/srv/irclogs.ubuntu.com/2012/06/18/#ubuntu-kernel.txt

=== BenC__ is now known as BenC
* smb yawns07:18
cooloneysmb: morning,07:21
smbcooloney, Morning, how are things going?07:21
cooloneysmb: i'm good, just went through the kernel config about cgroups, looks like there is not much change we need to do07:22
cooloneysmb: could you please share the link for me again about the review cgroups config task07:23
smbcooloney, Ah, good it is not too much. Though I take it some things are still missing... Sure, give me a sec07:23
cooloneysmb: thanks. 07:24
smbcooloney, It is not directly the link to the workitem, but this may be even more useful: http://status.ubuntu.com/ubuntu-quantal/u/cooloney.html07:26
smbThe link for the cgroups stuff is in the work item details07:27
cooloneysmb: wow, thanks a lot, i searched a lot but failed to find out this07:27
smbcooloney, It is a good place to look whether someone gave you a work item without telling you. ;)07:27
cooloneysmb: for the CFQ and deadline discussion, i plan to reply to the reporter about cking's evaluation07:27
cooloneysmb: and ask for some hard facts in detail07:28
smbcooloney, Is that your friends or theother reporter?07:28
smbI think from the other report it is at least clear that it is related to kvm and the emulated disk07:29
cooloneysmb: i sent out an email, did you see that07:29
smbcooloney, I just started my email and begin to drink some coffee, don't expect too much, yet... :-P07:29
cooloneysmb: echo 10 > /proc/sys/kernel/hung_task_timeout_secs07:31
cooloneyAnd start some parallel reads and writes. After a while dmesg will07:31
cooloneyhave some warnings about flusher being blocked.07:31
smbcooloney, But quickly looking, I am not sure I see it. Where did you send this? cc me somewhere or our mailing list(s)07:31
cooloneywe will got oops with CFQ but not with deadline07:31
smbcooloney, Well sure that is maybe not the ideal fairness but should not yet bet fatal. 07:31
cooloneysmb: yeah, not fatal.07:32
smbMeans somehow deadline still does not cause a latency of more that 10s while cfq des07:32
smbdoes07:32
cooloneysmb: i sent to c-k-t07:32
smbcooloney, Ohhh07:32
smbYou mean last week ...07:32
cooloneyyeah,07:32
smbThat one I did see. Thought it was something new07:32
cooloneyit should be last friday07:32
smbSorry, yeah, I saw that test. Surely I need to check on the latency values on the data I got.07:33
smbBut I have not yet done so07:33
cooloneysmb: np, do you think i can put cking's evalution report in LP bug?07:34
smbcooloney, I would maybe wait till we get some time to talk to him. I think for how it looks, test data may concentrate on the wrong details. This was mainly looking at the trhoughput07:35
smbBut the gelt problem seems rather that the latency goes quite high07:36
smbThis sounds a bit too bad for something called fair scheduler. I wonder whether it is something basic or really a bug in the code07:36
smbIf we add data, we should put up both views07:37
cooloneysmb: yeah, my friend in taobao.com said they originally used CFQ by default, but got some time out writing issue in their system07:37
cooloneythen they applied some patches from google to add deadline features in CFQ, 07:38
cooloneyso i'm looking for those patches07:38
cooloneybut not sure it's the same issue07:38
smbRight, initially I thought that it might just be the emulated disk controller but if there is something really wrong with cfq and it is possible to get it so badly behaving that there seems to be no progress, this sounds like a real problem with cfq07:40
smbI think it would be interesting to have several io tasks in parallel and see how much the latency and runtimes differ...07:41
ppisatimoin07:41
smbSince throughput seems not too bad, it looks a bit like that is done in chunks that potentially are long times apart07:42
smbppisati, morning07:42
cooloneyppisati: morning, man07:43
cooloneysmb: for parallel tasks, I just run bonnie++ 3 times in background07:44
smbcooloney, Right, now you only need to gather the data for that. Meaning probably a) how long do those tasks run, b) what is the throughput and c) how much latency each task sees on those tests...07:46
smbAnd at least while looking at the latency one needs to be careful as bonnie does change the scale07:47
cooloneysmb: ok, i will test it again, my computer will be quite slow when testing this. -:<07:50
smbcooloney, Hold on just a bit. I could do it as well on a different machine. Or it might be that cking has the data already but jut not yet put into graphs07:53
cooloneysmb: ok, np, actually, i have a server machine, which was occupied by my wife.07:54
smbcooloney, And you really think _was_? ;)07:54
cooloneysmb: it't not very powerful actually, and any new items for lxc or virt stuff you think i need to take care of?07:56
cooloneysmb: i found there is not much work for lxc,07:56
smbcooloney, Not directly, but making sure at least one of us knows all the parts on the kernel side to make it work (or to adjust if needed). And verify the test cases Serge has (are they covering enough, is something missed)07:59
smbEspecially interesting should be how they try to get the environment as secure as possible. 08:00
cooloneysmb: got it. thx08:09
* apw looks out on a new week08:10
smbapw, nothing to see. It is just preparing to lash out 08:17
cooloneyapw: morning08:20
apw moin08:20
smbcooloney, So cking would be helping to do some testing later on. He got lots of kit to let it run on08:20
cooloneysmb: cool, save my poor computer.08:21
* ppisati notes 3.5rcX redefines the concept of "f*cked up" for omap309:39
cooloneyppisati: what's the "f*cked up" crap in 3.5-rc for omap3?10:06
ppisaticooloney: it doesn't survive 2 boots in a row10:16
ppisaticooloney: secondo time the fs is badly corrupted10:16
tgardnerhenrix, did you see the email on the kteam list about the hp infrared RC6 remote regression ?12:17
henrixtgardner: checking...12:18
henrixtgardner: ah, ok. will reply in a minute12:18
henrixtgardner: i'm actually looking at a similar bug with same kernel12:19
tgardnerhenrix, ack12:19
=== tgardner is now known as rtg
* ogasawara back in 2014:47
=== yofel_ is now known as yofel
ppisatibrb14:58
sisari want to file a bug about touchpad, i should file it under "ACPI" or "Drivers" ? (https://bugzilla.kernel.org/enter_bug.cgi)15:10
rtgogasawara, herton, apw: I need to reboot gomeisa to clean up some sbuild carnage.15:36
hertonrtg, ack15:36
ogasawarartg: gimme 2min to copy some build logs off15:36
rtgogasawara, ack15:36
ogasawarartg: ok, reboot whenever you're ready15:37
rtgherton, ownership on gomeisa seems to have gotten totally scrogged when I installed sbuild. its gonna take awhile to restore ownership settings I think15:48
hertonrtg, yep heard that, I probably will go to tangerine for now too15:48
apwrtg, ack15:53
* ppisati -> gym, back in a bit16:04
* herton -> lunch16:06
bambeehi17:23
ogasawarabambee: hi, thanks for joining the channel.  you mentioned you were interested in getting involved and I suggested bug triage is a nice place to start as it gets you familiar with Launchpad, how our bugs are handled, and our typical work flows in general.17:26
ogasawarabambee: if this is something you'd like to start with I'd actually point you to jsalisbury and he can help you get started.17:27
jsalisburybambee, o/17:27
bambeejsalisbury: hi 17:28
jsalisburybambee, if you are interested in triage there is a few good wiki pages that is good to review:17:28
jsalisburybambee, https://wiki.ubuntu.com/Bugs/HowToTriage17:28
bambeeI would be interested in kernel/testing and debugging, I think or debugging+triaging17:29
jsalisburybambee, there is allot of info there.17:29
jsalisburybambee, we can always use help testing :-D17:29
jsalisburybambee, and debugging/triaging17:30
bambeeoh17:30
jsalisburybambee, the best way to get started is to take a look at some of the wiki docs.  Then go through the process of triaging a new bug, which I can help you with.  And if you find a bug with similar hardware that you have, testing for those types of bugs really help.17:32
* bambee is reading the wiki docs17:32
bambeeok it's noted, thanks for your explanations ;)17:34
jsalisburybambee, sure np.  just let me know if you run into anything you have questions on.17:35
bambeefor a beginning, you're right, I think that's probably better to start with triaging and testing (to be familiar with launchpad and your development process). Later I think I would like to help with development, and bugs fixing (this is a task reserved for guru, probably).17:39
=== rtg is now known as rtg-lunch
=== rtg-lunch is now known as rtg
bdmurrayapw: any news on bug 882147?19:53
ubot2Launchpad bug 882147 in linux "overlayfs does not implement inotify interfaces correctly" [High,Triaged] https://launchpad.net/bugs/88214719:53
* rtg -> eod21:02
henrixjsalisbury: ping21:30
jsalisburyhenrix, pong21:30
henrixjsalisbury: regarding bug #101480021:30
ubot2Launchpad bug 1014800 in linux "infrared remote doesnt work anymore (3.2.0-26)" [Medium,Confirmed] https://launchpad.net/bugs/101480021:30
henrixjsalisbury: mind if i step in? i believe i know the issue, so i've built a test kernel21:30
jsalisburyhenrix, absolutely :-D  Thanks for the help!21:31
henrixjsalisbury: thanks. its related with a patch i submitted21:31
jsalisburyhenrix, ok, thanks21:31
henrixjsalisbury: so, i'll just "readjust" the patch and it should fix it21:31
jsalisburyhenrix, good stuff21:32
henrixjsalisbury: ;)21:32

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