[00:04:57] --- kaj has left [00:14:31] --- kaj has become available [00:23:43] --- Russ has left: Disconnected [00:50:18] --- abo has left [01:11:34] --- jaltman has left: Replaced by new connection [01:11:35] --- jaltman has become available [01:27:41] --- abo has become available [02:29:01] --- haba has left [02:43:06] --- Jeffrey Altman has become available [03:22:22] --- Simon Wilkinson has left [04:07:20] --- kaj has left [04:09:58] --- kaj has become available [05:16:12] --- haba has become available [07:08:52] --- deason has become available [07:09:25] --- Simon Wilkinson has become available [07:25:17] --- Simon Wilkinson has left [07:37:20] --- dev-zero@jabber.org has become available [07:37:48] --- dev-zero@jabber.org has left: offline [07:48:29] --- haba has left [07:58:54] --- reuteras has left [07:59:01] --- Simon Wilkinson has become available [08:02:50] --- Simon Wilkinson has left [08:39:26] --- jaltman has left: Disconnected [08:43:53] --- clc31 has become available [08:44:33] --- clc31 has left [09:03:39] --- Simon Wilkinson has become available [09:31:48] --- jaltman has become available [09:43:41] --- jaltman has left: Disconnected [09:46:33] --- kaj has left [09:48:08] --- jaltman has become available [09:59:02] --- Simon Wilkinson has left [10:02:32] --- Simon Wilkinson has become available [10:02:51] --- meffie has become available [10:12:41] --- Simon Wilkinson has left [10:19:23] --- Simon Wilkinson has become available [10:28:26] --- Kevin Sumner has become available [10:47:24] --- Russ has become available [11:00:07] jaltman: i managed to get a test case perl script that can trigger the issue in RT125192. i'm trying this with libraries linked against the two git changes you pointed me at. [11:01:22] if i can figure out how to append to rt tickets, i'll throw the script on it for reference [11:01:42] never mind, i just dumped core. [11:01:55] kula: I can probably make it so you can add to that RT ticket, if you like. [11:01:56] --- kaj has become available [11:02:29] the script would be very useful [11:02:39] simonwilkinson: that would be handy, i can at least provide the script [11:03:01] jaltman: would you like the core dump i just got? if so, i can get it to you after the meeting i'm in [11:03:11] yay! [11:03:22] kula: You can already - you're the requestor. [11:03:31] a backtrace for the core as a irst step is fine [11:04:00] i'll throw that in the ticket as wlel. but now, meeeting. [11:04:42] You need to send an email to openafs-bugs@openafs.org with a subject containing [grand.central.org #125192] [11:04:42] (exactly as written there, including the square brackets) [11:05:21] will do [11:05:46] Seeing all the things MIT have broken by turning off DES is quite amusing. [11:06:05] We needn't fear about folks rushing to other filesystems, given that they've broken NFSv4 and Samba too :) [11:06:32] --- Jeffrey Altman has left: Disconnected [11:07:57] --- deason has left [11:08:31] was it NFSv4 or NFSv3 with Kerberos auth? [11:08:32] --- deason has become available [11:09:45] --- dev-zero@jabber.org has become available [11:10:10] NFSv4 only does DES on Linux. [11:10:39] Getting anything stronger than DES into the kernel has stalled for reasons that aren't entirely clear - but Umich produced the code several years ago. [11:10:54] nice :) [11:11:27] they have assumed that because AFS only does fcrypt that DES is sufficient [11:11:31] :) [11:12:26] It'll be interesting to see what the vendors do - I know Debian shipped with DES disabled (hence all of this fallout). I suspect other vendors will turn it back on. [11:25:21] --- Simon Wilkinson has left [12:00:07] In Debian, we're going to patch aklog to re-enable DES for it. [12:00:11] You can't do that with gssd. [12:00:15] It will be interesting to see what happens there. [12:02:29] --- Simon Wilkinson has become available [12:10:24] --- dev-zero@jabber.org has left [12:10:29] --- dev-zero@jabber.org has become available [12:45:08] --- dev-zero@jabber.org has left [12:45:21] --- dev-zero@jabber.org has become available [13:10:22] what is debian doing about kerberized telnet? does it not distribute it? [13:11:48] You'd have to ask Sam. [13:12:17] I would imagine they'll ship with whatever ships in the Kerberised application bundle that MIT's been trying to offload for ages. [13:27:07] that package is useless without DES since MIT does not support [13:27:11] 3DES ENCRYPT [13:40:33] --- Simon Wilkinson has left [13:43:51] --- abo has left [13:44:25] --- abo has become available [13:44:38] We're shipping what's currently in the app tree. [13:44:45] It works if you set allow_weak_crypto to true. [13:45:03] So far as I know, there isn't anyone involved in Debian package maintenance who really cares about Kerberized telnet. (I certainly don't.) [13:56:30] --- haba has become available [14:38:00] kula: can you tell me what the value of p->flags is on the packet in rxi_AllocPacketTSFPQ [14:49:20] gimmie a sec [14:50:36] assuming you mean frame 0 of thread 1, p->flags is "$1 = 2 '\002'" [14:54:05] actually, I was thinking of frame 6 from Thread 2 since that is where the osi_Panic is being called from [14:56:02] i'm not seeing anything that is obviously a packet in that frame --- is it something buried in rx_ts_info? [14:58:37] the panic is being called from within nest macros which are referencing a stack variable "struct rx_packet *p" [14:58:57] I suspect the variable might be optimized out by the compiler [14:59:16] either that, or my gdb-fu is weak [15:10:25] what I think is going on in this core is that rx_Listener is running before rx_Init and rx_InitHost have completed. As a result, thread safety is questionable. [15:11:07] --- Simon Wilkinson has become available [15:11:07] This may simply be a problem with the AFS Perl wrapper [15:11:42] --- mdionne has become available [15:12:17] Problems with the AFS perl wrapper wouldn't surprise me. [15:15:26] However, what should stop rx_Listener from starting before rx_Init has completed? [15:37:34] it looks like there is a separate rx_Init call for each of the operation classes. (pts, vos, bos, ...) which should be fine if the locking is working properly [15:46:26] Simon: that is a very good question [15:50:12] yeah, i'm pretty sure there are issues with the AFS perl wrapper. whether it's causing this problem or not, i do not know, but i'd certainly do the module differently. [15:50:44] I took a quick look at some of its code, and it doesn't seem to start up RX any differently than any other client would. [16:11:33] --- Simon Wilkinson has left: Lost connection [16:11:33] --- dev-zero@jabber.org has left: Lost connection [16:12:15] --- deason has left [16:12:36] --- Simon Wilkinson has become available [16:16:42] --- dev-zero@jabber.org has become available [17:12:52] I think our current definition of xdr_free() may be a mistake. It has a different signature than the 'normal' xdr_free, and the normal one's behaviour (of recursively freeing an XDR structure) would actually be of more use. [17:32:10] --- dev-zero@jabber.org has left [18:21:45] --- Russ has left: Disconnected [18:22:39] --- meffie has left [18:30:04] --- jaltman has left: Disconnected [18:30:40] --- jaltman has become available [18:40:47] --- Russ has become available [18:56:21] --- mdionne has left [19:04:52] --- mdionne has become available [19:36:00] --- mdionne has left [19:36:00] --- mdionne has become available [19:36:04] --- mdionne has left [20:34:51] --- deason has become available [22:15:10] --- reuteras has become available [22:15:13] --- deason has left [22:52:08] --- kaj has left