[00:42:42] --- Rasmus Kaj has become available [01:18:20] --- dev-zero@jabber.org has left [02:07:03] --- dev-zero@jabber.org has become available [02:23:24] --- dev-zero@jabber.org has left [02:23:35] --- dev-zero@jabber.org has become available [04:58:03] --- Jeffrey Altman has become available [05:47:10] --- reuteras has left [05:55:41] --- meffie has left [06:12:19] --- deason has become available [06:25:09] --- reuteras has become available [06:35:03] I've just added documentation about the ChangeId hook to http://www.dementia.org/twiki/bin/view/AFSLore/GitDevelopers if folk would like to review and comment on it. [06:50:51] --- Simon Wilkinson has left [06:50:52] --- Simon Wilkinson has become available [06:54:24] --- Simon Wilkinson has left [06:56:55] --- Simon Wilkinson has become available [07:20:09] --- reuteras has left [08:01:04] shadow: if it makes a difference, 995 was without dynroot [08:01:33] Does you root.afs have 'rl' permissions? [08:02:17] (Derrick and I were just discussing that, and thought that the breakage would only be _with_ dynroot, so without is interesting.) [08:02:35] s:anyuser rl, yeah [08:03:11] Hmmm. So we're getting something wrong then, because that should pass the getattr check. [08:04:53] I can dig a bit more now; got at least a little time before more meetings [08:05:57] We never VerifyVCache [08:07:02] Ah no, eventually we do. [08:09:40] err, doesn't afs_AccessOK return 1 for success? [08:10:42] Actuall, yes. Bing, we have a winner... [08:27:39] when i finish this round with a kernel debugger i can see wtf [08:28:15] deason: Your comment on 995 - as in, with your change, we still leak stat information? [08:28:27] yes [08:31:22] awesome. Code Review - Unexpected error. com.google.gwt.user.client.rpc.Invocation (and a block which is a bad JSON response which appears to be an encapsulated entire patch) [08:31:40] not reproducible at all. [08:31:45] Nice. [08:32:10] I should upgrade gerrit at some point over the holidays - 2.1 is out, which changes a fair few things [08:32:47] the router in my coffeehouse of record was offline and had to be reset, maybe something weir when i switched from tethered back to wifi [08:33:12] *facepalm*. well, i see the likely issue with the other code i'm touching [08:38:21] --- dev-zero@jabber.org has left [08:41:00] Hmmm. The fileserver doesn't seem to apply any permission checks if the file whose status is being fetched is a directory. [08:59:03] we may be bypassing the afs_getattr checks, from the check in afs_linux_revalidate [08:59:42] but if I make it always call afs_getattr, I can't even list the directory contents with l and no r [08:59:43] --- Rasmus Kaj has left [09:00:18] Yeh. I think we need to just bypass access checks if what we're accessing is a directory. That's certainly what the fileserver does. [09:01:03] well, if we don't have l... [09:01:44] As far as I can see, the fileserver will still deliver FetchStatus results for a directory, even if you don't have l for the parent. [09:02:44] Whether the readdir will work or not is another matter, though. [09:05:51] okay, adding a '&& vType(vcp) == VDIR' to the afs_linux_revalidate check seems to correctly hide the stat info [09:06:03] we still get the file type, though, I presume from the readdir [09:06:23] We should probably do that in afs_getattr - to match what the fileserver is prepared to tell us. [09:07:23] I'd _really_ like to change the Linux bits of this code, though. If we could make getattr avoid the GLOCK(), then we could hugely improve web server performance. [09:07:24] the restrictiveness should match [09:08:29] > As far as I can see, the fileserver will still deliver FetchStatus results for a directory, even if you don't have l for the parent. isn't that what it should do? conceptually I thought 'l' just got you the mapping of name to fid [09:08:38] if you already have the fid, you might as well have had l [09:10:04] It's not at all clear to me how FetchStatus results should be restricted... All I want to do at present is make the cached case match what happens if you talk to the fileserver. [09:19:04] --- dev-zero@jabber.org has become available [09:28:54] --- dev-zero@jabber.org has left [09:29:11] --- dev-zero@jabber.org has become available [09:56:15] we really ought to replace rxk_Listener with so_upcall. sigh. [10:09:04] --- haba has left [11:15:55] --- jaltman has become available [11:25:19] --- Russ has become available [11:29:11] --- dev-zero@jabber.org has left [11:29:21] --- dev-zero@jabber.org has become available [12:38:31] deason: Probably best not to have a long discussion in gerrit. [12:38:43] (assuming you're reading here ...) [12:38:44] yes, sorry [12:39:10] On Linux, we only lookup when the entry in the dentry cache is invalid. Otherwise, lookups only happen the first time the object is encountered. [12:39:43] So an object may be looked up by user A, which populates a dentry for that name->inode mapping. [12:40:01] User B can then come along and use that dentry, without invoking afs_lookup themselves. [12:41:07] okay [12:42:43] --- haba has become available [12:44:32] I'm just a little confused why this stat-access logic is in afs_lookup; does the 'lookup operation' fill in stat info? [12:45:48] I guess it might on some platforms, where there's a system inode that needs to be populated. [12:50:34] --- jaltman has left: Disconnected [13:43:13] --- Rasmus Kaj has become available [13:59:32] --- jaltman has become available [14:19:55] --- Rasmus Kaj has left [14:35:53] --- summatusmentis has left [14:45:55] --- shadow@gmail.com/owl3F032C83 has left [14:47:33] --- RedBear has left [14:48:15] --- matt has become available [14:48:21] --- dwbotsch has become available [15:06:50] --- dev-zero@jabber.org has left [15:07:07] --- dev-zero@jabber.org has become available [15:16:09] --- deason has left [15:34:14] --- Rasmus Kaj has become available [15:44:20] --- deason has become available [15:46:57] --- dev-zero@jabber.org has left [15:47:03] --- dev-zero@jabber.org has become available [15:47:23] --- dev-zero@jabber.org has left [15:47:27] --- dev-zero@jabber.org has become available [15:52:49] --- dev-zero@jabber.org has left: Lost connection [16:10:07] --- Rasmus Kaj has left [16:19:56] --- Rasmus Kaj has become available [16:21:19] --- dev-zero@jabber.org has become available [16:39:52] --- Rasmus Kaj has left [16:49:51] --- Rasmus Kaj has become available [17:09:47] --- Rasmus Kaj has left [17:11:10] --- matt has left [17:51:07] --- jaltman has left: Disconnected [19:56:45] --- deason has left [20:28:20] --- jaltman has become available [21:37:22] --- summatusmentis has become available [21:38:30] --- shadow@gmail.com/owl8EBEC36C has become available [22:22:47] --- reuteras has become available [22:41:36] --- Rasmus Kaj has become available [22:43:38] --- Rasmus Kaj has left [23:52:36] --- dev-zero@jabber.org has left