[00:15:50] --- reuteras has left
[00:27:09] --- Russ has become available
[03:16:08] --- Russ has left: Disconnected
[04:04:21] --- steven.jenkins has left
[04:33:24] --- geekosaur has left
[04:58:53] --- geekosaur has become available
[05:18:33] --- reuteras has become available
[05:52:44] --- meffie has become available
[05:52:57] --- meffie has left
[05:53:01] --- meffie has become available
[07:30:59] --- deason has become available
[07:39:34] --- reuteras has left
[08:54:54] --- rra has become available
[08:56:54] --- jaltman has left: Replaced by new connection
[08:56:55] --- jaltman has become available
[08:57:35] --- jaltman has left: Disconnected
[08:59:00] --- jaltman has become available
[10:52:49] --- jaltman has left: Disconnected
[10:53:02] --- jaltman has become available
[11:05:43] --- meffie has left
[11:33:08] --- meffie has become available
[12:19:56] --- phalenor has left
[12:31:49] --- phalenor has become available
[14:08:54] --- phalenor has left
[14:18:55] --- phalenor has become available
[14:55:58] --- meffie has left
[15:36:42] --- deason has left
[16:16:31] --- deason has become available
[17:25:43] --- geekosaur has left
[17:25:49] --- geekosaur has become available
[17:53:39] --- rra has left: Disconnected
[18:13:34] --- Russ has become available
[21:22:10] <kaduk@mit.edu/barnowl> I am rather disturbed that this removes the trivial deadlock that I
can see in lookup ...
diff --git a/src/afs/FBSD/osi_vnodeops.c b/src/afs/FBSD/osi_vnodeops.c
index cbd94d7..5215642 100644
--- a/src/afs/FBSD/osi_vnodeops.c
+++ b/src/afs/FBSD/osi_vnodeops.c
@@ -531,7 +531,9 @@ afs_vop_lookup(ap)
      * we also always return the vnode locked. */
 
     if (flags & ISDOTDOT) {
+       MA_VOP_UNLOCK(dvp, 0, p);
        ma_vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
+       ma_vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p);
        /* always return the child locked */
        if (lockparent && (flags & ISLASTCN)
            && (error = ma_vn_lock(dvp, LK_EXCLUSIVE, p))) {

(I also think lookup should probably be rewritten from scratch, but
...)
[21:32:57] --- jaltman has left: Replaced by new connection
[21:32:58] --- jaltman has become available
[21:44:05] --- deason has left
[21:51:03] <shadow@gmail.com/owl256771E3> why? lock order matters
[21:55:40] --- jaltman has left: Disconnected
[21:55:58] --- jaltman has become available
[22:12:14] --- jaltman has left: Disconnected
[22:19:39] --- jaltman has become available
[22:19:48] <kaduk@mit.edu/barnowl> Sure.  But it feels like fixing the symptom instead of fixing the problem.
I kind of want to see dvp unlocked much earlier in the ISDOTDOT case, maybe even
before the rpc.
[23:21:10] --- reuteras has become available