Merge branch 'parisc-4.4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull parisc update from Helge Deller:
 "This patchset adds Huge Page and HUGETLBFS support for parisc"

Honestly, the hugepage support should have gone through in the merge
window, and is not really an rc-time fix.  But it only touches
arch/parisc, and I cannot find it in myself to care.  If one of the
three parisc users notices a breakage, I will point at Helge and make
rude farting noises.

* 'parisc-4.4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Map kernel text and data on huge pages
  parisc: Add Huge Page and HUGETLBFS support
  parisc: Use long branch to do_syscall_trace_exit
  parisc: Increase initial kernel mapping to 32MB on 64bit kernel
  parisc: Initialize the fault vector earlier in the boot process.
  parisc: Add defines for Huge page support
  parisc: Drop unused MADV_xxxK_PAGES flags from asm/mman.h
  parisc: Drop definition of start_thread_som for HP-UX SOM binaries
  parisc: Fix wrong comment regarding first pmd entry flags

17 files changed, 382 insertions, 116 deletions

diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index c365469..729f891 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -108,6 +108,9 @@ config PGTABLE_LEVELS
 	default 3 if 64BIT && PARISC_PAGE_SIZE_4KB
 	default 2
 
+config SYS_SUPPORTS_HUGETLBFS
+	def_bool y if PA20
+
 source "init/Kconfig"
 
 source "kernel/Kconfig.freezer"
diff --git a/arch/parisc/include/asm/hugetlb.h b/arch/parisc/include/asm/hugetlb.h
new file mode 100644
index 0000000..7d56a9c
--- /dev/null
+++ b/arch/parisc/include/asm/hugetlb.h
@@ -0,0 +1,85 @@
+#ifndef _ASM_PARISC64_HUGETLB_H
+#define _ASM_PARISC64_HUGETLB_H
+
+#include <asm/page.h>
+#include <asm-generic/hugetlb.h>
+
+
+void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+		     pte_t *ptep, pte_t pte);
+
+pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
+			      pte_t *ptep);
+
+static inline int is_hugepage_only_range(struct mm_struct *mm,
+					 unsigned long addr,
+					 unsigned long len) {
+	return 0;
+}
+
+/*
+ * If the arch doesn't supply something else, assume that hugepage
+ * size aligned regions are ok without further preparation.
+ */
+static inline int prepare_hugepage_range(struct file *file,
+			unsigned long addr, unsigned long len)
+{
+	if (len & ~HPAGE_MASK)
+		return -EINVAL;
+	if (addr & ~HPAGE_MASK)
+		return -EINVAL;
+	return 0;
+}
+
+static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
+					  unsigned long addr, unsigned long end,
+					  unsigned long floor,
+					  unsigned long ceiling)
+{
+	free_pgd_range(tlb, addr, end, floor, ceiling);
+}
+
+static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
+					 unsigned long addr, pte_t *ptep)
+{
+}
+
+static inline int huge_pte_none(pte_t pte)
+{
+	return pte_none(pte);
+}
+
+static inline pte_t huge_pte_wrprotect(pte_t pte)
+{
+	return pte_wrprotect(pte);
+}
+
+static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
+					   unsigned long addr, pte_t *ptep)
+{
+	pte_t old_pte = *ptep;
+	set_huge_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
+}
+
+static inline int huge_ptep_set_access_flags(struct vm_area_struct *vma,
+					     unsigned long addr, pte_t *ptep,
+					     pte_t pte, int dirty)
+{
+	int changed = !pte_same(*ptep, pte);
+	if (changed) {
+		set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+		flush_tlb_page(vma, addr);
+	}
+	return changed;
+}
+
+static inline pte_t huge_ptep_get(pte_t *ptep)
+{
+	return *ptep;
+}
+
+static inline void arch_clear_hugepage_flags(struct page *page)
+{
+}
+
+#endif /* _ASM_PARISC64_HUGETLB_H */
diff --git a/arch/parisc/include/asm/page.h b/arch/parisc/include/asm/page.h
index 60d5d17..80e742a 100644
--- a/arch/parisc/include/asm/page.h
+++ b/arch/parisc/include/asm/page.h
@@ -145,11 +145,22 @@ extern int npmem_ranges;
 #endif /* CONFIG_DISCONTIGMEM */
 
 #ifdef CONFIG_HUGETLB_PAGE
-#define HPAGE_SHIFT		22	/* 4MB (is this fixed?) */
+#define HPAGE_SHIFT		PMD_SHIFT /* fixed for transparent huge pages */
 #define HPAGE_SIZE      	((1UL) << HPAGE_SHIFT)
 #define HPAGE_MASK		(~(HPAGE_SIZE - 1))
 #define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
+
+#if defined(CONFIG_64BIT) && defined(CONFIG_PARISC_PAGE_SIZE_4KB)
+# define REAL_HPAGE_SHIFT	20 /* 20 = 1MB */
+# define _HUGE_PAGE_SIZE_ENCODING_DEFAULT _PAGE_SIZE_ENCODING_1M
+#elif !defined(CONFIG_64BIT) && defined(CONFIG_PARISC_PAGE_SIZE_4KB)
+# define REAL_HPAGE_SHIFT	22 /* 22 = 4MB */
+# define _HUGE_PAGE_SIZE_ENCODING_DEFAULT _PAGE_SIZE_ENCODING_4M
+#else
+# define REAL_HPAGE_SHIFT	24 /* 24 = 16MB */
+# define _HUGE_PAGE_SIZE_ENCODING_DEFAULT _PAGE_SIZE_ENCODING_16M
 #endif
+#endif /* CONFIG_HUGETLB_PAGE */
 
 #define virt_addr_valid(kaddr)	pfn_valid(__pa(kaddr) >> PAGE_SHIFT)
 
diff --git a/arch/parisc/include/asm/pgalloc.h b/arch/parisc/include/asm/pgalloc.h
index 3edbb9f..f2fd327 100644
--- a/arch/parisc/include/asm/pgalloc.h
+++ b/arch/parisc/include/asm/pgalloc.h
@@ -35,7 +35,7 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm)
 				        PxD_FLAG_VALID | 
 					PxD_FLAG_ATTACHED) 
 			+ (__u32)(__pa((unsigned long)pgd) >> PxD_VALUE_SHIFT));
-		/* The first pmd entry also is marked with _PAGE_GATEWAY as
+		/* The first pmd entry also is marked with PxD_FLAG_ATTACHED as
 		 * a signal that this pmd may not be freed */
 		__pgd_val_set(*pgd, PxD_FLAG_ATTACHED);
 #endif
diff --git a/arch/parisc/include/asm/pgtable.h b/arch/parisc/include/asm/pgtable.h
index f93c4a4..d8534f9 100644
--- a/arch/parisc/include/asm/pgtable.h
+++ b/arch/parisc/include/asm/pgtable.h
@@ -83,7 +83,11 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr)
 	printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, (unsigned long)pgd_val(e))
 
 /* This is the size of the initially mapped kernel memory */
-#define KERNEL_INITIAL_ORDER	24	/* 0 to 1<<24 = 16MB */
+#ifdef CONFIG_64BIT
+#define KERNEL_INITIAL_ORDER	25	/* 1<<25 = 32MB */
+#else
+#define KERNEL_INITIAL_ORDER	24	/* 1<<24 = 16MB */
+#endif
 #define KERNEL_INITIAL_SIZE	(1 << KERNEL_INITIAL_ORDER)
 
 #if CONFIG_PGTABLE_LEVELS == 3
@@ -167,7 +171,7 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr)
 #define _PAGE_NO_CACHE_BIT 24   /* (0x080) Uncached Page (U bit) */
 #define _PAGE_ACCESSED_BIT 23   /* (0x100) Software: Page Accessed */
 #define _PAGE_PRESENT_BIT  22   /* (0x200) Software: translation valid */
-/* bit 21 was formerly the FLUSH bit but is now unused */
+#define _PAGE_HPAGE_BIT    21   /* (0x400) Software: Huge Page */
 #define _PAGE_USER_BIT     20   /* (0x800) Software: User accessible page */
 
 /* N.B. The bits are defined in terms of a 32 bit word above, so the */
@@ -194,6 +198,7 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr)
 #define _PAGE_NO_CACHE (1 << xlate_pabit(_PAGE_NO_CACHE_BIT))
 #define _PAGE_ACCESSED (1 << xlate_pabit(_PAGE_ACCESSED_BIT))
 #define _PAGE_PRESENT  (1 << xlate_pabit(_PAGE_PRESENT_BIT))
+#define _PAGE_HUGE     (1 << xlate_pabit(_PAGE_HPAGE_BIT))
 #define _PAGE_USER     (1 << xlate_pabit(_PAGE_USER_BIT))
 
 #define _PAGE_TABLE	(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE |  _PAGE_DIRTY | _PAGE_ACCESSED)
@@ -217,7 +222,7 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr)
 #define PxD_FLAG_VALID    (1 << xlate_pabit(_PxD_VALID_BIT))
 #define PxD_FLAG_MASK     (0xf)
 #define PxD_FLAG_SHIFT    (4)
-#define PxD_VALUE_SHIFT   (8) /* (PAGE_SHIFT-PxD_FLAG_SHIFT) */
+#define PxD_VALUE_SHIFT   (PFN_PTE_SHIFT-PxD_FLAG_SHIFT)
 
 #ifndef __ASSEMBLY__
 
@@ -363,6 +368,18 @@ static inline pte_t pte_mkwrite(pte_t pte)	{ pte_val(pte) |= _PAGE_WRITE; return
 static inline pte_t pte_mkspecial(pte_t pte)	{ return pte; }
 
 /*
+ * Huge pte definitions.
+ */
+#ifdef CONFIG_HUGETLB_PAGE
+#define pte_huge(pte)           (pte_val(pte) & _PAGE_HUGE)
+#define pte_mkhuge(pte)         (__pte(pte_val(pte) | _PAGE_HUGE))
+#else
+#define pte_huge(pte)           (0)
+#define pte_mkhuge(pte)         (pte)
+#endif
+
+
+/*
  * Conversion functions: convert a page and protection to a page entry,
  * and a page entry and page directory to the page they refer to.
  */
@@ -410,8 +427,9 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 /* Find an entry in the second-level page table.. */
 
 #if CONFIG_PGTABLE_LEVELS == 3
+#define pmd_index(addr)         (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))
 #define pmd_offset(dir,address) \
-((pmd_t *) pgd_page_vaddr(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1)))
+((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(address))
 #else
 #define pmd_offset(dir,addr) ((pmd_t *) dir)
 #endif
diff --git a/arch/parisc/include/asm/processor.h b/arch/parisc/include/asm/processor.h
index 54adb60..7e759ec 100644
--- a/arch/parisc/include/asm/processor.h
+++ b/arch/parisc/include/asm/processor.h
@@ -192,33 +192,6 @@ void show_trace(struct task_struct *task, unsigned long *stack);
  */
 typedef unsigned int elf_caddr_t;
 
-#define start_thread_som(regs, new_pc, new_sp) do {	\
-	unsigned long *sp = (unsigned long *)new_sp;	\
-	__u32 spaceid = (__u32)current->mm->context;	\
-	unsigned long pc = (unsigned long)new_pc;	\
-	/* offset pc for priv. level */			\
-	pc |= 3;					\
-							\
-	regs->iasq[0] = spaceid;			\
-	regs->iasq[1] = spaceid;			\
-	regs->iaoq[0] = pc;				\
-	regs->iaoq[1] = pc + 4;                         \
-	regs->sr[2] = LINUX_GATEWAY_SPACE;              \
-	regs->sr[3] = 0xffff;				\
-	regs->sr[4] = spaceid;				\
-	regs->sr[5] = spaceid;				\
-	regs->sr[6] = spaceid;				\
-	regs->sr[7] = spaceid;				\
-	regs->gr[ 0] = USER_PSW;                        \
-	regs->gr[30] = ((new_sp)+63)&~63;		\
-	regs->gr[31] = pc;				\
-							\
-	get_user(regs->gr[26],&sp[0]);			\
-	get_user(regs->gr[25],&sp[-1]); 		\
-	get_user(regs->gr[24],&sp[-2]); 		\
-	get_user(regs->gr[23],&sp[-3]); 		\
-} while(0)
-
 /* The ELF abi wants things done a "wee bit" differently than
  * som does.  Supporting this behavior here avoids
  * having our own version of create_elf_tables.
diff --git a/arch/parisc/include/uapi/asm/mman.h b/arch/parisc/include/uapi/asm/mman.h
index ecc3ae1..dd4d187 100644
--- a/arch/parisc/include/uapi/asm/mman.h
+++ b/arch/parisc/include/uapi/asm/mman.h
@@ -49,16 +49,6 @@
 #define MADV_DONTFORK	10		/* don't inherit across fork */
 #define MADV_DOFORK	11		/* do inherit across fork */
 
-/* The range 12-64 is reserved for page size specification. */
-#define MADV_4K_PAGES   12              /* Use 4K pages  */
-#define MADV_16K_PAGES  14              /* Use 16K pages */
-#define MADV_64K_PAGES  16              /* Use 64K pages */
-#define MADV_256K_PAGES 18              /* Use 256K pages */
-#define MADV_1M_PAGES   20              /* Use 1 Megabyte pages */
-#define MADV_4M_PAGES   22              /* Use 4 Megabyte pages */
-#define MADV_16M_PAGES  24              /* Use 16 Megabyte pages */
-#define MADV_64M_PAGES  26              /* Use 64 Megabyte pages */
-
 #define MADV_MERGEABLE   65		/* KSM may merge identical pages */
 #define MADV_UNMERGEABLE 66		/* KSM may not merge identical pages */
 
diff --git a/arch/parisc/kernel/asm-offsets.c b/arch/parisc/kernel/asm-offsets.c
index 59001ce..d2f6257 100644
--- a/arch/parisc/kernel/asm-offsets.c
+++ b/arch/parisc/kernel/asm-offsets.c
@@ -290,6 +290,14 @@ int main(void)
 	DEFINE(ASM_PFN_PTE_SHIFT, PFN_PTE_SHIFT);
 	DEFINE(ASM_PT_INITIAL, PT_INITIAL);
 	BLANK();
+	/* HUGEPAGE_SIZE is only used in vmlinux.lds.S to align kernel text
+	 * and kernel data on physical huge pages */
+#ifdef CONFIG_HUGETLB_PAGE
+	DEFINE(HUGEPAGE_SIZE, 1UL << REAL_HPAGE_SHIFT);
+#else
+	DEFINE(HUGEPAGE_SIZE, PAGE_SIZE);
+#endif
+	BLANK();
 	DEFINE(EXCDATA_IP, offsetof(struct exception_data, fault_ip));
 	DEFINE(EXCDATA_SPACE, offsetof(struct exception_data, fault_space));
 	DEFINE(EXCDATA_ADDR, offsetof(struct exception_data, fault_addr));
diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S
index c5ef408..623496c 100644
--- a/arch/parisc/kernel/entry.S
+++ b/arch/parisc/kernel/entry.S
@@ -502,21 +502,38 @@
 	STREG		\pte,0(\ptp)
 	.endm
 
+	/* We have (depending on the page size):
+	 * - 38 to 52-bit Physical Page Number
+	 * - 12 to 26-bit page offset
+	 */
 	/* bitshift difference between a PFN (based on kernel's PAGE_SIZE)
 	 * to a CPU TLB 4k PFN (4k => 12 bits to shift) */
-	#define PAGE_ADD_SHIFT  (PAGE_SHIFT-12)
+	#define PAGE_ADD_SHIFT		(PAGE_SHIFT-12)
+	#define PAGE_ADD_HUGE_SHIFT	(REAL_HPAGE_SHIFT-12)
 
 	/* Drop prot bits and convert to page addr for iitlbt and idtlbt */
-	.macro		convert_for_tlb_insert20 pte
+	.macro		convert_for_tlb_insert20 pte,tmp
+#ifdef CONFIG_HUGETLB_PAGE
+	copy		\pte,\tmp
+	extrd,u		\tmp,(63-ASM_PFN_PTE_SHIFT)+(63-58)+PAGE_ADD_SHIFT,\
+				64-PAGE_SHIFT-PAGE_ADD_SHIFT,\pte
+
+	depdi		_PAGE_SIZE_ENCODING_DEFAULT,63,\
+				(63-58)+PAGE_ADD_SHIFT,\pte
+	extrd,u,*=	\tmp,_PAGE_HPAGE_BIT+32,1,%r0
+	depdi		_HUGE_PAGE_SIZE_ENCODING_DEFAULT,63,\
+				(63-58)+PAGE_ADD_HUGE_SHIFT,\pte
+#else /* Huge pages disabled */
 	extrd,u		\pte,(63-ASM_PFN_PTE_SHIFT)+(63-58)+PAGE_ADD_SHIFT,\
 				64-PAGE_SHIFT-PAGE_ADD_SHIFT,\pte
 	depdi		_PAGE_SIZE_ENCODING_DEFAULT,63,\
 				(63-58)+PAGE_ADD_SHIFT,\pte
+#endif
 	.endm
 
 	/* Convert the pte and prot to tlb insertion values.  How
 	 * this happens is quite subtle, read below */
-	.macro		make_insert_tlb	spc,pte,prot
+	.macro		make_insert_tlb	spc,pte,prot,tmp
 	space_to_prot   \spc \prot        /* create prot id from space */
 	/* The following is the real subtlety.  This is depositing
 	 * T <-> _PAGE_REFTRAP
@@ -553,7 +570,7 @@
 	depdi		1,12,1,\prot
 
 	/* Drop prot bits and convert to page addr for iitlbt and idtlbt */
-	convert_for_tlb_insert20 \pte
+	convert_for_tlb_insert20 \pte \tmp
 	.endm
 
 	/* Identical macro to make_insert_tlb above, except it
@@ -646,17 +663,12 @@
 
 
 	/*
-	 * Align fault_vector_20 on 4K boundary so that both
-	 * fault_vector_11 and fault_vector_20 are on the
-	 * same page. This is only necessary as long as we
-	 * write protect the kernel text, which we may stop
-	 * doing once we use large page translations to cover
-	 * the static part of the kernel address space.
+	 * Fault_vectors are architecturally required to be aligned on a 2K
+	 * boundary
 	 */
 
 	.text
-
-	.align 4096
+	.align 2048
 
 ENTRY(fault_vector_20)
 	/* First vector is invalid (0) */
@@ -1147,7 +1159,7 @@ dtlb_miss_20w:
 	tlb_lock	spc,ptp,pte,t0,t1,dtlb_check_alias_20w
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 	
 	idtlbt          pte,prot
 
@@ -1173,7 +1185,7 @@ nadtlb_miss_20w:
 	tlb_lock	spc,ptp,pte,t0,t1,nadtlb_check_alias_20w
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	idtlbt          pte,prot
 
@@ -1267,7 +1279,7 @@ dtlb_miss_20:
 	tlb_lock	spc,ptp,pte,t0,t1,dtlb_check_alias_20
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	f_extend	pte,t1
 
@@ -1295,7 +1307,7 @@ nadtlb_miss_20:
 	tlb_lock	spc,ptp,pte,t0,t1,nadtlb_check_alias_20
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	f_extend	pte,t1
 	
@@ -1404,7 +1416,7 @@ itlb_miss_20w:
 	tlb_lock	spc,ptp,pte,t0,t1,itlb_fault
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 	
 	iitlbt          pte,prot
 
@@ -1428,7 +1440,7 @@ naitlb_miss_20w:
 	tlb_lock	spc,ptp,pte,t0,t1,naitlb_check_alias_20w
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	iitlbt          pte,prot
 
@@ -1514,7 +1526,7 @@ itlb_miss_20:
 	tlb_lock	spc,ptp,pte,t0,t1,itlb_fault
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	f_extend	pte,t1
 
@@ -1534,7 +1546,7 @@ naitlb_miss_20:
 	tlb_lock	spc,ptp,pte,t0,t1,naitlb_check_alias_20
 	update_accessed	ptp,pte,t0,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	f_extend	pte,t1
 
@@ -1566,7 +1578,7 @@ dbit_trap_20w:
 	tlb_lock	spc,ptp,pte,t0,t1,dbit_fault
 	update_dirty	ptp,pte,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 		
 	idtlbt          pte,prot
 
@@ -1610,7 +1622,7 @@ dbit_trap_20:
 	tlb_lock	spc,ptp,pte,t0,t1,dbit_fault
 	update_dirty	ptp,pte,t1
 
-	make_insert_tlb	spc,pte,prot
+	make_insert_tlb	spc,pte,prot,t1
 
 	f_extend	pte,t1
 	
diff --git a/arch/parisc/kernel/head.S b/arch/parisc/kernel/head.S
index e7d6452..75aa0db 100644
--- a/arch/parisc/kernel/head.S
+++ b/arch/parisc/kernel/head.S
@@ -69,7 +69,7 @@ $bss_loop:
 	stw,ma          %arg2,4(%r1)
 	stw,ma          %arg3,4(%r1)
 
-	/* Initialize startup VM. Just map first 8/16 MB of memory */
+	/* Initialize startup VM. Just map first 16/32 MB of memory */
 	load32		PA(swapper_pg_dir),%r4
 	mtctl		%r4,%cr24	/* Initialize kernel root pointer */
 	mtctl		%r4,%cr25	/* Initialize user root pointer */
@@ -107,7 +107,7 @@ $bss_loop:
 	/* Now initialize the PTEs themselves.  We use RWX for
 	 * everything ... it will get remapped correctly later */
 	ldo		0+_PAGE_KERNEL_RWX(%r0),%r3 /* Hardwired 0 phys addr start */
-	ldi		(1<<(KERNEL_INITIAL_ORDER-PAGE_SHIFT)),%r11 /* PFN count */
+	load32		(1<<(KERNEL_INITIAL_ORDER-PAGE_SHIFT)),%r11 /* PFN count */
 	load32		PA(pg0),%r1
 
 $pgt_fill_loop:
diff --git a/arch/parisc/kernel/setup.c b/arch/parisc/kernel/setup.c
index 72a3c65..f7ea626 100644
--- a/arch/parisc/kernel/setup.c
+++ b/arch/parisc/kernel/setup.c
@@ -130,7 +130,16 @@ void __init setup_arch(char **cmdline_p)
 	printk(KERN_INFO "The 32-bit Kernel has started...\n");
 #endif
 
-	printk(KERN_INFO "Default page size is %dKB.\n", (int)(PAGE_SIZE / 1024));
+	printk(KERN_INFO "Kernel default page size is %d KB. Huge pages ",
+		(int)(PAGE_SIZE / 1024));
+#ifdef CONFIG_HUGETLB_PAGE
+	printk(KERN_CONT "enabled with %d MB physical and %d MB virtual size",
+		 1 << (REAL_HPAGE_SHIFT - 20), 1 << (HPAGE_SHIFT - 20));
+#else
+	printk(KERN_CONT "disabled");
+#endif
+	printk(KERN_CONT ".\n");
+
 
 	pdc_console_init();
 
@@ -377,6 +386,7 @@ arch_initcall(parisc_init);
 void start_parisc(void)
 {
 	extern void start_kernel(void);
+	extern void early_trap_init(void);
 
 	int ret, cpunum;
 	struct pdc_coproc_cfg coproc_cfg;
@@ -397,6 +407,8 @@ void start_parisc(void)
 		panic("must have an fpu to boot linux");
 	}
 
+	early_trap_init(); /* initialize checksum of fault_vector */
+
 	start_kernel();
 	// not reached
 }
diff --git a/arch/parisc/kernel/syscall.S b/arch/parisc/kernel/syscall.S
index 0b8d26d..3fbd725 100644
--- a/arch/parisc/kernel/syscall.S
+++ b/arch/parisc/kernel/syscall.S
@@ -369,7 +369,7 @@ tracesys_exit:
 	ldo	-16(%r30),%r29			/* Reference param save area */
 #endif
 	ldo	TASK_REGS(%r1),%r26
-	bl	do_syscall_trace_exit,%r2
+	BL	do_syscall_trace_exit,%r2
 	STREG   %r28,TASK_PT_GR28(%r1)          /* save return value now */
 	ldo     -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1      /* get task ptr */
 	LDREG	TI_TASK(%r1), %r1
@@ -390,7 +390,7 @@ tracesys_sigexit:
 #ifdef CONFIG_64BIT
 	ldo	-16(%r30),%r29			/* Reference param save area */
 #endif
-	bl	do_syscall_trace_exit,%r2
+	BL	do_syscall_trace_exit,%r2
 	ldo	TASK_REGS(%r1),%r26
 
 	ldil	L%syscall_exit_rfi,%r1
diff --git a/arch/parisc/kernel/traps.c b/arch/parisc/kernel/traps.c
index b99b39f..553b098 100644
--- a/arch/parisc/kernel/traps.c
+++ b/arch/parisc/kernel/traps.c
@@ -807,7 +807,7 @@ void notrace handle_interruption(int code, struct pt_regs *regs)
 }
 
 
-int __init check_ivt(void *iva)
+void __init initialize_ivt(const void *iva)
 {
 	extern u32 os_hpmc_size;
 	extern const u32 os_hpmc[];
@@ -818,8 +818,8 @@ int __init check_ivt(void *iva)
 	u32 *hpmcp;
 	u32 length;
 
-	if (strcmp((char *)iva, "cows can fly"))
-		return -1;
+	if (strcmp((const char *)iva, "cows can fly"))
+		panic("IVT invalid");
 
 	ivap = (u32 *)iva;
 
@@ -839,28 +839,23 @@ int __init check_ivt(void *iva)
 	    check += ivap[i];
 
 	ivap[5] = -check;
-
-	return 0;
 }
 	
-#ifndef CONFIG_64BIT
-extern const void fault_vector_11;
-#endif
-extern const void fault_vector_20;
 
-void __init trap_init(void)
+/* early_trap_init() is called before we set up kernel mappings and
+ * write-protect the kernel */
+void  __init early_trap_init(void)
 {
-	void *iva;
+	extern const void fault_vector_20;
 
-	if (boot_cpu_data.cpu_type >= pcxu)
-		iva = (void *) &fault_vector_20;
-	else
-#ifdef CONFIG_64BIT
-		panic("Can't boot 64-bit OS on PA1.1 processor!");
-#else
-		iva = (void *) &fault_vector_11;
+#ifndef CONFIG_64BIT
+	extern const void fault_vector_11;
+	initialize_ivt(&fault_vector_11);
 #endif
 
-	if (check_ivt(iva))
-		panic("IVT invalid");
+	initialize_ivt(&fault_vector_20);
+}
+
+void __init trap_init(void)
+{
 }
diff --git a/arch/parisc/kernel/vmlinux.lds.S b/arch/parisc/kernel/vmlinux.lds.S
index 0dacc5c..308f290 100644
--- a/arch/parisc/kernel/vmlinux.lds.S
+++ b/arch/parisc/kernel/vmlinux.lds.S
@@ -60,7 +60,7 @@ SECTIONS
 		EXIT_DATA
 	}
 	PERCPU_SECTION(8)
-	. = ALIGN(PAGE_SIZE);
+	. = ALIGN(HUGEPAGE_SIZE);
 	__init_end = .;
 	/* freed after init ends here */
 
@@ -116,7 +116,7 @@ SECTIONS
 	 * that we can properly leave these
 	 * as writable
 	 */
-	. = ALIGN(PAGE_SIZE);
+	. = ALIGN(HUGEPAGE_SIZE);
 	data_start = .;
 
 	EXCEPTION_TABLE(8)
@@ -135,8 +135,11 @@ SECTIONS
 	_edata = .;
 
 	/* BSS */
-	BSS_SECTION(PAGE_SIZE, PAGE_SIZE, 8)
+	BSS_SECTION(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE)
+
+	/* bootmap is allocated in setup_bootmem() directly behind bss. */
 
+	. = ALIGN(HUGEPAGE_SIZE);
 	_end = . ;
 
 	STABS_DEBUG
diff --git a/arch/parisc/mm/Makefile b/arch/parisc/mm/Makefile
index 758ceef..134393d 100644
--- a/arch/parisc/mm/Makefile
+++ b/arch/parisc/mm/Makefile
@@ -3,3 +3,4 @@
 #
 
 obj-y	 := init.o fault.o ioremap.o
+obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
diff --git a/arch/parisc/mm/hugetlbpage.c b/arch/parisc/mm/hugetlbpage.c
new file mode 100644
index 0000000..f6fdc77
--- /dev/null
+++ b/arch/parisc/mm/hugetlbpage.c
@@ -0,0 +1,161 @@
+/*
+ * PARISC64 Huge TLB page support.
+ *
+ * This parisc implementation is heavily based on the SPARC and x86 code.
+ *
+ * Copyright (C) 2015 Helge Deller <deller@gmx.de>
+ */
+
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/pagemap.h>
+#include <linux/sysctl.h>
+
+#include <asm/mman.h>
+#include <asm/pgalloc.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+#include <asm/cacheflush.h>
+#include <asm/mmu_context.h>
+
+
+unsigned long
+hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
+		unsigned long len, unsigned long pgoff, unsigned long flags)
+{
+	struct hstate *h = hstate_file(file);
+
+	if (len & ~huge_page_mask(h))
+		return -EINVAL;
+	if (len > TASK_SIZE)
+		return -ENOMEM;
+
+	if (flags & MAP_FIXED)
+		if (prepare_hugepage_range(file, addr, len))
+			return -EINVAL;
+
+	if (addr)
+		addr = ALIGN(addr, huge_page_size(h));
+
+	/* we need to make sure the colouring is OK */
+	return arch_get_unmapped_area(file, addr, len, pgoff, flags);
+}
+
+
+pte_t *huge_pte_alloc(struct mm_struct *mm,
+			unsigned long addr, unsigned long sz)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+
+	/* We must align the address, because our caller will run
+	 * set_huge_pte_at() on whatever we return, which writes out
+	 * all of the sub-ptes for the hugepage range.  So we have
+	 * to give it the first such sub-pte.
+	 */
+	addr &= HPAGE_MASK;
+
+	pgd = pgd_offset(mm, addr);
+	pud = pud_alloc(mm, pgd, addr);
+	if (pud) {
+		pmd = pmd_alloc(mm, pud, addr);
+		if (pmd)
+			pte = pte_alloc_map(mm, NULL, pmd, addr);
+	}
+	return pte;
+}
+
+pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+
+	addr &= HPAGE_MASK;
+
+	pgd = pgd_offset(mm, addr);
+	if (!pgd_none(*pgd)) {
+		pud = pud_offset(pgd, addr);
+		if (!pud_none(*pud)) {
+			pmd = pmd_offset(pud, addr);
+			if (!pmd_none(*pmd))
+				pte = pte_offset_map(pmd, addr);
+		}
+	}
+	return pte;
+}
+
+/* Purge data and instruction TLB entries.  Must be called holding
+ * the pa_tlb_lock.  The TLB purge instructions are slow on SMP
+ * machines since the purge must be broadcast to all CPUs.
+ */
+static inline void purge_tlb_entries_huge(struct mm_struct *mm, unsigned long addr)
+{
+	int i;
+
+	/* We may use multiple physical huge pages (e.g. 2x1 MB) to emulate
+	 * Linux standard huge pages (e.g. 2 MB) */
+	BUILD_BUG_ON(REAL_HPAGE_SHIFT > HPAGE_SHIFT);
+
+	addr &= HPAGE_MASK;
+	addr |= _HUGE_PAGE_SIZE_ENCODING_DEFAULT;
+
+	for (i = 0; i < (1 << (HPAGE_SHIFT-REAL_HPAGE_SHIFT)); i++) {
+		mtsp(mm->context, 1);
+		pdtlb(addr);
+		if (unlikely(split_tlb))
+			pitlb(addr);
+		addr += (1UL << REAL_HPAGE_SHIFT);
+	}
+}
+
+void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+		     pte_t *ptep, pte_t entry)
+{
+	unsigned long addr_start;
+	int i;
+
+	addr &= HPAGE_MASK;
+	addr_start = addr;
+
+	for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) {
+		/* Directly write pte entry.  We could call set_pte_at(mm, addr, ptep, entry)
+		 * instead, but then we get double locking on pa_tlb_lock. */
+		*ptep = entry;
+		ptep++;
+
+		/* Drop the PAGE_SIZE/non-huge tlb entry */
+		purge_tlb_entries(mm, addr);
+
+		addr += PAGE_SIZE;
+		pte_val(entry) += PAGE_SIZE;
+	}
+
+	purge_tlb_entries_huge(mm, addr_start);
+}
+
+
+pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
+			      pte_t *ptep)
+{
+	pte_t entry;
+
+	entry = *ptep;
+	set_huge_pte_at(mm, addr, ptep, __pte(0));
+
+	return entry;
+}
+
+int pmd_huge(pmd_t pmd)
+{
+	return 0;
+}
+
+int pud_huge(pud_t pud)
+{
+	return 0;
+}
diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c
index c5fec48..1b366c4 100644
--- a/arch/parisc/mm/init.c
+++ b/arch/parisc/mm/init.c
@@ -409,15 +409,11 @@ static void __init map_pages(unsigned long start_vaddr,
 	unsigned long vaddr;
 	unsigned long ro_start;
 	unsigned long ro_end;
-	unsigned long fv_addr;
-	unsigned long gw_addr;
-	extern const unsigned long fault_vector_20;
-	extern void * const linux_gateway_page;
+	unsigned long kernel_end;
 
 	ro_start = __pa((unsigned long)_text);
 	ro_end   = __pa((unsigned long)&data_start);
-	fv_addr  = __pa((unsigned long)&fault_vector_20) & PAGE_MASK;
-	gw_addr  = __pa((unsigned long)&linux_gateway_page) & PAGE_MASK;
+	kernel_end  = __pa((unsigned long)&_end);
 
 	end_paddr = start_paddr + size;
 
@@ -475,24 +471,25 @@ static void __init map_pages(unsigned long start_vaddr,
 			for (tmp2 = start_pte; tmp2 < PTRS_PER_PTE; tmp2++, pg_table++) {
 				pte_t pte;
 
-				/*
-				 * Map the fault vector writable so we can
-				 * write the HPMC checksum.
-				 */
 				if (force)
 					pte =  __mk_pte(address, pgprot);
-				else if (parisc_text_address(vaddr) &&
-					 address != fv_addr)
+				else if (parisc_text_address(vaddr)) {
 					pte = __mk_pte(address, PAGE_KERNEL_EXEC);
+					if (address >= ro_start && address < kernel_end)
+						pte = pte_mkhuge(pte);
+				}
 				else
 #if defined(CONFIG_PARISC_PAGE_SIZE_4KB)
-				if (address >= ro_start && address < ro_end
-							&& address != fv_addr
-							&& address != gw_addr)
-					pte = __mk_pte(address, PAGE_KERNEL_RO);
-				else
+				if (address >= ro_start && address < ro_end) {
+					pte = __mk_pte(address, PAGE_KERNEL_EXEC);
+					pte = pte_mkhuge(pte);
+				} else
 #endif
+				{
 					pte = __mk_pte(address, pgprot);
+					if (address >= ro_start && address < kernel_end)
+						pte = pte_mkhuge(pte);
+				}
 
 				if (address >= end_paddr) {
 					if (force)
@@ -536,15 +533,12 @@ void free_initmem(void)
 
 	/* force the kernel to see the new TLB entries */
 	__flush_tlb_range(0, init_begin, init_end);
-	/* Attempt to catch anyone trying to execute code here
-	 * by filling the page with BRK insns.
-	 */
-	memset((void *)init_begin, 0x00, init_end - init_begin);
+
 	/* finally dump all the instructions which were cached, since the
 	 * pages are no-longer executable */
 	flush_icache_range(init_begin, init_end);
 	
-	free_initmem_default(-1);
+	free_initmem_default(POISON_FREE_INITMEM);
 
 	/* set up a new led state on systems shipped LED State panel */
 	pdc_chassis_send_status(PDC_CHASSIS_DIRECT_BCOMPLETE);
@@ -728,8 +722,8 @@ static void __init pagetable_init(void)
 		unsigned long size;
 
 		start_paddr = pmem_ranges[range].start_pfn << PAGE_SHIFT;
-		end_paddr = start_paddr + (pmem_ranges[range].pages << PAGE_SHIFT);
 		size = pmem_ranges[range].pages << PAGE_SHIFT;
+		end_paddr = start_paddr + size;
 
 		map_pages((unsigned long)__va(start_paddr), start_paddr,
 			  size, PAGE_KERNEL, 0);