mm, hugetlb: convert hugetlbfs to use split pmd lock
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Thu, 14 Nov 2013 22:31:02 +0000 (14:31 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 15 Nov 2013 00:32:14 +0000 (09:32 +0900)
Hugetlb supports multiple page sizes. We use split lock only for PMD
level, but not for PUD.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Tested-by: Alex Thorlton <athorlton@sgi.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "Eric W . Biederman" <ebiederm@xmission.com>
Cc: "Paul E . McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Jones <davej@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Robin Holt <robinmholt@gmail.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/meminfo.c
include/linux/hugetlb.h
include/linux/swapops.h
mm/hugetlb.c
mm/mempolicy.c
mm/migrate.c
mm/rmap.c

index c805d5b69ba125a5ab57cee445520cc6b81c8522..a77d2b2991998187843271a714ab241adacfc0b5 100644 (file)
@@ -1,8 +1,8 @@
 #include <linux/fs.h>
-#include <linux/hugetlb.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/mm.h>
+#include <linux/hugetlb.h>
 #include <linux/mman.h>
 #include <linux/mmzone.h>
 #include <linux/proc_fs.h>
index 0393270466c3fd8a59d24781cf54ab5ffd0c5a57..acd2010328f3044b1cf6dd788e4ac09ab9262925 100644 (file)
@@ -392,6 +392,15 @@ static inline int hugepage_migration_support(struct hstate *h)
        return pmd_huge_support() && (huge_page_shift(h) == PMD_SHIFT);
 }
 
+static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
+                                          struct mm_struct *mm, pte_t *pte)
+{
+       if (huge_page_size(h) == PMD_SIZE)
+               return pmd_lockptr(mm, (pmd_t *) pte);
+       VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
+       return &mm->page_table_lock;
+}
+
 #else  /* CONFIG_HUGETLB_PAGE */
 struct hstate {};
 #define alloc_huge_page_node(h, nid) NULL
@@ -401,6 +410,7 @@ struct hstate {};
 #define hstate_sizelog(s) NULL
 #define hstate_vma(v) NULL
 #define hstate_inode(i) NULL
+#define page_hstate(page) NULL
 #define huge_page_size(h) PAGE_SIZE
 #define huge_page_mask(h) PAGE_MASK
 #define vma_kernel_pagesize(v) PAGE_SIZE
@@ -421,6 +431,22 @@ static inline pgoff_t basepage_index(struct page *page)
 #define dissolve_free_huge_pages(s, e) do {} while (0)
 #define pmd_huge_support()     0
 #define hugepage_migration_support(h)  0
+
+static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
+                                          struct mm_struct *mm, pte_t *pte)
+{
+       return &mm->page_table_lock;
+}
 #endif /* CONFIG_HUGETLB_PAGE */
 
+static inline spinlock_t *huge_pte_lock(struct hstate *h,
+                                       struct mm_struct *mm, pte_t *pte)
+{
+       spinlock_t *ptl;
+
+       ptl = huge_pte_lockptr(h, mm, pte);
+       spin_lock(ptl);
+       return ptl;
+}
+
 #endif /* _LINUX_HUGETLB_H */
index 8d4fa82bfb913d0021d93c49e71dab940804fdf7..c0f75261a728bf18b2ab930dbbee6bd35acf571c 100644 (file)
@@ -139,7 +139,8 @@ static inline void make_migration_entry_read(swp_entry_t *entry)
 
 extern void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
                                        unsigned long address);
-extern void migration_entry_wait_huge(struct mm_struct *mm, pte_t *pte);
+extern void migration_entry_wait_huge(struct vm_area_struct *vma,
+               struct mm_struct *mm, pte_t *pte);
 #else
 
 #define make_migration_entry(page, write) swp_entry(0, 0)
@@ -151,8 +152,8 @@ static inline int is_migration_entry(swp_entry_t swp)
 static inline void make_migration_entry_read(swp_entry_t *entryp) { }
 static inline void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
                                         unsigned long address) { }
-static inline void migration_entry_wait_huge(struct mm_struct *mm,
-                                       pte_t *pte) { }
+static inline void migration_entry_wait_huge(struct vm_area_struct *vma,
+               struct mm_struct *mm, pte_t *pte) { }
 static inline int is_write_migration_entry(swp_entry_t entry)
 {
        return 0;
index 0b7656e804d126cf0fcfa04a8b427396af86deb1..7d57af21f49e920776979dfe11023780ebff6afb 100644 (file)
@@ -2376,6 +2376,7 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
        cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
 
        for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
+               spinlock_t *src_ptl, *dst_ptl;
                src_pte = huge_pte_offset(src, addr);
                if (!src_pte)
                        continue;
@@ -2387,8 +2388,9 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
                if (dst_pte == src_pte)
                        continue;
 
-               spin_lock(&dst->page_table_lock);
-               spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING);
+               dst_ptl = huge_pte_lock(h, dst, dst_pte);
+               src_ptl = huge_pte_lockptr(h, src, src_pte);
+               spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
                if (!huge_pte_none(huge_ptep_get(src_pte))) {
                        if (cow)
                                huge_ptep_set_wrprotect(src, addr, src_pte);
@@ -2398,8 +2400,8 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
                        page_dup_rmap(ptepage);
                        set_huge_pte_at(dst, addr, dst_pte, entry);
                }
-               spin_unlock(&src->page_table_lock);
-               spin_unlock(&dst->page_table_lock);
+               spin_unlock(src_ptl);
+               spin_unlock(dst_ptl);
        }
        return 0;
 
@@ -2442,6 +2444,7 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
        unsigned long address;
        pte_t *ptep;
        pte_t pte;
+       spinlock_t *ptl;
        struct page *page;
        struct hstate *h = hstate_vma(vma);
        unsigned long sz = huge_page_size(h);
@@ -2455,25 +2458,25 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
        tlb_start_vma(tlb, vma);
        mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
 again:
-       spin_lock(&mm->page_table_lock);
        for (address = start; address < end; address += sz) {
                ptep = huge_pte_offset(mm, address);
                if (!ptep)
                        continue;
 
+               ptl = huge_pte_lock(h, mm, ptep);
                if (huge_pmd_unshare(mm, &address, ptep))
-                       continue;
+                       goto unlock;
 
                pte = huge_ptep_get(ptep);
                if (huge_pte_none(pte))
-                       continue;
+                       goto unlock;
 
                /*
                 * HWPoisoned hugepage is already unmapped and dropped reference
                 */
                if (unlikely(is_hugetlb_entry_hwpoisoned(pte))) {
                        huge_pte_clear(mm, address, ptep);
-                       continue;
+                       goto unlock;
                }
 
                page = pte_page(pte);
@@ -2484,7 +2487,7 @@ again:
                 */
                if (ref_page) {
                        if (page != ref_page)
-                               continue;
+                               goto unlock;
 
                        /*
                         * Mark the VMA as having unmapped its page so that
@@ -2501,13 +2504,18 @@ again:
 
                page_remove_rmap(page);
                force_flush = !__tlb_remove_page(tlb, page);
-               if (force_flush)
+               if (force_flush) {
+                       spin_unlock(ptl);
                        break;
+               }
                /* Bail out after unmapping reference page if supplied */
-               if (ref_page)
+               if (ref_page) {
+                       spin_unlock(ptl);
                        break;
+               }
+unlock:
+               spin_unlock(ptl);
        }
-       spin_unlock(&mm->page_table_lock);
        /*
         * mmu_gather ran out of room to batch pages, we break out of
         * the PTE lock to avoid doing the potential expensive TLB invalidate
@@ -2613,7 +2621,7 @@ static int unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
  */
 static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
                        unsigned long address, pte_t *ptep, pte_t pte,
-                       struct page *pagecache_page)
+                       struct page *pagecache_page, spinlock_t *ptl)
 {
        struct hstate *h = hstate_vma(vma);
        struct page *old_page, *new_page;
@@ -2647,8 +2655,8 @@ retry_avoidcopy:
 
        page_cache_get(old_page);
 
-       /* Drop page_table_lock as buddy allocator may be called */
-       spin_unlock(&mm->page_table_lock);
+       /* Drop page table lock as buddy allocator may be called */
+       spin_unlock(ptl);
        new_page = alloc_huge_page(vma, address, outside_reserve);
 
        if (IS_ERR(new_page)) {
@@ -2666,13 +2674,13 @@ retry_avoidcopy:
                        BUG_ON(huge_pte_none(pte));
                        if (unmap_ref_private(mm, vma, old_page, address)) {
                                BUG_ON(huge_pte_none(pte));
-                               spin_lock(&mm->page_table_lock);
+                               spin_lock(ptl);
                                ptep = huge_pte_offset(mm, address & huge_page_mask(h));
                                if (likely(pte_same(huge_ptep_get(ptep), pte)))
                                        goto retry_avoidcopy;
                                /*
-                                * race occurs while re-acquiring page_table_lock, and
-                                * our job is done.
+                                * race occurs while re-acquiring page table
+                                * lock, and our job is done.
                                 */
                                return 0;
                        }
@@ -2680,7 +2688,7 @@ retry_avoidcopy:
                }
 
                /* Caller expects lock to be held */
-               spin_lock(&mm->page_table_lock);
+               spin_lock(ptl);
                if (err == -ENOMEM)
                        return VM_FAULT_OOM;
                else
@@ -2695,7 +2703,7 @@ retry_avoidcopy:
                page_cache_release(new_page);
                page_cache_release(old_page);
                /* Caller expects lock to be held */
-               spin_lock(&mm->page_table_lock);
+               spin_lock(ptl);
                return VM_FAULT_OOM;
        }
 
@@ -2707,10 +2715,10 @@ retry_avoidcopy:
        mmun_end = mmun_start + huge_page_size(h);
        mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
        /*
-        * Retake the page_table_lock to check for racing updates
+        * Retake the page table lock to check for racing updates
         * before the page tables are altered
         */
-       spin_lock(&mm->page_table_lock);
+       spin_lock(ptl);
        ptep = huge_pte_offset(mm, address & huge_page_mask(h));
        if (likely(pte_same(huge_ptep_get(ptep), pte))) {
                ClearPagePrivate(new_page);
@@ -2724,13 +2732,13 @@ retry_avoidcopy:
                /* Make the old page be freed below */
                new_page = old_page;
        }
-       spin_unlock(&mm->page_table_lock);
+       spin_unlock(ptl);
        mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
        page_cache_release(new_page);
        page_cache_release(old_page);
 
        /* Caller expects lock to be held */
-       spin_lock(&mm->page_table_lock);
+       spin_lock(ptl);
        return 0;
 }
 
@@ -2778,6 +2786,7 @@ static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
        struct page *page;
        struct address_space *mapping;
        pte_t new_pte;
+       spinlock_t *ptl;
 
        /*
         * Currently, we are forced to kill the process in the event the
@@ -2864,7 +2873,8 @@ retry:
                        goto backout_unlocked;
                }
 
-       spin_lock(&mm->page_table_lock);
+       ptl = huge_pte_lockptr(h, mm, ptep);
+       spin_lock(ptl);
        size = i_size_read(mapping->host) >> huge_page_shift(h);
        if (idx >= size)
                goto backout;
@@ -2885,16 +2895,16 @@ retry:
 
        if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
                /* Optimization, do the COW without a second fault */
-               ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page);
+               ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page, ptl);
        }
 
-       spin_unlock(&mm->page_table_lock);
+       spin_unlock(ptl);
        unlock_page(page);
 out:
        return ret;
 
 backout:
-       spin_unlock(&mm->page_table_lock);
+       spin_unlock(ptl);
 backout_unlocked:
        unlock_page(page);
        put_page(page);
@@ -2906,6 +2916,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 {
        pte_t *ptep;
        pte_t entry;
+       spinlock_t *ptl;
        int ret;
        struct page *page = NULL;
        struct page *pagecache_page = NULL;
@@ -2918,7 +2929,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
        if (ptep) {
                entry = huge_ptep_get(ptep);
                if (unlikely(is_hugetlb_entry_migration(entry))) {
-                       migration_entry_wait_huge(mm, ptep);
+                       migration_entry_wait_huge(vma, mm, ptep);
                        return 0;
                } else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
                        return VM_FAULT_HWPOISON_LARGE |
@@ -2974,17 +2985,18 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
        if (page != pagecache_page)
                lock_page(page);
 
-       spin_lock(&mm->page_table_lock);
+       ptl = huge_pte_lockptr(h, mm, ptep);
+       spin_lock(ptl);
        /* Check for a racing update before calling hugetlb_cow */
        if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
-               goto out_page_table_lock;
+               goto out_ptl;
 
 
        if (flags & FAULT_FLAG_WRITE) {
                if (!huge_pte_write(entry)) {
                        ret = hugetlb_cow(mm, vma, address, ptep, entry,
-                                                       pagecache_page);
-                       goto out_page_table_lock;
+                                       pagecache_page, ptl);
+                       goto out_ptl;
                }
                entry = huge_pte_mkdirty(entry);
        }
@@ -2993,8 +3005,8 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
                                                flags & FAULT_FLAG_WRITE))
                update_mmu_cache(vma, address, ptep);
 
-out_page_table_lock:
-       spin_unlock(&mm->page_table_lock);
+out_ptl:
+       spin_unlock(ptl);
 
        if (pagecache_page) {
                unlock_page(pagecache_page);
@@ -3020,9 +3032,9 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
        unsigned long remainder = *nr_pages;
        struct hstate *h = hstate_vma(vma);
 
-       spin_lock(&mm->page_table_lock);
        while (vaddr < vma->vm_end && remainder) {
                pte_t *pte;
+               spinlock_t *ptl = NULL;
                int absent;
                struct page *page;
 
@@ -3030,8 +3042,12 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
                 * Some archs (sparc64, sh*) have multiple pte_ts to
                 * each hugepage.  We have to make sure we get the
                 * first, for the page indexing below to work.
+                *
+                * Note that page table lock is not held when pte is null.
                 */
                pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
+               if (pte)
+                       ptl = huge_pte_lock(h, mm, pte);
                absent = !pte || huge_pte_none(huge_ptep_get(pte));
 
                /*
@@ -3043,6 +3059,8 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
                 */
                if (absent && (flags & FOLL_DUMP) &&
                    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
+                       if (pte)
+                               spin_unlock(ptl);
                        remainder = 0;
                        break;
                }
@@ -3062,10 +3080,10 @@ long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
                      !huge_pte_write(huge_ptep_get(pte)))) {
                        int ret;
 
-                       spin_unlock(&mm->page_table_lock);
+                       if (pte)
+                               spin_unlock(ptl);
                        ret = hugetlb_fault(mm, vma, vaddr,
                                (flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
-                       spin_lock(&mm->page_table_lock);
                        if (!(ret & VM_FAULT_ERROR))
                                continue;
 
@@ -3096,8 +3114,8 @@ same_page:
                         */
                        goto same_page;
                }
+               spin_unlock(ptl);
        }
-       spin_unlock(&mm->page_table_lock);
        *nr_pages = remainder;
        *position = vaddr;
 
@@ -3118,13 +3136,15 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
        flush_cache_range(vma, address, end);
 
        mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
-       spin_lock(&mm->page_table_lock);
        for (; address < end; address += huge_page_size(h)) {
+               spinlock_t *ptl;
                ptep = huge_pte_offset(mm, address);
                if (!ptep)
                        continue;
+               ptl = huge_pte_lock(h, mm, ptep);
                if (huge_pmd_unshare(mm, &address, ptep)) {
                        pages++;
+                       spin_unlock(ptl);
                        continue;
                }
                if (!huge_pte_none(huge_ptep_get(ptep))) {
@@ -3134,8 +3154,8 @@ unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
                        set_huge_pte_at(mm, address, ptep, pte);
                        pages++;
                }
+               spin_unlock(ptl);
        }
-       spin_unlock(&mm->page_table_lock);
        /*
         * Must flush TLB before releasing i_mmap_mutex: x86's huge_pmd_unshare
         * may have cleared our pud entry and done put_page on the page table:
@@ -3298,6 +3318,7 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
        unsigned long saddr;
        pte_t *spte = NULL;
        pte_t *pte;
+       spinlock_t *ptl;
 
        if (!vma_shareable(vma, addr))
                return (pte_t *)pmd_alloc(mm, pud, addr);
@@ -3320,13 +3341,14 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
        if (!spte)
                goto out;
 
-       spin_lock(&mm->page_table_lock);
+       ptl = huge_pte_lockptr(hstate_vma(vma), mm, spte);
+       spin_lock(ptl);
        if (pud_none(*pud))
                pud_populate(mm, pud,
                                (pmd_t *)((unsigned long)spte & PAGE_MASK));
        else
                put_page(virt_to_page(spte));
-       spin_unlock(&mm->page_table_lock);
+       spin_unlock(ptl);
 out:
        pte = (pte_t *)pmd_alloc(mm, pud, addr);
        mutex_unlock(&mapping->i_mmap_mutex);
@@ -3340,7 +3362,7 @@ out:
  * indicated by page_count > 1, unmap is achieved by clearing pud and
  * decrementing the ref count. If count == 1, the pte page is not shared.
  *
- * called with vma->vm_mm->page_table_lock held.
+ * called with page table lock held.
  *
  * returns: 1 successfully unmapped a shared pte page
  *         0 the underlying pte page is not shared, or it is the last user
index 4cc19f6ab6c6f84da6bb99955f5026a9bc066a76..c4403cdf3433ddc79515dac25e5eedd0b39fa320 100644 (file)
@@ -525,8 +525,9 @@ static void queue_pages_hugetlb_pmd_range(struct vm_area_struct *vma,
 #ifdef CONFIG_HUGETLB_PAGE
        int nid;
        struct page *page;
+       spinlock_t *ptl;
 
-       spin_lock(&vma->vm_mm->page_table_lock);
+       ptl = huge_pte_lock(hstate_vma(vma), vma->vm_mm, (pte_t *)pmd);
        page = pte_page(huge_ptep_get((pte_t *)pmd));
        nid = page_to_nid(page);
        if (node_isset(nid, *nodes) == !!(flags & MPOL_MF_INVERT))
@@ -536,7 +537,7 @@ static void queue_pages_hugetlb_pmd_range(struct vm_area_struct *vma,
            (flags & MPOL_MF_MOVE && page_mapcount(page) == 1))
                isolate_huge_page(page, private);
 unlock:
-       spin_unlock(&vma->vm_mm->page_table_lock);
+       spin_unlock(ptl);
 #else
        BUG();
 #endif
index dfc8300ecbb273fe3d7f5ef37e21e20a63b58c84..01f45cefa4cdf465c5b085fe9c4eaa3d3508fbdc 100644 (file)
@@ -130,7 +130,7 @@ static int remove_migration_pte(struct page *new, struct vm_area_struct *vma,
                ptep = huge_pte_offset(mm, addr);
                if (!ptep)
                        goto out;
-               ptl = &mm->page_table_lock;
+               ptl = huge_pte_lockptr(hstate_vma(vma), mm, ptep);
        } else {
                pmd = mm_find_pmd(mm, addr);
                if (!pmd)
@@ -249,9 +249,10 @@ void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
        __migration_entry_wait(mm, ptep, ptl);
 }
 
-void migration_entry_wait_huge(struct mm_struct *mm, pte_t *pte)
+void migration_entry_wait_huge(struct vm_area_struct *vma,
+               struct mm_struct *mm, pte_t *pte)
 {
-       spinlock_t *ptl = &(mm)->page_table_lock;
+       spinlock_t *ptl = huge_pte_lockptr(hstate_vma(vma), mm, pte);
        __migration_entry_wait(mm, pte, ptl);
 }
 
index b59d741dcf65f255f7a0a835d51b845d36d5632f..55c8b8dc9ffb0c349eb63ad20a8d7bbcc2e9b25d 100644 (file)
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -601,7 +601,7 @@ pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
 
        if (unlikely(PageHuge(page))) {
                pte = huge_pte_offset(mm, address);
-               ptl = &mm->page_table_lock;
+               ptl = huge_pte_lockptr(page_hstate(page), mm, pte);
                goto check;
        }