Commit 1dc954bd2f85144b9a2959139d41a5f6b9b34712

Authored by Aneesh Kumar K.V
Committed by Benjamin Herrenschmidt
1 parent 9000c17dc0

powerpc/mm: NUMA pte should be handled via slow path in get_user_pages_fast()

We need to handle numa pte via the slow path

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Showing 1 changed file with 13 additions and 0 deletions Side-by-side Diff

arch/powerpc/mm/gup.c
... ... @@ -36,6 +36,11 @@
36 36 do {
37 37 pte_t pte = ACCESS_ONCE(*ptep);
38 38 struct page *page;
  39 + /*
  40 + * Similar to the PMD case, NUMA hinting must take slow path
  41 + */
  42 + if (pte_numa(pte))
  43 + return 0;
39 44  
40 45 if ((pte_val(pte) & mask) != result)
41 46 return 0;
... ... @@ -75,6 +80,14 @@
75 80 if (pmd_none(pmd) || pmd_trans_splitting(pmd))
76 81 return 0;
77 82 if (pmd_huge(pmd) || pmd_large(pmd)) {
  83 + /*
  84 + * NUMA hinting faults need to be handled in the GUP
  85 + * slowpath for accounting purposes and so that they
  86 + * can be serialised against THP migration.
  87 + */
  88 + if (pmd_numa(pmd))
  89 + return 0;
  90 +
78 91 if (!gup_hugepte((pte_t *)pmdp, PMD_SIZE, addr, next,
79 92 write, pages, nr))
80 93 return 0;