[00:09:45] --- Russ has left: Disconnected [00:10:05] --- Russ has become available [00:21:02] --- haba has become available [00:27:53] --- dev-zero@jabber.org has become available [00:29:30] --- Russ has left: Disconnected [00:29:32] --- Russ has become available [00:40:47] --- Russ has left: Replaced by new connection [00:40:47] --- Russ has become available [00:58:42] --- Russ has left: Disconnected [01:24:04] --- haba has left [01:41:32] --- haba has become available [02:28:47] --- Jeffrey Altman has left: Replaced by new connection [02:41:57] --- haba has left [04:19:25] --- haba has become available [04:24:29] --- Jeffrey Altman has become available [05:15:16] --- abo has left [05:18:54] --- abo has become available [05:26:28] --- Jeffrey Altman has left: Replaced by new connection [05:50:00] --- reuteras has become available [06:15:42] --- Jeffrey Altman has become available [06:26:22] --- reuteras has left [07:03:34] --- deason has become available [07:11:12] --- Simon Wilkinson has left [07:23:58] --- Simon Wilkinson has become available [07:30:37] --- haba has left [07:56:46] --- haba has become available [08:45:42] --- brantgurga has become available [09:39:10] --- haba has left [10:15:30] --- dev-zero@jabber.org has left [10:41:32] --- Russ has become available [11:45:16] --- brantgurga has left [11:46:57] --- dev-zero@jabber.org has become available [11:51:31] --- deason has left [11:51:45] --- deason has become available [12:02:01] --- deason has left [12:02:05] --- deason has become available [12:05:36] hmm, unix cm question: updates to afs_blocksUsed in afs_AdjustSize don't seem to be serialized; shouldn't that be protected by some lock? [12:05:50] the struct dcache is locked, but couldn't two dcaches hit that at the same time? [12:08:16] The GLOCK will catch it, I would have thought. [12:08:29] So, it should get fixed, but its not likely to go wrong at the moment. [12:11:49] GLOCK is always held in there? [12:12:38] I thought I saw callers holding dc->lock before calling [12:13:28] You have to hold the GLOCK to be able to manipulate any AFS lock. [12:13:58] We pretty much get the GLOCK at the kernel/us boundary, and hold it unless we sleep or call back into the kernel. [12:17:40] okay, but "it should get fixed" still? I wouldn't see how it's a problem, then [12:17:56] Because at some point GLOCK will go away. [12:18:01] (in my dreams, anyway) [12:18:13] And we'll rely upon the finer grained locks to protect our datastructures. [12:18:30] okay [12:18:39] I assume we are not near that point [12:19:10] Not currently. And on Linux, we also get the BKL on entry, so that serialises us, too. [12:19:22] Our SMP story is not a pleasant one ... [12:21:23] i'd love to ive in a GLOCK-free universe [12:23:10] * Russ secretly replaces all of the GLOCKS with Smith and Wessons and sees if anyone notices. [12:23:48] My suspicion is that we could get rid of the BKL if we fixed our inode locking when using the disk cache. [12:26:21] The GLOCK is harder. I'd really love to have access to some kind of static lock checking tool so that we could make sure we hold the right locks. Because at present, I'm pretty sure we don't. [13:04:25] --- haba has become available [13:16:40] --- brantgurga has become available [13:26:34] --- dwbotsch has left [13:27:15] --- dwbotsch has become available [14:05:17] --- dev-zero@jabber.org has left: Replaced by new connection [14:05:17] --- dev-zero@jabber.org has become available [14:24:53] --- brantgurga has left [15:01:36] 279 (DAFS salvage change) also looks good to me. I'm confirming that it compiles now. It looks like Derrick already approved; if I confirm it builds, are we set to submit it, or did we want more specific testing? [15:05:36] --- dev-zero@jabber.org has left [15:20:01] --- deason has left [15:37:16] --- summatusmentis has become available [15:42:12] 279: i wanted to see if i, like, worked. on a real server. [15:42:46] Okay, that would be more specific testing. :) [15:43:06] * Russ should get around to setting up a test cell for verifying patches. [15:43:35] i have one. i am just busy with making macos something later than 10.6 work [15:45:27] Oh, yeah, I understand -- I meant set up my own so that you don't have to do it. [15:45:53] I think 216 (GNU-style long options) never made it to openafs-devel. [15:45:58] i'll give you admin in dementia if you want. wtf do i care... [15:59:13] I wish -1's weren't persistant after uploading a new patch, although I can see how that would go either way. [15:59:32] 120 looks good to me, although I probably shouldn't submit that without building a kernel module that I actually use for something. [16:00:13] heh [16:47:32] --- deason has become available [17:26:10] --- brantgurga has become available [17:39:52] Brant, are you around? [17:40:19] mmc.exe is searching for HKCU\Software\Classes\CLSID\{4A65D267-1539-4BD1-921D-1C49B3E58EB7} and not finding it [17:41:01] same for HKCR\CLSID\{4A65D267-1539-4BD1-921D-1C49B3E58EB7}\AppID [18:36:43] --- brantgurga has left [18:49:56] --- brantgurga has become available [20:05:00] 215 (adding delayed aborts to rxdebug stats) looks good to me. Is there any other concern to check there before applying it? Any problem with assigning those two additional reserved ints in the struct to that purpose? [20:05:07] --- dwbotsch has left: Lost connection [20:05:07] --- Rrrrred has left: Lost connection [20:05:44] it's visible on the wire, yes? didn't i comment on this? [20:07:18] It's visible on the wire, yeah, good point. So we need a protocol review. [20:07:35] There are no comments. [20:08:01] damn. i thought i sent "talk to afs3-stds" [20:09:02] * Russ is doing so now. [20:09:19] --- Rrrrred has become available [20:09:20] --- dwbotsch has become available [21:26:31] --- reuteras has become available [22:02:37] --- deason has left [23:11:57] --- Russ has left: Disconnected [23:27:08] --- dev-zero@jabber.org has become available [23:43:26] --- brantgurga has left [23:51:18] --- haba has left