[02:29:43] --- Simon Wilkinson has become available [02:32:33] andersk: gdb your kernel image and list *vfs_readdir+0x7f [02:36:15] At a guess, that's probably going to be the inode_mutex, which is taken with mutex_lock_killable() [03:11:52] Actually, yeah. d_alloc_and_lookup requires that the inode mutex must be held on the parent of the object being allocated, which is the one that vfs_readdir is trying to acquire. So, this is a Linux kernel issue - looking at the cache manager isn't going to help us. [03:12:16] The best we could do to improve things is make the sleep waiting for AFSDB records interruptible. [06:44:05] --- deason has become available [06:52:52] --- Simon Wilkinson has left [09:15:40] --- Roman Mitz has become available [09:21:59] --- mfelliott has become available [09:41:26] --- rra has become available [11:18:04] --- Simon Wilkinson has become available [12:24:58] --- deason has left [14:01:39] --- Simon Wilkinson has left [14:01:40] --- Simon Wilkinson has become available [14:42:15] --- phalenor has left [16:25:17] --- Roman Mitz has left [16:25:18] --- Roman Mitz has become available [17:40:26] --- rra has left: Disconnected [17:58:16] --- Russ has become available [18:23:33] --- jaltman/FrogsLeap has left: Disconnected [18:23:58] --- jaltman/FrogsLeap has become available [19:41:04] --- phalenor has become available [20:06:35] --- Roman Mitz has left [21:01:37] --- shadow@gmail.com/barnowlFF509CC1 has left [21:02:09] --- shadow@gmail.com/barnowlFF509CC1 has become available [21:03:38] --- shadow@gmail.com/barnowlFF509CC1 has left [21:05:02] --- shadow@gmail.com/barnowlFF509CC1 has become available [21:30:30] --- Russ has left: Disconnected