[00:24:17] --- reuteras has become available [00:42:23] --- jaltman has left: Disconnected [04:27:48] --- jaltman has become available [04:45:01] --- jaltman has left: Disconnected [04:50:49] --- meffie has become available [05:21:01] --- reuteras has left [06:00:17] --- jaltman has become available [07:05:19] --- jaltman has left: Disconnected [07:10:29] --- jaltman has become available [07:26:01] --- jaltman has left: Disconnected [07:26:10] --- jaltman has become available [07:28:20] --- deason has become available [08:21:29] er, does gerrit 3267 need to go on master, too? [08:21:58] potpourri? probably. doh. [08:45:09] --- mfelliott has left [08:51:10] --- jaltman has left: Disconnected [09:52:24] --- mfelliott has become available [10:08:20] --- jaltman has become available [11:24:52] --- jaltman has left: Disconnected [11:36:37] --- kula has left [11:39:01] --- jaltman has become available [12:04:01] I used the specfile from the 1.4.12-1.1.2 SRPM. It builds the openafs-kpasswd package (which, for this purpose, I don't want). But the i386 version also puts kpasswd in the 'openafs' package, which makes it conflict with krb5-workstation [12:56:57] There’s no Git tag for 1.5.78. [12:57:46] uh. [12:57:51] hang on [12:58:40] have a tag. (i forgot to push it) [13:30:58] > afs_root ... bogus PutVCache ... panic Hm, but if I'm running dynroot, should the release of root.afs effect me? [13:31:51] no, it shouldn't. but changing the configuariuon (newly discovered cell) will [13:32:01] e.g. AFSDB or SRV finds new to you cell [13:38:45] I'm the only user on the machine, and only touch one volume in one cell. [13:39:41] then unlikely [13:42:43] --- meffie has left [13:42:50] (Unless make is doing something dumb getting to the builddir, but it seems unlikely) [13:43:13] well, it could. [13:43:19] fstrace will tell you [13:45:27] I actually haven't used fstrace, yet. Maybe I will play around in the zone cell to see if I can trigger this with volume releases. [13:45:42] fstrace setset cm -active [13:45:46] do stuff [13:45:50] fstrace dump cm [13:45:59] make sure afszcm.cat is findable where fstrace expects it [13:46:45] I'm not sure how that helps me when the direct event of interest is a panic? [13:47:10] true [13:47:17] arrange to not panic :\ [13:53:42] DECLARE_MUTEX is gone in Linux 2.6.37; it should be replaced with DEFINE_SEMAPHORE or with real mutexes: http://git.kernel.org/linus/4882720b [13:54:14] got sha1? [13:54:52] Like 4882720b? [13:54:58] 4882720b267b7b1d1b0ce08334b205f0329d4615 if you want the rest of it. [13:55:36] cool [13:55:52] wait. i meant in our gerrit [13:56:27] Oh, as in, where's the patch? [13:56:33] :) [13:57:20] I can write one later if nobody gets to it first or has already. [13:58:07] i likely won't by the time you do. maybe marc will [14:06:02] This doesn't affect DEFINE_MUTEX, right? [14:09:31] DEFINE_MUTEX is for real mutexes. [14:09:53] (as opposed to DECLARE_MUTEX, which was for using semaphores as mutexes) [14:10:24] Then it looks like all the changes are limited to osi_alloc.c [14:31:24] --- rra has become available [14:48:45] --- mdionne has become available [14:55:15] Okay, there's a patch in gerrit. [15:02:33] that patch looks ok. but there's other issues to fix for 2.6.37 with llseek ops. I don't have a working module yet [15:31:46] --- deason has left [15:43:28] --- deason has become available [15:47:29] --- mdionne has left [17:05:58] Hm, looking more carefully at this dump, that goto (in afs_root) may not actually be in the buggy path. [17:28:53] WTF? afs_globalVp is NULL, but it seems to be what was passed to PutVCache, and the underlying vputx has a valid vnode with a valid v_data ... [17:59:05] --- rra has left: Disconnected [18:16:39] --- Russ has become available [19:55:00] --- kula has become available [20:46:09] Yeah, there are two threads calling afs_PutVCache(afs_globalVp) at the same time (and a third waiting for the GLOCK in afs_root()). So presumably the but you spotted from inspection is the one I'm tickling. [21:37:06] --- Russ has left: Disconnected [23:08:46] so just patch the bug :)