/srv/irclogs.ubuntu.com/2011/05/17/#ubuntu-kernel.txt

eruditehermithey, does anyone know why amd64 versions are no longer being built in the kernel ppa?00:30
eruditehermithey04:13
eruditehermitdoes anyone know why the ubuntu kernel ppa doesn't have amd64 builds for the newer kernels?04:13
BenCericm|ubuntu: could be a build failure04:18
BenCeruditehermit: ^^04:24
eruditehermitBenC, how do I check?04:39
BenCThe buildd reports for the latest package in the ppa04:41
apweruditehermit, which kernel specifically is missing the build07:54
* apw yawns loudly08:19
jk-hey apw08:19
apwjk- hey, hows it hangng08:19
jk-sleepy08:20
jk-you?08:20
apwalways sleepy ... UDS is a killer08:21
ohsixhm what sort of pages exactly are these page allocation warnings about; i get them a lot when running perf while swapping08:22
apwohsix, nominally they are reported against a process which is consuming them08:26
ohsixis it fatal to that process or is it just advisory08:27
ohsixperf doesn't die or seem to malfunction; neither does any other thing thats warned about08:27
apwmost cases unless we see the oom killer stepping in you are fine, though they are telling you the kernel is struggling to get what people are asking for08:28
apwit is probabal that perf is falling back to smaller available blocks in that case08:29
apwand reducing efficiency of reporting to userspace or similar08:29
ohsixhm i'll have to check how much overcommit there is next time it happens; i know there was swap space still available in all cases that i saw08:32
ohsixi thought it was waiting for some sort of special pages but it's pretty agnostic with respect to what it is, happens to perf a lot though08:33
ohsixooh found one, http://paste.ubuntu.com/608856/08:36
apwohsix, perf is likely to put demands on real memory which cannot be paged08:36
apwMay 15 17:43:36 krang kernel: [491791.506159] perf_2.6.38-9: page allocation failure. order:4, mode:0xc0d008:36
apwso perf is asking for large allocations, 16 pages at a time and being denied08:37
ohsixokie dokie, i'll look more closely at the report it generates next time it's not perf; thanks for the info08:39
ohsixi have one other but its from the intel gpu dump tool running, might be similar situation to perf08:40
apwyeah probabally needs a big buffer to get the state08:41
ohsixhttp://paste.ubuntu.com/608862/ happened in quick succession as apport was trying to report a render error08:42
ohsixi see some suggestions about tweaking min_free_kbytes, i thought with overcommit stuff could sleep until allocations succeeded? and stuff could be ejected to satisfy it08:44
apwohsix, yes at order 3 and below that occurs, above that there is no likelyhood of success of the eviction08:46
ohsixwhats the "hi: " on per-cpu? all my reports have 186 and so do the ones i can see on google; fixed size area right?08:46
apwanyone wanting to allocate at that size is expected to handle failure gracefully08:46
ohsixahh08:46
ohsixinteresting08:47
apwhi: is the high watermark on the per node page allocation caches08:47
apwso when you free the 186 order 0 page on a cpu a batch will be given back to the global pool08:47
ohsixso where it's expected to be seen it shouldn't be a problem, unless it is and the tool needs to be fixed to accomodate the error?08:48
apwright, and you are seeing no symptoms other than the failures i think, perf keeps working so it must be moving to smaller buffers08:51
ohsixi'll have to see why it raises the error; i don't know how an app would get a chunk of memory or vm without __NOWARN ono it08:53
ohsix__GPF_NOWARN08:53
ohsixok night; thanks for the info08:53
apwohsix, yeah it is probabaly they should have that if they know they are coping08:54
apwsmb, are we expecting any work items on your other-kernel-o-server-requirements blueprint09:12
smbapw, We could place them there though the discussion happened in the other session(s)09:14
smbI have not yet checked what is visible where09:14
apwseems reasonable to pull them onto that one, and you can own the delivery of them09:14
smbSo I probably should go over things you put into your blueprints09:14
apwhttps://blueprints.launchpad.net/ubuntu/+spec/tr-oneiric-kernel-team09:16
apwsmb, the above blueprint seems to have been made with the sole purpose of attaching all kernel blueprints to it, useful to find out blueprints though09:16
smbapw, neat. yeah I think there might be some workitems I would be adding to the server-requirement blueprint if those are not already somewhere else09:18
apwsmb, where items belong on that once do move them, i moved one from config to delta cause it was more appropriate there09:18
smbapw, ok, will do09:19
ogra_hmm, is there something like skipabi=1 for the aliases check ? my custom package always dies in that part 09:20
apwcking, didn't you have some blueprints on fwts ?09:21
smbogra_, aliases? Not sure I know what you mean. There would be skipmodules=1 for the modules check09:22
ogra_with an s ?09:22
* ogra_ has skipmodule=true on his cmdline09:23
smbogra_, maybe without. It is usually so long time between usages I look into the code09:23
ckingapw, minimal one09:23
apwcking, i don't see it at all on our list, whats it called ?09:23
ckinglemme recall..09:24
smbogra_, Ok, it is without the s09:24
apwcking, also do you have a launchpad team for your team ?09:24
ppisatiogra_: watch out, sometimes it's skipmodule, others skipmodules09:24
ogra_damned, i though i had simply typoed 09:24
ckinghttps://blueprints.launchpad.net/ubuntu/+spec/hwe-o-fwts-enhancements09:24
ogra_ppisati, sometimes ? 09:25
ckingapw,  a LP team for my team??09:25
ppisatiogra_: wait09:25
ogra_debian/rules suggests it is without s09:26
ckingapw, https://launchpad.net/~canonical-hwe-team perchance09:26
apwcking, was trying to work out whi i can't find the thing09:26
smbogra_, Yep, that is right09:26
apwor more why it cannot find you09:26
ppisatiwatch out09:26
ppisati[flag@newluxor ubuntu-natty]$ git co ti-omap409:26
ppisatiSwitched to branch 'ti-omap4'09:26
ppisati[flag@newluxor ubuntu-natty]$ grep -Ri skipmodule debian.*09:26
ppisatidebian.master/rules.d/ppc64.mk:skipmodule       = true09:26
ppisatidebian.master/rules.d/powerpc.mk:skipmodule     = true09:26
ppisatidebian.master/rules.d/armhf.mk:skipmodules      = true09:26
ppisatiit bite me once already09:26
smbppisati, that looks a bit like arm sets it wrong...09:27
ogra_that seems to be a bug in the armhf port 09:27
ogra_which we dont use yet09:27
ppisatidunno, but yes, it's not consistent09:27
smbogra_, Which sort of brings us back to what exactly is the failure?09:28
ogra_Checking for dupe aliases in ac100...09:29
ogra_Could not open /home/ogra/kernel/linux-ac100-2.6.37.6/debian/linux-image-2.6.37.6-1-ac100/lib/modules/2.6.37.6-1-ac100/modules.alias at debian/tests/check-aliases line 10.09:29
ogra_run-parts: debian/tests/check-aliases exited with return code 209:29
ogra_what would create modules.alias ?09:29
ogra_seems there is no skip option for it 09:30
apwisn't that a depmod output09:30
smbThat was what I was thinking too09:30
apwthat is very fatal, we don't want any kernels with bad aliases, so i am not supprised there is no skip for it09:30
apwwithout the module.aliases your package is likely no good regardless09:31
ogra_weird, why wouldnt it be called then ... i copied my debian dirs from another package and the compile runs fine up to that point09:31
apwi think we'd need the whole log to know much more09:31
smbAnd probably know what exactly "copied the debian dirs from another package" means. Somehow I have a bad feeling there...09:32
ogra_bah09:32
ogra_ogra@isis:~/kernel/linux-ac100-2.6.37.6$ ls debian/linux-image-2.6.37.6-1-ac100/lib/modules/09:32
ogra_2.6.37-1-ac100  2.6.37.6-1-ac10009:32
ogra_the latter dir is empty ... the former has the missing bits09:33
apwcking, ahh you were only proposed for oneiric09:33
ckingwhat should it be then?09:33
apwogra_, thats a version problem09:33
apwwhat is the version number in your changelog09:34
ogra_yeah i guessed as much now09:34
ogra_2.6.37.6-1.109:34
apwi am not sure you can have the .6 in there09:34
apwdon't think that works09:34
ogra_k09:34
smbogra_, maybe missing clean run09:34
ogra_upstream builds use it09:34
smbto update the control files 09:34
ogra_i'll drop it and re-run 09:35
apwogra_, do you mean the mainline ones ?09:35
ogra_apw, the tree i'm building from 09:35
ogra_its a chromeos based branch with ac100 stuff on top09:35
ogra_with the last mainline merge the .6 appeaderd09:36
apwnot sure how it ever built for them09:36
apw.6 in Makefile is fine, its in debian.*/changelog that its not09:36
ogra_*appeared09:36
ogra_k09:36
* ogra_ starts over09:37
diwica quick question, how do I skip building the pae kernel?09:45
smbdiwic, I don't know whether you can skip individual kernels, but you can only build one flavour09:46
diwicsmb, right, that's the semantic I meant to say09:47
smbdiwic, with "fakeroot debian/rules binary-<flavour>"09:47
smbdiwic, You may want to additionally run binary-headers once to get the generic headers part09:47
smbI mean general09:47
diwichmm. I mistakenly thought that the backport modules shared the same infrastructure, but probably they don't (No rule to make target binary-generic)09:49
diwicAnyway I'll build both then09:49
smbdiwic, No, that tends to be a bit more painful09:49
diwicargh09:50
smbFirst you have to get the headers from the kernel build and extract them somewhere, and then you would run the build with binary-modules-<flavour> and also need to set KDIR (I believe) to the dir where the headers are09:50
diwicI'll build the entire kernel, that's probably simplest09:51
diwicI'm making patches to the include stuff as well09:51
smbIt is a bit simpler in a chroot where you can install packages. Then you can install the headers package from the kernel build09:51
smbapw, So I added some workitems to my blueprint. The target may move and I need to speak to  jjohansen about him being assigned to one10:01
apwheh you are in charge of server for the kernel, so feel free to abuse him :)10:01
smbHe seemed to be more on the task there anyway. So I hope it is not really abuse. :)10:02
ogra_dpkg-deb: building package `linux-image-2.6.37-1-ac100' in `../linux-image-2.6.37-1-ac100_2.6.37-1.1_armel.deb'.10:54
ogra_...10:55
ogra_dpkg-deb: building package `linux-headers-2.6.37-1-ac100' in `../linux-headers-2.6.37-1-ac100_2.6.37-1.1_armel.deb'.10:55
ogra_...10:55
* ogra_ hugs apw10:55
apwogra_, yay10:55
ogra_btw, that build took 70min locally on the tegra 10:56
apwthats pretty quick10:57
ogra_i guess even if i do a complete build in chroot it will beat out builders10:57
ogra_by hours10:57
amitkogra_: -ac100?10:57
ogra_amitk, yep10:57
ogra_rolling an image in oneiric :)10:58
amitkogra_: congrats!10:58
diwicgaah...what's the fakeroot target for changing config options?10:58
apwglad to see you are past a .31 kernel10:58
ogra_well, the kernel is far from being done, but having an upgardeable package will help a lot to get more people working on it 10:58
diwicfdr target 10:59
apwdiwic, i normally just add the option to the leaf i want it in and run fdr updateconfigs10:59
apwi think there is an fdr editconfigs but thats rather boring10:59
ogra_and david has some secret plans to equip more people with ac100's10:59
smbwhich is probably not so secret anymore now. :-P11:00
ogra_hehe :)11:00
ogra_LOL11:02
ogra_http://bellard.org/jslinux/ finally i can emulate x86 on arm 11:02
sorenFabrice is nuts.11:05
ogra_soren, well, that will solve all virtualization issues we have on arm ... we can just set up a webserver now *g*11:06
sorenogra_: You don't even need that, afaict. Just a javascript engine.11:08
apwogra_, this thing eats CPU :)11:11
ogra_apw, we'll just add a "dont-eat-CPU" CSS :)11:11
sorenIt's an emulator written in javascript. What did you expect? :)11:12
apwsoren, i expected it to be a fake ...11:12
apwheh even the compiler works11:12
sorenapw: This is Fabrice Bellard, we're talking about.11:12
sorenHe does not have a sense of humour we're aware of.11:13
sorenIf he says he wrote an emulator in javascript, he means it :)11:13
apw:)11:15
ogra_apw, do you usually call debsign ../*.changes manually or are my debian dirs broken ? my package doesnt ask for a passphrase here12:13
ogra_(during source package build)12:14
apwdpkg-buildpackage -S -rfakeroot -I.bzr -I.bzr-builddeb -I.git -I.gitignore -i'\.git.*'12:14
apwogra_, i use that and it does ask yes12:14
apwthough in my common cause it fails cause i don't have the keys where i build and then i debsign -r anyhow12:15
ogra_dpkg-buildpackage -rfakeroot -S -sa -kogra@ubuntu.com12:15
ogra_thats what i use (no git bits in my source)12:15
apwso yes i would expect it to ask, cirtainly asked me last time i maked one12:16
ogra_weird12:16
ogra_it doesnt with the n900 debian dirs i used as template 12:17
apwisn't asking a something you can turn on/off by default in your personal config12:19
apwogra_, do you have gpg installed ?12:21
apwas it won't ask you to sign unless you have gpg or pgp installed where you make the package12:21
ogra_yes, indeed i do, i use this machine for development regulary12:22
ogra_though let me check12:22
apwthen i am stumped12:22
ogra_BAH !12:22
ogra_ogra@isis:~/kernel/linux-ac100-2.6.37$ LANG=C ls ~/.gnupg12:23
ogra_ls: cannot access /home/ogra/.gnupg: No such file or directory12:23
* ogra_ hangs his head in shame12:23
* ppisati -> lunch12:34
Q-FUNKcould anybody help me with bug #783989   ?14:09
ubot2Launchpad bug 783989 in linux "spurrious EXT4 errors remount /home read-only" [Undecided,New] https://launchpad.net/bugs/78398914:09
Q-FUNKthe end of dmesg seems to indicate what the cause might be, except that I don't know ext4 error codes.14:10
ogasawarakernel dudes, did we want to resume our weekly IRC meetings today?  If so, I'll prep the agenda in bjf's absence.  If not, we'll defer and start back up next week.14:15
* ogasawara back in 2014:16
tgardnerogasawara, I think next week should be fine.14:16
* smb does not object to push to next week14:16
apwogasawara, yeah we have no idea whats what till the blueprints settle down14:26
sconklintgardner: do you want to see whether you have any opinions on this before I revert the patch? https://bugs.launchpad.net/ubuntu/+source/linux/+bug/73517116:09
ubot2Launchpad bug 735171 in linux "driver ath9k is too slow or not responding" [Undecided,Fix committed]16:09
sconklinThe Natty commit is 0ce260b37b523766f7fa0dc33b6981f0d85de9e216:09
sconklinno huge hurry, I'm about to go to the dentist and it may be tomorrow before I actually rip it16:10
tgardnersconklin, looking16:11
tgardnersforshee, can you have a look at bug #735171 ? I'm running out of time today, and sconklin thinks the patch should be reverted. '(pre stable) ath9k_hw: partially revert "fix dma descriptor rx error bit parsing' - but it may already be upstream stable.16:25
ubot2Launchpad bug 735171 in linux "driver ath9k is too slow or not responding" [Undecided,Fix committed] https://launchpad.net/bugs/73517116:25
sforsheetgardner, will do16:26
jcastrojjohansen: here's our bug: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/78212716:37
ubot2Launchpad bug 782127 in linux "RTL8192CE causes a lockup on a Thinkpad X120e" [Undecided,Confirmed]16:37
jcastrojjohansen: I am trying with an oneiric kernel now, it hasn't frozen yet.16:37
jjohansenjcastro: yeah but your home again, the real test is a noisy environment16:44
jjohansenjcastro: that said I installed an oneiric kernel too :)16:44
jcastrojjohansen: I will endevaour to find a noisy spot.16:45
jjohansen:)16:45
jjohansensmb: re sever work item, I had assumed I was doing that anyways as I have an apparmor work item for it from the security side of things :)16:47
smbjjohansen, I was assuming that, so I felt less guilty when volunteering you to the task. :)16:47
ppisatiguys, do we have a list of options that should be turned on to make a vanilla kernel happy with an ubuntu userland?16:54
apwppisati, nothing definative i know of16:56
apwppisati, things we absolutly require are added to the enforce file, but thats only things we have added more recently16:57
ppisatiuhm16:58
ppisatik16:58
sforsheesconklin, re bug #735171, the patch is already in the .38.5 stable release, and a couple of testers reported that the kernel in proposed fixed their problems, so I'd think we should keep the patch16:58
ubot2Launchpad bug 735171 in linux "driver ath9k is too slow or not responding" [Undecided,Fix committed] https://launchpad.net/bugs/73517116:58
sforsheeprobably a case of similar symptoms with different causes16:59
mfilipewill kernel-2.6.39  be released in natty?17:00
apwmfilipe, there are no plans for that no17:02
mfilipehum...17:02
mfilipethanks17:02
apwmfilipe, we only backport kernels to the previous LTS and only then after release17:03
mfilipeI want solve a problem with my intel graphic-card but the ickle releases patch to only 2.6.39 upstream :S17:05
apwmfilipe, we do do unsupported builds of mainline17:06
apwand if you can find the fix, we can try and get that fix into the natty kernel17:06
mfilipehttps://bugs.freedesktop.org/show_bug.cgi?id=3659917:09
ubot2Freedesktop bug 36599 in Driver/intel "[i915 VGA] VGA output blurred" [Normal,Resolved: duplicate]17:09
mfilipeI tried apply the patch in kernel but the intel_bios.c is out-of-date 17:10
apwmfilipe, is there an ubuntu bug yet?17:10
mfilipeyeah... I one second that I will get the link in launchpad17:10
mfilipeapw, https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/67192117:13
ubot2Launchpad bug 671921 in xserver-xorg-video-intel "External monitor using VGA has sync issues with KMS" [Undecided,New]17:13
=== yofel_ is now known as yofel
ogra_apw, so i'm using your patch to lower the console loglevel ... if i set it to 2 as you do, i still get a lot of messages, only setting 1 quietens it, am i missing any additional bits you use with loglevel 2 ?18:46
ogra_or is my kernel just a noisy beast ?18:47
apwlikely the latter i am afraig ogra_ 18:49
ogra_ok, i'll just patch it to 1 then18:49
apwheh18:49
* ogra_ is really proud, the first kernel package i ever created and it builds *and* even works on first go18:50
ogra_if one looks into the build stuff its actually easier than it appears on first sight 18:51
apwogra_, its a hairy beast, but it gives good hugs if you are nice to it18:51
ogra_hehe, well put18:51
=== smb is now known as smb-afk
bryceh_apw, fugly console in oneiric @ 78421720:22
jjohansenbryceh_: now that is some nice corruption20:26
lamontdid 2.6.37 make getrlimit() privileged such that a non-$foo process cannot determine, among other things, it's open file limit?21:21
=== hallyn_ is now known as hallyn
jjohansenlamont: not that I know of, there was some rlimit work to rework how its done in the kernel, change the locking and move around where the security check is done but getrlimit shouldn't be privileged22:02
lamontjjohansen: so bugs.debian.org/626928 (and 627026) should be referred to the kernel there?22:13
* jjohansen looks22:13
jjohansenlamont: commit fc832ad3645f0507f24d11752544525a50a83c71 at least at first glance seems to be the fix22:21
jjohansenso yeah forward to their kernel with that commit #22:21
jjohansenlamont: or it might be commit aa5bd67dcfdf9af34c7fa36ebc87d4e1f7e91873 that fixed it22:26

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