releases/5.0.18/mm-hugetlb.c-don-t-put_page-in-lock-of-hugetlb_lock.patch

   1 From 2bf753e64b4a702e27ce26ff520c59563c62f96b Mon Sep 17 00:00:00 2001
   2 From: Kai Shen <shenkai8@huawei.com>
   3 Date: Mon, 13 May 2019 17:15:37 -0700
   4 Subject: mm/hugetlb.c: don't put_page in lock of hugetlb_lock
   5
   6 From: Kai Shen <shenkai8@huawei.com>
   7
   8 commit 2bf753e64b4a702e27ce26ff520c59563c62f96b upstream.
   9
  10 spinlock recursion happened when do LTP test:
  11 #!/bin/bash
  12 ./runltp -p -f hugetlb &
  13 ./runltp -p -f hugetlb &
  14 ./runltp -p -f hugetlb &
  15 ./runltp -p -f hugetlb &
  16 ./runltp -p -f hugetlb &
  17
  18 The dtor returned by get_compound_page_dtor in __put_compound_page may be
  19 the function of free_huge_page which will lock the hugetlb_lock, so don't
  20 put_page in lock of hugetlb_lock.
  21
  22  BUG: spinlock recursion on CPU#0, hugemmap05/1079
  23   lock: hugetlb_lock+0x0/0x18, .magic: dead4ead, .owner: hugemmap05/1079, .owner_cpu: 0
  24  Call trace:
  25   dump_backtrace+0x0/0x198
  26   show_stack+0x24/0x30
  27   dump_stack+0xa4/0xcc
  28   spin_dump+0x84/0xa8
  29   do_raw_spin_lock+0xd0/0x108
  30   _raw_spin_lock+0x20/0x30
  31   free_huge_page+0x9c/0x260
  32   __put_compound_page+0x44/0x50
  33   __put_page+0x2c/0x60
  34   alloc_surplus_huge_page.constprop.19+0xf0/0x140
  35   hugetlb_acct_memory+0x104/0x378
  36   hugetlb_reserve_pages+0xe0/0x250
  37   hugetlbfs_file_mmap+0xc0/0x140
  38   mmap_region+0x3e8/0x5b0
  39   do_mmap+0x280/0x460
  40   vm_mmap_pgoff+0xf4/0x128
  41   ksys_mmap_pgoff+0xb4/0x258
  42   __arm64_sys_mmap+0x34/0x48
  43   el0_svc_common+0x78/0x130
  44   el0_svc_handler+0x38/0x78
  45   el0_svc+0x8/0xc
  46
  47 Link: http://lkml.kernel.org/r/b8ade452-2d6b-0372-32c2-703644032b47@huawei.com
  48 Fixes: 9980d744a0 ("mm, hugetlb: get rid of surplus page accounting tricks")
  49 Signed-off-by: Kai Shen <shenkai8@huawei.com>
  50 Signed-off-by: Feilong Lin <linfeilong@huawei.com>
  51 Reported-by: Wang Wang <wangwang2@huawei.com>
  52 Reviewed-by: Oscar Salvador <osalvador@suse.de>
  53 Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
  54 Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
  55 Acked-by: Michal Hocko <mhocko@suse.com>
  56 Cc: <stable@vger.kernel.org>
  57 Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
  58 Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
  59 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  60
  61 ---
  62  mm/hugetlb.c |    3 ++-
  63  1 file changed, 2 insertions(+), 1 deletion(-)
  64
  65 --- a/mm/hugetlb.c
  66 +++ b/mm/hugetlb.c
  67 @@ -1573,8 +1573,9 @@ static struct page *alloc_surplus_huge_p
  68          */
  69         if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
  70                 SetPageHugeTemporary(page);
  71 +               spin_unlock(&hugetlb_lock);
  72                 put_page(page);
  73 -               page = NULL;
  74 +               return NULL;
  75         } else {
  76                 h->surplus_huge_pages++;
  77                 h->surplus_huge_pages_node[page_to_nid(page)]++;