drbd: fix potential access after free

Occasionally, if we disconnect, we triggered this assert:
  block drbd7: ASSERT FAILED tl_hash[27] == c30b0f04, expected NULL

hlist_del() happens only on master bio completion.

We used to wait for pending IO to complete before freeing tl_hash
on disconnect. We no longer do so, since we learned to "freeze"
IO on disconnect.

If the local disk is too slow, we may reach C_STANDALONE early,
and there are still some requests pending locally when we call
drbd_free_tl_hash().

If we now free the tl_hash, and later the local IO completion completes
the master bio, which then does hlist_del() and clobbers freed memory.

Do hlist_del_init() and hlist_add_fake() before kfree(tl_hash),
so the hlist_del() on master bio completion is harmless.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

1 files changed, 12 insertions, 5 deletions

diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index ea4836e0ae98..83d99133f94b 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -3801,11 +3801,18 @@ void drbd_free_tl_hash(struct drbd_conf *mdev)
 	mdev->ee_hash = NULL;
 	mdev->ee_hash_s = 0;
 
-	/* paranoia code */
-	for (h = mdev->tl_hash; h < mdev->tl_hash + mdev->tl_hash_s; h++)
-		if (h->first)
-			dev_err(DEV, "ASSERT FAILED tl_hash[%u] == %p, expected NULL\n",
-				(int)(h - mdev->tl_hash), h->first);
+	/* We may not have had the chance to wait for all locally pending
+	 * application requests. The hlist_add_fake() prevents access after
+	 * free on master bio completion. */
+	for (h = mdev->tl_hash; h < mdev->tl_hash + mdev->tl_hash_s; h++) {
+		struct drbd_request *req;
+		struct hlist_node *pos, *n;
+		hlist_for_each_entry_safe(req, pos, n, h, collision) {
+			hlist_del_init(&req->collision);
+			hlist_add_fake(&req->collision);
+		}
+	}
+
 	kfree(mdev->tl_hash);
 	mdev->tl_hash = NULL;
 	mdev->tl_hash_s = 0;