/srv/irclogs.ubuntu.com/2012/11/15/#ubuntu-kernel.txt

=== BenC__ is now known as BenC
=== dlbike76 is now known as dlbike76_
dlbike76_Hi I'm running kernel 3.5.0-18-generic on quantal and am seeing instances where the kernel is killing the chromium tasks05:20
dlbike76_Is this a known problem?05:20
=== dlbike76_ is now known as dlbike76
dlbike76I'm running lubuntu on a computer with limited RAM, and I didn't notice the problem while on precise.05:22
dlbike76I'm seeing the kernel sending sigkill to the various chromium tasks in the kern.log once memory usage gets to a certain point.05:23
RAOFdlbike76: That would presumably be the OOM killer?05:25
dlbike76RAOF:  OOM?05:26
RAOFOut of memory killer.05:26
dlbike76Gotcha.05:26
RAOFie: We've overcommitted, and now something wants to cash the check for memory that we don't have, so we need to kill something.05:26
dlbike76Should that happen even if swap hasn't been fully utilized?05:26
ohsixyea ...05:27
ohsixswap can't always service memory requests05:27
ohsixyou can adjust the overcommit ratio05:27
ohsixthough at best it will probably get you a machine that runs slower for a long time before something is finally killed :]05:28
ohsixchrome seems faster in part due to the way it uses memory to do its work, you might get more out of the situation with a browser not based on webkit05:29
dlbike76It's working pretty good otherwise.  I'll just have to multitask less...05:29
ohsixyou can tell firefox how many images to keep decoded before discarding them, and how much of the back/forward cache to keep around; lots of knobs that won't get you around not having enough memory05:30
dlbike76I've been experimenting with various browsers to see which ones work best for the limited RAM that I currently have.05:31
dlbike76My biggest problem is that I went from a system with over 3GB RAM to a system with approx 300MB.05:32
ohsixfun, i had to use my netbook when my laptop broke once :] could barely start the firefox session i regularly used, and do anything else05:33
dlbike76So is OMM Killer something new in the 3.5 kernel? 05:36
dlbike76I don't remember this ever happening in 3.2, but I do remember the system slowing to a crawl at times.05:37
ohsixit's not new, it's probably just a coincidence that you're noticing it kill chrome05:38
dlbike76Okay at least I know that it is working correctly.  Thanks for the help ohsix and RAOF.05:44
=== jk-- is now known as jk-
=== smb` is now known as smb
smbinfinity, Hiya, somehow I think there was not time to look at the xen package?09:06
* apw yawns09:10
* smb remembers to make more tea09:14
infinitysmb: Oh, uhm.  Indeed.  I got sidetracked with a bunch of other stuff.09:37
infinitysmb: Say, you have a new core-dev on your team, you should make him review and sponsor. ;)09:38
smbinfinity, I asked him about another package and he hasn't done it for a whole week... :-P09:38
smbapw, ^ ;)09:38
infinitysmb: A whole week?  Man, what a slacker.09:38
apwyou all smell09:39
infinity(... says the man who's had a review pending for Andy for months)09:39
Laneyhello kernelers ... :-)09:59
LaneyI noticed that with the new raring kernel I don't have aufs or overlayfs or whateveritis any more ... is that deliberate?10:00
Laneyalso modprobe says this "FATAL: Could not load /lib/modules/3.7.0-0-generic/modules.dep: No such file or directory"10:00
apwLaney, i suspect its expected, undesired, i'll check a bit later10:05
Laneyapw: OK, I'll go back to 3.5 then for now, cheers10:06
petantikHello, I'm trying to debug the ubuntu kernel with kgdb over serial. How can I get a vmlinux file from the vmlinuz file so I can load it into gdb? or does the vmlinuz work fine?10:09
einonmpetantik: Are you not building the kernel from source yourself?10:17
petantikeinonm: I am but not using upstream kernel sources 10:18
petantikeinonm: einonm using the instructions here https://help.ubuntu.com/community/Kernel/Compile#Alternate_Build_Method:_The_Old-Fashioned_Debian_Way10:19
einonmSo do you have a vmlinux image in your build directory. I'm not overly familiar with that way of building the kernel, but there should be a vmlinux produced after building. 10:21
petantikwhat I'm trying to figure out is how to generate a vmlinux file instead of a vmlinuz file10:21
petantikeinonm: let me check10:21
einonmThe vmlinux file is generated before the vmlinuz file, and is used to generate the vmlinuz file10:21
petantikeinonm: you're right, definitely a pebkac moment.10:22
petantikcheers10:22
einonmnp :)10:22
* henrix just received a build failure email for a build that is still pending...!?!?12:35
henrixapw: any idea what happened here? ^^12:35
henrixapw: it was an armel build for 3.2.0-34.5312:36
apwhenrix, the build may have failed due to buildd failure (say they rebooted a buildd) and then someone put the build back for rebuilding12:36
apwso what you see if it pending for rebuild12:36
apwhenrix, got a pointer to the build12:36
henrixapw: yeah, but the link sends me to the "Lost something?" thing :)12:37
apwhenrix, well that probabally means it was indeed retried, as we do lose the errors then12:37
henrixapw: ok, so i'l just wait and see12:38
apwwhich build was it, which series12:38
henrixi've just build tested armel again and it builds without an error12:38
henrixit was precise12:38
henrix3.2.0-34.5312:38
apwhenrix, yeah it seems to be building right now, so it must have been put back12:39
henrixapw: ok, thanks. let's wait and see what happens12:39
rtgapw, henrix: I restarted it 'cause it was a compiler seg fault13:03
apwheh there you go then13:03
apwrtg, i am working on fixing aufs and overlayfs in raring, so don't upload without me13:04
rtgcking, could you give me a quick review of your batter patch on git://kernel.ubuntu.com/ubuntu/ubuntu-nexus7.git master-next ? I've rebased against a new release and there were some conflicts in your battery status patches.13:05
rtgapw, ack. did you re-upload signed or do you plan to wait ?13:06
apwrtg, no i didn't ... i will do that right now13:07
apwoh perhaps i won't if that is holding us from breaking the images ... hmmm13:07
henrixrtg: ok, thanks13:07
* henrix -> lunch13:08
rtgapw, can't build dailies without overlay or aufs ?13:08
apwrtg, they won't boot without as far as i know13:10
apwrtg, i have some test builds ongoing, i am doing overlayfs first, as thats the default13:10
apwif that works we can upload that13:10
rtgapw, ok13:10
* ppisati switches to the haswell box (back in a bit)14:02
* henrix realises that herton is referred in lwn! \o/14:30
apwhenrix, link ?15:07
henrixapw: it's on this week edition, kernel section: https://lwn.net/Articles/524304/15:08
henrixapw: it's just the announment of the stable tree for 3.5 :)15:09
apwheh ... good stuff15:10
apwrtg, ogasawara, anything you want thrown in the next upload for raring ... i don't want to have any rebases in here really, as this enabled a heap of code15:17
ogasawaraapw: nothing here15:18
ppisatiogasawara: did you put somewhere the list of patches that you showed during the delta review?15:21
apwppisati, normally that is in the spec for the blueprint15:21
ogasawarappisati: yah, in the spec.  lemme find the url for you.15:22
ogasawarappisati: https://wiki.ubuntu.com/KernelTeam/Specs/RaringKernelDeltaReview15:22
ppisaticool, thanks15:23
apwogasawara, i have generated the 'per dev patch review' WIs and added them15:44
ogasawaraapw: sweet, thanks15:44
ogasawaraapw: I need to go through and update our milestone tracking too in our BP's since the new monthly checkpoints were added to LP15:44
apwogasawara, monthly what now ?15:45
* rtg actually made a nexus7 kernel that boots15:46
ogasawaraapw: there's the new monthly milestones now since we don't have alpha's15:46
ogasawaraapw: eg ubuntu-13.04-month-1 15:46
* apw hans rtg a pint ... well done15:46
ogasawaraapw: hrm, I'm also now seeing the alpha's there too.  I swore they were not there yesterday.15:47
apwogasawara, yeah, so we should be trying to fan out our WIs to the appropriate months15:47
apwogasawara, now all i need is some feeling for what month-1 etc means, any idea when they are15:49
ogasawaraapw: nope.  I checked yesterday to see if the release schedule had any specific date for what the month-1 deadline was, but found nothing.15:49
apwogasawara, the milestones themselves will have them, if i could remember where they are15:50
apwhttps://bugs.launchpad.net/ubuntu/raring/+milestones15:50
apwogasawara, oh the alphas etc are 'optin' now15:51
apwogasawara, anyhow so its the 18th, -month-1 is like saturday, and so on15:51
ogasawaraapw: ah ok.  /me makes a note15:51
ogasawaraI never knew of the /+milestones url, that's helpful15:53
apwogasawara, me either, i just guessed, right for once15:54
ckingis that a new LP feature?15:54
apwcking, don't think so no15:55
bjfapw, due to the elegance and consistency of LP you can just naturally anticipate those features15:55
apwit is the very first time that my guess found something, i am somewhat chuffed15:55
ogasawaraapw: you just happened to stumble onto a LP easter egg15:56
ckingI though the way to find out LP features was to grok the source15:56
cking*thought15:56
* ogasawara back in 2015:57
ckinghttps://launchpad.net/launchpad/+specs  is interesting, all the eggs aren't in one basket15:59
apwrtg, ogasawara, ok i have a kernel here with working overlayfs and aufs, so i am proposing to upload it now16:06
rtgapw, lock and load....16:07
apwBenC, i have just pushed a new raring master-next with working overlayfs which you need for your images to boot16:11
BenCapw: Ah, thanks16:11
apwrtg, when you build source pacakges do you do so on 32 or 64 bit chroots ?16:11
rtgapw, generally both16:12
BenCapw: Is that -2.8?16:12
apwBenC, yes16:12
apwBenC, i have literally just pushed the master repo, packages are being prepped now16:13
apwthough i am running it here on my kit16:13
apwBenC, oh you may want to fetch again so that the master branch is updated, else the rebase script may not work, as i found out when i did lowlatency :)16:27
rtgapw, what script ? ppc is a straightforward rebase against master isn't it ?16:28
apwrtg, well it cannot be a strightforward rebase just because master is rebase as well, for lowlatency i use an update-from-master script which knows how to find the bottom of what to rebase etc16:29
BenCMine seemed to find the correct places to rebase against using the tag16:31
BenCMaybe I'm lucky16:31
BenCapw: Starting a build, so I'll follow up in a couple hours16:34
rtgogra_, are the nexus7 udeb files of any use ? the installer just blats an image onto flash, right ?16:35
apwrtg, the d-i and netboot images use them don't they?  regardless of final installation method16:36
ckingI installed my n7 kernel  image from the .debs  - but hey, I don't know the official way its done16:37
rtgapw, neither of which are useable by the nexus 716:37
apwrtg, then it seems unlikely they can be being used indeed16:37
rtglemme check with janimo16:37
rtg^^16:37
* ppisati goes out for a bit...16:43
stgraberhello kernel people. The new linux-libc-dev broke libnl and in turn a bunch of universe packages.16:55
stgraberI'm now working on a patch to libnl's netlink-kernel.h to deal with the change and would appreciate one of you confirming it's the correct approach16:56
stgraberthe debdiff for libnl would be: http://paste.ubuntu.com/1360658/16:56
rtgapw is likely the most informed about header issues16:56
apwstgraber, hi16:57
stgraberI'm currently doing a test rebuild of sssd (one of the affected package) with the updated version of libnl to see if that does the trick16:57
stgraberapw: hi16:57
apwstgraber, oh god that is utterly vile16:57
apwstgraber, it is clearly doing something which it should not, relying on the names of _ prefixed defines16:58
apwstgraber, what you have done seems as appropriate for the new headers as for the old, but *barf*16:58
stgraberhehe, well, upstream fixed that in libnl3 (different api) and redhat is in the process of porting sssd to using libnl316:59
stgraberso hopefully we can kill libnl1 soon and won't have that kind of problem again...16:59
apwstgraber, but yeah i think you have no choice in your situation and you appear to have done something which means the right thing16:59
stgrabergood and the test build just succeeded so I'll push that to the archive now. Thanks for taking a look at the fix.17:00
apwrtg, ogasawara, not sure who did the fold down last time we rebased, but when we did squashing aufs into its base patches (aufs-base and aufs-standalone) made it real complex to unpick and update17:15
apwi've attempted to mark the ones which are annoying to lose separation on (no-squash) for next time17:16
rtgapw, prolly me17:16
janimortg, feel free to drop udebs17:44
janimoI suppose ubiquity does not use them either in case we ever want a proper ubuntu installer17:45
rtgjanimo, already did, just uploaded to raring directly17:45
janimortg, great17:45
rtgjanimo, also created nexus7 branch in ubuntu-raring-meta17:45
janimortg, should I push the firmware package from PPA to raring NEW too or Are you looking at that too?17:45
janimortg, nice, had no git tree for meta so far17:46
rtgjanimo, where is the firmware package ?17:47
janimoin the ubuntu-nexus7 staging ppa raring pocket17:47
rtgjanimo, lemme have a look...17:48
* rtg -> lunch18:15
=== yofel_ is now known as yofel
* henrix -> EOW19:33
slangasekogasawara: bug #1079385 looks like a duplicate of the overlayfs support issue; is there a master bug for this?20:52
ubot2Launchpad bug 1079385 in upstart "init: can't start after load iso image on hyper-v server 2008 r2 sp1" [Undecided,New] https://launchpad.net/bugs/107938520:52
slangaseklooks like 107919320:57
ogasawaraslangasek: yep, that's the one we shoved in the changelog20:58
slangasekok, thanks20:58
BenCapw: Can you help me out using insert-ubuntu-changes.pl?21:17
BenCnm, figured it out21:19
=== BenC__ is now known as BenC
infinityBenC: Oh hey.  I wanted to chat with you.21:27
infinityBenC: Specifically about the udebs you dropped from PPC...21:27
BenCinfinity: Yes?21:28
infinityBenC: The pcmcia stuff is absolutely usable on Apple hardware.  Perhaps only disabling those for the e500 flavours would have made more sense than across the board?21:28
infinityBenC: (And the 8250 stuff also worked fine in the past, and should still do...)21:28
BenCinfinity: unfortunately, there's no way to do that on a per flavor build21:28
BenCI have it on my todo list to re-enable that once I create a mechanism that does what we need it to21:29
infinityHrm.21:29
BenCThe udeb's are empty on e500* so It's not like I could just ignore a few modules21:29
infinityYet another argument for "powerpc should have stayed in master, and only e500 should be out-of-tree"? :P21:29
infinityHaving e500 randomly break previously-working non-e500 setups seems suboptimal. :/21:30
BenCI've not heard of any other arguments for such a case :)21:30
BenCI did say I plan to fix it :)21:30
BenCIMO, it's kernel-wedge that is suboptimal21:30
infinityYeah, I heard you.  I'm just thinking in the more general case.21:30
BenCAlso, having 8250 (for I assume serial console) be modular is less than desirable anyway…I can't imagine anyone using that in a sane way21:31
BenCMy plan was to build it in on non-e500* to match up and not lose that udeb functionality, but I haven't decided yet21:32
infinityIt wouldn't generally be for consoles, most PPC hardware I know has onboard serial that isn't 8250.21:32
infinity(Which, sure, also makes it less likely to be a useful udeb...)21:32
BenCI couldn't really think of a legit use case21:32
infinityKeeping the module but dropping the udeb is probably fine for the serial one.21:33
BenCBut I'll take your word for it that it's important :)21:33
infinityFor pcmcia, not so much.  Those are kinda useful.21:33
BenCI've never heard of people using PCMCIA from powerbook's for things needed during install21:33
BenCExcept maybe networking, but is that something not available on most PowerBooks anyway?21:34
infinityI could dream up a scenario where one might want an 8250 udeb to do some serial debugging or something weird, but it's pretty corner case, and not something I suspect anyone really cares about.21:34
infinityPCMCIA for networking would absolutely be the use-case.21:34
BenCOh, wait, is airport on things like G5's attached to a PCMCIA port internally?21:34
BenCI think it is…so there's a good reason for having it21:35
BenCG5/G421:35
infinityThat's also possible, I dunno, but I was just thinking of the ton of powerbooks with crap internal NICs and PCMCIA instead.21:35
infinity(Which was not an uncommon setup)21:35
infinityEither way, we don't make a habit of telling people "if you don't use the internal NIC, you no get networking during install".21:35
BenCOk, I'll definitely work on getting PCMCIA udeb's back, and I'll mull over 825021:35
infinityHence why we even have usb-network-udeb, etc.21:35
infinity(Or whatever it's called)21:36
infinityBenC: Of course, a rebase to -2 master is probably slightly more urgent, due to the overlayfs/aufs oopses in -1. :P21:36
infinityBenC: No pressure, welcome back to the kernel team, etc.21:37
* infinity mails Ben a beer.21:37
BenCinfinity: sure thing, and you missed the opportunity to buy me a real beer21:38
BenCOddly, not many were willing to drink with me in Copenhagen21:38
infinityYou were a hard man to find after hours at UDS.  I could have just beered you at the smoke doors between sessions. :P21:38
infinity(Or maybe I was the hard one to find?  I'm not sure...)21:39
BenCI believe I was being avoided due to the bar having jaeger on hand21:40
apwBenC, this new upload didn't get an abi bump, i cannot see how it could not have been; do you still have skipabi/skipmodules turned on ?22:40
apw(not that it matters anywhere near as much on ppc as you likely have no dkms, but ...)22:41
apwinfinity, this whole kernels are (new) and linux-signed can't build till they are done crap really makes getting kernels built a long arduious process23:01
apwit would be good to get the signed in, so we get new kernels in the live's23:01
infinityapw: Yeah, yeah.  Hold your horses. ;)23:03
apwheh thanks :)23:15

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