mm/hugetlb: add more arch-defined huge_pte functions
Gerald Schaefer [Mon, 29 Apr 2013 22:07:23 +0000 (15:07 -0700)]
Commit abf09bed3cce ("s390/mm: implement software dirty bits")
introduced another difference in the pte layout vs.  the pmd layout on
s390, thoroughly breaking the s390 support for hugetlbfs.  This requires
replacing some more pte_xxx functions in mm/hugetlbfs.c with a
huge_pte_xxx version.

This patch introduces those huge_pte_xxx functions and their generic
implementation in asm-generic/hugetlb.h, which will now be included on
all architectures supporting hugetlbfs apart from s390.  This change
will be a no-op for those architectures.

[akpm@linux-foundation.org: fix warning]
Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Hillf Danton <dhillf@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.cz> [for !s390 parts]
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

12 files changed:
arch/ia64/include/asm/hugetlb.h
arch/mips/include/asm/hugetlb.h
arch/powerpc/include/asm/hugetlb.h
arch/s390/include/asm/hugetlb.h
arch/s390/include/asm/pgtable.h
arch/s390/mm/hugetlbpage.c
arch/sh/include/asm/hugetlb.h
arch/sparc/include/asm/hugetlb.h
arch/tile/include/asm/hugetlb.h
arch/x86/include/asm/hugetlb.h
include/asm-generic/hugetlb.h [new file with mode: 0644]
mm/hugetlb.c

index 94eaa5b..aa91005 100644 (file)
@@ -2,6 +2,7 @@
 #define _ASM_IA64_HUGETLB_H
 
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 
 void hugetlb_free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
index ef99db9..fe0d15d 100644 (file)
@@ -10,6 +10,7 @@
 #define __ASM_HUGETLB_H
 
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 
 static inline int is_hugepage_only_range(struct mm_struct *mm,
index 62e11a3..4fcbd6b 100644 (file)
@@ -3,6 +3,7 @@
 
 #ifdef CONFIG_HUGETLB_PAGE
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 extern struct kmem_cache *hugepte_cache;
 
index 593753e..bd90359 100644 (file)
@@ -114,7 +114,7 @@ static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
 #define huge_ptep_set_wrprotect(__mm, __addr, __ptep)                  \
 ({                                                                     \
        pte_t __pte = huge_ptep_get(__ptep);                            \
-       if (pte_write(__pte)) {                                         \
+       if (huge_pte_write(__pte)) {                                    \
                huge_ptep_invalidate(__mm, __addr, __ptep);             \
                set_huge_pte_at(__mm, __addr, __ptep,                   \
                                huge_pte_wrprotect(__pte));             \
@@ -127,4 +127,58 @@ static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
        huge_ptep_invalidate(vma->vm_mm, address, ptep);
 }
 
+static inline pte_t mk_huge_pte(struct page *page, pgprot_t pgprot)
+{
+       pte_t pte;
+       pmd_t pmd;
+
+       pmd = mk_pmd_phys(page_to_phys(page), pgprot);
+       pte_val(pte) = pmd_val(pmd);
+       return pte;
+}
+
+static inline int huge_pte_write(pte_t pte)
+{
+       pmd_t pmd;
+
+       pmd_val(pmd) = pte_val(pte);
+       return pmd_write(pmd);
+}
+
+static inline int huge_pte_dirty(pte_t pte)
+{
+       /* No dirty bit in the segment table entry. */
+       return 0;
+}
+
+static inline pte_t huge_pte_mkwrite(pte_t pte)
+{
+       pmd_t pmd;
+
+       pmd_val(pmd) = pte_val(pte);
+       pte_val(pte) = pmd_val(pmd_mkwrite(pmd));
+       return pte;
+}
+
+static inline pte_t huge_pte_mkdirty(pte_t pte)
+{
+       /* No dirty bit in the segment table entry. */
+       return pte;
+}
+
+static inline pte_t huge_pte_modify(pte_t pte, pgprot_t newprot)
+{
+       pmd_t pmd;
+
+       pmd_val(pmd) = pte_val(pte);
+       pte_val(pte) = pmd_val(pmd_modify(pmd, newprot));
+       return pte;
+}
+
+static inline void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
+                                 pte_t *ptep)
+{
+       pmd_clear((pmd_t *) ptep);
+}
+
 #endif /* _ASM_S390_HUGETLB_H */
index 4a64c0e..b462291 100644 (file)
@@ -424,6 +424,13 @@ extern unsigned long MODULES_END;
 #define __S110 PAGE_RW
 #define __S111 PAGE_RW
 
+/*
+ * Segment entry (large page) protection definitions.
+ */
+#define SEGMENT_NONE   __pgprot(_HPAGE_TYPE_NONE)
+#define SEGMENT_RO     __pgprot(_HPAGE_TYPE_RO)
+#define SEGMENT_RW     __pgprot(_HPAGE_TYPE_RW)
+
 static inline int mm_exclusive(struct mm_struct *mm)
 {
        return likely(mm == current->active_mm &&
@@ -914,26 +921,6 @@ static inline pte_t pte_mkspecial(pte_t pte)
 #ifdef CONFIG_HUGETLB_PAGE
 static inline pte_t pte_mkhuge(pte_t pte)
 {
-       /*
-        * PROT_NONE needs to be remapped from the pte type to the ste type.
-        * The HW invalid bit is also different for pte and ste. The pte
-        * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE
-        * bit, so we don't have to clear it.
-        */
-       if (pte_val(pte) & _PAGE_INVALID) {
-               if (pte_val(pte) & _PAGE_SWT)
-                       pte_val(pte) |= _HPAGE_TYPE_NONE;
-               pte_val(pte) |= _SEGMENT_ENTRY_INV;
-       }
-       /*
-        * Clear SW pte bits, there are no SW bits in a segment table entry.
-        */
-       pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX | _PAGE_SWC |
-                         _PAGE_SWR | _PAGE_SWW);
-       /*
-        * Also set the change-override bit because we don't need dirty bit
-        * tracking for hugetlbfs pages.
-        */
        pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
        return pte;
 }
@@ -1278,31 +1265,7 @@ static inline void __pmd_idte(unsigned long address, pmd_t *pmdp)
        }
 }
 
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
-
-#define SEGMENT_NONE   __pgprot(_HPAGE_TYPE_NONE)
-#define SEGMENT_RO     __pgprot(_HPAGE_TYPE_RO)
-#define SEGMENT_RW     __pgprot(_HPAGE_TYPE_RW)
-
-#define __HAVE_ARCH_PGTABLE_DEPOSIT
-extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
-
-#define __HAVE_ARCH_PGTABLE_WITHDRAW
-extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm);
-
-static inline int pmd_trans_splitting(pmd_t pmd)
-{
-       return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
-}
-
-static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
-                             pmd_t *pmdp, pmd_t entry)
-{
-       if (!(pmd_val(entry) & _SEGMENT_ENTRY_INV) && MACHINE_HAS_EDAT1)
-               pmd_val(entry) |= _SEGMENT_ENTRY_CO;
-       *pmdp = entry;
-}
-
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
 {
        /*
@@ -1323,10 +1286,11 @@ static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
        return pmd;
 }
 
-static inline pmd_t pmd_mkhuge(pmd_t pmd)
+static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
 {
-       pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
-       return pmd;
+       pmd_t __pmd;
+       pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
+       return __pmd;
 }
 
 static inline pmd_t pmd_mkwrite(pmd_t pmd)
@@ -1336,6 +1300,34 @@ static inline pmd_t pmd_mkwrite(pmd_t pmd)
                pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO;
        return pmd;
 }
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+
+#define __HAVE_ARCH_PGTABLE_DEPOSIT
+extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
+
+#define __HAVE_ARCH_PGTABLE_WITHDRAW
+extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm);
+
+static inline int pmd_trans_splitting(pmd_t pmd)
+{
+       return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
+}
+
+static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
+                             pmd_t *pmdp, pmd_t entry)
+{
+       if (!(pmd_val(entry) & _SEGMENT_ENTRY_INV) && MACHINE_HAS_EDAT1)
+               pmd_val(entry) |= _SEGMENT_ENTRY_CO;
+       *pmdp = entry;
+}
+
+static inline pmd_t pmd_mkhuge(pmd_t pmd)
+{
+       pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
+       return pmd;
+}
 
 static inline pmd_t pmd_wrprotect(pmd_t pmd)
 {
@@ -1432,13 +1424,6 @@ static inline void pmdp_set_wrprotect(struct mm_struct *mm,
        }
 }
 
-static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
-{
-       pmd_t __pmd;
-       pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
-       return __pmd;
-}
-
 #define pfn_pmd(pfn, pgprot)   mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
 #define mk_pmd(page, pgprot)   pfn_pmd(page_to_pfn(page), (pgprot))
 
index 532525e..121089d 100644 (file)
@@ -39,7 +39,7 @@ int arch_prepare_hugepage(struct page *page)
        if (!ptep)
                return -ENOMEM;
 
-       pte = mk_pte(page, PAGE_RW);
+       pte_val(pte) = addr;
        for (i = 0; i < PTRS_PER_PTE; i++) {
                set_pte_at(&init_mm, addr + i * PAGE_SIZE, ptep + i, pte);
                pte_val(pte) += PAGE_SIZE;
index b3808c7..699255d 100644 (file)
@@ -3,6 +3,7 @@
 
 #include <asm/cacheflush.h>
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 
 static inline int is_hugepage_only_range(struct mm_struct *mm,
index 7eb57d2..e4cab46 100644 (file)
@@ -2,6 +2,7 @@
 #define _ASM_SPARC64_HUGETLB_H
 
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 
 void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
index 0f885af..3257733 100644 (file)
@@ -16,6 +16,7 @@
 #define _ASM_TILE_HUGETLB_H
 
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 
 static inline int is_hugepage_only_range(struct mm_struct *mm,
index bdd35db..a809121 100644 (file)
@@ -2,6 +2,7 @@
 #define _ASM_X86_HUGETLB_H
 
 #include <asm/page.h>
+#include <asm-generic/hugetlb.h>
 
 
 static inline int is_hugepage_only_range(struct mm_struct *mm,
diff --git a/include/asm-generic/hugetlb.h b/include/asm-generic/hugetlb.h
new file mode 100644 (file)
index 0000000..d06079c
--- /dev/null
@@ -0,0 +1,40 @@
+#ifndef _ASM_GENERIC_HUGETLB_H
+#define _ASM_GENERIC_HUGETLB_H
+
+static inline pte_t mk_huge_pte(struct page *page, pgprot_t pgprot)
+{
+       return mk_pte(page, pgprot);
+}
+
+static inline int huge_pte_write(pte_t pte)
+{
+       return pte_write(pte);
+}
+
+static inline int huge_pte_dirty(pte_t pte)
+{
+       return pte_dirty(pte);
+}
+
+static inline pte_t huge_pte_mkwrite(pte_t pte)
+{
+       return pte_mkwrite(pte);
+}
+
+static inline pte_t huge_pte_mkdirty(pte_t pte)
+{
+       return pte_mkdirty(pte);
+}
+
+static inline pte_t huge_pte_modify(pte_t pte, pgprot_t newprot)
+{
+       return pte_modify(pte, newprot);
+}
+
+static inline void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
+                                 pte_t *ptep)
+{
+       pte_clear(mm, addr, ptep);
+}
+
+#endif /* _ASM_GENERIC_HUGETLB_H */
index 1a12f5b..73b864a 100644 (file)
@@ -2247,10 +2247,11 @@ static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
        pte_t entry;
 
        if (writable) {
-               entry =
-                   pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
+               entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
+                                        vma->vm_page_prot)));
        } else {
-               entry = huge_pte_wrprotect(mk_pte(page, vma->vm_page_prot));
+               entry = huge_pte_wrprotect(mk_huge_pte(page,
+                                          vma->vm_page_prot));
        }
        entry = pte_mkyoung(entry);
        entry = pte_mkhuge(entry);
@@ -2264,7 +2265,7 @@ static void set_huge_ptep_writable(struct vm_area_struct *vma,
 {
        pte_t entry;
 
-       entry = pte_mkwrite(pte_mkdirty(huge_ptep_get(ptep)));
+       entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
        if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
                update_mmu_cache(vma, address, ptep);
 }
@@ -2379,7 +2380,7 @@ again:
                 * HWPoisoned hugepage is already unmapped and dropped reference
                 */
                if (unlikely(is_hugetlb_entry_hwpoisoned(pte))) {
-                       pte_clear(mm, address, ptep);
+                       huge_pte_clear(mm, address, ptep);
                        continue;
                }
 
@@ -2403,7 +2404,7 @@ again:
 
                pte = huge_ptep_get_and_clear(mm, address, ptep);
                tlb_remove_tlb_entry(tlb, ptep, address);
-               if (pte_dirty(pte))
+               if (huge_pte_dirty(pte))
                        set_page_dirty(page);
 
                page_remove_rmap(page);
@@ -2856,7 +2857,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
         * page now as it is used to determine if a reservation has been
         * consumed.
         */
-       if ((flags & FAULT_FLAG_WRITE) && !pte_write(entry)) {
+       if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
                if (vma_needs_reservation(h, vma, address) < 0) {
                        ret = VM_FAULT_OOM;
                        goto out_mutex;
@@ -2886,12 +2887,12 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 
 
        if (flags & FAULT_FLAG_WRITE) {
-               if (!pte_write(entry)) {
+               if (!huge_pte_write(entry)) {
                        ret = hugetlb_cow(mm, vma, address, ptep, entry,
                                                        pagecache_page);
                        goto out_page_table_lock;
                }
-               entry = pte_mkdirty(entry);
+               entry = huge_pte_mkdirty(entry);
        }
        entry = pte_mkyoung(entry);
        if (huge_ptep_set_access_flags(vma, address, ptep, entry,
@@ -2972,7 +2973,8 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
                 * directly from any kind of swap entries.
                 */
                if (absent || is_swap_pte(huge_ptep_get(pte)) ||
-                   ((flags & FOLL_WRITE) && !pte_write(huge_ptep_get(pte)))) {
+                   ((flags & FOLL_WRITE) &&
+                     !huge_pte_write(huge_ptep_get(pte)))) {
                        int ret;
 
                        spin_unlock(&mm->page_table_lock);
@@ -3042,7 +3044,7 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
                }
                if (!huge_pte_none(huge_ptep_get(ptep))) {
                        pte = huge_ptep_get_and_clear(mm, address, ptep);
-                       pte = pte_mkhuge(pte_modify(pte, newprot));
+                       pte = pte_mkhuge(huge_pte_modify(pte, newprot));
                        pte = arch_make_huge_pte(pte, vma, NULL, 0);
                        set_huge_pte_at(mm, address, ptep, pte);
                        pages++;