Quoth Keith M Wesolowski on Wed, Aug 29, 2007 at 10:39:58AM -0700: > It appears that in rc_node_delete() near rc_node.c:3359, we drop our > node lock while checking permissions. I'm trying to understand > whether this is a rule, as I can see no other reason for dropping the > lock there. If so, three calls to rc_node_modify_permission_check() > seem to be in error; they are made while holding a node lock, despite > that function's non-use of anything which could be protected by it.
configd drops locks around perm_granted() because it might do name service lookups, which could take a long time. I would not be surprised if that rule was not applied consistently or thoroughly, or if it isn't the right rule in the first place. David
