| === \sh_away is now known as \sh | ||
| abogani | cking: Are you around? | 07:36 |
|---|---|---|
| cking | abogani: Hi there.. | 07:36 |
| abogani | cking: About Bug #177895 | 07:38 |
| ubotu | Launchpad bug 177895 in linux "Kernel 2.6.24-2 causing ~1000 wakeups by "Rescheduling Interrupts"" [Medium,In progress] https://launchpad.net/bugs/177895 | 07:38 |
| abogani | cking: Interesting commit is 62fb185130e4d420f71a30ff59d8b16b74ef5d2b in mainline. | 07:38 |
| cking | Yes, I am very familiar with this one :-) | 07:38 |
| === doko_ is now known as doko | ||
| cking | Yep - apparently it removes two earlier commits | 07:39 |
| abogani | So I let thi Bug to you :-) | 07:39 |
| cking | ..well I had a look at it and I was not sure if these commits are actually in our kernel | 07:39 |
| abogani | It isn't a trivial merge. | 07:39 |
| abogani | cking: Do you have hw that incurr evidently in this bug? | 07:40 |
| abogani | cking: What is your TZ? | 07:41 |
| cking | abogani: I have a Centrino duo which I see 200-300+ Rescheduling Interrupts. But not 5000+ as some are seeing | 07:41 |
| cking | cking: My TZ is UTC. I start early :-) | 07:41 |
| abogani | cking:so Good Morning! :-) | 07:42 |
| abogani | cking: Mainline give the same result on my Centrino duo! | 07:42 |
| cking | abogani: Back to commit 62fb185130e4d420f71a30ff59d8b16b74ef5d2b... I believe it reverts 58e2d4ca581167c2a079f4ee02be2f0bc52e8729 and 6b2d7700266b9402e12824e11e0099ae6a4a6a79 | 07:43 |
| cking | abogani: however, I was unsure if these commits were actually made to our current kernel tree | 07:44 |
| cking | abogani: have you investigated this to any depth yourself? | 07:44 |
| abogani | cking: I'm already cherry-picked and my laptop works perfectly. | 07:44 |
| cking | abogani: do you have a patch that I can pick from so that I can give it a run through? | 07:45 |
| abogani | cking: But i don't have hw that expose evidently thus bug. I'm not sure that this fix... :-( | 07:45 |
| abogani | cking: Do you prefer git or email? | 07:45 |
| cking | abogani: yes this is the same kind of problem I am facing too. And the patch is a bit "radical" as it does touch a lot of the important parts of the scheduler | 07:46 |
| abogani | Agreed | 07:46 |
| cking | abogani: git - if that's OK, but email if that's easier for you. | 07:46 |
| cking | abogani: I have some tests with powertop and so forth that I can apply to see what's going on. | 07:47 |
| abogani | cking: Ok i'll push it in my git tree. Please Let me some minutes... | 07:47 |
| cking | abogani: The main thing is to see if the rescheduling interrupts are legitimate - a lot of them really are due to correct load balancing | 07:48 |
| cking | abogani: so it will take me 3-4 hours to shove in some diagnostics and see if the fix really is OK | 07:48 |
| cking | abogani: OK - much appreciated - this saves me a lot of work. :-) | 07:49 |
| === asac_ is now known as asac | ||
| abogani | cking: Sorry I don't know why but my git don't work anymore! :-( | 09:16 |
| cking | abogani: Ah. Problem :-( | 09:17 |
| abogani | cking: Mhhhh .gitignore conflict | 09:18 |
| abogani | No way to understand why git stop work... | 09:27 |
| cking | abogani: perhaps I should try the patch on my git tree and toy with the fix that way | 09:34 |
| abogani | cking: Are you registered user on Freenode? | 09:34 |
| cking | abogani: I thought so - why? | 09:35 |
| abogani | cking: I'm trying to send you a file | 09:35 |
| cking | abogani: perhaps try Emailing it to colin.king "at" ubuntu.com | 09:36 |
| cking | abogani: meanwhile I check my Freenode and IRC settings | 09:37 |
| abogani | cking: Mailed. I send you my adaptation of the cherry-pick fix | 09:38 |
| cking | abogani: many thanks! | 09:38 |
| abogani | Sorry but my git is completely break ... :-( | 09:39 |
| cking | abogani: ..I know the feeling, when it breaks, it *really* breaks - and usually at the worst time. | 09:39 |
| abogani | :-) | 09:40 |
| abogani | Murphy docet ;-) | 09:41 |
| amitk | abogani: what seems to be the problem with git? | 09:46 |
| === setmora is now known as eradicus | ||
| abogani | amitk: Hi Amit! I cherry-pick a commit from a remote tracked repo, use git-mergetool and git-status. And all is ok. I execute git-commit and all things disappear. git-log and git-status don't show nothing! Work seems completely lost! | 09:49 |
| abogani | Seems to me a index problem... | 09:49 |
| abogani | s/index/git index/ | 09:49 |
| amitk | abogani: interesting. And you weren't working on a branch? | 09:52 |
| abogani | amitk: No. Just created with 'git-checkout -b' | 09:54 |
| amitk | abogani: that is a branch | 09:55 |
| cking | abogani: Hi again.. | 10:09 |
| abogani | cking: Hi | 10:09 |
| cking | can you gzip the patch and re-send it to me as a gzip attachment. My mail client is silently putting in white spaces that cause git-apply to fail :-( | 10:10 |
| === rikai__ is now known as rikai | ||
| cking | ..by the way I can reproduce the 10,000+ Rescheduling Interrupts quite predictably now - so it will be straight forward to test the patch. | 10:11 |
| abogani | cking: The fault is my webmail! | 10:14 |
| abogani | cking: Mailed. | 10:15 |
| cking | abogani: no worries - thanks again for the speedy response! | 10:15 |
| abogani | cking: How can you reproduce the bug? Is it necessary something special actions? | 10:16 |
| cking | abogani: It's a case of getting the right suite of apps to generate loads of timer actions and get enough free CPU cycles so that both cores are incorrectly balanced... | 10:17 |
| cking | ...then one can see the scheduler working hard to try to rebalance the load and cause zillions of IPI events. | 10:17 |
| Ng | cking: is that going to be easily fixable for hardy? :) | 10:17 |
| abogani | Not easily... | 10:18 |
| cking | ..I hope so.. it will take a few hours of analysing the IPI events under different loads before I can say. I think it's not easily fixable.. | 10:19 |
| cking | ...I've been looking at an answer on and off for a few weeks on this one - and tinkering with the scheduler is risky for Beta | 10:19 |
| Ng | yeah | 10:20 |
| Ng | just curious because my laptop seems to do several hundred wakeups a second for that | 10:20 |
| cking | After a lot of consideration I believe a lot of the rescheduling interrupts are legitimate. | 10:20 |
| cking | A lot of apps do very frequent timer events .. they wake up and the scheduler tries to spread the load across cores | 10:21 |
| cking | For laptops with Centrino Duo cores it is better to spread the load across cores so that they are busy rather than have one core running at full speed | 10:23 |
| amitk | cking: what happens if IRQ balancing is disabled? the interrupts only go to the core on which the app is running? | 10:23 |
| cking | amitk: Not sure if this is just a IRQ balancing. My understanding is that the scheduler is rebalancing the load across cores in probably too agressive a manner | 10:25 |
| cking | ..OR... | 10:25 |
| cking | ..that we can now see the IPI events because the Rescheduling Interrupts are more visible | 10:25 |
| cking | amitk: Any wise input on this? | 10:27 |
| amitk | cking: not really. Though I wonder if the rebalancing behaviour will change if the scheduler is changed. | 10:28 |
| cking | amitk: looking at all the scheduler tweaking in 2.6.25+ I am very concerned that this type of issue has a lot of life in it.. | 10:29 |
| cking | amitk: my big worry is that the problem may be resolvable for Hardy, but need a lot more work for later kernels. | 10:31 |
| cking | ..and also the scheduler is not a trivial piece of code to tinker with - if one goofs up, one goofs up big time. | 10:32 |
| amitk | cking: how so? | 10:32 |
| cking | amitk: sorry.. I misunderstand: how so what? | 10:32 |
| amitk | cking: why would it be harder to fix for later kernels? It should be fixed upstream, right?> | 10:34 |
| cking | amitk: true | 10:36 |
| cking | I'm just concerned with a fix now in Hardy LTS which makes the scheduler diverge from upstream in a (perhaps significant and) unmaintainable way | 10:38 |
| amitk | cking: i see you point now. But with LTS, it is unlikely that we will upgrade to a newer kernel. | 10:40 |
| amitk | *your | 10:40 |
| cking | amitk: indeed.. but I'm always keen to reduce risk.. especially wrt key kernel components. | 10:42 |
| amitk | cking: agreed | 10:43 |
| cking | Hi BenC | 11:01 |
| BenC | hey | 11:02 |
| cking | another day, another bug :-) | 11:02 |
| BenC | hehe | 11:04 |
| BenC | cking, amitk: If either of you wants to follow up on my work on cpu1 losing cpufreq capa, I found some interesting bits after some debug last night | 11:09 |
| BenC | after suspend, when bringing up cpu1, there are some acpi errors about unknown op codes | 11:10 |
| cking | BenC: Sounds like something meaty that I could get my teeth into.. | 11:10 |
| BenC | bringing cpu1 down/up before suspend/resume works fine (the cpufreq symlink returns in sysfs) | 11:10 |
| BenC | doing so after suspend resume repeats the acpi errors though | 11:11 |
| * amitk gives up the bug to cking reluctantly - not :) | 11:11 | |
| BenC | mjg59: Does the acpi subsystem reread the acpi tables after suspend/resume, or rely on it's internal copy? | 11:12 |
| cking | BenC: Sounds like https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.22/+bug/183033 | 11:12 |
| ubotu | Launchpad bug 183033 in linux-source-2.6.22 "Intel Core 2 Duo - Resume from suspend, CPU Frequency Scaling is gone on CPU1" [Undecided,Confirmed] | 11:12 |
| BenC | cking: That's the one | 11:12 |
| mjg59 | BenC: They're never re-read | 11:12 |
| mjg59 | Hm. In principle, SSDTs can be loaded and unloaded at runtime. | 11:12 |
| cking | mjg59: Not sure what happens in reality though. I will look into it. | 11:13 |
| mjg59 | BenC: Curious. acpidump and iasl -d should let you figure out where that object is meant to live | 11:13 |
| BenC | mjg59: Then my first guess is that the internal copy is getting corrupt somehow | 11:15 |
| BenC | maybe acpidump before and after suspend will confirm that | 11:16 |
| mjg59 | BenC: acpidump will dump direct from the hardware, not the kernel's representation | 11:16 |
| mjg59 | /proc/dsdt /might/ give you the kernel version of the dsdt | 11:16 |
| mjg59 | /proc/acpi/dsdt, that is | 11:17 |
| mjg59 | BenC: What hardware are you seeing this on? | 11:17 |
| mjg59 | A d630 has literally just turned up at my front door, so I can poke it there | 11:17 |
| cking | mjg59: This bug also occurs on my Lenovo 3000N200 Centrino Duo | 11:17 |
| BenC | mjg59: Lots of systems | 11:17 |
| cking | BenC: so at least I can dig into the bug on my hardware straight away | 11:18 |
| mjg59 | Ok. I'm not seeing it on my HP. | 11:19 |
| BenC | I reproduced it on two of my dell laptops | 11:19 |
| Ng | fwiw, not seeing it on my thinkpad | 11:19 |
| BenC | hoping this isn't a BIOS bug | 11:20 |
| BenC | or if it is, hopefully we can work around it somehow | 11:20 |
| BenC | mjg59: Should acpidump and /proc/acpi/dsdt match? | 11:22 |
| mjg59 | BenC: The dsdt hunk of acpidump should I believe, yes | 11:22 |
| mjg59 | Certainly after passing them through iasl -d | 11:22 |
| BenC | mjg59: Are the CPU related bits we are concerned about in the dsdt? | 11:23 |
| mjg59 | BenC: Not sure. If you disassemble the dsdt and find the _PCT methods, then yes :) | 11:23 |
| BenC | mjg59: Thanks :) | 11:24 |
| BenC | cking: Ok, sounds like you've got plenty to go on...I expect a patch in upstream kernel and a full report when I return in 3 hours :) | 11:24 |
| mjg59 | Otherwise, probably in an SSDT | 11:24 |
| cking | BenC: Of course ;-) | 11:24 |
| BenC | See you guys in a bit | 11:25 |
| adinc | if i download the packaged ubuntu kernel sources, where does it install them? | 11:45 |
| abogani | adinc: apt-get source linux-image-2.6.24-12-generic will install source in current dir | 12:09 |
| adinc | abogani: no it did install to AAAAAAAAAAAAAAAAAAAAAAAA/AAAAAABBBBBBBBBBB | 12:19 |
| adinc | pardon to /usr/src | 12:19 |
| adinc | can i disable CGROUPS on a running kernel, or do i have to disable it in the kernel configuration before compiling? | 13:53 |
| rtg | adinc: its a compilation config macro, so I don't think you can change the way the scheduler works at runtime. | 13:55 |
| adinc | is this essential in ubuntu kernel? or can i safely disable it | 13:55 |
| rtg | adinc: you can set it however you like. the current value is targeted to desktop single user environments. | 13:58 |
| adinc | where can i get the complete config file for the hardy kernel packages kernel? | 13:58 |
| adinc | the config file which is in /boot is unfortunately not complete | 13:59 |
| rtg | adinc: install a headers package, then look for a .config in /usr/src/linux-headers* | 13:59 |
| adinc | which one is complete linux-headers-2.6.24-12 or linux-headers-2.6.24-12-generic | 14:00 |
| adinc | generic | 14:00 |
| adinc | ... | 14:00 |
| rtg | adinc: alternatively you could 'debian/rules prepare-generic', then look in debian/build/build-generic for a .config for that specific flavour. | 14:01 |
| adinc | rtg: no this can't be the complete .config file since the .config in linux-headers is IWL3945 module missing, i mean it is not set | 14:01 |
| mjg59 | adinc: iwl3945 comes from linux-ubuntu-modules | 14:02 |
| mjg59 | Not linux-image | 14:02 |
| rtg | adinc: thats because iwlwifi is built in LUM, not the kernel. | 14:02 |
| adinc | so how do i get a complete config file | 14:02 |
| mjg59 | adinc: That is a complete config file | 14:02 |
| adinc | i don't understand | 14:03 |
| rtg | adinc: its the complete kernel config. | 14:03 |
| mjg59 | adinc: The Ubuntu kernel does not include iwl3945. It's in a separate package. | 14:03 |
| adinc | but the kernel needs to be prepared in order to accept this module, doesnt it | 14:03 |
| mjg59 | adinc: No | 14:03 |
| rtg | adinc: have you read https://wiki.ubuntu.com/KernelMaintenance? | 14:04 |
| adinc | i've read several wiki pages, but let me see if i know this aswell | 14:04 |
| adinc | i'm thankfull for any information | 14:04 |
| adinc | this would mean that i could compile this particular module only with the kernel headers, but without the source. only this particular modul, is this right? | 14:07 |
| adinc | rtg: thank you for the link, here it tells that the config files are concatenated into one file | 14:08 |
| === \sh is now known as \sh_away | ||
| tseliot | BenC: did you review my patch? | 16:50 |
| lamont | BenC: which version of things was that vmmon, I wonder... no networking here.. FTL | 17:32 |
| infinity | BenC: Is the patch from #201591 queued for the next kernel upload, by any chance? | 18:26 |
| infinity | BenC: I'd love to be able to see myself type in consoles. :) | 18:26 |
| rtg | infinity: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-hardy.git;a=commit;h=4cbe826672e05ace08a4848e53135d652772feae will appear with the next upload. | 18:36 |
| infinity | rtg: Danke. | 18:38 |
| infinity | rtg: (For the record, despite the bug title, it's a regression for all *fb drivers, not just atyfb... I have the same bug on vesafb) | 18:38 |
| rtg | infinity: you are correct, the commit title is disingenuous, but the code is correct. | 18:41 |
| infinity | rtg: Yeah, I know the code is correct, the heads up was just for the Debian changelog to be accurate. :) | 18:42 |
| rtg | infinity: I'll try to remember when I do the upload. | 18:43 |
| infinity | rtg: Is there a timeframe on that? There's nothing scarier for me than running a devel release without consoles. :) | 18:46 |
| rtg | infinity: by next Friday for sure, perhaps sooner. | 18:49 |
| cradek | rtg: thanks, that patch is the one that fixed my system too | 19:00 |
| BenC | rtg: I'm doing an upload this evening | 19:04 |
| BenC | or, soon after on the weekend | 19:04 |
| tseliot | BenC: did you review my patch (or read my email)? | 19:15 |
| rtg | BenC: good. I was gonna ask Steve if there was any reason I shouldn't do an upload today. | 19:17 |
| deb | Am i connected | 21:30 |
| deb | I have downloaded 8.04 and found it is not a viable distro for me, and many others. | 21:31 |
| deb | The reason is the lack of internet apps | 21:31 |
| alex_joni | you do know that on the CD there's only a tiny amount of installed packages | 21:31 |
| alex_joni | there are about 18-19000 other apps in repositories, which you can easily install | 21:32 |
| deb | Pardon? There are more apps on the CD | 21:32 |
| deb | ? | 21:32 |
| deb | Oh, that. That is my point. As presently setup, I cannot get to the repositories with 8.04 | 21:33 |
| deb | Reason is the internet access apps are so skimpy. | 21:33 |
| infinity | Define "internet access apps". | 21:33 |
| infinity | Also, please define it in another channel (#ubuntu, perhaps?), this is A) not a support channel, and B) dedicated to kernel discussion. | 21:34 |
| deb | I have two locations and two different ISP's that I connect through - as setup I cannot connect from either with 8.04 | 21:34 |
| deb | One location has a wireless network and the main computer on it has a D-Link DWA-552 wireless card, which 8.04 does not recognize (There is a GPL driver for it beause Sabayon uses it) | 21:35 |
| deb | There is also no "ndiswrapper." | 21:35 |
| deb | So I can use the Windows driver | 21:35 |
| deb | The other location uses PPPOE -complete with user name and password - for each log on, there are no apps to allow me to do that (they are available as PCLinuxOS uses thenm)( | 21:36 |
| alex_joni | then use PCLinuxOS | 21:37 |
| alex_joni | as infinity pointed out, this is a place to discuss kernel development, which you obviously aren't interested in | 21:37 |
| deb | A less than "sensitive" anwer | 21:37 |
| infinity | We ship ndiswrapper and ppoeconf, both installed by default no less. | 21:38 |
| deb | No I am not interested in kernel development, but your folks on the desktop sent me here | 21:38 |
| deb | Where are they, I looked but did not find them. | 21:39 |
| infinity | deb: pppoeconf is in /usr/sbin, ndiswrapper modules are provided by the default kernel setup, though you might need ndiswrapper-utils-1.9, if you need the userspace apps. | 21:41 |
| soren | ndisgtk is.. Oh, he buggered off. | 21:43 |
| soren | *shrug* | 21:43 |
| infinity | It's always unforunate when we lose a "IS THIS THE KERNEL CHANNEL, GIVE ME FREE SUPPORT NOW, UR DISTRO SUCKS, U ALL SUXORS!!" type. | 21:44 |
| alex_joni | infinity: you forgot "I HAVE NO CLUE WHAT I'M DOING, BUT THERE IS SOMETHING WRONG WITH WHAT YOU DID" | 21:46 |
| infinity | alex_joni: I'm fine with people lacking clue but, yes, I agree that it's irksome when they blame their lack of clue on us clearly having broken their computer from afar. | 21:46 |
| alex_joni | and sometimes even meaning to do that (breaking their computers) | 21:47 |
| zul | if everyone based their stuff off of pclinuxos then a world will be a happier place | 21:54 |
| alex_joni | how do you guys debug machine crashes.. if there's nothing in syslog? | 22:02 |
| desrt | hai! i can has kernel hacker? | 23:32 |
| * desrt requires patch application | 23:41 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!