mm,hwpoison: double-check page count in __get_any_page()
authorNaoya Horiguchi <naoya.horiguchi@nec.com>
Fri, 16 Oct 2020 03:07:25 +0000 (20:07 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 16 Oct 2020 18:11:17 +0000 (11:11 -0700)
Soft offlining could fail with EIO due to the race condition with hugepage
migration.  This issuse became visible due to the change by previous patch
that makes soft offline handler take page refcount by its own.  We have no
way to directly pin zero refcount page, and the page considered as a zero
refcount page could be allocated just after the first check.

This patch adds the second check to find the race and gives us chance to
handle it more reliably.

Reported-by: Qian Cai <cai@lca.pw>
Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Aristeu Rozanski <aris@ruivo.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Dmitry Yakunin <zeil@yandex-team.ru>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Oscar Salvador <osalvador@suse.com>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20200922135650.1634-14-osalvador@suse.de
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memory-failure.c

index f9fa9982b5d4a3971961eaa47462ac8dbeb9d4a1..7c63ba9ff6e77bcc6cf695e56ef7a62dc7940a71 100644 (file)
@@ -1707,6 +1707,9 @@ static int __get_any_page(struct page *p, unsigned long pfn, int flags)
                } else if (is_free_buddy_page(p)) {
                        pr_info("%s: %#lx free buddy page\n", __func__, pfn);
                        ret = 0;
+               } else if (page_count(p)) {
+                       /* raced with allocation */
+                       ret = -EBUSY;
                } else {
                        pr_info("%s: %#lx: unknown zero refcount page type %lx\n",
                                __func__, pfn, p->flags);
@@ -1723,6 +1726,9 @@ static int get_any_page(struct page *page, unsigned long pfn, int flags)
 {
        int ret = __get_any_page(page, pfn, flags);
 
+       if (ret == -EBUSY)
+               ret = __get_any_page(page, pfn, flags);
+
        if (ret == 1 && !PageHuge(page) &&
            !PageLRU(page) && !__PageMovable(page)) {
                /*