[01:29:32] --- reuteras has left [01:29:43] --- reuteras has become available [05:22:10] --- SecureEndpoints has left: Replaced by new connection [06:00:11] --- mmeffie has become available [06:00:37] --- matt has become available [06:07:57] --- cg2v has become available [06:09:32] --- Simon Wilkinson has left [07:14:26] --- SecureEndpoints has become available [07:18:25] --- Russ has become available [07:18:54] --- reuteras has left [07:18:54] --- reuteras has become available [07:18:54] --- reuteras has left [07:19:47] --- reuteras has become available [07:51:19] --- SecureEndpoints has left [08:00:38] --- SecureEndpoints has become available [08:05:20] --- Simon Wilkinson has become available [08:44:31] --- reuteras has left [08:53:43] --- cg2v has left [08:54:17] --- stevenjenkins has left [08:54:30] --- cg2v has become available [08:57:13] --- stevenjenkins has become available [09:02:23] --- mmeffie has left [09:05:39] --- mmeffie has become available [09:18:01] --- mmeffie has left [10:24:09] --- Simon Wilkinson has left [10:41:51] --- Simon Wilkinson has become available [11:07:46] --- cg2v has left [11:09:21] --- dev-zero@jabber.org has become available [11:09:25] --- dev-zero@jabber.org has left: offline [11:39:09] --- dev-zero@jabber.org has become available [11:39:11] --- dev-zero@jabber.org has left: offline [13:30:20] --- mdionne has become available [13:41:37] Simon: when you have a minute, I'd like to clarify a few points that came up in the release-team meeting this morning (linux cache) [13:41:59] he's idle, but i'm curious what you wish to talk about [13:45:19] there are no locking issues (that I've seen) with the "open by file handle" code. In fact I've done a lot of testing and it's solid on every filesystem I threw at it (xfs, resiser, tmpfs, btrfs, etc.) [13:45:58] istr he got it to deadlock but hasn't had time to collect details [13:46:24] where I did see problems was with "open by path" with some filesystems - xfs and reiserfs, although I didn't try many others. worked fine with ext2,3 [13:48:36] so he has done testing of the exportfs based code? I thought he was talking about opening by path [13:49:27] i don't know, we'll have to wait for him to reappear, which may be sunday [13:49:59] Ok, I'll catch him later [13:55:49] --- Russ has left: Disconnected [14:18:35] --- matt has left [15:56:58] mdionne: I think I was mispeaking. I was remembering a conversation we'd had a while back where you'd mentioned issues with deadlocks with other filesystems. [15:58:29] Hi Simon - OK, that's what I assumed [15:59:00] If those issues aren't there any more, than that's great! [15:59:26] like I said earlier, the file handle stuff is pretty solid in my testing [16:00:16] Excellent. I shall amend my thinking - sorry for the misconception. [16:01:18] in particular I found that on my system (not that much memory), tmpfs as a disk cache has similar speed to memcache (no particular tuning), but behaves MUCH better [16:02:29] That's the case that haba is interested in. I would imagine that will make him very happy. [16:02:50] It would be good to land my patch in http://rt.central.org/rt/Ticket/Display.html?id=124184 (or similar) at some point, too. [16:03:02] (That's the cache inode type abstraction stuff) [16:03:59] yes, if you look at http://rt.central.org/rt/Ticket/Display.html?id=123620 , I updated my code based on that [16:06:07] Yeh. We need to convince Derrick to take those for 1.5 [16:06:56] that patch removes the build option and makes file handle access the default method if iget() is not available (so 2.6.25+) [16:11:25] That's perhaps scarier. But 1.5.x is currently "here be dragons", so it's probably not any worse than anything else. [16:12:50] but for those kernels we're already using file handles to open files - generated from the inode number. so i don't think the change is that dramatic [16:13:12] True. [16:14:15] what changes is that we collect the real file handle from the filesystem instead of generating it ourselves. which is more correct anyway. [16:34:34] --- Russ has become available [19:01:55] --- mdionne has left [19:40:31] --- Russ has left: Disconnected [23:56:28] --- kula has left