ceph: queue cap_snaps once per realm

We were forming a dirty list, and then queueing cap_snaps for each realm
_and_ its children, regardless of whether the children were already in the
dirty list.  This meant we did it twice for some realms.  Which in turn
meant we corrupted mdsc->snap_flush_list when the cap_snap was re-added to
the list it was already on, and could trigger an infinite loop.

We were also using recursion to do reach all the children, a no-no when
stack is limited.

Instead, (re)queue any children on the dirty list, avoiding processing
anything twice and avoiding any recursion.

Signed-off-by: Sage Weil <sage@newdream.net>

diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
index 39c243a..f40b913 100644 (file)
--- a/fs/ceph/snap.c
+++ b/fs/ceph/snap.c
@@ -584,10 +584,14 @@ static void queue_realm_cap_snaps(struct ceph_snap_realm *realm)
        if (lastinode)
                iput(lastinode);
 
-       dout("queue_realm_cap_snaps %p %llx children\n", realm, realm->ino);
-       list_for_each_entry(child, &realm->children, child_item)
-               queue_realm_cap_snaps(child);
+       list_for_each_entry(child, &realm->children, child_item) {
+               dout("queue_realm_cap_snaps %p %llx queue child %p %llx\n",
+                    realm, realm->ino, child, child->ino);
+               list_del_init(&child->dirty_item);
+               list_add(&child->dirty_item, &realm->dirty_item);
+       }
 
+       list_del_init(&realm->dirty_item);
        dout("queue_realm_cap_snaps %p %llx done\n", realm, realm->ino);
 }
 
@@ -683,7 +687,9 @@ more:
         * queue cap snaps _after_ we've built the new snap contexts,
         * so that i_head_snapc can be set appropriately.
         */
-       list_for_each_entry(realm, &dirty_realms, dirty_item) {
+       while (!list_empty(&dirty_realms)) {
+               realm = list_first_entry(&dirty_realms, struct ceph_snap_realm,
+                                        dirty_item);
                queue_realm_cap_snaps(realm);
        }