=== asac_ is now known as asac | ||
BenC | mjg59: well it stops if I kill acpid (and it happens even at the terminal without X running) | 00:36 |
---|---|---|
BenC | mjg59: Actually, I should have answered "no" since I get two actual keyboard events if acpid is running | 00:37 |
BenC | mjg59: noticed by running showkeys | 00:38 |
cathyal | Hi | 03:24 |
cathyal | is the ubuntu kernel based off the debian kernel | 03:24 |
TheMuso | cathyal: No. | 03:45 |
cathyal | figures | 03:47 |
=== rametux is now known as RameTux | ||
=== rametux_ is now known as RameTux | ||
CarlFK | I can cause a kernel panic - what's the app I run to help with the bug report? | 06:25 |
=== thegodfather is now known as fabbione | ||
=== rametux is now known as RameTux | ||
=== rametux_ is now known as RameTux | ||
=== rametux_ is now known as RameTux | ||
=== Keybuk_ is now known as Keybuk | ||
Kano | hi, did you know about the new ndiswrapper problem: http://www.mail-archive.com/frugalware-git@frugalware.org/msg22366.html | 13:07 |
Kano | i see no fix for that in 8.17, or do i miss something? | 13:08 |
rtg | Kano: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-intrepid.git;a=commit;h=fb54a3e61f44d37b07426f2e7a86d7ca0a99e90a | 13:18 |
Kano | ok, thanks | 13:23 |
NCommander | smb_tp, thanks! (I saw your email) | 13:44 |
Kano | rtg: will you do a 2.6.28 test git soon? | 13:44 |
rtg | Kano: soon, being sometime in the next few weeks. | 13:46 |
=== rametux is now known as RameTux | ||
CarlFK | "kerneloops is a daemon that collects kernel crash information..." does that include panic ? | 14:54 |
rtg | an oops is a panic. | 14:55 |
CarlFK | neat. it's magic. :) | 14:55 |
CarlFK | I am guessing I have to do something? kerneloops.org is not loading | 14:56 |
rtg | amitk is out kerneloops expert | 14:57 |
rtg | s/out/our/ | 14:57 |
amitk | CarlFK: once you isntalled kerneloops, a gnome applet should be running in the background | 15:05 |
CarlFK | amitk: what if I am runing on u-server? | 15:05 |
amitk | CarlFK: no UI? | 15:05 |
CarlFK | right | 15:05 |
CarlFK | well, no gui :) | 15:05 |
amitk | CarlFK: ps aux | grep kerneloops | 15:06 |
CarlFK | if it is a huge problem running it without X, I can load it up - this box is just to hack on this | 15:06 |
CarlFK | ps aux | grep kerneloops = nothing but the grep... | 15:06 |
amitk | just run it manually then | 15:07 |
amitk | with --nodaemon | 15:08 |
amitk | if you want to know what it is doing | 15:08 |
CarlFK | gah - "Connection to cp666 closed by remote host." | 15:11 |
amitk | CarlFK: I guess it was designed mostly for desktop users. | 15:11 |
amitk | cp666? | 15:11 |
CarlFK | I am working with a memory leak, and it kills processes | 15:11 |
CarlFK | compaq 666mhz | 15:11 |
CarlFK | hostname | 15:11 |
amitk | CarlFK: run it manually and let me know what happens. | 15:12 |
* amitk -> steps out for 30mins | 15:12 | |
rtg | amitk: what do we have to do to prep the Jaunty kernel to be a good kerneloops citizen? | 15:44 |
amitk | rtg: add something in the VERSION to identify it as Ubuntu | 15:52 |
amitk | rtg: and info d-i and buildd folks about the same | 15:53 |
rtg | amitk: lets make sure it gets done early, cause its gonna break some stuff, right? | 15:54 |
amitk | rtg: right | 15:54 |
amitk | rtg: who is preparing the tree? you on BenC ? | 15:54 |
amitk | s/on/or | 15:54 |
rtg | haven't started yet. isn't pgraner keeping a list of UDS topics somewhere? | 15:55 |
BenC | amitk, rtg: I'm doing the rebase for jaunty this weekend | 15:57 |
CarlFK | amitk: as memory gets used, processes are being killed. I bet kerneloops gets killed. is there some way to prevent that? (I am guessing best to disable the killing processes process... | 16:42 |
CarlFK | how can I set 80x50 to get more lines on the console ? | 16:45 |
nxvl | hi! | 17:29 |
nxvl | is there any way i can have my own branch in kernel.ubuntu.com? | 17:29 |
nxvl | and how can i create one | 17:29 |
NCommander | nxvl, you need a kernel.ubuntu.com (aka zinc) account | 17:30 |
nxvl | oh | 17:30 |
amitk | CarlFK: use sysctl to write '2' to /proc/sys/vm/overcommit_memory to turn off overcommit. That should keep OOM killer at bay. | 18:20 |
CarlFK | amitk: given I have to keep rebooting the box, what's the kernel param to do it? | 18:30 |
amitk | CarlFK: shove it into /etc/sysctl.conf | 18:30 |
CarlFK | amitk: didn't help: juser@cp666:~$ cat /proc/sys/vm/overcommit_memory ; 2 | 18:41 |
CarlFK | still killed lots of processes before the panic | 18:41 |
amitk | CarlFK: how do you it kills processes? | 18:43 |
CarlFK | amitk: on the console: "out of memory, kill process (123)..." | 18:44 |
CarlFK | I am about to run it in a vm, so I should be abele get screen shots | 18:46 |
amitk | CarlFK: is this a custom app leaking memory? | 18:56 |
CarlFK | amitk: I think it is the vivi v4l2 driver (that is part of vanilla kernel) | 18:57 |
CarlFK | the app I use is the V4L2 sample: caputre.c | 18:57 |
amitk | CarlFK: ohh.. a kernel driver | 18:57 |
CarlFK | yes | 18:57 |
NCommander | amitk, https://launchpad.net/bugs/290153 - do you know if anyone been working on this? | 18:57 |
amitk | NCommander: not that I know of. Make sure ogasawara knows about it. She is our task master. | 19:00 |
amitk | CarlFK: so running the capture app causes the oops? | 19:00 |
NCommander | ogasawara, https://launchpad.net/bugs/290153 - mind taking a look, I know a few people who are having this one | 19:00 |
CarlFK | amitk: yep | 19:01 |
CarlFK | amitk: I am building up docs on it: http://linuxtv.org/v4lwiki/index.php/Test_Suite#memory_leak_2 | 19:01 |
amitk | CarlFK: you might try running ulimit on the capture process to limit its memory | 19:02 |
amitk | man bash -> grep ulimit | 19:02 |
CarlFK | amitk: will that help fix the bug, or just prevent it? | 19:02 |
amitk | CarlFK: prevent the capture process from exceeding its size. If it is the one leaking memory, it should crash first, not other processes. | 19:03 |
CarlFK | amitk: given the memory isn't freed when the app exites, will ulimit help? | 19:04 |
amitk | CarlFK: the app exits correctly? | 19:05 |
CarlFK | amitk: exit(EXIT_FAILURE); line 54 http://linuxtv.org/hg/v4l-dvb/file/b45ffc93fb82/v4l2-apps/test/capture_example.c | 19:06 |
CarlFK | if i run the app 8-12 times (depending on box) I get various flavors of crash that require a reboot | 19:07 |
amitk | CarlFK: can't get into it right now (9pm here). Could you please file a bug with all the info consolidated and how to reproduce it. | 19:08 |
CarlFK | amitk: on lp, right? | 19:11 |
CarlFK | what package? | 19:11 |
amitk | CarlFK: yeah. File it against 'linux' now since you suspect the kernel is oopsing | 19:13 |
CarlFK | amitk: I have a few screen shots of the console - wana see? | 19:13 |
CarlFK | http://dev.personnelware.com/carl/a/qemu_2.png | 19:13 |
CarlFK | 0-4 are the whole sequence | 19:14 |
CarlFK | er, stops at 3. 4 is this chat window... opps. | 19:14 |
amitk | CarlFK: I find it very surprising that turning off overcommit_memory still triggers OOM | 19:28 |
amitk | CarlFK: what is the line you put in sysctl.conf? | 19:29 |
CarlFK | for that I ran sysctl | 19:33 |
CarlFK | amitk: oh hell... I forgot to sudo | 19:34 |
amitk | CarlFK: you don't need sudo there. Just add vm.overcommit_memory = 2 to a new line in the file | 19:35 |
amitk | and reboot | 19:35 |
CarlFK | well... takes a while for qemu to reboot | 19:36 |
CarlFK | so I need to sudo for kerneloops --nodaemon ? | 19:36 |
CarlFK | do I... ? | 19:36 |
amitk | CarlFK: you shouldn't need to. But no harm in it. | 19:38 |
CarlFK | amitk: kerneloops --nodaemon means if the shell is killed, kerneloops goes too, right? | 19:42 |
amitk | CarlFK: yeah | 19:44 |
amitk | CarlFK: kerneloops isn't really a solution to your current problem. If you are able to reproduce this in a VM, then see if you can get a serial console going there. | 19:47 |
CarlFK | amitk: ok. I have done it, but will take a while to figure that out, so get out of here while you have a chance :) | 19:48 |
amitk | heh | 19:51 |
=== thegodfather is now known as fabbione | ||
ogasawara_ | NCommander: thanks for the note about bug 290153 | 20:35 |
CarlFK | Is this the OOM killer, or something else? [ 289.211607] Out of memory: kill process 3980 (sshd) score 2809 or a child [ 289.215599] Killed process 3982 (bash) | 21:47 |
soren | CarlFK: Yes, that's the OOM killer doing its thing. | 21:55 |
CarlFK | thanks. | 21:55 |
soren | CarlFK: It sees you're out of memory and using a set of heuristics, it selects a process and kills it to free memory. | 21:56 |
CarlFK | any idea how to turn it off? | 21:58 |
soren | Stop running out of memory. | 21:58 |
CarlFK | I am trying to gather info about a kernel panic, | 21:58 |
CarlFK | heh - good plan. | 21:58 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!