thp: keep huge zero page pinned until tlb flush
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Thu, 28 Apr 2016 23:18:27 +0000 (16:18 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 29 Apr 2016 02:34:04 +0000 (19:34 -0700)
Andrea has found[1] a race condition on MMU-gather based TLB flush vs
split_huge_page() or shrinker which frees huge zero under us (patch 1/2
and 2/2 respectively).

With new THP refcounting, we don't need patch 1/2: mmu_gather keeps the
page pinned until flush is complete and the pin prevents the page from
being split under us.

We still need patch 2/2.  This is simplified version of Andrea's patch.
We don't need fancy encoding.

[1] http://lkml.kernel.org/r/1447938052-22165-1-git-send-email-aarcange@redhat.com

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/huge_mm.h
mm/huge_memory.c
mm/swap.c

index 7008623e24b19bfef175d32773fa5c9855d36e76..d7b9e5346fba0390890a4e477c8952485319e887 100644 (file)
@@ -152,6 +152,7 @@ static inline bool is_huge_zero_pmd(pmd_t pmd)
 }
 
 struct page *get_huge_zero_page(void);
+void put_huge_zero_page(void);
 
 #else /* CONFIG_TRANSPARENT_HUGEPAGE */
 #define HPAGE_PMD_SHIFT ({ BUILD_BUG(); 0; })
@@ -208,6 +209,10 @@ static inline bool is_huge_zero_page(struct page *page)
        return false;
 }
 
+static inline void put_huge_zero_page(void)
+{
+       BUILD_BUG();
+}
 
 static inline struct page *follow_devmap_pmd(struct vm_area_struct *vma,
                unsigned long addr, pmd_t *pmd, int flags)
index 86f9f8b82f8ecfc47b92554e3a82501346b0fcb8..5346de05f471f4060c618ca07b3406c9eae76245 100644 (file)
@@ -232,7 +232,7 @@ retry:
        return READ_ONCE(huge_zero_page);
 }
 
-static void put_huge_zero_page(void)
+void put_huge_zero_page(void)
 {
        /*
         * Counter should never go to zero here. Only shrinker can put
@@ -1684,12 +1684,12 @@ int zap_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma,
        if (vma_is_dax(vma)) {
                spin_unlock(ptl);
                if (is_huge_zero_pmd(orig_pmd))
-                       put_huge_zero_page();
+                       tlb_remove_page(tlb, pmd_page(orig_pmd));
        } else if (is_huge_zero_pmd(orig_pmd)) {
                pte_free(tlb->mm, pgtable_trans_huge_withdraw(tlb->mm, pmd));
                atomic_long_dec(&tlb->mm->nr_ptes);
                spin_unlock(ptl);
-               put_huge_zero_page();
+               tlb_remove_page(tlb, pmd_page(orig_pmd));
        } else {
                struct page *page = pmd_page(orig_pmd);
                page_remove_rmap(page, true);
index a0bc206b4ac6705e839bd8e7164d522ea454a5e5..03aacbcb013f2be47efbb15d78387c8b368e2604 100644 (file)
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -728,6 +728,11 @@ void release_pages(struct page **pages, int nr, bool cold)
                        zone = NULL;
                }
 
+               if (is_huge_zero_page(page)) {
+                       put_huge_zero_page();
+                       continue;
+               }
+
                page = compound_head(page);
                if (!put_page_testzero(page))
                        continue;