From 5d44670facd3205212f8fe89eb422e3b5f309612 Mon Sep 17 00:00:00 2001 From: Marcus Meissner Date: Thu, 5 May 2011 10:44:11 -0700 Subject: ocfs2: Initialize data_ac (might be used uninitialized) CLANG found that there is a path that has data_ac uninitialized, this place 2917 /* This gets us the dx_root */ 2918 ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &meta_ac); 2919 if (ret) { 3 Taking true branch 2920 mlog_errno(ret); 2921 goto out; 4 Control jumps to line 3168 2922 } Goes to the out: label without data_ac being initialized. Ciao, Marcus Signed-Off-By: Marcus Meissner Signed-off-by: Mark Fasheh Signed-off-by: Joel Becker diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c index 9fe5b8fd..8582e3f 100644 --- a/fs/ocfs2/dir.c +++ b/fs/ocfs2/dir.c @@ -2868,7 +2868,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, bytes = blocks_wanted << sb->s_blocksize_bits; struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); struct ocfs2_inode_info *oi = OCFS2_I(dir); - struct ocfs2_alloc_context *data_ac; + struct ocfs2_alloc_context *data_ac = NULL; struct ocfs2_alloc_context *meta_ac = NULL; struct buffer_head *dirdata_bh = NULL; struct buffer_head *dx_root_bh = NULL; -- cgit v0.10.2 From 9a790ba1ec02bbae0933e7ebd576c0bc329e9796 Mon Sep 17 00:00:00 2001 From: Tristan Ye Date: Thu, 12 May 2011 20:47:07 +0800 Subject: ocfs2: skip existing hole when removing the last extent_rec in punching-hole codes. In the case of removing a partial extent record which covers a hole, current punching-hole logic will try to remove more than the length of whole extent record, which leads to the failure of following assert(fs/ocfs2/alloc.c): 5507 BUG_ON(cpos < le32_to_cpu(rec->e_cpos) || trunc_range > rec_range); This patch tries to skip existing hole at the last attempt of removing a partial extent record, what's more, it also adds some necessary comments for better understanding of punching-hole codes. Signed-off-by: Tristan Ye Signed-off-by: Joel Becker diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 41565ae..89659d6 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -1607,6 +1607,9 @@ static void ocfs2_calc_trunc_pos(struct inode *inode, range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec); if (le32_to_cpu(rec->e_cpos) >= trunc_start) { + /* + * remove an entire extent record. + */ *trunc_cpos = le32_to_cpu(rec->e_cpos); /* * Skip holes if any. @@ -1617,7 +1620,16 @@ static void ocfs2_calc_trunc_pos(struct inode *inode, *blkno = le64_to_cpu(rec->e_blkno); *trunc_end = le32_to_cpu(rec->e_cpos); } else if (range > trunc_start) { + /* + * remove a partial extent record, which means we're + * removing the last extent record. + */ *trunc_cpos = trunc_start; + /* + * skip hole if any. + */ + if (range < *trunc_end) + *trunc_end = range; *trunc_len = *trunc_end - trunc_start; coff = trunc_start - le32_to_cpu(rec->e_cpos); *blkno = le64_to_cpu(rec->e_blkno) + -- cgit v0.10.2 From 4da6dc293604f55d156148b8f60b94053e3195fc Mon Sep 17 00:00:00 2001 From: Sunil Mushran Date: Wed, 4 May 2011 10:27:10 -0700 Subject: ocfs2/dlm: Use negotiated o2dlm protocol version Patch fixes a bug in the o2dlm protocol negotiation in that it is using the builtin version rather than the negotiated version during the domain join. This causes join errors when a node having kernel >= 2.6.37 joins a cluster with nodes having kernels < 2.6.37. This only affects the o2cb cluster stack. Signed-off-by: Sunil Mushran Reported-by: Jacek Stepniewski Acked-by: Mark Fasheh Signed-off-by: Joel Becker diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c index 7540a49..3b179d6 100644 --- a/fs/ocfs2/dlm/dlmdomain.c +++ b/fs/ocfs2/dlm/dlmdomain.c @@ -1614,7 +1614,8 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm) spin_unlock(&dlm->spinlock); /* Support for global heartbeat and node info was added in 1.1 */ - if (dlm_protocol.pv_major > 1 || dlm_protocol.pv_minor > 0) { + if (dlm->dlm_locking_proto.pv_major > 1 || + dlm->dlm_locking_proto.pv_minor > 0) { status = dlm_send_nodeinfo(dlm, ctxt->yes_resp_map); if (status) { mlog_errno(status); -- cgit v0.10.2 From 76d9fc2954d057b19bf5d7b854df2b621b00fdec Mon Sep 17 00:00:00 2001 From: Sunil Mushran Date: Wed, 4 May 2011 10:28:00 -0700 Subject: ocfs2/cluster: Increase the live threshold for global heartbeat We have seen isolated cases (very few, I might add) of o2hb not detecting all live nodes on startup. One plausible reasoning for it is that other node had a hb io delay at the same time. The live threshold set at 2 (as low as it can be) could be increased to ameliorate the situation. But increasing the threshold directly affects mount time. Currently it takes around 5 secs to mount a volume in o2cb cluster with local heartbeat. Increasing the threshold will make mounts even slower. As the issue itself is rare, we have left things as they are for the local heartbeat mode. However we can improve the situation for global heartbeat mode as in that mode, we start the heartbeat much before the mount. This patch doubles the live threshold for the start of the first region in global heartbeat mode. Addresses internal Oracle bug#10635585. Signed-off-by: Sunil Mushran Acked-by: Mark Fasheh Signed-off-by: Joel Becker diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c index 6437202..1d28505 100644 --- a/fs/ocfs2/cluster/heartbeat.c +++ b/fs/ocfs2/cluster/heartbeat.c @@ -1690,6 +1690,7 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, struct file *filp = NULL; struct inode *inode = NULL; ssize_t ret = -EINVAL; + int live_threshold; if (reg->hr_bdev) goto out; @@ -1766,8 +1767,18 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, * A node is considered live after it has beat LIVE_THRESHOLD * times. We're not steady until we've given them a chance * _after_ our first read. + * The default threshold is bare minimum so as to limit the delay + * during mounts. For global heartbeat, the threshold doubled for the + * first region. */ - atomic_set(®->hr_steady_iterations, O2HB_LIVE_THRESHOLD + 1); + live_threshold = O2HB_LIVE_THRESHOLD; + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + if (o2hb_pop_count(&o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1) + live_threshold <<= 1; + spin_unlock(&o2hb_live_lock); + } + atomic_set(®->hr_steady_iterations, live_threshold + 1); hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s", reg->hr_item.ci_name); -- cgit v0.10.2 From 33c12a5436464f8d4f56d68e5e79e24a3a1f11aa Mon Sep 17 00:00:00 2001 From: Sunil Mushran Date: Wed, 4 May 2011 10:28:01 -0700 Subject: ocfs2/cluster: Heartbeat mismatch message improved If o2hb finds unexpected values in the heartbeat slot, it prints a message "ERROR: Device "dm-6": another node is heartbeating in our slot!" This message could be misleading. This patch adds two more messages to help users better diagnose the problem. Signed-off-by: Sunil Mushran Acked-by: Mark Fasheh Signed-off-by: Joel Becker diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c index 1d28505..9a3e6bb 100644 --- a/fs/ocfs2/cluster/heartbeat.c +++ b/fs/ocfs2/cluster/heartbeat.c @@ -539,25 +539,41 @@ static int o2hb_verify_crc(struct o2hb_region *reg, /* We want to make sure that nobody is heartbeating on top of us -- * this will help detect an invalid configuration. */ -static int o2hb_check_last_timestamp(struct o2hb_region *reg) +static void o2hb_check_last_timestamp(struct o2hb_region *reg) { - int node_num, ret; struct o2hb_disk_slot *slot; struct o2hb_disk_heartbeat_block *hb_block; + char *errstr; - node_num = o2nm_this_node(); - - ret = 1; - slot = ®->hr_slots[node_num]; + slot = ®->hr_slots[o2nm_this_node()]; /* Don't check on our 1st timestamp */ - if (slot->ds_last_time) { - hb_block = slot->ds_raw_block; + if (!slot->ds_last_time) + return; - if (le64_to_cpu(hb_block->hb_seq) != slot->ds_last_time) - ret = 0; - } + hb_block = slot->ds_raw_block; + if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time && + le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation && + hb_block->hb_node == slot->ds_node_num) + return; - return ret; +#define ERRSTR1 "Another node is heartbeating on device" +#define ERRSTR2 "Heartbeat generation mismatch on device" +#define ERRSTR3 "Heartbeat sequence mismatch on device" + + if (hb_block->hb_node != slot->ds_node_num) + errstr = ERRSTR1; + else if (le64_to_cpu(hb_block->hb_generation) != + slot->ds_last_generation) + errstr = ERRSTR2; + else + errstr = ERRSTR3; + + mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), " + "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name, + slot->ds_node_num, (unsigned long long)slot->ds_last_generation, + (unsigned long long)slot->ds_last_time, hb_block->hb_node, + (unsigned long long)le64_to_cpu(hb_block->hb_generation), + (unsigned long long)le64_to_cpu(hb_block->hb_seq)); } static inline void o2hb_prepare_block(struct o2hb_region *reg, @@ -983,9 +999,7 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) /* With an up to date view of the slots, we can check that no * other node has been improperly configured to heartbeat in * our slot. */ - if (!o2hb_check_last_timestamp(reg)) - mlog(ML_ERROR, "Device \"%s\": another node is heartbeating " - "in our slot!\n", reg->hr_dev_name); + o2hb_check_last_timestamp(reg); /* fill in the proper info for our next heartbeat */ o2hb_prepare_block(reg, reg->hr_generation); @@ -999,8 +1013,8 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) } i = -1; - while((i = find_next_bit(configured_nodes, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { - + while((i = find_next_bit(configured_nodes, + O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { change |= o2hb_check_slot(reg, ®->hr_slots[i]); } -- cgit v0.10.2 From 10b3dd76117a327557b8cb898b41c18afd08dc86 Mon Sep 17 00:00:00 2001 From: Sunil Mushran Date: Wed, 4 May 2011 10:28:02 -0700 Subject: ocfs2: Skip mount recovery for hard-ro mounts Patch skips mount recovery for hard-ro mounts which otherwise leads to an oops. Signed-off-by: Sunil Mushran Acked-by: Mark Fasheh Signed-off-by: Joel Becker diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c index b141a44..295d564 100644 --- a/fs/ocfs2/journal.c +++ b/fs/ocfs2/journal.c @@ -1260,6 +1260,9 @@ void ocfs2_complete_mount_recovery(struct ocfs2_super *osb) { struct ocfs2_journal *journal = osb->journal; + if (ocfs2_is_hard_readonly(osb)) + return; + /* No need to queue up our truncate_log as regular cleanup will catch * that */ ocfs2_queue_recovery_completion(journal, osb->slot_num, -- cgit v0.10.2 From df016c665b10ae80d8db67ec8103b50c5c234e5c Mon Sep 17 00:00:00 2001 From: Sunil Mushran Date: Wed, 4 May 2011 10:28:07 -0700 Subject: ocfs2/dlm: Target node death during resource migration leads to thread spin During resource migration, if the target node were to die, the thread doing the migration spins until the target node is not removed from the domain map. This patch slows the spin by making the thread wait for the recovery to kick in. Signed-off-by: Sunil Mushran Signed-off-by: Joel Becker diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c index fede57e..84d1663 100644 --- a/fs/ocfs2/dlm/dlmmaster.c +++ b/fs/ocfs2/dlm/dlmmaster.c @@ -2574,6 +2574,9 @@ fail: res->state &= ~DLM_LOCK_RES_MIGRATING; wake = 1; spin_unlock(&res->spinlock); + if (dlm_is_host_down(ret)) + dlm_wait_for_node_death(dlm, target, + DLM_NODE_DEATH_WAIT_MAX); goto leave; } -- cgit v0.10.2 From df7f99670a4c76f269ae57ce91876b309417a316 Mon Sep 17 00:00:00 2001 From: Joel Becker Date: Tue, 22 Feb 2011 01:09:49 -0800 Subject: configfs: Don't try to d_delete() negative dentries. When configfs is faking mkdir() on its subsystem or default group objects, it starts by adding a negative dentry. It then tries to instantiate the group. If that should fail, it must clean up after itself. I was using d_delete() here, but configfs_attach_group() promises to return an empty dentry on error. d_delete() explodes with the entry dentry. Let's try d_drop() instead. The unhashing is what we want for our dentry. Signed-off-by: Joel Becker diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c index 3313dd1..b11d734 100644 --- a/fs/configfs/dir.c +++ b/fs/configfs/dir.c @@ -689,7 +689,8 @@ static int create_default_group(struct config_group *parent_group, sd = child->d_fsdata; sd->s_type |= CONFIGFS_USET_DEFAULT; } else { - d_delete(child); + BUG_ON(child->d_inode); + d_drop(child); dput(child); } } @@ -1683,7 +1684,8 @@ int configfs_register_subsystem(struct configfs_subsystem *subsys) err = configfs_attach_group(sd->s_element, &group->cg_item, dentry); if (err) { - d_delete(dentry); + BUG_ON(dentry->d_inode); + d_drop(dentry); dput(dentry); } else { spin_lock(&configfs_dirent_lock); -- cgit v0.10.2 From 24307aa1e707b31613be92deaba7990e16bc1aec Mon Sep 17 00:00:00 2001 From: Joel Becker Date: Wed, 18 May 2011 04:08:16 -0700 Subject: configfs: Fix race between configfs_readdir() and configfs_d_iput() configfs_readdir() will use the existing inode numbers of inodes in the dcache, but it makes them up for attribute files that aren't currently instantiated. There is a race where a closing attribute file can be tearing down at the same time as configfs_readdir() is trying to get its inode number. We want to get the inode number of open attribute files, because they should match while instantiated. We can't lock down the transition where dentry->d_inode is set to NULL, so we just check for NULL there. We can, however, ensure that an inode we find isn't iput() in configfs_d_iput() until after we've accessed it. Signed-off-by: Joel Becker diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c index b11d734..9a37a9b 100644 --- a/fs/configfs/dir.c +++ b/fs/configfs/dir.c @@ -53,11 +53,14 @@ DEFINE_SPINLOCK(configfs_dirent_lock); static void configfs_d_iput(struct dentry * dentry, struct inode * inode) { - struct configfs_dirent * sd = dentry->d_fsdata; + struct configfs_dirent *sd = dentry->d_fsdata; if (sd) { BUG_ON(sd->s_dentry != dentry); + /* Coordinate with configfs_readdir */ + spin_lock(&configfs_dirent_lock); sd->s_dentry = NULL; + spin_unlock(&configfs_dirent_lock); configfs_put(sd); } iput(inode); @@ -1546,7 +1549,7 @@ static int configfs_readdir(struct file * filp, void * dirent, filldir_t filldir struct configfs_dirent * parent_sd = dentry->d_fsdata; struct configfs_dirent *cursor = filp->private_data; struct list_head *p, *q = &cursor->s_sibling; - ino_t ino; + ino_t ino = 0; int i = filp->f_pos; switch (i) { @@ -1574,6 +1577,7 @@ static int configfs_readdir(struct file * filp, void * dirent, filldir_t filldir struct configfs_dirent *next; const char * name; int len; + struct inode *inode = NULL; next = list_entry(p, struct configfs_dirent, s_sibling); @@ -1582,9 +1586,28 @@ static int configfs_readdir(struct file * filp, void * dirent, filldir_t filldir name = configfs_get_name(next); len = strlen(name); - if (next->s_dentry) - ino = next->s_dentry->d_inode->i_ino; - else + + /* + * We'll have a dentry and an inode for + * PINNED items and for open attribute + * files. We lock here to prevent a race + * with configfs_d_iput() clearing + * s_dentry before calling iput(). + * + * Why do we go to the trouble? If + * someone has an attribute file open, + * the inode number should match until + * they close it. Beyond that, we don't + * care. + */ + spin_lock(&configfs_dirent_lock); + dentry = next->s_dentry; + if (dentry) + inode = dentry->d_inode; + if (inode) + ino = inode->i_ino; + spin_unlock(&configfs_dirent_lock); + if (!inode) ino = iunique(configfs_sb, 2); if (filldir(dirent, name, len, filp->f_pos, ino, -- cgit v0.10.2