mm: page_alloc: revert inadvertent !__GFP_FS retry behavior change
authorJohannes Weiner <hannes@cmpxchg.org>
Fri, 27 Feb 2015 23:52:09 +0000 (15:52 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 28 Feb 2015 17:57:51 +0000 (09:57 -0800)
Historically, !__GFP_FS allocations were not allowed to invoke the OOM
killer once reclaim had failed, but nevertheless kept looping in the
allocator.

Commit 9879de7373fc ("mm: page_alloc: embed OOM killing naturally into
allocation slowpath"), which should have been a simple cleanup patch,
accidentally changed the behavior to aborting the allocation at that
point.  This creates problems with filesystem callers (?) that currently
rely on the allocator waiting for other tasks to intervene.

Revert the behavior as it shouldn't have been changed as part of a
cleanup patch.

Fixes: 9879de7373fc ("mm: page_alloc: embed OOM killing naturally into allocation slowpath")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Dave Chinner <david@fromorbit.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: <stable@vger.kernel.org> [3.19.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/page_alloc.c

index a47f0b229a1aca202b15195c88ab93d52e65064f..7abfa70cdc1ae8767fd663d905372447a7fe6864 100644 (file)
@@ -2353,8 +2353,15 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
                if (ac->high_zoneidx < ZONE_NORMAL)
                        goto out;
                /* The OOM killer does not compensate for light reclaim */
-               if (!(gfp_mask & __GFP_FS))
+               if (!(gfp_mask & __GFP_FS)) {
+                       /*
+                        * XXX: Page reclaim didn't yield anything,
+                        * and the OOM killer can't be invoked, but
+                        * keep looping as per should_alloc_retry().
+                        */
+                       *did_some_progress = 1;
                        goto out;
+               }
                /*
                 * GFP_THISNODE contains __GFP_NORETRY and we never hit this.
                 * Sanity check for bare calls of __GFP_THISNODE, not real OOM.