/srv/irclogs.ubuntu.com/2013/09/30/#ubuntu-kernel.txt

linuxRhello, I discover total system freezes when working with eclipse (ubuntu 12.04 with all updates installed). can someone help me to analyze this? thanks01:26
=== DavidDuffey is now known as dduffey
=== DarkPlayer_ is now known as DarkPlayer
=== smb` is now known as smb
ppisatimoin07:50
apwppisati, niu07:56
* smb moans07:58
=== fmasi_afk is now known as fmasi
=== fmasi is now known as fmasi_afk
=== fmasi_afk is now known as fmasi
=== fmasi is now known as fmasi_afk
=== fmasi_afk is now known as fmasi
linuxRHi. I experience system freezes in context of intel video and kernel 3.2.0-54-generic (kernel log: [drm:gen6_sanitize_pm] *ERROR* Power management discrepancy: GEN6_RP_INTERRUPT_LIMITS expected 00070000, was 16000000 freeze) ... is this problem being worked on? I think it possibly affects a large number of users09:02
apwlinuxR, is there a bug filed?09:17
apwand what is the symptom ofther than the report in dmesg09:18
linuxRapw, the symptom is a complete system crash (not just X server)09:20
linuxRI think there's a number of bugs related to this: 1194329 , 1168467, http://ubuntuforums.org/showthread.php?t=213552209:23
apwbug #1194329, bug #116846709:23
apwOI ubot2`,09:23
linuxRit seems to be a kernel problem introduced with 3.2.0-4009:24
apwlinuxR, is it easy to reproduce for you ?09:25
linuxRapw, yes. I just need to open a few files in "eclipse" and switch tabs between these files...and boom it goes09:27
apwlinuxR, those errors in other bugs seems to be non-fatal, at least from a running point of view09:27
apwlinuxR, so it is not clear that those error messages are relevant or not09:27
apwlinuxR, therefore, as you have an easy way to reproduce it, then the right thing for you to do09:28
apwlinuxR, is to file a bug from your system, and then we can get a proper bisection started for it09:30
linuxRokay, I will open a bug09:30
linuxRI'd be glad to provide further anaylsis when someone can guide me through09:31
apwget the bug filed, and then lets get a run of 3.2.0-40.64 and 3.2.0-41.65 to confirm the first one is good and second one is bad09:31
apwhttps://launchpad.net/ubuntu/precise/+source/linux/3.2.0-41.6509:32
apwhttps://launchpad.net/ubuntu/precise/+source/linux/3.2.0-40.6409:32
linuxRapw, can I just install all these kernels along each other? how can I select one for boot?09:45
jpdslinuxR: In GRUB.09:49
apwlinuxR, yes you can have many kernels installed they should all appear in grub menus09:55
linuxRcan I just instal older kernels with apt?09:59
linuxRyou referred to a source package?10:00
smblinuxR, you use "dpkg -i *.deb" with those10:02
linuxRsmb, but then I'd have to configure/compile it myself?10:03
smblinuxR, And if you comment out the hidden grub variables in /etc/default/grub and run update-grub you will get a visible selection screen without having to press left-shift10:03
smblinuxR, apw did point to the lp page which has the source but also the build deb packages10:04
apwlinuxR, as smb says, lower down on the pager are per build pages, which one you need depends on your machine10:05
smbYou will need linux-headers...*all.deb, linux headers i386 or amd64 (depedns on what you got installed) the linux-image for that and linux-image-extra if that exists10:05
linuxRokay, will try that, thanks10:06
smblinuxR, Btw, the all headers is in the i386 build only10:07
linuxRwould it not be a good idea to also try a current kernel (e.g. 3.8) and see if the problem was maybe fixed already?10:09
apwlinuxR, you can do that if you wish indeed.  if we want to find the commit which broke it though (which is the easiest way to find out what fixed it later if it is fixed) is to prove the pair of kernels which bracket the breakage10:11
apwas then we can bisect to find the actual failing patch10:11
apwlinuxR, when you have a bug, can we have the bug number please10:11
ppisatibrb10:12
linuxRapw, the bug is: https://bugs.launchpad.net/ubuntu/+bug/1233086 10:30
ubot2`linuxR: Error: launchpad bug 1233086 not found10:30
apwlinuxR, did you file that as a security issue or something?  as it seems to be 'hidden' (assuming you got the bug # right)10:31
linuxRapw, yes...this is security-relevant, isnt it?10:31
linuxRshould I disclose it?10:31
apwlinuxR, i thought your machine was crashing ?10:31
linuxRapw, thats correct. Thats why I thought this could possibly be exploited10:32
apwlocal only though, so not really any more exploitable than any other crasher or oops imo, but your call10:33
apwif its a security issue only the security team can see it10:34
linuxRI'll change that asap10:34
linuxRhttps://bugs.launchpad.net/ubuntu/+bug/123308610:36
apwlinuxR, ok i have put a summary of our discussion on IRC on there, and subscribed our defect analyst who can help with the bisection once you have a pair of kernels where it appeared10:39
apwlinuxR, if you discover an old kernel which it does not appear in, then we have a regression so let me know if you find one10:40
linuxRapw, yes that would be the next steps to do. I'll give an update as soon as I have tested this with other kernels.10:42
=== fmasi is now known as fmasi_afk
=== fmasi_afk is now known as fmasi
=== fmasi is now known as fmasi_afk
=== fmasi_afk is now known as fmasi
=== fmasi is now known as fmasi_afk
linuxRapw, ok I'm ready to install and try the other kernels..but I have difficulties to determine the exact package name to install12:18
linuxRcan you give me a hint?12:19
linuxRapt-cache search 3.2.0-40 yields 34 packages12:20
linuxRlinux-image-3.2.0-40-generic. is that the correct one?12:21
=== fmasi_afk is now known as fmasi
apwlinuxR, smb listed the exact ones to download 12:26
smblinuxR, You probably need 4. uname -m will tell you whether amd64 or i386, uname -r would tell you which flavour but very likely generic. Then12:26
smbi386 build: linux-headers-3.2.0-40_3.2.0-40.64_all.deb 12:27
smblinux-headers-3.2.0-40-<flavour>_3.2.0-40.64_<arch>.deb12:28
smblinux-image-3.2.0-40-<flavour>_3.2.0-40.64_<arch>.deb 12:28
=== fmasi is now known as fmasi_afk
smblinux-image-extra-3.2.0-40-<flavour>_3.2.0-40.64_<arch>.deb 12:28
linuxRokay, thanks. since I'm going to reboot now, I could trigger the freeze again. Is there any additional information I could get from this?12:31
ppisatiapw: i was wrong, qemu doesn't emulate the virtualization extension, so we need real hw12:49
ppisatismb: ^12:49
apwppisati, that is a bit of a shitter12:50
smbppisati, Ok, but yeah would have been nice...12:50
ppisatiapw: either calxeda ecx2000 or samsung exynos512:50
apwppisati, i guess we can use the 'fast' model12:50
ppisatiapw: you mean the arm provided simulator?12:53
ppisatiapw: could be, never tried12:53
smbhenrix, would ext4 issues on online-resizefs related to set_flexbg_block_bitmap trigger any ringing bells on pending stable updates? (apparently in 3.8 but not 3.5 which maybe makes it SEP)12:55
henrixsmb: i don't remember seeing anything related with that, but let me have a look12:57
henrixsmb: do you have a link, sha1, ... ?12:57
smbhenrix, I only saw some inode count overflow things12:57
smbhenrix, https://bugs.launchpad.net/ubuntu/+bug/123308612:57
smberr12:58
smbhenrix, wrong number orry12:58
smbhttps://bugs.launchpad.net/ubuntu/+source/linux/+bug/123307512:58
henrixsmb: heh, i thought so :)12:58
smbhenrix, Hm, though who knows maybe it is related to counts as the grow is large12:59
henrixsmb: yeah, from < 3G to 2TB... that's a huge resize :)13:00
smbhenrix, Apparently it did work in Q. I already asked to try raring release and S to narrow things a bit13:02
linuxRapw, smb : I could not find the additonal kernel in the grub boot menu, although the installation script reported that grub configuration would have been updated..ideas?13:04
smblinuxR, Did you look under "advanced options" or so?13:05
apwlinuxR, what menu options _did_ you have13:05
smbThe older kernels get sorted in there (the naming is a bit .. unhelpful)13:05
henrixsmb: do you have any idea on the kernel version where this issue was seen?13:06
linuxRapw, just the current kernel (3.2.0-54-generic)13:06
smbhenrix, Well, just "latest from today" for R13:06
henrixsmb: ah, ok. i was looking at 3f8a6411fbada1fa482276591e037f3b1adcf55b but this is already in R for some time13:07
linuxRi'll ckeck again13:08
smblinuxR, The advanced menu does not show any kernel versions nitially13:08
smbjust when selecting it13:08
apwlinuxR, yeah "ubuntu" boots direct and "ubuntu advanced" is a menu13:09
apwit is most ... interesting semantics wise13:09
linuxRi'll check..thx13:09
=== fmasi_afk is now known as fmasi
rtgapw, pushed -rc3 rebase. thanks for the overlayfs fixes.14:29
apwrtg, np14:30
caribouapw: FYI, the mempolicy bug number is 123317514:30
caribouapw: bug: #123317514:31
* apw pokes ubot2` in the eye14:31
xnoxrtg: goldfish kernel now boot fine, had to fix up qemu =)14:31
caribouapw: :)14:31
rtgxnox, cool14:31
xnoxrtg: (armhf that is, haven't started on x86 yet)14:31
caribouapw: one interesting fact about this one is that servers running lucid with similar workload *do* *not* show this behavior14:32
apwcaribou, did you manage to repro it ?14:32
caribouapw: no, not yet14:32
caribouapw: but I was looking at Lucid's __mpol_put() and it is sensibly simpler14:32
apwcaribou, t14:33
apwcaribou, there is even less locking in lucid than in precise, i a supprised it is not worse14:33
caribouapw: nm, they're both the same :-/14:34
apwcaribou, well there is a task lock round it in precise, which would likel change timing, perhaps it makes things more likely14:35
ppisatiapw: it seems we need a license to get the RTSM14:38
ppisatiapw: so no luck14:38
diwicRead The Sucking Manual ?14:38
ppisatiReal-Time System Model14:39
ppisati"RTSMs are simulation models of ARM Hardware Platforms..."14:39
ppisatietcetc14:39
diwicok14:39
linuxRapw, smb : I'm working on  3.2.0-40-generic now for some hours, have not been able to trigger the error again since14:53
linuxRstill see the " [drm:gen6_sanitize_pm] *ERROR* Power management discrepancy" message in kernel log..possibly unrelated to the freeze problem14:54
* ppisati goes out for a bit15:03
apwlinuxR, that is as waht i suspected might happen, so now you need to step forward to the next version and see if it breaks, and keeps doing that till it does,15:05
riso64bithello15:08
jsalisburylinuxR, I can help you with the bisect.  I'm going to post some comments and kernels to test in the bug.15:11
caribouIs it possible to 'simulate' a NUMA architecture on a KVM virtual machine using non-NUMA hardware ?15:17
apwriso64bit, welcome15:17
apwcaribou, hmm it might be, but not something i have ever tried, and you won't get the real effects clearly if you don't have a real numa under the hood15:18
caribouapw: I just want to exercise the numa-aware code using numactl15:18
riso64biti have a problem whit the driver of touchpad on HPMini 210 & I have already open a ticket on launchpad (_https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1188965) but Christopher M. Penalver (penalvch) write me to contact the Driver manteiner for help to fix the problem (https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1188965/comments/95).Yesterday i have sent the mail but i don't have received any 15:19
riso64bitresponse and so I don't now if i have made the right procedure.15:19
apwcaribou, not sure is my best answer then :)  smb might have played so15:19
riso64bitsorry i'm not English15:19
riso64bit:D15:19
caribouapw: ok, I'll try to ping him, thanks15:19
smbapw, caribou Not exactly numa. Its possible to define the layout through libvirt cores/threads/cpu ...15:20
apwit might be enough to be usable for caribou's purposes then15:21
caribousmb: yep, I did that15:21
caribouok, I'll dig that out, thanks15:21
smbYeah, probably if two sockets automatically means NUMA like15:21
apwriso64bit, i'd not be expecting a response necessarily very quickly one doesn't even know if they are 'at work' any particular day15:22
riso64bityes i know, but i don't know if i have made the right procedure :D15:23
apwriso64bit, there is little correct proceedure with upstream, other than donning a fire proof suit before emailing15:25
smbcaribou, There seems to be -numa for kvm... Question is whether and if how libvirt would do things. But you should be able to look for the qemu process that gets started15:25
caribousmb: oh, thanks I'll check that out15:26
riso64bitok, i will wait. 15:27
smbcaribou, Hm, maybe if you hand tune the xml through virsh (see http://www.libvirt.org/formatdomain.html#elementsCPU)15:36
caribousmb: that looks good :<numa><cell... maybe what I'm after15:37
caribousmb: thanks15:38
=== rtg is now known as rtg-afk
=== fmasi is now known as fmasi_afk
=== fmasi_afk is now known as fmasi
=== kentb-out is now known as kentb
=== fmasi is now known as fmasi_afk
apwppisati, do you get any spinning irg/* processses on your omap4 with -generic16:43
ppisatiapw: not that i recall16:48
apw   76 root     -51   0     0    0    0 S  24.8  0.0   1188:03 irq/88-48070000                                          16:48
apw   84 root     -51   0     0    0    0 R  23.1  0.0   1134:30 irq/151-twl6040                                          16:48
apwppisati, ^16:49
ppisatiapw: let me setup my board16:50
apwppisati, i am going to try rebooting it to see if it is transient, i did get an L3 error over the weekend16:50
apwppisati, yeah things look a lot better following a reboot, i wonder if the L3 error handler did not clean up right16:52
=== jdstrand_ is now known as jdstrand
=== psivaa is now known as psivaa-afk
apwtjaalton, hey ... this 2 minutes to get dash thing with atoms and similar, are we expecting to do anything with ti17:10
apwtjaalton, bug #122260217:11
tjaaltonapw: still no word from upstream, but maybe the change could be reverted or make it so that for gen3 it forces the opengl version to 1.417:19
apwtjaalton, either sounds appealing over where we are17:20
tjaaltonheh, sure17:20
apwtjaalton, though i will say =1.4 does not make dash pretty, if appearing same day17:20
tjaaltonwith the workaround?17:25
apwtjaalton, yeah with the workaround the dash is not the right colours so teh gallback is not working quite right17:26
Sarvattapw: can you run driconf and pastebinit the ~/.drirc it creates?17:26
apwSarvatt, with or without the workaround17:27
Sarvattdoesn't matter17:27
tjaaltonhmm ok17:27
tjaaltonwell force it to whatever it was before :)17:27
apwFloating point exception (core dumped)17:27
apwquality s/w17:28
Sarvatthah17:28
Sarvattapw: ah nevermind, they ripped out the options completely17:28
apwSarvatt, Floating point exception (core dumped)17:29
apwarse17:29
Sarvatttjaalton: thats pretty damn safe to just revert17:29
apwSarvatt, http://paste.ubuntu.com/6176427/17:29
Sarvattapw: sorry about that, I didn't see that they ripped out the options that forced it to 2.1 completely so they aren't in there to disable17:30
apwSarvatt, heh17:30
tjaaltonSarvatt: ok then.. still would be good to know what upstream was thinking.. :/17:34
* apw seculates ... la la la hasn't that shit h/w died yet ?17:34
tjaaltonSarvatt: was it idr?17:34
Sarvatti915g supports it and we dont give a crap about i915 anymore, lets just do it17:34
=== fmasi_afk is now known as fmasi
Sarvattwhipping up a patch to revert it now, files got shuffled around a bit17:35
tjaaltonwell i guess it allows the hw to pass a few piglit tests more, even if they take an eternity to run :)17:35
tjaalton^ motivation17:36
=== fmasi is now known as fmasi_afk
Sarvatttjaalton: ok to push this to git? http://paste.ubuntu.com/6176506/17:47
apwSarvatt, if you have some binaries to test, i am happy to do so17:48
Sarvattsure thing, i'll throw it in a ppa17:48
tjaaltonSarvatt: yup go ahead17:49
Sarvattoh nice, i already have a ppa named apw :)17:51
apwSarvatt, :)17:51
Sarvattapw: its uploading to https://launchpad.net/~sarvatt/+archive/apw now, should be about 45 minutes till its built17:51
apwSarvatt, will watch out for it17:52
apwSarvatt, 'maverick' packages ... an old one17:52
Sarvatt148 weeks, wonder what it was about17:53
Sarvattahh right, actually sandybridge acceleration for 10.10 where it was disabled in the archive17:53
Sarvattdang, 38 minute builder queue even at urgency=critical17:55
apwSarvatt, i'll slurp it down and try a local build then18:04
Sarvattapw: amd64 machine?18:04
Sarvattif you have multiarch mesa installed thats gonna be a headache and dont bother18:05
apwSarvatt, this is a 32 bit atom, so i think not18:05
apwSarvatt, and by local i mean a builder build on my kernel builder18:05
=== rtg-afk is now known as rtg
apwSarvatt, that FTBFS for me18:30
tjaaltonlog?18:31
apwtjaalton, http://paste.ubuntu.com/6176677/18:32
apwSarvatt, doesn't that need like {} round the case 3: contents if you have variables in it18:36
rkrishnaHi, I am getting error "unable to handle kernel paging request" trying to install a video capture card, any ideas?19:04
Sarvattapw: its built in the ppa, sorry about all that20:53
linuxRapw, jsalisbury: I tried to track down the error to a specific kernel version. I thought 3.4.42 was stable (working for multiple hours) ;tried 3.4.44 which crashed few minutes after boot, then tried 3.4.42 again which then also crashed almost instantly. Any ideas?21:10
* rtg -> EOD21:55

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!