Back to Home / #uml / 2007 / 12 / Prev Day | Next Day
#uml IRC Logs for 2007-12-18

---Logopened Tue Dec 18 00:00:07 2007
01:47|-|balbir [~balbir@59.145.136.1] has joined #uml
07:41|-|cl4sh [~cl4sh@qik.ds.pg.gda.pl] has joined #uml
07:45|-|cl4sh [~cl4sh@qik.ds.pg.gda.pl] has quit []
08:20|-|balbir [~balbir@59.145.136.1] has quit [Ping timeout: 480 seconds]
08:39|-|dang [~dang@nemesis.fprintf.net] has quit [Quit: Leaving.]
08:48|-|real0ne [~test@41.251.5.131] has quit [autokilled: We do not support illegal activity on this network. If you feel an error has been made, please contact support@oftc.net. (2007-12-18 13:48:20)]
08:51|-|krau [~cktakahas@200.184.118.132] has quit [Ping timeout: 480 seconds]
09:04|-|krau [~cktakahas@200.184.118.132] has joined #uml
09:04|-|dang [~dang@aa-redwall.nexthop.com] has joined #uml
09:39|-|Magotari [~karol@chello089076073248.chello.pl] has quit [Ping timeout: 480 seconds]
09:55|-|karol [~karol@chello089076073248.chello.pl] has joined #uml
09:59|-|dan1 [~dang@aa-redwall.nexthop.com] has joined #uml
09:59|-|dang [~dang@aa-redwall.nexthop.com] has quit [Read error: Connection reset by peer]
11:22|-|jdike [~jdike@pool-72-93-105-51.bstnma.fios.verizon.net] has joined #uml
11:22<jdike-#uml->>Hi guys
11:33|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has joined #uml
11:38[~]jdike #uml zeros in on the execshield patch#uml-> zeros in on the execshield patch
12:09|-|Infinito [argos@201-3-115-208.gnace701.dsl.brasiltelecom.net.br] has joined #uml
12:16<karol-#uml->>My super dirty hacked in feature is working. When you don't specify an umid when running uml_mconsole, and only one uml is running, it switches to it by default.
12:17<karol-#uml->>To add: Multiple uml commands at once, different switching for when only two umls are running, notices about new uml and old uml becoming inactive.
12:17<karol-#uml->>Great practice. Useless, but we all start somewhere...
12:27<jdike-#uml->>those all sound useful
12:28<peterz-#uml->>I'd love it when uml_mconsole would just work
12:28<peterz-#uml->>one instance is all I ever run anyway
12:29<jdike-#uml->>peterz, I don't see it being that big a deal
12:29<karol-#uml->>'./myco' is all you need to auto-get control if one instance is up.
12:29<peterz-#uml->>jdike: nah, I have it scripted to look up the umid
12:29<jdike-#uml->>shell history eliminates msot of the typing if you always use the same umid
12:30<peterz-#uml->>yeah, I should add a umid to my ./linux script
12:32<karol-#uml->>jdike: Kernel panic - not syncing: switch_mm_skas - PTRACE_SWITCH_MM failed, errno = 22
12:32<karol-#uml->>This is the failure I get with skas3.
12:32<karol-#uml->>skas0 actually works sometimes.
12:37|-|ctrace [~chris@barfly.umd.edu] has joined #uml
12:37<jdike-#uml->>karol, that was the failure with the original skas3 patch on 2.6.23
12:37<jdike-#uml->>before I fixed it
12:38<karol-#uml->>karol@BlackBox ~/expe $ uname -a
12:38<karol-#uml->>Linux BlackBox 2.6.23skas3-v9-pre9-skas3-v9-pre9 #3 SMP Thu Dec 13 19:36:33 CET 2007 i686 Intel(R) Pentium(R) 4 CPU 3.40GHz GenuineIntel GNU/Linux
12:38<karol-#uml->>karol@BlackBox ~/expe $
12:38<karol-#uml->>This is with the latest and the greatest skas3 from uml-devel.
12:38<jdike-#uml->>yeah
12:39<jdike-#uml->>you're using BB's skas3 patch, not mine
12:39<jdike-#uml->>I think
12:39<karol-#uml->>Oh? I'm quite positive about it being the uml_devel. Maybe I am confusing kernels. Gonna just redo it cleanly.
12:39<jdike-#uml->>you're sure that's the one I sent out?
12:40<karol-#uml->>80% sure. But we will know for 100% sure once I redo it all.
12:40<ctrace-#uml->>hiya, i noticed i have been able to repeatedly get a UML instance to crash when I do something that triggers the bell (e.g. holding down TAB in a shell)
12:40<jdike-#uml->>stack trace and version?
12:40<ctrace-#uml->>strace'ing the hanging process reveals:
12:40<ctrace-#uml->>ioctl(0, SNDCTL_TMR_TIMEBASE or TCGETS, {B38400 -opost -isig -icanon -echo ...}) = 0
12:40<ctrace-#uml->>ioctl(0, SNDCTL_TMR_CONTINUE or TCSETSF <unfinished ...>
12:40<jdike-#uml->>crash or hang?
12:40<ctrace-#uml->>i am using 2.6.23.9 for the uml kernel and 2.6.18-4-686 for the host kernel (stock debian 4.0 box)
12:41<ctrace-#uml->>hmm i guess i would say hang, the process sticks around on the host..
12:41<karol-#uml->>Do you get a stack trace/panic printed?
12:41<karol-#uml->>If yes, crash. If no, hang.
12:41<ctrace-#uml->>i was able to attach gdb to the process and get a trace, yup
12:41<ctrace-#uml->>the first few lines are:
12:42<ctrace-#uml->>#0 0xb7f04410 in ?? ()
12:42<ctrace-#uml->>#1 0x0822fbc0 in cpu0_irqstack ()
12:42<ctrace-#uml->>#2 0x0822fb90 in cpu0_irqstack ()
12:42<ctrace-#uml->>#3 0x00005404 in ?? ()
12:42<ctrace-#uml->>#4 0xb7e8fab6 in tcsetattr () from /lib/tls/i686/cmov/libc.so.6
12:42<ctrace-#uml->>#5 0x0805ad17 in fd_close (fd=0, d=0x8908198) at arch/um/drivers/fd.c:72
12:42<jdike-#uml->>and that last ioctl just hangs?
12:42<ctrace-#uml->>yup for the longest time i could not re-create
12:42<ctrace-#uml->>then i realized it was related to the bell, and by holding down TAB and triggering the bell at a shell prompt i can now trigger it repeatedly
12:42<jdike-#uml->>what are the next few lines above that?
12:43<ctrace-#uml->>here's the next 5:
12:43<ctrace-#uml->>#6 0x0805b129 in close_one_chan (chan=0x828e708, delay_free_irq=1) at arch/um/drivers/chan_kern.c:297
12:43<ctrace-#uml->>#7 0x0805b155 in close_chan (chans=0x8231710, delay_free_irq=1) at arch/um/drivers/chan_kern.c:313
12:43<ctrace-#uml->>#8 0x0805b84d in chan_interrupt (chans=0x8231710, task=0x823172c, tty=0x869a050, irq=2) at arch/um/drivers/chan_kern.c:673
12:43<ctrace-#uml->>#9 0x0805bbfb in line_interrupt (irq=2, data=0x828e708) at arch/um/drivers/line.c:30
12:43<ctrace-#uml->>#10 0x0808d9f9 in handle_IRQ_event (irq=2, action=0x87ee7c0) at kernel/irq/handle.c:140
12:43<ctrace-#uml->>if more would help just let me know ;-)
12:45<jdike-#uml->>OK, give me a few minutes
12:48|-|tyler29 [~tyler@ARennes-257-1-52-67.w81-53.abo.wanadoo.fr] has joined #uml
12:50<ctrace-#uml->>there is probably one other thing i should mention -- after I boot a UML instance, I am running a command like 'screen -dmS narb1 /dev/pts/1' on the UML host..
12:50<ctrace-#uml->>then logging into the UML instance using 'screen -r narb1', then I hold down TAB to make the bell ring
12:51<ctrace-#uml->>(i am going to try to force the crash w/o screen being involved now..)
12:51|-|balbir [~balbir@122.167.205.122] has joined #uml
12:51<jdike-#uml->>does the host show the UML as being busy or idle?
12:53<ctrace-#uml->>in terms of CPU usage ?
12:53<jdike-#uml->>yes
12:53<ctrace-#uml->>one sec, will check
12:58<ctrace-#uml->>ok sorry took a few mins to force the crash that time
12:58<ctrace-#uml->>cpu usage of PID ~/.uml/narb1/pid is very low, about 1.7%
12:58<jdike-#uml->>OK
12:58<jdike-#uml->>and when you strace it, it's just sitting in that tcsetattr?
13:00<ctrace-#uml->>yes, actually we had a discussion awhile ago (my nick was chtracymax) and i think this problem may actually be very similar, perhaps it is something to do with the process that is consuming the kernel console output...
13:01<ctrace-#uml->>this was back around Nov 2006 on this channel
13:01<ctrace-#uml->>strace just sits there tho, yes..
13:01<jdike-#uml->>other people have seen hangs when the main console is a pts device, they are attached to it with screen, and something generates console output
13:01<jdike-#uml->>but in that case, UML is spinning
13:01<jdike-#uml->>that's a host bug, and there's a UML workaround for it
13:03|-|Infinito [argos@201-3-115-208.gnace701.dsl.brasiltelecom.net.br] has quit [Quit: Quitte]
13:04<ctrace-#uml->>hmm we are using python to manage instances to create virtual networking environments, python launches the instances, logs kernel output, then runs a screen command on the host so we can attach to it
13:04<jdike-#uml->>that exactly the sort of environment in which I would expect to see this problem
13:04<ctrace-#uml->>i am going to launch the UML instance manually from the shell and w/o screen
13:05<ctrace-#uml->>i wonder what is being output on the console? :-)
13:08<ctrace-#uml->>i believe you mentioned that one workaround may be to tell syslog not to send stuff to the console ?
13:09<jdike-#uml->>no
13:23<ctrace-#uml->>thanks for your help, will do a bit more investigating here, my initial attempts do seem to show that i cannot force the crash when i launch the UML instance manually from the shell
13:24<jdike-#uml->>try this patch to UML
13:24<jdike-#uml->>http://rafb.net/p/0TxZrn43.txt
13:25<ctrace-#uml->>will do, thanks for the pointer!
13:49|-|balbir [~balbir@122.167.205.122] has quit [Ping timeout: 480 seconds]
14:20<karol-#uml->>*sigh* You know you are bored when you enter 'quietness' as a variable name. To this date some of my stuff has variables like 'buddha' in it...
14:21|-|Infinito [argos@200-101-124-139.gnace701.dsl.brasiltelecom.net.br] has joined #uml
14:34|-|tyler29 [~tyler@ARennes-257-1-52-67.w81-53.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
14:50|-|tyler29 [~tyler@ARennes-257-1-4-61.w81-250.abo.wanadoo.fr] has joined #uml
14:57|-|Infinito [argos@200-101-124-139.gnace701.dsl.brasiltelecom.net.br] has quit [Quit: Quitte]
15:12|-|tyler29 [~tyler@ARennes-257-1-4-61.w81-250.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
15:18<karol-#uml->>(KjLHbm) * config ubda
15:18<karol-#uml->>stoptest --> OK kilogram.boom
15:18<karol-#uml->>KjLHbm --> OK arch.raw
15:18<karol-#uml->>(KjLHbm)
15:20<jdike-#uml->>what's that doing?
15:20<karol-#uml->>Sending a command to all live umls.
15:20<jdike-#uml->>ah
15:20<jdike-#uml->>neat
15:21<karol-#uml->>Very buggy at this point, and make of hacks on top of hacks. But this is a prototype.
15:21<karol-#uml->>I know from experience, I am never able to code something good the first time around.
15:21<jdike-#uml->>no one is
15:28|-|tyler29 [~tyler@ARennes-257-1-74-95.w81-53.abo.wanadoo.fr] has joined #uml
15:35<karol-#uml->>jdike: If you want I can send you my changes after I am done. The only problem I see is that the first thing I do after a cvs checkout is running "indent -kr -i8". Also I am going to change quite a bit in the code.
15:35<jdike-#uml->>yeah
16:25<jdike-#uml->>gawd
16:25[~]jdike #uml is way too clever for his own good#uml-> is way too clever for his own good
16:29<karol-#uml->>What happened?
16:30<jdike-#uml->>can_do_skas3 is supposed to check for skas3 support on the host
16:30<jdike-#uml->>so I added if(!can_do_skas4()) can_do_skas3()
16:30<jdike-#uml->>and UML starts dying horribly
16:31<jdike-#uml->>it turns out there was a little critical bit of initialization done under can_do_skas3
16:31<jdike-#uml->>which I put there because it was convenient
16:31<jdike-#uml->>and then forgot about
16:32<jdike-#uml->>to make matters worse, the symptoms were possibly causable by bugs in the skas4 code I'm working on
16:32<karol-#uml->>Ouch.
16:33<jdike-#uml->>so I spent time investigating that angle
16:36<karol-#uml->>I imagine.
16:42<fo0bar-#uml->>jdike: that patch from yesterday (re: 32-bit guest on 64-bit host) worked fine, thanks
16:42<jdike-#uml->>figured sop
16:42<jdike-#uml->>so
16:44<fo0bar-#uml->>jdike: is IA32_EMULATION not implemented because of lack of user interest, and/or horrible horrible complexity to do?
16:44[~]fo0bar #uml is starting to look at the company's future x86_64 strategy, if it's not obvious :)#uml-> is starting to look at the company's future x86_64 strategy, if it's not obvious :)
16:45<jdike-#uml->>hehe
16:45<jdike-#uml->>it's not horrible
16:45<jdike-#uml->>I just don't get enough complaints about it
16:48<fo0bar-#uml->>it hasn't happened before, but I just anticipate if we switch to 64-bit, all of a sudden we'll need a to support a 32-bit binary blob somewhere
16:49<fo0bar-#uml->>plus it would be useful to only maintain a 64-bit guest kernel, and if I needed to move a physical 32-bit machine's userland to UML, while using the 64-bit kernel
16:49<jdike-#uml->>no arguments it's useful
16:52<fo0bar-#uml->>yeah. and I'm not asking you to run out and implement it, but maybe I'll follow up on the mailing list and see if there's interest or anyone offering coding support
16:57<jdike-#uml->>I actually have a very old patch that does it
16:57<jdike-#uml->>it needs some serious integration and updating
16:59<fo0bar-#uml->>from the points where IA32_EMULATION is referenced (I'm looking 2.6.23 now), it looks like the x86 merge will make that even more fun
17:00<jdike-#uml->>the work should all be in arch/um
17:00<jdike-#uml->>a lot of that code is generic and just needs to be hooked up
17:00<fo0bar-#uml->>oh right
17:00<jdike-#uml->>see fs/compat_ioctl.h
17:01<jdike-#uml->>whoops, .c
17:34|-|Infinito [argos@200-101-125-12.gnace701.dsl.brasiltelecom.net.br] has joined #uml
17:41|-|dan1 [~dang@aa-redwall.nexthop.com] has quit [Ping timeout: 480 seconds]
17:53|-|tyler29 [~tyler@ARennes-257-1-74-95.w81-53.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
18:38|-|dang [~dang@nemesis.fprintf.net] has joined #uml
19:23|-|Micksa [~mslade@203.55.194.182] has quit [Ping timeout: 480 seconds]
20:30|-|balbir [~balbir@122.167.205.122] has joined #uml
20:47|-|Infinito [argos@200-101-125-12.gnace701.dsl.brasiltelecom.net.br] has quit [Quit: Quitte]
21:17|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
21:38|-|jdike [~jdike@pool-72-93-105-51.bstnma.fios.verizon.net] has quit [Quit: Leaving]
21:45|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has joined #uml
21:59|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
22:04|-|balbir [~balbir@122.167.205.122] has quit [Ping timeout: 480 seconds]
22:58|-|aindilis [~aindilis@75.146.96.197] has quit [Remote host closed the connection]
23:59|-|VS_ChanLog [~stats@ns.theshore.net] has left #uml [Rotating Logs]
23:59|-|VS_ChanLog [~stats@ns.theshore.net] has joined #uml
---Logclosed Wed Dec 19 00:00:48 2007