[00:31] Simon (~sgarner@210.54.177.190) joined #vserver.
[00:51] JonB (~jon@129.142.112.33) joined #vserver.
[01:05] Nick change: riel -> unriel
[01:33] Nick change: Bertl -> Bertl_zZ
[01:50] infowolfe (infowolfe@pcp04891550pcs.frnkmd01.md.comcast.net) left irc: Quit: Trillian (http://www.ceruleanstudios.com)
[02:02] surriel (~riel@66.92.77.98) joined #vserver.
[02:13] JonB (~jon@129.142.112.33) left irc: Quit: Client exiting
[06:11] infowolfe (infowolfe@68.33.215.209) joined #vserver.
[08:05] infowolfe (infowolfe@68.33.215.209) left irc: Quit: Trillian (http://www.ceruleanstudios.com)
[08:05] infowolfe (infowolfe@68.33.215.209) joined #vserver.
[08:59] shuri (~ipv6@207.236.226.187) joined #vserver.
[09:08] shuri (~ipv6@207.236.226.187) left irc: Quit: ipv6
[10:15] infowolfe (infowolfe@68.33.215.209) left #vserver.
[10:36] virtuoso (~shisha@ip114-115.adsl.wplus.ru) got netsplit.
[10:36] ensc (~ircensc@ultra.csn.tu-chemnitz.de) got netsplit.
[10:36] Bertl_zZ (~herbert@MAIL.13thfloor.at) got netsplit.
[10:36] maja|ipv6 (maharaja@ipax.tk) got netsplit.
[10:47] maja|ipv6 (maharaja@ipax.tk) got lost in the net-split.
[10:47] Bertl_zZ (~herbert@MAIL.13thfloor.at) got lost in the net-split.
[10:47] ensc (~ircensc@ultra.csn.tu-chemnitz.de) got lost in the net-split.
[10:47] virtuoso (~shisha@ip114-115.adsl.wplus.ru) got lost in the net-split.
[10:47] virtuoso (~shisha@ip114-115.adsl.wplus.ru) joined #vserver.
[10:47] ensc (~ircensc@ultra.csn.tu-chemnitz.de) joined #vserver.
[10:47] Bertl_zZ (~herbert@MAIL.13thfloor.at) joined #vserver.
[10:47] maja|ipv6 (maharaja@ipax.tk) joined #vserver.
[11:24] say (~say@212.86.243.154) joined #vserver.
[11:24] say_ (~say@212.86.243.154) left irc: Read error: Connection reset by peer
[11:36] <Bertl_zZ> nick Bertl
[11:36] Nick change: Bertl_zZ -> Bertl
[11:36] <Bertl> good morning ;)
[11:44] <virtuoso> Hi Herbert.
[11:44] <Bertl> hi!
[12:58] BobR (~georg@oglgogl.BMTP.AKH-Wien.ac.at) joined #vserver.
[12:58] <Bertl> hi bob!
[13:02] <kestrel_> hello there
[13:02] <Bertl> hi!
[13:02] <kestrel_> hey herbert, how are you?
[13:02] <Bertl> fine, thanks, how are you?
[13:03] <kestrel_> pretty good
[13:03] <kestrel_> been in training for the week, so i've had plenty of time to think and plan :)
[13:04] <kestrel_> my god, training sucks
[13:04] <Bertl> good ;)
[13:04] <kestrel_> what do you do when you're not hacking vserver herbert?
[13:04] <Bertl> I'm currently working at our general hospital ... but usually having fun or doing linux consulting ;)
[13:05] <kestrel_> heh
[13:05] <kestrel_> are you making a living doing linux consulting?
[13:05] <Bertl> hmm .. well not really .. but I probably could if I wouln't do other stuff ;)
[13:06] <kestrel_> damn that other stuff
[13:06] <kestrel_> :)
[13:06] <kestrel_> what do you do at the hospital?
[13:06] <Bertl> curently I'm talking to you ;)
[13:06] <kestrel_> hahaha :)
[13:06] Action: kestrel_ understands
[13:08] <Bertl> well, actually I'm evaluating/installing software to map functional image data (with focus on the brain) on brain templates for statistical evaluations ... (on linux ;)
[13:09] <kestrel_> that sounds like fun
[13:09] <kestrel_> are you evaluating commercial software, free, or both?
[13:09] <Bertl> in this case GPL software ...
[13:30] say (~say@212.86.243.154) left irc: Ping timeout: 483 seconds
[13:34] Simon (~sgarner@210.54.177.190) left irc: Quit: so long, and thanks for all the fish
[14:18] serving (~serving@213.186.189.225) left irc: Ping timeout: 493 seconds
[14:24] Nick change: Bertl -> Bertl_oO
[14:29] Nick change: BobR -> BobR_oO
[14:51] beardfactor10 (~jonathan@212.69.216.20) joined #vserver.
[15:21] BobR_oO (~georg@oglgogl.BMTP.AKH-Wien.ac.at) left irc: Ping timeout: 480 seconds
[15:22] say (~say@212.86.243.154) joined #vserver.
[15:47] BobR_oO (~georg@oglgogl.BMTP.AKH-Wien.ac.at) joined #vserver.
[15:47] Nick change: Bertl_oO -> Bertl
[15:47] <Bertl> hi say!
[15:54] <beardfactor10> so it's linking againts syscall.o
[15:54] <Bertl> okay, can we replace this with the following ...
[15:55] <Bertl> #define __NR_vserver273
[15:55] <Bertl> static inline 
[15:55] <Bertl> _syscall3(int, vserver, uint32_t, cmd, uint32_t, id, void *, data);
[15:55] <Bertl> or the aproriate for the stable version?
[15:55] Nick change: BobR_oO -> BobR
[15:55] <beardfactor10> yup
[15:56] <Bertl> I hope this doesn't change the effect ;)
[15:56] <beardfactor10> I'll be delighted if it does... but also very surprised :)
[15:56] <Bertl> what is the 'idea' behind this program?
[15:57] <Bertl> I mean, did you develop it to 'kill' the system, or was it just an accident?
[15:57] <beardfactor10> it's called "killer.cc" in my tree :)
[15:57] <Bertl> so I assume latter ...
[15:58] Action: beardfactor10 laughs
[15:59] <beardfactor10> I was going to develop into a stress rig to try and reproduce the wild problem
[15:59] <beardfactor10> but it worked with just ht esingle syscall
[15:59] <Bertl> okay, I'll be back in 1-2 hours, please if possible, make the test program independant from other code .. I'll test it then ... okay?
[15:59] <beardfactor10> sure
[15:59] <Bertl> will you be there, later?
[16:00] Nick change: BobR -> BobR_oO
[16:00] <Bertl> okay have to leave now ...
[16:00] Nick change: Bertl -> Bertl_oO
[16:01] infowolfe (infowolfe@68.33.215.209) joined #vserver.
[16:03] <beardfactor10> yes
[16:03] <beardfactor10> for another four hours at least
[16:08] BobR_oO (~georg@oglgogl.BMTP.AKH-Wien.ac.at) left irc: Ping timeout: 480 seconds
[16:10] serving (~serving@213.186.190.195) joined #vserver.
[16:21] <beardfactor10> gtkBitchX-1.0c20cvs+ by panasync - Linux 2.4.21-pre5
[16:23] beardfactor10 (~jonathan@212.69.216.20) left #vserver.
[16:25] beardfactor10 (~jonathan@212.69.216.20) joined #vserver.
[17:04] mhepp (~mhepp@213.211.38.19) joined #vserver.
[17:04] mhepp (~mhepp@213.211.38.19) left irc: Client Quit
[17:18] Nick change: Bertl_oO -> Bertl
[17:19] <Bertl> okay, back, Jonathan?
[17:21] <beardfactor10> working on modifying killer.c for 1.1.5
[17:21] Action: beardfactor10 needs (more) caffeine
[17:22] <beardfactor10> btw herbert, I like the code tidy for 1.1.5
[17:31] <Bertl> okay, you know the syscall interface changed for 1.1.x ..
[17:31] <beardfactor10> hence the modiciations, yes
[17:32] <Bertl> okay, you can get the 'syscall' stuff on my page ...
[17:32] <Bertl> http://vserver.13thfloor.at/Experimental/vkill-0.01.tar.bz2
[17:32] <Bertl> for an example ...
[17:39] <virtuoso> Bertl: I'm sniffing 13thfloor for a test patch for 2.6..
[17:39] <Bertl> hmm, and?
[17:39] <virtuoso> And I can't see it. :)
[17:39] <Bertl> http://vserver.13thfloor.at/Experimental/patch-2.6.0-test9-vs0.01.diff
[17:40] <virtuoso> Thx.
[17:47] mhepp (~mhepp@213.211.38.19) joined #vserver.
[17:47] mhepp (~mhepp@213.211.38.19) left irc: Remote host closed the connection
[17:54] <beardfactor10> grrr
[17:54] Action: beardfactor10 reddens with IRC newbness
[17:55] <beardfactor10> ... same behaviour on 1.1.5 anyhow
[17:58] <Bertl> oaky, via email?
[17:58] <beardfactor10> better yet, I'm adding a vserver "page" to my company vs <gasp>
[18:01] <Bertl> good idea ;)
[18:10] <beardfactor10> http://jonathan.dsvr.co.uk/vserver/killer.c
[18:32] <beardfactor10> any advance on "oh. that's nasty"?
[18:32] <Bertl> well .. yes, I'm currently at ctx: 57757
[18:33] <beardfactor10> on an SMP box?
[18:33] <Bertl> and I expect it to hang at 65535 ;)
[18:35] <Bertl> it's running in an emulation ;)
[18:35] <beardfactor10> ah. which sort of emu? 
[18:35] <Bertl> QEMU 
[18:36] <beardfactor10> i don't think that that's going to be the basis of a viable workaround :)
[18:37] <Bertl> hmm .. well it didn't hang ... ;)
[18:38] <Bertl> but I have a patch for you anyway ... could you try it?
[18:38] <beardfactor10> sure
[18:38] <Bertl> http://vserver.13thfloor.at/Experimental/patch-dynamic-fix-02.diff
[18:39] <Bertl> this is ontop of vs1.1.5 ...
[18:39] <Bertl> what systems do you use to test, and why/how do you stress them?
[18:42] <virtuoso> Bertl: I have 2.6 on my workstation only. I planned to install it at home but didn't manage to yet.
[18:43] <beardfactor10> well, the wild problem occurs on a range of dul processor boxes, from  PIII 450's to Xeon 2.6 Ghz
[18:43] <beardfactor10> in test terms, it's usally lower end systems (same old, same old)
[18:43] <virtuoso> Bertl: Oh, that wasn't for me, sorry. :)
[18:43] <Bertl> no problem ;)
[18:44] <Bertl> I'm always interested in information ...
[18:44] <beardfactor10> usefull stressing is hard
[18:45] <beardfactor10> I've used my wonderful s_context perl module to knock up a script which launches and monitors disk, memory and cpu hogs in various vses
[18:45] <beardfactor10> but that's not live data
[18:46] <Bertl> @jonathan just to refresh my memory, why are you testing this?
[18:46] <Bertl> don't get me wrong, I'm more than happy that you are testing ;)
[18:46] <beardfactor10> the wild problem didn't occur on my test machine (indedd some machines have been running _that_ kernel for months in the wild, whereas a range of others locked up within hours/days)
[18:47] <Bertl> hmm, do you have any statistics, which systems are 'especially' prune to this lock?
[18:48] Action: beardfactor10 giggles
[18:48] <beardfactor10> I kept a grid of this and there was no correlation in hardware
[18:48] Action: Bertl jsut realized why *smile*
[18:49] <Bertl> well, no native speaker, you know ... u/o ...
[18:50] <beardfactor10> my guess is that higher load levels of the syscall activity are an indicator
[18:50] <beardfactor10> that's the main difference between DSVR's usage and vserver userspace utils
[18:50] <Bertl> this sugegst a lock/race of the syscall with itself ...
[18:50] <beardfactor10> yes, but I've not found one
[18:51] <beardfactor10> more to the point, why is it locking the scheduler
[18:51] <Bertl> that is easy to accomplish ... did you test my patch yet?
[18:51] <beardfactor10> I've tried adding traces to the locks involved, buit obviously didn't find anything
[18:51] <beardfactor10> no, not yet
[18:51] <Bertl> please do so if possible ...
[18:52] <beardfactor10> within five minutes
[18:52] <Bertl> this would rule out something, I don't want to analyze further ...
[18:58] <beardfactor10> you're trying to rule our a wrap around scenario?
[18:58] <Bertl> yes, but I guess I already know the reason for those deadlocks ...
[18:59] <beardfactor10> and ... :)
[18:59] <Bertl> well, yes, it is obvious ...
[18:59] <Bertl> but I wouldn't have looked there ...
[18:59] <Bertl> vc_new_s_context()
[18:59] Action: beardfactor10 snorts
[18:59] <Bertl> take a look at the if (ctx == -1) case ...
[19:00] <Bertl> we here take the alloc_ctx_lock (spinlock)
[19:00] <Bertl> and then a few lines later ...
[19:01] <Bertl> switch_user_struct ...
[19:01] <Bertl> which does the alloc_uid
[19:01] <Bertl> which probably sleeps ... 
[19:03] <beardfactor10> ... in the calls to  spin_lock(&uidhash_lock);
[19:04] <Bertl> okay, that's probably the reason for _a lot of_ different lockups ...
[19:04] <beardfactor10> patch-dynamic-fix-02.diff still locks up
[19:05] <Bertl> okay ... I'll medidate over a smart solution ... probably available today or tomorrow ...
[19:05] <Bertl> I would appreciate if you could test it then ...
[19:05] <beardfactor10> cool
[19:05] <beardfactor10> sure
[19:05] <Bertl> I could be wrong on that, but it looks very suspicious ...
[19:06] <Bertl> hmm, actually we can do a simple test for that ...
[19:07] <Bertl> could you for a test, just comment out the spin_lock(&alloc_ctx_lock); and unlock?
[19:08] <Bertl> line 222 and 259 after the patch, and see if this doesn't lock ... it will behave a little strange, but that should not hurt ...
[19:08] <Bertl> in vcontext.c ;)
[19:09] <beardfactor10> okay, recompling
[19:15] <beardfactor10> without that spinlock, killer does NOT lockup
[19:23] <Bertl> sounds good to me ;)
[19:46] <beardfactor10> hmmn....
[19:47] <Bertl> okay ... will be back in 3-4 hours ...
[19:47] Nick change: Bertl -> Bertl_oO
[20:46] kloo_ (~kloo@213.84.79.23) left irc: Remote host closed the connection
[23:07] shuri (~ipv6@207.236.226.187) joined #vserver.
[23:10] <shuri> hello
[23:18] unriel (~riel@66.187.230.200) got netsplit.
[23:18] linas (~linas@67.100.217.179) got netsplit.
[23:18] mcp (~hightower@81.17.110.148) got netsplit.
[23:19] unriel (~riel@66.187.230.200) returned to #vserver.
[23:19] mcp (~hightower@81.17.110.148) returned to #vserver.
[23:19] linas (~linas@67.100.217.179) returned to #vserver.
[23:26] gaertner (~gaertner@212.68.83.129) left #vserver.
[23:33] JonB (~jon@129.142.112.33) joined #vserver.
[00:00] --- Thu Nov 27 2003