,7 +727,6 @@ relock:
 		list_splice(&referenced, &sb->s_dentry_lru);
 	spin_unlock(&dcache_lru_lock);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 }
 
 /**
@@ -788,7 +748,6 @@ static void prune_dcache(int count)
 
 	if (unused == 0 || count == 0)
 		return;
-	spin_lock(&dcache_lock);
 	if (count >= unused)
 		prune_ratio = 1;
 	else
@@ -825,11 +784,9 @@ static void prune_dcache(int count)
 		if (down_read_trylock(&sb->s_umount)) {
 			if ((sb->s_root != NULL) &&
 			    (!list_empty(&sb->s_dentry_lru))) {
-				spin_unlock(&dcache_lock);
 				__shrink_dcache_sb(sb, &w_count,
 						DCACHE_REFERENCED);
 				pruned -= w_count;
-				spin_lock(&dcache_lock);
 			}
 			up_read(&sb->s_umount);
 		}
@@ -845,7 +802,6 @@ static void prune_dcache(int count)
 	if (p)
 		__put_super(p);
 	spin_unlock(&sb_lock);
-	spin_unlock(&dcache_lock);
 }
 
 /**
@@ -859,7 +815,6 @@ void shrink_dcache_sb(struct super_block *sb)
 {
 	LIST_HEAD(tmp);
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
@@ -868,7 +823,6 @@ void shrink_dcache_sb(struct super_block *sb)
 	}
 	spin_unlock(&dcache_lru_lock);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(shrink_dcache_sb);
 
@@ -885,12 +839,10 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 	BUG_ON(!IS_ROOT(dentry));
 
 	/* detach this root from the system */
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	dentry_lru_del(dentry);
 	__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 
 	for (;;) {
 		/* descend to the first leaf in the current subtree */
@@ -899,7 +851,6 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 
 			/* this is a branch with children - detach all of them
 			 * from the system in one go */
-			spin_lock(&dcache_lock);
 			spin_lock(&dentry->d_lock);
 			list_for_each_entry(loop, &dentry->d_subdirs,
 					    d_u.d_child) {
@@ -910,7 +861,6 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 				spin_unlock(&loop->d_lock);
 			}
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 
 			/* move to the first child */
 			dentry = list_entry(dentry->d_subdirs.next,
@@ -977,8 +927,7 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 
 /*
  * destroy the dentries attached to a superblock on unmounting
- * - we don't need to use dentry->d_lock, and only need dcache_lock when
- *   removing the dentry from the system lists and hashes because:
+ * - we don't need to use dentry->d_lock because:
  *   - the superblock is detached from all mountings and open files, so the
  *     dentry trees will not be rearranged by the VFS
  *   - s_umount is write-locked, so the memory pressure shrinker will ignore
@@ -1029,7 +978,6 @@ rename_retry:
 	this_parent = parent;
 	seq = read_seqbegin(&rename_lock);
 
-	spin_lock(&dcache_lock);
 	if (d_mountpoint(parent))
 		goto positive;
 	spin_lock(&this_parent->d_lock);
@@ -1075,7 +1023,6 @@ resume:
 		if (this_parent != child->d_parent ||
 				read_seqretry(&rename_lock, seq)) {
 			spin_unlock(&this_parent->d_lock);
-			spin_unlock(&dcache_lock);
 			rcu_read_unlock();
 			goto rename_retry;
 		}
@@ -1084,12 +1031,10 @@ resume:
 		goto resume;
 	}
 	spin_unlock(&this_parent->d_lock);
-	spin_unlock(&dcache_lock);
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
 	return 0; /* No mount points found in tree */
 positive:
-	spin_unlock(&dcache_lock);
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
 	return 1;
@@ -1121,7 +1066,6 @@ rename_retry:
 	this_parent = parent;
 	seq = read_seqbegin(&rename_lock);
 
-	spin_lock(&dcache_lock);
 	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
@@ -1185,7 +1129,6 @@ resume:
 		if (this_parent != child->d_parent ||
 				read_seqretry(&rename_lock, seq)) {
 			spin_unlock(&this_parent->d_lock);
-			spin_unlock(&dcache_lock);
 			rcu_read_unlock();
 			goto rename_retry;
 		}
@@ -1195,7 +1138,6 @@ resume:
 	}
 out:
 	spin_unlock(&this_parent->d_lock);
-	spin_unlock(&dcache_lock);
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
 	return found;
@@ -1297,7 +1239,6 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	INIT_LIST_HEAD(&dentry->d_u.d_child);
 
 	if (parent) {
-		spin_lock(&dcache_lock);
 		spin_lock(&parent->d_lock);
 		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 		dentry->d_parent = dget_dlock(parent);
@@ -1305,7 +1246,6 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 		list_add(&dentry->d_u.d_child, &parent->d_subdirs);
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&parent->d_lock);
-		spin_unlock(&dcache_lock);
 	}
 
 	this_cpu_inc(nr_dentry);
@@ -1325,7 +1265,6 @@ struct dentry *d_alloc_name(struct dentry *parent, const char *name)
 }
 EXPORT_SYMBOL(d_alloc_name);
 
-/* the caller must hold dcache_lock */
 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 {
 	spin_lock(&dentry->d_lock);
@@ -1354,11 +1293,9 @@ static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 void d_instantiate(struct dentry *entry, struct inode * inode)
 {
 	BUG_ON(!list_empty(&entry->d_alias));
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	__d_instantiate(entry, inode);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	security_d_instantiate(entry, inode);
 }
 EXPORT_SYMBOL(d_instantiate);
@@ -1422,11 +1359,9 @@ struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
 
 	BUG_ON(!list_empty(&entry->d_alias));
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	result = __d_instantiate_unique(entry, inode);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 
 	if (!result) {
 		security_d_instantiate(entry, inode);
@@ -1515,12 +1450,11 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	}
 	tmp->d_parent = tmp; /* make sure dput doesn't croak */
 
-	spin_lock(&dcache_lock);
+
 	spin_lock(&dcache_inode_lock);
 	res = __d_find_alias(inode, 0);
 	if (res) {
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 		dput(tmp);
 		goto out_iput;
 	}
@@ -1538,7 +1472,6 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	spin_unlock(&tmp->d_lock);
 	spin_unlock(&dcache_inode_lock);
 
-	spin_unlock(&dcache_lock);
 	return tmp;
 
  out_iput:
@@ -1568,21 +1501,18 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 	struct dentry *new = NULL;
 
 	if (inode && S_ISDIR(inode->i_mode)) {
-		spin_lock(&dcache_lock);
 		spin_lock(&dcache_inode_lock);
 		new = __d_find_alias(inode, 1);
 		if (new) {
 			BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
 			spin_unlock(&dcache_inode_lock);
-			spin_unlock(&dcache_lock);
 			security_d_instantiate(new, inode);
 			d_move(new, dentry);
 			iput(inode);
 		} else {
-			/* already taking dcache_lock, so d_add() by hand */
+			/* already taking dcache_inode_lock, so d_add() by hand */
 			__d_instantiate(dentry, inode);
 			spin_unlock(&dcache_inode_lock);
-			spin_unlock(&dcache_lock);
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
 		}
@@ -1655,12 +1585,10 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	 * Negative dentry: instantiate it unless the inode is a directory and
 	 * already has a dentry.
 	 */
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	if (!S_ISDIR(inode->i_mode) || list_empty(&inode->i_dentry)) {
 		__d_instantiate(found, inode);
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 		security_d_instantiate(found, inode);
 		return found;
 	}
@@ -1672,7 +1600,6 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	new = list_entry(inode->i_dentry.next, struct dentry, d_alias);
 	dget_locked(new);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	security_d_instantiate(found, inode);
 	d_move(new, found);
 	iput(inode);
@@ -1843,7 +1770,6 @@ int d_validate(struct dentry *dentry, struct dentry *dparent)
 {
 	struct dentry *child;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dparent->d_lock);
 	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
 		if (dentry == child) {
@@ -1851,12 +1777,10 @@ int d_validate(struct dentry *dentry, struct dentry *dparent)
 			__dget_locked_dlock(dentry);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dparent->d_lock);
-			spin_unlock(&dcache_lock);
 			return 1;
 		}
 	}
 	spin_unlock(&dparent->d_lock);
-	spin_unlock(&dcache_lock);
 
 	return 0;
 }
@@ -1889,7 +1813,6 @@ void d_delete(struct dentry * dentry)
 	/*
 	 * Are we the only user?
 	 */
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	spin_lock(&dentry->d_lock);
 	isdir = S_ISDIR(dentry->d_inode->i_mode);
@@ -1905,7 +1828,6 @@ void d_delete(struct dentry * dentry)
 
 	spin_unlock(&dentry->d_lock);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 
 	fsnotify_nameremove(dentry, isdir);
 }
@@ -1932,13 +1854,11 @@ static void _d_rehash(struct dentry * entry)
  
 void d_rehash(struct dentry * entry)
 {
-	spin_lock(&dcache_lock);
 	spin_lock(&entry->d_lock);
 	spin_lock(&dcache_hash_lock);
 	_d_rehash(entry);
 	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&entry->d_lock);
-	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(d_rehash);
 
@@ -1961,11 +1881,9 @@ void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
 	BUG_ON(!mutex_is_locked(&dentry->d_inode->i_mutex));
 	BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 }
 EXPORT_SYMBOL(dentry_update_name_case);
 
@@ -2058,14 +1976,14 @@ static void dentry_unlock_parents_for_move(struct dentry *dentry,
  * The hash value has to match the hash queue that the dentry is on..
  */
 /*
- * d_move_locked - move a dentry
+ * d_move - move a dentry
  * @dentry: entry to move
  * @target: new dentry
  *
  * Update the dcache to reflect the move of a file name. Negative
  * dcache entries should not be moved in this way.
  */
-static void d_move_locked(struct dentry * dentry, struct dentry * target)
+void d_move(struct dentry * dentry, struct dentry * target)
 {
 	if (!dentry->d_inode)
 		printk(KERN_WARNING "VFS: moving negative dcache entry\n");
@@ -2114,22 +2032,6 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	spin_unlock(&dentry->d_lock);
 	write_sequnlock(&rename_lock);
 }
-
-/**
- * d_move - move a dentry
- * @dentry: entry to move
- * @target: new dentry
- *
- * Update the dcache to reflect the move of a file name. Negative
- * dcache entries should not be moved in this way.
- */
-
-void d_move(struct dentry * dentry, struct dentry * target)
-{
-	spin_lock(&dcache_lock);
-	d_move_locked(dentry, target);
-	spin_unlock(&dcache_lock);
-}
 EXPORT_SYMBOL(d_move);
 
 /**
@@ -2155,13 +2057,12 @@ struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
  * This helper attempts to cope with remotely renamed directories
  *
  * It assumes that the caller is already holding
- * dentry->d_parent->d_inode->i_mutex and the dcache_lock
+ * dentry->d_parent->d_inode->i_mutex and the dcache_inode_lock
  *
  * Note: If ever the locking in lock_rename() changes, then please
  * remember to update this too...
  */
 static struct dentry *__d_unalias(struct dentry *dentry, struct dentry *alias)
-	__releases(dcache_lock)
 	__releases(dcache_inode_lock)
 {
 	struct mutex *m1 = NULL, *m2 = NULL;
@@ -2185,11 +2086,10 @@ static struct dentry *__d_unalias(struct dentry *dentry, struct dentry *alias)
 		goto out_err;
 	m2 = &alias->d_parent->d_inode->i_mutex;
 out_unalias:
-	d_move_locked(alias, dentry);
+	d_move(alias, dentry);
 	ret = alias;
 out_err:
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	if (m2)
 		mutex_unlock(m2);
 	if (m1)
@@ -2249,7 +2149,6 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 
 	BUG_ON(!d_unhashed(dentry));
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 
 	if (!inode) {
@@ -2295,7 +2194,6 @@ found:
 	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&actual->d_lock);
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 out_nolock:
 	if (actual == dentry) {
 		security_d_instantiate(dentry, inode);
@@ -2307,7 +2205,6 @@ out_nolock:
 
 shouldnt_be_hashed:
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 	BUG();
 }
 EXPORT_SYMBOL_GPL(d_materialise_unique);
@@ -2421,11 +2318,9 @@ char *__d_path(const struct path *path, struct path *root,
 	int error;
 
 	prepend(&res, &buflen, "\0", 1);
-	spin_lock(&dcache_lock);
 	write_seqlock(&rename_lock);
 	error = prepend_path(path, root, &res, &buflen);
 	write_sequnlock(&rename_lock);
-	spin_unlock(&dcache_lock);
 
 	if (error)
 		return ERR_PTR(error);
@@ -2487,14 +2382,12 @@ char *d_path(const struct path *path, char *buf, int buflen)
 		return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
 
 	get_fs_root(current->fs, &root);
-	spin_lock(&dcache_lock);
 	write_seqlock(&rename_lock);
 	tmp = root;
 	error = path_with_deleted(path, &tmp, &res, &buflen);
 	if (error)
 		res = ERR_PTR(error);
 	write_sequnlock(&rename_lock);
-	spin_unlock(&dcache_lock);
 	path_put(&root);
 	return res;
 }
@@ -2520,14 +2413,12 @@ char *d_path_with_unreachable(const struct path *path, char *buf, int buflen)
 		return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
 
 	get_fs_root(current->fs, &root);
-	spin_lock(&dcache_lock);
 	write_seqlock(&rename_lock);
 	tmp = root;
 	error = path_with_deleted(path, &tmp, &res, &buflen);
 	if (!error && !path_equal(&tmp, &root))
 		error = prepend_unreachable(&res, &buflen);
 	write_sequnlock(&rename_lock);
-	spin_unlock(&dcache_lock);
 	path_put(&root);
 	if (error)
 		res =  ERR_PTR(error);
@@ -2594,11 +2485,9 @@ char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
 {
 	char *retval;
 
-	spin_lock(&dcache_lock);
 	write_seqlock(&rename_lock);
 	retval = __dentry_path(dentry, buf, buflen);
 	write_sequnlock(&rename_lock);
-	spin_unlock(&dcache_lock);
 
 	return retval;
 }
@@ -2609,7 +2498,6 @@ char *dentry_path(struct dentry *dentry, char *buf, int buflen)
 	char *p = NULL;
 	char *retval;
 
-	spin_lock(&dcache_lock);
 	write_seqlock(&rename_lock);
 	if (d_unlinked(dentry)) {
 		p = buf + buflen;
@@ -2619,12 +2507,10 @@ char *dentry_path(struct dentry *dentry, char *buf, int buflen)
 	}
 	retval = __dentry_path(dentry, buf, buflen);
 	write_sequnlock(&rename_lock);
-	spin_unlock(&dcache_lock);
 	if (!IS_ERR(retval) && p)
 		*p = '/';	/* restore '/' overriden with '\0' */
 	return retval;
 Elong:
-	spin_unlock(&dcache_lock);
 	return ERR_PTR(-ENAMETOOLONG);
 }
 
@@ -2658,7 +2544,6 @@ SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
 	get_fs_root_and_pwd(current->fs, &root, &pwd);
 
 	error = -ENOENT;
-	spin_lock(&dcache_lock);
 	write_seqlock(&rename_lock);
 	if (!d_unlinked(pwd.dentry)) {
 		unsigned long len;
@@ -2669,7 +2554,6 @@ SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
 		prepend(&cwd, &buflen, "\0", 1);
 		error = prepend_path(&pwd, &tmp, &cwd, &buflen);
 		write_sequnlock(&rename_lock);
-		spin_unlock(&dcache_lock);
 
 		if (error)
 			goto out;
@@ -2690,7 +2574,6 @@ SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
 		}
 	} else {
 		write_sequnlock(&rename_lock);
-		spin_unlock(&dcache_lock);
 	}
 
 out:
@@ -2776,7 +2659,6 @@ void d_genocide(struct dentry *root)
 rename_retry:
 	this_parent = root;
 	seq = read_seqbegin(&rename_lock);
-	spin_lock(&dcache_lock);
 	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
@@ -2823,7 +2705,6 @@ resume:
 		if (this_parent != child->d_parent ||
 				read_seqretry(&rename_lock, seq)) {
 			spin_unlock(&this_parent->d_lock);
-			spin_unlock(&dcache_lock);
 			rcu_read_unlock();
 			goto rename_retry;
 		}
@@ -2832,7 +2713,6 @@ resume:
 		goto resume;
 	}
 	spin_unlock(&this_parent->d_lock);
-	spin_unlock(&dcache_lock);
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
 }
diff --git a/fs/exportfs/expfs.c b/fs/exportfs/expfs.c
index 84b8c46..53a5c08 100644
--- a/fs/exportfs/expfs.c
+++ b/fs/exportfs/expfs.c
@@ -47,24 +47,20 @@ find_acceptable_alias(struct dentry *result,
 	if (acceptable(context, result))
 		return result;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &result->d_inode->i_dentry, d_alias) {
 		dget_locked(dentry);
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 		if (toput)
 			dput(toput);
 		if (dentry != result && acceptable(context, dentry)) {
 			dput(result);
 			return dentry;
 		}
-		spin_lock(&dcache_lock);
 		spin_lock(&dcache_inode_lock);
 		toput = dentry;
 	}
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 
 	if (toput)
 		dput(toput);
diff --git a/fs/libfs.c b/fs/libfs.c
index cc47949..28b3666 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -100,7 +100,6 @@ loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
 			struct dentry *cursor = file->private_data;
 			loff_t n = file->f_pos - 2;
 
-			spin_lock(&dcache_lock);
 			spin_lock(&dentry->d_lock);
 			/* d_lock not required for cursor */
 			list_del(&cursor->d_u.d_child);
@@ -116,7 +115,6 @@ loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
 			}
 			list_add_tail(&cursor->d_u.d_child, p);
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 		}
 	}
 	mutex_unlock(&dentry->d_inode->i_mutex);
@@ -159,7 +157,6 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 			i++;
 			/* fallthrough */
 		default:
-			spin_lock(&dcache_lock);
 			spin_lock(&dentry->d_lock);
 			if (filp->f_pos == 2)
 				list_move(q, &dentry->d_subdirs);
@@ -175,13 +172,11 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 
 				spin_unlock(&next->d_lock);
 				spin_unlock(&dentry->d_lock);
-				spin_unlock(&dcache_lock);
 				if (filldir(dirent, next->d_name.name, 
 					    next->d_name.len, filp->f_pos, 
 					    next->d_inode->i_ino, 
 					    dt_type(next->d_inode)) < 0)
 					return 0;
-				spin_lock(&dcache_lock);
 				spin_lock(&dentry->d_lock);
 				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
 				/* next is still alive */
@@ -191,7 +186,6 @@ int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 				filp->f_pos++;
 			}
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 	}
 	return 0;
 }
@@ -285,7 +279,6 @@ int simple_empty(struct dentry *dentry)
 	struct dentry *child;
 	int ret = 0;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
@@ -298,7 +291,6 @@ int simple_empty(struct dentry *dentry)
 	ret = 1;
 out:
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 	return ret;
 }
 
diff --git a/fs/namei.c b/fs/namei.c
index cbfa5fb..5642bc2 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -612,8 +612,8 @@ int follow_up(struct path *path)
 	return 1;
 }
 
-/* no need for dcache_lock, as serialization is taken care in
- * namespace.c
+/*
+ * serialization is taken care of in namespace.c
  */
 static int __follow_mount(struct path *path)
 {
@@ -645,9 +645,6 @@ static void follow_mount(struct path *path)
 	}
 }
 
-/* no need for dcache_lock, as serialization is taken care in
- * namespace.c
- */
 int follow_down(struct path *path)
 {
 	struct vfsmount *mounted;
@@ -2131,12 +2128,10 @@ void dentry_unhash(struct dentry *dentry)
 {
 	dget(dentry);
 	shrink_dcache_parent(dentry);
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	if (dentry->d_count == 2)
 		__d_drop(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 int vfs_rmdir(struct inode *dir, struct dentry *dentry)
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index 102278e..de15c53 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -391,7 +391,6 @@ ncp_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 	}
 
 	/* If a pointer is invalid, we search the dentry. */
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
@@ -402,13 +401,11 @@ ncp_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 			else
 				dent = NULL;
 			spin_unlock(&parent->d_lock);
-			spin_unlock(&dcache_lock);
 			goto out;
 		}
 		next = next->next;
 	}
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 	return NULL;
 
 out:
diff --git a/fs/ncpfs/ncplib_kernel.h b/fs/ncpfs/ncplib_kernel.h
index c4b718f..1220df7 100644
--- a/fs/ncpfs/ncplib_kernel.h
+++ b/fs/ncpfs/ncplib_kernel.h
@@ -193,7 +193,6 @@ ncp_renew_dentries(struct dentry *parent)
 	struct list_head *next;
 	struct dentry *dentry;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
@@ -207,7 +206,6 @@ ncp_renew_dentries(struct dentry *parent)
 		next = next->next;
 	}
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 static inline void
@@ -217,7 +215,6 @@ ncp_invalidate_dircache_entries(struct dentry *parent)
 	struct list_head *next;
 	struct dentry *dentry;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&parent->d_lock);
 	next = parent->d_subdirs.next;
 	while (next != &parent->d_subdirs) {
@@ -227,7 +224,6 @@ ncp_invalidate_dircache_entries(struct dentry *parent)
 		next = next->next;
 	}
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 struct ncp_cache_head {
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 12de824..eb77471 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1718,11 +1718,9 @@ static int nfs_unlink(struct inode *dir, struct dentry *dentry)
 	dfprintk(VFS, "NFS: unlink(%s/%ld, %s)\n", dir->i_sb->s_id,
 		dir->i_ino, dentry->d_name.name);
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	if (dentry->d_count > 1) {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_lock);
 		/* Start asynchronous writeout of the inode */
 		write_inode_now(dentry->d_inode, 0);
 		error = nfs_sillyrename(dir, dentry);
@@ -1733,7 +1731,6 @@ static int nfs_unlink(struct inode *dir, struct dentry *dentry)
 		need_rehash = 1;
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 	error = nfs_safe_remove(dentry);
 	if (!error || error == -ENOENT) {
 		nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c
index 850f67d..b3e36c3 100644
--- a/fs/nfs/getroot.c
+++ b/fs/nfs/getroot.c
@@ -63,13 +63,11 @@ static int nfs_superblock_set_dummy_root(struct super_block *sb, struct inode *i
 		 * This again causes shrink_dcache_for_umount_subtree() to
 		 * Oops, since the test for IS_ROOT() will fail.
 		 */
-		spin_lock(&dcache_lock);
 		spin_lock(&dcache_inode_lock);
 		spin_lock(&sb->s_root->d_lock);
 		list_del_init(&sb->s_root->d_alias);
 		spin_unlock(&sb->s_root->d_lock);
 		spin_unlock(&dcache_inode_lock);
-		spin_unlock(&dcache_lock);
 	}
 	return 0;
 }
diff --git a/fs/nfs/namespace.c b/fs/nfs/namespace.c
index 78c0ebb..74aaf39 100644
--- a/fs/nfs/namespace.c
+++ b/fs/nfs/namespace.c
@@ -60,7 +60,6 @@ rename_retry:
 
 	seq = read_seqbegin(&rename_lock);
 	rcu_read_lock();
-	spin_lock(&dcache_lock);
 	while (!IS_ROOT(dentry) && dentry != droot) {
 		namelen = dentry->d_name.len;
 		buflen -= namelen + 1;
@@ -71,7 +70,6 @@ rename_retry:
 		*--end = '/';
 		dentry = dentry->d_parent;
 	}
-	spin_unlock(&dcache_lock);
 	rcu_read_unlock();
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
@@ -91,7 +89,6 @@ rename_retry:
 	memcpy(end, base, namelen);
 	return end;
 Elong_unlock:
-	spin_unlock(&dcache_lock);
 	rcu_read_unlock();
 	if (read_seqretry(&rename_lock, seq))
 		goto rename_retry;
diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index ae769fc..9be6ec1 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -59,7 +59,6 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 	/* determine if the children should tell inode about their events */
 	watched = fsnotify_inode_watches_children(inode);
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	/* run all of the dentries associated with this inode.  Since this is a
 	 * directory, there damn well better only be one item on this list */
@@ -84,7 +83,6 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 		spin_unlock(&alias->d_lock);
 	}
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 }
 
 /* Notify this dentry's parent about a child's events. */
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index c31b5c6..b7de749 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -169,7 +169,6 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 	struct list_head *p;
 	struct dentry *dentry = NULL;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&dcache_inode_lock);
 	list_for_each(p, &inode->i_dentry) {
 		dentry = list_entry(p, struct dentry, d_alias);
@@ -189,7 +188,6 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 	}
 
 	spin_unlock(&dcache_inode_lock);
-	spin_unlock(&dcache_lock);
 
 	return dentry;
 }
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index c963eba..a2ceb94 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -183,7 +183,6 @@ struct dentry_operations {
 #define DCACHE_GENOCIDE		0x0200
 
 extern spinlock_t dcache_inode_lock;
-extern spinlock_t dcache_lock;
 extern seqlock_t rename_lock;
 
 static inline int dname_external(struct dentry *dentry)
@@ -296,8 +295,8 @@ extern char *dentry_path(struct dentry *, char *, int);
  *	destroyed when it has references. dget() should never be
  *	called for dentries with zero reference counter. For these cases
  *	(preferably none, functions in dcache.c are sufficient for normal
- *	needs and they take necessary precautions) you should hold dcache_lock
- *	and call dget_locked() instead of dget().
+ *	needs and they take necessary precautions) you should hold d_lock
+ *	and call dget_dlock() instead of dget().
  */
 static inline struct dentry *dget_dlock(struct dentry *dentry)
 {
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 090f0ea..296cf2f 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1378,7 +1378,7 @@ struct super_block {
 #else
 	struct list_head	s_files;
 #endif
-	/* s_dentry_lru and s_nr_dentry_unused are protected by dcache_lock */
+	/* s_dentry_lru, s_nr_dentry_unused protected by dcache.c lru locks */
 	struct list_head	s_dentry_lru;	/* unused dentry lru */
 	int			s_nr_dentry_unused;	/* # of dentry on lru */
 
@@ -2446,6 +2446,10 @@ static inline ino_t parent_ino(struct dentry *dentry)
 {
 	ino_t res;
 
+	/*
+	 * Don't strictly need d_lock here? If the parent ino could change
+	 * then surely we'd have a deeper race in the caller?
+	 */
 	spin_lock(&dentry->d_lock);
 	res = dentry->d_parent->d_inode->i_ino;
 	spin_unlock(&dentry->d_lock);
diff --git a/include/linux/fsnotify.h b/include/linux/fsnotify.h
index b10bcde..2a53f10 100644
--- a/include/linux/fsnotify.h
+++ b/include/linux/fsnotify.h
@@ -17,7 +17,6 @@
 
 /*
  * fsnotify_d_instantiate - instantiate a dentry for inode
- * Called with dcache_lock held.
  */
 static inline void fsnotify_d_instantiate(struct dentry *dentry,
 					  struct inode *inode)
@@ -62,7 +61,6 @@ static inline int fsnotify_perm(struct file *file, int mask)
 
 /*
  * fsnotify_d_move - dentry has been moved
- * Called with dcache_lock and dentry->d_lock held.
  */
 static inline void fsnotify_d_move(struct dentry *dentry)
 {
diff --git a/include/linux/fsnotify_backend.h b/include/linux/fsnotify_backend.h
index 7380763..69ad89b 100644
--- a/include/linux/fsnotify_backend.h
+++ b/include/linux/fsnotify_backend.h
@@ -329,9 +329,15 @@ static inline void __fsnotify_update_dcache_flags(struct dentry *dentry)
 {
 	struct dentry *parent;
 
-	assert_spin_locked(&dcache_lock);
 	assert_spin_locked(&dentry->d_lock);
 
+	/*
+	 * Serialisation of setting PARENT_WATCHED on the dentries is provided
+	 * by d_lock. If inotify_inode_watched changes after we have taken
+	 * d_lock, the following __fsnotify_update_child_dentry_flags call will
+	 * find our entry, so it will spin until we complete here, and update
+	 * us with the new state.
+	 */
 	parent = dentry->d_parent;
 	if (parent->d_inode && fsnotify_inode_watches_children(parent->d_inode))
 		dentry->d_flags |= DCACHE_FSNOTIFY_PARENT_WATCHED;
@@ -341,15 +347,12 @@ static inline void __fsnotify_update_dcache_flags(struct dentry *dentry)
 
 /*
  * fsnotify_d_instantiate - instantiate a dentry for inode
- * Called with dcache_lock held.
  */
 static inline void __fsnotify_d_instantiate(struct dentry *dentry, struct inode *inode)
 {
 	if (!inode)
 		return;
 
-	assert_spin_locked(&dcache_lock);
-
 	spin_lock(&dentry->d_lock);
 	__fsnotify_update_dcache_flags(dentry);
 	spin_unlock(&dentry->d_lock);
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 05b441d..aec730b 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -41,7 +41,6 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
  *  - require a directory
  *  - ending slashes ok even for nonexistent files
  *  - internal "there are more path components" flag
- *  - locked when lookup done with dcache_lock held
  *  - dentry cache is untrusted; force a real lookup
  */
 #define LOOKUP_FOLLOW		 1
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 7b4705b..1864cb6 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -876,7 +876,6 @@ static void cgroup_clear_directory(struct dentry *dentry)
 	struct list_head *node;
 
 	BUG_ON(!mutex_is_locked(&dentry->d_inode->i_mutex));
-	spin_lock(&dcache_lock);
 	spin_lock(&dentry->d_lock);
 	node = dentry->d_subdirs.next;
 	while (node != &dentry->d_subdirs) {
@@ -891,18 +890,15 @@ static void cgroup_clear_directory(struct dentry *dentry)
 			dget_locked_dlock(d);
 			spin_unlock(&d->d_lock);
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_lock);
 			d_delete(d);
 			simple_unlink(dentry->d_inode, d);
 			dput(d);
-			spin_lock(&dcache_lock);
 			spin_lock(&dentry->d_lock);
 		} else
 			spin_unlock(&d->d_lock);
 		node = dentry->d_subdirs.next;
 	}
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 /*
@@ -914,14 +910,12 @@ static void cgroup_d_remove_dir(struct dentry *dentry)
 
 	cgroup_clear_directory(dentry);
 
-	spin_lock(&dcache_lock);
 	parent = dentry->d_parent;
 	spin_lock(&parent->d_lock);
 	spin_lock(&dentry->d_lock);
 	list_del_init(&dentry->d_u.d_child);
 	spin_unlock(&dentry->d_lock);
 	spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_lock);
 	remove_dir(dentry);
 }
 
diff --git a/mm/filemap.c b/mm/filemap.c
index 6b9aee2..ca38939 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -102,9 +102,6 @@
  *    ->inode_lock		(zap_pte_range->set_page_dirty)
  *    ->private_lock		(zap_pte_range->__set_page_dirty_buffers)
  *
- *  ->task->proc_lock
- *    ->dcache_lock		(proc_pid_lookup)
- *
  *  (code doesn't rely on that order, so you could switch it around)
  *  ->tasklist_lock             (memory_failure, collect_procs_ao)
  *    ->i_mmap_lock
diff --git a/security/selinux/selinuxfs.c b/security/selinux/selinuxfs.c
index 017ec09..2285d69 100644
--- a/security/selinux/selinuxfs.c
+++ b/security/selinux/selinuxfs.c
@@ -1145,7 +1145,6 @@ static void sel_remove_entries(struct dentry *de)
 {
 	struct list_head *node;
 
-	spin_lock(&dcache_lock);
 	spin_lock(&de->d_lock);
 	node = de->d_subdirs.next;
 	while (node != &de->d_subdirs) {
@@ -1158,11 +1157,9 @@ static void sel_remove_entries(struct dentry *de)
 			dget_locked_dlock(d);
 			spin_unlock(&de->d_lock);
 			spin_unlock(&d->d_lock);
-			spin_unlock(&dcache_lock);
 			d_delete(d);
 			simple_unlink(de->d_inode, d);
 			dput(d);
-			spin_lock(&dcache_lock);
 			spin_lock(&de->d_lock);
 		} else
 			spin_unlock(&d->d_lock);
@@ -1170,7 +1167,6 @@ static void sel_remove_entries(struct dentry *de)
 	}
 
 	spin_unlock(&de->d_lock);
-	spin_unlock(&dcache_lock);
 }
 
 #define BOOL_DIR_NAME "booleans"
-- 
cgit v0.10.2


From 58db63d086790eec2ed433f9d8c4962239809cf8 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:39 +1100
Subject: fs: dcache avoid starvation in dcache multi-step operations

Long lived dcache "multi-step" operations which retry on rename seq can
be starved with a lot of rename activity. If they fail after the 1st pass,
take the rename_lock for writing to avoid further starvation.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 0dbae053b..bf6294a 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -973,10 +973,11 @@ int have_submounts(struct dentry *parent)
 	struct dentry *this_parent;
 	struct list_head *next;
 	unsigned seq;
+	int locked = 0;
 
-rename_retry:
-	this_parent = parent;
 	seq = read_seqbegin(&rename_lock);
+again:
+	this_parent = parent;
 
 	if (d_mountpoint(parent))
 		goto positive;
@@ -1021,7 +1022,7 @@ resume:
 		/* might go back up the wrong parent if we have had a rename
 		 * or deletion */
 		if (this_parent != child->d_parent ||
-				read_seqretry(&rename_lock, seq)) {
+			 (!locked && read_seqretry(&rename_lock, seq))) {
 			spin_unlock(&this_parent->d_lock);
 			rcu_read_unlock();
 			goto rename_retry;
@@ -1031,13 +1032,22 @@ resume:
 		goto resume;
 	}
 	spin_unlock(&this_parent->d_lock);
-	if (read_seqretry(&rename_lock, seq))
+	if (!locked && read_seqretry(&rename_lock, seq))
 		goto rename_retry;
+	if (locked)
+		write_sequnlock(&rename_lock);
 	return 0; /* No mount points found in tree */
 positive:
-	if (read_seqretry(&rename_lock, seq))
+	if (!locked && read_seqretry(&rename_lock, seq))
 		goto rename_retry;
+	if (locked)
+		write_sequnlock(&rename_lock);
 	return 1;
+
+rename_retry:
+	locked = 1;
+	write_seqlock(&rename_lock);
+	goto again;
 }
 EXPORT_SYMBOL(have_submounts);
 
@@ -1061,11 +1071,11 @@ static int select_parent(struct dentry * parent)
 	struct list_head *next;
 	unsigned seq;
 	int found = 0;
+	int locked = 0;
 
-rename_retry:
-	this_parent = parent;
 	seq = read_seqbegin(&rename_lock);
-
+again:
+	this_parent = parent;
 	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
@@ -1127,7 +1137,7 @@ resume:
 		/* might go back up the wrong parent if we have had a rename
 		 * or deletion */
 		if (this_parent != child->d_parent ||
-				read_seqretry(&rename_lock, seq)) {
+			(!locked && read_seqretry(&rename_lock, seq))) {
 			spin_unlock(&this_parent->d_lock);
 			rcu_read_unlock();
 			goto rename_retry;
@@ -1138,9 +1148,18 @@ resume:
 	}
 out:
 	spin_unlock(&this_parent->d_lock);
-	if (read_seqretry(&rename_lock, seq))
+	if (!locked && read_seqretry(&rename_lock, seq))
 		goto rename_retry;
+	if (locked)
+		write_sequnlock(&rename_lock);
 	return found;
+
+rename_retry:
+	if (found)
+		return found;
+	locked = 1;
+	write_seqlock(&rename_lock);
+	goto again;
 }
 
 /**
@@ -2655,10 +2674,11 @@ void d_genocide(struct dentry *root)
 	struct dentry *this_parent;
 	struct list_head *next;
 	unsigned seq;
+	int locked = 0;
 
-rename_retry:
-	this_parent = root;
 	seq = read_seqbegin(&rename_lock);
+again:
+	this_parent = root;
 	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
@@ -2703,7 +2723,7 @@ resume:
 		/* might go back up the wrong parent if we have had a rename
 		 * or deletion */
 		if (this_parent != child->d_parent ||
-				read_seqretry(&rename_lock, seq)) {
+			 (!locked && read_seqretry(&rename_lock, seq))) {
 			spin_unlock(&this_parent->d_lock);
 			rcu_read_unlock();
 			goto rename_retry;
@@ -2713,8 +2733,16 @@ resume:
 		goto resume;
 	}
 	spin_unlock(&this_parent->d_lock);
-	if (read_seqretry(&rename_lock, seq))
+	if (!locked && read_seqretry(&rename_lock, seq))
 		goto rename_retry;
+	if (locked)
+		write_sequnlock(&rename_lock);
+	return;
+
+rename_retry:
+	locked = 1;
+	write_seqlock(&rename_lock);
+	goto again;
 }
 
 /**
-- 
cgit v0.10.2


From 61f3dee4af09528997a970280da240577bf60721 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:40 +1100
Subject: fs: dcache reduce dput locking

It is possible to run dput without taking data structure locks up-front. In
many cases where we don't kill the dentry anyway, these locks are not required.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index bf6294a..98a0569 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -323,35 +323,16 @@ repeat:
 	if (dentry->d_count == 1)
 		might_sleep();
 	spin_lock(&dentry->d_lock);
-	if (IS_ROOT(dentry))
-		parent = NULL;
-	else
-		parent = dentry->d_parent;
-	if (dentry->d_count == 1) {
-		if (!spin_trylock(&dcache_inode_lock)) {
-drop2:
-			spin_unlock(&dentry->d_lock);
-			goto repeat;
-		}
-		if (parent && !spin_trylock(&parent->d_lock)) {
-			spin_unlock(&dcache_inode_lock);
-			goto drop2;
-		}
-	}
-	dentry->d_count--;
-	if (dentry->d_count) {
+	BUG_ON(!dentry->d_count);
+	if (dentry->d_count > 1) {
+		dentry->d_count--;
 		spin_unlock(&dentry->d_lock);
-		if (parent)
-			spin_unlock(&parent->d_lock);
 		return;
 	}
 
-	/*
-	 * AV: ->d_delete() is _NOT_ allowed to block now.
-	 */
 	if (dentry->d_op && dentry->d_op->d_delete) {
 		if (dentry->d_op->d_delete(dentry))
-			goto unhash_it;
+			goto kill_it;
 	}
 
 	/* Unreachable? Get rid of it */
@@ -362,17 +343,30 @@ drop2:
 	dentry->d_flags |= DCACHE_REFERENCED;
 	dentry_lru_add(dentry);
 
- 	spin_unlock(&dentry->d_lock);
-	if (parent)
-		spin_unlock(&parent->d_lock);
-	spin_unlock(&dcache_inode_lock);
+	dentry->d_count--;
+	spin_unlock(&dentry->d_lock);
 	return;
 
-unhash_it:
-	__d_drop(dentry);
 kill_it:
+	if (!spin_trylock(&dcache_inode_lock)) {
+relock:
+		spin_unlock(&dentry->d_lock);
+		cpu_relax();
+		goto repeat;
+	}
+	if (IS_ROOT(dentry))
+		parent = NULL;
+	else
+		parent = dentry->d_parent;
+	if (parent && !spin_trylock(&parent->d_lock)) {
+		spin_unlock(&dcache_inode_lock);
+		goto relock;
+	}
+	dentry->d_count--;
 	/* if dentry was on the d_lru list delete it from there */
 	dentry_lru_del(dentry);
+	/* if it was on the hash (d_delete case), then remove it */
+	__d_drop(dentry);
 	dentry = d_kill(dentry, parent);
 	if (dentry)
 		goto repeat;
-- 
cgit v0.10.2


From 89ad485f01fd83c47f17a128db3bd7b89c0f244f Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:41 +1100
Subject: fs: dcache reduce locking in d_alloc

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 98a0569..ccdc5c2 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1253,11 +1253,13 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 
 	if (parent) {
 		spin_lock(&parent->d_lock);
-		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+		/*
+		 * don't need child lock because it is not subject
+		 * to concurrency here
+		 */
 		dentry->d_parent = dget_dlock(parent);
 		dentry->d_sb = parent->d_sb;
 		list_add(&dentry->d_u.d_child, &parent->d_subdirs);
-		spin_unlock(&dentry->d_lock);
 		spin_unlock(&parent->d_lock);
 	}
 
-- 
cgit v0.10.2


From 357f8e658bba8a085c4a5d4331e30894be8096b8 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:42 +1100
Subject: fs: dcache reduce dcache_inode_lock

dcache_inode_lock can be avoided in d_delete() and d_materialise_unique()
in cases where it is not required.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index ccdc5c2..01f0167 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1828,10 +1828,15 @@ void d_delete(struct dentry * dentry)
 	/*
 	 * Are we the only user?
 	 */
-	spin_lock(&dcache_inode_lock);
+again:
 	spin_lock(&dentry->d_lock);
 	isdir = S_ISDIR(dentry->d_inode->i_mode);
 	if (dentry->d_count == 1) {
+		if (!spin_trylock(&dcache_inode_lock)) {
+			spin_unlock(&dentry->d_lock);
+			cpu_relax();
+			goto again;
+		}
 		dentry->d_flags &= ~DCACHE_CANT_MOUNT;
 		dentry_iput(dentry);
 		fsnotify_nameremove(dentry, isdir);
@@ -1842,7 +1847,6 @@ void d_delete(struct dentry * dentry)
 		__d_drop(dentry);
 
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_inode_lock);
 
 	fsnotify_nameremove(dentry, isdir);
 }
@@ -2164,14 +2168,15 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 
 	BUG_ON(!d_unhashed(dentry));
 
-	spin_lock(&dcache_inode_lock);
-
 	if (!inode) {
 		actual = dentry;
 		__d_instantiate(dentry, NULL);
-		goto found_lock;
+		d_rehash(actual);
+		goto out_nolock;
 	}
 
+	spin_lock(&dcache_inode_lock);
+
 	if (S_ISDIR(inode->i_mode)) {
 		struct dentry *alias;
 
@@ -2198,10 +2203,9 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 	actual = __d_instantiate_unique(dentry, inode);
 	if (!actual)
 		actual = dentry;
-	else if (unlikely(!d_unhashed(actual)))
-		goto shouldnt_be_hashed;
+	else
+		BUG_ON(!d_unhashed(actual));
 
-found_lock:
 	spin_lock(&actual->d_lock);
 found:
 	spin_lock(&dcache_hash_lock);
@@ -2217,10 +2221,6 @@ out_nolock:
 
 	iput(inode);
 	return actual;
-
-shouldnt_be_hashed:
-	spin_unlock(&dcache_inode_lock);
-	BUG();
 }
 EXPORT_SYMBOL_GPL(d_materialise_unique);
 
-- 
cgit v0.10.2


From dc0474be3e27463d4d4a2793f82366eed906f223 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:43 +1100
Subject: fs: dcache rationalise dget variants

dget_locked was a shortcut to avoid the lazy lru manipulation when we already
held dcache_lock (lru manipulation was relatively cheap at that point).
However, how that the lru lock is an innermost one, we never hold it at any
caller, so the lock cost can now be avoided. We already have well working lazy
dcache LRU, so it should be fine to defer LRU manipulations to scan time.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/arch/powerpc/platforms/cell/spufs/inode.c b/arch/powerpc/platforms/cell/spufs/inode.c
index 2662b50..03185de 100644
--- a/arch/powerpc/platforms/cell/spufs/inode.c
+++ b/arch/powerpc/platforms/cell/spufs/inode.c
@@ -161,7 +161,7 @@ static void spufs_prune_dir(struct dentry *dir)
 	list_for_each_entry_safe(dentry, tmp, &dir->d_subdirs, d_u.d_child) {
 		spin_lock(&dentry->d_lock);
 		if (!(d_unhashed(dentry)) && dentry->d_inode) {
-			dget_locked_dlock(dentry);
+			dget_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			simple_unlink(dir->d_inode, dentry);
diff --git a/drivers/infiniband/hw/ipath/ipath_fs.c b/drivers/infiniband/hw/ipath/ipath_fs.c
index 925e882..31ae1b1 100644
--- a/drivers/infiniband/hw/ipath/ipath_fs.c
+++ b/drivers/infiniband/hw/ipath/ipath_fs.c
@@ -279,7 +279,7 @@ static int remove_file(struct dentry *parent, char *name)
 
 	spin_lock(&tmp->d_lock);
 	if (!(d_unhashed(tmp) && tmp->d_inode)) {
-		dget_locked_dlock(tmp);
+		dget_dlock(tmp);
 		__d_drop(tmp);
 		spin_unlock(&tmp->d_lock);
 		simple_unlink(parent->d_inode, tmp);
diff --git a/drivers/infiniband/hw/qib/qib_fs.c b/drivers/infiniband/hw/qib/qib_fs.c
index 49af4a6..df7fa25 100644
--- a/drivers/infiniband/hw/qib/qib_fs.c
+++ b/drivers/infiniband/hw/qib/qib_fs.c
@@ -455,7 +455,7 @@ static int remove_file(struct dentry *parent, char *name)
 
 	spin_lock(&tmp->d_lock);
 	if (!(d_unhashed(tmp) && tmp->d_inode)) {
-		dget_locked_dlock(tmp);
+		dget_dlock(tmp);
 		__d_drop(tmp);
 		spin_unlock(&tmp->d_lock);
 		simple_unlink(parent->d_inode, tmp);
diff --git a/drivers/staging/smbfs/cache.c b/drivers/staging/smbfs/cache.c
index 75dfd40..f2a1323 100644
--- a/drivers/staging/smbfs/cache.c
+++ b/drivers/staging/smbfs/cache.c
@@ -102,7 +102,7 @@ smb_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 		dent = list_entry(next, struct dentry, d_u.d_child);
 		if ((unsigned long)dent->d_fsdata == fpos) {
 			if (dent->d_inode)
-				dget_locked(dent);
+				dget(dent);
 			else
 				dent = NULL;
 			goto out_unlock;
diff --git a/fs/configfs/inode.c b/fs/configfs/inode.c
index fb3a55f..c83f476 100644
--- a/fs/configfs/inode.c
+++ b/fs/configfs/inode.c
@@ -252,7 +252,7 @@ void configfs_drop_dentry(struct configfs_dirent * sd, struct dentry * parent)
 	if (dentry) {
 		spin_lock(&dentry->d_lock);
 		if (!(d_unhashed(dentry) && dentry->d_inode)) {
-			dget_locked_dlock(dentry);
+			dget_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			simple_unlink(parent->d_inode, dentry);
diff --git a/fs/dcache.c b/fs/dcache.c
index 01f0167..b4d2e28 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -429,32 +429,17 @@ int d_invalidate(struct dentry * dentry)
 EXPORT_SYMBOL(d_invalidate);
 
 /* This must be called with d_lock held */
-static inline struct dentry * __dget_locked_dlock(struct dentry *dentry)
+static inline void __dget_dlock(struct dentry *dentry)
 {
 	dentry->d_count++;
-	dentry_lru_del(dentry);
-	return dentry;
 }
 
-/* This must be called with d_lock held */
-static inline struct dentry * __dget_locked(struct dentry *dentry)
+static inline void __dget(struct dentry *dentry)
 {
 	spin_lock(&dentry->d_lock);
-	__dget_locked_dlock(dentry);
+	__dget_dlock(dentry);
 	spin_unlock(&dentry->d_lock);
-	return dentry;
-}
-
-struct dentry * dget_locked_dlock(struct dentry *dentry)
-{
-	return __dget_locked_dlock(dentry);
-}
-
-struct dentry * dget_locked(struct dentry *dentry)
-{
-	return __dget_locked(dentry);
 }
-EXPORT_SYMBOL(dget_locked);
 
 struct dentry *dget_parent(struct dentry *dentry)
 {
@@ -512,7 +497,7 @@ again:
 			    (alias->d_flags & DCACHE_DISCONNECTED)) {
 				discon_alias = alias;
 			} else if (!want_discon) {
-				__dget_locked_dlock(alias);
+				__dget_dlock(alias);
 				spin_unlock(&alias->d_lock);
 				return alias;
 			}
@@ -525,7 +510,7 @@ again:
 		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
 			if (IS_ROOT(alias) &&
 			    (alias->d_flags & DCACHE_DISCONNECTED)) {
-				__dget_locked_dlock(alias);
+				__dget_dlock(alias);
 				spin_unlock(&alias->d_lock);
 				return alias;
 			}
@@ -561,7 +546,7 @@ restart:
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		spin_lock(&dentry->d_lock);
 		if (!dentry->d_count) {
-			__dget_locked_dlock(dentry);
+			__dget_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_inode_lock);
@@ -1257,7 +1242,8 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 		 * don't need child lock because it is not subject
 		 * to concurrency here
 		 */
-		dentry->d_parent = dget_dlock(parent);
+		__dget_dlock(parent);
+		dentry->d_parent = parent;
 		dentry->d_sb = parent->d_sb;
 		list_add(&dentry->d_u.d_child, &parent->d_subdirs);
 		spin_unlock(&parent->d_lock);
@@ -1360,7 +1346,7 @@ static struct dentry *__d_instantiate_unique(struct dentry *entry,
 			continue;
 		if (memcmp(qstr->name, name, len))
 			continue;
-		dget_locked(alias);
+		__dget(alias);
 		return alias;
 	}
 
@@ -1613,7 +1599,7 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	 * reference to it, move it in place and use it.
 	 */
 	new = list_entry(inode->i_dentry.next, struct dentry, d_alias);
-	dget_locked(new);
+	__dget(new);
 	spin_unlock(&dcache_inode_lock);
 	security_d_instantiate(found, inode);
 	d_move(new, found);
@@ -1789,7 +1775,7 @@ int d_validate(struct dentry *dentry, struct dentry *dparent)
 	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
 		if (dentry == child) {
 			spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
-			__dget_locked_dlock(dentry);
+			__dget_dlock(dentry);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dparent->d_lock);
 			return 1;
diff --git a/fs/exportfs/expfs.c b/fs/exportfs/expfs.c
index 53a5c08..f06a940 100644
--- a/fs/exportfs/expfs.c
+++ b/fs/exportfs/expfs.c
@@ -49,7 +49,7 @@ find_acceptable_alias(struct dentry *result,
 
 	spin_lock(&dcache_inode_lock);
 	list_for_each_entry(dentry, &result->d_inode->i_dentry, d_alias) {
-		dget_locked(dentry);
+		dget(dentry);
 		spin_unlock(&dcache_inode_lock);
 		if (toput)
 			dput(toput);
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index de15c53..0ba3cdc 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -397,7 +397,7 @@ ncp_dget_fpos(struct dentry *dentry, struct dentry *parent, unsigned long fpos)
 		dent = list_entry(next, struct dentry, d_u.d_child);
 		if ((unsigned long)dent->d_fsdata == fpos) {
 			if (dent->d_inode)
-				dget_locked(dent);
+				dget(dent);
 			else
 				dent = NULL;
 			spin_unlock(&parent->d_lock);
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index b7de749..4d54c60 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -178,7 +178,7 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 			mlog(0, "dentry found: %.*s\n",
 			     dentry->d_name.len, dentry->d_name.name);
 
-			dget_locked_dlock(dentry);
+			dget_dlock(dentry);
 			spin_unlock(&dentry->d_lock);
 			break;
 		}
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index a2ceb94..ca64868 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -287,23 +287,17 @@ extern char *dentry_path(struct dentry *, char *, int);
 /* Allocation counts.. */
 
 /**
- *	dget, dget_locked	-	get a reference to a dentry
+ *	dget, dget_dlock -	get a reference to a dentry
  *	@dentry: dentry to get a reference to
  *
  *	Given a dentry or %NULL pointer increment the reference count
  *	if appropriate and return the dentry. A dentry will not be 
- *	destroyed when it has references. dget() should never be
- *	called for dentries with zero reference counter. For these cases
- *	(preferably none, functions in dcache.c are sufficient for normal
- *	needs and they take necessary precautions) you should hold d_lock
- *	and call dget_dlock() instead of dget().
+ *	destroyed when it has references.
  */
 static inline struct dentry *dget_dlock(struct dentry *dentry)
 {
-	if (dentry) {
-		BUG_ON(!dentry->d_count);
+	if (dentry)
 		dentry->d_count++;
-	}
 	return dentry;
 }
 
@@ -317,9 +311,6 @@ static inline struct dentry *dget(struct dentry *dentry)
 	return dentry;
 }
 
-extern struct dentry * dget_locked(struct dentry *);
-extern struct dentry * dget_locked_dlock(struct dentry *);
-
 extern struct dentry *dget_parent(struct dentry *dentry);
 
 /**
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 1864cb6..9f41470 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -887,7 +887,7 @@ static void cgroup_clear_directory(struct dentry *dentry)
 			/* This should never be called on a cgroup
 			 * directory with child cgroups */
 			BUG_ON(d->d_inode->i_mode & S_IFDIR);
-			dget_locked_dlock(d);
+			dget_dlock(d);
 			spin_unlock(&d->d_lock);
 			spin_unlock(&dentry->d_lock);
 			d_delete(d);
diff --git a/security/selinux/selinuxfs.c b/security/selinux/selinuxfs.c
index 2285d69..43deac2 100644
--- a/security/selinux/selinuxfs.c
+++ b/security/selinux/selinuxfs.c
@@ -1154,7 +1154,7 @@ static void sel_remove_entries(struct dentry *de)
 		list_del_init(node);
 
 		if (d->d_inode) {
-			dget_locked_dlock(d);
+			dget_dlock(d);
 			spin_unlock(&de->d_lock);
 			spin_unlock(&d->d_lock);
 			d_delete(d);
-- 
cgit v0.10.2


From a734eb458ab2bd11479a27dd54f48e1b26a55845 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:44 +1100
Subject: fs: dcache reduce d_parent locking

Use RCU to simplify locking in dget_parent.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index b4d2e28..a8f8976 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -446,24 +446,27 @@ struct dentry *dget_parent(struct dentry *dentry)
 	struct dentry *ret;
 
 repeat:
-	spin_lock(&dentry->d_lock);
+	/*
+	 * Don't need rcu_dereference because we re-check it was correct under
+	 * the lock.
+	 */
+	rcu_read_lock();
 	ret = dentry->d_parent;
-	if (!ret)
-		goto out;
-	if (dentry == ret) {
-		ret->d_count++;
+	if (!ret) {
+		rcu_read_unlock();
 		goto out;
 	}
-	if (!spin_trylock(&ret->d_lock)) {
-		spin_unlock(&dentry->d_lock);
-		cpu_relax();
+	spin_lock(&ret->d_lock);
+	if (unlikely(ret != dentry->d_parent)) {
+		spin_unlock(&ret->d_lock);
+		rcu_read_unlock();
 		goto repeat;
 	}
+	rcu_read_unlock();
 	BUG_ON(!ret->d_count);
 	ret->d_count++;
 	spin_unlock(&ret->d_lock);
 out:
-	spin_unlock(&dentry->d_lock);
 	return ret;
 }
 EXPORT_SYMBOL(dget_parent);
-- 
cgit v0.10.2


From 89e6054836a7b1e7500cd70a14b5579e752c9250 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:45 +1100
Subject: fs: dcache reduce prune_one_dentry locking

prune_one_dentry can avoid quite a bit of locking in the common case where
ancestors have an elevated refcount. Alternatively, we could have gone the
other way and made fewer trylocks in the case where d_count goes to zero, but
is probably less common.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index a8f8976..1957063 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -582,26 +582,29 @@ static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
 	 * Prune ancestors.
 	 */
 	while (dentry) {
-		spin_lock(&dcache_inode_lock);
-again:
+relock:
 		spin_lock(&dentry->d_lock);
+		if (dentry->d_count > 1) {
+			dentry->d_count--;
+			spin_unlock(&dentry->d_lock);
+			return;
+		}
+		if (!spin_trylock(&dcache_inode_lock)) {
+relock2:
+			spin_unlock(&dentry->d_lock);
+			cpu_relax();
+			goto relock;
+		}
+
 		if (IS_ROOT(dentry))
 			parent = NULL;
 		else
 			parent = dentry->d_parent;
 		if (parent && !spin_trylock(&parent->d_lock)) {
-			spin_unlock(&dentry->d_lock);
-			goto again;
-		}
-		dentry->d_count--;
-		if (dentry->d_count) {
-			if (parent)
-				spin_unlock(&parent->d_lock);
-			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_inode_lock);
-			return;
+			goto relock2;
 		}
-
+		dentry->d_count--;
 		dentry_lru_del(dentry);
 		__d_drop(dentry);
 		dentry = d_kill(dentry, parent);
-- 
cgit v0.10.2


From be182bff72fae6a3eb25624b39170c40b72f0909 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:46 +1100
Subject: fs: reduce dcache_inode_lock width in lru scanning

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 1957063..fe8f28a 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -621,7 +621,7 @@ static void shrink_dentry_list(struct list_head *list)
 		dentry = list_entry(list->prev, struct dentry, d_lru);
 
 		if (!spin_trylock(&dentry->d_lock)) {
-relock:
+relock1:
 			spin_unlock(&dcache_lru_lock);
 			cpu_relax();
 			spin_lock(&dcache_lru_lock);
@@ -638,20 +638,24 @@ relock:
 			spin_unlock(&dentry->d_lock);
 			continue;
 		}
+		if (!spin_trylock(&dcache_inode_lock)) {
+relock2:
+			spin_unlock(&dentry->d_lock);
+			goto relock1;
+		}
 		if (IS_ROOT(dentry))
 			parent = NULL;
 		else
 			parent = dentry->d_parent;
 		if (parent && !spin_trylock(&parent->d_lock)) {
-			spin_unlock(&dentry->d_lock);
-			goto relock;
+			spin_unlock(&dcache_inode_lock);
+			goto relock2;
 		}
 		__dentry_lru_del(dentry);
 		spin_unlock(&dcache_lru_lock);
 
 		prune_one_dentry(dentry, parent);
 		/* dcache_inode_lock and dentry->d_lock dropped */
-		spin_lock(&dcache_inode_lock);
 		spin_lock(&dcache_lru_lock);
 	}
 }
@@ -672,7 +676,6 @@ static void __shrink_dcache_sb(struct super_block *sb, int *count, int flags)
 	LIST_HEAD(tmp);
 	int cnt = *count;
 
-	spin_lock(&dcache_inode_lock);
 relock:
 	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
@@ -711,7 +714,6 @@ relock:
 	if (!list_empty(&referenced))
 		list_splice(&referenced, &sb->s_dentry_lru);
 	spin_unlock(&dcache_lru_lock);
-	spin_unlock(&dcache_inode_lock);
 }
 
 /**
@@ -800,14 +802,12 @@ void shrink_dcache_sb(struct super_block *sb)
 {
 	LIST_HEAD(tmp);
 
-	spin_lock(&dcache_inode_lock);
 	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
 		list_splice_init(&sb->s_dentry_lru, &tmp);
 		shrink_dentry_list(&tmp);
 	}
 	spin_unlock(&dcache_lru_lock);
-	spin_unlock(&dcache_inode_lock);
 }
 EXPORT_SYMBOL(shrink_dcache_sb);
 
-- 
cgit v0.10.2


From ec33679d78f9d653a44ddba10b5fb824c06330a1 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:47 +1100
Subject: fs: use RCU in shrink_dentry_list to reduce lock nesting

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index fe8f28a..d1840b3 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -615,16 +615,16 @@ static void shrink_dentry_list(struct list_head *list)
 {
 	struct dentry *dentry;
 
-	while (!list_empty(list)) {
+	rcu_read_lock();
+	for (;;) {
 		struct dentry *parent;
 
-		dentry = list_entry(list->prev, struct dentry, d_lru);
-
-		if (!spin_trylock(&dentry->d_lock)) {
-relock1:
-			spin_unlock(&dcache_lru_lock);
-			cpu_relax();
-			spin_lock(&dcache_lru_lock);
+		dentry = list_entry_rcu(list->prev, struct dentry, d_lru);
+		if (&dentry->d_lru == list)
+			break; /* empty */
+		spin_lock(&dentry->d_lock);
+		if (dentry != list_entry(list->prev, struct dentry, d_lru)) {
+			spin_unlock(&dentry->d_lock);
 			continue;
 		}
 
@@ -634,14 +634,16 @@ relock1:
 		 * it - just keep it off the LRU list.
 		 */
 		if (dentry->d_count) {
-			__dentry_lru_del(dentry);
+			dentry_lru_del(dentry);
 			spin_unlock(&dentry->d_lock);
 			continue;
 		}
+
 		if (!spin_trylock(&dcache_inode_lock)) {
-relock2:
+relock:
 			spin_unlock(&dentry->d_lock);
-			goto relock1;
+			cpu_relax();
+			continue;
 		}
 		if (IS_ROOT(dentry))
 			parent = NULL;
@@ -649,15 +651,15 @@ relock2:
 			parent = dentry->d_parent;
 		if (parent && !spin_trylock(&parent->d_lock)) {
 			spin_unlock(&dcache_inode_lock);
-			goto relock2;
+			goto relock;
 		}
-		__dentry_lru_del(dentry);
-		spin_unlock(&dcache_lru_lock);
+		dentry_lru_del(dentry);
 
+		rcu_read_unlock();
 		prune_one_dentry(dentry, parent);
-		/* dcache_inode_lock and dentry->d_lock dropped */
-		spin_lock(&dcache_lru_lock);
+		rcu_read_lock();
 	}
+	rcu_read_unlock();
 }
 
 /**
@@ -705,15 +707,15 @@ relock:
 			if (!--cnt)
 				break;
 		}
-		/* XXX: re-add cond_resched_lock when dcache_lock goes away */
+		cond_resched_lock(&dcache_lru_lock);
 	}
-
-	*count = cnt;
-	shrink_dentry_list(&tmp);
-
 	if (!list_empty(&referenced))
 		list_splice(&referenced, &sb->s_dentry_lru);
 	spin_unlock(&dcache_lru_lock);
+
+	shrink_dentry_list(&tmp);
+
+	*count = cnt;
 }
 
 /**
@@ -805,7 +807,9 @@ void shrink_dcache_sb(struct super_block *sb)
 	spin_lock(&dcache_lru_lock);
 	while (!list_empty(&sb->s_dentry_lru)) {
 		list_splice_init(&sb->s_dentry_lru, &tmp);
+		spin_unlock(&dcache_lru_lock);
 		shrink_dentry_list(&tmp);
+		spin_lock(&dcache_lru_lock);
 	}
 	spin_unlock(&dcache_lru_lock);
 }
-- 
cgit v0.10.2


From 77812a1ef139d84270d27faacc0630c887411013 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:48 +1100
Subject: fs: consolidate dentry kill sequence

The tricky locking for disposing of a dentry is duplicated 3 times in the
dcache (dput, pruning a dentry from the LRU, and pruning its ancestors).
Consolidate them all into a single function dentry_kill.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index d1840b3..dc0551c 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -284,6 +284,40 @@ void d_drop(struct dentry *dentry)
 }
 EXPORT_SYMBOL(d_drop);
 
+/*
+ * Finish off a dentry we've decided to kill.
+ * dentry->d_lock must be held, returns with it unlocked.
+ * If ref is non-zero, then decrement the refcount too.
+ * Returns dentry requiring refcount drop, or NULL if we're done.
+ */
+static inline struct dentry *dentry_kill(struct dentry *dentry, int ref)
+	__releases(dentry->d_lock)
+{
+	struct dentry *parent;
+
+	if (!spin_trylock(&dcache_inode_lock)) {
+relock:
+		spin_unlock(&dentry->d_lock);
+		cpu_relax();
+		return dentry; /* try again with same dentry */
+	}
+	if (IS_ROOT(dentry))
+		parent = NULL;
+	else
+		parent = dentry->d_parent;
+	if (parent && !spin_trylock(&parent->d_lock)) {
+		spin_unlock(&dcache_inode_lock);
+		goto relock;
+	}
+	if (ref)
+		dentry->d_count--;
+	/* if dentry was on the d_lru list delete it from there */
+	dentry_lru_del(dentry);
+	/* if it was on the hash then remove it */
+	__d_drop(dentry);
+	return d_kill(dentry, parent);
+}
+
 /* 
  * This is dput
  *
@@ -309,13 +343,9 @@ EXPORT_SYMBOL(d_drop);
  * call the dentry unlink method as well as removing it from the queues and
  * releasing its resources. If the parent dentries were scheduled for release
  * they too may now get deleted.
- *
- * no dcache lock, please.
  */
-
 void dput(struct dentry *dentry)
 {
-	struct dentry *parent;
 	if (!dentry)
 		return;
 
@@ -348,26 +378,7 @@ repeat:
 	return;
 
 kill_it:
-	if (!spin_trylock(&dcache_inode_lock)) {
-relock:
-		spin_unlock(&dentry->d_lock);
-		cpu_relax();
-		goto repeat;
-	}
-	if (IS_ROOT(dentry))
-		parent = NULL;
-	else
-		parent = dentry->d_parent;
-	if (parent && !spin_trylock(&parent->d_lock)) {
-		spin_unlock(&dcache_inode_lock);
-		goto relock;
-	}
-	dentry->d_count--;
-	/* if dentry was on the d_lru list delete it from there */
-	dentry_lru_del(dentry);
-	/* if it was on the hash (d_delete case), then remove it */
-	__d_drop(dentry);
-	dentry = d_kill(dentry, parent);
+	dentry = dentry_kill(dentry, 1);
 	if (dentry)
 		goto repeat;
 }
@@ -563,51 +574,43 @@ restart:
 EXPORT_SYMBOL(d_prune_aliases);
 
 /*
- * Throw away a dentry - free the inode, dput the parent.  This requires that
- * the LRU list has already been removed.
+ * Try to throw away a dentry - free the inode, dput the parent.
+ * Requires dentry->d_lock is held, and dentry->d_count == 0.
+ * Releases dentry->d_lock.
  *
- * Try to prune ancestors as well.  This is necessary to prevent
- * quadratic behavior of shrink_dcache_parent(), but is also expected
- * to be beneficial in reducing dentry cache fragmentation.
+ * This may fail if locks cannot be acquired no problem, just try again.
  */
-static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
+static void try_prune_one_dentry(struct dentry *dentry)
 	__releases(dentry->d_lock)
-	__releases(parent->d_lock)
-	__releases(dcache_inode_lock)
 {
-	__d_drop(dentry);
-	dentry = d_kill(dentry, parent);
+	struct dentry *parent;
 
+	parent = dentry_kill(dentry, 0);
 	/*
-	 * Prune ancestors.
+	 * If dentry_kill returns NULL, we have nothing more to do.
+	 * if it returns the same dentry, trylocks failed. In either
+	 * case, just loop again.
+	 *
+	 * Otherwise, we need to prune ancestors too. This is necessary
+	 * to prevent quadratic behavior of shrink_dcache_parent(), but
+	 * is also expected to be beneficial in reducing dentry cache
+	 * fragmentation.
 	 */
+	if (!parent)
+		return;
+	if (parent == dentry)
+		return;
+
+	/* Prune ancestors. */
+	dentry = parent;
 	while (dentry) {
-relock:
 		spin_lock(&dentry->d_lock);
 		if (dentry->d_count > 1) {
 			dentry->d_count--;
 			spin_unlock(&dentry->d_lock);
 			return;
 		}
-		if (!spin_trylock(&dcache_inode_lock)) {
-relock2:
-			spin_unlock(&dentry->d_lock);
-			cpu_relax();
-			goto relock;
-		}
-
-		if (IS_ROOT(dentry))
-			parent = NULL;
-		else
-			parent = dentry->d_parent;
-		if (parent && !spin_trylock(&parent->d_lock)) {
-			spin_unlock(&dcache_inode_lock);
-			goto relock2;
-		}
-		dentry->d_count--;
-		dentry_lru_del(dentry);
-		__d_drop(dentry);
-		dentry = d_kill(dentry, parent);
+		dentry = dentry_kill(dentry, 1);
 	}
 }
 
@@ -617,8 +620,6 @@ static void shrink_dentry_list(struct list_head *list)
 
 	rcu_read_lock();
 	for (;;) {
-		struct dentry *parent;
-
 		dentry = list_entry_rcu(list->prev, struct dentry, d_lru);
 		if (&dentry->d_lru == list)
 			break; /* empty */
@@ -639,24 +640,10 @@ static void shrink_dentry_list(struct list_head *list)
 			continue;
 		}
 
-		if (!spin_trylock(&dcache_inode_lock)) {
-relock:
-			spin_unlock(&dentry->d_lock);
-			cpu_relax();
-			continue;
-		}
-		if (IS_ROOT(dentry))
-			parent = NULL;
-		else
-			parent = dentry->d_parent;
-		if (parent && !spin_trylock(&parent->d_lock)) {
-			spin_unlock(&dcache_inode_lock);
-			goto relock;
-		}
-		dentry_lru_del(dentry);
-
 		rcu_read_unlock();
-		prune_one_dentry(dentry, parent);
+
+		try_prune_one_dentry(dentry);
+
 		rcu_read_lock();
 	}
 	rcu_read_unlock();
-- 
cgit v0.10.2


From fa0d7e3de6d6fc5004ad9dea0dd6b286af8f03e9 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:49 +1100
Subject: fs: icache RCU free inodes

RCU free the struct inode. This will allow:

- Subsequent store-free path walking patch. The inode must be consulted for
  permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
  to take i_lock no longer need to take sb_inode_list_lock to walk the list in
  the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
  page lock to follow page->mapping.

The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.

In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.

The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index 1eb7695..ccf0ce7 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -346,3 +346,17 @@ look at examples of other filesystems) for guidance.
 for details of what locks to replace dcache_lock with in order to protect
 particular things. Most of the time, a filesystem only needs ->d_lock, which
 protects *all* the dcache state of a given dentry.
+
+--
+[mandatory]
+
+	Filesystems must RCU-free their inodes, if they can have been accessed
+via rcu-walk path walk (basically, if the file can have had a path name in the
+vfs namespace).
+
+	i_dentry and i_rcu share storage in a union, and the vfs expects
+i_dentry to be reinitialized before it is freed, so an:
+
+  INIT_LIST_HEAD(&inode->i_dentry);
+
+must be done in the RCU callback.
diff --git a/arch/powerpc/platforms/cell/spufs/inode.c b/arch/powerpc/platforms/cell/spufs/inode.c
index 03185de..856e9c3 100644
--- a/arch/powerpc/platforms/cell/spufs/inode.c
+++ b/arch/powerpc/platforms/cell/spufs/inode.c
@@ -71,12 +71,18 @@ spufs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void
-spufs_destroy_inode(struct inode *inode)
+static void spufs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(spufs_inode_cache, SPUFS_I(inode));
 }
 
+static void spufs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, spufs_i_callback);
+}
+
 static void
 spufs_init_once(void *p)
 {
diff --git a/drivers/staging/pohmelfs/inode.c b/drivers/staging/pohmelfs/inode.c
index 61685cc..cc8d2840 100644
--- a/drivers/staging/pohmelfs/inode.c
+++ b/drivers/staging/pohmelfs/inode.c
@@ -826,6 +826,13 @@ const struct address_space_operations pohmelfs_aops = {
 	.set_page_dirty 	= __set_page_dirty_nobuffers,
 };
 
+static void pohmelfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(pohmelfs_inode_cache, POHMELFS_I(inode));
+}
+
 /*
  * ->detroy_inode() callback. Deletes inode from the caches
  *  and frees private data.
@@ -842,8 +849,8 @@ static void pohmelfs_destroy_inode(struct inode *inode)
 
 	dprintk("%s: pi: %p, inode: %p, ino: %llu.\n",
 		__func__, pi, &pi->vfs_inode, pi->ino);
-	kmem_cache_free(pohmelfs_inode_cache, pi);
 	atomic_long_dec(&psb->total_inodes);
+	call_rcu(&inode->i_rcu, pohmelfs_i_callback);
 }
 
 /*
diff --git a/drivers/staging/smbfs/inode.c b/drivers/staging/smbfs/inode.c
index 540a984..244319d 100644
--- a/drivers/staging/smbfs/inode.c
+++ b/drivers/staging/smbfs/inode.c
@@ -62,11 +62,18 @@ static struct inode *smb_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void smb_destroy_inode(struct inode *inode)
+static void smb_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(smb_inode_cachep, SMB_I(inode));
 }
 
+static void smb_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, smb_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct smb_inode_info *ei = (struct smb_inode_info *) foo;
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 1073bca..f6f9081 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -237,10 +237,17 @@ struct inode *v9fs_alloc_inode(struct super_block *sb)
  *
  */
 
-void v9fs_destroy_inode(struct inode *inode)
+static void v9fs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(vcookie_cache, v9fs_inode2cookie(inode));
 }
+
+void v9fs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, v9fs_i_callback);
+}
 #endif
 
 /**
diff --git a/fs/adfs/super.c b/fs/adfs/super.c
index 959dbff..47dffc5 100644
--- a/fs/adfs/super.c
+++ b/fs/adfs/super.c
@@ -240,11 +240,18 @@ static struct inode *adfs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void adfs_destroy_inode(struct inode *inode)
+static void adfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(adfs_inode_cachep, ADFS_I(inode));
 }
 
+static void adfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, adfs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct adfs_inode_info *ei = (struct adfs_inode_info *) foo;
diff --git a/fs/affs/super.c b/fs/affs/super.c
index 0cf7f43..4c18fcf 100644
--- a/fs/affs/super.c
+++ b/fs/affs/super.c
@@ -95,11 +95,18 @@ static struct inode *affs_alloc_inode(struct super_block *sb)
 	return &i->vfs_inode;
 }
 
-static void affs_destroy_inode(struct inode *inode)
+static void affs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(affs_inode_cachep, AFFS_I(inode));
 }
 
+static void affs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, affs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct affs_inode_info *ei = (struct affs_inode_info *) foo;
diff --git a/fs/afs/super.c b/fs/afs/super.c
index 27201cf..f901a9d 100644
--- a/fs/afs/super.c
+++ b/fs/afs/super.c
@@ -498,6 +498,14 @@ static struct inode *afs_alloc_inode(struct super_block *sb)
 	return &vnode->vfs_inode;
 }
 
+static void afs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct afs_vnode *vnode = AFS_FS_I(inode);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(afs_inode_cachep, vnode);
+}
+
 /*
  * destroy an AFS inode struct
  */
@@ -511,7 +519,7 @@ static void afs_destroy_inode(struct inode *inode)
 
 	ASSERTCMP(vnode->server, ==, NULL);
 
-	kmem_cache_free(afs_inode_cachep, vnode);
+	call_rcu(&inode->i_rcu, afs_i_callback);
 	atomic_dec(&afs_count_active_inodes);
 }
 
diff --git a/fs/befs/linuxvfs.c b/fs/befs/linuxvfs.c
index aa4e7c7..de93581 100644
--- a/fs/befs/linuxvfs.c
+++ b/fs/befs/linuxvfs.c
@@ -284,12 +284,18 @@ befs_alloc_inode(struct super_block *sb)
         return &bi->vfs_inode;
 }
 
-static void
-befs_destroy_inode(struct inode *inode)
+static void befs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
         kmem_cache_free(befs_inode_cachep, BEFS_I(inode));
 }
 
+static void befs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, befs_i_callback);
+}
+
 static void init_once(void *foo)
 {
         struct befs_inode_info *bi = (struct befs_inode_info *) foo;
diff --git a/fs/bfs/inode.c b/fs/bfs/inode.c
index 76db6d7..a8e37f8 100644
--- a/fs/bfs/inode.c
+++ b/fs/bfs/inode.c
@@ -248,11 +248,18 @@ static struct inode *bfs_alloc_inode(struct super_block *sb)
 	return &bi->vfs_inode;
 }
 
-static void bfs_destroy_inode(struct inode *inode)
+static void bfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(bfs_inode_cachep, BFS_I(inode));
 }
 
+static void bfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, bfs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct bfs_inode_info *bi = foo;
diff --git a/fs/block_dev.c b/fs/block_dev.c
index 4230252..771f235 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -409,13 +409,20 @@ static struct inode *bdev_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void bdev_destroy_inode(struct inode *inode)
+static void bdev_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
 	struct bdev_inode *bdi = BDEV_I(inode);
 
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(bdev_cachep, bdi);
 }
 
+static void bdev_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, bdev_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct bdev_inode *ei = (struct bdev_inode *) foo;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 7ce9f89..f9d2994 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -6495,6 +6495,13 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	return inode;
 }
 
+static void btrfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
+}
+
 void btrfs_destroy_inode(struct inode *inode)
 {
 	struct btrfs_ordered_extent *ordered;
@@ -6564,7 +6571,7 @@ void btrfs_destroy_inode(struct inode *inode)
 	inode_tree_del(inode);
 	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
 free:
-	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
+	call_rcu(&inode->i_rcu, btrfs_i_callback);
 }
 
 int btrfs_drop_inode(struct inode *inode)
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 2a48caf..47f8c8b 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -368,6 +368,15 @@ struct inode *ceph_alloc_inode(struct super_block *sb)
 	return &ci->vfs_inode;
 }
 
+static void ceph_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct ceph_inode_info *ci = ceph_inode(inode);
+
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(ceph_inode_cachep, ci);
+}
+
 void ceph_destroy_inode(struct inode *inode)
 {
 	struct ceph_inode_info *ci = ceph_inode(inode);
@@ -407,7 +416,7 @@ void ceph_destroy_inode(struct inode *inode)
 	if (ci->i_xattrs.prealloc_blob)
 		ceph_buffer_put(ci->i_xattrs.prealloc_blob);
 
-	kmem_cache_free(ceph_inode_cachep, ci);
+	call_rcu(&inode->i_rcu, ceph_i_callback);
 }
 
 
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index 3936aa7..223717d 100644
--- a/fs/cifs/cifsfs.c
+++ b/fs/cifs/cifsfs.c
@@ -334,10 +334,17 @@ cifs_alloc_inode(struct super_block *sb)
 	return &cifs_inode->vfs_inode;
 }
 
+static void cifs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(cifs_inode_cachep, CIFS_I(inode));
+}
+
 static void
 cifs_destroy_inode(struct inode *inode)
 {
-	kmem_cache_free(cifs_inode_cachep, CIFS_I(inode));
+	call_rcu(&inode->i_rcu, cifs_i_callback);
 }
 
 static void
diff --git a/fs/coda/inode.c b/fs/coda/inode.c
index 5ea57c8..50dc7d1 100644
--- a/fs/coda/inode.c
+++ b/fs/coda/inode.c
@@ -56,11 +56,18 @@ static struct inode *coda_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void coda_destroy_inode(struct inode *inode)
+static void coda_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(coda_inode_cachep, ITOC(inode));
 }
 
+static void coda_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, coda_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct coda_inode_info *ei = (struct coda_inode_info *) foo;
diff --git a/fs/ecryptfs/super.c b/fs/ecryptfs/super.c
index 2720178..3042fe1 100644
--- a/fs/ecryptfs/super.c
+++ b/fs/ecryptfs/super.c
@@ -62,6 +62,16 @@ out:
 	return inode;
 }
 
+static void ecryptfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct ecryptfs_inode_info *inode_info;
+	inode_info = ecryptfs_inode_to_private(inode);
+
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(ecryptfs_inode_info_cache, inode_info);
+}
+
 /**
  * ecryptfs_destroy_inode
  * @inode: The ecryptfs inode
@@ -88,7 +98,7 @@ static void ecryptfs_destroy_inode(struct inode *inode)
 		}
 	}
 	ecryptfs_destroy_crypt_stat(&inode_info->crypt_stat);
-	kmem_cache_free(ecryptfs_inode_info_cache, inode_info);
+	call_rcu(&inode->i_rcu, ecryptfs_i_callback);
 }
 
 /**
diff --git a/fs/efs/super.c b/fs/efs/super.c
index 5073a07..0f31acb 100644
--- a/fs/efs/super.c
+++ b/fs/efs/super.c
@@ -65,11 +65,18 @@ static struct inode *efs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void efs_destroy_inode(struct inode *inode)
+static void efs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(efs_inode_cachep, INODE_INFO(inode));
 }
 
+static void efs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, efs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct efs_inode_info *ei = (struct efs_inode_info *) foo;
diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 79c3ae6..8c6c466 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -150,12 +150,19 @@ static struct inode *exofs_alloc_inode(struct super_block *sb)
 	return &oi->vfs_inode;
 }
 
+static void exofs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
+}
+
 /*
  * Remove an inode from the cache
  */
 static void exofs_destroy_inode(struct inode *inode)
 {
-	kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
+	call_rcu(&inode->i_rcu, exofs_i_callback);
 }
 
 /*
diff --git a/fs/ext2/super.c b/fs/ext2/super.c
index d89e0b6..e0c6380 100644
--- a/fs/ext2/super.c
+++ b/fs/ext2/super.c
@@ -161,11 +161,18 @@ static struct inode *ext2_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void ext2_destroy_inode(struct inode *inode)
+static void ext2_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 }
 
+static void ext2_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, ext2_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
diff --git a/fs/ext3/super.c b/fs/ext3/super.c
index acf8695..77ce161 100644
--- a/fs/ext3/super.c
+++ b/fs/ext3/super.c
@@ -479,6 +479,13 @@ static struct inode *ext3_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
+static void ext3_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
+}
+
 static void ext3_destroy_inode(struct inode *inode)
 {
 	if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
@@ -489,7 +496,7 @@ static void ext3_destroy_inode(struct inode *inode)
 				false);
 		dump_stack();
 	}
-	kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
+	call_rcu(&inode->i_rcu, ext3_i_callback);
 }
 
 static void init_once(void *foo)
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index fb15c9c..cd37f9d 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -841,6 +841,13 @@ static int ext4_drop_inode(struct inode *inode)
 	return drop;
 }
 
+static void ext4_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
+}
+
 static void ext4_destroy_inode(struct inode *inode)
 {
 	ext4_ioend_wait(inode);
@@ -853,7 +860,7 @@ static void ext4_destroy_inode(struct inode *inode)
 				true);
 		dump_stack();
 	}
-	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
+	call_rcu(&inode->i_rcu, ext4_i_callback);
 }
 
 static void init_once(void *foo)
diff --git a/fs/fat/inode.c b/fs/fat/inode.c
index ad6998a..8cccfeb 100644
--- a/fs/fat/inode.c
+++ b/fs/fat/inode.c
@@ -514,11 +514,18 @@ static struct inode *fat_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void fat_destroy_inode(struct inode *inode)
+static void fat_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(fat_inode_cachep, MSDOS_I(inode));
 }
 
+static void fat_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, fat_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct msdos_inode_info *ei = (struct msdos_inode_info *)foo;
diff --git a/fs/freevxfs/vxfs_inode.c b/fs/freevxfs/vxfs_inode.c
index 8c04eac..2ba6719 100644
--- a/fs/freevxfs/vxfs_inode.c
+++ b/fs/freevxfs/vxfs_inode.c
@@ -337,6 +337,13 @@ vxfs_iget(struct super_block *sbp, ino_t ino)
 	return ip;
 }
 
+static void vxfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(vxfs_inode_cachep, inode->i_private);
+}
+
 /**
  * vxfs_evict_inode - remove inode from main memory
  * @ip:		inode to discard.
@@ -350,5 +357,5 @@ vxfs_evict_inode(struct inode *ip)
 {
 	truncate_inode_pages(&ip->i_data, 0);
 	end_writeback(ip);
-	kmem_cache_free(vxfs_inode_cachep, ip->i_private);
+	call_rcu(&ip->i_rcu, vxfs_i_callback);
 }
diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index cfce3ad..44e0a6c 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -99,6 +99,13 @@ static struct inode *fuse_alloc_inode(struct super_block *sb)
 	return inode;
 }
 
+static void fuse_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(fuse_inode_cachep, inode);
+}
+
 static void fuse_destroy_inode(struct inode *inode)
 {
 	struct fuse_inode *fi = get_fuse_inode(inode);
@@ -106,7 +113,7 @@ static void fuse_destroy_inode(struct inode *inode)
 	BUG_ON(!list_empty(&fi->queued_writes));
 	if (fi->forget_req)
 		fuse_request_free(fi->forget_req);
-	kmem_cache_free(fuse_inode_cachep, inode);
+	call_rcu(&inode->i_rcu, fuse_i_callback);
 }
 
 void fuse_send_forget(struct fuse_conn *fc, struct fuse_req *req,
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 2b2c499..16c2eca 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1405,11 +1405,18 @@ static struct inode *gfs2_alloc_inode(struct super_block *sb)
 	return &ip->i_inode;
 }
 
-static void gfs2_destroy_inode(struct inode *inode)
+static void gfs2_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(gfs2_inode_cachep, inode);
 }
 
+static void gfs2_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, gfs2_i_callback);
+}
+
 const struct super_operations gfs2_super_ops = {
 	.alloc_inode		= gfs2_alloc_inode,
 	.destroy_inode		= gfs2_destroy_inode,
diff --git a/fs/hfs/super.c b/fs/hfs/super.c
index 4824c27..ef4ee57 100644
--- a/fs/hfs/super.c
+++ b/fs/hfs/super.c
@@ -167,11 +167,18 @@ static struct inode *hfs_alloc_inode(struct super_block *sb)
 	return i ? &i->vfs_inode : NULL;
 }
 
-static void hfs_destroy_inode(struct inode *inode)
+static void hfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(hfs_inode_cachep, HFS_I(inode));
 }
 
+static void hfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, hfs_i_callback);
+}
+
 static const struct super_operations hfs_super_operations = {
 	.alloc_inode	= hfs_alloc_inode,
 	.destroy_inode	= hfs_destroy_inode,
diff --git a/fs/hfsplus/super.c b/fs/hfsplus/super.c
index 52cc746..182e83a 100644
--- a/fs/hfsplus/super.c
+++ b/fs/hfsplus/super.c
@@ -488,11 +488,19 @@ static struct inode *hfsplus_alloc_inode(struct super_block *sb)
 	return i ? &i->vfs_inode : NULL;
 }
 
-static void hfsplus_destroy_inode(struct inode *inode)
+static void hfsplus_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(hfsplus_inode_cachep, HFSPLUS_I(inode));
 }
 
+static void hfsplus_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, hfsplus_i_callback);
+}
+
 #define HFSPLUS_INODE_SIZE	sizeof(struct hfsplus_inode_info)
 
 static struct dentry *hfsplus_mount(struct file_system_type *fs_type,
diff --git a/fs/hostfs/hostfs_kern.c b/fs/hostfs/hostfs_kern.c
index 39dc505..861113f 100644
--- a/fs/hostfs/hostfs_kern.c
+++ b/fs/hostfs/hostfs_kern.c
@@ -247,11 +247,18 @@ static void hostfs_evict_inode(struct inode *inode)
 	}
 }
 
-static void hostfs_destroy_inode(struct inode *inode)
+static void hostfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kfree(HOSTFS_I(inode));
 }
 
+static void hostfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, hostfs_i_callback);
+}
+
 static int hostfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 {
 	const char *root_path = vfs->mnt_sb->s_fs_info;
diff --git a/fs/hpfs/super.c b/fs/hpfs/super.c
index 6c5f015..49935ba 100644
--- a/fs/hpfs/super.c
+++ b/fs/hpfs/super.c
@@ -177,11 +177,18 @@ static struct inode *hpfs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void hpfs_destroy_inode(struct inode *inode)
+static void hpfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(hpfs_inode_cachep, hpfs_i(inode));
 }
 
+static void hpfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, hpfs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct hpfs_inode_info *ei = (struct hpfs_inode_info *) foo;
diff --git a/fs/hppfs/hppfs.c b/fs/hppfs/hppfs.c
index f702b5f..87ed48e 100644
--- a/fs/hppfs/hppfs.c
+++ b/fs/hppfs/hppfs.c
@@ -632,11 +632,18 @@ void hppfs_evict_inode(struct inode *ino)
 	mntput(ino->i_sb->s_fs_info);
 }
 
-static void hppfs_destroy_inode(struct inode *inode)
+static void hppfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kfree(HPPFS_I(inode));
 }
 
+static void hppfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, hppfs_i_callback);
+}
+
 static const struct super_operations hppfs_sbops = {
 	.alloc_inode	= hppfs_alloc_inode,
 	.destroy_inode	= hppfs_destroy_inode,
diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index a5fe681..9885082 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -663,11 +663,18 @@ static struct inode *hugetlbfs_alloc_inode(struct super_block *sb)
 	return &p->vfs_inode;
 }
 
+static void hugetlbfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode));
+}
+
 static void hugetlbfs_destroy_inode(struct inode *inode)
 {
 	hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb));
 	mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy);
-	kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode));
+	call_rcu(&inode->i_rcu, hugetlbfs_i_callback);
 }
 
 static const struct address_space_operations hugetlbfs_aops = {
diff --git a/fs/inode.c b/fs/inode.c
index 5a0a898..6751dfe 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -272,6 +272,13 @@ void __destroy_inode(struct inode *inode)
 }
 EXPORT_SYMBOL(__destroy_inode);
 
+static void i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(inode_cachep, inode);
+}
+
 static void destroy_inode(struct inode *inode)
 {
 	BUG_ON(!list_empty(&inode->i_lru));
@@ -279,7 +286,7 @@ static void destroy_inode(struct inode *inode)
 	if (inode->i_sb->s_op->destroy_inode)
 		inode->i_sb->s_op->destroy_inode(inode);
 	else
-		kmem_cache_free(inode_cachep, (inode));
+		call_rcu(&inode->i_rcu, i_callback);
 }
 
 /*
@@ -432,6 +439,7 @@ void end_writeback(struct inode *inode)
 	BUG_ON(!(inode->i_state & I_FREEING));
 	BUG_ON(inode->i_state & I_CLEAR);
 	inode_sync_wait(inode);
+	/* don't need i_lock here, no concurrent mods to i_state */
 	inode->i_state = I_FREEING | I_CLEAR;
 }
 EXPORT_SYMBOL(end_writeback);
diff --git a/fs/isofs/inode.c b/fs/isofs/inode.c
index d204ee4..d8f3a65 100644
--- a/fs/isofs/inode.c
+++ b/fs/isofs/inode.c
@@ -81,11 +81,18 @@ static struct inode *isofs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void isofs_destroy_inode(struct inode *inode)
+static void isofs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(isofs_inode_cachep, ISOFS_I(inode));
 }
 
+static void isofs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, isofs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct iso_inode_info *ei = foo;
diff --git a/fs/jffs2/super.c b/fs/jffs2/super.c
index c86041b..853b8e3 100644
--- a/fs/jffs2/super.c
+++ b/fs/jffs2/super.c
@@ -40,11 +40,18 @@ static struct inode *jffs2_alloc_inode(struct super_block *sb)
 	return &f->vfs_inode;
 }
 
-static void jffs2_destroy_inode(struct inode *inode)
+static void jffs2_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(jffs2_inode_cachep, JFFS2_INODE_INFO(inode));
 }
 
+static void jffs2_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, jffs2_i_callback);
+}
+
 static void jffs2_i_init_once(void *foo)
 {
 	struct jffs2_inode_info *f = foo;
diff --git a/fs/jfs/super.c b/fs/jfs/super.c
index 0669fc1..b715b0f 100644
--- a/fs/jfs/super.c
+++ b/fs/jfs/super.c
@@ -115,6 +115,14 @@ static struct inode *jfs_alloc_inode(struct super_block *sb)
 	return &jfs_inode->vfs_inode;
 }
 
+static void jfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct jfs_inode_info *ji = JFS_IP(inode);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(jfs_inode_cachep, ji);
+}
+
 static void jfs_destroy_inode(struct inode *inode)
 {
 	struct jfs_inode_info *ji = JFS_IP(inode);
@@ -128,7 +136,7 @@ static void jfs_destroy_inode(struct inode *inode)
 		ji->active_ag = -1;
 	}
 	spin_unlock_irq(&ji->ag_lock);
-	kmem_cache_free(jfs_inode_cachep, ji);
+	call_rcu(&inode->i_rcu, jfs_i_callback);
 }
 
 static int jfs_statfs(struct dentry *dentry, struct kstatfs *buf)
diff --git a/fs/logfs/inode.c b/fs/logfs/inode.c
index d8c71ec..03b8c24 100644
--- a/fs/logfs/inode.c
+++ b/fs/logfs/inode.c
@@ -141,13 +141,20 @@ struct inode *logfs_safe_iget(struct super_block *sb, ino_t ino, int *is_cached)
 	return __logfs_iget(sb, ino);
 }
 
+static void logfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(logfs_inode_cache, logfs_inode(inode));
+}
+
 static void __logfs_destroy_inode(struct inode *inode)
 {
 	struct logfs_inode *li = logfs_inode(inode);
 
 	BUG_ON(li->li_block);
 	list_del(&li->li_freeing_list);
-	kmem_cache_free(logfs_inode_cache, li);
+	call_rcu(&inode->i_rcu, logfs_i_callback);
 }
 
 static void logfs_destroy_inode(struct inode *inode)
diff --git a/fs/minix/inode.c b/fs/minix/inode.c
index fb20208..ae0b83f 100644
--- a/fs/minix/inode.c
+++ b/fs/minix/inode.c
@@ -68,11 +68,18 @@ static struct inode *minix_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void minix_destroy_inode(struct inode *inode)
+static void minix_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(minix_inode_cachep, minix_i(inode));
 }
 
+static void minix_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, minix_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct minix_inode_info *ei = (struct minix_inode_info *) foo;
diff --git a/fs/ncpfs/inode.c b/fs/ncpfs/inode.c
index 8fb93b6..60047db 100644
--- a/fs/ncpfs/inode.c
+++ b/fs/ncpfs/inode.c
@@ -58,11 +58,18 @@ static struct inode *ncp_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void ncp_destroy_inode(struct inode *inode)
+static void ncp_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(ncp_inode_cachep, NCP_FINFO(inode));
 }
 
+static void ncp_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, ncp_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct ncp_inode_info *ei = (struct ncp_inode_info *) foo;
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index e67e31c..017daa3 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -1438,11 +1438,18 @@ struct inode *nfs_alloc_inode(struct super_block *sb)
 	return &nfsi->vfs_inode;
 }
 
-void nfs_destroy_inode(struct inode *inode)
+static void nfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
 }
 
+void nfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, nfs_i_callback);
+}
+
 static inline void nfs4_init_once(struct nfs_inode *nfsi)
 {
 #ifdef CONFIG_NFS_V4
diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
index d36fc7e..e2dcc9c 100644
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@@ -162,10 +162,13 @@ struct inode *nilfs_alloc_inode(struct super_block *sb)
 	return &ii->vfs_inode;
 }
 
-void nilfs_destroy_inode(struct inode *inode)
+static void nilfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
 	struct nilfs_mdt_info *mdi = NILFS_MDT(inode);
 
+	INIT_LIST_HEAD(&inode->i_dentry);
+
 	if (mdi) {
 		kfree(mdi->mi_bgl); /* kfree(NULL) is safe */
 		kfree(mdi);
@@ -173,6 +176,11 @@ void nilfs_destroy_inode(struct inode *inode)
 	kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
 }
 
+void nilfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, nilfs_i_callback);
+}
+
 static int nilfs_sync_super(struct nilfs_sb_info *sbi, int flag)
 {
 	struct the_nilfs *nilfs = sbi->s_nilfs;
diff --git a/fs/ntfs/inode.c b/fs/ntfs/inode.c
index 93622b1..a627ed8 100644
--- a/fs/ntfs/inode.c
+++ b/fs/ntfs/inode.c
@@ -332,6 +332,13 @@ struct inode *ntfs_alloc_big_inode(struct super_block *sb)
 	return NULL;
 }
 
+static void ntfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
+}
+
 void ntfs_destroy_big_inode(struct inode *inode)
 {
 	ntfs_inode *ni = NTFS_I(inode);
@@ -340,7 +347,7 @@ void ntfs_destroy_big_inode(struct inode *inode)
 	BUG_ON(ni->page);
 	if (!atomic_dec_and_test(&ni->count))
 		BUG();
-	kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
+	call_rcu(&inode->i_rcu, ntfs_i_callback);
 }
 
 static inline ntfs_inode *ntfs_alloc_extent_inode(void)
diff --git a/fs/ocfs2/dlmfs/dlmfs.c b/fs/ocfs2/dlmfs/dlmfs.c
index b2df490..8c5c0ed 100644
--- a/fs/ocfs2/dlmfs/dlmfs.c
+++ b/fs/ocfs2/dlmfs/dlmfs.c
@@ -351,11 +351,18 @@ static struct inode *dlmfs_alloc_inode(struct super_block *sb)
 	return &ip->ip_vfs_inode;
 }
 
-static void dlmfs_destroy_inode(struct inode *inode)
+static void dlmfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(dlmfs_inode_cache, DLMFS_I(inode));
 }
 
+static void dlmfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, dlmfs_i_callback);
+}
+
 static void dlmfs_evict_inode(struct inode *inode)
 {
 	int status;
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index cfeab7c..17ff46f 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -569,11 +569,18 @@ static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 	return &oi->vfs_inode;
 }
 
-static void ocfs2_destroy_inode(struct inode *inode)
+static void ocfs2_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 }
 
+static void ocfs2_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, ocfs2_i_callback);
+}
+
 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 						unsigned int cbits)
 {
diff --git a/fs/openpromfs/inode.c b/fs/openpromfs/inode.c
index 911e61f..a2a5bff 100644
--- a/fs/openpromfs/inode.c
+++ b/fs/openpromfs/inode.c
@@ -343,11 +343,18 @@ static struct inode *openprom_alloc_inode(struct super_block *sb)
 	return &oi->vfs_inode;
 }
 
-static void openprom_destroy_inode(struct inode *inode)
+static void openprom_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(op_inode_cachep, OP_I(inode));
 }
 
+static void openprom_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, openprom_i_callback);
+}
+
 static struct inode *openprom_iget(struct super_block *sb, ino_t ino)
 {
 	struct inode *inode;
diff --git a/fs/proc/inode.c b/fs/proc/inode.c
index 3ddb606..6bcb926 100644
--- a/fs/proc/inode.c
+++ b/fs/proc/inode.c
@@ -65,11 +65,18 @@ static struct inode *proc_alloc_inode(struct super_block *sb)
 	return inode;
 }
 
-static void proc_destroy_inode(struct inode *inode)
+static void proc_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(proc_inode_cachep, PROC_I(inode));
 }
 
+static void proc_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, proc_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct proc_inode *ei = (struct proc_inode *) foo;
diff --git a/fs/qnx4/inode.c b/fs/qnx4/inode.c
index fcada42..e63b417 100644
--- a/fs/qnx4/inode.c
+++ b/fs/qnx4/inode.c
@@ -425,11 +425,18 @@ static struct inode *qnx4_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void qnx4_destroy_inode(struct inode *inode)
+static void qnx4_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(qnx4_inode_cachep, qnx4_i(inode));
 }
 
+static void qnx4_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, qnx4_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct qnx4_inode_info *ei = (struct qnx4_inode_info *) foo;
diff --git a/fs/reiserfs/super.c b/fs/reiserfs/super.c
index b243117..2575682 100644
--- a/fs/reiserfs/super.c
+++ b/fs/reiserfs/super.c
@@ -529,11 +529,18 @@ static struct inode *reiserfs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void reiserfs_destroy_inode(struct inode *inode)
+static void reiserfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(reiserfs_inode_cachep, REISERFS_I(inode));
 }
 
+static void reiserfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, reiserfs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct reiserfs_inode_info *ei = (struct reiserfs_inode_info *)foo;
diff --git a/fs/romfs/super.c b/fs/romfs/super.c
index 6647f90..2305e31 100644
--- a/fs/romfs/super.c
+++ b/fs/romfs/super.c
@@ -400,11 +400,18 @@ static struct inode *romfs_alloc_inode(struct super_block *sb)
 /*
  * return a spent inode to the slab cache
  */
-static void romfs_destroy_inode(struct inode *inode)
+static void romfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(romfs_inode_cachep, ROMFS_I(inode));
 }
 
+static void romfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, romfs_i_callback);
+}
+
 /*
  * get filesystem statistics
  */
diff --git a/fs/squashfs/super.c b/fs/squashfs/super.c
index 24de30b..20700b9 100644
--- a/fs/squashfs/super.c
+++ b/fs/squashfs/super.c
@@ -440,11 +440,18 @@ static struct inode *squashfs_alloc_inode(struct super_block *sb)
 }
 
 
-static void squashfs_destroy_inode(struct inode *inode)
+static void squashfs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(squashfs_inode_cachep, squashfs_i(inode));
 }
 
+static void squashfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, squashfs_i_callback);
+}
+
 
 static struct file_system_type squashfs_fs_type = {
 	.owner = THIS_MODULE,
diff --git a/fs/sysv/inode.c b/fs/sysv/inode.c
index de44d06..0630eb9 100644
--- a/fs/sysv/inode.c
+++ b/fs/sysv/inode.c
@@ -333,11 +333,18 @@ static struct inode *sysv_alloc_inode(struct super_block *sb)
 	return &si->vfs_inode;
 }
 
-static void sysv_destroy_inode(struct inode *inode)
+static void sysv_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(sysv_inode_cachep, SYSV_I(inode));
 }
 
+static void sysv_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, sysv_i_callback);
+}
+
 static void init_once(void *p)
 {
 	struct sysv_inode_info *si = (struct sysv_inode_info *)p;
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index 91fac54..6e11c29 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -272,12 +272,20 @@ static struct inode *ubifs_alloc_inode(struct super_block *sb)
 	return &ui->vfs_inode;
 };
 
+static void ubifs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct ubifs_inode *ui = ubifs_inode(inode);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(ubifs_inode_slab, ui);
+}
+
 static void ubifs_destroy_inode(struct inode *inode)
 {
 	struct ubifs_inode *ui = ubifs_inode(inode);
 
 	kfree(ui->data);
-	kmem_cache_free(ubifs_inode_slab, inode);
+	call_rcu(&inode->i_rcu, ubifs_i_callback);
 }
 
 /*
diff --git a/fs/udf/super.c b/fs/udf/super.c
index 4a5c7c6..b539d53 100644
--- a/fs/udf/super.c
+++ b/fs/udf/super.c
@@ -139,11 +139,18 @@ static struct inode *udf_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void udf_destroy_inode(struct inode *inode)
+static void udf_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(udf_inode_cachep, UDF_I(inode));
 }
 
+static void udf_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, udf_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct udf_inode_info *ei = (struct udf_inode_info *)foo;
diff --git a/fs/ufs/super.c b/fs/ufs/super.c
index 2c47dae..2c61ac5 100644
--- a/fs/ufs/super.c
+++ b/fs/ufs/super.c
@@ -1412,11 +1412,18 @@ static struct inode *ufs_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void ufs_destroy_inode(struct inode *inode)
+static void ufs_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(ufs_inode_cachep, UFS_I(inode));
 }
 
+static void ufs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, ufs_i_callback);
+}
+
 static void init_once(void *foo)
 {
 	struct ufs_inode_info *ei = (struct ufs_inode_info *) foo;
diff --git a/fs/xfs/xfs_iget.c b/fs/xfs/xfs_iget.c
index 0cdd269..d7de5a3 100644
--- a/fs/xfs/xfs_iget.c
+++ b/fs/xfs/xfs_iget.c
@@ -91,6 +91,17 @@ xfs_inode_alloc(
 	return ip;
 }
 
+STATIC void
+xfs_inode_free_callback(
+	struct rcu_head		*head)
+{
+	struct inode		*inode = container_of(head, struct inode, i_rcu);
+	struct xfs_inode	*ip = XFS_I(inode);
+
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_zone_free(xfs_inode_zone, ip);
+}
+
 void
 xfs_inode_free(
 	struct xfs_inode	*ip)
@@ -134,7 +145,7 @@ xfs_inode_free(
 	ASSERT(!spin_is_locked(&ip->i_flags_lock));
 	ASSERT(completion_done(&ip->i_flush));
 
-	kmem_zone_free(xfs_inode_zone, ip);
+	call_rcu(&ip->i_vnode.i_rcu, xfs_inode_free_callback);
 }
 
 /*
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 296cf2f..1ff4d0a 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -737,7 +737,10 @@ struct inode {
 	struct list_head	i_wb_list;	/* backing dev IO list */
 	struct list_head	i_lru;		/* inode LRU list */
 	struct list_head	i_sb_list;
-	struct list_head	i_dentry;
+	union {
+		struct list_head	i_dentry;
+		struct rcu_head		i_rcu;
+	};
 	unsigned long		i_ino;
 	atomic_t		i_count;
 	unsigned int		i_nlink;
diff --git a/include/linux/net.h b/include/linux/net.h
index 16faa13..06bde49 100644
--- a/include/linux/net.h
+++ b/include/linux/net.h
@@ -120,7 +120,6 @@ enum sock_shutdown_cmd {
 struct socket_wq {
 	wait_queue_head_t	wait;
 	struct fasync_struct	*fasync_list;
-	struct rcu_head		rcu;
 } ____cacheline_aligned_in_smp;
 
 /**
diff --git a/ipc/mqueue.c b/ipc/mqueue.c
index 035f439..14fb6d6 100644
--- a/ipc/mqueue.c
+++ b/ipc/mqueue.c
@@ -237,11 +237,18 @@ static struct inode *mqueue_alloc_inode(struct super_block *sb)
 	return &ei->vfs_inode;
 }
 
-static void mqueue_destroy_inode(struct inode *inode)
+static void mqueue_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
 }
 
+static void mqueue_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, mqueue_i_callback);
+}
+
 static void mqueue_evict_inode(struct inode *inode)
 {
 	struct mqueue_inode_info *info;
diff --git a/mm/shmem.c b/mm/shmem.c
index 47fdeeb..5ee67c9 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -2415,13 +2415,20 @@ static struct inode *shmem_alloc_inode(struct super_block *sb)
 	return &p->vfs_inode;
 }
 
+static void shmem_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(shmem_inode_cachep, SHMEM_I(inode));
+}
+
 static void shmem_destroy_inode(struct inode *inode)
 {
 	if ((inode->i_mode & S_IFMT) == S_IFREG) {
 		/* only struct inode is valid if it's an inline symlink */
 		mpol_free_shared_policy(&SHMEM_I(inode)->policy);
 	}
-	kmem_cache_free(shmem_inode_cachep, SHMEM_I(inode));
+	call_rcu(&inode->i_rcu, shmem_i_callback);
 }
 
 static void init_once(void *foo)
diff --git a/net/socket.c b/net/socket.c
index 088fb3f..97fff3a 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -262,20 +262,20 @@ static struct inode *sock_alloc_inode(struct super_block *sb)
 }
 
 
-static void wq_free_rcu(struct rcu_head *head)
+static void sock_free_rcu(struct rcu_head *head)
 {
-	struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct socket_alloc *ei = container_of(inode, struct socket_alloc,
+								vfs_inode);
 
-	kfree(wq);
+	kfree(ei->socket.wq);
+	INIT_LIST_HEAD(&inode->i_dentry);
+	kmem_cache_free(sock_inode_cachep, ei);
 }
 
 static void sock_destroy_inode(struct inode *inode)
 {
-	struct socket_alloc *ei;
-
-	ei = container_of(inode, struct socket_alloc, vfs_inode);
-	call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
-	kmem_cache_free(sock_inode_cachep, ei);
+	call_rcu(&inode->i_rcu, sock_free_rcu);
 }
 
 static void init_once(void *foo)
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index a0dc1a8..2899fe2 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -162,11 +162,19 @@ rpc_alloc_inode(struct super_block *sb)
 }
 
 static void
-rpc_destroy_inode(struct inode *inode)
+rpc_i_callback(struct rcu_head *head)
 {
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(rpc_inode_cachep, RPC_I(inode));
 }
 
+static void
+rpc_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, rpc_i_callback);
+}
+
 static int
 rpc_pipe_open(struct inode *inode, struct file *filp)
 {
-- 
cgit v0.10.2


From ff0c7d15f9787b7e8c601533c015295cc68329f8 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:50 +1100
Subject: fs: avoid inode RCU freeing for pseudo fs

Pseudo filesystems that don't put inode on RCU list or reachable by
rcu-walk dentries do not need to RCU free their inodes.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/inode.c b/fs/inode.c
index 6751dfe..da85e56 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -257,6 +257,12 @@ static struct inode *alloc_inode(struct super_block *sb)
 	return inode;
 }
 
+void free_inode_nonrcu(struct inode *inode)
+{
+	kmem_cache_free(inode_cachep, inode);
+}
+EXPORT_SYMBOL(free_inode_nonrcu);
+
 void __destroy_inode(struct inode *inode)
 {
 	BUG_ON(inode_has_buffers(inode));
diff --git a/fs/pipe.c b/fs/pipe.c
index 04629f3..ae3592d 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -1253,6 +1253,10 @@ out:
 	return ret;
 }
 
+static const struct super_operations pipefs_ops = {
+	.destroy_inode = free_inode_nonrcu,
+};
+
 /*
  * pipefs should _never_ be mounted by userland - too much of security hassle,
  * no real gain from having the whole whorehouse mounted. So we don't need
@@ -1262,7 +1266,7 @@ out:
 static struct dentry *pipefs_mount(struct file_system_type *fs_type,
 			 int flags, const char *dev_name, void *data)
 {
-	return mount_pseudo(fs_type, "pipe:", NULL, PIPEFS_MAGIC);
+	return mount_pseudo(fs_type, "pipe:", &pipefs_ops, PIPEFS_MAGIC);
 }
 
 static struct file_system_type pipe_fs_type = {
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 1ff4d0a..ea202ff 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2233,6 +2233,7 @@ extern void iget_failed(struct inode *);
 extern void end_writeback(struct inode *);
 extern void __destroy_inode(struct inode *);
 extern struct inode *new_inode(struct super_block *);
+extern void free_inode_nonrcu(struct inode *inode);
 extern int should_remove_suid(struct dentry *);
 extern int file_remove_suid(struct file *);
 
diff --git a/include/linux/net.h b/include/linux/net.h
index 06bde49..16faa13 100644
--- a/include/linux/net.h
+++ b/include/linux/net.h
@@ -120,6 +120,7 @@ enum sock_shutdown_cmd {
 struct socket_wq {
 	wait_queue_head_t	wait;
 	struct fasync_struct	*fasync_list;
+	struct rcu_head		rcu;
 } ____cacheline_aligned_in_smp;
 
 /**
diff --git a/net/socket.c b/net/socket.c
index 97fff3a..817dc92 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -262,20 +262,21 @@ static struct inode *sock_alloc_inode(struct super_block *sb)
 }
 
 
-static void sock_free_rcu(struct rcu_head *head)
+
+static void wq_free_rcu(struct rcu_head *head)
 {
-	struct inode *inode = container_of(head, struct inode, i_rcu);
-	struct socket_alloc *ei = container_of(inode, struct socket_alloc,
-								vfs_inode);
+	struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
 
-	kfree(ei->socket.wq);
-	INIT_LIST_HEAD(&inode->i_dentry);
-	kmem_cache_free(sock_inode_cachep, ei);
+	kfree(wq);
 }
 
 static void sock_destroy_inode(struct inode *inode)
 {
-	call_rcu(&inode->i_rcu, sock_free_rcu);
+	struct socket_alloc *ei;
+
+	ei = container_of(inode, struct socket_alloc, vfs_inode);
+	call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
+	kmem_cache_free(sock_inode_cachep, ei);
 }
 
 static void init_once(void *foo)
-- 
cgit v0.10.2


From 3c22cd5709e8143444a6d08682a87f4c57902df3 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:51 +1100
Subject: kernel: optimise seqlock

Add branch annotations for seqlock read fastpath, and introduce
__read_seqcount_begin and __read_seqcount_end functions, that can avoid the
smp_rmb() if used carefully. These will be used by store-free path walking
algorithm performance is critical and seqlocks are in use.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/include/linux/seqlock.h b/include/linux/seqlock.h
index 632205c..e98cd2e 100644
--- a/include/linux/seqlock.h
+++ b/include/linux/seqlock.h
@@ -107,7 +107,7 @@ static __always_inline int read_seqretry(const seqlock_t *sl, unsigned start)
 {
 	smp_rmb();
 
-	return (sl->sequence != start);
+	return unlikely(sl->sequence != start);
 }
 
 
@@ -125,14 +125,25 @@ typedef struct seqcount {
 #define SEQCNT_ZERO { 0 }
 #define seqcount_init(x)	do { *(x) = (seqcount_t) SEQCNT_ZERO; } while (0)
 
-/* Start of read using pointer to a sequence counter only.  */
-static inline unsigned read_seqcount_begin(const seqcount_t *s)
+/**
+ * __read_seqcount_begin - begin a seq-read critical section (without barrier)
+ * @s: pointer to seqcount_t
+ * Returns: count to be passed to read_seqcount_retry
+ *
+ * __read_seqcount_begin is like read_seqcount_begin, but has no smp_rmb()
+ * barrier. Callers should ensure that smp_rmb() or equivalent ordering is
+ * provided before actually loading any of the variables that are to be
+ * protected in this critical section.
+ *
+ * Use carefully, only in critical code, and comment how the barrier is
+ * provided.
+ */
+static inline unsigned __read_seqcount_begin(const seqcount_t *s)
 {
 	unsigned ret;
 
 repeat:
 	ret = s->sequence;
-	smp_rmb();
 	if (unlikely(ret & 1)) {
 		cpu_relax();
 		goto repeat;
@@ -140,14 +151,56 @@ repeat:
 	return ret;
 }
 
-/*
- * Test if reader processed invalid data because sequence number has changed.
+/**
+ * read_seqcount_begin - begin a seq-read critical section
+ * @s: pointer to seqcount_t
+ * Returns: count to be passed to read_seqcount_retry
+ *
+ * read_seqcount_begin opens a read critical section of the given seqcount.
+ * Validity of the critical section is tested by checking read_seqcount_retry
+ * function.
+ */
+static inline unsigned read_seqcount_begin(const seqcount_t *s)
+{
+	unsigned ret = __read_seqcount_begin(s);
+	smp_rmb();
+	return ret;
+}
+
+/**
+ * __read_seqcount_retry - end a seq-read critical section (without barrier)
+ * @s: pointer to seqcount_t
+ * @start: count, from read_seqcount_begin
+ * Returns: 1 if retry is required, else 0
+ *
+ * __read_seqcount_retry is like read_seqcount_retry, but has no smp_rmb()
+ * barrier. Callers should ensure that smp_rmb() or equivalent ordering is
+ * provided before actually loading any of the variables that are to be
+ * protected in this critical section.
+ *
+ * Use carefully, only in critical code, and comment how the barrier is
+ * provided.
+ */
+static inline int __read_seqcount_retry(const seqcount_t *s, unsigned start)
+{
+	return unlikely(s->sequence != start);
+}
+
+/**
+ * read_seqcount_retry - end a seq-read critical section
+ * @s: pointer to seqcount_t
+ * @start: count, from read_seqcount_begin
+ * Returns: 1 if retry is required, else 0
+ *
+ * read_seqcount_retry closes a read critical section of the given seqcount.
+ * If the critical section was invalid, it must be ignored (and typically
+ * retried).
  */
 static inline int read_seqcount_retry(const seqcount_t *s, unsigned start)
 {
 	smp_rmb();
 
-	return s->sequence != start;
+	return __read_seqcount_retry(s, start);
 }
 
 
@@ -167,6 +220,19 @@ static inline void write_seqcount_end(seqcount_t *s)
 	s->sequence++;
 }
 
+/**
+ * write_seqcount_barrier - invalidate in-progress read-side seq operations
+ * @s: pointer to seqcount_t
+ *
+ * After write_seqcount_barrier, no read-side seq operations will complete
+ * successfully and see data older than this.
+ */
+static inline void write_seqcount_barrier(seqcount_t *s)
+{
+	smp_wmb();
+	s->sequence+=2;
+}
+
 /*
  * Possible sw/hw IRQ protected versions of the interfaces.
  */
-- 
cgit v0.10.2


From 31e6b01f4183ff419a6d1f86177cbf4662347cec Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:52 +1100
Subject: fs: rcu-walk for path lookup

Perform common cases of path lookups without any stores or locking in the
ancestor dentry elements. This is called rcu-walk, as opposed to the current
algorithm which is a refcount based walk, or ref-walk.

This results in far fewer atomic operations on every path element,
significantly improving path lookup performance. It also avoids cacheline
bouncing on common dentries, significantly improving scalability.

The overall design is like this:
* LOOKUP_RCU is set in nd->flags, which distinguishes rcu-walk from ref-walk.
* Take the RCU lock for the entire path walk, starting with the acquiring
  of the starting path (eg. root/cwd/fd-path). So now dentry refcounts are
  not required for dentry persistence.
* synchronize_rcu is called when unregistering a filesystem, so we can
  access d_ops and i_ops during rcu-walk.
* Similarly take the vfsmount lock for the entire path walk. So now mnt
  refcounts are not required for persistence. Also we are free to perform mount
  lookups, and to assume dentry mount points and mount roots are stable up and
  down the path.
* Have a per-dentry seqlock to protect the dentry name, parent, and inode,
  so we can load this tuple atomically, and also check whether any of its
  members have changed.
* Dentry lookups (based on parent, candidate string tuple) recheck the parent
  sequence after the child is found in case anything changed in the parent
  during the path walk.
* inode is also RCU protected so we can load d_inode and use the inode for
  limited things.
* i_mode, i_uid, i_gid can be tested for exec permissions during path walk.
* i_op can be loaded.

When we reach the destination dentry, we lock it, recheck lookup sequence,
and increment its refcount and mountpoint refcount. RCU and vfsmount locks
are dropped. This is termed "dropping rcu-walk". If the dentry refcount does
not match, we can not drop rcu-walk gracefully at the current point in the
lokup, so instead return -ECHILD (for want of a better errno). This signals the
path walking code to re-do the entire lookup with a ref-walk.

Aside from the final dentry, there are other situations that may be encounted
where we cannot continue rcu-walk. In that case, we drop rcu-walk (ie. take
a reference on the last good dentry) and continue with a ref-walk. Again, if
we can drop rcu-walk gracefully, we return -ECHILD and do the whole lookup
using ref-walk. But it is very important that we can continue with ref-walk
for most cases, particularly to avoid the overhead of double lookups, and to
gain the scalability advantages on common path elements (like cwd and root).

The cases where rcu-walk cannot continue are:
* NULL dentry (ie. any uncached path element)
* parent with d_inode->i_op->permission or ACLs
* dentries with d_revalidate
* Following links

In future patches, permission checks and d_revalidate become rcu-walk aware. It
may be possible eventually to make following links rcu-walk aware.

Uncached path elements will always require dropping to ref-walk mode, at the
very least because i_mutex needs to be grabbed, and objects allocated.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/dentry-locking.txt b/Documentation/filesystems/dentry-locking.txt
deleted file mode 100644
index 30b6a40..0000000
--- a/Documentation/filesystems/dentry-locking.txt
+++ /dev/null
@@ -1,172 +0,0 @@
-RCU-based dcache locking model
-==============================
-
-On many workloads, the most common operation on dcache is to look up a
-dentry, given a parent dentry and the name of the child. Typically,
-for every open(), stat() etc., the dentry corresponding to the
-pathname will be looked up by walking the tree starting with the first
-component of the pathname and using that dentry along with the next
-component to look up the next level and so on. Since it is a frequent
-operation for workloads like multiuser environments and web servers,
-it is important to optimize this path.
-
-Prior to 2.5.10, dcache_lock was acquired in d_lookup and thus in
-every component during path look-up. Since 2.5.10 onwards, fast-walk
-algorithm changed this by holding the dcache_lock at the beginning and
-walking as many cached path component dentries as possible. This
-significantly decreases the number of acquisition of
-dcache_lock. However it also increases the lock hold time
-significantly and affects performance in large SMP machines. Since
-2.5.62 kernel, dcache has been using a new locking model that uses RCU
-to make dcache look-up lock-free.
-
-The current dcache locking model is not very different from the
-existing dcache locking model. Prior to 2.5.62 kernel, dcache_lock
-protected the hash chain, d_child, d_alias, d_lru lists as well as
-d_inode and several other things like mount look-up. RCU-based changes
-affect only the way the hash chain is protected. For everything else
-the dcache_lock must be taken for both traversing as well as
-updating. The hash chain updates too take the dcache_lock.  The
-significant change is the way d_lookup traverses the hash chain, it
-doesn't acquire the dcache_lock for this and rely on RCU to ensure
-that the dentry has not been *freed*.
-
-dcache_lock no longer exists, dentry locking is explained in fs/dcache.c
-
-Dcache locking details
-======================
-
-For many multi-user workloads, open() and stat() on files are very
-frequently occurring operations. Both involve walking of path names to
-find the dentry corresponding to the concerned file. In 2.4 kernel,
-dcache_lock was held during look-up of each path component. Contention
-and cache-line bouncing of this global lock caused significant
-scalability problems. With the introduction of RCU in Linux kernel,
-this was worked around by making the look-up of path components during
-path walking lock-free.
-
-
-Safe lock-free look-up of dcache hash table
-===========================================
-
-Dcache is a complex data structure with the hash table entries also
-linked together in other lists. In 2.4 kernel, dcache_lock protected
-all the lists. RCU dentry hash walking works like this:
-
-1. The deletion from hash chain is done using hlist_del_rcu() macro
-   which doesn't initialize next pointer of the deleted dentry and
-   this allows us to walk safely lock-free while a deletion is
-   happening. This is a standard hlist_rcu iteration.
-
-2. Insertion of a dentry into the hash table is done using
-   hlist_add_head_rcu() which take care of ordering the writes - the
-   writes to the dentry must be visible before the dentry is
-   inserted. This works in conjunction with hlist_for_each_rcu(),
-   which has since been replaced by hlist_for_each_entry_rcu(), while
-   walking the hash chain. The only requirement is that all
-   initialization to the dentry must be done before
-   hlist_add_head_rcu() since we don't have lock protection
-   while traversing the hash chain.
-
-3. The dentry looked up without holding locks cannot be returned for
-   walking if it is unhashed. It then may have a NULL d_inode or other
-   bogosity since RCU doesn't protect the other fields in the dentry. We
-   therefore use a flag DCACHE_UNHASHED to indicate unhashed dentries
-   and use this in conjunction with a per-dentry lock (d_lock). Once
-   looked up without locks, we acquire the per-dentry lock (d_lock) and
-   check if the dentry is unhashed. If so, the look-up is failed. If not,
-   the reference count of the dentry is increased and the dentry is
-   returned.
-
-4. Once a dentry is looked up, it must be ensured during the path walk
-   for that component it doesn't go away. In pre-2.5.10 code, this was
-   done holding a reference to the dentry. dcache_rcu does the same.
-   In some sense, dcache_rcu path walking looks like the pre-2.5.10
-   version.
-
-5. All dentry hash chain updates must take the per-dentry lock (see
-   fs/dcache.c). This excludes dput() to ensure that a dentry that has
-   been looked up concurrently does not get deleted before dget() can
-   take a ref.
-
-6. There are several ways to do reference counting of RCU protected
-   objects. One such example is in ipv4 route cache where deferred
-   freeing (using call_rcu()) is done as soon as the reference count
-   goes to zero. This cannot be done in the case of dentries because
-   tearing down of dentries require blocking (dentry_iput()) which
-   isn't supported from RCU callbacks. Instead, tearing down of
-   dentries happen synchronously in dput(), but actual freeing happens
-   later when RCU grace period is over. This allows safe lock-free
-   walking of the hash chains, but a matched dentry may have been
-   partially torn down. The checking of DCACHE_UNHASHED flag with
-   d_lock held detects such dentries and prevents them from being
-   returned from look-up.
-
-
-Maintaining POSIX rename semantics
-==================================
-
-Since look-up of dentries is lock-free, it can race against a
-concurrent rename operation. For example, during rename of file A to
-B, look-up of either A or B must succeed.  So, if look-up of B happens
-after A has been removed from the hash chain but not added to the new
-hash chain, it may fail.  Also, a comparison while the name is being
-written concurrently by a rename may result in false positive matches
-violating rename semantics.  Issues related to race with rename are
-handled as described below :
-
-1. Look-up can be done in two ways - d_lookup() which is safe from
-   simultaneous renames and __d_lookup() which is not.  If
-   __d_lookup() fails, it must be followed up by a d_lookup() to
-   correctly determine whether a dentry is in the hash table or
-   not. d_lookup() protects look-ups using a sequence lock
-   (rename_lock).
-
-2. The name associated with a dentry (d_name) may be changed if a
-   rename is allowed to happen simultaneously. To avoid memcmp() in
-   __d_lookup() go out of bounds due to a rename and false positive
-   comparison, the name comparison is done while holding the
-   per-dentry lock. This prevents concurrent renames during this
-   operation.
-
-3. Hash table walking during look-up may move to a different bucket as
-   the current dentry is moved to a different bucket due to rename.
-   But we use hlists in dcache hash table and they are
-   null-terminated.  So, even if a dentry moves to a different bucket,
-   hash chain walk will terminate. [with a list_head list, it may not
-   since termination is when the list_head in the original bucket is
-   reached].  Since we redo the d_parent check and compare name while
-   holding d_lock, lock-free look-up will not race against d_move().
-
-4. There can be a theoretical race when a dentry keeps coming back to
-   original bucket due to double moves. Due to this look-up may
-   consider that it has never moved and can end up in a infinite loop.
-   But this is not any worse that theoretical livelocks we already
-   have in the kernel.
-
-
-Important guidelines for filesystem developers related to dcache_rcu
-====================================================================
-
-1. Existing dcache interfaces (pre-2.5.62) exported to filesystem
-   don't change. Only dcache internal implementation changes. However
-   filesystems *must not* delete from the dentry hash chains directly
-   using the list macros like allowed earlier. They must use dcache
-   APIs like d_drop() or __d_drop() depending on the situation.
-
-2. d_flags is now protected by a per-dentry lock (d_lock). All access
-   to d_flags must be protected by it.
-
-3. For a hashed dentry, checking of d_count needs to be protected by
-   d_lock.
-
-
-Papers and other documentation on dcache locking
-================================================
-
-1. Scaling dcache with RCU (http://linuxjournal.com/article.php?sid=7124).
-
-2. http://lse.sourceforge.net/locking/dcache/dcache.html
-
-
-
diff --git a/Documentation/filesystems/path-lookup.txt b/Documentation/filesystems/path-lookup.txt
new file mode 100644
index 0000000..09b2878
--- /dev/null
+++ b/Documentation/filesystems/path-lookup.txt
@@ -0,0 +1,345 @@
+Path walking and name lookup locking
+====================================
+
+Path resolution is the finding a dentry corresponding to a path name string, by
+performing a path walk. Typically, for every open(), stat() etc., the path name
+will be resolved. Paths are resolved by walking the namespace tree, starting
+with the first component of the pathname (eg. root or cwd) with a known dentry,
+then finding the child of that dentry, which is named the next component in the
+path string. Then repeating the lookup from the child dentry and finding its
+child with the next element, and so on.
+
+Since it is a frequent operation for workloads like multiuser environments and
+web servers, it is important to optimize this code.
+
+Path walking synchronisation history:
+Prior to 2.5.10, dcache_lock was acquired in d_lookup (dcache hash lookup) and
+thus in every component during path look-up. Since 2.5.10 onwards, fast-walk
+algorithm changed this by holding the dcache_lock at the beginning and walking
+as many cached path component dentries as possible. This significantly
+decreases the number of acquisition of dcache_lock. However it also increases
+the lock hold time significantly and affects performance in large SMP machines.
+Since 2.5.62 kernel, dcache has been using a new locking model that uses RCU to
+make dcache look-up lock-free.
+
+All the above algorithms required taking a lock and reference count on the
+dentry that was looked up, so that may be used as the basis for walking the
+next path element. This is inefficient and unscalable. It is inefficient
+because of the locks and atomic operations required for every dentry element
+slows things down. It is not scalable because many parallel applications that
+are path-walk intensive tend to do path lookups starting from a common dentry
+(usually, the root "/" or current working directory). So contention on these
+common path elements causes lock and cacheline queueing.
+
+Since 2.6.38, RCU is used to make a significant part of the entire path walk
+(including dcache look-up) completely "store-free" (so, no locks, atomics, or
+even stores into cachelines of common dentries). This is known as "rcu-walk"
+path walking.
+
+Path walking overview
+=====================
+
+A name string specifies a start (root directory, cwd, fd-relative) and a
+sequence of elements (directory entry names), which together refer to a path in
+the namespace. A path is represented as a (dentry, vfsmount) tuple. The name
+elements are sub-strings, seperated by '/'.
+
+Name lookups will want to find a particular path that a name string refers to
+(usually the final element, or parent of final element). This is done by taking
+the path given by the name's starting point (which we know in advance -- eg.
+current->fs->cwd or current->fs->root) as the first parent of the lookup. Then
+iteratively for each subsequent name element, look up the child of the current
+parent with the given name and if it is not the desired entry, make it the
+parent for the next lookup.
+
+A parent, of course, must be a directory, and we must have appropriate
+permissions on the parent inode to be able to walk into it.
+
+Turning the child into a parent for the next lookup requires more checks and
+procedures. Symlinks essentially substitute the symlink name for the target
+name in the name string, and require some recursive path walking.  Mount points
+must be followed into (thus changing the vfsmount that subsequent path elements
+refer to), switching from the mount point path to the root of the particular
+mounted vfsmount. These behaviours are variously modified depending on the
+exact path walking flags.
+
+Path walking then must, broadly, do several particular things:
+- find the start point of the walk;
+- perform permissions and validity checks on inodes;
+- perform dcache hash name lookups on (parent, name element) tuples;
+- traverse mount points;
+- traverse symlinks;
+- lookup and create missing parts of the path on demand.
+
+Safe store-free look-up of dcache hash table
+============================================
+
+Dcache name lookup
+------------------
+In order to lookup a dcache (parent, name) tuple, we take a hash on the tuple
+and use that to select a bucket in the dcache-hash table. The list of entries
+in that bucket is then walked, and we do a full comparison of each entry
+against our (parent, name) tuple.
+
+The hash lists are RCU protected, so list walking is not serialised with
+concurrent updates (insertion, deletion from the hash). This is a standard RCU
+list application with the exception of renames, which will be covered below.
+
+Parent and name members of a dentry, as well as its membership in the dcache
+hash, and its inode are protected by the per-dentry d_lock spinlock. A
+reference is taken on the dentry (while the fields are verified under d_lock),
+and this stabilises its d_inode pointer and actual inode. This gives a stable
+point to perform the next step of our path walk against.
+
+These members are also protected by d_seq seqlock, although this offers
+read-only protection and no durability of results, so care must be taken when
+using d_seq for synchronisation (see seqcount based lookups, below).
+
+Renames
+-------
+Back to the rename case. In usual RCU protected lists, the only operations that
+will happen to an object is insertion, and then eventually removal from the
+list. The object will not be reused until an RCU grace period is complete.
+This ensures the RCU list traversal primitives can run over the object without
+problems (see RCU documentation for how this works).
+
+However when a dentry is renamed, its hash value can change, requiring it to be
+moved to a new hash list. Allocating and inserting a new alias would be
+expensive and also problematic for directory dentries. Latency would be far to
+high to wait for a grace period after removing the dentry and before inserting
+it in the new hash bucket. So what is done is to insert the dentry into the
+new list immediately.
+
+However, when the dentry's list pointers are updated to point to objects in the
+new list before waiting for a grace period, this can result in a concurrent RCU
+lookup of the old list veering off into the new (incorrect) list and missing
+the remaining dentries on the list.
+
+There is no fundamental problem with walking down the wrong list, because the
+dentry comparisons will never match. However it is fatal to miss a matching
+dentry. So a seqlock is used to detect when a rename has occurred, and so the
+lookup can be retried.
+
+         1      2      3
+        +---+  +---+  +---+
+hlist-->| N-+->| N-+->| N-+->
+head <--+-P |<-+-P |<-+-P |
+        +---+  +---+  +---+
+
+Rename of dentry 2 may require it deleted from the above list, and inserted
+into a new list. Deleting 2 gives the following list.
+
+         1             3
+        +---+         +---+     (don't worry, the longer pointers do not
+hlist-->| N-+-------->| N-+->    impose a measurable performance overhead
+head <--+-P |<--------+-P |      on modern CPUs)
+        +---+         +---+
+          ^      2      ^
+          |    +---+    |
+          |    | N-+----+
+          +----+-P |
+               +---+
+
+This is a standard RCU-list deletion, which leaves the deleted object's
+pointers intact, so a concurrent list walker that is currently looking at
+object 2 will correctly continue to object 3 when it is time to traverse the
+next object.
+
+However, when inserting object 2 onto a new list, we end up with this:
+
+         1             3
+        +---+         +---+
+hlist-->| N-+-------->| N-+->
+head <--+-P |<--------+-P |
+        +---+         +---+
+                 2
+               +---+
+               | N-+---->
+          <----+-P |
+               +---+
+
+Because we didn't wait for a grace period, there may be a concurrent lookup
+still at 2. Now when it follows 2's 'next' pointer, it will walk off into
+another list without ever having checked object 3.
+
+A related, but distinctly different, issue is that of rename atomicity versus
+lookup operations. If a file is renamed from 'A' to 'B', a lookup must only
+find either 'A' or 'B'. So if a lookup of 'A' returns NULL, a subsequent lookup
+of 'B' must succeed (note the reverse is not true).
+
+Between deleting the dentry from the old hash list, and inserting it on the new
+hash list, a lookup may find neither 'A' nor 'B' matching the dentry. The same
+rename seqlock is also used to cover this race in much the same way, by
+retrying a negative lookup result if a rename was in progress.
+
+Seqcount based lookups
+----------------------
+In refcount based dcache lookups, d_lock is used to serialise access to
+the dentry, stabilising it while comparing its name and parent and then
+taking a reference count (the reference count then gives a stable place to
+start the next part of the path walk from).
+
+As explained above, we would like to do path walking without taking locks or
+reference counts on intermediate dentries along the path. To do this, a per
+dentry seqlock (d_seq) is used to take a "coherent snapshot" of what the dentry
+looks like (its name, parent, and inode). That snapshot is then used to start
+the next part of the path walk. When loading the coherent snapshot under d_seq,
+care must be taken to load the members up-front, and use those pointers rather
+than reloading from the dentry later on (otherwise we'd have interesting things
+like d_inode going NULL underneath us, if the name was unlinked).
+
+Also important is to avoid performing any destructive operations (pretty much:
+no non-atomic stores to shared data), and to recheck the seqcount when we are
+"done" with the operation. Retry or abort if the seqcount does not match.
+Avoiding destructive or changing operations means we can easily unwind from
+failure.
+
+What this means is that a caller, provided they are holding RCU lock to
+protect the dentry object from disappearing, can perform a seqcount based
+lookup which does not increment the refcount on the dentry or write to
+it in any way. This returned dentry can be used for subsequent operations,
+provided that d_seq is rechecked after that operation is complete.
+
+Inodes are also rcu freed, so the seqcount lookup dentry's inode may also be
+queried for permissions.
+
+With this two parts of the puzzle, we can do path lookups without taking
+locks or refcounts on dentry elements.
+
+RCU-walk path walking design
+============================
+
+Path walking code now has two distinct modes, ref-walk and rcu-walk. ref-walk
+is the traditional[*] way of performing dcache lookups using d_lock to
+serialise concurrent modifications to the dentry and take a reference count on
+it. ref-walk is simple and obvious, and may sleep, take locks, etc while path
+walking is operating on each dentry. rcu-walk uses seqcount based dentry
+lookups, and can perform lookup of intermediate elements without any stores to
+shared data in the dentry or inode. rcu-walk can not be applied to all cases,
+eg. if the filesystem must sleep or perform non trivial operations, rcu-walk
+must be switched to ref-walk mode.
+
+[*] RCU is still used for the dentry hash lookup in ref-walk, but not the full
+    path walk.
+
+Where ref-walk uses a stable, refcounted ``parent'' to walk the remaining
+path string, rcu-walk uses a d_seq protected snapshot. When looking up a
+child of this parent snapshot, we open d_seq critical section on the child
+before closing d_seq critical section on the parent. This gives an interlocking
+ladder of snapshots to walk down.
+
+
+     proc 101
+      /----------------\
+     / comm:    "vi"    \
+    /  fs.root: dentry0  \
+    \  fs.cwd:  dentry2  /
+     \                  /
+      \----------------/
+
+So when vi wants to open("/home/npiggin/test.c", O_RDWR), then it will
+start from current->fs->root, which is a pinned dentry. Alternatively,
+"./test.c" would start from cwd; both names refer to the same path in
+the context of proc101.
+
+     dentry 0
+    +---------------------+   rcu-walk begins here, we note d_seq, check the
+    | name:    "/"        |   inode's permission, and then look up the next
+    | inode:   10         |   path element which is "home"...
+    | children:"home", ...|
+    +---------------------+
+              |
+     dentry 1 V
+    +---------------------+   ... which brings us here. We find dentry1 via
+    | name:    "home"     |   hash lookup, then note d_seq and compare name
+    | inode:   678        |   string and parent pointer. When we have a match,
+    | children:"npiggin"  |   we now recheck the d_seq of dentry0. Then we
+    +---------------------+   check inode and look up the next element.
+              |
+     dentry2  V
+    +---------------------+   Note: if dentry0 is now modified, lookup is
+    | name:    "npiggin"  |   not necessarily invalid, so we need only keep a
+    | inode:   543        |   parent for d_seq verification, and grandparents
+    | children:"a.c", ... |   can be forgotten.
+    +---------------------+
+              |
+     dentry3  V
+    +---------------------+   At this point we have our destination dentry.
+    | name:    "a.c"      |   We now take its d_lock, verify d_seq of this
+    | inode:   14221      |   dentry. If that checks out, we can increment
+    | children:NULL       |   its refcount because we're holding d_lock.
+    +---------------------+
+
+Taking a refcount on a dentry from rcu-walk mode, by taking its d_lock,
+re-checking its d_seq, and then incrementing its refcount is called
+"dropping rcu" or dropping from rcu-walk into ref-walk mode.
+
+It is, in some sense, a bit of a house of cards. If the seqcount check of the
+parent snapshot fails, the house comes down, because we had closed the d_seq
+section on the grandparent, so we have nothing left to stand on. In that case,
+the path walk must be fully restarted (which we do in ref-walk mode, to avoid
+live locks). It is costly to have a full restart, but fortunately they are
+quite rare.
+
+When we reach a point where sleeping is required, or a filesystem callout
+requires ref-walk, then instead of restarting the walk, we attempt to drop rcu
+at the last known good dentry we have. Avoiding a full restart in ref-walk in
+these cases is fundamental for performance and scalability because blocking
+operations such as creates and unlinks are not uncommon.
+
+The detailed design for rcu-walk is like this:
+* LOOKUP_RCU is set in nd->flags, which distinguishes rcu-walk from ref-walk.
+* Take the RCU lock for the entire path walk, starting with the acquiring
+  of the starting path (eg. root/cwd/fd-path). So now dentry refcounts are
+  not required for dentry persistence.
+* synchronize_rcu is called when unregistering a filesystem, so we can
+  access d_ops and i_ops during rcu-walk.
+* Similarly take the vfsmount lock for the entire path walk. So now mnt
+  refcounts are not required for persistence. Also we are free to perform mount
+  lookups, and to assume dentry mount points and mount roots are stable up and
+  down the path.
+* Have a per-dentry seqlock to protect the dentry name, parent, and inode,
+  so we can load this tuple atomically, and also check whether any of its
+  members have changed.
+* Dentry lookups (based on parent, candidate string tuple) recheck the parent
+  sequence after the child is found in case anything changed in the parent
+  during the path walk.
+* inode is also RCU protected so we can load d_inode and use the inode for
+  limited things.
+* i_mode, i_uid, i_gid can be tested for exec permissions during path walk.
+* i_op can be loaded.
+* When the destination dentry is reached, drop rcu there (ie. take d_lock,
+  verify d_seq, increment refcount).
+* If seqlock verification fails anywhere along the path, do a full restart
+  of the path lookup in ref-walk mode. -ECHILD tends to be used (for want of
+  a better errno) to signal an rcu-walk failure.
+
+The cases where rcu-walk cannot continue are:
+* NULL dentry (ie. any uncached path element)
+* parent with d_inode->i_op->permission or ACLs
+* dentries with d_revalidate
+* Following links
+
+In future patches, permission checks and d_revalidate become rcu-walk aware. It
+may be possible eventually to make following links rcu-walk aware.
+
+Uncached path elements will always require dropping to ref-walk mode, at the
+very least because i_mutex needs to be grabbed, and objects allocated.
+
+Final note:
+"store-free" path walking is not strictly store free. We take vfsmount lock
+and refcounts (both of which can be made per-cpu), and we also store to the
+stack (which is essentially CPU-local), and we also have to take locks and
+refcount on final dentry.
+
+The point is that shared data, where practically possible, is not locked
+or stored into. The result is massive improvements in performance and
+scalability of path resolution.
+
+
+Papers and other documentation on dcache locking
+================================================
+
+1. Scaling dcache with RCU (http://linuxjournal.com/article.php?sid=7124).
+
+2. http://lse.sourceforge.net/locking/dcache/dcache.html
diff --git a/fs/dcache.c b/fs/dcache.c
index dc0551c..187fea0 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -152,9 +152,23 @@ static void d_free(struct dentry *dentry)
 		call_rcu(&dentry->d_u.d_rcu, __d_free);
 }
 
+/**
+ * dentry_rcuwalk_barrier - invalidate in-progress rcu-walk lookups
+ * After this call, in-progress rcu-walk path lookup will fail. This
+ * should be called after unhashing, and after changing d_inode (if
+ * the dentry has not already been unhashed).
+ */
+static inline void dentry_rcuwalk_barrier(struct dentry *dentry)
+{
+	assert_spin_locked(&dentry->d_lock);
+	/* Go through a barrier */
+	write_seqcount_barrier(&dentry->d_seq);
+}
+
 /*
  * Release the dentry's inode, using the filesystem
- * d_iput() operation if defined.
+ * d_iput() operation if defined. Dentry has no refcount
+ * and is unhashed.
  */
 static void dentry_iput(struct dentry * dentry)
 	__releases(dentry->d_lock)
@@ -179,6 +193,28 @@ static void dentry_iput(struct dentry * dentry)
 }
 
 /*
+ * Release the dentry's inode, using the filesystem
+ * d_iput() operation if defined. dentry remains in-use.
+ */
+static void dentry_unlink_inode(struct dentry * dentry)
+	__releases(dentry->d_lock)
+	__releases(dcache_inode_lock)
+{
+	struct inode *inode = dentry->d_inode;
+	dentry->d_inode = NULL;
+	list_del_init(&dentry->d_alias);
+	dentry_rcuwalk_barrier(dentry);
+	spin_unlock(&dentry->d_lock);
+	spin_unlock(&dcache_inode_lock);
+	if (!inode->i_nlink)
+		fsnotify_inoderemove(inode);
+	if (dentry->d_op && dentry->d_op->d_iput)
+		dentry->d_op->d_iput(dentry, inode);
+	else
+		iput(inode);
+}
+
+/*
  * dentry_lru_(add|del|move_tail) must be called with d_lock held.
  */
 static void dentry_lru_add(struct dentry *dentry)
@@ -272,6 +308,7 @@ void __d_drop(struct dentry *dentry)
 		spin_lock(&dcache_hash_lock);
 		hlist_del_rcu(&dentry->d_hash);
 		spin_unlock(&dcache_hash_lock);
+		dentry_rcuwalk_barrier(dentry);
 	}
 }
 EXPORT_SYMBOL(__d_drop);
@@ -309,6 +346,7 @@ relock:
 		spin_unlock(&dcache_inode_lock);
 		goto relock;
 	}
+
 	if (ref)
 		dentry->d_count--;
 	/* if dentry was on the d_lru list delete it from there */
@@ -1221,6 +1259,7 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	dentry->d_count = 1;
 	dentry->d_flags = DCACHE_UNHASHED;
 	spin_lock_init(&dentry->d_lock);
+	seqcount_init(&dentry->d_seq);
 	dentry->d_inode = NULL;
 	dentry->d_parent = NULL;
 	dentry->d_sb = NULL;
@@ -1269,6 +1308,7 @@ static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 	if (inode)
 		list_add(&dentry->d_alias, &inode->i_dentry);
 	dentry->d_inode = inode;
+	dentry_rcuwalk_barrier(dentry);
 	spin_unlock(&dentry->d_lock);
 	fsnotify_d_instantiate(dentry, inode);
 }
@@ -1611,6 +1651,111 @@ err_out:
 EXPORT_SYMBOL(d_add_ci);
 
 /**
+ * __d_lookup_rcu - search for a dentry (racy, store-free)
+ * @parent: parent dentry
+ * @name: qstr of name we wish to find
+ * @seq: returns d_seq value at the point where the dentry was found
+ * @inode: returns dentry->d_inode when the inode was found valid.
+ * Returns: dentry, or NULL
+ *
+ * __d_lookup_rcu is the dcache lookup function for rcu-walk name
+ * resolution (store-free path walking) design described in
+ * Documentation/filesystems/path-lookup.txt.
+ *
+ * This is not to be used outside core vfs.
+ *
+ * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
+ * held, and rcu_read_lock held. The returned dentry must not be stored into
+ * without taking d_lock and checking d_seq sequence count against @seq
+ * returned here.
+ *
+ * A refcount may be taken on the found dentry with the __d_rcu_to_refcount
+ * function.
+ *
+ * Alternatively, __d_lookup_rcu may be called again to look up the child of
+ * the returned dentry, so long as its parent's seqlock is checked after the
+ * child is looked up. Thus, an interlocking stepping of sequence lock checks
+ * is formed, giving integrity down the path walk.
+ */
+struct dentry *__d_lookup_rcu(struct dentry *parent, struct qstr *name,
+				unsigned *seq, struct inode **inode)
+{
+	unsigned int len = name->len;
+	unsigned int hash = name->hash;
+	const unsigned char *str = name->name;
+	struct hlist_head *head = d_hash(parent, hash);
+	struct hlist_node *node;
+	struct dentry *dentry;
+
+	/*
+	 * Note: There is significant duplication with __d_lookup_rcu which is
+	 * required to prevent single threaded performance regressions
+	 * especially on architectures where smp_rmb (in seqcounts) are costly.
+	 * Keep the two functions in sync.
+	 */
+
+	/*
+	 * The hash list is protected using RCU.
+	 *
+	 * Carefully use d_seq when comparing a candidate dentry, to avoid
+	 * races with d_move().
+	 *
+	 * It is possible that concurrent renames can mess up our list
+	 * walk here and result in missing our dentry, resulting in the
+	 * false-negative result. d_lookup() protects against concurrent
+	 * renames using rename_lock seqlock.
+	 *
+	 * See Documentation/vfs/dcache-locking.txt for more details.
+	 */
+	hlist_for_each_entry_rcu(dentry, node, head, d_hash) {
+		struct inode *i;
+		const char *tname;
+		int tlen;
+
+		if (dentry->d_name.hash != hash)
+			continue;
+
+seqretry:
+		*seq = read_seqcount_begin(&dentry->d_seq);
+		if (dentry->d_parent != parent)
+			continue;
+		if (d_unhashed(dentry))
+			continue;
+		tlen = dentry->d_name.len;
+		tname = dentry->d_name.name;
+		i = dentry->d_inode;
+		/*
+		 * This seqcount check is required to ensure name and
+		 * len are loaded atomically, so as not to walk off the
+		 * edge of memory when walking. If we could load this
+		 * atomically some other way, we could drop this check.
+		 */
+		if (read_seqcount_retry(&dentry->d_seq, *seq))
+			goto seqretry;
+		if (parent->d_op && parent->d_op->d_compare) {
+			if (parent->d_op->d_compare(parent, *inode,
+						dentry, i,
+						tlen, tname, name))
+				continue;
+		} else {
+			if (tlen != len)
+				continue;
+			if (memcmp(tname, str, tlen))
+				continue;
+		}
+		/*
+		 * No extra seqcount check is required after the name
+		 * compare. The caller must perform a seqcount check in
+		 * order to do anything useful with the returned dentry
+		 * anyway.
+		 */
+		*inode = i;
+		return dentry;
+	}
+	return NULL;
+}
+
+/**
  * d_lookup - search for a dentry
  * @parent: parent dentry
  * @name: qstr of name we wish to find
@@ -1621,9 +1766,9 @@ EXPORT_SYMBOL(d_add_ci);
  * dentry is returned. The caller must use dput to free the entry when it has
  * finished using it. %NULL is returned if the dentry does not exist.
  */
-struct dentry * d_lookup(struct dentry * parent, struct qstr * name)
+struct dentry *d_lookup(struct dentry *parent, struct qstr *name)
 {
-	struct dentry * dentry = NULL;
+	struct dentry *dentry;
 	unsigned seq;
 
         do {
@@ -1636,7 +1781,7 @@ struct dentry * d_lookup(struct dentry * parent, struct qstr * name)
 }
 EXPORT_SYMBOL(d_lookup);
 
-/*
+/**
  * __d_lookup - search for a dentry (racy)
  * @parent: parent dentry
  * @name: qstr of name we wish to find
@@ -1651,17 +1796,24 @@ EXPORT_SYMBOL(d_lookup);
  *
  * __d_lookup callers must be commented.
  */
-struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
+struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
 {
 	unsigned int len = name->len;
 	unsigned int hash = name->hash;
 	const unsigned char *str = name->name;
 	struct hlist_head *head = d_hash(parent,hash);
-	struct dentry *found = NULL;
 	struct hlist_node *node;
+	struct dentry *found = NULL;
 	struct dentry *dentry;
 
 	/*
+	 * Note: There is significant duplication with __d_lookup_rcu which is
+	 * required to prevent single threaded performance regressions
+	 * especially on architectures where smp_rmb (in seqcounts) are costly.
+	 * Keep the two functions in sync.
+	 */
+
+	/*
 	 * The hash list is protected using RCU.
 	 *
 	 * Take d_lock when comparing a candidate dentry, to avoid races
@@ -1677,24 +1829,15 @@ struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
 	rcu_read_lock();
 	
 	hlist_for_each_entry_rcu(dentry, node, head, d_hash) {
-		struct qstr *qstr;
+		const char *tname;
+		int tlen;
 
 		if (dentry->d_name.hash != hash)
 			continue;
-		if (dentry->d_parent != parent)
-			continue;
 
 		spin_lock(&dentry->d_lock);
-
-		/*
-		 * Recheck the dentry after taking the lock - d_move may have
-		 * changed things. Don't bother checking the hash because
-		 * we're about to compare the whole name anyway.
-		 */
 		if (dentry->d_parent != parent)
 			goto next;
-
-		/* non-existing due to RCU? */
 		if (d_unhashed(dentry))
 			goto next;
 
@@ -1702,16 +1845,17 @@ struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
 		 * It is safe to compare names since d_move() cannot
 		 * change the qstr (protected by d_lock).
 		 */
-		qstr = &dentry->d_name;
+		tlen = dentry->d_name.len;
+		tname = dentry->d_name.name;
 		if (parent->d_op && parent->d_op->d_compare) {
 			if (parent->d_op->d_compare(parent, parent->d_inode,
 						dentry, dentry->d_inode,
-						qstr->len, qstr->name, name))
+						tlen, tname, name))
 				goto next;
 		} else {
-			if (qstr->len != len)
+			if (tlen != len)
 				goto next;
-			if (memcmp(qstr->name, str, len))
+			if (memcmp(tname, str, tlen))
 				goto next;
 		}
 
@@ -1821,7 +1965,7 @@ again:
 			goto again;
 		}
 		dentry->d_flags &= ~DCACHE_CANT_MOUNT;
-		dentry_iput(dentry);
+		dentry_unlink_inode(dentry);
 		fsnotify_nameremove(dentry, isdir);
 		return;
 	}
@@ -1884,7 +2028,9 @@ void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
 	BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */
 
 	spin_lock(&dentry->d_lock);
+	write_seqcount_begin(&dentry->d_seq);
 	memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
+	write_seqcount_end(&dentry->d_seq);
 	spin_unlock(&dentry->d_lock);
 }
 EXPORT_SYMBOL(dentry_update_name_case);
@@ -1997,6 +2143,9 @@ void d_move(struct dentry * dentry, struct dentry * target)
 
 	dentry_lock_for_move(dentry, target);
 
+	write_seqcount_begin(&dentry->d_seq);
+	write_seqcount_begin(&target->d_seq);
+
 	/* Move the dentry to the target hash queue, if on different bucket */
 	spin_lock(&dcache_hash_lock);
 	if (!d_unhashed(dentry))
@@ -2005,6 +2154,7 @@ void d_move(struct dentry * dentry, struct dentry * target)
 	spin_unlock(&dcache_hash_lock);
 
 	/* Unhash the target: dput() will then get rid of it */
+	/* __d_drop does write_seqcount_barrier, but they're OK to nest. */
 	__d_drop(target);
 
 	list_del(&dentry->d_u.d_child);
@@ -2028,6 +2178,9 @@ void d_move(struct dentry * dentry, struct dentry * target)
 
 	list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
 
+	write_seqcount_end(&target->d_seq);
+	write_seqcount_end(&dentry->d_seq);
+
 	dentry_unlock_parents_for_move(dentry, target);
 	spin_unlock(&target->d_lock);
 	fsnotify_d_move(dentry);
@@ -2110,6 +2263,9 @@ static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
 
 	dentry_lock_for_move(anon, dentry);
 
+	write_seqcount_begin(&dentry->d_seq);
+	write_seqcount_begin(&anon->d_seq);
+
 	dparent = dentry->d_parent;
 	aparent = anon->d_parent;
 
@@ -2130,6 +2286,9 @@ static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
 	else
 		INIT_LIST_HEAD(&anon->d_u.d_child);
 
+	write_seqcount_end(&dentry->d_seq);
+	write_seqcount_end(&anon->d_seq);
+
 	dentry_unlock_parents_for_move(anon, dentry);
 	spin_unlock(&dentry->d_lock);
 
diff --git a/fs/filesystems.c b/fs/filesystems.c
index 68ba492..751d6b2 100644
--- a/fs/filesystems.c
+++ b/fs/filesystems.c
@@ -115,6 +115,9 @@ int unregister_filesystem(struct file_system_type * fs)
 		tmp = &(*tmp)->next;
 	}
 	write_unlock(&file_systems_lock);
+
+	synchronize_rcu();
+
 	return -EINVAL;
 }
 
diff --git a/fs/namei.c b/fs/namei.c
index 5642bc2..8d3f15b 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -169,8 +169,8 @@ EXPORT_SYMBOL(putname);
 /*
  * This does basic POSIX ACL permission checking
  */
-static int acl_permission_check(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask))
+static inline int __acl_permission_check(struct inode *inode, int mask,
+		int (*check_acl)(struct inode *inode, int mask), int rcu)
 {
 	umode_t			mode = inode->i_mode;
 
@@ -180,9 +180,13 @@ static int acl_permission_check(struct inode *inode, int mask,
 		mode >>= 6;
 	else {
 		if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) {
-			int error = check_acl(inode, mask);
-			if (error != -EAGAIN)
-				return error;
+			if (rcu) {
+				return -ECHILD;
+			} else {
+				int error = check_acl(inode, mask);
+				if (error != -EAGAIN)
+					return error;
+			}
 		}
 
 		if (in_group_p(inode->i_gid))
@@ -197,6 +201,12 @@ static int acl_permission_check(struct inode *inode, int mask,
 	return -EACCES;
 }
 
+static inline int acl_permission_check(struct inode *inode, int mask,
+		int (*check_acl)(struct inode *inode, int mask))
+{
+	return __acl_permission_check(inode, mask, check_acl, 0);
+}
+
 /**
  * generic_permission  -  check for access rights on a Posix-like filesystem
  * @inode:	inode to check access rights for
@@ -375,6 +385,173 @@ void path_put(struct path *path)
 EXPORT_SYMBOL(path_put);
 
 /**
+ * nameidata_drop_rcu - drop this nameidata out of rcu-walk
+ * @nd: nameidata pathwalk data to drop
+ * @Returns: 0 on success, -ECHLID on failure
+ *
+ * Path walking has 2 modes, rcu-walk and ref-walk (see
+ * Documentation/filesystems/path-lookup.txt). __drop_rcu* functions attempt
+ * to drop out of rcu-walk mode and take normal reference counts on dentries
+ * and vfsmounts to transition to rcu-walk mode. __drop_rcu* functions take
+ * refcounts at the last known good point before rcu-walk got stuck, so
+ * ref-walk may continue from there. If this is not successful (eg. a seqcount
+ * has changed), then failure is returned and path walk restarts from the
+ * beginning in ref-walk mode.
+ *
+ * nameidata_drop_rcu attempts to drop the current nd->path and nd->root into
+ * ref-walk. Must be called from rcu-walk context.
+ */
+static int nameidata_drop_rcu(struct nameidata *nd)
+{
+	struct fs_struct *fs = current->fs;
+	struct dentry *dentry = nd->path.dentry;
+
+	BUG_ON(!(nd->flags & LOOKUP_RCU));
+	if (nd->root.mnt) {
+		spin_lock(&fs->lock);
+		if (nd->root.mnt != fs->root.mnt ||
+				nd->root.dentry != fs->root.dentry)
+			goto err_root;
+	}
+	spin_lock(&dentry->d_lock);
+	if (!__d_rcu_to_refcount(dentry, nd->seq))
+		goto err;
+	BUG_ON(nd->inode != dentry->d_inode);
+	spin_unlock(&dentry->d_lock);
+	if (nd->root.mnt) {
+		path_get(&nd->root);
+		spin_unlock(&fs->lock);
+	}
+	mntget(nd->path.mnt);
+
+	rcu_read_unlock();
+	br_read_unlock(vfsmount_lock);
+	nd->flags &= ~LOOKUP_RCU;
+	return 0;
+err:
+	spin_unlock(&dentry->d_lock);
+err_root:
+	if (nd->root.mnt)
+		spin_unlock(&fs->lock);
+	return -ECHILD;
+}
+
+/* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing.  */
+static inline int nameidata_drop_rcu_maybe(struct nameidata *nd)
+{
+	if (nd->flags & LOOKUP_RCU)
+		return nameidata_drop_rcu(nd);
+	return 0;
+}
+
+/**
+ * nameidata_dentry_drop_rcu - drop nameidata and dentry out of rcu-walk
+ * @nd: nameidata pathwalk data to drop
+ * @dentry: dentry to drop
+ * @Returns: 0 on success, -ECHLID on failure
+ *
+ * nameidata_dentry_drop_rcu attempts to drop the current nd->path and nd->root,
+ * and dentry into ref-walk. @dentry must be a path found by a do_lookup call on
+ * @nd. Must be called from rcu-walk context.
+ */
+static int nameidata_dentry_drop_rcu(struct nameidata *nd, struct dentry *dentry)
+{
+	struct fs_struct *fs = current->fs;
+	struct dentry *parent = nd->path.dentry;
+
+	BUG_ON(!(nd->flags & LOOKUP_RCU));
+	if (nd->root.mnt) {
+		spin_lock(&fs->lock);
+		if (nd->root.mnt != fs->root.mnt ||
+				nd->root.dentry != fs->root.dentry)
+			goto err_root;
+	}
+	spin_lock(&parent->d_lock);
+	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+	if (!__d_rcu_to_refcount(dentry, nd->seq))
+		goto err;
+	/*
+	 * If the sequence check on the child dentry passed, then the child has
+	 * not been removed from its parent. This means the parent dentry must
+	 * be valid and able to take a reference at this point.
+	 */
+	BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent);
+	BUG_ON(!parent->d_count);
+	parent->d_count++;
+	spin_unlock(&dentry->d_lock);
+	spin_unlock(&parent->d_lock);
+	if (nd->root.mnt) {
+		path_get(&nd->root);
+		spin_unlock(&fs->lock);
+	}
+	mntget(nd->path.mnt);
+
+	rcu_read_unlock();
+	br_read_unlock(vfsmount_lock);
+	nd->flags &= ~LOOKUP_RCU;
+	return 0;
+err:
+	spin_unlock(&dentry->d_lock);
+	spin_unlock(&parent->d_lock);
+err_root:
+	if (nd->root.mnt)
+		spin_unlock(&fs->lock);
+	return -ECHILD;
+}
+
+/* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing.  */
+static inline int nameidata_dentry_drop_rcu_maybe(struct nameidata *nd, struct dentry *dentry)
+{
+	if (nd->flags & LOOKUP_RCU)
+		return nameidata_dentry_drop_rcu(nd, dentry);
+	return 0;
+}
+
+/**
+ * nameidata_drop_rcu_last - drop nameidata ending path walk out of rcu-walk
+ * @nd: nameidata pathwalk data to drop
+ * @Returns: 0 on success, -ECHLID on failure
+ *
+ * nameidata_drop_rcu_last attempts to drop the current nd->path into ref-walk.
+ * nd->path should be the final element of the lookup, so nd->root is discarded.
+ * Must be called from rcu-walk context.
+ */
+static int nameidata_drop_rcu_last(struct nameidata *nd)
+{
+	struct dentry *dentry = nd->path.dentry;
+
+	BUG_ON(!(nd->flags & LOOKUP_RCU));
+	nd->flags &= ~LOOKUP_RCU;
+	nd->root.mnt = NULL;
+	spin_lock(&dentry->d_lock);
+	if (!__d_rcu_to_refcount(dentry, nd->seq))
+		goto err_unlock;
+	BUG_ON(nd->inode != dentry->d_inode);
+	spin_unlock(&dentry->d_lock);
+
+	mntget(nd->path.mnt);
+
+	rcu_read_unlock();
+	br_read_unlock(vfsmount_lock);
+
+	return 0;
+
+err_unlock:
+	spin_unlock(&dentry->d_lock);
+	rcu_read_unlock();
+	br_read_unlock(vfsmount_lock);
+	return -ECHILD;
+}
+
+/* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing.  */
+static inline int nameidata_drop_rcu_last_maybe(struct nameidata *nd)
+{
+	if (likely(nd->flags & LOOKUP_RCU))
+		return nameidata_drop_rcu_last(nd);
+	return 0;
+}
+
+/**
  * release_open_intent - free up open intent resources
  * @nd: pointer to nameidata
  */
@@ -459,26 +636,40 @@ force_reval_path(struct path *path, struct nameidata *nd)
  * short-cut DAC fails, then call ->permission() to do more
  * complete permission check.
  */
-static int exec_permission(struct inode *inode)
+static inline int __exec_permission(struct inode *inode, int rcu)
 {
 	int ret;
 
 	if (inode->i_op->permission) {
+		if (rcu)
+			return -ECHILD;
 		ret = inode->i_op->permission(inode, MAY_EXEC);
 		if (!ret)
 			goto ok;
 		return ret;
 	}
-	ret = acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl);
+	ret = __acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl, rcu);
 	if (!ret)
 		goto ok;
+	if (rcu && ret == -ECHILD)
+		return ret;
 
 	if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH))
 		goto ok;
 
 	return ret;
 ok:
-	return security_inode_permission(inode, MAY_EXEC);
+	return security_inode_exec_permission(inode, rcu);
+}
+
+static int exec_permission(struct inode *inode)
+{
+	return __exec_permission(inode, 0);
+}
+
+static int exec_permission_rcu(struct inode *inode)
+{
+	return __exec_permission(inode, 1);
 }
 
 static __always_inline void set_root(struct nameidata *nd)
@@ -489,8 +680,20 @@ static __always_inline void set_root(struct nameidata *nd)
 
 static int link_path_walk(const char *, struct nameidata *);
 
+static __always_inline void set_root_rcu(struct nameidata *nd)
+{
+	if (!nd->root.mnt) {
+		struct fs_struct *fs = current->fs;
+		spin_lock(&fs->lock);
+		nd->root = fs->root;
+		spin_unlock(&fs->lock);
+	}
+}
+
 static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link)
 {
+	int ret;
+
 	if (IS_ERR(link))
 		goto fail;
 
@@ -500,8 +703,10 @@ static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *l
 		nd->path = nd->root;
 		path_get(&nd->root);
 	}
+	nd->inode = nd->path.dentry->d_inode;
 
-	return link_path_walk(link, nd);
+	ret = link_path_walk(link, nd);
+	return ret;
 fail:
 	path_put(&nd->path);
 	return PTR_ERR(link);
@@ -516,11 +721,12 @@ static void path_put_conditional(struct path *path, struct nameidata *nd)
 
 static inline void path_to_nameidata(struct path *path, struct nameidata *nd)
 {
-	dput(nd->path.dentry);
-	if (nd->path.mnt != path->mnt) {
-		mntput(nd->path.mnt);
-		nd->path.mnt = path->mnt;
+	if (!(nd->flags & LOOKUP_RCU)) {
+		dput(nd->path.dentry);
+		if (nd->path.mnt != path->mnt)
+			mntput(nd->path.mnt);
 	}
+	nd->path.mnt = path->mnt;
 	nd->path.dentry = path->dentry;
 }
 
@@ -535,9 +741,11 @@ __do_follow_link(struct path *path, struct nameidata *nd, void **p)
 
 	if (path->mnt != nd->path.mnt) {
 		path_to_nameidata(path, nd);
+		nd->inode = nd->path.dentry->d_inode;
 		dget(dentry);
 	}
 	mntget(path->mnt);
+
 	nd->last_type = LAST_BIND;
 	*p = dentry->d_inode->i_op->follow_link(dentry, nd);
 	error = PTR_ERR(*p);
@@ -591,6 +799,20 @@ loop:
 	return err;
 }
 
+static int follow_up_rcu(struct path *path)
+{
+	struct vfsmount *parent;
+	struct dentry *mountpoint;
+
+	parent = path->mnt->mnt_parent;
+	if (parent == path->mnt)
+		return 0;
+	mountpoint = path->mnt->mnt_mountpoint;
+	path->dentry = mountpoint;
+	path->mnt = parent;
+	return 1;
+}
+
 int follow_up(struct path *path)
 {
 	struct vfsmount *parent;
@@ -615,6 +837,21 @@ int follow_up(struct path *path)
 /*
  * serialization is taken care of in namespace.c
  */
+static void __follow_mount_rcu(struct nameidata *nd, struct path *path,
+				struct inode **inode)
+{
+	while (d_mountpoint(path->dentry)) {
+		struct vfsmount *mounted;
+		mounted = __lookup_mnt(path->mnt, path->dentry, 1);
+		if (!mounted)
+			return;
+		path->mnt = mounted;
+		path->dentry = mounted->mnt_root;
+		nd->seq = read_seqcount_begin(&path->dentry->d_seq);
+		*inode = path->dentry->d_inode;
+	}
+}
+
 static int __follow_mount(struct path *path)
 {
 	int res = 0;
@@ -660,7 +897,42 @@ int follow_down(struct path *path)
 	return 0;
 }
 
-static __always_inline void follow_dotdot(struct nameidata *nd)
+static int follow_dotdot_rcu(struct nameidata *nd)
+{
+	struct inode *inode = nd->inode;
+
+	set_root_rcu(nd);
+
+	while(1) {
+		if (nd->path.dentry == nd->root.dentry &&
+		    nd->path.mnt == nd->root.mnt) {
+			break;
+		}
+		if (nd->path.dentry != nd->path.mnt->mnt_root) {
+			struct dentry *old = nd->path.dentry;
+			struct dentry *parent = old->d_parent;
+			unsigned seq;
+
+			seq = read_seqcount_begin(&parent->d_seq);
+			if (read_seqcount_retry(&old->d_seq, nd->seq))
+				return -ECHILD;
+			inode = parent->d_inode;
+			nd->path.dentry = parent;
+			nd->seq = seq;
+			break;
+		}
+		if (!follow_up_rcu(&nd->path))
+			break;
+		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
+		inode = nd->path.dentry->d_inode;
+	}
+	__follow_mount_rcu(nd, &nd->path, &inode);
+	nd->inode = inode;
+
+	return 0;
+}
+
+static void follow_dotdot(struct nameidata *nd)
 {
 	set_root(nd);
 
@@ -681,6 +953,7 @@ static __always_inline void follow_dotdot(struct nameidata *nd)
 			break;
 	}
 	follow_mount(&nd->path);
+	nd->inode = nd->path.dentry->d_inode;
 }
 
 /*
@@ -718,18 +991,17 @@ static struct dentry *d_alloc_and_lookup(struct dentry *parent,
  *  It _is_ time-critical.
  */
 static int do_lookup(struct nameidata *nd, struct qstr *name,
-		     struct path *path)
+			struct path *path, struct inode **inode)
 {
 	struct vfsmount *mnt = nd->path.mnt;
-	struct dentry *dentry, *parent;
+	struct dentry *dentry, *parent = nd->path.dentry;
 	struct inode *dir;
 	/*
 	 * See if the low-level filesystem might want
 	 * to use its own hash..
 	 */
-	if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) {
-		int err = nd->path.dentry->d_op->d_hash(nd->path.dentry,
-				nd->path.dentry->d_inode, name);
+	if (parent->d_op && parent->d_op->d_hash) {
+		int err = parent->d_op->d_hash(parent, nd->inode, name);
 		if (err < 0)
 			return err;
 	}
@@ -739,21 +1011,48 @@ static int do_lookup(struct nameidata *nd, struct qstr *name,
 	 * of a false negative due to a concurrent rename, we're going to
 	 * do the non-racy lookup, below.
 	 */
-	dentry = __d_lookup(nd->path.dentry, name);
-	if (!dentry)
-		goto need_lookup;
+	if (nd->flags & LOOKUP_RCU) {
+		unsigned seq;
+
+		*inode = nd->inode;
+		dentry = __d_lookup_rcu(parent, name, &seq, inode);
+		if (!dentry) {
+			if (nameidata_drop_rcu(nd))
+				return -ECHILD;
+			goto need_lookup;
+		}
+		/* Memory barrier in read_seqcount_begin of child is enough */
+		if (__read_seqcount_retry(&parent->d_seq, nd->seq))
+			return -ECHILD;
+
+		nd->seq = seq;
+		if (dentry->d_op && dentry->d_op->d_revalidate) {
+			/* We commonly drop rcu-walk here */
+			if (nameidata_dentry_drop_rcu(nd, dentry))
+				return -ECHILD;
+			goto need_revalidate;
+		}
+		path->mnt = mnt;
+		path->dentry = dentry;
+		__follow_mount_rcu(nd, path, inode);
+	} else {
+		dentry = __d_lookup(parent, name);
+		if (!dentry)
+			goto need_lookup;
 found:
-	if (dentry->d_op && dentry->d_op->d_revalidate)
-		goto need_revalidate;
+		if (dentry->d_op && dentry->d_op->d_revalidate)
+			goto need_revalidate;
 done:
-	path->mnt = mnt;
-	path->dentry = dentry;
-	__follow_mount(path);
+		path->mnt = mnt;
+		path->dentry = dentry;
+		__follow_mount(path);
+		*inode = path->dentry->d_inode;
+	}
 	return 0;
 
 need_lookup:
-	parent = nd->path.dentry;
 	dir = parent->d_inode;
+	BUG_ON(nd->inode != dir);
 
 	mutex_lock(&dir->i_mutex);
 	/*
@@ -815,7 +1114,6 @@ static inline int follow_on_final(struct inode *inode, unsigned lookup_flags)
 static int link_path_walk(const char *name, struct nameidata *nd)
 {
 	struct path next;
-	struct inode *inode;
 	int err;
 	unsigned int lookup_flags = nd->flags;
 	
@@ -824,18 +1122,28 @@ static int link_path_walk(const char *name, struct nameidata *nd)
 	if (!*name)
 		goto return_reval;
 
-	inode = nd->path.dentry->d_inode;
 	if (nd->depth)
 		lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE);
 
 	/* At this point we know we have a real path component. */
 	for(;;) {
+		struct inode *inode;
 		unsigned long hash;
 		struct qstr this;
 		unsigned int c;
 
 		nd->flags |= LOOKUP_CONTINUE;
-		err = exec_permission(inode);
+		if (nd->flags & LOOKUP_RCU) {
+			err = exec_permission_rcu(nd->inode);
+			if (err == -ECHILD) {
+				if (nameidata_drop_rcu(nd))
+					return -ECHILD;
+				goto exec_again;
+			}
+		} else {
+exec_again:
+			err = exec_permission(nd->inode);
+		}
  		if (err)
 			break;
 
@@ -866,37 +1174,44 @@ static int link_path_walk(const char *name, struct nameidata *nd)
 		if (this.name[0] == '.') switch (this.len) {
 			default:
 				break;
-			case 2:	
+			case 2:
 				if (this.name[1] != '.')
 					break;
-				follow_dotdot(nd);
-				inode = nd->path.dentry->d_inode;
+				if (nd->flags & LOOKUP_RCU) {
+					if (follow_dotdot_rcu(nd))
+						return -ECHILD;
+				} else
+					follow_dotdot(nd);
 				/* fallthrough */
 			case 1:
 				continue;
 		}
 		/* This does the actual lookups.. */
-		err = do_lookup(nd, &this, &next);
+		err = do_lookup(nd, &this, &next, &inode);
 		if (err)
 			break;
-
 		err = -ENOENT;
-		inode = next.dentry->d_inode;
 		if (!inode)
 			goto out_dput;
 
 		if (inode->i_op->follow_link) {
+			/* We commonly drop rcu-walk here */
+			if (nameidata_dentry_drop_rcu_maybe(nd, next.dentry))
+				return -ECHILD;
+			BUG_ON(inode != next.dentry->d_inode);
 			err = do_follow_link(&next, nd);
 			if (err)
 				goto return_err;
+			nd->inode = nd->path.dentry->d_inode;
 			err = -ENOENT;
-			inode = nd->path.dentry->d_inode;
-			if (!inode)
+			if (!nd->inode)
 				break;
-		} else
+		} else {
 			path_to_nameidata(&next, nd);
+			nd->inode = inode;
+		}
 		err = -ENOTDIR; 
-		if (!inode->i_op->lookup)
+		if (!nd->inode->i_op->lookup)
 			break;
 		continue;
 		/* here ends the main loop */
@@ -911,32 +1226,39 @@ last_component:
 		if (this.name[0] == '.') switch (this.len) {
 			default:
 				break;
-			case 2:	
+			case 2:
 				if (this.name[1] != '.')
 					break;
-				follow_dotdot(nd);
-				inode = nd->path.dentry->d_inode;
+				if (nd->flags & LOOKUP_RCU) {
+					if (follow_dotdot_rcu(nd))
+						return -ECHILD;
+				} else
+					follow_dotdot(nd);
 				/* fallthrough */
 			case 1:
 				goto return_reval;
 		}
-		err = do_lookup(nd, &this, &next);
+		err = do_lookup(nd, &this, &next, &inode);
 		if (err)
 			break;
-		inode = next.dentry->d_inode;
 		if (follow_on_final(inode, lookup_flags)) {
+			if (nameidata_dentry_drop_rcu_maybe(nd, next.dentry))
+				return -ECHILD;
+			BUG_ON(inode != next.dentry->d_inode);
 			err = do_follow_link(&next, nd);
 			if (err)
 				goto return_err;
-			inode = nd->path.dentry->d_inode;
-		} else
+			nd->inode = nd->path.dentry->d_inode;
+		} else {
 			path_to_nameidata(&next, nd);
+			nd->inode = inode;
+		}
 		err = -ENOENT;
-		if (!inode)
+		if (!nd->inode)
 			break;
 		if (lookup_flags & LOOKUP_DIRECTORY) {
 			err = -ENOTDIR; 
-			if (!inode->i_op->lookup)
+			if (!nd->inode->i_op->lookup)
 				break;
 		}
 		goto return_base;
@@ -958,6 +1280,8 @@ return_reval:
 		 */
 		if (nd->path.dentry && nd->path.dentry->d_sb &&
 		    (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) {
+			if (nameidata_drop_rcu_maybe(nd))
+				return -ECHILD;
 			err = -ESTALE;
 			/* Note: we do not d_invalidate() */
 			if (!nd->path.dentry->d_op->d_revalidate(
@@ -965,16 +1289,34 @@ return_reval:
 				break;
 		}
 return_base:
+		if (nameidata_drop_rcu_last_maybe(nd))
+			return -ECHILD;
 		return 0;
 out_dput:
-		path_put_conditional(&next, nd);
+		if (!(nd->flags & LOOKUP_RCU))
+			path_put_conditional(&next, nd);
 		break;
 	}
-	path_put(&nd->path);
+	if (!(nd->flags & LOOKUP_RCU))
+		path_put(&nd->path);
 return_err:
 	return err;
 }
 
+static inline int path_walk_rcu(const char *name, struct nameidata *nd)
+{
+	current->total_link_count = 0;
+
+	return link_path_walk(name, nd);
+}
+
+static inline int path_walk_simple(const char *name, struct nameidata *nd)
+{
+	current->total_link_count = 0;
+
+	return link_path_walk(name, nd);
+}
+
 static int path_walk(const char *name, struct nameidata *nd)
 {
 	struct path save = nd->path;
@@ -1000,6 +1342,88 @@ static int path_walk(const char *name, struct nameidata *nd)
 	return result;
 }
 
+static void path_finish_rcu(struct nameidata *nd)
+{
+	if (nd->flags & LOOKUP_RCU) {
+		/* RCU dangling. Cancel it. */
+		nd->flags &= ~LOOKUP_RCU;
+		nd->root.mnt = NULL;
+		rcu_read_unlock();
+		br_read_unlock(vfsmount_lock);
+	}
+	if (nd->file)
+		fput(nd->file);
+}
+
+static int path_init_rcu(int dfd, const char *name, unsigned int flags, struct nameidata *nd)
+{
+	int retval = 0;
+	int fput_needed;
+	struct file *file;
+
+	nd->last_type = LAST_ROOT; /* if there are only slashes... */
+	nd->flags = flags | LOOKUP_RCU;
+	nd->depth = 0;
+	nd->root.mnt = NULL;
+	nd->file = NULL;
+
+	if (*name=='/') {
+		struct fs_struct *fs = current->fs;
+
+		br_read_lock(vfsmount_lock);
+		rcu_read_lock();
+
+		spin_lock(&fs->lock);
+		nd->root = fs->root;
+		nd->path = nd->root;
+		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
+		spin_unlock(&fs->lock);
+
+	} else if (dfd == AT_FDCWD) {
+		struct fs_struct *fs = current->fs;
+
+		br_read_lock(vfsmount_lock);
+		rcu_read_lock();
+
+		spin_lock(&fs->lock);
+		nd->path = fs->pwd;
+		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
+		spin_unlock(&fs->lock);
+	} else {
+		struct dentry *dentry;
+
+		file = fget_light(dfd, &fput_needed);
+		retval = -EBADF;
+		if (!file)
+			goto out_fail;
+
+		dentry = file->f_path.dentry;
+
+		retval = -ENOTDIR;
+		if (!S_ISDIR(dentry->d_inode->i_mode))
+			goto fput_fail;
+
+		retval = file_permission(file, MAY_EXEC);
+		if (retval)
+			goto fput_fail;
+
+		nd->path = file->f_path;
+		if (fput_needed)
+			nd->file = file;
+
+		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
+		br_read_lock(vfsmount_lock);
+		rcu_read_lock();
+	}
+	nd->inode = nd->path.dentry->d_inode;
+	return 0;
+
+fput_fail:
+	fput_light(file, fput_needed);
+out_fail:
+	return retval;
+}
+
 static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd)
 {
 	int retval = 0;
@@ -1040,6 +1464,7 @@ static int path_init(int dfd, const char *name, unsigned int flags, struct namei
 
 		fput_light(file, fput_needed);
 	}
+	nd->inode = nd->path.dentry->d_inode;
 	return 0;
 
 fput_fail:
@@ -1052,16 +1477,53 @@ out_fail:
 static int do_path_lookup(int dfd, const char *name,
 				unsigned int flags, struct nameidata *nd)
 {
-	int retval = path_init(dfd, name, flags, nd);
-	if (!retval)
-		retval = path_walk(name, nd);
-	if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry &&
-				nd->path.dentry->d_inode))
-		audit_inode(name, nd->path.dentry);
+	int retval;
+
+	/*
+	 * Path walking is largely split up into 2 different synchronisation
+	 * schemes, rcu-walk and ref-walk (explained in
+	 * Documentation/filesystems/path-lookup.txt). These share much of the
+	 * path walk code, but some things particularly setup, cleanup, and
+	 * following mounts are sufficiently divergent that functions are
+	 * duplicated. Typically there is a function foo(), and its RCU
+	 * analogue, foo_rcu().
+	 *
+	 * -ECHILD is the error number of choice (just to avoid clashes) that
+	 * is returned if some aspect of an rcu-walk fails. Such an error must
+	 * be handled by restarting a traditional ref-walk (which will always
+	 * be able to complete).
+	 */
+	retval = path_init_rcu(dfd, name, flags, nd);
+	if (unlikely(retval))
+		return retval;
+	retval = path_walk_rcu(name, nd);
+	path_finish_rcu(nd);
 	if (nd->root.mnt) {
 		path_put(&nd->root);
 		nd->root.mnt = NULL;
 	}
+
+	if (unlikely(retval == -ECHILD || retval == -ESTALE)) {
+		/* slower, locked walk */
+		if (retval == -ESTALE)
+			flags |= LOOKUP_REVAL;
+		retval = path_init(dfd, name, flags, nd);
+		if (unlikely(retval))
+			return retval;
+		retval = path_walk(name, nd);
+		if (nd->root.mnt) {
+			path_put(&nd->root);
+			nd->root.mnt = NULL;
+		}
+	}
+
+	if (likely(!retval)) {
+		if (unlikely(!audit_dummy_context())) {
+			if (nd->path.dentry && nd->inode)
+				audit_inode(name, nd->path.dentry);
+		}
+	}
+
 	return retval;
 }
 
@@ -1104,10 +1566,11 @@ int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt,
 	path_get(&nd->path);
 	nd->root = nd->path;
 	path_get(&nd->root);
+	nd->inode = nd->path.dentry->d_inode;
 
 	retval = path_walk(name, nd);
 	if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry &&
-				nd->path.dentry->d_inode))
+				nd->inode))
 		audit_inode(name, nd->path.dentry);
 
 	path_put(&nd->root);
@@ -1488,6 +1951,7 @@ out_unlock:
 	mutex_unlock(&dir->d_inode->i_mutex);
 	dput(nd->path.dentry);
 	nd->path.dentry = path->dentry;
+
 	if (error)
 		return error;
 	/* Don't check for write permission, don't truncate */
@@ -1582,6 +2046,9 @@ exit:
 	return ERR_PTR(error);
 }
 
+/*
+ * Handle O_CREAT case for do_filp_open
+ */
 static struct file *do_last(struct nameidata *nd, struct path *path,
 			    int open_flag, int acc_mode,
 			    int mode, const char *pathname)
@@ -1603,42 +2070,16 @@ static struct file *do_last(struct nameidata *nd, struct path *path,
 		}
 		/* fallthrough */
 	case LAST_ROOT:
-		if (open_flag & O_CREAT)
-			goto exit;
-		/* fallthrough */
+		goto exit;
 	case LAST_BIND:
 		audit_inode(pathname, dir);
 		goto ok;
 	}
 
 	/* trailing slashes? */
-	if (nd->last.name[nd->last.len]) {
-		if (open_flag & O_CREAT)
-			goto exit;
-		nd->flags |= LOOKUP_DIRECTORY | LOOKUP_FOLLOW;
-	}
-
-	/* just plain open? */
-	if (!(open_flag & O_CREAT)) {
-		error = do_lookup(nd, &nd->last, path);
-		if (error)
-			goto exit;
-		error = -ENOENT;
-		if (!path->dentry->d_inode)
-			goto exit_dput;
-		if (path->dentry->d_inode->i_op->follow_link)
-			return NULL;
-		error = -ENOTDIR;
-		if (nd->flags & LOOKUP_DIRECTORY) {
-			if (!path->dentry->d_inode->i_op->lookup)
-				goto exit_dput;
-		}
-		path_to_nameidata(path, nd);
-		audit_inode(pathname, nd->path.dentry);
-		goto ok;
-	}
+	if (nd->last.name[nd->last.len])
+		goto exit;
 
-	/* OK, it's O_CREAT */
 	mutex_lock(&dir->d_inode->i_mutex);
 
 	path->dentry = lookup_hash(nd);
@@ -1709,8 +2150,9 @@ static struct file *do_last(struct nameidata *nd, struct path *path,
 		return NULL;
 
 	path_to_nameidata(path, nd);
+	nd->inode = path->dentry->d_inode;
 	error = -EISDIR;
-	if (S_ISDIR(path->dentry->d_inode->i_mode))
+	if (S_ISDIR(nd->inode->i_mode))
 		goto exit;
 ok:
 	filp = finish_open(nd, open_flag, acc_mode);
@@ -1741,7 +2183,7 @@ struct file *do_filp_open(int dfd, const char *pathname,
 	struct path path;
 	int count = 0;
 	int flag = open_to_namei_flags(open_flag);
-	int force_reval = 0;
+	int flags;
 
 	if (!(open_flag & O_CREAT))
 		mode = 0;
@@ -1770,54 +2212,84 @@ struct file *do_filp_open(int dfd, const char *pathname,
 	if (open_flag & O_APPEND)
 		acc_mode |= MAY_APPEND;
 
-	/* find the parent */
-reval:
-	error = path_init(dfd, pathname, LOOKUP_PARENT, &nd);
+	flags = LOOKUP_OPEN;
+	if (open_flag & O_CREAT) {
+		flags |= LOOKUP_CREATE;
+		if (open_flag & O_EXCL)
+			flags |= LOOKUP_EXCL;
+	}
+	if (open_flag & O_DIRECTORY)
+		flags |= LOOKUP_DIRECTORY;
+	if (!(open_flag & O_NOFOLLOW))
+		flags |= LOOKUP_FOLLOW;
+
+	filp = get_empty_filp();
+	if (!filp)
+		return ERR_PTR(-ENFILE);
+
+	filp->f_flags = open_flag;
+	nd.intent.open.file = filp;
+	nd.intent.open.flags = flag;
+	nd.intent.open.create_mode = mode;
+
+	if (open_flag & O_CREAT)
+		goto creat;
+
+	/* !O_CREAT, simple open */
+	error = do_path_lookup(dfd, pathname, flags, &nd);
+	if (unlikely(error))
+		goto out_filp;
+	error = -ELOOP;
+	if (!(nd.flags & LOOKUP_FOLLOW)) {
+		if (nd.inode->i_op->follow_link)
+			goto out_path;
+	}
+	error = -ENOTDIR;
+	if (nd.flags & LOOKUP_DIRECTORY) {
+		if (!nd.inode->i_op->lookup)
+			goto out_path;
+	}
+	audit_inode(pathname, nd.path.dentry);
+	filp = finish_open(&nd, open_flag, acc_mode);
+	return filp;
+
+creat:
+	/* OK, have to create the file. Find the parent. */
+	error = path_init_rcu(dfd, pathname,
+			LOOKUP_PARENT | (flags & LOOKUP_REVAL), &nd);
 	if (error)
-		return ERR_PTR(error);
-	if (force_reval)
-		nd.flags |= LOOKUP_REVAL;
+		goto out_filp;
+	error = path_walk_rcu(pathname, &nd);
+	path_finish_rcu(&nd);
+	if (unlikely(error == -ECHILD || error == -ESTALE)) {
+		/* slower, locked walk */
+		if (error == -ESTALE) {
+reval:
+			flags |= LOOKUP_REVAL;
+		}
+		error = path_init(dfd, pathname,
+				LOOKUP_PARENT | (flags & LOOKUP_REVAL), &nd);
+		if (error)
+			goto out_filp;
 
-	current->total_link_count = 0;
-	error = link_path_walk(pathname, &nd);
-	if (error) {
-		filp = ERR_PTR(error);
-		goto out;
+		error = path_walk_simple(pathname, &nd);
 	}
-	if (unlikely(!audit_dummy_context()) && (open_flag & O_CREAT))
+	if (unlikely(error))
+		goto out_filp;
+	if (unlikely(!audit_dummy_context()))
 		audit_inode(pathname, nd.path.dentry);
 
 	/*
 	 * We have the parent and last component.
 	 */
-
-	error = -ENFILE;
-	filp = get_empty_filp();
-	if (filp == NULL)
-		goto exit_parent;
-	nd.intent.open.file = filp;
-	filp->f_flags = open_flag;
-	nd.intent.open.flags = flag;
-	nd.intent.open.create_mode = mode;
-	nd.flags &= ~LOOKUP_PARENT;
-	nd.flags |= LOOKUP_OPEN;
-	if (open_flag & O_CREAT) {
-		nd.flags |= LOOKUP_CREATE;
-		if (open_flag & O_EXCL)
-			nd.flags |= LOOKUP_EXCL;
-	}
-	if (open_flag & O_DIRECTORY)
-		nd.flags |= LOOKUP_DIRECTORY;
-	if (!(open_flag & O_NOFOLLOW))
-		nd.flags |= LOOKUP_FOLLOW;
+	nd.flags = flags;
 	filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname);
 	while (unlikely(!filp)) { /* trailing symlink */
 		struct path holder;
-		struct inode *inode = path.dentry->d_inode;
 		void *cookie;
 		error = -ELOOP;
 		/* S_ISDIR part is a temporary automount kludge */
-		if (!(nd.flags & LOOKUP_FOLLOW) && !S_ISDIR(inode->i_mode))
+		if (!(nd.flags & LOOKUP_FOLLOW) && !S_ISDIR(nd.inode->i_mode))
 			goto exit_dput;
 		if (count++ == 32)
 			goto exit_dput;
@@ -1838,36 +2310,33 @@ reval:
 			goto exit_dput;
 		error = __do_follow_link(&path, &nd, &cookie);
 		if (unlikely(error)) {
+			if (!IS_ERR(cookie) && nd.inode->i_op->put_link)
+				nd.inode->i_op->put_link(path.dentry, &nd, cookie);
 			/* nd.path had been dropped */
-			if (!IS_ERR(cookie) && inode->i_op->put_link)
-				inode->i_op->put_link(path.dentry, &nd, cookie);
-			path_put(&path);
-			release_open_intent(&nd);
-			filp = ERR_PTR(error);
-			goto out;
+			nd.path = path;
+			goto out_path;
 		}
 		holder = path;
 		nd.flags &= ~LOOKUP_PARENT;
 		filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname);
-		if (inode->i_op->put_link)
-			inode->i_op->put_link(holder.dentry, &nd, cookie);
+		if (nd.inode->i_op->put_link)
+			nd.inode->i_op->put_link(holder.dentry, &nd, cookie);
 		path_put(&holder);
 	}
 out:
 	if (nd.root.mnt)
 		path_put(&nd.root);
-	if (filp == ERR_PTR(-ESTALE) && !force_reval) {
-		force_reval = 1;
+	if (filp == ERR_PTR(-ESTALE) && !(flags & LOOKUP_REVAL))
 		goto reval;
-	}
 	return filp;
 
 exit_dput:
 	path_put_conditional(&path, &nd);
+out_path:
+	path_put(&nd.path);
+out_filp:
 	if (!IS_ERR(nd.intent.open.file))
 		release_open_intent(&nd);
-exit_parent:
-	path_put(&nd.path);
 	filp = ERR_PTR(error);
 	goto out;
 }
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index ae4b0fd..998e3a7 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -402,6 +402,10 @@ static int proc_sys_compare(const struct dentry *parent,
 		const struct dentry *dentry, const struct inode *inode,
 		unsigned int len, const char *str, const struct qstr *name)
 {
+	/* Although proc doesn't have negative dentries, rcu-walk means
+	 * that inode here can be NULL */
+	if (!inode)
+		return 0;
 	if (name->len != len)
 		return 1;
 	if (memcmp(name->name, str, len))
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index ca64868..c2e7390 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -5,6 +5,7 @@
 #include <linux/list.h>
 #include <linux/rculist.h>
 #include <linux/spinlock.h>
+#include <linux/seqlock.h>
 #include <linux/cache.h>
 #include <linux/rcupdate.h>
 
@@ -90,6 +91,7 @@ struct dentry {
 	unsigned int d_count;		/* protected by d_lock */
 	unsigned int d_flags;		/* protected by d_lock */
 	spinlock_t d_lock;		/* per dentry lock */
+	seqcount_t d_seq;		/* per dentry seqlock */
 	int d_mounted;
 	struct inode *d_inode;		/* Where the name belongs to - NULL is
 					 * negative */
@@ -266,9 +268,33 @@ extern void d_move(struct dentry *, struct dentry *);
 extern struct dentry *d_ancestor(struct dentry *, struct dentry *);
 
 /* appendix may either be NULL or be used for transname suffixes */
-extern struct dentry * d_lookup(struct dentry *, struct qstr *);
-extern struct dentry * __d_lookup(struct dentry *, struct qstr *);
-extern struct dentry * d_hash_and_lookup(struct dentry *, struct qstr *);
+extern struct dentry *d_lookup(struct dentry *, struct qstr *);
+extern struct dentry *d_hash_and_lookup(struct dentry *, struct qstr *);
+extern struct dentry *__d_lookup(struct dentry *, struct qstr *);
+extern struct dentry *__d_lookup_rcu(struct dentry *parent, struct qstr *name,
+				unsigned *seq, struct inode **inode);
+
+/**
+ * __d_rcu_to_refcount - take a refcount on dentry if sequence check is ok
+ * @dentry: dentry to take a ref on
+ * @seq: seqcount to verify against
+ * @Returns: 0 on failure, else 1.
+ *
+ * __d_rcu_to_refcount operates on a dentry,seq pair that was returned
+ * by __d_lookup_rcu, to get a reference on an rcu-walk dentry.
+ */
+static inline int __d_rcu_to_refcount(struct dentry *dentry, unsigned seq)
+{
+	int ret = 0;
+
+	assert_spin_locked(&dentry->d_lock);
+	if (!read_seqcount_retry(&dentry->d_seq, seq)) {
+		ret = 1;
+		dentry->d_count++;
+	}
+
+	return ret;
+}
 
 /* validate "insecure" dentry pointer */
 extern int d_validate(struct dentry *, struct dentry *);
diff --git a/include/linux/namei.h b/include/linux/namei.h
index aec730b..18d06ad 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -19,7 +19,10 @@ struct nameidata {
 	struct path	path;
 	struct qstr	last;
 	struct path	root;
+	struct file	*file;
+	struct inode	*inode; /* path.dentry.d_inode */
 	unsigned int	flags;
+	unsigned	seq;
 	int		last_type;
 	unsigned	depth;
 	char *saved_names[MAX_NESTED_LINKS + 1];
@@ -43,11 +46,13 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
  *  - internal "there are more path components" flag
  *  - dentry cache is untrusted; force a real lookup
  */
-#define LOOKUP_FOLLOW		 1
-#define LOOKUP_DIRECTORY	 2
-#define LOOKUP_CONTINUE		 4
-#define LOOKUP_PARENT		16
-#define LOOKUP_REVAL		64
+#define LOOKUP_FOLLOW		0x0001
+#define LOOKUP_DIRECTORY	0x0002
+#define LOOKUP_CONTINUE		0x0004
+
+#define LOOKUP_PARENT		0x0010
+#define LOOKUP_REVAL		0x0020
+#define LOOKUP_RCU		0x0040
 /*
  * Intent data
  */
diff --git a/include/linux/security.h b/include/linux/security.h
index fd4d55f..ed954019 100644
--- a/include/linux/security.h
+++ b/include/linux/security.h
@@ -457,7 +457,6 @@ static inline void security_free_mnt_opts(struct security_mnt_opts *opts)
  *	called when the actual read/write operations are performed.
  *	@inode contains the inode structure to check.
  *	@mask contains the permission mask.
- *	@nd contains the nameidata (may be NULL).
  *	Return 0 if permission is granted.
  * @inode_setattr:
  *	Check permission before setting file attributes.  Note that the kernel
@@ -1713,6 +1712,7 @@ int security_inode_rename(struct inode *old_dir, struct dentry *old_dentry,
 int security_inode_readlink(struct dentry *dentry);
 int security_inode_follow_link(struct dentry *dentry, struct nameidata *nd);
 int security_inode_permission(struct inode *inode, int mask);
+int security_inode_exec_permission(struct inode *inode, unsigned int flags);
 int security_inode_setattr(struct dentry *dentry, struct iattr *attr);
 int security_inode_getattr(struct vfsmount *mnt, struct dentry *dentry);
 int security_inode_setxattr(struct dentry *dentry, const char *name,
@@ -2102,6 +2102,12 @@ static inline int security_inode_permission(struct inode *inode, int mask)
 	return 0;
 }
 
+static inline int security_inode_exec_permission(struct inode *inode,
+						  unsigned int flags)
+{
+	return 0;
+}
+
 static inline int security_inode_setattr(struct dentry *dentry,
 					  struct iattr *attr)
 {
diff --git a/security/security.c b/security/security.c
index 1b798d3..c645e26 100644
--- a/security/security.c
+++ b/security/security.c
@@ -513,6 +513,15 @@ int security_inode_permission(struct inode *inode, int mask)
 	return security_ops->inode_permission(inode, mask);
 }
 
+int security_inode_exec_permission(struct inode *inode, unsigned int flags)
+{
+	if (unlikely(IS_PRIVATE(inode)))
+		return 0;
+	if (flags)
+		return -ECHILD;
+	return security_ops->inode_permission(inode, MAY_EXEC);
+}
+
 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
 {
 	if (unlikely(IS_PRIVATE(dentry->d_inode)))
-- 
cgit v0.10.2


From c28cc36469554dc55540f059fbdc7fa22a2c31fc Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:53 +1100
Subject: fs: fs_struct use seqlock

Use a seqlock in the fs_struct to enable us to take an atomic copy of the
complete cwd and root paths. Use this in the RCU lookup path to avoid a
thread-shared spinlock in RCU lookup operations.

Multi-threaded apps may now perform path lookups with scalability matching
multi-process apps. Operations such as stat(2) become very scalable for
multi-threaded workload.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/fs_struct.c b/fs/fs_struct.c
index ed45a9c..60b8531 100644
--- a/fs/fs_struct.c
+++ b/fs/fs_struct.c
@@ -14,9 +14,11 @@ void set_fs_root(struct fs_struct *fs, struct path *path)
 	struct path old_root;
 
 	spin_lock(&fs->lock);
+	write_seqcount_begin(&fs->seq);
 	old_root = fs->root;
 	fs->root = *path;
 	path_get(path);
+	write_seqcount_end(&fs->seq);
 	spin_unlock(&fs->lock);
 	if (old_root.dentry)
 		path_put(&old_root);
@@ -31,9 +33,11 @@ void set_fs_pwd(struct fs_struct *fs, struct path *path)
 	struct path old_pwd;
 
 	spin_lock(&fs->lock);
+	write_seqcount_begin(&fs->seq);
 	old_pwd = fs->pwd;
 	fs->pwd = *path;
 	path_get(path);
+	write_seqcount_end(&fs->seq);
 	spin_unlock(&fs->lock);
 
 	if (old_pwd.dentry)
@@ -52,6 +56,7 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root)
 		fs = p->fs;
 		if (fs) {
 			spin_lock(&fs->lock);
+			write_seqcount_begin(&fs->seq);
 			if (fs->root.dentry == old_root->dentry
 			    && fs->root.mnt == old_root->mnt) {
 				path_get(new_root);
@@ -64,6 +69,7 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root)
 				fs->pwd = *new_root;
 				count++;
 			}
+			write_seqcount_end(&fs->seq);
 			spin_unlock(&fs->lock);
 		}
 		task_unlock(p);
@@ -88,8 +94,10 @@ void exit_fs(struct task_struct *tsk)
 		int kill;
 		task_lock(tsk);
 		spin_lock(&fs->lock);
+		write_seqcount_begin(&fs->seq);
 		tsk->fs = NULL;
 		kill = !--fs->users;
+		write_seqcount_end(&fs->seq);
 		spin_unlock(&fs->lock);
 		task_unlock(tsk);
 		if (kill)
@@ -105,6 +113,7 @@ struct fs_struct *copy_fs_struct(struct fs_struct *old)
 		fs->users = 1;
 		fs->in_exec = 0;
 		spin_lock_init(&fs->lock);
+		seqcount_init(&fs->seq);
 		fs->umask = old->umask;
 		get_fs_root_and_pwd(old, &fs->root, &fs->pwd);
 	}
@@ -144,6 +153,7 @@ EXPORT_SYMBOL(current_umask);
 struct fs_struct init_fs = {
 	.users		= 1,
 	.lock		= __SPIN_LOCK_UNLOCKED(init_fs.lock),
+	.seq		= SEQCNT_ZERO,
 	.umask		= 0022,
 };
 
diff --git a/fs/namei.c b/fs/namei.c
index 8d3f15b..c731b50 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -684,9 +684,12 @@ static __always_inline void set_root_rcu(struct nameidata *nd)
 {
 	if (!nd->root.mnt) {
 		struct fs_struct *fs = current->fs;
-		spin_lock(&fs->lock);
-		nd->root = fs->root;
-		spin_unlock(&fs->lock);
+		unsigned seq;
+
+		do {
+			seq = read_seqcount_begin(&fs->seq);
+			nd->root = fs->root;
+		} while (read_seqcount_retry(&fs->seq, seq));
 	}
 }
 
@@ -1369,26 +1372,31 @@ static int path_init_rcu(int dfd, const char *name, unsigned int flags, struct n
 
 	if (*name=='/') {
 		struct fs_struct *fs = current->fs;
+		unsigned seq;
 
 		br_read_lock(vfsmount_lock);
 		rcu_read_lock();
 
-		spin_lock(&fs->lock);
-		nd->root = fs->root;
-		nd->path = nd->root;
-		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
-		spin_unlock(&fs->lock);
+		do {
+			seq = read_seqcount_begin(&fs->seq);
+			nd->root = fs->root;
+			nd->path = nd->root;
+			nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq);
+		} while (read_seqcount_retry(&fs->seq, seq));
 
 	} else if (dfd == AT_FDCWD) {
 		struct fs_struct *fs = current->fs;
+		unsigned seq;
 
 		br_read_lock(vfsmount_lock);
 		rcu_read_lock();
 
-		spin_lock(&fs->lock);
-		nd->path = fs->pwd;
-		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
-		spin_unlock(&fs->lock);
+		do {
+			seq = read_seqcount_begin(&fs->seq);
+			nd->path = fs->pwd;
+			nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq);
+		} while (read_seqcount_retry(&fs->seq, seq));
+
 	} else {
 		struct dentry *dentry;
 
@@ -1411,7 +1419,7 @@ static int path_init_rcu(int dfd, const char *name, unsigned int flags, struct n
 		if (fput_needed)
 			nd->file = file;
 
-		nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq);
+		nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq);
 		br_read_lock(vfsmount_lock);
 		rcu_read_lock();
 	}
diff --git a/include/linux/fs_struct.h b/include/linux/fs_struct.h
index a42b5bf..003dc0f 100644
--- a/include/linux/fs_struct.h
+++ b/include/linux/fs_struct.h
@@ -2,10 +2,13 @@
 #define _LINUX_FS_STRUCT_H
 
 #include <linux/path.h>
+#include <linux/spinlock.h>
+#include <linux/seqlock.h>
 
 struct fs_struct {
 	int users;
 	spinlock_t lock;
+	seqcount_t seq;
 	int umask;
 	int in_exec;
 	struct path root, pwd;
-- 
cgit v0.10.2


From 5f57cbcc02cf18f6b22ef4066bb10afeb8f930ff Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:54 +1100
Subject: fs: dcache remove d_mounted

Rather than keep a d_mounted count in the dentry, set a dentry flag instead.
The flag can be cleared by checking the hash table to see if there are any
mounts left, which is not time critical because it is performed at detach time.

The mounted state of a dentry is only used to speculatively take a look in the
mount hash table if it is set -- before following the mount, vfsmount lock is
taken and mount re-checked without races.

This saves 4 bytes on 32-bit, nothing on 64-bit but it does provide a hole I
might use later (and some configs have larger than 32-bit spinlocks which might
make use of the hole).

Autofs4 conversion and changelog by Ian Kent <raven@themaw.net>:
In autofs4, when expring direct (or offset) mounts we need to ensure that we
block user path walks into the autofs mount, which is covered by another mount.
To do this we clear the mounted status so that follows stop before walking into
the mount and are essentially blocked until the expire is completed. The
automount daemon still finds the correct dentry for the umount due to the
follow mount logic in fs/autofs4/root.c:autofs4_follow_link(), which is set as
an inode operation for direct and offset mounts only and is called following
the lookup that stopped at the covered mount.

At the end of the expire the covering mount probably has gone away so the
mounted status need not be restored. But we need to check this and only restore
the mounted status if the expire failed.

XXX: autofs may not work right if we have other mounts go over the top of it?

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index 2f7951d..cc1d013 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -295,7 +295,9 @@ struct dentry *autofs4_expire_direct(struct super_block *sb,
 		struct autofs_info *ino = autofs4_dentry_ino(root);
 		if (d_mountpoint(root)) {
 			ino->flags |= AUTOFS_INF_MOUNTPOINT;
-			root->d_mounted--;
+			spin_lock(&root->d_lock);
+			root->d_flags &= ~DCACHE_MOUNTED;
+			spin_unlock(&root->d_lock);
 		}
 		ino->flags |= AUTOFS_INF_EXPIRING;
 		init_completion(&ino->expire_complete);
@@ -503,7 +505,14 @@ int autofs4_do_expire_multi(struct super_block *sb, struct vfsmount *mnt,
 
 		spin_lock(&sbi->fs_lock);
 		if (ino->flags & AUTOFS_INF_MOUNTPOINT) {
-			sb->s_root->d_mounted++;
+			spin_lock(&sb->s_root->d_lock);
+			/*
+			 * If we haven't been expired away, then reset
+			 * mounted status.
+			 */
+			if (mnt->mnt_parent != mnt)
+				sb->s_root->d_flags |= DCACHE_MOUNTED;
+			spin_unlock(&sb->s_root->d_lock);
 			ino->flags &= ~AUTOFS_INF_MOUNTPOINT;
 		}
 		ino->flags &= ~AUTOFS_INF_EXPIRING;
diff --git a/fs/dcache.c b/fs/dcache.c
index 187fea0..1d5cf51 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1265,7 +1265,6 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	dentry->d_sb = NULL;
 	dentry->d_op = NULL;
 	dentry->d_fsdata = NULL;
-	dentry->d_mounted = 0;
 	INIT_HLIST_NODE(&dentry->d_hash);
 	INIT_LIST_HEAD(&dentry->d_lru);
 	INIT_LIST_HEAD(&dentry->d_subdirs);
diff --git a/fs/namespace.c b/fs/namespace.c
index 3dbfc07..39a7d50 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -492,6 +492,27 @@ static void __touch_mnt_namespace(struct mnt_namespace *ns)
 }
 
 /*
+ * Clear dentry's mounted state if it has no remaining mounts.
+ * vfsmount_lock must be held for write.
+ */
+static void dentry_reset_mounted(struct vfsmount *mnt, struct dentry *dentry)
+{
+	unsigned u;
+
+	for (u = 0; u < HASH_SIZE; u++) {
+		struct vfsmount *p;
+
+		list_for_each_entry(p, &mount_hashtable[u], mnt_hash) {
+			if (p->mnt_mountpoint == dentry)
+				return;
+		}
+	}
+	spin_lock(&dentry->d_lock);
+	dentry->d_flags &= ~DCACHE_MOUNTED;
+	spin_unlock(&dentry->d_lock);
+}
+
+/*
  * vfsmount lock must be held for write
  */
 static void detach_mnt(struct vfsmount *mnt, struct path *old_path)
@@ -502,7 +523,7 @@ static void detach_mnt(struct vfsmount *mnt, struct path *old_path)
 	mnt->mnt_mountpoint = mnt->mnt_root;
 	list_del_init(&mnt->mnt_child);
 	list_del_init(&mnt->mnt_hash);
-	old_path->dentry->d_mounted--;
+	dentry_reset_mounted(old_path->mnt, old_path->dentry);
 }
 
 /*
@@ -513,7 +534,9 @@ void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry,
 {
 	child_mnt->mnt_parent = mntget(mnt);
 	child_mnt->mnt_mountpoint = dget(dentry);
-	dentry->d_mounted++;
+	spin_lock(&dentry->d_lock);
+	dentry->d_flags |= DCACHE_MOUNTED;
+	spin_unlock(&dentry->d_lock);
 }
 
 /*
@@ -1073,7 +1096,7 @@ void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill)
 		list_del_init(&p->mnt_child);
 		if (p->mnt_parent != p) {
 			p->mnt_parent->mnt_ghosts++;
-			p->mnt_mountpoint->d_mounted--;
+			dentry_reset_mounted(p->mnt_parent, p->mnt_mountpoint);
 		}
 		change_mnt_propagation(p, MS_PRIVATE);
 	}
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index c2e7390..e441469 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -92,7 +92,6 @@ struct dentry {
 	unsigned int d_flags;		/* protected by d_lock */
 	spinlock_t d_lock;		/* per dentry lock */
 	seqcount_t d_seq;		/* per dentry seqlock */
-	int d_mounted;
 	struct inode *d_inode;		/* Where the name belongs to - NULL is
 					 * negative */
 	/*
@@ -156,33 +155,34 @@ struct dentry_operations {
 
 /* d_flags entries */
 #define DCACHE_AUTOFS_PENDING 0x0001    /* autofs: "under construction" */
-#define DCACHE_NFSFS_RENAMED  0x0002    /* this dentry has been "silly
-					 * renamed" and has to be
-					 * deleted on the last dput()
-					 */
-#define	DCACHE_DISCONNECTED 0x0004
-     /* This dentry is possibly not currently connected to the dcache tree,
-      * in which case its parent will either be itself, or will have this
-      * flag as well.  nfsd will not use a dentry with this bit set, but will
-      * first endeavour to clear the bit either by discovering that it is
-      * connected, or by performing lookup operations.   Any filesystem which
-      * supports nfsd_operations MUST have a lookup function which, if it finds
-      * a directory inode with a DCACHE_DISCONNECTED dentry, will d_move
-      * that dentry into place and return that dentry rather than the passed one,
-      * typically using d_splice_alias.
-      */
+#define DCACHE_NFSFS_RENAMED  0x0002
+     /* this dentry has been "silly renamed" and has to be deleted on the last
+      * dput() */
+
+#define	DCACHE_DISCONNECTED	0x0004
+     /* This dentry is possibly not currently connected to the dcache tree, in
+      * which case its parent will either be itself, or will have this flag as
+      * well.  nfsd will not use a dentry with this bit set, but will first
+      * endeavour to clear the bit either by discovering that it is connected,
+      * or by performing lookup operations.   Any filesystem which supports
+      * nfsd_operations MUST have a lookup function which, if it finds a
+      * directory inode with a DCACHE_DISCONNECTED dentry, will d_move that
+      * dentry into place and return that dentry rather than the passed one,
+      * typically using d_splice_alias. */
 
 #define DCACHE_REFERENCED	0x0008  /* Recently used, don't discard. */
 #define DCACHE_UNHASHED		0x0010	
-
-#define DCACHE_INOTIFY_PARENT_WATCHED	0x0020 /* Parent inode is watched by inotify */
+#define DCACHE_INOTIFY_PARENT_WATCHED 0x0020
+     /* Parent inode is watched by inotify */
 
 #define DCACHE_COOKIE		0x0040	/* For use by dcookie subsystem */
-
-#define DCACHE_FSNOTIFY_PARENT_WATCHED	0x0080 /* Parent inode is watched by some fsnotify listener */
+#define DCACHE_FSNOTIFY_PARENT_WATCHED 0x0080
+     /* Parent inode is watched by some fsnotify listener */
 
 #define DCACHE_CANT_MOUNT	0x0100
 #define DCACHE_GENOCIDE		0x0200
+#define DCACHE_MOUNTED		0x0400	/* is a mountpoint */
+
 
 extern spinlock_t dcache_inode_lock;
 extern seqlock_t rename_lock;
@@ -372,7 +372,7 @@ extern void dput(struct dentry *);
 
 static inline int d_mountpoint(struct dentry *dentry)
 {
-	return dentry->d_mounted;
+	return dentry->d_flags & DCACHE_MOUNTED;
 }
 
 extern struct vfsmount *lookup_mnt(struct path *);
-- 
cgit v0.10.2


From fb045adb99d9b7c562dc7fef834857f78249daa1 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:55 +1100
Subject: fs: dcache reduce branches in lookup path

Reduce some branches and memory accesses in dcache lookup by adding dentry
flags to indicate common d_ops are set, rather than having to check them.
This saves a pointer memory access (dentry->d_op) in common path lookup
situations, and saves another pointer load and branch in cases where we
have d_op but not the particular operation.

Patched with:

git grep -E '[.>]([[:space:]])*d_op([[:space:]])*=' | xargs sed -e 's/\([^\t ]*\)->d_op = \(.*\);/d_set_d_op(\1, \2);/' -e 's/\([^\t ]*\)\.d_op = \(.*\);/d_set_d_op(\&\1, \2);/' -i

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index d39d8a5..5a24f40 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -2233,7 +2233,7 @@ pfm_alloc_file(pfm_context_t *ctx)
 	}
 	path.mnt = mntget(pfmfs_mnt);
 
-	path.dentry->d_op = &pfmfs_dentry_operations;
+	d_set_d_op(path.dentry, &pfmfs_dentry_operations);
 	d_add(path.dentry, inode);
 
 	file = alloc_file(&path, FMODE_READ, &pfm_file_ops);
diff --git a/drivers/staging/autofs/root.c b/drivers/staging/autofs/root.c
index 0fdec4b..b09adb5 100644
--- a/drivers/staging/autofs/root.c
+++ b/drivers/staging/autofs/root.c
@@ -237,7 +237,7 @@ static struct dentry *autofs_root_lookup(struct inode *dir, struct dentry *dentr
 	 *
 	 * We need to do this before we release the directory semaphore.
 	 */
-	dentry->d_op = &autofs_dentry_operations;
+	d_set_d_op(dentry, &autofs_dentry_operations);
 	dentry->d_flags |= DCACHE_AUTOFS_PENDING;
 	d_add(dentry, NULL);
 
diff --git a/drivers/staging/smbfs/dir.c b/drivers/staging/smbfs/dir.c
index 5f79799..78f0941 100644
--- a/drivers/staging/smbfs/dir.c
+++ b/drivers/staging/smbfs/dir.c
@@ -398,9 +398,9 @@ smb_new_dentry(struct dentry *dentry)
 	struct smb_sb_info *server = server_from_dentry(dentry);
 
 	if (server->mnt->flags & SMB_MOUNT_CASE)
-		dentry->d_op = &smbfs_dentry_operations_case;
+		d_set_d_op(dentry, &smbfs_dentry_operations_case);
 	else
-		dentry->d_op = &smbfs_dentry_operations;
+		d_set_d_op(dentry, &smbfs_dentry_operations);
 	dentry->d_time = jiffies;
 }
 
@@ -462,9 +462,9 @@ smb_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
 	add_entry:
 			server = server_from_dentry(dentry);
 			if (server->mnt->flags & SMB_MOUNT_CASE)
-				dentry->d_op = &smbfs_dentry_operations_case;
+				d_set_d_op(dentry, &smbfs_dentry_operations_case);
 			else
-				dentry->d_op = &smbfs_dentry_operations;
+				d_set_d_op(dentry, &smbfs_dentry_operations);
 
 			d_add(dentry, inode);
 			smb_renew_times(dentry);
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index f6f9081..df8bbb3 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -635,9 +635,9 @@ v9fs_create(struct v9fs_session_info *v9ses, struct inode *dir,
 	}
 
 	if (v9ses->cache)
-		dentry->d_op = &v9fs_cached_dentry_operations;
+		d_set_d_op(dentry, &v9fs_cached_dentry_operations);
 	else
-		dentry->d_op = &v9fs_dentry_operations;
+		d_set_d_op(dentry, &v9fs_dentry_operations);
 
 	d_instantiate(dentry, inode);
 	err = v9fs_fid_add(dentry, fid);
@@ -749,7 +749,7 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 				err);
 			goto error;
 		}
-		dentry->d_op = &v9fs_cached_dentry_operations;
+		d_set_d_op(dentry, &v9fs_cached_dentry_operations);
 		d_instantiate(dentry, inode);
 		err = v9fs_fid_add(dentry, fid);
 		if (err < 0)
@@ -767,7 +767,7 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 			err = PTR_ERR(inode);
 			goto error;
 		}
-		dentry->d_op = &v9fs_dentry_operations;
+		d_set_d_op(dentry, &v9fs_dentry_operations);
 		d_instantiate(dentry, inode);
 	}
 	/* Now set the ACL based on the default value */
@@ -956,7 +956,7 @@ static int v9fs_vfs_mkdir_dotl(struct inode *dir,
 				err);
 			goto error;
 		}
-		dentry->d_op = &v9fs_cached_dentry_operations;
+		d_set_d_op(dentry, &v9fs_cached_dentry_operations);
 		d_instantiate(dentry, inode);
 		err = v9fs_fid_add(dentry, fid);
 		if (err < 0)
@@ -973,7 +973,7 @@ static int v9fs_vfs_mkdir_dotl(struct inode *dir,
 			err = PTR_ERR(inode);
 			goto error;
 		}
-		dentry->d_op = &v9fs_dentry_operations;
+		d_set_d_op(dentry, &v9fs_dentry_operations);
 		d_instantiate(dentry, inode);
 	}
 	/* Now set the ACL based on the default value */
@@ -1041,9 +1041,9 @@ static struct dentry *v9fs_vfs_lookup(struct inode *dir, struct dentry *dentry,
 
 inst_out:
 	if (v9ses->cache)
-		dentry->d_op = &v9fs_cached_dentry_operations;
+		d_set_d_op(dentry, &v9fs_cached_dentry_operations);
 	else
-		dentry->d_op = &v9fs_dentry_operations;
+		d_set_d_op(dentry, &v9fs_dentry_operations);
 
 	d_add(dentry, inode);
 	return NULL;
@@ -1709,7 +1709,7 @@ v9fs_vfs_symlink_dotl(struct inode *dir, struct dentry *dentry,
 					err);
 			goto error;
 		}
-		dentry->d_op = &v9fs_cached_dentry_operations;
+		d_set_d_op(dentry, &v9fs_cached_dentry_operations);
 		d_instantiate(dentry, inode);
 		err = v9fs_fid_add(dentry, fid);
 		if (err < 0)
@@ -1722,7 +1722,7 @@ v9fs_vfs_symlink_dotl(struct inode *dir, struct dentry *dentry,
 			err = PTR_ERR(inode);
 			goto error;
 		}
-		dentry->d_op = &v9fs_dentry_operations;
+		d_set_d_op(dentry, &v9fs_dentry_operations);
 		d_instantiate(dentry, inode);
 	}
 
@@ -1856,7 +1856,7 @@ v9fs_vfs_link_dotl(struct dentry *old_dentry, struct inode *dir,
 		ihold(old_dentry->d_inode);
 	}
 
-	dentry->d_op = old_dentry->d_op;
+	d_set_d_op(dentry, old_dentry->d_op);
 	d_instantiate(dentry, old_dentry->d_inode);
 
 	return err;
@@ -1980,7 +1980,7 @@ v9fs_vfs_mknod_dotl(struct inode *dir, struct dentry *dentry, int omode,
 				err);
 			goto error;
 		}
-		dentry->d_op = &v9fs_cached_dentry_operations;
+		d_set_d_op(dentry, &v9fs_cached_dentry_operations);
 		d_instantiate(dentry, inode);
 		err = v9fs_fid_add(dentry, fid);
 		if (err < 0)
@@ -1996,7 +1996,7 @@ v9fs_vfs_mknod_dotl(struct inode *dir, struct dentry *dentry, int omode,
 			err = PTR_ERR(inode);
 			goto error;
 		}
-		dentry->d_op = &v9fs_dentry_operations;
+		d_set_d_op(dentry, &v9fs_dentry_operations);
 		d_instantiate(dentry, inode);
 	}
 	/* Now set the ACL based on the default value */
diff --git a/fs/adfs/dir.c b/fs/adfs/dir.c
index a11e5e1..bf7693c 100644
--- a/fs/adfs/dir.c
+++ b/fs/adfs/dir.c
@@ -276,7 +276,7 @@ adfs_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
 	struct object_info obj;
 	int error;
 
-	dentry->d_op = &adfs_dentry_operations;	
+	d_set_d_op(dentry, &adfs_dentry_operations);
 	lock_kernel();
 	error = adfs_dir_lookup_byname(dir, &dentry->d_name, &obj);
 	if (error == 0) {
diff --git a/fs/adfs/super.c b/fs/adfs/super.c
index 47dffc5..a4041b5 100644
--- a/fs/adfs/super.c
+++ b/fs/adfs/super.c
@@ -484,7 +484,7 @@ static int adfs_fill_super(struct super_block *sb, void *data, int silent)
 		adfs_error(sb, "get root inode failed\n");
 		goto error;
 	} else
-		sb->s_root->d_op = &adfs_dentry_operations;
+		d_set_d_op(sb->s_root, &adfs_dentry_operations);
 	unlock_kernel();
 	return 0;
 
diff --git a/fs/affs/namei.c b/fs/affs/namei.c
index 5aca08c..944a404 100644
--- a/fs/affs/namei.c
+++ b/fs/affs/namei.c
@@ -240,7 +240,7 @@ affs_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
 		if (IS_ERR(inode))
 			return ERR_CAST(inode);
 	}
-	dentry->d_op = AFFS_SB(sb)->s_flags & SF_INTL ? &affs_intl_dentry_operations : &affs_dentry_operations;
+	d_set_d_op(dentry, AFFS_SB(sb)->s_flags & SF_INTL ? &affs_intl_dentry_operations : &affs_dentry_operations);
 	d_add(dentry, inode);
 	return NULL;
 }
diff --git a/fs/affs/super.c b/fs/affs/super.c
index 4c18fcf..d39081b 100644
--- a/fs/affs/super.c
+++ b/fs/affs/super.c
@@ -482,7 +482,7 @@ got_root:
 		printk(KERN_ERR "AFFS: Get root inode failed\n");
 		goto out_error;
 	}
-	sb->s_root->d_op = &affs_dentry_operations;
+	d_set_d_op(sb->s_root, &affs_dentry_operations);
 
 	pr_debug("AFFS: s_flags=%lX\n",sb->s_flags);
 	return 0;
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index 2c18cde..b8bb7e7 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -581,7 +581,7 @@ static struct dentry *afs_lookup(struct inode *dir, struct dentry *dentry,
 	}
 
 success:
-	dentry->d_op = &afs_fs_dentry_operations;
+	d_set_d_op(dentry, &afs_fs_dentry_operations);
 
 	d_add(dentry, inode);
 	_leave(" = 0 { vn=%u u=%u } -> { ino=%lu v=%llu }",
diff --git a/fs/anon_inodes.c b/fs/anon_inodes.c
index 57ce55b..aca8806 100644
--- a/fs/anon_inodes.c
+++ b/fs/anon_inodes.c
@@ -113,7 +113,7 @@ struct file *anon_inode_getfile(const char *name,
 	 */
 	ihold(anon_inode_inode);
 
-	path.dentry->d_op = &anon_inodefs_dentry_operations;
+	d_set_d_op(path.dentry, &anon_inodefs_dentry_operations);
 	d_instantiate(path.dentry, anon_inode_inode);
 
 	error = -ENFILE;
diff --git a/fs/autofs4/inode.c b/fs/autofs4/inode.c
index ac87e49..a7bdb9d 100644
--- a/fs/autofs4/inode.c
+++ b/fs/autofs4/inode.c
@@ -309,7 +309,7 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
 		goto fail_iput;
 	pipe = NULL;
 
-	root->d_op = &autofs4_sb_dentry_operations;
+	d_set_d_op(root, &autofs4_sb_dentry_operations);
 	root->d_fsdata = ino;
 
 	/* Can this call block? */
diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
index 10ca68a..bfe3f2e 100644
--- a/fs/autofs4/root.c
+++ b/fs/autofs4/root.c
@@ -571,7 +571,7 @@ static struct dentry *autofs4_lookup(struct inode *dir, struct dentry *dentry, s
 		 * we check for the hashed dentry and return the newly
 		 * hashed dentry.
 		 */
-		dentry->d_op = &autofs4_root_dentry_operations;
+		d_set_d_op(dentry, &autofs4_root_dentry_operations);
 
 		/*
 		 * And we need to ensure that the same dentry is used for
@@ -710,9 +710,9 @@ static int autofs4_dir_symlink(struct inode *dir,
 	d_add(dentry, inode);
 
 	if (dir == dir->i_sb->s_root->d_inode)
-		dentry->d_op = &autofs4_root_dentry_operations;
+		d_set_d_op(dentry, &autofs4_root_dentry_operations);
 	else
-		dentry->d_op = &autofs4_dentry_operations;
+		d_set_d_op(dentry, &autofs4_dentry_operations);
 
 	dentry->d_fsdata = ino;
 	ino->dentry = dget(dentry);
@@ -845,9 +845,9 @@ static int autofs4_dir_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	d_add(dentry, inode);
 
 	if (dir == dir->i_sb->s_root->d_inode)
-		dentry->d_op = &autofs4_root_dentry_operations;
+		d_set_d_op(dentry, &autofs4_root_dentry_operations);
 	else
-		dentry->d_op = &autofs4_dentry_operations;
+		d_set_d_op(dentry, &autofs4_dentry_operations);
 
 	dentry->d_fsdata = ino;
 	ino->dentry = dget(dentry);
diff --git a/fs/btrfs/export.c b/fs/btrfs/export.c
index 659f532..0ccf9a8 100644
--- a/fs/btrfs/export.c
+++ b/fs/btrfs/export.c
@@ -110,7 +110,7 @@ static struct dentry *btrfs_get_dentry(struct super_block *sb, u64 objectid,
 
 	dentry = d_obtain_alias(inode);
 	if (!IS_ERR(dentry))
-		dentry->d_op = &btrfs_dentry_operations;
+		d_set_d_op(dentry, &btrfs_dentry_operations);
 	return dentry;
 fail:
 	srcu_read_unlock(&fs_info->subvol_srcu, index);
@@ -225,7 +225,7 @@ static struct dentry *btrfs_get_parent(struct dentry *child)
 	key.offset = 0;
 	dentry = d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root, NULL));
 	if (!IS_ERR(dentry))
-		dentry->d_op = &btrfs_dentry_operations;
+		d_set_d_op(dentry, &btrfs_dentry_operations);
 	return dentry;
 fail:
 	btrfs_free_path(path);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index f9d2994..63e4546 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4084,7 +4084,7 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 	int index;
 	int ret;
 
-	dentry->d_op = &btrfs_dentry_operations;
+	d_set_d_op(dentry, &btrfs_dentry_operations);
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 58abc3d..cc01cf8 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -42,11 +42,11 @@ int ceph_init_dentry(struct dentry *dentry)
 
 	if (dentry->d_parent == NULL ||   /* nfs fh_to_dentry */
 	    ceph_snap(dentry->d_parent->d_inode) == CEPH_NOSNAP)
-		dentry->d_op = &ceph_dentry_ops;
+		d_set_d_op(dentry, &ceph_dentry_ops);
 	else if (ceph_snap(dentry->d_parent->d_inode) == CEPH_SNAPDIR)
-		dentry->d_op = &ceph_snapdir_dentry_ops;
+		d_set_d_op(dentry, &ceph_snapdir_dentry_ops);
 	else
-		dentry->d_op = &ceph_snap_dentry_ops;
+		d_set_d_op(dentry, &ceph_snap_dentry_ops);
 
 	di = kmem_cache_alloc(ceph_dentry_cachep, GFP_NOFS | __GFP_ZERO);
 	if (!di)
diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index 88bfe68..e3b10ca 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -135,9 +135,9 @@ static void setup_cifs_dentry(struct cifsTconInfo *tcon,
 			      struct inode *newinode)
 {
 	if (tcon->nocase)
-		direntry->d_op = &cifs_ci_dentry_ops;
+		d_set_d_op(direntry, &cifs_ci_dentry_ops);
 	else
-		direntry->d_op = &cifs_dentry_ops;
+		d_set_d_op(direntry, &cifs_dentry_ops);
 	d_instantiate(direntry, newinode);
 }
 
@@ -421,9 +421,9 @@ int cifs_mknod(struct inode *inode, struct dentry *direntry, int mode,
 		rc = cifs_get_inode_info_unix(&newinode, full_path,
 						inode->i_sb, xid);
 		if (pTcon->nocase)
-			direntry->d_op = &cifs_ci_dentry_ops;
+			d_set_d_op(direntry, &cifs_ci_dentry_ops);
 		else
-			direntry->d_op = &cifs_dentry_ops;
+			d_set_d_op(direntry, &cifs_dentry_ops);
 
 		if (rc == 0)
 			d_instantiate(direntry, newinode);
@@ -604,9 +604,9 @@ cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
 
 	if ((rc == 0) && (newInode != NULL)) {
 		if (pTcon->nocase)
-			direntry->d_op = &cifs_ci_dentry_ops;
+			d_set_d_op(direntry, &cifs_ci_dentry_ops);
 		else
-			direntry->d_op = &cifs_dentry_ops;
+			d_set_d_op(direntry, &cifs_dentry_ops);
 		d_add(direntry, newInode);
 		if (posix_open) {
 			filp = lookup_instantiate_filp(nd, direntry,
@@ -634,9 +634,9 @@ cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
 		rc = 0;
 		direntry->d_time = jiffies;
 		if (pTcon->nocase)
-			direntry->d_op = &cifs_ci_dentry_ops;
+			d_set_d_op(direntry, &cifs_ci_dentry_ops);
 		else
-			direntry->d_op = &cifs_dentry_ops;
+			d_set_d_op(direntry, &cifs_dentry_ops);
 		d_add(direntry, NULL);
 	/*	if it was once a directory (but how can we tell?) we could do
 		shrink_dcache_parent(direntry); */
diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index 99b9a2c..2a239d8 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -1319,9 +1319,9 @@ int cifs_mkdir(struct inode *inode, struct dentry *direntry, int mode)
 	to set uid/gid */
 			inc_nlink(inode);
 			if (pTcon->nocase)
-				direntry->d_op = &cifs_ci_dentry_ops;
+				d_set_d_op(direntry, &cifs_ci_dentry_ops);
 			else
-				direntry->d_op = &cifs_dentry_ops;
+				d_set_d_op(direntry, &cifs_dentry_ops);
 
 			cifs_unix_basic_to_fattr(&fattr, pInfo, cifs_sb);
 			cifs_fill_uniqueid(inode->i_sb, &fattr);
@@ -1363,9 +1363,9 @@ mkdir_get_info:
 						 inode->i_sb, xid, NULL);
 
 		if (pTcon->nocase)
-			direntry->d_op = &cifs_ci_dentry_ops;
+			d_set_d_op(direntry, &cifs_ci_dentry_ops);
 		else
-			direntry->d_op = &cifs_dentry_ops;
+			d_set_d_op(direntry, &cifs_dentry_ops);
 		d_instantiate(direntry, newinode);
 		 /* setting nlink not necessary except in cases where we
 		  * failed to get it from the server or was set bogus */
diff --git a/fs/cifs/link.c b/fs/cifs/link.c
index 85cdbf8..fe2f6a9 100644
--- a/fs/cifs/link.c
+++ b/fs/cifs/link.c
@@ -525,9 +525,9 @@ cifs_symlink(struct inode *inode, struct dentry *direntry, const char *symname)
 			      rc);
 		} else {
 			if (pTcon->nocase)
-				direntry->d_op = &cifs_ci_dentry_ops;
+				d_set_d_op(direntry, &cifs_ci_dentry_ops);
 			else
-				direntry->d_op = &cifs_dentry_ops;
+				d_set_d_op(direntry, &cifs_dentry_ops);
 			d_instantiate(direntry, newinode);
 		}
 	}
diff --git a/fs/cifs/readdir.c b/fs/cifs/readdir.c
index ee463ae..ec5b68e 100644
--- a/fs/cifs/readdir.c
+++ b/fs/cifs/readdir.c
@@ -103,9 +103,9 @@ cifs_readdir_lookup(struct dentry *parent, struct qstr *name,
 	}
 
 	if (cifs_sb_master_tcon(CIFS_SB(sb))->nocase)
-		dentry->d_op = &cifs_ci_dentry_ops;
+		d_set_d_op(dentry, &cifs_ci_dentry_ops);
 	else
-		dentry->d_op = &cifs_dentry_ops;
+		d_set_d_op(dentry, &cifs_dentry_ops);
 
 	alias = d_materialise_unique(dentry, inode);
 	if (alias != NULL) {
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index 9e37e8b..aa40c81 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -125,7 +125,7 @@ static struct dentry *coda_lookup(struct inode *dir, struct dentry *entry, struc
 		return ERR_PTR(error);
 
 exit:
-	entry->d_op = &coda_dentry_operations;
+	d_set_d_op(entry, &coda_dentry_operations);
 
 	if (inode && (type & CODA_NOCACHE))
 		coda_flag_inode(inode, C_VATTR | C_PURGE);
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index e9acea4..36637a8 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -442,7 +442,7 @@ static int configfs_attach_attr(struct configfs_dirent * sd, struct dentry * den
 		return error;
 	}
 
-	dentry->d_op = &configfs_dentry_ops;
+	d_set_d_op(dentry, &configfs_dentry_ops);
 	d_rehash(dentry);
 
 	return 0;
@@ -489,7 +489,7 @@ static struct dentry * configfs_lookup(struct inode *dir,
 		 */
 		if (dentry->d_name.len > NAME_MAX)
 			return ERR_PTR(-ENAMETOOLONG);
-		dentry->d_op = &configfs_dentry_ops;
+		d_set_d_op(dentry, &configfs_dentry_ops);
 		d_add(dentry, NULL);
 		return NULL;
 	}
@@ -683,7 +683,7 @@ static int create_default_group(struct config_group *parent_group,
 	ret = -ENOMEM;
 	child = d_alloc(parent, &name);
 	if (child) {
-		child->d_op = &configfs_dentry_ops;
+		d_set_d_op(child, &configfs_dentry_ops);
 		d_add(child, NULL);
 
 		ret = configfs_attach_group(&parent_group->cg_item,
@@ -1681,7 +1681,7 @@ int configfs_register_subsystem(struct configfs_subsystem *subsys)
 	err = -ENOMEM;
 	dentry = d_alloc(configfs_sb->s_root, &name);
 	if (dentry) {
-		dentry->d_op = &configfs_dentry_ops;
+		d_set_d_op(dentry, &configfs_dentry_ops);
 		d_add(dentry, NULL);
 
 		err = configfs_attach_group(sd->s_element, &group->cg_item,
diff --git a/fs/dcache.c b/fs/dcache.c
index 1d5cf51..f9693da 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -398,7 +398,7 @@ repeat:
 		return;
 	}
 
-	if (dentry->d_op && dentry->d_op->d_delete) {
+	if (dentry->d_flags & DCACHE_OP_DELETE) {
 		if (dentry->d_op->d_delete(dentry))
 			goto kill_it;
 	}
@@ -1301,6 +1301,28 @@ struct dentry *d_alloc_name(struct dentry *parent, const char *name)
 }
 EXPORT_SYMBOL(d_alloc_name);
 
+void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
+{
+	BUG_ON(dentry->d_op);
+	BUG_ON(dentry->d_flags & (DCACHE_OP_HASH	|
+				DCACHE_OP_COMPARE	|
+				DCACHE_OP_REVALIDATE	|
+				DCACHE_OP_DELETE ));
+	dentry->d_op = op;
+	if (!op)
+		return;
+	if (op->d_hash)
+		dentry->d_flags |= DCACHE_OP_HASH;
+	if (op->d_compare)
+		dentry->d_flags |= DCACHE_OP_COMPARE;
+	if (op->d_revalidate)
+		dentry->d_flags |= DCACHE_OP_REVALIDATE;
+	if (op->d_delete)
+		dentry->d_flags |= DCACHE_OP_DELETE;
+
+}
+EXPORT_SYMBOL(d_set_d_op);
+
 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 {
 	spin_lock(&dentry->d_lock);
@@ -1731,7 +1753,7 @@ seqretry:
 		 */
 		if (read_seqcount_retry(&dentry->d_seq, *seq))
 			goto seqretry;
-		if (parent->d_op && parent->d_op->d_compare) {
+		if (parent->d_flags & DCACHE_OP_COMPARE) {
 			if (parent->d_op->d_compare(parent, *inode,
 						dentry, i,
 						tlen, tname, name))
@@ -1846,7 +1868,7 @@ struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
 		 */
 		tlen = dentry->d_name.len;
 		tname = dentry->d_name.name;
-		if (parent->d_op && parent->d_op->d_compare) {
+		if (parent->d_flags & DCACHE_OP_COMPARE) {
 			if (parent->d_op->d_compare(parent, parent->d_inode,
 						dentry, dentry->d_inode,
 						tlen, tname, name))
@@ -1887,7 +1909,7 @@ struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
 	 * routine may choose to leave the hash value unchanged.
 	 */
 	name->hash = full_name_hash(name->name, name->len);
-	if (dir->d_op && dir->d_op->d_hash) {
+	if (dir->d_flags & DCACHE_OP_HASH) {
 		if (dir->d_op->d_hash(dir, dir->d_inode, name) < 0)
 			goto out;
 	}
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index 5e5c7ec..f91b35d 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -441,7 +441,7 @@ static struct dentry *ecryptfs_lookup(struct inode *ecryptfs_dir_inode,
 	struct qstr lower_name;
 	int rc = 0;
 
-	ecryptfs_dentry->d_op = &ecryptfs_dops;
+	d_set_d_op(ecryptfs_dentry, &ecryptfs_dops);
 	if ((ecryptfs_dentry->d_name.len == 1
 	     && !strcmp(ecryptfs_dentry->d_name.name, "."))
 	    || (ecryptfs_dentry->d_name.len == 2
diff --git a/fs/ecryptfs/main.c b/fs/ecryptfs/main.c
index a9dbd62..3510386 100644
--- a/fs/ecryptfs/main.c
+++ b/fs/ecryptfs/main.c
@@ -189,7 +189,7 @@ int ecryptfs_interpose(struct dentry *lower_dentry, struct dentry *dentry,
 	if (special_file(lower_inode->i_mode))
 		init_special_inode(inode, lower_inode->i_mode,
 				   lower_inode->i_rdev);
-	dentry->d_op = &ecryptfs_dops;
+	d_set_d_op(dentry, &ecryptfs_dops);
 	fsstack_copy_attr_all(inode, lower_inode);
 	/* This size will be overwritten for real files w/ headers and
 	 * other metadata */
@@ -594,7 +594,7 @@ static struct dentry *ecryptfs_mount(struct file_system_type *fs_type, int flags
 		deactivate_locked_super(s);
 		goto out;
 	}
-	s->s_root->d_op = &ecryptfs_dops;
+	d_set_d_op(s->s_root, &ecryptfs_dops);
 	s->s_root->d_sb = s;
 	s->s_root->d_parent = s->s_root;
 
diff --git a/fs/fat/inode.c b/fs/fat/inode.c
index 8cccfeb..206351a 100644
--- a/fs/fat/inode.c
+++ b/fs/fat/inode.c
@@ -750,7 +750,7 @@ static struct dentry *fat_fh_to_dentry(struct super_block *sb,
 	 */
 	result = d_obtain_alias(inode);
 	if (!IS_ERR(result))
-		result->d_op = sb->s_root->d_op;
+		d_set_d_op(result, sb->s_root->d_op);
 	return result;
 }
 
@@ -800,7 +800,7 @@ static struct dentry *fat_get_parent(struct dentry *child)
 
 	parent = d_obtain_alias(inode);
 	if (!IS_ERR(parent))
-		parent->d_op = sb->s_root->d_op;
+		d_set_d_op(parent, sb->s_root->d_op);
 out:
 	unlock_super(sb);
 
diff --git a/fs/fat/namei_msdos.c b/fs/fat/namei_msdos.c
index 3b3e072..35ffe43 100644
--- a/fs/fat/namei_msdos.c
+++ b/fs/fat/namei_msdos.c
@@ -227,10 +227,10 @@ static struct dentry *msdos_lookup(struct inode *dir, struct dentry *dentry,
 	}
 out:
 	unlock_super(sb);
-	dentry->d_op = &msdos_dentry_operations;
+	d_set_d_op(dentry, &msdos_dentry_operations);
 	dentry = d_splice_alias(inode, dentry);
 	if (dentry)
-		dentry->d_op = &msdos_dentry_operations;
+		d_set_d_op(dentry, &msdos_dentry_operations);
 	return dentry;
 
 error:
@@ -673,7 +673,7 @@ static int msdos_fill_super(struct super_block *sb, void *data, int silent)
 	}
 
 	sb->s_flags |= MS_NOATIME;
-	sb->s_root->d_op = &msdos_dentry_operations;
+	d_set_d_op(sb->s_root, &msdos_dentry_operations);
 	unlock_super(sb);
 	return 0;
 }
diff --git a/fs/fat/namei_vfat.c b/fs/fat/namei_vfat.c
index 4fc0627..3be5ed7 100644
--- a/fs/fat/namei_vfat.c
+++ b/fs/fat/namei_vfat.c
@@ -766,11 +766,11 @@ static struct dentry *vfat_lookup(struct inode *dir, struct dentry *dentry,
 
 out:
 	unlock_super(sb);
-	dentry->d_op = sb->s_root->d_op;
+	d_set_d_op(dentry, sb->s_root->d_op);
 	dentry->d_time = dentry->d_parent->d_inode->i_version;
 	dentry = d_splice_alias(inode, dentry);
 	if (dentry) {
-		dentry->d_op = sb->s_root->d_op;
+		d_set_d_op(dentry, sb->s_root->d_op);
 		dentry->d_time = dentry->d_parent->d_inode->i_version;
 	}
 	return dentry;
@@ -1072,9 +1072,9 @@ static int vfat_fill_super(struct super_block *sb, void *data, int silent)
 	}
 
 	if (MSDOS_SB(sb)->options.name_check != 's')
-		sb->s_root->d_op = &vfat_ci_dentry_ops;
+		d_set_d_op(sb->s_root, &vfat_ci_dentry_ops);
 	else
-		sb->s_root->d_op = &vfat_dentry_ops;
+		d_set_d_op(sb->s_root, &vfat_dentry_ops);
 
 	unlock_super(sb);
 	return 0;
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index c9627c9..c9a8a42 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -347,7 +347,7 @@ static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry,
 	}
 
 	entry = newent ? newent : entry;
-	entry->d_op = &fuse_dentry_operations;
+	d_set_d_op(entry, &fuse_dentry_operations);
 	if (outarg_valid)
 		fuse_change_entry_timeout(entry, &outarg);
 	else
diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index 44e0a6c..a8b31da 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -626,7 +626,7 @@ static struct dentry *fuse_get_dentry(struct super_block *sb,
 
 	entry = d_obtain_alias(inode);
 	if (!IS_ERR(entry) && get_node_id(inode) != FUSE_ROOT_ID) {
-		entry->d_op = &fuse_dentry_operations;
+		d_set_d_op(entry, &fuse_dentry_operations);
 		fuse_invalidate_entry_cache(entry);
 	}
 
@@ -728,7 +728,7 @@ static struct dentry *fuse_get_parent(struct dentry *child)
 
 	parent = d_obtain_alias(inode);
 	if (!IS_ERR(parent) && get_node_id(inode) != FUSE_ROOT_ID) {
-		parent->d_op = &fuse_dentry_operations;
+		d_set_d_op(parent, &fuse_dentry_operations);
 		fuse_invalidate_entry_cache(parent);
 	}
 
diff --git a/fs/gfs2/export.c b/fs/gfs2/export.c
index 5ab3839..97012ec 100644
--- a/fs/gfs2/export.c
+++ b/fs/gfs2/export.c
@@ -130,7 +130,7 @@ static struct dentry *gfs2_get_parent(struct dentry *child)
 
 	dentry = d_obtain_alias(gfs2_lookupi(child->d_inode, &gfs2_qdotdot, 1));
 	if (!IS_ERR(dentry))
-		dentry->d_op = &gfs2_dops;
+		d_set_d_op(dentry, &gfs2_dops);
 	return dentry;
 }
 
@@ -158,7 +158,7 @@ static struct dentry *gfs2_get_dentry(struct super_block *sb,
 out_inode:
 	dentry = d_obtain_alias(inode);
 	if (!IS_ERR(dentry))
-		dentry->d_op = &gfs2_dops;
+		d_set_d_op(dentry, &gfs2_dops);
 	return dentry;
 }
 
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 3eb1393..2aeabd4 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -440,7 +440,7 @@ static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
 		iput(inode);
 		return -ENOMEM;
 	}
-	dentry->d_op = &gfs2_dops;
+	d_set_d_op(dentry, &gfs2_dops);
 	*dptr = dentry;
 	return 0;
 }
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index 12cbea7..f28f897 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -106,7 +106,7 @@ static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
 {
 	struct inode *inode = NULL;
 
-	dentry->d_op = &gfs2_dops;
+	d_set_d_op(dentry, &gfs2_dops);
 
 	inode = gfs2_lookupi(dir, &dentry->d_name, 0);
 	if (inode && IS_ERR(inode))
diff --git a/fs/hfs/dir.c b/fs/hfs/dir.c
index 2b3b861..ea4aefe 100644
--- a/fs/hfs/dir.c
+++ b/fs/hfs/dir.c
@@ -25,7 +25,7 @@ static struct dentry *hfs_lookup(struct inode *dir, struct dentry *dentry,
 	struct inode *inode = NULL;
 	int res;
 
-	dentry->d_op = &hfs_dentry_operations;
+	d_set_d_op(dentry, &hfs_dentry_operations);
 
 	hfs_find_init(HFS_SB(dir->i_sb)->cat_tree, &fd);
 	hfs_cat_build_key(dir->i_sb, fd.search_key, dir->i_ino, &dentry->d_name);
diff --git a/fs/hfs/super.c b/fs/hfs/super.c
index ef4ee57..0bef62a 100644
--- a/fs/hfs/super.c
+++ b/fs/hfs/super.c
@@ -434,7 +434,7 @@ static int hfs_fill_super(struct super_block *sb, void *data, int silent)
 	if (!sb->s_root)
 		goto bail_iput;
 
-	sb->s_root->d_op = &hfs_dentry_operations;
+	d_set_d_op(sb->s_root, &hfs_dentry_operations);
 
 	/* everything's okay */
 	return 0;
diff --git a/fs/hfsplus/dir.c b/fs/hfsplus/dir.c
index 9d59c05..ccab871 100644
--- a/fs/hfsplus/dir.c
+++ b/fs/hfsplus/dir.c
@@ -37,7 +37,7 @@ static struct dentry *hfsplus_lookup(struct inode *dir, struct dentry *dentry,
 
 	sb = dir->i_sb;
 
-	dentry->d_op = &hfsplus_dentry_operations;
+	d_set_d_op(dentry, &hfsplus_dentry_operations);
 	dentry->d_fsdata = NULL;
 	hfs_find_init(HFSPLUS_SB(sb)->cat_tree, &fd);
 	hfsplus_cat_build_key(sb, fd.search_key, dir->i_ino, &dentry->d_name);
diff --git a/fs/hfsplus/super.c b/fs/hfsplus/super.c
index 182e83a..ddf712e 100644
--- a/fs/hfsplus/super.c
+++ b/fs/hfsplus/super.c
@@ -419,7 +419,7 @@ static int hfsplus_fill_super(struct super_block *sb, void *data, int silent)
 		err = -ENOMEM;
 		goto cleanup;
 	}
-	sb->s_root->d_op = &hfsplus_dentry_operations;
+	d_set_d_op(sb->s_root, &hfsplus_dentry_operations);
 
 	str.len = sizeof(HFSP_HIDDENDIR_NAME) - 1;
 	str.name = HFSP_HIDDENDIR_NAME;
diff --git a/fs/hostfs/hostfs_kern.c b/fs/hostfs/hostfs_kern.c
index 861113f..0bc81cf 100644
--- a/fs/hostfs/hostfs_kern.c
+++ b/fs/hostfs/hostfs_kern.c
@@ -612,7 +612,7 @@ struct dentry *hostfs_lookup(struct inode *ino, struct dentry *dentry,
 		goto out_put;
 
 	d_add(dentry, inode);
-	dentry->d_op = &hostfs_dentry_ops;
+	d_set_d_op(dentry, &hostfs_dentry_ops);
 	return NULL;
 
  out_put:
diff --git a/fs/hpfs/dentry.c b/fs/hpfs/dentry.c
index 35526df..32c13a9 100644
--- a/fs/hpfs/dentry.c
+++ b/fs/hpfs/dentry.c
@@ -65,5 +65,5 @@ static const struct dentry_operations hpfs_dentry_operations = {
 
 void hpfs_set_dentry_operations(struct dentry *dentry)
 {
-	dentry->d_op = &hpfs_dentry_operations;
+	d_set_d_op(dentry, &hpfs_dentry_operations);
 }
diff --git a/fs/isofs/inode.c b/fs/isofs/inode.c
index d8f3a65..844a790 100644
--- a/fs/isofs/inode.c
+++ b/fs/isofs/inode.c
@@ -949,7 +949,7 @@ root_found:
 		table += 2;
 	if (opt.check == 'r')
 		table++;
-	s->s_root->d_op = &isofs_dentry_ops[table];
+	d_set_d_op(s->s_root, &isofs_dentry_ops[table]);
 
 	kfree(opt.iocharset);
 
diff --git a/fs/isofs/namei.c b/fs/isofs/namei.c
index 715f7d3..679a849 100644
--- a/fs/isofs/namei.c
+++ b/fs/isofs/namei.c
@@ -172,7 +172,7 @@ struct dentry *isofs_lookup(struct inode *dir, struct dentry *dentry, struct nam
 	struct inode *inode;
 	struct page *page;
 
-	dentry->d_op = dir->i_sb->s_root->d_op;
+	d_set_d_op(dentry, dir->i_sb->s_root->d_op);
 
 	page = alloc_page(GFP_USER);
 	if (!page)
diff --git a/fs/jfs/namei.c b/fs/jfs/namei.c
index 57f90da..a151cbd 100644
--- a/fs/jfs/namei.c
+++ b/fs/jfs/namei.c
@@ -1466,7 +1466,7 @@ static struct dentry *jfs_lookup(struct inode *dip, struct dentry *dentry, struc
 	jfs_info("jfs_lookup: name = %s", name);
 
 	if (JFS_SBI(dip->i_sb)->mntflag & JFS_OS2)
-		dentry->d_op = &jfs_ci_dentry_operations;
+		d_set_d_op(dentry, &jfs_ci_dentry_operations);
 
 	if ((name[0] == '.') && (len == 1))
 		inum = dip->i_ino;
@@ -1495,7 +1495,7 @@ static struct dentry *jfs_lookup(struct inode *dip, struct dentry *dentry, struc
 	dentry = d_splice_alias(ip, dentry);
 
 	if (dentry && (JFS_SBI(dip->i_sb)->mntflag & JFS_OS2))
-		dentry->d_op = &jfs_ci_dentry_operations;
+		d_set_d_op(dentry, &jfs_ci_dentry_operations);
 
 	return dentry;
 }
diff --git a/fs/jfs/super.c b/fs/jfs/super.c
index b715b0f..3150d76 100644
--- a/fs/jfs/super.c
+++ b/fs/jfs/super.c
@@ -525,7 +525,7 @@ static int jfs_fill_super(struct super_block *sb, void *data, int silent)
 		goto out_no_root;
 
 	if (sbi->mntflag & JFS_OS2)
-		sb->s_root->d_op = &jfs_ci_dentry_operations;
+		d_set_d_op(sb->s_root, &jfs_ci_dentry_operations);
 
 	/* logical blocks are represented by 40 bits in pxd_t, etc. */
 	sb->s_maxbytes = ((u64) sb->s_blocksize) << 40;
diff --git a/fs/libfs.c b/fs/libfs.c
index 28b3666..889311e 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -59,7 +59,7 @@ struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct na
 
 	if (dentry->d_name.len > NAME_MAX)
 		return ERR_PTR(-ENAMETOOLONG);
-	dentry->d_op = &simple_dentry_operations;
+	d_set_d_op(dentry, &simple_dentry_operations);
 	d_add(dentry, NULL);
 	return NULL;
 }
diff --git a/fs/minix/namei.c b/fs/minix/namei.c
index c0d35a3..1b9e077 100644
--- a/fs/minix/namei.c
+++ b/fs/minix/namei.c
@@ -23,7 +23,7 @@ static struct dentry *minix_lookup(struct inode * dir, struct dentry *dentry, st
 	struct inode * inode = NULL;
 	ino_t ino;
 
-	dentry->d_op = dir->i_sb->s_root->d_op;
+	d_set_d_op(dentry, dir->i_sb->s_root->d_op);
 
 	if (dentry->d_name.len > minix_sb(dir->i_sb)->s_namelen)
 		return ERR_PTR(-ENAMETOOLONG);
diff --git a/fs/namei.c b/fs/namei.c
index c731b50..90bd287 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -587,6 +587,17 @@ do_revalidate(struct dentry *dentry, struct nameidata *nd)
 	return dentry;
 }
 
+static inline int need_reval_dot(struct dentry *dentry)
+{
+	if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE)))
+		return 0;
+
+	if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)))
+		return 0;
+
+	return 1;
+}
+
 /*
  * force_reval_path - force revalidation of a dentry
  *
@@ -610,10 +621,9 @@ force_reval_path(struct path *path, struct nameidata *nd)
 
 	/*
 	 * only check on filesystems where it's possible for the dentry to
-	 * become stale. It's assumed that if this flag is set then the
-	 * d_revalidate op will also be defined.
+	 * become stale.
 	 */
-	if (!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))
+	if (!need_reval_dot(dentry))
 		return 0;
 
 	status = dentry->d_op->d_revalidate(dentry, nd);
@@ -1003,7 +1013,7 @@ static int do_lookup(struct nameidata *nd, struct qstr *name,
 	 * See if the low-level filesystem might want
 	 * to use its own hash..
 	 */
-	if (parent->d_op && parent->d_op->d_hash) {
+	if (unlikely(parent->d_flags & DCACHE_OP_HASH)) {
 		int err = parent->d_op->d_hash(parent, nd->inode, name);
 		if (err < 0)
 			return err;
@@ -1029,7 +1039,7 @@ static int do_lookup(struct nameidata *nd, struct qstr *name,
 			return -ECHILD;
 
 		nd->seq = seq;
-		if (dentry->d_op && dentry->d_op->d_revalidate) {
+		if (dentry->d_flags & DCACHE_OP_REVALIDATE) {
 			/* We commonly drop rcu-walk here */
 			if (nameidata_dentry_drop_rcu(nd, dentry))
 				return -ECHILD;
@@ -1043,7 +1053,7 @@ static int do_lookup(struct nameidata *nd, struct qstr *name,
 		if (!dentry)
 			goto need_lookup;
 found:
-		if (dentry->d_op && dentry->d_op->d_revalidate)
+		if (dentry->d_flags & DCACHE_OP_REVALIDATE)
 			goto need_revalidate;
 done:
 		path->mnt = mnt;
@@ -1281,8 +1291,7 @@ return_reval:
 		 * We bypassed the ordinary revalidation routines.
 		 * We may need to check the cached dentry for staleness.
 		 */
-		if (nd->path.dentry && nd->path.dentry->d_sb &&
-		    (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) {
+		if (need_reval_dot(nd->path.dentry)) {
 			if (nameidata_drop_rcu_maybe(nd))
 				return -ECHILD;
 			err = -ESTALE;
@@ -1602,7 +1611,7 @@ static struct dentry *__lookup_hash(struct qstr *name,
 	 * See if the low-level filesystem might want
 	 * to use its own hash..
 	 */
-	if (base->d_op && base->d_op->d_hash) {
+	if (base->d_flags & DCACHE_OP_HASH) {
 		err = base->d_op->d_hash(base, inode, name);
 		dentry = ERR_PTR(err);
 		if (err < 0)
@@ -1616,7 +1625,7 @@ static struct dentry *__lookup_hash(struct qstr *name,
 	 */
 	dentry = d_lookup(base, name);
 
-	if (dentry && dentry->d_op && dentry->d_op->d_revalidate)
+	if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE))
 		dentry = do_revalidate(dentry, nd);
 
 	if (!dentry)
@@ -2070,7 +2079,7 @@ static struct file *do_last(struct nameidata *nd, struct path *path,
 		follow_dotdot(nd);
 		dir = nd->path.dentry;
 	case LAST_DOT:
-		if (nd->path.mnt->mnt_sb->s_type->fs_flags & FS_REVAL_DOT) {
+		if (need_reval_dot(dir)) {
 			if (!dir->d_op->d_revalidate(dir, nd)) {
 				error = -ESTALE;
 				goto exit;
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index 0ba3cdc..4b9cbb2 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -633,7 +633,7 @@ ncp_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
 		entry->ino = iunique(dir->i_sb, 2);
 		inode = ncp_iget(dir->i_sb, entry);
 		if (inode) {
-			newdent->d_op = &ncp_dentry_operations;
+			d_set_d_op(newdent, &ncp_dentry_operations);
 			d_instantiate(newdent, inode);
 			if (!hashed)
 				d_rehash(newdent);
@@ -889,7 +889,7 @@ static struct dentry *ncp_lookup(struct inode *dir, struct dentry *dentry, struc
 	if (inode) {
 		ncp_new_dentry(dentry);
 add_entry:
-		dentry->d_op = &ncp_dentry_operations;
+		d_set_d_op(dentry, &ncp_dentry_operations);
 		d_add(dentry, inode);
 		error = 0;
 	}
diff --git a/fs/ncpfs/inode.c b/fs/ncpfs/inode.c
index 60047db..0c75a5f 100644
--- a/fs/ncpfs/inode.c
+++ b/fs/ncpfs/inode.c
@@ -717,7 +717,7 @@ static int ncp_fill_super(struct super_block *sb, void *raw_data, int silent)
 	sb->s_root = d_alloc_root(root_inode);
         if (!sb->s_root)
 		goto out_no_root;
-	sb->s_root->d_op = &ncp_root_dentry_operations;
+	d_set_d_op(sb->s_root, &ncp_root_dentry_operations);
 	return 0;
 
 out_no_root:
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index eb77471..37e0a8b 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -438,7 +438,7 @@ void nfs_prime_dcache(struct dentry *parent, struct nfs_entry *entry)
 	if (dentry == NULL)
 		return;
 
-	dentry->d_op = NFS_PROTO(dir)->dentry_ops;
+	d_set_d_op(dentry, NFS_PROTO(dir)->dentry_ops);
 	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr);
 	if (IS_ERR(inode))
 		goto out;
@@ -1188,7 +1188,7 @@ static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, stru
 	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
 		goto out;
 
-	dentry->d_op = NFS_PROTO(dir)->dentry_ops;
+	d_set_d_op(dentry, NFS_PROTO(dir)->dentry_ops);
 
 	/*
 	 * If we're doing an exclusive create, optimize away the lookup
@@ -1333,7 +1333,7 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry
 		res = ERR_PTR(-ENAMETOOLONG);
 		goto out;
 	}
-	dentry->d_op = NFS_PROTO(dir)->dentry_ops;
+	d_set_d_op(dentry, NFS_PROTO(dir)->dentry_ops);
 
 	/* Let vfs_create() deal with O_EXCL. Instantiate, but don't hash
 	 * the dentry. */
diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c
index b3e36c3..c3a5a11 100644
--- a/fs/nfs/getroot.c
+++ b/fs/nfs/getroot.c
@@ -121,7 +121,7 @@ struct dentry *nfs_get_root(struct super_block *sb, struct nfs_fh *mntfh)
 	security_d_instantiate(ret, inode);
 
 	if (ret->d_op == NULL)
-		ret->d_op = server->nfs_client->rpc_ops->dentry_ops;
+		d_set_d_op(ret, server->nfs_client->rpc_ops->dentry_ops);
 out:
 	nfs_free_fattr(fsinfo.fattr);
 	return ret;
@@ -228,7 +228,7 @@ struct dentry *nfs4_get_root(struct super_block *sb, struct nfs_fh *mntfh)
 	security_d_instantiate(ret, inode);
 
 	if (ret->d_op == NULL)
-		ret->d_op = server->nfs_client->rpc_ops->dentry_ops;
+		d_set_d_op(ret, server->nfs_client->rpc_ops->dentry_ops);
 
 out:
 	nfs_free_fattr(fattr);
diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c
index 19ad145..6adafa5 100644
--- a/fs/ocfs2/export.c
+++ b/fs/ocfs2/export.c
@@ -138,7 +138,7 @@ check_gen:
 
 	result = d_obtain_alias(inode);
 	if (!IS_ERR(result))
-		result->d_op = &ocfs2_dentry_ops;
+		d_set_d_op(result, &ocfs2_dentry_ops);
 	else
 		mlog_errno(PTR_ERR(result));
 
@@ -176,7 +176,7 @@ static struct dentry *ocfs2_get_parent(struct dentry *child)
 
 	parent = d_obtain_alias(ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0, 0));
 	if (!IS_ERR(parent))
-		parent->d_op = &ocfs2_dentry_ops;
+		d_set_d_op(parent, &ocfs2_dentry_ops);
 
 bail_unlock:
 	ocfs2_inode_unlock(dir, 0);
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index ff5744e..d14cad6 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -147,7 +147,7 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry,
 	spin_unlock(&oi->ip_lock);
 
 bail_add:
-	dentry->d_op = &ocfs2_dentry_ops;
+	d_set_d_op(dentry, &ocfs2_dentry_ops);
 	ret = d_splice_alias(inode, dentry);
 
 	if (inode) {
@@ -415,7 +415,7 @@ static int ocfs2_mknod(struct inode *dir,
 		mlog_errno(status);
 		goto leave;
 	}
-	dentry->d_op = &ocfs2_dentry_ops;
+	d_set_d_op(dentry, &ocfs2_dentry_ops);
 
 	status = ocfs2_add_entry(handle, dentry, inode,
 				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
@@ -743,7 +743,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 	}
 
 	ihold(inode);
-	dentry->d_op = &ocfs2_dentry_ops;
+	d_set_d_op(dentry, &ocfs2_dentry_ops);
 	d_instantiate(dentry, inode);
 
 out_commit:
@@ -1794,7 +1794,7 @@ static int ocfs2_symlink(struct inode *dir,
 		mlog_errno(status);
 		goto bail;
 	}
-	dentry->d_op = &ocfs2_dentry_ops;
+	d_set_d_op(dentry, &ocfs2_dentry_ops);
 
 	status = ocfs2_add_entry(handle, dentry, inode,
 				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
@@ -2459,7 +2459,7 @@ int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
 		goto out_commit;
 	}
 
-	dentry->d_op = &ocfs2_dentry_ops;
+	d_set_d_op(dentry, &ocfs2_dentry_ops);
 	d_instantiate(dentry, inode);
 	status = 0;
 out_commit:
diff --git a/fs/pipe.c b/fs/pipe.c
index ae3592d..01a7865 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -1004,7 +1004,7 @@ struct file *create_write_pipe(int flags)
 		goto err_inode;
 	path.mnt = mntget(pipe_mnt);
 
-	path.dentry->d_op = &pipefs_dentry_operations;
+	d_set_d_op(path.dentry, &pipefs_dentry_operations);
 	d_instantiate(path.dentry, inode);
 
 	err = -ENFILE;
diff --git a/fs/proc/base.c b/fs/proc/base.c
index d932fdb..85f0a80 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1969,7 +1969,7 @@ static struct dentry *proc_fd_instantiate(struct inode *dir,
 	inode->i_op = &proc_pid_link_inode_operations;
 	inode->i_size = 64;
 	ei->op.proc_get_link = proc_fd_link;
-	dentry->d_op = &tid_fd_dentry_operations;
+	d_set_d_op(dentry, &tid_fd_dentry_operations);
 	d_add(dentry, inode);
 	/* Close the race of the process dying before we return the dentry */
 	if (tid_fd_revalidate(dentry, NULL))
@@ -2137,7 +2137,7 @@ static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
 	ei->fd = fd;
 	inode->i_mode = S_IFREG | S_IRUSR;
 	inode->i_fop = &proc_fdinfo_file_operations;
-	dentry->d_op = &tid_fd_dentry_operations;
+	d_set_d_op(dentry, &tid_fd_dentry_operations);
 	d_add(dentry, inode);
 	/* Close the race of the process dying before we return the dentry */
 	if (tid_fd_revalidate(dentry, NULL))
@@ -2196,7 +2196,7 @@ static struct dentry *proc_pident_instantiate(struct inode *dir,
 	if (p->fop)
 		inode->i_fop = p->fop;
 	ei->op = p->op;
-	dentry->d_op = &pid_dentry_operations;
+	d_set_d_op(dentry, &pid_dentry_operations);
 	d_add(dentry, inode);
 	/* Close the race of the process dying before we return the dentry */
 	if (pid_revalidate(dentry, NULL))
@@ -2615,7 +2615,7 @@ static struct dentry *proc_base_instantiate(struct inode *dir,
 	if (p->fop)
 		inode->i_fop = p->fop;
 	ei->op = p->op;
-	dentry->d_op = &proc_base_dentry_operations;
+	d_set_d_op(dentry, &proc_base_dentry_operations);
 	d_add(dentry, inode);
 	error = NULL;
 out:
@@ -2926,7 +2926,7 @@ static struct dentry *proc_pid_instantiate(struct inode *dir,
 	inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
 		ARRAY_SIZE(tgid_base_stuff));
 
-	dentry->d_op = &pid_dentry_operations;
+	d_set_d_op(dentry, &pid_dentry_operations);
 
 	d_add(dentry, inode);
 	/* Close the race of the process dying before we return the dentry */
@@ -3169,7 +3169,7 @@ static struct dentry *proc_task_instantiate(struct inode *dir,
 	inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
 		ARRAY_SIZE(tid_base_stuff));
 
-	dentry->d_op = &pid_dentry_operations;
+	d_set_d_op(dentry, &pid_dentry_operations);
 
 	d_add(dentry, inode);
 	/* Close the race of the process dying before we return the dentry */
diff --git a/fs/proc/generic.c b/fs/proc/generic.c
index 1d607be..f766be2 100644
--- a/fs/proc/generic.c
+++ b/fs/proc/generic.c
@@ -439,7 +439,7 @@ struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
 out_unlock:
 
 	if (inode) {
-		dentry->d_op = &proc_dentry_operations;
+		d_set_d_op(dentry, &proc_dentry_operations);
 		d_add(dentry, inode);
 		return NULL;
 	}
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 998e3a7..35efd85 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -120,7 +120,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 		goto out;
 
 	err = NULL;
-	dentry->d_op = &proc_sys_dentry_operations;
+	d_set_d_op(dentry, &proc_sys_dentry_operations);
 	d_add(dentry, inode);
 
 out:
@@ -201,7 +201,7 @@ static int proc_sys_fill_cache(struct file *filp, void *dirent,
 				dput(child);
 				return -ENOMEM;
 			} else {
-				child->d_op = &proc_sys_dentry_operations;
+				d_set_d_op(child, &proc_sys_dentry_operations);
 				d_add(child, inode);
 			}
 		} else {
diff --git a/fs/reiserfs/xattr.c b/fs/reiserfs/xattr.c
index 5d04a78..e0f0d7e 100644
--- a/fs/reiserfs/xattr.c
+++ b/fs/reiserfs/xattr.c
@@ -990,7 +990,7 @@ int reiserfs_lookup_privroot(struct super_block *s)
 				strlen(PRIVROOT_NAME));
 	if (!IS_ERR(dentry)) {
 		REISERFS_SB(s)->priv_root = dentry;
-		dentry->d_op = &xattr_lookup_poison_ops;
+		d_set_d_op(dentry, &xattr_lookup_poison_ops);
 		if (dentry->d_inode)
 			dentry->d_inode->i_flags |= S_PRIVATE;
 	} else
diff --git a/fs/sysfs/dir.c b/fs/sysfs/dir.c
index 27e1102..3e076ca 100644
--- a/fs/sysfs/dir.c
+++ b/fs/sysfs/dir.c
@@ -701,7 +701,7 @@ static struct dentry * sysfs_lookup(struct inode *dir, struct dentry *dentry,
 	/* instantiate and hash dentry */
 	ret = d_find_alias(inode);
 	if (!ret) {
-		dentry->d_op = &sysfs_dentry_ops;
+		d_set_d_op(dentry, &sysfs_dentry_ops);
 		dentry->d_fsdata = sysfs_get(sd);
 		d_add(dentry, inode);
 	} else {
diff --git a/fs/sysv/namei.c b/fs/sysv/namei.c
index 7507aeb..b5e68da 100644
--- a/fs/sysv/namei.c
+++ b/fs/sysv/namei.c
@@ -48,7 +48,7 @@ static struct dentry *sysv_lookup(struct inode * dir, struct dentry * dentry, st
 	struct inode * inode = NULL;
 	ino_t ino;
 
-	dentry->d_op = dir->i_sb->s_root->d_op;
+	d_set_d_op(dentry, dir->i_sb->s_root->d_op);
 	if (dentry->d_name.len > SYSV_NAMELEN)
 		return ERR_PTR(-ENAMETOOLONG);
 	ino = sysv_inode_by_name(dentry);
diff --git a/fs/sysv/super.c b/fs/sysv/super.c
index 3d9c62b..76712ae 100644
--- a/fs/sysv/super.c
+++ b/fs/sysv/super.c
@@ -346,7 +346,7 @@ static int complete_read_super(struct super_block *sb, int silent, int size)
 	if (sbi->s_forced_ro)
 		sb->s_flags |= MS_RDONLY;
 	if (sbi->s_truncate)
-		sb->s_root->d_op = &sysv_dentry_operations;
+		d_set_d_op(sb->s_root, &sysv_dentry_operations);
 	return 1;
 }
 
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index e441469..f4b40a7 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -183,6 +183,11 @@ struct dentry_operations {
 #define DCACHE_GENOCIDE		0x0200
 #define DCACHE_MOUNTED		0x0400	/* is a mountpoint */
 
+#define DCACHE_OP_HASH		0x1000
+#define DCACHE_OP_COMPARE	0x2000
+#define DCACHE_OP_REVALIDATE	0x4000
+#define DCACHE_OP_DELETE	0x8000
+
 
 extern spinlock_t dcache_inode_lock;
 extern seqlock_t rename_lock;
@@ -201,6 +206,7 @@ extern struct dentry * d_materialise_unique(struct dentry *, struct inode *);
 extern void __d_drop(struct dentry *dentry);
 extern void d_drop(struct dentry *dentry);
 extern void d_delete(struct dentry *);
+extern void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op);
 
 /* allocate/de-allocate */
 extern struct dentry * d_alloc(struct dentry *, const struct qstr *);
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 9f41470..51cddc1 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -2222,7 +2222,7 @@ static struct dentry *cgroup_lookup(struct inode *dir,
 
 	if (dentry->d_name.len > NAME_MAX)
 		return ERR_PTR(-ENAMETOOLONG);
-	dentry->d_op = &cgroup_dentry_operations;
+	d_set_d_op(dentry, &cgroup_dentry_operations);
 	d_add(dentry, NULL);
 	return NULL;
 }
diff --git a/net/socket.c b/net/socket.c
index 817dc92..991e266 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -368,7 +368,7 @@ static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 	}
 	path.mnt = mntget(sock_mnt);
 
-	path.dentry->d_op = &sockfs_dentry_operations;
+	d_set_d_op(path.dentry, &sockfs_dentry_operations);
 	d_instantiate(path.dentry, SOCK_INODE(sock));
 	SOCK_INODE(sock)->i_fop = &socket_file_ops;
 
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index 2899fe2..09f01f4 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -591,7 +591,7 @@ static struct dentry *__rpc_lookup_create(struct dentry *parent,
 		}
 	}
 	if (!dentry->d_inode)
-		dentry->d_op = &rpc_dentry_operations;
+		d_set_d_op(dentry, &rpc_dentry_operations);
 out_err:
 	return dentry;
 }
-- 
cgit v0.10.2


From 44a7d7a878c9cbb74f236ea755b25b6b2e26a9a9 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:56 +1100
Subject: fs: cache optimise dentry and inode for rcu-walk

Put dentry and inode fields into top of data structure.  This allows RCU path
traversal to perform an RCU dentry lookup in a path walk by touching only the
first 56 bytes of the dentry.

We also fit in 8 bytes of inline name in the first 64 bytes, so for short
names, only 64 bytes needs to be touched to perform the lookup. We should
get rid of the hash->prev pointer from the first 64 bytes, and fit 16 bytes
of name in there, which will take care of 81% rather than 32% of the kernel
tree.

inode is also rearranged so that RCU lookup will only touch a single cacheline
in the inode, plus one in the i_ops structure.

This is important for directory component lookups in RCU path walking. In the
kernel source, directory names average is around 6 chars, so this works.

When we reach the last element of the lookup, we need to lock it and take its
refcount which requires another cacheline access.

Align dentry and inode operations structs, so members will be at predictable
offsets and we can group common operations into head of structure.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index f9693da..07d1f68 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -83,8 +83,6 @@ EXPORT_SYMBOL(dcache_inode_lock);
 
 static struct kmem_cache *dentry_cache __read_mostly;
 
-#define DNAME_INLINE_LEN (sizeof(struct dentry)-offsetof(struct dentry,d_iname))
-
 /*
  * This is the single most critical data structure when it comes
  * to the dcache: the hashtable for lookups. Somebody should try
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index f4b40a7..b1aeda0 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -82,25 +82,33 @@ full_name_hash(const unsigned char *name, unsigned int len)
  * large memory footprint increase).
  */
 #ifdef CONFIG_64BIT
-#define DNAME_INLINE_LEN_MIN 32 /* 192 bytes */
+# define DNAME_INLINE_LEN 32 /* 192 bytes */
 #else
-#define DNAME_INLINE_LEN_MIN 40 /* 128 bytes */
+# ifdef CONFIG_SMP
+#  define DNAME_INLINE_LEN 36 /* 128 bytes */
+# else
+#  define DNAME_INLINE_LEN 40 /* 128 bytes */
+# endif
 #endif
 
 struct dentry {
-	unsigned int d_count;		/* protected by d_lock */
+	/* RCU lookup touched fields */
 	unsigned int d_flags;		/* protected by d_lock */
-	spinlock_t d_lock;		/* per dentry lock */
 	seqcount_t d_seq;		/* per dentry seqlock */
-	struct inode *d_inode;		/* Where the name belongs to - NULL is
-					 * negative */
-	/*
-	 * The next three fields are touched by __d_lookup.  Place them here
-	 * so they all fit in a cache line.
-	 */
 	struct hlist_node d_hash;	/* lookup hash list */
 	struct dentry *d_parent;	/* parent directory */
 	struct qstr d_name;
+	struct inode *d_inode;		/* Where the name belongs to - NULL is
+					 * negative */
+	unsigned char d_iname[DNAME_INLINE_LEN];	/* small names */
+
+	/* Ref lookup also touches following */
+	unsigned int d_count;		/* protected by d_lock */
+	spinlock_t d_lock;		/* per dentry lock */
+	const struct dentry_operations *d_op;
+	struct super_block *d_sb;	/* The root of the dentry tree */
+	unsigned long d_time;		/* used by d_revalidate */
+	void *d_fsdata;			/* fs-specific data */
 
 	struct list_head d_lru;		/* LRU list */
 	/*
@@ -112,12 +120,6 @@ struct dentry {
 	} d_u;
 	struct list_head d_subdirs;	/* our children */
 	struct list_head d_alias;	/* inode alias list */
-	unsigned long d_time;		/* used by d_revalidate */
-	const struct dentry_operations *d_op;
-	struct super_block *d_sb;	/* The root of the dentry tree */
-	void *d_fsdata;			/* fs-specific data */
-
-	unsigned char d_iname[DNAME_INLINE_LEN_MIN];	/* small names */
 };
 
 /*
@@ -143,7 +145,7 @@ struct dentry_operations {
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
 	char *(*d_dname)(struct dentry *, char *, int);
-};
+} ____cacheline_aligned;
 
 /*
  * Locking rules for dentry_operations callbacks are to be found in
diff --git a/include/linux/fs.h b/include/linux/fs.h
index ea202ff..a04aa96 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -733,6 +733,20 @@ struct posix_acl;
 #define ACL_NOT_CACHED ((void *)(-1))
 
 struct inode {
+	/* RCU path lookup touches following: */
+	umode_t			i_mode;
+	uid_t			i_uid;
+	gid_t			i_gid;
+	const struct inode_operations	*i_op;
+	struct super_block	*i_sb;
+
+	spinlock_t		i_lock;	/* i_blocks, i_bytes, maybe i_size */
+	unsigned int		i_flags;
+	struct mutex		i_mutex;
+
+	unsigned long		i_state;
+	unsigned long		dirtied_when;	/* jiffies of first dirtying */
+
 	struct hlist_node	i_hash;
 	struct list_head	i_wb_list;	/* backing dev IO list */
 	struct list_head	i_lru;		/* inode LRU list */
@@ -744,8 +758,6 @@ struct inode {
 	unsigned long		i_ino;
 	atomic_t		i_count;
 	unsigned int		i_nlink;
-	uid_t			i_uid;
-	gid_t			i_gid;
 	dev_t			i_rdev;
 	unsigned int		i_blkbits;
 	u64			i_version;
@@ -758,13 +770,8 @@ struct inode {
 	struct timespec		i_ctime;
 	blkcnt_t		i_blocks;
 	unsigned short          i_bytes;
-	umode_t			i_mode;
-	spinlock_t		i_lock;	/* i_blocks, i_bytes, maybe i_size */
-	struct mutex		i_mutex;
 	struct rw_semaphore	i_alloc_sem;
-	const struct inode_operations	*i_op;
 	const struct file_operations	*i_fop;	/* former ->i_op->default_file_ops */
-	struct super_block	*i_sb;
 	struct file_lock	*i_flock;
 	struct address_space	*i_mapping;
 	struct address_space	i_data;
@@ -785,11 +792,6 @@ struct inode {
 	struct hlist_head	i_fsnotify_marks;
 #endif
 
-	unsigned long		i_state;
-	unsigned long		dirtied_when;	/* jiffies of first dirtying */
-
-	unsigned int		i_flags;
-
 #ifdef CONFIG_IMA
 	/* protected by i_lock */
 	unsigned int		i_readcount; /* struct files open RO */
@@ -1549,8 +1551,15 @@ struct file_operations {
 };
 
 struct inode_operations {
-	int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
 	struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *);
+	void * (*follow_link) (struct dentry *, struct nameidata *);
+	int (*permission) (struct inode *, int);
+	int (*check_acl)(struct inode *, int);
+
+	int (*readlink) (struct dentry *, char __user *,int);
+	void (*put_link) (struct dentry *, struct nameidata *, void *);
+
+	int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
 	int (*link) (struct dentry *,struct inode *,struct dentry *);
 	int (*unlink) (struct inode *,struct dentry *);
 	int (*symlink) (struct inode *,struct dentry *,const char *);
@@ -1559,12 +1568,7 @@ struct inode_operations {
 	int (*mknod) (struct inode *,struct dentry *,int,dev_t);
 	int (*rename) (struct inode *, struct dentry *,
 			struct inode *, struct dentry *);
-	int (*readlink) (struct dentry *, char __user *,int);
-	void * (*follow_link) (struct dentry *, struct nameidata *);
-	void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int);
-	int (*check_acl)(struct inode *, int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
@@ -1576,7 +1580,7 @@ struct inode_operations {
 			  loff_t len);
 	int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
 		      u64 len);
-};
+} ____cacheline_aligned;
 
 struct seq_file;
 
-- 
cgit v0.10.2


From 34286d6662308d82aed891852d04c7c3a2649b16 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:57 +1100
Subject: fs: rcu-walk aware d_revalidate method

Require filesystems be aware of .d_revalidate being called in rcu-walk
mode (nd->flags & LOOKUP_RCU). For now do a simple push down, returning
-ECHILD from all implementations.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index bdad641..e90ffe6 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -9,7 +9,7 @@ be able to use diff(1).
 
 --------------------------- dentry_operations --------------------------
 prototypes:
-	int (*d_revalidate)(struct dentry *, int);
+	int (*d_revalidate)(struct dentry *, struct nameidata *);
 	int (*d_hash)(const struct dentry *, const struct inode *,
 			struct qstr *);
 	int (*d_compare)(const struct dentry *, const struct inode *,
@@ -21,14 +21,14 @@ prototypes:
 	char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
 
 locking rules:
-		rename_lock	->d_lock	may block
-d_revalidate:	no		no		yes
-d_hash		no		no		no
-d_compare:	yes		no		no
-d_delete:	no		yes		no
-d_release:	no		no		yes
-d_iput:		no		no		yes
-d_dname:	no		no		no
+		rename_lock	->d_lock	may block	rcu-walk
+d_revalidate:	no		no		yes (ref-walk)	maybe
+d_hash		no		no		no		maybe
+d_compare:	yes		no		no		maybe
+d_delete:	no		yes		no		no
+d_release:	no		no		yes		no
+d_iput:		no		no		yes		no
+d_dname:	no		no		no		no
 
 --------------------------- inode_operations --------------------------- 
 prototypes:
diff --git a/Documentation/filesystems/path-lookup.txt b/Documentation/filesystems/path-lookup.txt
index 09b2878..8789d18 100644
--- a/Documentation/filesystems/path-lookup.txt
+++ b/Documentation/filesystems/path-lookup.txt
@@ -317,11 +317,10 @@ The detailed design for rcu-walk is like this:
 The cases where rcu-walk cannot continue are:
 * NULL dentry (ie. any uncached path element)
 * parent with d_inode->i_op->permission or ACLs
-* dentries with d_revalidate
 * Following links
 
-In future patches, permission checks and d_revalidate become rcu-walk aware. It
-may be possible eventually to make following links rcu-walk aware.
+In future patches, permission checks become rcu-walk aware. It may be possible
+eventually to make following links rcu-walk aware.
 
 Uncached path elements will always require dropping to ref-walk mode, at the
 very least because i_mutex needs to be grabbed, and objects allocated.
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index ccf0ce7..cd9756a 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -360,3 +360,23 @@ i_dentry to be reinitialized before it is freed, so an:
   INIT_LIST_HEAD(&inode->i_dentry);
 
 must be done in the RCU callback.
+
+--
+[recommended]
+	vfs now tries to do path walking in "rcu-walk mode", which avoids
+atomic operations and scalability hazards on dentries and inodes (see
+Documentation/filesystems/path-walk.txt). d_hash and d_compare changes (above)
+are examples of the changes required to support this. For more complex
+filesystem callbacks, the vfs drops out of rcu-walk mode before the fs call, so
+no changes are required to the filesystem. However, this is costly and loses
+the benefits of rcu-walk mode. We will begin to add filesystem callbacks that
+are rcu-walk aware, shown below. Filesystems should take advantage of this
+where possible.
+
+--
+[mandatory]
+	d_revalidate is a callback that is made on every path element (if
+the filesystem provides it), which requires dropping out of rcu-walk mode. This
+may now be called in rcu-walk mode (nd->flags & LOOKUP_RCU). -ECHILD should be
+returned if the filesystem cannot handle rcu-walk. See
+Documentation/filesystems/vfs.txt for more details.
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index 69b10ff..c936b49 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -863,6 +863,15 @@ struct dentry_operations {
 	dcache. Most filesystems leave this as NULL, because all their
 	dentries in the dcache are valid
 
+	d_revalidate may be called in rcu-walk mode (nd->flags & LOOKUP_RCU).
+	If in rcu-walk mode, the filesystem must revalidate the dentry without
+	blocking or storing to the dentry, d_parent and d_inode should not be
+	used without care (because they can go NULL), instead nd->inode should
+	be used.
+
+	If a situation is encountered that rcu-walk cannot handle, return
+	-ECHILD and it will be called again in ref-walk mode.
+
   d_hash: called when the VFS adds a dentry to the hash table. The first
 	dentry passed to d_hash is the parent directory that the name is
 	to be hashed into. The inode is the dentry's inode.
diff --git a/drivers/staging/autofs/root.c b/drivers/staging/autofs/root.c
index b09adb5..bf0e975 100644
--- a/drivers/staging/autofs/root.c
+++ b/drivers/staging/autofs/root.c
@@ -154,13 +154,16 @@ static int try_to_fill_dentry(struct dentry *dentry, struct super_block *sb, str
  * yet completely filled in, and revalidate has to delay such
  * lookups..
  */
-static int autofs_revalidate(struct dentry * dentry, struct nameidata *nd)
+static int autofs_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
 	struct inode * dir;
 	struct autofs_sb_info *sbi;
 	struct autofs_dir_ent *ent;
 	int res;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	lock_kernel();
 	dir = dentry->d_parent->d_inode;
 	sbi = autofs_sbi(dir->i_sb);
diff --git a/drivers/staging/smbfs/dir.c b/drivers/staging/smbfs/dir.c
index 78f0941..dd612f5 100644
--- a/drivers/staging/smbfs/dir.c
+++ b/drivers/staging/smbfs/dir.c
@@ -14,6 +14,7 @@
 #include <linux/ctype.h>
 #include <linux/net.h>
 #include <linux/sched.h>
+#include <linux/namei.h>
 
 #include "smb_fs.h"
 #include "smb_mount.h"
@@ -301,13 +302,20 @@ static const struct dentry_operations smbfs_dentry_operations_case =
  * This is the callback when the dcache has a lookup hit.
  */
 static int
-smb_lookup_validate(struct dentry * dentry, struct nameidata *nd)
+smb_lookup_validate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct smb_sb_info *server = server_from_dentry(dentry);
-	struct inode * inode = dentry->d_inode;
-	unsigned long age = jiffies - dentry->d_time;
+	struct smb_sb_info *server;
+	struct inode *inode;
+	unsigned long age;
 	int valid;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	server = server_from_dentry(dentry);
+	inode = dentry->d_inode;
+	age = jiffies - dentry->d_time;
+
 	/*
 	 * The default validation is based on dentry age:
 	 * we believe in dentries for a few seconds.  (But each
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index b8bb7e7..34a3263 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -13,6 +13,7 @@
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/fs.h>
+#include <linux/namei.h>
 #include <linux/pagemap.h>
 #include <linux/ctype.h>
 #include <linux/sched.h>
@@ -607,6 +608,9 @@ static int afs_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 	void *dir_version;
 	int ret;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	vnode = AFS_FS_I(dentry->d_inode);
 
 	if (dentry->d_inode)
diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
index bfe3f2e..651e4ef 100644
--- a/fs/autofs4/root.c
+++ b/fs/autofs4/root.c
@@ -315,12 +315,19 @@ out_error:
  */
 static int autofs4_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct inode *dir = dentry->d_parent->d_inode;
-	struct autofs_sb_info *sbi = autofs4_sbi(dir->i_sb);
-	int oz_mode = autofs4_oz_mode(sbi);
+	struct inode *dir;
+	struct autofs_sb_info *sbi;
+	int oz_mode;
 	int flags = nd ? nd->flags : 0;
 	int status = 1;
 
+	if (flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	dir = dentry->d_parent->d_inode;
+	sbi = autofs4_sbi(dir->i_sb);
+	oz_mode = autofs4_oz_mode(sbi);
+
 	/* Pending dentry */
 	spin_lock(&sbi->fs_lock);
 	if (autofs4_ispending(dentry)) {
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index cc01cf8..fa7ca04 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -990,7 +990,12 @@ static int dir_lease_is_valid(struct inode *dir, struct dentry *dentry)
  */
 static int ceph_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct inode *dir = dentry->d_parent->d_inode;
+	struct inode *dir;
+
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	dir = dentry->d_parent->d_inode;
 
 	dout("d_revalidate %p '%.*s' inode %p offset %lld\n", dentry,
 	     dentry->d_name.len, dentry->d_name.name, dentry->d_inode,
diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index e3b10ca..db2a58c 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -656,6 +656,9 @@ lookup_out:
 static int
 cifs_d_revalidate(struct dentry *direntry, struct nameidata *nd)
 {
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	if (direntry->d_inode) {
 		if (cifs_revalidate_dentry(direntry))
 			return 0;
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index aa40c81..619a830 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -18,6 +18,7 @@
 #include <linux/errno.h>
 #include <linux/string.h>
 #include <linux/spinlock.h>
+#include <linux/namei.h>
 
 #include <asm/uaccess.h>
 
@@ -541,9 +542,13 @@ out:
 /* called when a cache lookup succeeds */
 static int coda_dentry_revalidate(struct dentry *de, struct nameidata *nd)
 {
-	struct inode *inode = de->d_inode;
+	struct inode *inode;
 	struct coda_inode_info *cii;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = de->d_inode;
 	if (!inode || coda_isroot(inode))
 		goto out;
 	if (is_bad_inode(inode))
diff --git a/fs/ecryptfs/dentry.c b/fs/ecryptfs/dentry.c
index 906e803..6fc4f31 100644
--- a/fs/ecryptfs/dentry.c
+++ b/fs/ecryptfs/dentry.c
@@ -44,12 +44,17 @@
  */
 static int ecryptfs_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct dentry *lower_dentry = ecryptfs_dentry_to_lower(dentry);
-	struct vfsmount *lower_mnt = ecryptfs_dentry_to_lower_mnt(dentry);
+	struct dentry *lower_dentry;
+	struct vfsmount *lower_mnt;
 	struct dentry *dentry_save;
 	struct vfsmount *vfsmount_save;
 	int rc = 1;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	lower_dentry = ecryptfs_dentry_to_lower(dentry);
+	lower_mnt = ecryptfs_dentry_to_lower_mnt(dentry);
 	if (!lower_dentry->d_op || !lower_dentry->d_op->d_revalidate)
 		goto out;
 	dentry_save = nd->path.dentry;
diff --git a/fs/fat/namei_vfat.c b/fs/fat/namei_vfat.c
index 3be5ed7..e3ffc5e 100644
--- a/fs/fat/namei_vfat.c
+++ b/fs/fat/namei_vfat.c
@@ -43,6 +43,9 @@ static int vfat_revalidate_shortname(struct dentry *dentry)
 
 static int vfat_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	/* This is not negative dentry. Always valid. */
 	if (dentry->d_inode)
 		return 1;
@@ -51,6 +54,9 @@ static int vfat_revalidate(struct dentry *dentry, struct nameidata *nd)
 
 static int vfat_revalidate_ci(struct dentry *dentry, struct nameidata *nd)
 {
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	/*
 	 * This is not negative dentry. Always valid.
 	 *
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index c9a8a42..07f4b5e 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -156,8 +156,12 @@ u64 fuse_get_attr_version(struct fuse_conn *fc)
  */
 static int fuse_dentry_revalidate(struct dentry *entry, struct nameidata *nd)
 {
-	struct inode *inode = entry->d_inode;
+	struct inode *inode;
+
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
 
+	inode = entry->d_inode;
 	if (inode && is_bad_inode(inode))
 		return 0;
 	else if (fuse_dentry_time(entry) < get_jiffies_64()) {
diff --git a/fs/gfs2/dentry.c b/fs/gfs2/dentry.c
index 50497f6..4a45633 100644
--- a/fs/gfs2/dentry.c
+++ b/fs/gfs2/dentry.c
@@ -11,6 +11,7 @@
 #include <linux/completion.h>
 #include <linux/buffer_head.h>
 #include <linux/gfs2_ondisk.h>
+#include <linux/namei.h>
 #include <linux/crc32.h>
 
 #include "gfs2.h"
@@ -34,15 +35,23 @@
 
 static int gfs2_drevalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct dentry *parent = dget_parent(dentry);
-	struct gfs2_sbd *sdp = GFS2_SB(parent->d_inode);
-	struct gfs2_inode *dip = GFS2_I(parent->d_inode);
-	struct inode *inode = dentry->d_inode;
+	struct dentry *parent;
+	struct gfs2_sbd *sdp;
+	struct gfs2_inode *dip;
+	struct inode *inode;
 	struct gfs2_holder d_gh;
 	struct gfs2_inode *ip = NULL;
 	int error;
 	int had_lock = 0;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	parent = dget_parent(dentry);
+	sdp = GFS2_SB(parent->d_inode);
+	dip = GFS2_I(parent->d_inode);
+	inode = dentry->d_inode;
+
 	if (inode) {
 		if (is_bad_inode(inode))
 			goto invalid;
diff --git a/fs/hfs/sysdep.c b/fs/hfs/sysdep.c
index 7478f5c..19cf291 100644
--- a/fs/hfs/sysdep.c
+++ b/fs/hfs/sysdep.c
@@ -8,15 +8,20 @@
  * This file contains the code to do various system dependent things.
  */
 
+#include <linux/namei.h>
 #include "hfs_fs.h"
 
 /* dentry case-handling: just lowercase everything */
 
 static int hfs_revalidate_dentry(struct dentry *dentry, struct nameidata *nd)
 {
-	struct inode *inode = dentry->d_inode;
+	struct inode *inode;
 	int diff;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = dentry->d_inode;
 	if(!inode)
 		return 1;
 
diff --git a/fs/jfs/namei.c b/fs/jfs/namei.c
index a151cbd..4414e3a 100644
--- a/fs/jfs/namei.c
+++ b/fs/jfs/namei.c
@@ -1608,6 +1608,8 @@ out:
 
 static int jfs_ci_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
 	/*
 	 * This is not negative dentry. Always valid.
 	 *
diff --git a/fs/namei.c b/fs/namei.c
index 90bd287..6e27536 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -563,10 +563,26 @@ void release_open_intent(struct nameidata *nd)
 		fput(nd->intent.open.file);
 }
 
+static int d_revalidate(struct dentry *dentry, struct nameidata *nd)
+{
+	int status;
+
+	status = dentry->d_op->d_revalidate(dentry, nd);
+	if (status == -ECHILD) {
+		if (nameidata_dentry_drop_rcu(nd, dentry))
+			return status;
+		status = dentry->d_op->d_revalidate(dentry, nd);
+	}
+
+	return status;
+}
+
 static inline struct dentry *
 do_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	int status = dentry->d_op->d_revalidate(dentry, nd);
+	int status;
+
+	status = d_revalidate(dentry, nd);
 	if (unlikely(status <= 0)) {
 		/*
 		 * The dentry failed validation.
@@ -574,14 +590,20 @@ do_revalidate(struct dentry *dentry, struct nameidata *nd)
 		 * the dentry otherwise d_revalidate is asking us
 		 * to return a fail status.
 		 */
-		if (!status) {
+		if (status < 0) {
+			/* If we're in rcu-walk, we don't have a ref */
+			if (!(nd->flags & LOOKUP_RCU))
+				dput(dentry);
+			dentry = ERR_PTR(status);
+
+		} else {
+			/* Don't d_invalidate in rcu-walk mode */
+			if (nameidata_dentry_drop_rcu_maybe(nd, dentry))
+				return ERR_PTR(-ECHILD);
 			if (!d_invalidate(dentry)) {
 				dput(dentry);
 				dentry = NULL;
 			}
-		} else {
-			dput(dentry);
-			dentry = ERR_PTR(status);
 		}
 	}
 	return dentry;
@@ -626,7 +648,7 @@ force_reval_path(struct path *path, struct nameidata *nd)
 	if (!need_reval_dot(dentry))
 		return 0;
 
-	status = dentry->d_op->d_revalidate(dentry, nd);
+	status = d_revalidate(dentry, nd);
 	if (status > 0)
 		return 0;
 
@@ -1039,12 +1061,8 @@ static int do_lookup(struct nameidata *nd, struct qstr *name,
 			return -ECHILD;
 
 		nd->seq = seq;
-		if (dentry->d_flags & DCACHE_OP_REVALIDATE) {
-			/* We commonly drop rcu-walk here */
-			if (nameidata_dentry_drop_rcu(nd, dentry))
-				return -ECHILD;
+		if (dentry->d_flags & DCACHE_OP_REVALIDATE)
 			goto need_revalidate;
-		}
 		path->mnt = mnt;
 		path->dentry = dentry;
 		__follow_mount_rcu(nd, path, inode);
@@ -1292,12 +1310,11 @@ return_reval:
 		 * We may need to check the cached dentry for staleness.
 		 */
 		if (need_reval_dot(nd->path.dentry)) {
-			if (nameidata_drop_rcu_maybe(nd))
-				return -ECHILD;
-			err = -ESTALE;
 			/* Note: we do not d_invalidate() */
-			if (!nd->path.dentry->d_op->d_revalidate(
-					nd->path.dentry, nd))
+			err = d_revalidate(nd->path.dentry, nd);
+			if (!err)
+				err = -ESTALE;
+			if (err < 0)
 				break;
 		}
 return_base:
@@ -2080,10 +2097,11 @@ static struct file *do_last(struct nameidata *nd, struct path *path,
 		dir = nd->path.dentry;
 	case LAST_DOT:
 		if (need_reval_dot(dir)) {
-			if (!dir->d_op->d_revalidate(dir, nd)) {
+			error = d_revalidate(nd->path.dentry, nd);
+			if (!error)
 				error = -ESTALE;
+			if (error < 0)
 				goto exit;
-			}
 		}
 		/* fallthrough */
 	case LAST_ROOT:
diff --git a/fs/ncpfs/dir.c b/fs/ncpfs/dir.c
index 4b9cbb2..28f136d 100644
--- a/fs/ncpfs/dir.c
+++ b/fs/ncpfs/dir.c
@@ -17,6 +17,7 @@
 #include <linux/kernel.h>
 #include <linux/vmalloc.h>
 #include <linux/mm.h>
+#include <linux/namei.h>
 #include <asm/uaccess.h>
 #include <asm/byteorder.h>
 
@@ -308,6 +309,9 @@ ncp_lookup_validate(struct dentry *dentry, struct nameidata *nd)
 	int res, val = 0, len;
 	__u8 __name[NCP_MAXPATHLEN + 1];
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	parent = dget_parent(dentry);
 	dir = parent->d_inode;
 
diff --git a/fs/ncpfs/inode.c b/fs/ncpfs/inode.c
index 0c75a5f..9531c05 100644
--- a/fs/ncpfs/inode.c
+++ b/fs/ncpfs/inode.c
@@ -29,6 +29,7 @@
 #include <linux/vfs.h>
 #include <linux/mount.h>
 #include <linux/seq_file.h>
+#include <linux/namei.h>
 
 #include <linux/ncp_fs.h>
 
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 37e0a8b..58beace 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -938,7 +938,8 @@ static int nfs_check_verifier(struct inode *dir, struct dentry *dentry)
  * component of the path.
  * We check for this using LOOKUP_CONTINUE and LOOKUP_PARENT.
  */
-static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd, unsigned int mask)
+static inline unsigned int nfs_lookup_check_intent(struct nameidata *nd,
+						unsigned int mask)
 {
 	if (nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))
 		return 0;
@@ -1018,7 +1019,7 @@ int nfs_neg_need_reval(struct inode *dir, struct dentry *dentry,
  * If the parent directory is seen to have changed, we throw out the
  * cached dentry and do a new lookup.
  */
-static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd)
+static int nfs_lookup_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
 	struct inode *dir;
 	struct inode *inode;
@@ -1027,6 +1028,9 @@ static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd)
 	struct nfs_fattr *fattr = NULL;
 	int error;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
 	parent = dget_parent(dentry);
 	dir = parent->d_inode;
 	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index 4d54c60..0310b16 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -52,9 +52,15 @@ void ocfs2_dentry_attach_gen(struct dentry *dentry)
 static int ocfs2_dentry_revalidate(struct dentry *dentry,
 				   struct nameidata *nd)
 {
-	struct inode *inode = dentry->d_inode;
+	struct inode *inode;
 	int ret = 0;    /* if all else fails, just return false */
-	struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
+	struct ocfs2_super *osb;
+
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = dentry->d_inode;
+	osb = OCFS2_SB(dentry->d_sb);
 
 	mlog_entry("(0x%p, '%.*s')\n", dentry,
 		   dentry->d_name.len, dentry->d_name.name);
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 85f0a80..dc5b2fc 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1719,10 +1719,16 @@ static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat
  */
 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct inode *inode = dentry->d_inode;
-	struct task_struct *task = get_proc_task(inode);
+	struct inode *inode;
+	struct task_struct *task;
 	const struct cred *cred;
 
+	if (nd && nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = dentry->d_inode;
+	task = get_proc_task(inode);
+
 	if (task) {
 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
 		    task_dumpable(task)) {
@@ -1888,12 +1894,19 @@ static int proc_fd_link(struct inode *inode, struct path *path)
 
 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct inode *inode = dentry->d_inode;
-	struct task_struct *task = get_proc_task(inode);
-	int fd = proc_fd(inode);
+	struct inode *inode;
+	struct task_struct *task;
+	int fd;
 	struct files_struct *files;
 	const struct cred *cred;
 
+	if (nd && nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = dentry->d_inode;
+	task = get_proc_task(inode);
+	fd = proc_fd(inode);
+
 	if (task) {
 		files = get_files_struct(task);
 		if (files) {
@@ -2563,8 +2576,14 @@ static const struct pid_entry proc_base_stuff[] = {
  */
 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct inode *inode = dentry->d_inode;
-	struct task_struct *task = get_proc_task(inode);
+	struct inode *inode;
+	struct task_struct *task;
+
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	inode = dentry->d_inode;
+	task = get_proc_task(inode);
 	if (task) {
 		put_task_struct(task);
 		return 1;
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 35efd85..c9097f4 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -5,6 +5,7 @@
 #include <linux/sysctl.h>
 #include <linux/proc_fs.h>
 #include <linux/security.h>
+#include <linux/namei.h>
 #include "internal.h"
 
 static const struct dentry_operations proc_sys_dentry_operations;
@@ -389,6 +390,8 @@ static const struct inode_operations proc_sys_dir_operations = {
 
 static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
 	return !PROC_I(dentry->d_inode)->sysctl->unregistering;
 }
 
diff --git a/fs/reiserfs/xattr.c b/fs/reiserfs/xattr.c
index e0f0d7e..9ea22a5 100644
--- a/fs/reiserfs/xattr.c
+++ b/fs/reiserfs/xattr.c
@@ -972,6 +972,8 @@ int reiserfs_permission(struct inode *inode, int mask)
 
 static int xattr_hide_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
 	return -EPERM;
 }
 
diff --git a/fs/sysfs/dir.c b/fs/sysfs/dir.c
index 3e076ca..ea9120a 100644
--- a/fs/sysfs/dir.c
+++ b/fs/sysfs/dir.c
@@ -239,9 +239,13 @@ static int sysfs_dentry_delete(const struct dentry *dentry)
 
 static int sysfs_dentry_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
-	struct sysfs_dirent *sd = dentry->d_fsdata;
+	struct sysfs_dirent *sd;
 	int is_dir;
 
+	if (nd->flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	sd = dentry->d_fsdata;
 	mutex_lock(&sysfs_mutex);
 
 	/* The sysfs dirent has been deleted */
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index b1aeda0..8b2064d 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -190,7 +190,6 @@ struct dentry_operations {
 #define DCACHE_OP_REVALIDATE	0x4000
 #define DCACHE_OP_DELETE	0x8000
 
-
 extern spinlock_t dcache_inode_lock;
 extern seqlock_t rename_lock;
 
-- 
cgit v0.10.2


From b74c79e99389cd79b31fcc08f82c24e492e63c7e Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:58 +1100
Subject: fs: provide rcu-walk aware permission i_ops

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index e90ffe6..977d891 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -47,8 +47,8 @@ ata *);
 	void * (*follow_link) (struct dentry *, struct nameidata *);
 	void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int, struct nameidata *);
-	int (*check_acl)(struct inode *, int);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
@@ -76,7 +76,7 @@ follow_link:	no
 put_link:	no
 truncate:	yes		(see below)
 setattr:	yes
-permission:	no
+permission:	no (may not block if called in rcu-walk mode)
 check_acl:	no
 getattr:	no
 setxattr:	yes
diff --git a/Documentation/filesystems/path-lookup.txt b/Documentation/filesystems/path-lookup.txt
index 8789d18..eb59c8b 100644
--- a/Documentation/filesystems/path-lookup.txt
+++ b/Documentation/filesystems/path-lookup.txt
@@ -316,11 +316,9 @@ The detailed design for rcu-walk is like this:
 
 The cases where rcu-walk cannot continue are:
 * NULL dentry (ie. any uncached path element)
-* parent with d_inode->i_op->permission or ACLs
 * Following links
 
-In future patches, permission checks become rcu-walk aware. It may be possible
-eventually to make following links rcu-walk aware.
+It may be possible eventually to make following links rcu-walk aware.
 
 Uncached path elements will always require dropping to ref-walk mode, at the
 very least because i_mutex needs to be grabbed, and objects allocated.
@@ -336,9 +334,49 @@ or stored into. The result is massive improvements in performance and
 scalability of path resolution.
 
 
+Interesting statistics
+======================
+
+The following table gives rcu lookup statistics for a few simple workloads
+(2s12c24t Westmere, debian non-graphical system). Ungraceful are attempts to
+drop rcu that fail due to d_seq failure and requiring the entire path lookup
+again. Other cases are successful rcu-drops that are required before the final
+element, nodentry for missing dentry, revalidate for filesystem revalidate
+routine requiring rcu drop, permission for permission check requiring drop,
+and link for symlink traversal requiring drop.
+
+     rcu-lookups     restart  nodentry          link  revalidate  permission
+bootup     47121           0      4624          1010       10283        7852
+dbench  25386793           0   6778659(26.7%)     55         549        1156
+kbuild   2696672          10     64442(2.3%)  108764(4.0%)     1        1590
+git diff   39605           0        28             2           0         106
+vfstest 24185492        4945    708725(2.9%) 1076136(4.4%)     0        2651
+
+What this shows is that failed rcu-walk lookups, ie. ones that are restarted
+entirely with ref-walk, are quite rare. Even the "vfstest" case which
+specifically has concurrent renames/mkdir/rmdir/ creat/unlink/etc to excercise
+such races is not showing a huge amount of restarts.
+
+Dropping from rcu-walk to ref-walk mean that we have encountered a dentry where
+the reference count needs to be taken for some reason. This is either because
+we have reached the target of the path walk, or because we have encountered a
+condition that can't be resolved in rcu-walk mode.  Ideally, we drop rcu-walk
+only when we have reached the target dentry, so the other statistics show where
+this does not happen.
+
+Note that a graceful drop from rcu-walk mode due to something such as the
+dentry not existing (which can be common) is not necessarily a failure of
+rcu-walk scheme, because some elements of the path may have been walked in
+rcu-walk mode. The further we get from common path elements (such as cwd or
+root), the less contended the dentry is likely to be. The closer we are to
+common path elements, the more likely they will exist in dentry cache.
+
+
 Papers and other documentation on dcache locking
 ================================================
 
 1. Scaling dcache with RCU (http://linuxjournal.com/article.php?sid=7124).
 
 2. http://lse.sourceforge.net/locking/dcache/dcache.html
+
+
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index cd9756a..07a32b4 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -380,3 +380,8 @@ the filesystem provides it), which requires dropping out of rcu-walk mode. This
 may now be called in rcu-walk mode (nd->flags & LOOKUP_RCU). -ECHILD should be
 returned if the filesystem cannot handle rcu-walk. See
 Documentation/filesystems/vfs.txt for more details.
+
+	permission and check_acl are inode permission checks that are called
+on many or all directory inodes on the way down a path walk (to check for
+exec permission). These must now be rcu-walk aware (flags & IPERM_RCU). See
+Documentation/filesystems/vfs.txt for more details.
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index c936b49..fbb324e 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -325,7 +325,8 @@ struct inode_operations {
         void * (*follow_link) (struct dentry *, struct nameidata *);
         void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int, struct nameidata *);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
@@ -414,6 +415,13 @@ otherwise noted.
   permission: called by the VFS to check for access rights on a POSIX-like
   	filesystem.
 
+	May be called in rcu-walk mode (flags & IPERM_RCU). If in rcu-walk
+	mode, the filesystem must check the permission without blocking or
+	storing to the inode.
+
+	If a situation is encountered that rcu-walk cannot handle, return
+	-ECHILD and it will be called again in ref-walk mode.
+
   setattr: called by the VFS to set attributes for a file. This method
   	is called by chmod(2) and related system calls.
 
diff --git a/drivers/staging/smbfs/file.c b/drivers/staging/smbfs/file.c
index 5dcd19c..31372e7 100644
--- a/drivers/staging/smbfs/file.c
+++ b/drivers/staging/smbfs/file.c
@@ -407,11 +407,14 @@ smb_file_release(struct inode *inode, struct file * file)
  * privileges, so we need our own check for this.
  */
 static int
-smb_file_permission(struct inode *inode, int mask)
+smb_file_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int mode = inode->i_mode;
 	int error = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	VERBOSE("mode=%x, mask=%x\n", mode, mask);
 
 	/* Look at user permissions */
diff --git a/fs/9p/acl.c b/fs/9p/acl.c
index 12d6023..6e58c4c 100644
--- a/fs/9p/acl.c
+++ b/fs/9p/acl.c
@@ -91,11 +91,14 @@ static struct posix_acl *v9fs_get_cached_acl(struct inode *inode, int type)
 	return acl;
 }
 
-int v9fs_check_acl(struct inode *inode, int mask)
+int v9fs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	struct v9fs_session_info *v9ses;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	v9ses = v9fs_inode2v9ses(inode);
 	if ((v9ses->flags & V9FS_ACCESS_MASK) != V9FS_ACCESS_CLIENT) {
 		/*
diff --git a/fs/9p/acl.h b/fs/9p/acl.h
index 59e18c2..7ef3ac9 100644
--- a/fs/9p/acl.h
+++ b/fs/9p/acl.h
@@ -16,7 +16,7 @@
 
 #ifdef CONFIG_9P_FS_POSIX_ACL
 extern int v9fs_get_acl(struct inode *, struct p9_fid *);
-extern int v9fs_check_acl(struct inode *inode, int mask);
+extern int v9fs_check_acl(struct inode *inode, int mask, unsigned int flags);
 extern int v9fs_acl_chmod(struct dentry *);
 extern int v9fs_set_create_acl(struct dentry *,
 			       struct posix_acl *, struct posix_acl *);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index cca8eef..6d4bc1c 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -624,7 +624,7 @@ extern void afs_clear_permits(struct afs_vnode *);
 extern void afs_cache_permit(struct afs_vnode *, struct key *, long);
 extern void afs_zap_permits(struct rcu_head *);
 extern struct key *afs_request_key(struct afs_cell *);
-extern int afs_permission(struct inode *, int);
+extern int afs_permission(struct inode *, int, unsigned int);
 
 /*
  * server.c
diff --git a/fs/afs/security.c b/fs/afs/security.c
index bb4ed14..f44b9d3 100644
--- a/fs/afs/security.c
+++ b/fs/afs/security.c
@@ -285,13 +285,16 @@ static int afs_check_permit(struct afs_vnode *vnode, struct key *key,
  * - AFS ACLs are attached to directories only, and a file is controlled by its
  *   parent directory's ACL
  */
-int afs_permission(struct inode *inode, int mask)
+int afs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct afs_vnode *vnode = AFS_FS_I(inode);
 	afs_access_t uninitialized_var(access);
 	struct key *key;
 	int ret;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	_enter("{{%x:%u},%lx},%x,",
 	       vnode->fid.vid, vnode->fid.vnode, vnode->flags, mask);
 
@@ -347,7 +350,7 @@ int afs_permission(struct inode *inode, int mask)
 	}
 
 	key_put(key);
-	ret = generic_permission(inode, mask, NULL);
+	ret = generic_permission(inode, mask, flags, NULL);
 	_leave(" = %d", ret);
 	return ret;
 
diff --git a/fs/bad_inode.c b/fs/bad_inode.c
index f024d8a..9ad2369 100644
--- a/fs/bad_inode.c
+++ b/fs/bad_inode.c
@@ -229,8 +229,11 @@ static int bad_inode_readlink(struct dentry *dentry, char __user *buffer,
 	return -EIO;
 }
 
-static int bad_inode_permission(struct inode *inode, int mask)
+static int bad_inode_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	return -EIO;
 }
 
diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index 2222d16..cb518a4 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -185,13 +185,15 @@ static int btrfs_xattr_acl_set(struct dentry *dentry, const char *name,
 	return ret;
 }
 
-int btrfs_check_acl(struct inode *inode, int mask)
+int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error = -EAGAIN;
 
-	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index af52f6d..a142d20 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2544,7 +2544,7 @@ int btrfs_sync_fs(struct super_block *sb, int wait);
 
 /* acl.c */
 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
-int btrfs_check_acl(struct inode *inode, int mask);
+int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags);
 #else
 #define btrfs_check_acl NULL
 #endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 63e4546..5cf0db0 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -7211,11 +7211,14 @@ static int btrfs_set_page_dirty(struct page *page)
 	return __set_page_dirty_nobuffers(page);
 }
 
-static int btrfs_permission(struct inode *inode, int mask)
+static int btrfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
 		return -EACCES;
-	return generic_permission(inode, mask, btrfs_check_acl);
+	return generic_permission(inode, mask, flags, btrfs_check_acl);
 }
 
 static const struct inode_operations btrfs_dir_inode_operations = {
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 47f8c8b..e61de4f 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1781,12 +1781,17 @@ int ceph_do_getattr(struct inode *inode, int mask)
  * Check inode permissions.  We verify we have a valid value for
  * the AUTH cap, then call the generic handler.
  */
-int ceph_permission(struct inode *inode, int mask)
+int ceph_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	int err = ceph_do_getattr(inode, CEPH_CAP_AUTH_SHARED);
+	int err;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	err = ceph_do_getattr(inode, CEPH_CAP_AUTH_SHARED);
 
 	if (!err)
-		err = generic_permission(inode, mask, NULL);
+		err = generic_permission(inode, mask, flags, NULL);
 	return err;
 }
 
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 7f01728..4553d88 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -665,7 +665,7 @@ extern void ceph_queue_invalidate(struct inode *inode);
 extern void ceph_queue_writeback(struct inode *inode);
 
 extern int ceph_do_getattr(struct inode *inode, int mask);
-extern int ceph_permission(struct inode *inode, int mask);
+extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 			struct kstat *stat);
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index 223717d..8e21e0f 100644
--- a/fs/cifs/cifsfs.c
+++ b/fs/cifs/cifsfs.c
@@ -283,10 +283,13 @@ cifs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	return 0;
 }
 
-static int cifs_permission(struct inode *inode, int mask)
+static int cifs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct cifs_sb_info *cifs_sb;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	cifs_sb = CIFS_SB(inode->i_sb);
 
 	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NO_PERM) {
@@ -298,7 +301,7 @@ static int cifs_permission(struct inode *inode, int mask)
 		on the client (above and beyond ACL on servers) for
 		servers which do not support setting and viewing mode bits,
 		so allowing client to check permissions is useful */
-		return generic_permission(inode, mask, NULL);
+		return generic_permission(inode, mask, flags, NULL);
 }
 
 static struct kmem_cache *cifs_inode_cachep;
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index 619a830..29badd9 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -135,10 +135,13 @@ exit:
 }
 
 
-int coda_permission(struct inode *inode, int mask)
+int coda_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int error;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	mask &= MAY_READ | MAY_WRITE | MAY_EXEC;
  
 	if (!mask)
diff --git a/fs/coda/pioctl.c b/fs/coda/pioctl.c
index 2fd89b5..741f0bd 100644
--- a/fs/coda/pioctl.c
+++ b/fs/coda/pioctl.c
@@ -24,7 +24,7 @@
 #include <linux/coda_psdev.h>
 
 /* pioctl ops */
-static int coda_ioctl_permission(struct inode *inode, int mask);
+static int coda_ioctl_permission(struct inode *inode, int mask, unsigned int flags);
 static long coda_pioctl(struct file *filp, unsigned int cmd,
 			unsigned long user_data);
 
@@ -41,8 +41,10 @@ const struct file_operations coda_ioctl_operations = {
 };
 
 /* the coda pioctl inode ops */
-static int coda_ioctl_permission(struct inode *inode, int mask)
+static int coda_ioctl_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	return (mask & MAY_EXEC) ? -EACCES : 0;
 }
 
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index f91b35d..337352a 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -980,8 +980,10 @@ int ecryptfs_truncate(struct dentry *dentry, loff_t new_length)
 }
 
 static int
-ecryptfs_permission(struct inode *inode, int mask)
+ecryptfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	return inode_permission(ecryptfs_inode_to_lower(inode), mask);
 }
 
diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
index 2bcc043..dd9bb3f 100644
--- a/fs/ext2/acl.c
+++ b/fs/ext2/acl.c
@@ -232,10 +232,14 @@ ext2_set_acl(struct inode *inode, int type, struct posix_acl *acl)
 }
 
 int
-ext2_check_acl(struct inode *inode, int mask)
+ext2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/ext2/acl.h b/fs/ext2/acl.h
index 3ff6cbb..c939b7b 100644
--- a/fs/ext2/acl.h
+++ b/fs/ext2/acl.h
@@ -54,7 +54,7 @@ static inline int ext2_acl_count(size_t size)
 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext2_check_acl (struct inode *, int);
+extern int ext2_check_acl (struct inode *, int, unsigned int);
 extern int ext2_acl_chmod (struct inode *);
 extern int ext2_init_acl (struct inode *, struct inode *);
 
diff --git a/fs/ext3/acl.c b/fs/ext3/acl.c
index 8a11fe2..9e49da8 100644
--- a/fs/ext3/acl.c
+++ b/fs/ext3/acl.c
@@ -240,10 +240,14 @@ ext3_set_acl(handle_t *handle, struct inode *inode, int type,
 }
 
 int
-ext3_check_acl(struct inode *inode, int mask)
+ext3_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/ext3/acl.h b/fs/ext3/acl.h
index 5973346..5faf804 100644
--- a/fs/ext3/acl.h
+++ b/fs/ext3/acl.h
@@ -54,7 +54,7 @@ static inline int ext3_acl_count(size_t size)
 #ifdef CONFIG_EXT3_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext3_check_acl (struct inode *, int);
+extern int ext3_check_acl (struct inode *, int, unsigned int);
 extern int ext3_acl_chmod (struct inode *);
 extern int ext3_init_acl (handle_t *, struct inode *, struct inode *);
 
diff --git a/fs/ext4/acl.c b/fs/ext4/acl.c
index 5e2ed45..373dcae 100644
--- a/fs/ext4/acl.c
+++ b/fs/ext4/acl.c
@@ -238,10 +238,14 @@ ext4_set_acl(handle_t *handle, struct inode *inode, int type,
 }
 
 int
-ext4_check_acl(struct inode *inode, int mask)
+ext4_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/ext4/acl.h b/fs/ext4/acl.h
index 9d843d5..dec8211 100644
--- a/fs/ext4/acl.h
+++ b/fs/ext4/acl.h
@@ -54,7 +54,7 @@ static inline int ext4_acl_count(size_t size)
 #ifdef CONFIG_EXT4_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext4_check_acl(struct inode *, int);
+extern int ext4_check_acl(struct inode *, int, unsigned int);
 extern int ext4_acl_chmod(struct inode *);
 extern int ext4_init_acl(handle_t *, struct inode *, struct inode *);
 
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index 07f4b5e..f738599 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -985,12 +985,15 @@ static int fuse_access(struct inode *inode, int mask)
  * access request is sent.  Execute permission is still checked
  * locally based on file mode.
  */
-static int fuse_permission(struct inode *inode, int mask)
+static int fuse_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct fuse_conn *fc = get_fuse_conn(inode);
 	bool refreshed = false;
 	int err = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	if (!fuse_allow_task(fc, current))
 		return -EACCES;
 
@@ -1005,7 +1008,7 @@ static int fuse_permission(struct inode *inode, int mask)
 	}
 
 	if (fc->flags & FUSE_DEFAULT_PERMISSIONS) {
-		err = generic_permission(inode, mask, NULL);
+		err = generic_permission(inode, mask, flags, NULL);
 
 		/* If permission is denied, try to refresh file
 		   attributes.  This is also needed, because the root
@@ -1013,7 +1016,8 @@ static int fuse_permission(struct inode *inode, int mask)
 		if (err == -EACCES && !refreshed) {
 			err = fuse_do_getattr(inode, NULL, NULL);
 			if (!err)
-				err = generic_permission(inode, mask, NULL);
+				err = generic_permission(inode, mask,
+							flags, NULL);
 		}
 
 		/* Note: the opposite of the above test does not
diff --git a/fs/generic_acl.c b/fs/generic_acl.c
index 6bc9e3a..6280042 100644
--- a/fs/generic_acl.c
+++ b/fs/generic_acl.c
@@ -190,10 +190,14 @@ generic_acl_chmod(struct inode *inode)
 }
 
 int
-generic_check_acl(struct inode *inode, int mask)
+generic_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = get_cached_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = get_cached_acl(inode, ACL_TYPE_ACCESS);
 	if (acl) {
 		int error = posix_acl_permission(inode, acl, mask);
 		posix_acl_release(acl);
diff --git a/fs/gfs2/acl.c b/fs/gfs2/acl.c
index 48171f4..7118f1a 100644
--- a/fs/gfs2/acl.c
+++ b/fs/gfs2/acl.c
@@ -75,11 +75,14 @@ static struct posix_acl *gfs2_acl_get(struct gfs2_inode *ip, int type)
  * Returns: errno
  */
 
-int gfs2_check_acl(struct inode *inode, int mask)
+int gfs2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	acl = gfs2_acl_get(GFS2_I(inode), ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
diff --git a/fs/gfs2/acl.h b/fs/gfs2/acl.h
index b522b0c..a93907c 100644
--- a/fs/gfs2/acl.h
+++ b/fs/gfs2/acl.h
@@ -16,7 +16,7 @@
 #define GFS2_POSIX_ACL_DEFAULT		"posix_acl_default"
 #define GFS2_ACL_MAX_ENTRIES		25
 
-extern int gfs2_check_acl(struct inode *inode, int mask);
+extern int gfs2_check_acl(struct inode *inode, int mask, unsigned int);
 extern int gfs2_acl_create(struct gfs2_inode *dip, struct inode *inode);
 extern int gfs2_acl_chmod(struct gfs2_inode *ip, struct iattr *attr);
 extern const struct xattr_handler gfs2_xattr_system_handler;
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index aa99647..fca6689 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -241,7 +241,7 @@ static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask)
 	    !capable(CAP_LINUX_IMMUTABLE))
 		goto out;
 	if (!IS_IMMUTABLE(inode)) {
-		error = gfs2_permission(inode, MAY_WRITE);
+		error = gfs2_permission(inode, MAY_WRITE, 0);
 		if (error)
 			goto out;
 	}
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index e1213f7..6163be9 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -509,7 +509,7 @@ struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
 	}
 
 	if (!is_root) {
-		error = gfs2_permission(dir, MAY_EXEC);
+		error = gfs2_permission(dir, MAY_EXEC, 0);
 		if (error)
 			goto out;
 	}
@@ -539,7 +539,7 @@ static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
 {
 	int error;
 
-	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
+	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, 0);
 	if (error)
 		return error;
 
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index d8499fa..732a183 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -113,7 +113,7 @@ extern struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
 extern struct inode *gfs2_createi(struct gfs2_holder *ghs,
 				  const struct qstr *name,
 				  unsigned int mode, dev_t dev);
-extern int gfs2_permission(struct inode *inode, int mask);
+extern int gfs2_permission(struct inode *inode, int mask, unsigned int flags);
 extern int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr);
 extern struct inode *gfs2_lookup_simple(struct inode *dip, const char *name);
 extern void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf);
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index f28f897..5c63a06 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -166,7 +166,7 @@ static int gfs2_link(struct dentry *old_dentry, struct inode *dir,
 	if (error)
 		goto out_child;
 
-	error = gfs2_permission(dir, MAY_WRITE | MAY_EXEC);
+	error = gfs2_permission(dir, MAY_WRITE | MAY_EXEC, 0);
 	if (error)
 		goto out_gunlock;
 
@@ -289,7 +289,7 @@ static int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name,
 	if (IS_APPEND(&dip->i_inode))
 		return -EPERM;
 
-	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC);
+	error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, 0);
 	if (error)
 		return error;
 
@@ -822,7 +822,7 @@ static int gfs2_rename(struct inode *odir, struct dentry *odentry,
 			}
 		}
 	} else {
-		error = gfs2_permission(ndir, MAY_WRITE | MAY_EXEC);
+		error = gfs2_permission(ndir, MAY_WRITE | MAY_EXEC, 0);
 		if (error)
 			goto out_gunlock;
 
@@ -857,7 +857,7 @@ static int gfs2_rename(struct inode *odir, struct dentry *odentry,
 	/* Check out the dir to be renamed */
 
 	if (dir_rename) {
-		error = gfs2_permission(odentry->d_inode, MAY_WRITE);
+		error = gfs2_permission(odentry->d_inode, MAY_WRITE, 0);
 		if (error)
 			goto out_gunlock;
 	}
@@ -1041,13 +1041,17 @@ static void gfs2_put_link(struct dentry *dentry, struct nameidata *nd, void *p)
  * Returns: errno
  */
 
-int gfs2_permission(struct inode *inode, int mask)
+int gfs2_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_inode *ip;
 	struct gfs2_holder i_gh;
 	int error;
 	int unlock = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	ip = GFS2_I(inode);
 	if (gfs2_glock_is_locked_by_me(ip->i_gl) == NULL) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
 		if (error)
@@ -1058,7 +1062,7 @@ int gfs2_permission(struct inode *inode, int mask)
 	if ((mask & MAY_WRITE) && IS_IMMUTABLE(inode))
 		error = -EACCES;
 	else
-		error = generic_permission(inode, mask, gfs2_check_acl);
+		error = generic_permission(inode, mask, flags, gfs2_check_acl);
 	if (unlock)
 		gfs2_glock_dq_uninit(&i_gh);
 
diff --git a/fs/hostfs/hostfs_kern.c b/fs/hostfs/hostfs_kern.c
index 0bc81cf..d3244d9 100644
--- a/fs/hostfs/hostfs_kern.c
+++ b/fs/hostfs/hostfs_kern.c
@@ -749,11 +749,14 @@ int hostfs_rename(struct inode *from_ino, struct dentry *from,
 	return err;
 }
 
-int hostfs_permission(struct inode *ino, int desired)
+int hostfs_permission(struct inode *ino, int desired, unsigned int flags)
 {
 	char *name;
 	int r = 0, w = 0, x = 0, err;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	if (desired & MAY_READ) r = 1;
 	if (desired & MAY_WRITE) w = 1;
 	if (desired & MAY_EXEC) x = 1;
@@ -768,7 +771,7 @@ int hostfs_permission(struct inode *ino, int desired)
 		err = access_file(name, r, w, x);
 	__putname(name);
 	if (!err)
-		err = generic_permission(ino, desired, NULL);
+		err = generic_permission(ino, desired, flags, NULL);
 	return err;
 }
 
diff --git a/fs/hpfs/namei.c b/fs/hpfs/namei.c
index 11c2b40..f4ad9e3 100644
--- a/fs/hpfs/namei.c
+++ b/fs/hpfs/namei.c
@@ -419,7 +419,7 @@ again:
 			unlock_kernel();
 			return -ENOSPC;
 		}
-		if (generic_permission(inode, MAY_WRITE, NULL) ||
+		if (generic_permission(inode, MAY_WRITE, 0, NULL) ||
 		    !S_ISREG(inode->i_mode) ||
 		    get_write_access(inode)) {
 			d_rehash(dentry);
diff --git a/fs/jffs2/acl.c b/fs/jffs2/acl.c
index 54a92fd..95b7967 100644
--- a/fs/jffs2/acl.c
+++ b/fs/jffs2/acl.c
@@ -259,11 +259,14 @@ static int jffs2_set_acl(struct inode *inode, int type, struct posix_acl *acl)
 	return rc;
 }
 
-int jffs2_check_acl(struct inode *inode, int mask)
+int jffs2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int rc;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	acl = jffs2_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
diff --git a/fs/jffs2/acl.h b/fs/jffs2/acl.h
index 5e42de8..3119f59 100644
--- a/fs/jffs2/acl.h
+++ b/fs/jffs2/acl.h
@@ -26,7 +26,7 @@ struct jffs2_acl_header {
 
 #ifdef CONFIG_JFFS2_FS_POSIX_ACL
 
-extern int jffs2_check_acl(struct inode *, int);
+extern int jffs2_check_acl(struct inode *, int, unsigned int);
 extern int jffs2_acl_chmod(struct inode *);
 extern int jffs2_init_acl_pre(struct inode *, struct inode *, int *);
 extern int jffs2_init_acl_post(struct inode *);
diff --git a/fs/jfs/acl.c b/fs/jfs/acl.c
index 1057a49..e5de942 100644
--- a/fs/jfs/acl.c
+++ b/fs/jfs/acl.c
@@ -114,10 +114,14 @@ out:
 	return rc;
 }
 
-int jfs_check_acl(struct inode *inode, int mask)
+int jfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = jfs_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = jfs_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/jfs/jfs_acl.h b/fs/jfs/jfs_acl.h
index 54e0755..f9285c4 100644
--- a/fs/jfs/jfs_acl.h
+++ b/fs/jfs/jfs_acl.h
@@ -20,7 +20,7 @@
 
 #ifdef CONFIG_JFS_POSIX_ACL
 
-int jfs_check_acl(struct inode *, int);
+int jfs_check_acl(struct inode *, int, unsigned int flags);
 int jfs_init_acl(tid_t, struct inode *, struct inode *);
 int jfs_acl_chmod(struct inode *inode);
 
diff --git a/fs/logfs/dir.c b/fs/logfs/dir.c
index 409dfd6..f9ddf0c 100644
--- a/fs/logfs/dir.c
+++ b/fs/logfs/dir.c
@@ -555,9 +555,11 @@ static int logfs_symlink(struct inode *dir, struct dentry *dentry,
 	return __logfs_create(dir, dentry, inode, target, destlen);
 }
 
-static int logfs_permission(struct inode *inode, int mask)
+static int logfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	return generic_permission(inode, mask, NULL);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+	return generic_permission(inode, mask, flags, NULL);
 }
 
 static int logfs_link(struct dentry *old_dentry, struct inode *dir,
diff --git a/fs/namei.c b/fs/namei.c
index 6e27536..4e957bf 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -169,8 +169,8 @@ EXPORT_SYMBOL(putname);
 /*
  * This does basic POSIX ACL permission checking
  */
-static inline int __acl_permission_check(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask), int rcu)
+static int acl_permission_check(struct inode *inode, int mask, unsigned int flags,
+		int (*check_acl)(struct inode *inode, int mask, unsigned int flags))
 {
 	umode_t			mode = inode->i_mode;
 
@@ -180,13 +180,9 @@ static inline int __acl_permission_check(struct inode *inode, int mask,
 		mode >>= 6;
 	else {
 		if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) {
-			if (rcu) {
-				return -ECHILD;
-			} else {
-				int error = check_acl(inode, mask);
-				if (error != -EAGAIN)
-					return error;
-			}
+			int error = check_acl(inode, mask, flags);
+			if (error != -EAGAIN)
+				return error;
 		}
 
 		if (in_group_p(inode->i_gid))
@@ -201,32 +197,31 @@ static inline int __acl_permission_check(struct inode *inode, int mask,
 	return -EACCES;
 }
 
-static inline int acl_permission_check(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask))
-{
-	return __acl_permission_check(inode, mask, check_acl, 0);
-}
-
 /**
- * generic_permission  -  check for access rights on a Posix-like filesystem
+ * generic_permission -  check for access rights on a Posix-like filesystem
  * @inode:	inode to check access rights for
  * @mask:	right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
  * @check_acl:	optional callback to check for Posix ACLs
+ * @flags	IPERM_FLAG_ flags.
  *
  * Used to check for read/write/execute permissions on a file.
  * We use "fsuid" for this, letting us set arbitrary permissions
  * for filesystem access without changing the "normal" uids which
- * are used for other things..
+ * are used for other things.
+ *
+ * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk
+ * request cannot be satisfied (eg. requires blocking or too much complexity).
+ * It would then be called again in ref-walk mode.
  */
-int generic_permission(struct inode *inode, int mask,
-		int (*check_acl)(struct inode *inode, int mask))
+int generic_permission(struct inode *inode, int mask, unsigned int flags,
+	int (*check_acl)(struct inode *inode, int mask, unsigned int flags))
 {
 	int ret;
 
 	/*
 	 * Do the basic POSIX ACL permission checks.
 	 */
-	ret = acl_permission_check(inode, mask, check_acl);
+	ret = acl_permission_check(inode, mask, flags, check_acl);
 	if (ret != -EACCES)
 		return ret;
 
@@ -281,9 +276,10 @@ int inode_permission(struct inode *inode, int mask)
 	}
 
 	if (inode->i_op->permission)
-		retval = inode->i_op->permission(inode, mask);
+		retval = inode->i_op->permission(inode, mask, 0);
 	else
-		retval = generic_permission(inode, mask, inode->i_op->check_acl);
+		retval = generic_permission(inode, mask, 0,
+				inode->i_op->check_acl);
 
 	if (retval)
 		return retval;
@@ -668,22 +664,19 @@ force_reval_path(struct path *path, struct nameidata *nd)
  * short-cut DAC fails, then call ->permission() to do more
  * complete permission check.
  */
-static inline int __exec_permission(struct inode *inode, int rcu)
+static inline int exec_permission(struct inode *inode, unsigned int flags)
 {
 	int ret;
 
 	if (inode->i_op->permission) {
-		if (rcu)
-			return -ECHILD;
-		ret = inode->i_op->permission(inode, MAY_EXEC);
-		if (!ret)
-			goto ok;
-		return ret;
+		ret = inode->i_op->permission(inode, MAY_EXEC, flags);
+	} else {
+		ret = acl_permission_check(inode, MAY_EXEC, flags,
+				inode->i_op->check_acl);
 	}
-	ret = __acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl, rcu);
-	if (!ret)
+	if (likely(!ret))
 		goto ok;
-	if (rcu && ret == -ECHILD)
+	if (ret == -ECHILD)
 		return ret;
 
 	if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH))
@@ -691,17 +684,7 @@ static inline int __exec_permission(struct inode *inode, int rcu)
 
 	return ret;
 ok:
-	return security_inode_exec_permission(inode, rcu);
-}
-
-static int exec_permission(struct inode *inode)
-{
-	return __exec_permission(inode, 0);
-}
-
-static int exec_permission_rcu(struct inode *inode)
-{
-	return __exec_permission(inode, 1);
+	return security_inode_exec_permission(inode, flags);
 }
 
 static __always_inline void set_root(struct nameidata *nd)
@@ -1165,7 +1148,7 @@ static int link_path_walk(const char *name, struct nameidata *nd)
 
 		nd->flags |= LOOKUP_CONTINUE;
 		if (nd->flags & LOOKUP_RCU) {
-			err = exec_permission_rcu(nd->inode);
+			err = exec_permission(nd->inode, IPERM_FLAG_RCU);
 			if (err == -ECHILD) {
 				if (nameidata_drop_rcu(nd))
 					return -ECHILD;
@@ -1173,7 +1156,7 @@ static int link_path_walk(const char *name, struct nameidata *nd)
 			}
 		} else {
 exec_again:
-			err = exec_permission(nd->inode);
+			err = exec_permission(nd->inode, 0);
 		}
  		if (err)
 			break;
@@ -1620,7 +1603,7 @@ static struct dentry *__lookup_hash(struct qstr *name,
 	struct dentry *dentry;
 	int err;
 
-	err = exec_permission(inode);
+	err = exec_permission(inode, 0);
 	if (err)
 		return ERR_PTR(err);
 
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 58beace..d33da53 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -2189,11 +2189,14 @@ int nfs_may_open(struct inode *inode, struct rpc_cred *cred, int openflags)
 	return nfs_do_access(inode, cred, nfs_open_permission_mask(openflags));
 }
 
-int nfs_permission(struct inode *inode, int mask)
+int nfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct rpc_cred *cred;
 	int res = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	nfs_inc_stats(inode, NFSIOS_VFSACCESS);
 
 	if ((mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
@@ -2241,7 +2244,7 @@ out:
 out_notsup:
 	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
 	if (res == 0)
-		res = generic_permission(inode, mask, NULL);
+		res = generic_permission(inode, mask, flags, NULL);
 	goto out;
 }
 
diff --git a/fs/nilfs2/inode.c b/fs/nilfs2/inode.c
index 71d4bc8..77b48c8 100644
--- a/fs/nilfs2/inode.c
+++ b/fs/nilfs2/inode.c
@@ -785,15 +785,19 @@ out_err:
 	return err;
 }
 
-int nilfs_permission(struct inode *inode, int mask)
+int nilfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	struct nilfs_root *root = NILFS_I(inode)->i_root;
+	struct nilfs_root *root;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	root = NILFS_I(inode)->i_root;
 	if ((mask & MAY_WRITE) && root &&
 	    root->cno != NILFS_CPTREE_CURRENT_CNO)
 		return -EROFS; /* snapshot is not writable */
 
-	return generic_permission(inode, mask, NULL);
+	return generic_permission(inode, mask, flags, NULL);
 }
 
 int nilfs_load_inode_block(struct nilfs_sb_info *sbi, struct inode *inode,
diff --git a/fs/nilfs2/nilfs.h b/fs/nilfs2/nilfs.h
index f7560da..0ca9882 100644
--- a/fs/nilfs2/nilfs.h
+++ b/fs/nilfs2/nilfs.h
@@ -256,7 +256,7 @@ extern void nilfs_update_inode(struct inode *, struct buffer_head *);
 extern void nilfs_truncate(struct inode *);
 extern void nilfs_evict_inode(struct inode *);
 extern int nilfs_setattr(struct dentry *, struct iattr *);
-int nilfs_permission(struct inode *inode, int mask);
+int nilfs_permission(struct inode *inode, int mask, unsigned int flags);
 extern int nilfs_load_inode_block(struct nilfs_sb_info *, struct inode *,
 				  struct buffer_head **);
 extern int nilfs_inode_dirty(struct inode *);
diff --git a/fs/ocfs2/acl.c b/fs/ocfs2/acl.c
index 3919150..704f6b1 100644
--- a/fs/ocfs2/acl.c
+++ b/fs/ocfs2/acl.c
@@ -291,13 +291,17 @@ static int ocfs2_set_acl(handle_t *handle,
 	return ret;
 }
 
-int ocfs2_check_acl(struct inode *inode, int mask)
+int ocfs2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_super *osb;
 	struct buffer_head *di_bh = NULL;
 	struct posix_acl *acl;
 	int ret = -EAGAIN;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	osb = OCFS2_SB(inode->i_sb);
 	if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL))
 		return ret;
 
diff --git a/fs/ocfs2/acl.h b/fs/ocfs2/acl.h
index 5c5d31f..4fe7c9c 100644
--- a/fs/ocfs2/acl.h
+++ b/fs/ocfs2/acl.h
@@ -26,7 +26,7 @@ struct ocfs2_acl_entry {
 	__le32 e_id;
 };
 
-extern int ocfs2_check_acl(struct inode *, int);
+extern int ocfs2_check_acl(struct inode *, int, unsigned int);
 extern int ocfs2_acl_chmod(struct inode *);
 extern int ocfs2_init_acl(handle_t *, struct inode *, struct inode *,
 			  struct buffer_head *, struct buffer_head *,
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index f6cba56..bdadbae0 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -1307,10 +1307,13 @@ bail:
 	return err;
 }
 
-int ocfs2_permission(struct inode *inode, int mask)
+int ocfs2_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int ret;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	mlog_entry_void();
 
 	ret = ocfs2_inode_lock(inode, NULL, 0);
@@ -1320,7 +1323,7 @@ int ocfs2_permission(struct inode *inode, int mask)
 		goto out;
 	}
 
-	ret = generic_permission(inode, mask, ocfs2_check_acl);
+	ret = generic_permission(inode, mask, flags, ocfs2_check_acl);
 
 	ocfs2_inode_unlock(inode, 0);
 out:
diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h
index 97bf761..f5afbbe 100644
--- a/fs/ocfs2/file.h
+++ b/fs/ocfs2/file.h
@@ -61,7 +61,7 @@ int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
 int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
 		  struct kstat *stat);
-int ocfs2_permission(struct inode *inode, int mask);
+int ocfs2_permission(struct inode *inode, int mask, unsigned int flags);
 
 int ocfs2_should_update_atime(struct inode *inode,
 			      struct vfsmount *vfsmnt);
diff --git a/fs/proc/base.c b/fs/proc/base.c
index dc5b2fc..b953d41 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2114,11 +2114,13 @@ static const struct file_operations proc_fd_operations = {
  * /proc/pid/fd needs a special permission handler so that a process can still
  * access /proc/self/fd after it has executed a setuid().
  */
-static int proc_fd_permission(struct inode *inode, int mask)
+static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int rv;
 
-	rv = generic_permission(inode, mask, NULL);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+	rv = generic_permission(inode, mask, flags, NULL);
 	if (rv == 0)
 		return 0;
 	if (task_pid(current) == proc_pid(inode))
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index c9097f4..09a1f92 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -295,7 +295,7 @@ out:
 	return ret;
 }
 
-static int proc_sys_permission(struct inode *inode, int mask)
+static int proc_sys_permission(struct inode *inode, int mask,unsigned int flags)
 {
 	/*
 	 * sysctl entries that are not writeable,
@@ -305,6 +305,9 @@ static int proc_sys_permission(struct inode *inode, int mask)
 	struct ctl_table *table;
 	int error;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	/* Executable files are not allowed under /proc/sys/ */
 	if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))
 		return -EACCES;
diff --git a/fs/reiserfs/xattr.c b/fs/reiserfs/xattr.c
index 9ea22a5..3cfb2e9 100644
--- a/fs/reiserfs/xattr.c
+++ b/fs/reiserfs/xattr.c
@@ -870,11 +870,14 @@ out:
 	return err;
 }
 
-static int reiserfs_check_acl(struct inode *inode, int mask)
+static int reiserfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error = -EAGAIN; /* do regular unix permission checks by default */
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	acl = reiserfs_get_acl(inode, ACL_TYPE_ACCESS);
 
 	if (acl) {
@@ -951,8 +954,10 @@ static int xattr_mount_check(struct super_block *s)
 	return 0;
 }
 
-int reiserfs_permission(struct inode *inode, int mask)
+int reiserfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	/*
 	 * We don't do permission checks on the internal objects.
 	 * Permissions are determined by the "owning" object.
@@ -965,9 +970,10 @@ int reiserfs_permission(struct inode *inode, int mask)
 	 * Stat data v1 doesn't support ACLs.
 	 */
 	if (get_inode_sd_version(inode) != STAT_DATA_V1)
-		return generic_permission(inode, mask, reiserfs_check_acl);
+		return generic_permission(inode, mask, flags,
+					reiserfs_check_acl);
 #endif
-	return generic_permission(inode, mask, NULL);
+	return generic_permission(inode, mask, flags, NULL);
 }
 
 static int xattr_hide_revalidate(struct dentry *dentry, struct nameidata *nd)
diff --git a/fs/sysfs/inode.c b/fs/sysfs/inode.c
index cffb1fd..30ac273 100644
--- a/fs/sysfs/inode.c
+++ b/fs/sysfs/inode.c
@@ -348,13 +348,18 @@ int sysfs_hash_and_remove(struct sysfs_dirent *dir_sd, const void *ns, const cha
 		return -ENOENT;
 }
 
-int sysfs_permission(struct inode *inode, int mask)
+int sysfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	struct sysfs_dirent *sd = inode->i_private;
+	struct sysfs_dirent *sd;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	sd = inode->i_private;
 
 	mutex_lock(&sysfs_mutex);
 	sysfs_refresh_inode(sd, inode);
 	mutex_unlock(&sysfs_mutex);
 
-	return generic_permission(inode, mask, NULL);
+	return generic_permission(inode, mask, flags, NULL);
 }
diff --git a/fs/sysfs/sysfs.h b/fs/sysfs/sysfs.h
index d9be60a..ffaaa81 100644
--- a/fs/sysfs/sysfs.h
+++ b/fs/sysfs/sysfs.h
@@ -200,7 +200,7 @@ static inline void __sysfs_put(struct sysfs_dirent *sd)
 struct inode *sysfs_get_inode(struct super_block *sb, struct sysfs_dirent *sd);
 void sysfs_evict_inode(struct inode *inode);
 int sysfs_sd_setattr(struct sysfs_dirent *sd, struct iattr *iattr);
-int sysfs_permission(struct inode *inode, int mask);
+int sysfs_permission(struct inode *inode, int mask, unsigned int flags);
 int sysfs_setattr(struct dentry *dentry, struct iattr *iattr);
 int sysfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
 int sysfs_setxattr(struct dentry *dentry, const char *name, const void *value,
diff --git a/fs/xfs/linux-2.6/xfs_acl.c b/fs/xfs/linux-2.6/xfs_acl.c
index b277186..4b11eaf 100644
--- a/fs/xfs/linux-2.6/xfs_acl.c
+++ b/fs/xfs/linux-2.6/xfs_acl.c
@@ -219,12 +219,16 @@ xfs_set_acl(struct inode *inode, int type, struct posix_acl *acl)
 }
 
 int
-xfs_check_acl(struct inode *inode, int mask)
+xfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct xfs_inode *ip = XFS_I(inode);
+	struct xfs_inode *ip;
 	struct posix_acl *acl;
 	int error = -EAGAIN;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	ip = XFS_I(inode);
 	trace_xfs_check_acl(ip);
 
 	/*
diff --git a/fs/xfs/xfs_acl.h b/fs/xfs/xfs_acl.h
index 0135e2a..11dd720 100644
--- a/fs/xfs/xfs_acl.h
+++ b/fs/xfs/xfs_acl.h
@@ -42,7 +42,7 @@ struct xfs_acl {
 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
 
 #ifdef CONFIG_XFS_POSIX_ACL
-extern int xfs_check_acl(struct inode *inode, int mask);
+extern int xfs_check_acl(struct inode *inode, int mask, unsigned int flags);
 extern struct posix_acl *xfs_get_acl(struct inode *inode, int type);
 extern int xfs_inherit_acl(struct inode *inode, struct posix_acl *default_acl);
 extern int xfs_acl_chmod(struct inode *inode);
diff --git a/include/linux/coda_linux.h b/include/linux/coda_linux.h
index 2e914d0..4ccc59c 100644
--- a/include/linux/coda_linux.h
+++ b/include/linux/coda_linux.h
@@ -37,7 +37,7 @@ extern const struct file_operations coda_ioctl_operations;
 /* operations shared over more than one file */
 int coda_open(struct inode *i, struct file *f);
 int coda_release(struct inode *i, struct file *f);
-int coda_permission(struct inode *inode, int mask);
+int coda_permission(struct inode *inode, int mask, unsigned int flags);
 int coda_revalidate_inode(struct dentry *);
 int coda_getattr(struct vfsmount *, struct dentry *, struct kstat *);
 int coda_setattr(struct dentry *, struct iattr *);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index a04aa96..d5a4d42 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1550,11 +1550,13 @@ struct file_operations {
 	int (*setlease)(struct file *, long, struct file_lock **);
 };
 
+#define IPERM_FLAG_RCU	0x0001
+
 struct inode_operations {
 	struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *);
 	void * (*follow_link) (struct dentry *, struct nameidata *);
-	int (*permission) (struct inode *, int);
-	int (*check_acl)(struct inode *, int);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 
 	int (*readlink) (struct dentry *, char __user *,int);
 	void (*put_link) (struct dentry *, struct nameidata *, void *);
@@ -2165,8 +2167,8 @@ extern sector_t bmap(struct inode *, sector_t);
 #endif
 extern int notify_change(struct dentry *, struct iattr *);
 extern int inode_permission(struct inode *, int);
-extern int generic_permission(struct inode *, int,
-		int (*check_acl)(struct inode *, int));
+extern int generic_permission(struct inode *, int, unsigned int,
+		int (*check_acl)(struct inode *, int, unsigned int));
 
 static inline bool execute_ok(struct inode *inode)
 {
diff --git a/include/linux/generic_acl.h b/include/linux/generic_acl.h
index 574bea4..0437e37 100644
--- a/include/linux/generic_acl.h
+++ b/include/linux/generic_acl.h
@@ -10,6 +10,6 @@ extern const struct xattr_handler generic_acl_default_handler;
 
 int generic_acl_init(struct inode *, struct inode *);
 int generic_acl_chmod(struct inode *);
-int generic_check_acl(struct inode *inode, int mask);
+int generic_check_acl(struct inode *inode, int mask, unsigned int flags);
 
 #endif /* LINUX_GENERIC_ACL_H */
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 29d504d..0779bb8 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -351,7 +351,7 @@ extern int nfs_refresh_inode(struct inode *, struct nfs_fattr *);
 extern int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr);
 extern int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fattr);
 extern int nfs_getattr(struct vfsmount *, struct dentry *, struct kstat *);
-extern int nfs_permission(struct inode *, int);
+extern int nfs_permission(struct inode *, int, unsigned int);
 extern int nfs_open(struct inode *, struct file *);
 extern int nfs_release(struct inode *, struct file *);
 extern int nfs_attribute_timeout(struct inode *inode);
diff --git a/include/linux/reiserfs_xattr.h b/include/linux/reiserfs_xattr.h
index b2cf208..3b94c91 100644
--- a/include/linux/reiserfs_xattr.h
+++ b/include/linux/reiserfs_xattr.h
@@ -41,7 +41,7 @@ int reiserfs_xattr_init(struct super_block *sb, int mount_flags);
 int reiserfs_lookup_privroot(struct super_block *sb);
 int reiserfs_delete_xattrs(struct inode *inode);
 int reiserfs_chown_xattrs(struct inode *inode, struct iattr *attrs);
-int reiserfs_permission(struct inode *inode, int mask);
+int reiserfs_permission(struct inode *inode, int mask, unsigned int flags);
 
 #ifdef CONFIG_REISERFS_FS_XATTR
 #define has_xattr_dir(inode) (REISERFS_I(inode)->i_flags & i_has_xattr_dir)
-- 
cgit v0.10.2


From 1e1743ebe35ec7e3c1fa732408358fbc614cbbe5 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:49:59 +1100
Subject: fs: provide simple rcu-walk generic_check_acl implementation

This simple implementation just checks for no ACLs on the inode, and
if so, then the rcu-walk may proceed, otherwise fail it.

This could easily be extended to put acls under RCU and check them
under seqlock, if need be. But this implementation is enough to show
the rcu-walk aware permissions code for path lookups is working, and
will handle cases where there are no ACLs or ACLs in just the final
element.

This patch implicity converts tmpfs to rcu-aware permission check.
Subsequent patches onvert ext*, xfs, and, btrfs. Each of these uses
acl/permission code in a different way, so convert them all to provide
templates and proof of concept.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/generic_acl.c b/fs/generic_acl.c
index 6280042..06c48a8 100644
--- a/fs/generic_acl.c
+++ b/fs/generic_acl.c
@@ -192,16 +192,18 @@ generic_acl_chmod(struct inode *inode)
 int
 generic_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl;
-
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
-
-	acl = get_cached_acl(inode, ACL_TYPE_ACCESS);
-	if (acl) {
-		int error = posix_acl_permission(inode, acl, mask);
-		posix_acl_release(acl);
-		return error;
+	if (flags & IPERM_FLAG_RCU) {
+		if (!negative_cached_acl(inode, ACL_TYPE_ACCESS))
+			return -ECHILD;
+	} else {
+		struct posix_acl *acl;
+
+		acl = get_cached_acl(inode, ACL_TYPE_ACCESS);
+		if (acl) {
+			int error = posix_acl_permission(inode, acl, mask);
+			posix_acl_release(acl);
+			return error;
+		}
 	}
 	return -EAGAIN;
 }
diff --git a/include/linux/posix_acl.h b/include/linux/posix_acl.h
index 6760816..d68283a 100644
--- a/include/linux/posix_acl.h
+++ b/include/linux/posix_acl.h
@@ -108,6 +108,25 @@ static inline struct posix_acl *get_cached_acl(struct inode *inode, int type)
 	return acl;
 }
 
+static inline int negative_cached_acl(struct inode *inode, int type)
+{
+	struct posix_acl **p, *acl;
+	switch (type) {
+	case ACL_TYPE_ACCESS:
+		p = &inode->i_acl;
+		break;
+	case ACL_TYPE_DEFAULT:
+		p = &inode->i_default_acl;
+		break;
+	default:
+		BUG();
+	}
+	acl = ACCESS_ONCE(*p);
+	if (acl)
+		return 0;
+	return 1;
+}
+
 static inline void set_cached_acl(struct inode *inode,
 				  int type,
 				  struct posix_acl *acl)
-- 
cgit v0.10.2


From 73598611ade7c85f0c3d52ba5130103f6709c6d3 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:00 +1100
Subject: ext2,3,4: provide simple rcu-walk ACL implementation

This simple implementation just checks for no ACLs on the inode, and
if so, then the rcu-walk may proceed, otherwise fail it.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
index dd9bb3f..7b41805 100644
--- a/fs/ext2/acl.c
+++ b/fs/ext2/acl.c
@@ -236,8 +236,11 @@ ext2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
+	if (flags & IPERM_FLAG_RCU) {
+		if (!negative_cached_acl(inode, ACL_TYPE_ACCESS))
+			return -ECHILD;
+		return -EAGAIN;
+	}
 
 	acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
diff --git a/fs/ext3/acl.c b/fs/ext3/acl.c
index 9e49da8..e4fa49e 100644
--- a/fs/ext3/acl.c
+++ b/fs/ext3/acl.c
@@ -244,8 +244,11 @@ ext3_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
+	if (flags & IPERM_FLAG_RCU) {
+		if (!negative_cached_acl(inode, ACL_TYPE_ACCESS))
+			return -ECHILD;
+		return -EAGAIN;
+	}
 
 	acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
diff --git a/fs/ext4/acl.c b/fs/ext4/acl.c
index 373dcae..e0270d1 100644
--- a/fs/ext4/acl.c
+++ b/fs/ext4/acl.c
@@ -242,8 +242,11 @@ ext4_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
+	if (flags & IPERM_FLAG_RCU) {
+		if (!negative_cached_acl(inode, ACL_TYPE_ACCESS))
+			return -ECHILD;
+		return -EAGAIN;
+	}
 
 	acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
-- 
cgit v0.10.2


From 258a5aa8dfc6294f5f7df892023ee4d3e57c9841 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:01 +1100
Subject: btrfs: provide simple rcu-walk ACL implementation

This simple implementation just checks for no ACLs on the inode, and
if so, then the rcu-walk may proceed, otherwise fail it.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index cb518a4..6ae2c8c 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -187,18 +187,21 @@ static int btrfs_xattr_acl_set(struct dentry *dentry, const char *name,
 
 int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl;
 	int error = -EAGAIN;
 
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
+	if (flags & IPERM_FLAG_RCU) {
+		if (!negative_cached_acl(inode, ACL_TYPE_ACCESS))
+			error = -ECHILD;
 
-	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
-	if (IS_ERR(acl))
-		return PTR_ERR(acl);
-	if (acl) {
-		error = posix_acl_permission(inode, acl, mask);
-		posix_acl_release(acl);
+	} else {
+		struct posix_acl *acl;
+		acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
+		if (IS_ERR(acl))
+			return PTR_ERR(acl);
+		if (acl) {
+			error = posix_acl_permission(inode, acl, mask);
+			posix_acl_release(acl);
+		}
 	}
 
 	return error;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 5cf0db0..a0ff46a 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -7213,9 +7213,6 @@ static int btrfs_set_page_dirty(struct page *page)
 
 static int btrfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
-
 	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
 		return -EACCES;
 	return generic_permission(inode, mask, flags, btrfs_check_acl);
-- 
cgit v0.10.2


From 880566e17c4fce0d998d92610d32fcb9506aa6dd Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:02 +1100
Subject: xfs: provide simple rcu-walk ACL implementation

This simple implementation just checks for no ACLs on the inode, and
if so, then the rcu-walk may proceed, otherwise fail it.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/xfs/linux-2.6/xfs_acl.c b/fs/xfs/linux-2.6/xfs_acl.c
index 4b11eaf..39f4f80 100644
--- a/fs/xfs/linux-2.6/xfs_acl.c
+++ b/fs/xfs/linux-2.6/xfs_acl.c
@@ -225,9 +225,6 @@ xfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 	struct posix_acl *acl;
 	int error = -EAGAIN;
 
-	if (flags & IPERM_FLAG_RCU)
-		return -ECHILD;
-
 	ip = XFS_I(inode);
 	trace_xfs_check_acl(ip);
 
@@ -238,6 +235,12 @@ xfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 	if (!XFS_IFORK_Q(ip))
 		return -EAGAIN;
 
+	if (flags & IPERM_FLAG_RCU) {
+		if (!negative_cached_acl(inode, ACL_TYPE_ACCESS))
+			return -ECHILD;
+		return -EAGAIN;
+	}
+
 	acl = xfs_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
-- 
cgit v0.10.2


From 4e35e6070b1ceed89c3bba2af4216c286fb1dafd Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:03 +1100
Subject: kernel: add bl_list

Introduce a type of hlist that can support the use of the lowest bit in the
hlist_head. This will be subsequently used to implement per-bucket bit spinlock
for inode and dentry hashes, and may be useful in other cases such as network
hashes.

Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/include/linux/list_bl.h b/include/linux/list_bl.h
new file mode 100644
index 0000000..9ee97e7
--- /dev/null
+++ b/include/linux/list_bl.h
@@ -0,0 +1,144 @@
+#ifndef _LINUX_LIST_BL_H
+#define _LINUX_LIST_BL_H
+
+#include <linux/list.h>
+
+/*
+ * Special version of lists, where head of the list has a lock in the lowest
+ * bit. This is useful for scalable hash tables without increasing memory
+ * footprint overhead.
+ *
+ * For modification operations, the 0 bit of hlist_bl_head->first
+ * pointer must be set.
+ *
+ * With some small modifications, this can easily be adapted to store several
+ * arbitrary bits (not just a single lock bit), if the need arises to store
+ * some fast and compact auxiliary data.
+ */
+
+#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
+#define LIST_BL_LOCKMASK	1UL
+#else
+#define LIST_BL_LOCKMASK	0UL
+#endif
+
+#ifdef CONFIG_DEBUG_LIST
+#define LIST_BL_BUG_ON(x) BUG_ON(x)
+#else
+#define LIST_BL_BUG_ON(x)
+#endif
+
+
+struct hlist_bl_head {
+	struct hlist_bl_node *first;
+};
+
+struct hlist_bl_node {
+	struct hlist_bl_node *next, **pprev;
+};
+#define INIT_HLIST_BL_HEAD(ptr) \
+	((ptr)->first = NULL)
+
+static inline void INIT_HLIST_BL_NODE(struct hlist_bl_node *h)
+{
+	h->next = NULL;
+	h->pprev = NULL;
+}
+
+#define hlist_bl_entry(ptr, type, member) container_of(ptr,type,member)
+
+static inline int hlist_bl_unhashed(const struct hlist_bl_node *h)
+{
+	return !h->pprev;
+}
+
+static inline struct hlist_bl_node *hlist_bl_first(struct hlist_bl_head *h)
+{
+	return (struct hlist_bl_node *)
+		((unsigned long)h->first & ~LIST_BL_LOCKMASK);
+}
+
+static inline void hlist_bl_set_first(struct hlist_bl_head *h,
+					struct hlist_bl_node *n)
+{
+	LIST_BL_BUG_ON((unsigned long)n & LIST_BL_LOCKMASK);
+	LIST_BL_BUG_ON(!((unsigned long)h->first & LIST_BL_LOCKMASK));
+	h->first = (struct hlist_bl_node *)((unsigned long)n | LIST_BL_LOCKMASK);
+}
+
+static inline int hlist_bl_empty(const struct hlist_bl_head *h)
+{
+	return !((unsigned long)h->first & ~LIST_BL_LOCKMASK);
+}
+
+static inline void hlist_bl_add_head(struct hlist_bl_node *n,
+					struct hlist_bl_head *h)
+{
+	struct hlist_bl_node *first = hlist_bl_first(h);
+
+	n->next = first;
+	if (first)
+		first->pprev = &n->next;
+	n->pprev = &h->first;
+	hlist_bl_set_first(h, n);
+}
+
+static inline void __hlist_bl_del(struct hlist_bl_node *n)
+{
+	struct hlist_bl_node *next = n->next;
+	struct hlist_bl_node **pprev = n->pprev;
+
+	LIST_BL_BUG_ON((unsigned long)n & LIST_BL_LOCKMASK);
+
+	/* pprev may be `first`, so be careful not to lose the lock bit */
+	*pprev = (struct hlist_bl_node *)
+			((unsigned long)next |
+			 ((unsigned long)*pprev & LIST_BL_LOCKMASK));
+	if (next)
+		next->pprev = pprev;
+}
+
+static inline void hlist_bl_del(struct hlist_bl_node *n)
+{
+	__hlist_bl_del(n);
+	n->next = LIST_POISON1;
+	n->pprev = LIST_POISON2;
+}
+
+static inline void hlist_bl_del_init(struct hlist_bl_node *n)
+{
+	if (!hlist_bl_unhashed(n)) {
+		__hlist_bl_del(n);
+		INIT_HLIST_BL_NODE(n);
+	}
+}
+
+/**
+ * hlist_bl_for_each_entry	- iterate over list of given type
+ * @tpos:	the type * to use as a loop cursor.
+ * @pos:	the &struct hlist_node to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the hlist_node within the struct.
+ *
+ */
+#define hlist_bl_for_each_entry(tpos, pos, head, member)		\
+	for (pos = hlist_bl_first(head);				\
+	     pos &&							\
+		({ tpos = hlist_bl_entry(pos, typeof(*tpos), member); 1;}); \
+	     pos = pos->next)
+
+/**
+ * hlist_bl_for_each_entry_safe - iterate over list of given type safe against removal of list entry
+ * @tpos:	the type * to use as a loop cursor.
+ * @pos:	the &struct hlist_node to use as a loop cursor.
+ * @n:		another &struct hlist_node to use as temporary storage
+ * @head:	the head for your list.
+ * @member:	the name of the hlist_node within the struct.
+ */
+#define hlist_bl_for_each_entry_safe(tpos, pos, n, head, member)	 \
+	for (pos = hlist_bl_first(head);				 \
+	     pos && ({ n = pos->next; 1; }) && 				 \
+		({ tpos = hlist_bl_entry(pos, typeof(*tpos), member); 1;}); \
+	     pos = n)
+
+#endif
diff --git a/include/linux/rculist_bl.h b/include/linux/rculist_bl.h
new file mode 100644
index 0000000..b872b49
--- /dev/null
+++ b/include/linux/rculist_bl.h
@@ -0,0 +1,127 @@
+#ifndef _LINUX_RCULIST_BL_H
+#define _LINUX_RCULIST_BL_H
+
+/*
+ * RCU-protected bl list version. See include/linux/list_bl.h.
+ */
+#include <linux/list_bl.h>
+#include <linux/rcupdate.h>
+
+static inline void hlist_bl_set_first_rcu(struct hlist_bl_head *h,
+					struct hlist_bl_node *n)
+{
+	LIST_BL_BUG_ON((unsigned long)n & LIST_BL_LOCKMASK);
+	LIST_BL_BUG_ON(!((unsigned long)h->first & LIST_BL_LOCKMASK));
+	rcu_assign_pointer(h->first,
+		(struct hlist_bl_node *)((unsigned long)n | LIST_BL_LOCKMASK));
+}
+
+static inline struct hlist_bl_node *hlist_bl_first_rcu(struct hlist_bl_head *h)
+{
+	return (struct hlist_bl_node *)
+		((unsigned long)rcu_dereference(h->first) & ~LIST_BL_LOCKMASK);
+}
+
+/**
+ * hlist_bl_del_init_rcu - deletes entry from hash list with re-initialization
+ * @n: the element to delete from the hash list.
+ *
+ * Note: hlist_bl_unhashed() on the node returns true after this. It is
+ * useful for RCU based read lockfree traversal if the writer side
+ * must know if the list entry is still hashed or already unhashed.
+ *
+ * In particular, it means that we can not poison the forward pointers
+ * that may still be used for walking the hash list and we can only
+ * zero the pprev pointer so list_unhashed() will return true after
+ * this.
+ *
+ * The caller must take whatever precautions are necessary (such as
+ * holding appropriate locks) to avoid racing with another
+ * list-mutation primitive, such as hlist_bl_add_head_rcu() or
+ * hlist_bl_del_rcu(), running on this same list.  However, it is
+ * perfectly legal to run concurrently with the _rcu list-traversal
+ * primitives, such as hlist_bl_for_each_entry_rcu().
+ */
+static inline void hlist_bl_del_init_rcu(struct hlist_bl_node *n)
+{
+	if (!hlist_bl_unhashed(n)) {
+		__hlist_bl_del(n);
+		n->pprev = NULL;
+	}
+}
+
+/**
+ * hlist_bl_del_rcu - deletes entry from hash list without re-initialization
+ * @n: the element to delete from the hash list.
+ *
+ * Note: hlist_bl_unhashed() on entry does not return true after this,
+ * the entry is in an undefined state. It is useful for RCU based
+ * lockfree traversal.
+ *
+ * In particular, it means that we can not poison the forward
+ * pointers that may still be used for walking the hash list.
+ *
+ * The caller must take whatever precautions are necessary
+ * (such as holding appropriate locks) to avoid racing
+ * with another list-mutation primitive, such as hlist_bl_add_head_rcu()
+ * or hlist_bl_del_rcu(), running on this same list.
+ * However, it is perfectly legal to run concurrently with
+ * the _rcu list-traversal primitives, such as
+ * hlist_bl_for_each_entry().
+ */
+static inline void hlist_bl_del_rcu(struct hlist_bl_node *n)
+{
+	__hlist_bl_del(n);
+	n->pprev = LIST_POISON2;
+}
+
+/**
+ * hlist_bl_add_head_rcu
+ * @n: the element to add to the hash list.
+ * @h: the list to add to.
+ *
+ * Description:
+ * Adds the specified element to the specified hlist_bl,
+ * while permitting racing traversals.
+ *
+ * The caller must take whatever precautions are necessary
+ * (such as holding appropriate locks) to avoid racing
+ * with another list-mutation primitive, such as hlist_bl_add_head_rcu()
+ * or hlist_bl_del_rcu(), running on this same list.
+ * However, it is perfectly legal to run concurrently with
+ * the _rcu list-traversal primitives, such as
+ * hlist_bl_for_each_entry_rcu(), used to prevent memory-consistency
+ * problems on Alpha CPUs.  Regardless of the type of CPU, the
+ * list-traversal primitive must be guarded by rcu_read_lock().
+ */
+static inline void hlist_bl_add_head_rcu(struct hlist_bl_node *n,
+					struct hlist_bl_head *h)
+{
+	struct hlist_bl_node *first;
+
+	/* don't need hlist_bl_first_rcu because we're under lock */
+	first = hlist_bl_first(h);
+
+	n->next = first;
+	if (first)
+		first->pprev = &n->next;
+	n->pprev = &h->first;
+
+	/* need _rcu because we can have concurrent lock free readers */
+	hlist_bl_set_first_rcu(h, n);
+}
+/**
+ * hlist_bl_for_each_entry_rcu - iterate over rcu list of given type
+ * @tpos:	the type * to use as a loop cursor.
+ * @pos:	the &struct hlist_bl_node to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the hlist_bl_node within the struct.
+ *
+ */
+#define hlist_bl_for_each_entry_rcu(tpos, pos, head, member)		\
+	for (pos = hlist_bl_first_rcu(head);				\
+		pos &&							\
+		({ tpos = hlist_bl_entry(pos, typeof(*tpos), member); 1; }); \
+		pos = rcu_dereference_raw(pos->next))
+
+#endif
-- 
cgit v0.10.2


From 626d607435617cc0f033522083e2bb195b81813c Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:04 +1100
Subject: bit_spinlock: add required includes

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/include/linux/bit_spinlock.h b/include/linux/bit_spinlock.h
index 7113a32..e612575 100644
--- a/include/linux/bit_spinlock.h
+++ b/include/linux/bit_spinlock.h
@@ -1,6 +1,10 @@
 #ifndef __LINUX_BIT_SPINLOCK_H
 #define __LINUX_BIT_SPINLOCK_H
 
+#include <linux/kernel.h>
+#include <linux/preempt.h>
+#include <asm/atomic.h>
+
 /*
  *  bit-based spin_lock()
  *
-- 
cgit v0.10.2


From ceb5bdc2d246f6d81cf61ed70f325308a11821d2 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:05 +1100
Subject: fs: dcache per-bucket dcache hash locking

We can turn the dcache hash locking from a global dcache_hash_lock into
per-bucket locking.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 07d1f68..9f04e1b 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -33,14 +33,18 @@
 #include <linux/bootmem.h>
 #include <linux/fs_struct.h>
 #include <linux/hardirq.h>
+#include <linux/bit_spinlock.h>
+#include <linux/rculist_bl.h>
 #include "internal.h"
 
 /*
  * Usage:
  * dcache_inode_lock protects:
  *   - i_dentry, d_alias, d_inode
- * dcache_hash_lock protects:
- *   - the dcache hash table, s_anon lists
+ * dcache_hash_bucket lock protects:
+ *   - the dcache hash table
+ * s_anon bl list spinlock protects:
+ *   - the s_anon list (see __d_drop)
  * dcache_lru_lock protects:
  *   - the dcache lru lists and counters
  * d_lock protects:
@@ -57,7 +61,8 @@
  * dcache_inode_lock
  *   dentry->d_lock
  *     dcache_lru_lock
- *     dcache_hash_lock
+ *     dcache_hash_bucket lock
+ *     s_anon lock
  *
  * If there is an ancestor relationship:
  * dentry->d_parent->...->d_parent->d_lock
@@ -74,7 +79,6 @@ int sysctl_vfs_cache_pressure __read_mostly = 100;
 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
 
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_inode_lock);
-static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_hash_lock);
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
@@ -96,7 +100,29 @@ static struct kmem_cache *dentry_cache __read_mostly;
 
 static unsigned int d_hash_mask __read_mostly;
 static unsigned int d_hash_shift __read_mostly;
-static struct hlist_head *dentry_hashtable __read_mostly;
+
+struct dcache_hash_bucket {
+	struct hlist_bl_head head;
+};
+static struct dcache_hash_bucket *dentry_hashtable __read_mostly;
+
+static inline struct dcache_hash_bucket *d_hash(struct dentry *parent,
+					unsigned long hash)
+{
+	hash += ((unsigned long) parent ^ GOLDEN_RATIO_PRIME) / L1_CACHE_BYTES;
+	hash = hash ^ ((hash ^ GOLDEN_RATIO_PRIME) >> D_HASHBITS);
+	return dentry_hashtable + (hash & D_HASHMASK);
+}
+
+static inline void spin_lock_bucket(struct dcache_hash_bucket *b)
+{
+	bit_spin_lock(0, (unsigned long *)&b->head.first);
+}
+
+static inline void spin_unlock_bucket(struct dcache_hash_bucket *b)
+{
+	__bit_spin_unlock(0, (unsigned long *)&b->head.first);
+}
 
 /* Statistics gathering. */
 struct dentry_stat_t dentry_stat = {
@@ -144,7 +170,7 @@ static void d_free(struct dentry *dentry)
 		dentry->d_op->d_release(dentry);
 
 	/* if dentry was never inserted into hash, immediate free is OK */
-	if (hlist_unhashed(&dentry->d_hash))
+	if (hlist_bl_unhashed(&dentry->d_hash))
 		__d_free(&dentry->d_u.d_rcu);
 	else
 		call_rcu(&dentry->d_u.d_rcu, __d_free);
@@ -302,11 +328,27 @@ static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 void __d_drop(struct dentry *dentry)
 {
 	if (!(dentry->d_flags & DCACHE_UNHASHED)) {
-		dentry->d_flags |= DCACHE_UNHASHED;
-		spin_lock(&dcache_hash_lock);
-		hlist_del_rcu(&dentry->d_hash);
-		spin_unlock(&dcache_hash_lock);
-		dentry_rcuwalk_barrier(dentry);
+		if (unlikely(dentry->d_flags & DCACHE_DISCONNECTED)) {
+			bit_spin_lock(0,
+				(unsigned long *)&dentry->d_sb->s_anon.first);
+			dentry->d_flags |= DCACHE_UNHASHED;
+			hlist_bl_del_init(&dentry->d_hash);
+			__bit_spin_unlock(0,
+				(unsigned long *)&dentry->d_sb->s_anon.first);
+		} else {
+			struct dcache_hash_bucket *b;
+			b = d_hash(dentry->d_parent, dentry->d_name.hash);
+			spin_lock_bucket(b);
+			/*
+			 * We may not actually need to put DCACHE_UNHASHED
+			 * manipulations under the hash lock, but follow
+			 * the principle of least surprise.
+			 */
+			dentry->d_flags |= DCACHE_UNHASHED;
+			hlist_bl_del_rcu(&dentry->d_hash);
+			spin_unlock_bucket(b);
+			dentry_rcuwalk_barrier(dentry);
+		}
 	}
 }
 EXPORT_SYMBOL(__d_drop);
@@ -961,8 +1003,8 @@ void shrink_dcache_for_umount(struct super_block *sb)
 	spin_unlock(&dentry->d_lock);
 	shrink_dcache_for_umount_subtree(dentry);
 
-	while (!hlist_empty(&sb->s_anon)) {
-		dentry = hlist_entry(sb->s_anon.first, struct dentry, d_hash);
+	while (!hlist_bl_empty(&sb->s_anon)) {
+		dentry = hlist_bl_entry(hlist_bl_first(&sb->s_anon), struct dentry, d_hash);
 		shrink_dcache_for_umount_subtree(dentry);
 	}
 }
@@ -1263,7 +1305,7 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	dentry->d_sb = NULL;
 	dentry->d_op = NULL;
 	dentry->d_fsdata = NULL;
-	INIT_HLIST_NODE(&dentry->d_hash);
+	INIT_HLIST_BL_NODE(&dentry->d_hash);
 	INIT_LIST_HEAD(&dentry->d_lru);
 	INIT_LIST_HEAD(&dentry->d_subdirs);
 	INIT_LIST_HEAD(&dentry->d_alias);
@@ -1459,14 +1501,6 @@ struct dentry * d_alloc_root(struct inode * root_inode)
 }
 EXPORT_SYMBOL(d_alloc_root);
 
-static inline struct hlist_head *d_hash(struct dentry *parent,
-					unsigned long hash)
-{
-	hash += ((unsigned long) parent ^ GOLDEN_RATIO_PRIME) / L1_CACHE_BYTES;
-	hash = hash ^ ((hash ^ GOLDEN_RATIO_PRIME) >> D_HASHBITS);
-	return dentry_hashtable + (hash & D_HASHMASK);
-}
-
 /**
  * d_obtain_alias - find or allocate a dentry for a given inode
  * @inode: inode to allocate the dentry for
@@ -1521,11 +1555,11 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	tmp->d_sb = inode->i_sb;
 	tmp->d_inode = inode;
 	tmp->d_flags |= DCACHE_DISCONNECTED;
-	tmp->d_flags &= ~DCACHE_UNHASHED;
 	list_add(&tmp->d_alias, &inode->i_dentry);
-	spin_lock(&dcache_hash_lock);
-	hlist_add_head(&tmp->d_hash, &inode->i_sb->s_anon);
-	spin_unlock(&dcache_hash_lock);
+	bit_spin_lock(0, (unsigned long *)&tmp->d_sb->s_anon.first);
+	tmp->d_flags &= ~DCACHE_UNHASHED;
+	hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
+	__bit_spin_unlock(0, (unsigned long *)&tmp->d_sb->s_anon.first);
 	spin_unlock(&tmp->d_lock);
 	spin_unlock(&dcache_inode_lock);
 
@@ -1567,7 +1601,7 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 			d_move(new, dentry);
 			iput(inode);
 		} else {
-			/* already taking dcache_inode_lock, so d_add() by hand */
+			/* already got dcache_inode_lock, so d_add() by hand */
 			__d_instantiate(dentry, inode);
 			spin_unlock(&dcache_inode_lock);
 			security_d_instantiate(dentry, inode);
@@ -1702,8 +1736,8 @@ struct dentry *__d_lookup_rcu(struct dentry *parent, struct qstr *name,
 	unsigned int len = name->len;
 	unsigned int hash = name->hash;
 	const unsigned char *str = name->name;
-	struct hlist_head *head = d_hash(parent, hash);
-	struct hlist_node *node;
+	struct dcache_hash_bucket *b = d_hash(parent, hash);
+	struct hlist_bl_node *node;
 	struct dentry *dentry;
 
 	/*
@@ -1726,7 +1760,7 @@ struct dentry *__d_lookup_rcu(struct dentry *parent, struct qstr *name,
 	 *
 	 * See Documentation/vfs/dcache-locking.txt for more details.
 	 */
-	hlist_for_each_entry_rcu(dentry, node, head, d_hash) {
+	hlist_bl_for_each_entry_rcu(dentry, node, &b->head, d_hash) {
 		struct inode *i;
 		const char *tname;
 		int tlen;
@@ -1820,8 +1854,8 @@ struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
 	unsigned int len = name->len;
 	unsigned int hash = name->hash;
 	const unsigned char *str = name->name;
-	struct hlist_head *head = d_hash(parent,hash);
-	struct hlist_node *node;
+	struct dcache_hash_bucket *b = d_hash(parent, hash);
+	struct hlist_bl_node *node;
 	struct dentry *found = NULL;
 	struct dentry *dentry;
 
@@ -1847,7 +1881,7 @@ struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
 	 */
 	rcu_read_lock();
 	
-	hlist_for_each_entry_rcu(dentry, node, head, d_hash) {
+	hlist_bl_for_each_entry_rcu(dentry, node, &b->head, d_hash) {
 		const char *tname;
 		int tlen;
 
@@ -1998,11 +2032,13 @@ again:
 }
 EXPORT_SYMBOL(d_delete);
 
-static void __d_rehash(struct dentry * entry, struct hlist_head *list)
+static void __d_rehash(struct dentry * entry, struct dcache_hash_bucket *b)
 {
-
+	BUG_ON(!d_unhashed(entry));
+	spin_lock_bucket(b);
  	entry->d_flags &= ~DCACHE_UNHASHED;
- 	hlist_add_head_rcu(&entry->d_hash, list);
+	hlist_bl_add_head_rcu(&entry->d_hash, &b->head);
+	spin_unlock_bucket(b);
 }
 
 static void _d_rehash(struct dentry * entry)
@@ -2020,9 +2056,7 @@ static void _d_rehash(struct dentry * entry)
 void d_rehash(struct dentry * entry)
 {
 	spin_lock(&entry->d_lock);
-	spin_lock(&dcache_hash_lock);
 	_d_rehash(entry);
-	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&entry->d_lock);
 }
 EXPORT_SYMBOL(d_rehash);
@@ -2165,15 +2199,16 @@ void d_move(struct dentry * dentry, struct dentry * target)
 	write_seqcount_begin(&dentry->d_seq);
 	write_seqcount_begin(&target->d_seq);
 
-	/* Move the dentry to the target hash queue, if on different bucket */
-	spin_lock(&dcache_hash_lock);
-	if (!d_unhashed(dentry))
-		hlist_del_rcu(&dentry->d_hash);
+	/* __d_drop does write_seqcount_barrier, but they're OK to nest. */
+
+	/*
+	 * Move the dentry to the target hash queue. Don't bother checking
+	 * for the same hash queue because of how unlikely it is.
+	 */
+	__d_drop(dentry);
 	__d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
-	spin_unlock(&dcache_hash_lock);
 
 	/* Unhash the target: dput() will then get rid of it */
-	/* __d_drop does write_seqcount_barrier, but they're OK to nest. */
 	__d_drop(target);
 
 	list_del(&dentry->d_u.d_child);
@@ -2369,9 +2404,7 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 
 	spin_lock(&actual->d_lock);
 found:
-	spin_lock(&dcache_hash_lock);
 	_d_rehash(actual);
-	spin_unlock(&dcache_hash_lock);
 	spin_unlock(&actual->d_lock);
 	spin_unlock(&dcache_inode_lock);
 out_nolock:
@@ -2953,7 +2986,7 @@ static void __init dcache_init_early(void)
 
 	dentry_hashtable =
 		alloc_large_system_hash("Dentry cache",
-					sizeof(struct hlist_head),
+					sizeof(struct dcache_hash_bucket),
 					dhash_entries,
 					13,
 					HASH_EARLY,
@@ -2962,7 +2995,7 @@ static void __init dcache_init_early(void)
 					0);
 
 	for (loop = 0; loop < (1 << d_hash_shift); loop++)
-		INIT_HLIST_HEAD(&dentry_hashtable[loop]);
+		INIT_HLIST_BL_HEAD(&dentry_hashtable[loop].head);
 }
 
 static void __init dcache_init(void)
@@ -2985,7 +3018,7 @@ static void __init dcache_init(void)
 
 	dentry_hashtable =
 		alloc_large_system_hash("Dentry cache",
-					sizeof(struct hlist_head),
+					sizeof(struct dcache_hash_bucket),
 					dhash_entries,
 					13,
 					0,
@@ -2994,7 +3027,7 @@ static void __init dcache_init(void)
 					0);
 
 	for (loop = 0; loop < (1 << d_hash_shift); loop++)
-		INIT_HLIST_HEAD(&dentry_hashtable[loop]);
+		INIT_HLIST_BL_HEAD(&dentry_hashtable[loop].head);
 }
 
 /* SLAB cache for __getname() consumers */
diff --git a/fs/super.c b/fs/super.c
index ca69615..968ba01 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -30,6 +30,7 @@
 #include <linux/idr.h>
 #include <linux/mutex.h>
 #include <linux/backing-dev.h>
+#include <linux/rculist_bl.h>
 #include "internal.h"
 
 
@@ -71,7 +72,7 @@ static struct super_block *alloc_super(struct file_system_type *type)
 		INIT_LIST_HEAD(&s->s_files);
 #endif
 		INIT_LIST_HEAD(&s->s_instances);
-		INIT_HLIST_HEAD(&s->s_anon);
+		INIT_HLIST_BL_HEAD(&s->s_anon);
 		INIT_LIST_HEAD(&s->s_inodes);
 		INIT_LIST_HEAD(&s->s_dentry_lru);
 		init_rwsem(&s->s_umount);
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 8b2064d..5f0392e 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -4,6 +4,7 @@
 #include <asm/atomic.h>
 #include <linux/list.h>
 #include <linux/rculist.h>
+#include <linux/rculist_bl.h>
 #include <linux/spinlock.h>
 #include <linux/seqlock.h>
 #include <linux/cache.h>
@@ -95,7 +96,7 @@ struct dentry {
 	/* RCU lookup touched fields */
 	unsigned int d_flags;		/* protected by d_lock */
 	seqcount_t d_seq;		/* per dentry seqlock */
-	struct hlist_node d_hash;	/* lookup hash list */
+	struct hlist_bl_node d_hash;	/* lookup hash list */
 	struct dentry *d_parent;	/* parent directory */
 	struct qstr d_name;
 	struct inode *d_inode;		/* Where the name belongs to - NULL is
diff --git a/include/linux/fs.h b/include/linux/fs.h
index d5a4d42..baf3e55 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -392,6 +392,7 @@ struct inodes_stat_t {
 #include <linux/capability.h>
 #include <linux/semaphore.h>
 #include <linux/fiemap.h>
+#include <linux/rculist_bl.h>
 
 #include <asm/atomic.h>
 #include <asm/byteorder.h>
@@ -1377,7 +1378,7 @@ struct super_block {
 	const struct xattr_handler **s_xattr;
 
 	struct list_head	s_inodes;	/* all inodes */
-	struct hlist_head	s_anon;		/* anonymous dentries for (nfs) exporting */
+	struct hlist_bl_head	s_anon;		/* anonymous dentries for (nfs) exporting */
 #ifdef CONFIG_SMP
 	struct list_head __percpu *s_files;
 #else
-- 
cgit v0.10.2


From 873feea09ebc980cbd3631b767356ce1eee65ec1 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:06 +1100
Subject: fs: dcache per-inode inode alias locking

dcache_inode_lock can be replaced with per-inode locking. Use existing
inode->i_lock for this. This is slightly non-trivial because we sometimes
need to find the inode from the dentry, which requires d_inode to be
stabilised (either with refcount or d_lock).

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index df8bbb3..5978298 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -277,11 +277,11 @@ static struct dentry *v9fs_dentry_from_dir_inode(struct inode *inode)
 {
 	struct dentry *dentry;
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	/* Directory should have only one entry. */
 	BUG_ON(S_ISDIR(inode->i_mode) && !list_is_singular(&inode->i_dentry));
 	dentry = list_entry(inode->i_dentry.next, struct dentry, d_alias);
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 	return dentry;
 }
 
diff --git a/fs/affs/amigaffs.c b/fs/affs/amigaffs.c
index 600101a..3a4557e 100644
--- a/fs/affs/amigaffs.c
+++ b/fs/affs/amigaffs.c
@@ -128,7 +128,7 @@ affs_fix_dcache(struct dentry *dentry, u32 entry_ino)
 	void *data = dentry->d_fsdata;
 	struct list_head *head, *next;
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	head = &inode->i_dentry;
 	next = head->next;
 	while (next != head) {
@@ -139,7 +139,7 @@ affs_fix_dcache(struct dentry *dentry, u32 entry_ino)
 		}
 		next = next->next;
 	}
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 }
 
 
diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index 2a239d8..a853a89 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -809,14 +809,14 @@ inode_has_hashed_dentries(struct inode *inode)
 {
 	struct dentry *dentry;
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		if (!d_unhashed(dentry) || IS_ROOT(dentry)) {
-			spin_unlock(&dcache_inode_lock);
+			spin_unlock(&inode->i_lock);
 			return true;
 		}
 	}
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 	return false;
 }
 
diff --git a/fs/dcache.c b/fs/dcache.c
index 9f04e1b..09ec945 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -39,8 +39,8 @@
 
 /*
  * Usage:
- * dcache_inode_lock protects:
- *   - i_dentry, d_alias, d_inode
+ * dcache->d_inode->i_lock protects:
+ *   - i_dentry, d_alias, d_inode of aliases
  * dcache_hash_bucket lock protects:
  *   - the dcache hash table
  * s_anon bl list spinlock protects:
@@ -58,7 +58,7 @@
  *   - d_alias, d_inode
  *
  * Ordering:
- * dcache_inode_lock
+ * dentry->d_inode->i_lock
  *   dentry->d_lock
  *     dcache_lru_lock
  *     dcache_hash_bucket lock
@@ -78,12 +78,10 @@
 int sysctl_vfs_cache_pressure __read_mostly = 100;
 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
 
-__cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_inode_lock);
 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
 
 EXPORT_SYMBOL(rename_lock);
-EXPORT_SYMBOL(dcache_inode_lock);
 
 static struct kmem_cache *dentry_cache __read_mostly;
 
@@ -196,14 +194,14 @@ static inline void dentry_rcuwalk_barrier(struct dentry *dentry)
  */
 static void dentry_iput(struct dentry * dentry)
 	__releases(dentry->d_lock)
-	__releases(dcache_inode_lock)
+	__releases(dentry->d_inode->i_lock)
 {
 	struct inode *inode = dentry->d_inode;
 	if (inode) {
 		dentry->d_inode = NULL;
 		list_del_init(&dentry->d_alias);
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_inode_lock);
+		spin_unlock(&inode->i_lock);
 		if (!inode->i_nlink)
 			fsnotify_inoderemove(inode);
 		if (dentry->d_op && dentry->d_op->d_iput)
@@ -212,7 +210,6 @@ static void dentry_iput(struct dentry * dentry)
 			iput(inode);
 	} else {
 		spin_unlock(&dentry->d_lock);
-		spin_unlock(&dcache_inode_lock);
 	}
 }
 
@@ -222,14 +219,14 @@ static void dentry_iput(struct dentry * dentry)
  */
 static void dentry_unlink_inode(struct dentry * dentry)
 	__releases(dentry->d_lock)
-	__releases(dcache_inode_lock)
+	__releases(dentry->d_inode->i_lock)
 {
 	struct inode *inode = dentry->d_inode;
 	dentry->d_inode = NULL;
 	list_del_init(&dentry->d_alias);
 	dentry_rcuwalk_barrier(dentry);
 	spin_unlock(&dentry->d_lock);
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 	if (!inode->i_nlink)
 		fsnotify_inoderemove(inode);
 	if (dentry->d_op && dentry->d_op->d_iput)
@@ -295,7 +292,7 @@ static void dentry_lru_move_tail(struct dentry *dentry)
 static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
 	__releases(parent->d_lock)
-	__releases(dcache_inode_lock)
+	__releases(dentry->d_inode->i_lock)
 {
 	dentry->d_parent = NULL;
 	list_del(&dentry->d_u.d_child);
@@ -370,9 +367,11 @@ EXPORT_SYMBOL(d_drop);
 static inline struct dentry *dentry_kill(struct dentry *dentry, int ref)
 	__releases(dentry->d_lock)
 {
+	struct inode *inode;
 	struct dentry *parent;
 
-	if (!spin_trylock(&dcache_inode_lock)) {
+	inode = dentry->d_inode;
+	if (inode && !spin_trylock(&inode->i_lock)) {
 relock:
 		spin_unlock(&dentry->d_lock);
 		cpu_relax();
@@ -383,7 +382,8 @@ relock:
 	else
 		parent = dentry->d_parent;
 	if (parent && !spin_trylock(&parent->d_lock)) {
-		spin_unlock(&dcache_inode_lock);
+		if (inode)
+			spin_unlock(&inode->i_lock);
 		goto relock;
 	}
 
@@ -618,9 +618,9 @@ struct dentry *d_find_alias(struct inode *inode)
 	struct dentry *de = NULL;
 
 	if (!list_empty(&inode->i_dentry)) {
-		spin_lock(&dcache_inode_lock);
+		spin_lock(&inode->i_lock);
 		de = __d_find_alias(inode, 0);
-		spin_unlock(&dcache_inode_lock);
+		spin_unlock(&inode->i_lock);
 	}
 	return de;
 }
@@ -634,20 +634,20 @@ void d_prune_aliases(struct inode *inode)
 {
 	struct dentry *dentry;
 restart:
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		spin_lock(&dentry->d_lock);
 		if (!dentry->d_count) {
 			__dget_dlock(dentry);
 			__d_drop(dentry);
 			spin_unlock(&dentry->d_lock);
-			spin_unlock(&dcache_inode_lock);
+			spin_unlock(&inode->i_lock);
 			dput(dentry);
 			goto restart;
 		}
 		spin_unlock(&dentry->d_lock);
 	}
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 }
 EXPORT_SYMBOL(d_prune_aliases);
 
@@ -1392,9 +1392,11 @@ static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 void d_instantiate(struct dentry *entry, struct inode * inode)
 {
 	BUG_ON(!list_empty(&entry->d_alias));
-	spin_lock(&dcache_inode_lock);
+	if (inode)
+		spin_lock(&inode->i_lock);
 	__d_instantiate(entry, inode);
-	spin_unlock(&dcache_inode_lock);
+	if (inode)
+		spin_unlock(&inode->i_lock);
 	security_d_instantiate(entry, inode);
 }
 EXPORT_SYMBOL(d_instantiate);
@@ -1458,9 +1460,11 @@ struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
 
 	BUG_ON(!list_empty(&entry->d_alias));
 
-	spin_lock(&dcache_inode_lock);
+	if (inode)
+		spin_lock(&inode->i_lock);
 	result = __d_instantiate_unique(entry, inode);
-	spin_unlock(&dcache_inode_lock);
+	if (inode)
+		spin_unlock(&inode->i_lock);
 
 	if (!result) {
 		security_d_instantiate(entry, inode);
@@ -1542,10 +1546,10 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	tmp->d_parent = tmp; /* make sure dput doesn't croak */
 
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	res = __d_find_alias(inode, 0);
 	if (res) {
-		spin_unlock(&dcache_inode_lock);
+		spin_unlock(&inode->i_lock);
 		dput(tmp);
 		goto out_iput;
 	}
@@ -1561,7 +1565,7 @@ struct dentry *d_obtain_alias(struct inode *inode)
 	hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
 	__bit_spin_unlock(0, (unsigned long *)&tmp->d_sb->s_anon.first);
 	spin_unlock(&tmp->d_lock);
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 
 	return tmp;
 
@@ -1592,18 +1596,18 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 	struct dentry *new = NULL;
 
 	if (inode && S_ISDIR(inode->i_mode)) {
-		spin_lock(&dcache_inode_lock);
+		spin_lock(&inode->i_lock);
 		new = __d_find_alias(inode, 1);
 		if (new) {
 			BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
-			spin_unlock(&dcache_inode_lock);
+			spin_unlock(&inode->i_lock);
 			security_d_instantiate(new, inode);
 			d_move(new, dentry);
 			iput(inode);
 		} else {
-			/* already got dcache_inode_lock, so d_add() by hand */
+			/* already taking inode->i_lock, so d_add() by hand */
 			__d_instantiate(dentry, inode);
-			spin_unlock(&dcache_inode_lock);
+			spin_unlock(&inode->i_lock);
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
 		}
@@ -1676,10 +1680,10 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	 * Negative dentry: instantiate it unless the inode is a directory and
 	 * already has a dentry.
 	 */
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	if (!S_ISDIR(inode->i_mode) || list_empty(&inode->i_dentry)) {
 		__d_instantiate(found, inode);
-		spin_unlock(&dcache_inode_lock);
+		spin_unlock(&inode->i_lock);
 		security_d_instantiate(found, inode);
 		return found;
 	}
@@ -1690,7 +1694,7 @@ struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
 	 */
 	new = list_entry(inode->i_dentry.next, struct dentry, d_alias);
 	__dget(new);
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 	security_d_instantiate(found, inode);
 	d_move(new, found);
 	iput(inode);
@@ -2004,15 +2008,17 @@ EXPORT_SYMBOL(d_validate);
  
 void d_delete(struct dentry * dentry)
 {
+	struct inode *inode;
 	int isdir = 0;
 	/*
 	 * Are we the only user?
 	 */
 again:
 	spin_lock(&dentry->d_lock);
-	isdir = S_ISDIR(dentry->d_inode->i_mode);
+	inode = dentry->d_inode;
+	isdir = S_ISDIR(inode->i_mode);
 	if (dentry->d_count == 1) {
-		if (!spin_trylock(&dcache_inode_lock)) {
+		if (inode && !spin_trylock(&inode->i_lock)) {
 			spin_unlock(&dentry->d_lock);
 			cpu_relax();
 			goto again;
@@ -2266,13 +2272,13 @@ struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
  * This helper attempts to cope with remotely renamed directories
  *
  * It assumes that the caller is already holding
- * dentry->d_parent->d_inode->i_mutex and the dcache_inode_lock
+ * dentry->d_parent->d_inode->i_mutex and the inode->i_lock
  *
  * Note: If ever the locking in lock_rename() changes, then please
  * remember to update this too...
  */
-static struct dentry *__d_unalias(struct dentry *dentry, struct dentry *alias)
-	__releases(dcache_inode_lock)
+static struct dentry *__d_unalias(struct inode *inode,
+		struct dentry *dentry, struct dentry *alias)
 {
 	struct mutex *m1 = NULL, *m2 = NULL;
 	struct dentry *ret;
@@ -2298,7 +2304,7 @@ out_unalias:
 	d_move(alias, dentry);
 	ret = alias;
 out_err:
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 	if (m2)
 		mutex_unlock(m2);
 	if (m1)
@@ -2371,7 +2377,7 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 		goto out_nolock;
 	}
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 
 	if (S_ISDIR(inode->i_mode)) {
 		struct dentry *alias;
@@ -2388,7 +2394,7 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 				goto found;
 			}
 			/* Nope, but we must(!) avoid directory aliasing */
-			actual = __d_unalias(dentry, alias);
+			actual = __d_unalias(inode, dentry, alias);
 			if (IS_ERR(actual))
 				dput(alias);
 			goto out_nolock;
@@ -2406,7 +2412,7 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 found:
 	_d_rehash(actual);
 	spin_unlock(&actual->d_lock);
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 out_nolock:
 	if (actual == dentry) {
 		security_d_instantiate(dentry, inode);
diff --git a/fs/exportfs/expfs.c b/fs/exportfs/expfs.c
index f06a940..4b68257 100644
--- a/fs/exportfs/expfs.c
+++ b/fs/exportfs/expfs.c
@@ -43,24 +43,26 @@ find_acceptable_alias(struct dentry *result,
 		void *context)
 {
 	struct dentry *dentry, *toput = NULL;
+	struct inode *inode;
 
 	if (acceptable(context, result))
 		return result;
 
-	spin_lock(&dcache_inode_lock);
-	list_for_each_entry(dentry, &result->d_inode->i_dentry, d_alias) {
+	inode = result->d_inode;
+	spin_lock(&inode->i_lock);
+	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
 		dget(dentry);
-		spin_unlock(&dcache_inode_lock);
+		spin_unlock(&inode->i_lock);
 		if (toput)
 			dput(toput);
 		if (dentry != result && acceptable(context, dentry)) {
 			dput(result);
 			return dentry;
 		}
-		spin_lock(&dcache_inode_lock);
+		spin_lock(&inode->i_lock);
 		toput = dentry;
 	}
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 
 	if (toput)
 		dput(toput);
diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c
index c3a5a11..5596c6a 100644
--- a/fs/nfs/getroot.c
+++ b/fs/nfs/getroot.c
@@ -63,11 +63,11 @@ static int nfs_superblock_set_dummy_root(struct super_block *sb, struct inode *i
 		 * This again causes shrink_dcache_for_umount_subtree() to
 		 * Oops, since the test for IS_ROOT() will fail.
 		 */
-		spin_lock(&dcache_inode_lock);
+		spin_lock(&sb->s_root->d_inode->i_lock);
 		spin_lock(&sb->s_root->d_lock);
 		list_del_init(&sb->s_root->d_alias);
 		spin_unlock(&sb->s_root->d_lock);
-		spin_unlock(&dcache_inode_lock);
+		spin_unlock(&sb->s_root->d_inode->i_lock);
 	}
 	return 0;
 }
diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index 9be6ec1..79b47cb 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -59,7 +59,7 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 	/* determine if the children should tell inode about their events */
 	watched = fsnotify_inode_watches_children(inode);
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	/* run all of the dentries associated with this inode.  Since this is a
 	 * directory, there damn well better only be one item on this list */
 	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
@@ -82,7 +82,7 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 		}
 		spin_unlock(&alias->d_lock);
 	}
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 }
 
 /* Notify this dentry's parent about a child's events. */
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index 0310b16..6d80ecc 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -175,7 +175,7 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 	struct list_head *p;
 	struct dentry *dentry = NULL;
 
-	spin_lock(&dcache_inode_lock);
+	spin_lock(&inode->i_lock);
 	list_for_each(p, &inode->i_dentry) {
 		dentry = list_entry(p, struct dentry, d_alias);
 
@@ -193,7 +193,7 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode,
 		dentry = NULL;
 	}
 
-	spin_unlock(&dcache_inode_lock);
+	spin_unlock(&inode->i_lock);
 
 	return dentry;
 }
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index 5f0392e..d719e4d 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -191,7 +191,6 @@ struct dentry_operations {
 #define DCACHE_OP_REVALIDATE	0x4000
 #define DCACHE_OP_DELETE	0x8000
 
-extern spinlock_t dcache_inode_lock;
 extern seqlock_t rename_lock;
 
 static inline int dname_external(struct dentry *dentry)
-- 
cgit v0.10.2


From 4b936885ab04dc6e0bb0ef35e0e23c1a7364d9e5 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:07 +1100
Subject: fs: improve scalability of pseudo filesystems

Regardless of how much we possibly try to scale dcache, there is likely
always going to be some fundamental contention when adding or removing children
under the same parent. Pseudo filesystems do not seem need to have connected
dentries because by definition they are disconnected.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/anon_inodes.c b/fs/anon_inodes.c
index aca8806..9d92b33 100644
--- a/fs/anon_inodes.c
+++ b/fs/anon_inodes.c
@@ -102,7 +102,7 @@ struct file *anon_inode_getfile(const char *name,
 	this.name = name;
 	this.len = strlen(name);
 	this.hash = 0;
-	path.dentry = d_alloc(anon_inode_mnt->mnt_sb->s_root, &this);
+	path.dentry = d_alloc_pseudo(anon_inode_mnt->mnt_sb, &this);
 	if (!path.dentry)
 		goto err_module;
 
diff --git a/fs/dcache.c b/fs/dcache.c
index 09ec945..9e6e6db 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1330,6 +1330,18 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 }
 EXPORT_SYMBOL(d_alloc);
 
+struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
+{
+	struct dentry *dentry = d_alloc(NULL, name);
+	if (dentry) {
+		dentry->d_sb = sb;
+		dentry->d_parent = dentry;
+		dentry->d_flags |= DCACHE_DISCONNECTED;
+	}
+	return dentry;
+}
+EXPORT_SYMBOL(d_alloc_pseudo);
+
 struct dentry *d_alloc_name(struct dentry *parent, const char *name)
 {
 	struct qstr q;
diff --git a/fs/pipe.c b/fs/pipe.c
index 01a7865..cfe3a7f 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -999,7 +999,7 @@ struct file *create_write_pipe(int flags)
 		goto err;
 
 	err = -ENOMEM;
-	path.dentry = d_alloc(pipe_mnt->mnt_sb->s_root, &name);
+	path.dentry = d_alloc_pseudo(pipe_mnt->mnt_sb, &name);
 	if (!path.dentry)
 		goto err_inode;
 	path.mnt = mntget(pipe_mnt);
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index d719e4d..c0a2ca9 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -211,6 +211,7 @@ extern void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op
 
 /* allocate/de-allocate */
 extern struct dentry * d_alloc(struct dentry *, const struct qstr *);
+extern struct dentry * d_alloc_pseudo(struct super_block *, const struct qstr *);
 extern struct dentry * d_splice_alias(struct inode *, struct dentry *);
 extern struct dentry * d_add_ci(struct dentry *, struct inode *, struct qstr *);
 extern struct dentry * d_obtain_alias(struct inode *);
diff --git a/net/socket.c b/net/socket.c
index 991e266..0ee74c3 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -361,7 +361,7 @@ static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 	if (unlikely(fd < 0))
 		return fd;
 
-	path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
+	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 	if (unlikely(!path.dentry)) {
 		put_unused_fd(fd);
 		return -ENOMEM;
-- 
cgit v0.10.2


From e1bb57826381199cc79fbf44e9dfeee58fc7b339 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:08 +1100
Subject: fs: prefetch inode data in dcache lookup

This makes single threaded git diff -1.25% +/- 0.05% elapsed time on my
2s12c24t Westmere system, and -0.86% +/- 0.05% on my 2s8c Barcelona, by
prefetching the important first cacheline of the inode in while we do the
actual name compare and other operations on the dentry.

There was no measurable slowdown in the single file stat case, or the creat
case (where negative dentries would be common).

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 9e6e6db..2a4ce7d 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1793,6 +1793,9 @@ seqretry:
 		tlen = dentry->d_name.len;
 		tname = dentry->d_name.name;
 		i = dentry->d_inode;
+		prefetch(tname);
+		if (i)
+			prefetch(i);
 		/*
 		 * This seqcount check is required to ensure name and
 		 * len are loaded atomically, so as not to walk off the
-- 
cgit v0.10.2


From 9d55c369bb5e695e629bc35cba2ef607755b3bee Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:09 +1100
Subject: fs: implement faster dentry memcmp

The standard memcmp function on a Westmere system shows up hot in
profiles in the `git diff` workload (both parallel and single threaded),
and it is likely due to the costs associated with trapping into
microcode, and little opportunity to improve memory access (dentry
name is not likely to take up more than a cacheline).

So replace it with an open-coded byte comparison. This increases code
size by 8 bytes in the critical __d_lookup_rcu function, but the
speedup is huge, averaging 10 runs of each:

git diff st   user   sys   elapsed  CPU
before        1.15   2.57  3.82      97.1
after         1.14   2.35  3.61      96.8

git diff mt   user   sys   elapsed  CPU
before        1.27   3.85  1.46     349
after         1.26   3.54  1.43     333

Elapsed time for single threaded git diff at 95.0% confidence:
        -0.21  +/- 0.01
        -5.45% +/- 0.24%

It's -0.66% +/- 0.06% elapsed time on my Opteron, so rep cmp costs on the
fam10h seem to be relatively smaller, but there is still a win.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/dcache.c b/fs/dcache.c
index 2a4ce7d..5699d4c 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1454,9 +1454,7 @@ static struct dentry *__d_instantiate_unique(struct dentry *entry,
 			continue;
 		if (alias->d_parent != entry->d_parent)
 			continue;
-		if (qstr->len != len)
-			continue;
-		if (memcmp(qstr->name, name, len))
+		if (dentry_cmp(qstr->name, qstr->len, name, len))
 			continue;
 		__dget(alias);
 		return alias;
@@ -1810,9 +1808,7 @@ seqretry:
 						tlen, tname, name))
 				continue;
 		} else {
-			if (tlen != len)
-				continue;
-			if (memcmp(tname, str, tlen))
+			if (dentry_cmp(tname, tlen, str, len))
 				continue;
 		}
 		/*
@@ -1925,9 +1921,7 @@ struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
 						tlen, tname, name))
 				goto next;
 		} else {
-			if (tlen != len)
-				goto next;
-			if (memcmp(tname, str, tlen))
+			if (dentry_cmp(tname, tlen, str, len))
 				goto next;
 		}
 
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index c0a2ca9..bd07758 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -47,6 +47,27 @@ struct dentry_stat_t {
 };
 extern struct dentry_stat_t dentry_stat;
 
+/*
+ * Compare 2 name strings, return 0 if they match, otherwise non-zero.
+ * The strings are both count bytes long, and count is non-zero.
+ */
+static inline int dentry_cmp(const unsigned char *cs, size_t scount,
+				const unsigned char *ct, size_t tcount)
+{
+	int ret;
+	if (scount != tcount)
+		return 1;
+	do {
+		ret = (*cs != *ct);
+		if (ret)
+			break;
+		cs++;
+		ct++;
+		tcount--;
+	} while (tcount);
+	return ret;
+}
+
 /* Name hashing routines. Initial hash value */
 /* Hash courtesy of the R5 hash in reiserfs modulo sign bits */
 #define init_name_hash()		0
-- 
cgit v0.10.2


From c6653a838b1b2738561aff0b8c0f62a9b714bdd9 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:10 +1100
Subject: fs: rename vfsmount counter helpers

Suggested by Andreas, mnt_ prefix is clearer namespace, follows kernel
conventions better, and is easier for tab complete. I introduced these
names so I'll admit they were not good choices.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/fs/namespace.c b/fs/namespace.c
index 39a7d50..03b8235 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -216,7 +216,7 @@ int __mnt_is_readonly(struct vfsmount *mnt)
 }
 EXPORT_SYMBOL_GPL(__mnt_is_readonly);
 
-static inline void inc_mnt_writers(struct vfsmount *mnt)
+static inline void mnt_inc_writers(struct vfsmount *mnt)
 {
 #ifdef CONFIG_SMP
 	(*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))++;
@@ -225,7 +225,7 @@ static inline void inc_mnt_writers(struct vfsmount *mnt)
 #endif
 }
 
-static inline void dec_mnt_writers(struct vfsmount *mnt)
+static inline void mnt_dec_writers(struct vfsmount *mnt)
 {
 #ifdef CONFIG_SMP
 	(*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))--;
@@ -234,7 +234,7 @@ static inline void dec_mnt_writers(struct vfsmount *mnt)
 #endif
 }
 
-static unsigned int count_mnt_writers(struct vfsmount *mnt)
+static unsigned int mnt_get_writers(struct vfsmount *mnt)
 {
 #ifdef CONFIG_SMP
 	unsigned int count = 0;
@@ -273,9 +273,9 @@ int mnt_want_write(struct vfsmount *mnt)
 	int ret = 0;
 
 	preempt_disable();
-	inc_mnt_writers(mnt);
+	mnt_inc_writers(mnt);
 	/*
-	 * The store to inc_mnt_writers must be visible before we pass
+	 * The store to mnt_inc_writers must be visible before we pass
 	 * MNT_WRITE_HOLD loop below, so that the slowpath can see our
 	 * incremented count after it has set MNT_WRITE_HOLD.
 	 */
@@ -289,7 +289,7 @@ int mnt_want_write(struct vfsmount *mnt)
 	 */
 	smp_rmb();
 	if (__mnt_is_readonly(mnt)) {
-		dec_mnt_writers(mnt);
+		mnt_dec_writers(mnt);
 		ret = -EROFS;
 		goto out;
 	}
@@ -317,7 +317,7 @@ int mnt_clone_write(struct vfsmount *mnt)
 	if (__mnt_is_readonly(mnt))
 		return -EROFS;
 	preempt_disable();
-	inc_mnt_writers(mnt);
+	mnt_inc_writers(mnt);
 	preempt_enable();
 	return 0;
 }
@@ -351,7 +351,7 @@ EXPORT_SYMBOL_GPL(mnt_want_write_file);
 void mnt_drop_write(struct vfsmount *mnt)
 {
 	preempt_disable();
-	dec_mnt_writers(mnt);
+	mnt_dec_writers(mnt);
 	preempt_enable();
 }
 EXPORT_SYMBOL_GPL(mnt_drop_write);
@@ -384,7 +384,7 @@ static int mnt_make_readonly(struct vfsmount *mnt)
 	 * MNT_WRITE_HOLD, so it can't be decremented by another CPU while
 	 * we're counting up here.
 	 */
-	if (count_mnt_writers(mnt) > 0)
+	if (mnt_get_writers(mnt) > 0)
 		ret = -EBUSY;
 	else
 		mnt->mnt_flags |= MNT_READONLY;
@@ -663,9 +663,9 @@ static inline void __mntput(struct vfsmount *mnt)
 	 */
 	/*
 	 * atomic_dec_and_lock() used to deal with ->mnt_count decrements
-	 * provides barriers, so count_mnt_writers() below is safe.  AV
+	 * provides barriers, so mnt_get_writers() below is safe.  AV
 	 */
-	WARN_ON(count_mnt_writers(mnt));
+	WARN_ON(mnt_get_writers(mnt));
 	fsnotify_vfsmount_delete(mnt);
 	dput(mnt->mnt_root);
 	free_vfsmnt(mnt);
-- 
cgit v0.10.2


From b3e19d924b6eaf2ca7d22cba99a517c5171007b6 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@kernel.dk>
Date: Fri, 7 Jan 2011 17:50:11 +1100
Subject: fs: scale mntget/mntput

The problem that this patch aims to fix is vfsmount refcounting scalability.
We need to take a reference on the vfsmount for every successful path lookup,
which often go to the same mount point.

The fundamental difficulty is that a "simple" reference count can never be made
scalable, because any time a reference is dropped, we must check whether that
was the last reference. To do that requires communication with all other CPUs
that may have taken a reference count.

We can make refcounts more scalable in a couple of ways, involving keeping
distributed counters, and checking for the global-zero condition less
frequently.

- check the global sum once every interval (this will delay zero detection
  for some interval, so it's probably a showstopper for vfsmounts).

- keep a local count and only taking the global sum when local reaches 0 (this
  is difficult for vfsmounts, because we can't hold preempt off for the life of
  a reference, so a counter would need to be per-thread or tied strongly to a
  particular CPU which requires more locking).

- keep a local difference of increments and decrements, which allows us to sum
  the total difference and hence find the refcount when summing all CPUs. Then,
  keep a single integer "long" refcount for slow and long lasting references,
  and only take the global sum of local counters when the long refcount is 0.

This last scheme is what I implemented here. Attached mounts and process root
and working directory references are "long" references, and everything else is
a short reference.

This allows scalable vfsmount references during path walking over mounted
subtrees and unattached (lazy umounted) mounts with processes still running
in them.

This results in one fewer atomic op in the fastpath: mntget is now just a
per-CPU inc, rather than an atomic inc; and mntput just requires a spinlock
and non-atomic decrement in the common case. However code is otherwise bigger
and heavier, so single threaded performance is basically a wash.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index 5a24f40..f099b82 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -1542,7 +1542,7 @@ pfm_exit_smpl_buffer(pfm_buffer_fmt_t *fmt)
  * any operations on the root directory. However, we need a non-trivial
  * d_name - pfm: will go nicely and kill the special-casing in procfs.
  */
-static struct vfsmount *pfmfs_mnt;
+static struct vfsmount *pfmfs_mnt __read_mostly;
 
 static int __init
 init_pfm_fs(void)
diff --git a/drivers/mtd/mtdchar.c b/drivers/mtd/mtdchar.c
index 4759d82..f511dd1 100644
--- a/drivers/mtd/mtdchar.c
+++ b/drivers/mtd/mtdchar.c
@@ -1201,7 +1201,7 @@ err_unregister_chdev:
 static void __exit cleanup_mtdchar(void)
 {
 	unregister_mtd_user(&mtdchar_notifier);
-	mntput(mtd_inode_mnt);
+	mntput_long(mtd_inode_mnt);
 	unregister_filesystem(&mtd_inodefs_type);
 	__unregister_chrdev(MTD_CHAR_MAJOR, 0, 1 << MINORBITS, "mtd");
 }
diff --git a/fs/anon_inodes.c b/fs/anon_inodes.c
index 9d92b33..5fd38112a 100644
--- a/fs/anon_inodes.c
+++ b/fs/anon_inodes.c
@@ -232,7 +232,7 @@ static int __init anon_inode_init(void)
 	return 0;
 
 err_mntput:
-	mntput(anon_inode_mnt);
+	mntput_long(anon_inode_mnt);
 err_unregister_filesystem:
 	unregister_filesystem(&anon_inode_fs_type);
 err_exit:
diff --git a/fs/fs_struct.c b/fs/fs_struct.c
index 60b8531..68ca487 100644
--- a/fs/fs_struct.c
+++ b/fs/fs_struct.c
@@ -17,11 +17,11 @@ void set_fs_root(struct fs_struct *fs, struct path *path)
 	write_seqcount_begin(&fs->seq);
 	old_root = fs->root;
 	fs->root = *path;
-	path_get(path);
+	path_get_long(path);
 	write_seqcount_end(&fs->seq);
 	spin_unlock(&fs->lock);
 	if (old_root.dentry)
-		path_put(&old_root);
+		path_put_long(&old_root);
 }
 
 /*
@@ -36,12 +36,12 @@ void set_fs_pwd(struct fs_struct *fs, struct path *path)
 	write_seqcount_begin(&fs->seq);
 	old_pwd = fs->pwd;
 	fs->pwd = *path;
-	path_get(path);
+	path_get_long(path);
 	write_seqcount_end(&fs->seq);
 	spin_unlock(&fs->lock);
 
 	if (old_pwd.dentry)
-		path_put(&old_pwd);
+		path_put_long(&old_pwd);
 }
 
 void chroot_fs_refs(struct path *old_root, struct path *new_root)
@@ -59,13 +59,13 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root)
 			write_seqcount_begin(&fs->seq);
 			if (fs->root.dentry == old_root->dentry
 			    && fs->root.mnt == old_root->mnt) {
-				path_get(new_root);
+				path_get_long(new_root);
 				fs->root = *new_root;
 				count++;
 			}
 			if (fs->pwd.dentry == old_root->dentry
 			    && fs->pwd.mnt == old_root->mnt) {
-				path_get(new_root);
+				path_get_long(new_root);
 				fs->pwd = *new_root;
 				count++;
 			}
@@ -76,13 +76,13 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root)
 	} while_each_thread(g, p);
 	read_unlock(&tasklist_lock);
 	while (count--)
-		path_put(old_root);
+		path_put_long(old_root);
 }
 
 void free_fs_struct(struct fs_struct *fs)
 {
-	path_put(&fs->root);
-	path_put(&fs->pwd);
+	path_put_long(&fs->root);
+	path_put_long(&fs->pwd);
 	kmem_cache_free(fs_cachep, fs);
 }
 
@@ -115,7 +115,13 @@ struct fs_struct *copy_fs_struct(struct fs_struct *old)
 		spin_lock_init(&fs->lock);
 		seqcount_init(&fs->seq);
 		fs->umask = old->umask;
-		get_fs_root_and_pwd(old, &fs->root, &fs->pwd);
+
+		spin_lock(&old->lock);
+		fs->root = old->root;
+		path_get_long(&fs->root);
+		fs->pwd = old->pwd;
+		path_get_long(&fs->pwd);
+		spin_unlock(&old->lock);
 	}
 	return fs;
 }
diff --git a/fs/internal.h b/fs/internal.h
index e43b9a4..9687c2e 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -63,6 +63,7 @@ extern int copy_mount_string(const void __user *, char **);
 
 extern void free_vfsmnt(struct vfsmount *);
 extern struct vfsmount *alloc_vfsmnt(const char *);
+extern unsigned int mnt_get_count(struct vfsmount *mnt);
 extern struct vfsmount *__lookup_mnt(struct vfsmount *, struct dentry *, int);
 extern void mnt_set_mountpoint(struct vfsmount *, struct dentry *,
 				struct vfsmount *);
diff --git a/fs/namei.c b/fs/namei.c
index 4e957bf..19433cd 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -368,6 +368,18 @@ void path_get(struct path *path)
 EXPORT_SYMBOL(path_get);
 
 /**
+ * path_get_long - get a long reference to a path
+ * @path: path to get the reference to
+ *
+ * Given a path increment the reference count to the dentry and the vfsmount.
+ */
+void path_get_long(struct path *path)
+{
+	mntget_long(path->mnt);
+	dget(path->dentry);
+}
+
+/**
  * path_put - put a reference to a path
  * @path: path to put the reference to
  *
@@ -381,6 +393,18 @@ void path_put(struct path *path)
 EXPORT_SYMBOL(path_put);
 
 /**
+ * path_put_long - put a long reference to a path
+ * @path: path to put the reference to
+ *
+ * Given a path decrement the reference count to the dentry and the vfsmount.
+ */
+void path_put_long(struct path *path)
+{
+	dput(path->dentry);
+	mntput_long(path->mnt);
+}
+
+/**
  * nameidata_drop_rcu - drop this nameidata out of rcu-walk
  * @nd: nameidata pathwalk data to drop
  * @Returns: 0 on success, -ECHLID on failure
diff --git a/fs/namespace.c b/fs/namespace.c
index 03b8235..3ddfd90 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -138,6 +138,64 @@ void mnt_release_group_id(struct vfsmount *mnt)
 	mnt->mnt_group_id = 0;
 }
 
+/*
+ * vfsmount lock must be held for read
+ */
+static inline void mnt_add_count(struct vfsmount *mnt, int n)
+{
+#ifdef CONFIG_SMP
+	this_cpu_add(mnt->mnt_pcp->mnt_count, n);
+#else
+	preempt_disable();
+	mnt->mnt_count += n;
+	preempt_enable();
+#endif
+}
+
+static inline void mnt_set_count(struct vfsmount *mnt, int n)
+{
+#ifdef CONFIG_SMP
+	this_cpu_write(mnt->mnt_pcp->mnt_count, n);
+#else
+	mnt->mnt_count = n;
+#endif
+}
+
+/*
+ * vfsmount lock must be held for read
+ */
+static inline void mnt_inc_count(struct vfsmount *mnt)
+{
+	mnt_add_count(mnt, 1);
+}
+
+/*
+ * vfsmount lock must be held for read
+ */
+static inline void mnt_dec_count(struct vfsmount *mnt)
+{
+	mnt_add_count(mnt, -1);
+}
+
+/*
+ * vfsmount lock must be held for write
+ */
+unsigned int mnt_get_count(struct vfsmount *mnt)
+{
+#ifdef CONFIG_SMP
+	unsigned int count = atomic_read(&mnt->mnt_longrefs);
+	int cpu;
+
+	for_each_possible_cpu(cpu) {
+		count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count;
+	}
+
+	return count;
+#else
+	return mnt->mnt_count;
+#endif
+}
+
 struct vfsmount *alloc_vfsmnt(const char *name)
 {
 	struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL);
@@ -154,7 +212,17 @@ struct vfsmount *alloc_vfsmnt(const char *name)
 				goto out_free_id;
 		}
 
-		atomic_set(&mnt->mnt_count, 1);
+#ifdef CONFIG_SMP
+		mnt->mnt_pcp = alloc_percpu(struct mnt_pcp);
+		if (!mnt->mnt_pcp)
+			goto out_free_devname;
+
+		atomic_set(&mnt->mnt_longrefs, 1);
+#else
+		mnt->mnt_count = 1;
+		mnt->mnt_writers = 0;
+#endif
+
 		INIT_LIST_HEAD(&mnt->mnt_hash);
 		INIT_LIST_HEAD(&mnt->mnt_child);
 		INIT_LIST_HEAD(&mnt->mnt_mounts);
@@ -166,13 +234,6 @@ struct vfsmount *alloc_vfsmnt(const char *name)
 #ifdef CONFIG_FSNOTIFY
 		INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks);
 #endif
-#ifdef CONFIG_SMP
-		mnt->mnt_writers = alloc_percpu(int);
-		if (!mnt->mnt_writers)
-			goto out_free_devname;
-#else
-		mnt->mnt_writers = 0;
-#endif
 	}
 	return mnt;
 
@@ -219,7 +280,7 @@ EXPORT_SYMBOL_GPL(__mnt_is_readonly);
 static inline void mnt_inc_writers(struct vfsmount *mnt)
 {
 #ifdef CONFIG_SMP
-	(*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))++;
+	this_cpu_inc(mnt->mnt_pcp->mnt_writers);
 #else
 	mnt->mnt_writers++;
 #endif
@@ -228,7 +289,7 @@ static inline void mnt_inc_writers(struct vfsmount *mnt)
 static inline void mnt_dec_writers(struct vfsmount *mnt)
 {
 #ifdef CONFIG_SMP
-	(*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))--;
+	this_cpu_dec(mnt->mnt_pcp->mnt_writers);
 #else
 	mnt->mnt_writers--;
 #endif
@@ -241,7 +302,7 @@ static unsigned int mnt_get_writers(struct vfsmount *mnt)
 	int cpu;
 
 	for_each_possible_cpu(cpu) {
-		count += *per_cpu_ptr(mnt->mnt_writers, cpu);
+		count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers;
 	}
 
 	return count;
@@ -418,7 +479,7 @@ void free_vfsmnt(struct vfsmount *mnt)
 	kfree(mnt->mnt_devname);
 	mnt_free_id(mnt);
 #ifdef CONFIG_SMP
-	free_percpu(mnt->mnt_writers);
+	free_percpu(mnt->mnt_pcp);
 #endif
 	kmem_cache_free(mnt_cache, mnt);
 }
@@ -652,9 +713,10 @@ static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root,
 	return NULL;
 }
 
-static inline void __mntput(struct vfsmount *mnt)
+static inline void mntfree(struct vfsmount *mnt)
 {
 	struct super_block *sb = mnt->mnt_sb;
+
 	/*
 	 * This probably indicates that somebody messed
 	 * up a mnt_want/drop_write() pair.  If this
@@ -662,8 +724,8 @@ static inline void __mntput(struct vfsmount *mnt)
 	 * to make r/w->r/o transitions.
 	 */
 	/*
-	 * atomic_dec_and_lock() used to deal with ->mnt_count decrements
-	 * provides barriers, so mnt_get_writers() below is safe.  AV
+	 * The locking used to deal with mnt_count decrement provides barriers,
+	 * so mnt_get_writers() below is safe.
 	 */
 	WARN_ON(mnt_get_writers(mnt));
 	fsnotify_vfsmount_delete(mnt);
@@ -672,28 +734,113 @@ static inline void __mntput(struct vfsmount *mnt)
 	deactivate_super(sb);
 }
 
-void mntput_no_expire(struct vfsmount *mnt)
-{
-repeat:
-	if (atomic_add_unless(&mnt->mnt_count, -1, 1))
-		return;
+#ifdef CONFIG_SMP
+static inline void __mntput(struct vfsmount *mnt, int longrefs)
+{
+	if (!longrefs) {
+put_again:
+		br_read_lock(vfsmount_lock);
+		if (likely(atomic_read(&mnt->mnt_longrefs))) {
+			mnt_dec_count(mnt);
+			br_read_unlock(vfsmount_lock);
+			return;
+		}
+		br_read_unlock(vfsmount_lock);
+	} else {
+		BUG_ON(!atomic_read(&mnt->mnt_longrefs));
+		if (atomic_add_unless(&mnt->mnt_longrefs, -1, 1))
+			return;
+	}
+
 	br_write_lock(vfsmount_lock);
-	if (!atomic_dec_and_test(&mnt->mnt_count)) {
+	if (!longrefs)
+		mnt_dec_count(mnt);
+	else
+		atomic_dec(&mnt->mnt_longrefs);
+	if (mnt_get_count(mnt)) {
 		br_write_unlock(vfsmount_lock);
 		return;
 	}
-	if (likely(!mnt->mnt_pinned)) {
+	if (unlikely(mnt->mnt_pinned)) {
+		mnt_add_count(mnt, mnt->mnt_pinned + 1);
+		mnt->mnt_pinned = 0;
 		br_write_unlock(vfsmount_lock);
-		__mntput(mnt);
+		acct_auto_close_mnt(mnt);
+		goto put_again;
+	}
+	br_write_unlock(vfsmount_lock);
+	mntfree(mnt);
+}
+#else
+static inline void __mntput(struct vfsmount *mnt, int longrefs)
+{
+put_again:
+	mnt_dec_count(mnt);
+	if (likely(mnt_get_count(mnt)))
 		return;
+	br_write_lock(vfsmount_lock);
+	if (unlikely(mnt->mnt_pinned)) {
+		mnt_add_count(mnt, mnt->mnt_pinned + 1);
+		mnt->mnt_pinned = 0;
+		br_write_unlock(vfsmount_lock);
+		acct_auto_close_mnt(mnt);
+		goto put_again;
 	}
-	atomic_add(mnt->mnt_pinned + 1, &mnt->mnt_count);
-	mnt->mnt_pinned = 0;
 	br_write_unlock(vfsmount_lock);
-	acct_auto_close_mnt(mnt);
-	goto repeat;
+	mntfree(mnt);
+}
+#endif
+
+static void mntput_no_expire(struct vfsmount *mnt)
+{
+	__mntput(mnt, 0);
+}
+
+void mntput(struct vfsmount *mnt)
+{
+	if (mnt) {
+		/* avoid cacheline pingpong, hope gcc doesn't get "smart" */
+		if (unlikely(mnt->mnt_expiry_mark))
+			mnt->mnt_expiry_mark = 0;
+		__mntput(mnt, 0);
+	}
+}
+EXPORT_SYMBOL(mntput);
+
+struct vfsmount *mntget(struct vfsmount *mnt)
+{
+	if (mnt)
+		mnt_inc_count(mnt);
+	return mnt;
+}
+EXPORT_SYMBOL(mntget);
+
+void mntput_long(struct vfsmount *mnt)
+{
+#ifdef CONFIG_SMP
+	if (mnt) {
+		/* avoid cacheline pingpong, hope gcc doesn't get "smart" */
+		if (unlikely(mnt->mnt_expiry_mark))
+			mnt->mnt_expiry_mark = 0;
+		__mntput(mnt, 1);
+	}
+#else
+	mntput(mnt);
+#endif
 }
-EXPORT_SYMBOL(mntput_no_expire);
+EXPORT_SYMBOL(mntput_long);
+
+struct vfsmount *mntget_long(struct vfsmount *mnt)
+{
+#ifdef CONFIG_SMP
+	if (mnt)
+		atomic_inc(&mnt->mnt_longrefs);
+	return mnt;
+#else
+	return mntget(mnt);
+#endif
+}
+EXPORT_SYMBOL(mntget_long);
 
 void mnt_pin(struct vfsmount *mnt)
 {
@@ -701,19 +848,17 @@ void mnt_pin(struct vfsmount *mnt)
 	mnt->mnt_pinned++;
 	br_write_unlock(vfsmount_lock);
 }
-
 EXPORT_SYMBOL(mnt_pin);
 
 void mnt_unpin(struct vfsmount *mnt)
 {
 	br_write_lock(vfsmount_lock);
 	if (mnt->mnt_pinned) {
-		atomic_inc(&mnt->mnt_count);
+		mnt_inc_count(mnt);
 		mnt->mnt_pinned--;
 	}
 	br_write_unlock(vfsmount_lock);
 }
-
 EXPORT_SYMBOL(mnt_unpin);
 
 static inline void mangle(struct seq_file *m, const char *s)
@@ -1008,12 +1153,13 @@ int may_umount_tree(struct vfsmount *mnt)
 	int minimum_refs = 0;
 	struct vfsmount *p;
 
-	br_read_lock(vfsmount_lock);
+	/* write lock needed for mnt_get_count */
+	br_write_lock(vfsmount_lock);
 	for (p = mnt; p; p = next_mnt(p, mnt)) {
-		actual_refs += atomic_read(&p->mnt_count);
+		actual_refs += mnt_get_count(p);
 		minimum_refs += 2;
 	}
-	br_read_unlock(vfsmount_lock);
+	br_write_unlock(vfsmount_lock);
 
 	if (actual_refs > minimum_refs)
 		return 0;
@@ -1040,10 +1186,10 @@ int may_umount(struct vfsmount *mnt)
 {
 	int ret = 1;
 	down_read(&namespace_sem);
-	br_read_lock(vfsmount_lock);
+	br_write_lock(vfsmount_lock);
 	if (propagate_mount_busy(mnt, 2))
 		ret = 0;
-	br_read_unlock(vfsmount_lock);
+	br_write_unlock(vfsmount_lock);
 	up_read(&namespace_sem);
 	return ret;
 }
@@ -1070,7 +1216,7 @@ void release_mounts(struct list_head *head)
 			dput(dentry);
 			mntput(m);
 		}
-		mntput(mnt);
+		mntput_long(mnt);
 	}
 }
 
@@ -1125,8 +1271,16 @@ static int do_umount(struct vfsmount *mnt, int flags)
 		    flags & (MNT_FORCE | MNT_DETACH))
 			return -EINVAL;
 
-		if (atomic_read(&mnt->mnt_count) != 2)
+		/*
+		 * probably don't strictly need the lock here if we examined
+		 * all race cases, but it's a slowpath.
+		 */
+		br_write_lock(vfsmount_lock);
+		if (mnt_get_count(mnt) != 2) {
+			br_write_lock(vfsmount_lock);
 			return -EBUSY;
+		}
+		br_write_unlock(vfsmount_lock);
 
 		if (!xchg(&mnt->mnt_expiry_mark, 1))
 			return -EAGAIN;
@@ -1815,7 +1969,7 @@ int do_add_mount(struct vfsmount *newmnt, struct path *path,
 
 unlock:
 	up_write(&namespace_sem);
-	mntput(newmnt);
+	mntput_long(newmnt);
 	return err;
 }
 
@@ -2148,11 +2302,11 @@ static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
 		if (fs) {
 			if (p == fs->root.mnt) {
 				rootmnt = p;
-				fs->root.mnt = mntget(q);
+				fs->root.mnt = mntget_long(q);
 			}
 			if (p == fs->pwd.mnt) {
 				pwdmnt = p;
-				fs->pwd.mnt = mntget(q);
+				fs->pwd.mnt = mntget_long(q);
 			}
 		}
 		p = next_mnt(p, mnt_ns->root);
@@ -2161,9 +2315,9 @@ static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
 	up_write(&namespace_sem);
 
 	if (rootmnt)
-		mntput(rootmnt);
+		mntput_long(rootmnt);
 	if (pwdmnt)
-		mntput(pwdmnt);
+		mntput_long(pwdmnt);
 
 	return new_ns;
 }
@@ -2350,6 +2504,7 @@ SYSCALL_DEFINE2(pivot_root, const char __user *, new_root,
 	touch_mnt_namespace(current->nsproxy->mnt_ns);
 	br_write_unlock(vfsmount_lock);
 	chroot_fs_refs(&root, &new);
+
 	error = 0;
 	path_put(&root_parent);
 	path_put(&parent_path);
@@ -2376,6 +2531,7 @@ static void __init init_mount_tree(void)
 	mnt = do_kern_mount("rootfs", 0, "rootfs", NULL);
 	if (IS_ERR(mnt))
 		panic("Can't create rootfs");
+
 	ns = create_mnt_ns(mnt);
 	if (IS_ERR(ns))
 		panic("Can't allocate initial namespace");
diff --git a/fs/pipe.c b/fs/pipe.c
index cfe3a7f..68f1f8e 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -1292,7 +1292,7 @@ static int __init init_pipe_fs(void)
 static void __exit exit_pipe_fs(void)
 {
 	unregister_filesystem(&pipe_fs_type);
-	mntput(pipe_mnt);
+	mntput_long(pipe_mnt);
 }
 
 fs_initcall(init_pipe_fs);
diff --git a/fs/pnode.c b/fs/pnode.c
index 8066b8d..d42514e 100644
--- a/fs/pnode.c
+++ b/fs/pnode.c
@@ -288,7 +288,7 @@ out:
  */
 static inline int do_refcount_check(struct vfsmount *mnt, int count)
 {
-	int mycount = atomic_read(&mnt->mnt_count) - mnt->mnt_ghosts;
+	int mycount = mnt_get_count(mnt) - mnt->mnt_ghosts;
 	return (mycount > count);
 }
 
@@ -300,7 +300,7 @@ static inline int do_refcount_check(struct vfsmount *mnt, int count)
  * Check if any of these mounts that **do not have submounts**
  * have more references than 'refcnt'. If so return busy.
  *
- * vfsmount lock must be held for read or write
+ * vfsmount lock must be held for write
  */
 int propagate_mount_busy(struct vfsmount *mnt, int refcnt)
 {
diff --git a/fs/super.c b/fs/super.c
index 968ba01..823e061 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -1140,7 +1140,7 @@ static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
 	return mnt;
 
  err:
-	mntput(mnt);
+	mntput_long(mnt);
 	return ERR_PTR(err);
 }
 
diff --git a/include/linux/mount.h b/include/linux/mount.h
index 5e7a594..1869ea2 100644
--- a/include/linux/mount.h
+++ b/include/linux/mount.h
@@ -13,6 +13,7 @@
 #include <linux/list.h>
 #include <linux/nodemask.h>
 #include <linux/spinlock.h>
+#include <linux/seqlock.h>
 #include <asm/atomic.h>
 
 struct super_block;
@@ -46,12 +47,24 @@ struct mnt_namespace;
 
 #define MNT_INTERNAL	0x4000
 
+struct mnt_pcp {
+	int mnt_count;
+	int mnt_writers;
+};
+
 struct vfsmount {
 	struct list_head mnt_hash;
 	struct vfsmount *mnt_parent;	/* fs we are mounted on */
 	struct dentry *mnt_mountpoint;	/* dentry of mountpoint */
 	struct dentry *mnt_root;	/* root of the mounted tree */
 	struct super_block *mnt_sb;	/* pointer to superblock */
+#ifdef CONFIG_SMP
+	struct mnt_pcp __percpu *mnt_pcp;
+	atomic_t mnt_longrefs;
+#else
+	int mnt_count;
+	int mnt_writers;
+#endif
 	struct list_head mnt_mounts;	/* list of children, anchored here */
 	struct list_head mnt_child;	/* and going through their mnt_child */
 	int mnt_flags;
@@ -70,57 +83,25 @@ struct vfsmount {
 	struct mnt_namespace *mnt_ns;	/* containing namespace */
 	int mnt_id;			/* mount identifier */
 	int mnt_group_id;		/* peer group identifier */
-	/*
-	 * We put mnt_count & mnt_expiry_mark at the end of struct vfsmount
-	 * to let these frequently modified fields in a separate cache line
-	 * (so that reads of mnt_flags wont ping-pong on SMP machines)
-	 */
-	atomic_t mnt_count;
 	int mnt_expiry_mark;		/* true if marked for expiry */
 	int mnt_pinned;
 	int mnt_ghosts;
-#ifdef CONFIG_SMP
-	int __percpu *mnt_writers;
-#else
-	int mnt_writers;
-#endif
 };
 
-static inline int *get_mnt_writers_ptr(struct vfsmount *mnt)
-{
-#ifdef CONFIG_SMP
-	return mnt->mnt_writers;
-#else
-	return &mnt->mnt_writers;
-#endif
-}
-
-static inline struct vfsmount *mntget(struct vfsmount *mnt)
-{
-	if (mnt)
-		atomic_inc(&mnt->mnt_count);
-	return mnt;
-}
-
 struct file; /* forward dec */
 
 extern int mnt_want_write(struct vfsmount *mnt);
 extern int mnt_want_write_file(struct file *file);
 extern int mnt_clone_write(struct vfsmount *mnt);
 extern void mnt_drop_write(struct vfsmount *mnt);
-extern void mntput_no_expire(struct vfsmount *mnt);
+extern void mntput(struct vfsmount *mnt);
+extern struct vfsmount *mntget(struct vfsmount *mnt);
+extern void mntput_long(struct vfsmount *mnt);
+extern struct vfsmount *mntget_long(struct vfsmount *mnt);
 extern void mnt_pin(struct vfsmount *mnt);
 extern void mnt_unpin(struct vfsmount *mnt);
 extern int __mnt_is_readonly(struct vfsmount *mnt);
 
-static inline void mntput(struct vfsmount *mnt)
-{
-	if (mnt) {
-		mnt->mnt_expiry_mark = 0;
-		mntput_no_expire(mnt);
-	}
-}
-
 extern struct vfsmount *do_kern_mount(const char *fstype, int flags,
 				      const char *name, void *data);
 
diff --git a/include/linux/path.h b/include/linux/path.h
index edc98de..a581e8c 100644
--- a/include/linux/path.h
+++ b/include/linux/path.h
@@ -10,7 +10,9 @@ struct path {
 };
 
 extern void path_get(struct path *);
+extern void path_get_long(struct path *);
 extern void path_put(struct path *);
+extern void path_put_long(struct path *);
 
 static inline int path_equal(const struct path *path1, const struct path *path2)
 {
diff --git a/net/socket.c b/net/socket.c
index 0ee74c3..815bba3 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -2390,6 +2390,8 @@ EXPORT_SYMBOL(sock_unregister);
 
 static int __init sock_init(void)
 {
+	int err;
+
 	/*
 	 *      Initialize sock SLAB cache.
 	 */
@@ -2406,8 +2408,15 @@ static int __init sock_init(void)
 	 */
 
 	init_inodecache();
-	register_filesystem(&sock_fs_type);
+
+	err = register_filesystem(&sock_fs_type);
+	if (err)
+		goto out_fs;
 	sock_mnt = kern_mount(&sock_fs_type);
+	if (IS_ERR(sock_mnt)) {
+		err = PTR_ERR(sock_mnt);
+		goto out_mount;
+	}
 
 	/* The real protocol initialization is performed in later initcalls.
 	 */
@@ -2420,7 +2429,13 @@ static int __init sock_init(void)
 	skb_timestamping_init();
 #endif
 
-	return 0;
+out:
+	return err;
+
+out_mount:
+	unregister_filesystem(&sock_fs_type);
+out_fs:
+	goto out;
 }
 
 core_initcall(sock_init);	/* early initcall */
-- 
cgit v0.10.2