drm/radeon: make radeon_mn_get wait for mmap_sem killable
authorMichal Hocko <mhocko@suse.com>
Mon, 23 May 2016 23:26:14 +0000 (16:26 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 24 May 2016 00:04:14 +0000 (17:04 -0700)
radeon_mn_get which is called during ioct path relies on mmap_sem for
write.  If the waiting task gets killed by the oom killer it would block
oom_reaper from asynchronous address space reclaim and reduce the
chances of timely OOM resolving.  Wait for the lock in the killable mode
and return with EINTR if the task got killed while waiting.

Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: David Airlie <airlied@linux.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/gpu/drm/radeon/radeon_mn.c

index eef006c..896f2cf 100644 (file)
@@ -186,7 +186,9 @@ static struct radeon_mn *radeon_mn_get(struct radeon_device *rdev)
        struct radeon_mn *rmn;
        int r;
 
-       down_write(&mm->mmap_sem);
+       if (down_write_killable(&mm->mmap_sem))
+               return ERR_PTR(-EINTR);
+
        mutex_lock(&rdev->mn_lock);
 
        hash_for_each_possible(rdev->mn_hash, rmn, node, (unsigned long)mm)