Probe vdevs before marking removed

Before allowing the ZED to mark a vdev as REMOVED due to a
hotplug event confirm that it is non-responsive with probe.
Any device which can be successfully probed should be left
ONLINE to prevent a healthy pool from being incorrectly
SUSPENDED.  This may occur for at least the following two
scenarios.

1) Drive expansion (zpool online -e) in VMware environments.
   If, during the partition resize operation, a partition is
   removed and re-created then udev will send a removed event.

2) Re-scanning the namespaces of an NVMe device (nvme ns-rescan)
   may result in a udev remove and add event being delivered.

Finally, update the ZED to only kick in a spare when the
removal was successful.

Reviewed-by: Ameer Hamza <[email protected]>
Reviewed-by: Tony Hutter <[email protected]>
Reviewed-by: Richard Yao <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Issue #14859
Closes #14861 

2 files changed, 14 insertions, 5 deletions

diff --git a/cmd/zed/agents/zfs_retire.c b/cmd/zed/agents/zfs_retire.c
index 28714ec29..f83ae0925 100644
--- a/cmd/zed/agents/zfs_retire.c
+++ b/cmd/zed/agents/zfs_retire.c
@@ -445,14 +445,16 @@ zfs_retire_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
 			return;
 
 		/* Remove the vdev since device is unplugged */
+		int remove_status = 0;
 		if (l2arc || (strcmp(class, "resource.fs.zfs.removed") == 0)) {
-			int status = zpool_vdev_remove_wanted(zhp, devname);
+			remove_status = zpool_vdev_remove_wanted(zhp, devname);
 			fmd_hdl_debug(hdl, "zpool_vdev_remove_wanted '%s'"
-			    ", ret:%d", devname, status);
+			    ", err:%d", devname, libzfs_errno(zhdl));
 		}
 
 		/* Replace the vdev with a spare if its not a l2arc */
-		if (!l2arc && (!fmd_prop_get_int32(hdl, "spare_on_remove") ||
+		if (!l2arc && !remove_status &&
+		    (!fmd_prop_get_int32(hdl, "spare_on_remove") ||
 		    replace_with_spare(hdl, zhp, vdev) == B_FALSE)) {
 			/* Could not handle with spare */
 			fmd_hdl_debug(hdl, "no spare for '%s'", devname);
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 4bfd95861..c243dddb7 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -4097,11 +4097,18 @@ vdev_remove_wanted(spa_t *spa, uint64_t guid)
 		return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));
 
 	/*
-	 * If the vdev is already removed, then don't do anything.
+	 * If the vdev is already removed, or expanding which can trigger
+	 * repartition add/remove events, then don't do anything.
 	 */
-	if (vd->vdev_removed)
+	if (vd->vdev_removed || vd->vdev_expanding)
 		return (spa_vdev_state_exit(spa, NULL, 0));
 
+	/*
+	 * Confirm the vdev has been removed, otherwise don't do anything.
+	 */
+	if (vd->vdev_ops->vdev_op_leaf && !zio_wait(vdev_probe(vd, NULL)))
+		return (spa_vdev_state_exit(spa, NULL, SET_ERROR(EEXIST)));
+
 	vd->vdev_remove_wanted = B_TRUE;
 	spa_async_request(spa, SPA_ASYNC_REMOVE);