FAILED: patch "[PATCH] rbd: avoid a deadlock on header_rwsem when flushing notifies" failed to apply to 4.9-stable tree

PATCH [email protected] 1 participant 1 message Apr 21, 2020

Subsystems (3)

Stable branch Block layer RADOS block device (RBD)

lore

Share & tools

lore

[email protected]

author maintainer

2020-04-21 17:19 UTC #0 lore raw

The patch below does not apply to the 4.9-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 0e4e1de5b63fa423b13593337a27fd2d2b0bcf77 Mon Sep 17 00:00:00 2001
From: Ilya Dryomov <[email protected]>
Date: Fri, 13 Mar 2020 11:20:51 +0100
Subject: [PATCH] rbd: avoid a deadlock on header_rwsem when flushing notifies

rbd_unregister_watch() flushes notifies and therefore cannot be called under header_rwsem because a header update notify takes header_rwsem to synchronize with "rbd map". If mapping an image fails after the watch is established and a header update notify sneaks in, we deadlock when erroring out from rbd_dev_image_probe().

Move watch registration and unregistration out of the critical section. The only reason they were put there was to make header_rwsem management slightly more obvious.

Fixes 811c66887746 ("rbd: fix rbd map vs notify races")

Signed-off-by Ilya Dryomov <[email protected]>

Reviewed-by Jason Dillaman <[email protected]>

drivers/block/rbd.c +13 -4

index 1e0a6b19ae0d..ff2377e6d12c 100644
@@ -4527,6 +4527,10 @@ static void cancel_tasks_sync(struct rbd_device *rbd_dev)
 	cancel_work_sync(&rbd_dev->unlock_work);
 }
 
+/*
+ * header_rwsem must not be held to avoid a deadlock with
+ * rbd_dev_refresh() when flushing notifies.
+ */
 static void rbd_unregister_watch(struct rbd_device *rbd_dev)
 {
 	cancel_tasks_sync(rbd_dev);

@@ -6907,6 +6911,9 @@ static void rbd_dev_image_release(struct rbd_device *rbd_dev)
  * device.  If this image is the one being mapped (i.e., not a
  * parent), initiate a watch on its header object before using that
  * object to get detailed information about the rbd image.
+ *
+ * On success, returns with header_rwsem held for write if called
+ * with @depth == 0.
  */
 static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
 {

@@ -6936,6 +6943,9 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
 		}
 	}
 
+	if (!depth)
+		down_write(&rbd_dev->header_rwsem);
+
 	ret = rbd_dev_header_info(rbd_dev);
 	if (ret) {
 		if (ret == -ENOENT && !need_watch)

@@ -6987,6 +6997,8 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
 err_out_probe:
 	rbd_dev_unprobe(rbd_dev);
 err_out_watch:
+	if (!depth)
+		up_write(&rbd_dev->header_rwsem);
 	if (need_watch)
 		rbd_unregister_watch(rbd_dev);
 err_out_format:

@@ -7050,12 +7062,9 @@ static ssize_t do_rbd_add(struct bus_type *bus,
 		goto err_out_rbd_dev;
 	}
 
-	down_write(&rbd_dev->header_rwsem);
 	rc = rbd_dev_image_probe(rbd_dev, 0);
-	if (rc < 0) {
-		up_write(&rbd_dev->header_rwsem);
+	if (rc < 0)
 		goto err_out_rbd_dev;
-	}
 
 	if (rbd_dev->opts->alloc_size > rbd_dev->layout.object_size) {
 		rbd_warn(rbd_dev, "alloc_size adjusted to %u",