ceph: flush msgr_wq during mds_client shutdown

The release method for mds connections uses a backpointer to the
mds_client, so we need to flush the workqueue of any pending work (and
ceph_connection references) prior to freeing the mds_client.  This fixes
an oops easily triggered under UML by

 while true ; do mount ... ; umount ... ; done

Also fix an outdated comment: the flush in ceph_destroy_client only flushes
OSD connections out.  This bug is basically an artifact of the ceph ->
ceph+libceph conversion.
Signed-off-by: Sage Weil <sage@newdream.net>

fs/ceph/mds_client.c

@@ -3215,9 +3215,15 @@ void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
 {
 	struct ceph_mds_client *mdsc = fsc->mdsc;
 
+	dout("mdsc_destroy %p\n", mdsc);
 	ceph_mdsc_stop(mdsc);
+
+	/* flush out any connection work with references to us */
+	ceph_msgr_flush();
+
 	fsc->mdsc = NULL;
 	kfree(mdsc);
+	dout("mdsc_destroy %p done\n", mdsc);
 }
 
 

net/ceph/ceph_common.c

@@ -394,8 +394,8 @@ void ceph_destroy_client(struct ceph_client *client)
 	ceph_osdc_stop(&client->osdc);
 
 	/*
-	 * make sure mds and osd connections close out before destroying
-	 * the auth module, which is needed to free those connections'
+	 * make sure osd connections close out before destroying the
+	 * auth module, which is needed to free those connections'
 	 * ceph_authorizers.
 	 */
 	ceph_msgr_flush();