mm/z3fold.c: fix z3fold_destroy_pool() ordering
authorHenry Burns <henryburns@google.com>
Tue, 13 Aug 2019 22:37:21 +0000 (15:37 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 13 Aug 2019 23:06:52 +0000 (16:06 -0700)
The constraint from the zpool use of z3fold_destroy_pool() is there are
no outstanding handles to memory (so no active allocations), but it is
possible for there to be outstanding work on either of the two wqs in
the pool.

If there is work queued on pool->compact_workqueue when it is called,
z3fold_destroy_pool() will do:

   z3fold_destroy_pool()
     destroy_workqueue(pool->release_wq)
     destroy_workqueue(pool->compact_wq)
       drain_workqueue(pool->compact_wq)
         do_compact_page(zhdr)
           kref_put(&zhdr->refcount)
             __release_z3fold_page(zhdr, ...)
               queue_work_on(pool->release_wq, &pool->work) *BOOM*

So compact_wq needs to be destroyed before release_wq.

Link: http://lkml.kernel.org/r/20190726224810.79660-1-henryburns@google.com
Fixes: 5d03a6613957 ("mm/z3fold.c: use kref to prevent page free/compact race")
Signed-off-by: Henry Burns <henryburns@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Reviewed-by: Jonathan Adams <jwadams@google.com>
Cc: Vitaly Vul <vitaly.vul@sony.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Al Viro <viro@zeniv.linux.org.uk
Cc: Henry Burns <henrywolfeburns@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/z3fold.c

index 1a029a7432ee56635f9dcde8f4c5ad2b1e2984ed..43de92f529617fd18c143002b60a50fbccaa4df2 100644 (file)
@@ -818,8 +818,15 @@ static void z3fold_destroy_pool(struct z3fold_pool *pool)
 {
        kmem_cache_destroy(pool->c_handle);
        z3fold_unregister_migration(pool);
-       destroy_workqueue(pool->release_wq);
+
+       /*
+        * We need to destroy pool->compact_wq before pool->release_wq,
+        * as any pending work on pool->compact_wq will call
+        * queue_work(pool->release_wq, &pool->work).
+        */
+
        destroy_workqueue(pool->compact_wq);
+       destroy_workqueue(pool->release_wq);
        kfree(pool);
 }