hugetlb: do not demote poisoned hugetlb pages
authorMike Kravetz <mike.kravetz@oracle.com>
Fri, 15 Apr 2022 02:13:52 +0000 (19:13 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 20 Apr 2022 07:36:24 +0000 (09:36 +0200)
commit 5a317412ef884763fdf7aa17f9f3636959d11d8f upstream.

It is possible for poisoned hugetlb pages to reside on the free lists.
The huge page allocation routines which dequeue entries from the free
lists make a point of avoiding poisoned pages.  There is no such check
and avoidance in the demote code path.

If a hugetlb page on the is on a free list, poison will only be set in
the head page rather then the page with the actual error.  If such a
page is demoted, then the poison flag may follow the wrong page.  A page
without error could have poison set, and a page with poison could not
have the flag set.

Check for poison before attempting to demote a hugetlb page.  Also,
return -EBUSY to the caller if only poisoned pages are on the free list.

Link: https://lkml.kernel.org/r/20220307215707.50916-1-mike.kravetz@oracle.com
Fixes: 8531fc6f52f5 ("hugetlb: add hugetlb demote page support")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mm/hugetlb.c

index f294db835f4bc46b737cc363ec99e7579ad3fde4..a1da8757cc9cc5e6f5a41f2713714289fe515e23 100644 (file)
@@ -3469,7 +3469,6 @@ static int demote_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
 {
        int nr_nodes, node;
        struct page *page;
-       int rc = 0;
 
        lockdep_assert_held(&hugetlb_lock);
 
@@ -3480,15 +3479,19 @@ static int demote_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
        }
 
        for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
-               if (!list_empty(&h->hugepage_freelists[node])) {
-                       page = list_entry(h->hugepage_freelists[node].next,
-                                       struct page, lru);
-                       rc = demote_free_huge_page(h, page);
-                       break;
+               list_for_each_entry(page, &h->hugepage_freelists[node], lru) {
+                       if (PageHWPoison(page))
+                               continue;
+
+                       return demote_free_huge_page(h, page);
                }
        }
 
-       return rc;
+       /*
+        * Only way to get here is if all pages on free lists are poisoned.
+        * Return -EBUSY so that caller will not retry.
+        */
+       return -EBUSY;
 }
 
 #define HSTATE_ATTR_RO(_name) \