[02:54:39] --- dwbotsch has left [02:56:31] --- dwbotsch has become available [03:51:12] --- jaltman has left: Replaced by new connection [03:51:13] --- jaltman has become available [05:37:55] --- matt has left [05:38:19] --- matt has become available [05:55:04] ben: honestly that patch was focused on 70. also, I don't think any vrefcounting change in osi_vm by me was final. [06:21:05] also, wrt mp, you might look into mp safety. if we're breaking assertions, that's the bug, we want to be mp safe [06:21:14] (different values of mp, here) [09:26:06] --- deason has become available [09:28:46] --- matt has left [11:02:56] Okay, I'll try enabling the PutVCache calls for 8 (9, actually) and see if anything seems to break. [11:04:00] At some point I'll build a DEBUG_VFS_LOCKS kernel and see how bad things are. And now that I have working kernel dumps again, tracking down mp-non-safety may get easier. [11:31:03] ... fascinating. I'll have to do more testing, but it looks like putting the PutVCache calls back for afs_vnop_lookup makes this panic in statat go away. I am getting my tokens discarded after some time of using them, though, which may be fun to track down. [11:42:29] --- Simon Wilkinson has become available [11:52:48] --- matt has become available [12:00:41] --- Russ has left: Disconnected [12:00:47] --- Russ has become available [12:23:04] --- Simon Wilkinson has left [12:38:30] uh. if you were getting panics on stat? yeah, you had the same issue macos did that i just fixed. i bet if you turned off bulkstat they'd go away. i should look through the PutVCache code [13:04:39] Maybe that's related, but I don't trust (that part of) my change that Ben's referring to. [13:30:20] --- jaltman has left: Replaced by new connection [13:30:21] --- jaltman has become available [15:08:52] --- mdionne has become available [15:25:38] --- matt has left [16:10:29] --- matt has become available [16:30:47] --- mdionne has left [17:20:07] --- jaltman has left: Disconnected [17:20:23] --- jaltman has become available [18:19:45] --- jaltman has left: Disconnected [19:56:05] --- jaltman has become available [19:58:04] Derrick: I posted some comments on the panic I was seeing to this room. Nothing from AFS was in the trace, but vfs_unbusy wanted to lock the mount point lock, and was panicing that the lock was not a (?) spin lock (it was in fact not a valid lock type at all). This was because the mount point pointer was pointing at garbage (and I have absolutely no idea why; our vfs_root vnop shouldn't have been able to touch that). Note that adding in PutVCache calls caused things to get better (or at least seems to do so). [20:27:48] --- Born Fool has become available [21:29:04] --- Born Fool has left [22:09:55] --- reuteras has become available [22:15:35] --- deason has left [23:09:33] --- Russ has left