[00:04] wtf [00:07] can it_rsa hold multiple private keys? [00:12] sure [00:12] but you probably wanted id_rsa ;) [00:13] er, yeah, that. [00:13] thaks [00:14] actually, i donno about id_dsa [00:15] I would think ssh would have an option to specify which key it tried to use [00:15] if it supported multiple keys in one file [00:16] or perhaps it tries all of them ;) [00:16] but i dont quite understand why -kernel is the place to ask about ssh [00:19] it isn't. I got confused :) [00:20] ... for linux? made me think this was #linuxhelp :) [00:20] ok ;) === Zhenech_ is now known as Zhenech [08:08] How do I pass a paramater to the modules that you're now compiling into the kernel? [08:09] .= on the kernel command line, perhaps? [08:21] soren, put them into a file in /etc/modprobe.d. Format "options =" [08:21] soren, Oh forget it [08:21] Should get my eyes open [08:21] Yeah, yours is correct [08:35] So there's no way to change these values, since I can't rmmod and then modprobe with new options? [08:35] *sob* [08:37] soren, If the code has no /proc or sysfs interface I fear not. [08:37] Oh, right. [08:37] AFAIUI, all module paramaters are exposed in sysfs these days? [08:37] Althought, they might be read-only. :( [08:37] soren: /sys/module/modulename/parameters [08:38] Should be there even if it's statically linked [08:38] mjg59: But that doesn't necessarily mean I can change the value. Some might require a reinitialisation. [08:40] soren: Right [08:40] If it's marked read/write then it should work on the fly [08:41] Hmm.. I don't even see processor.max_cstate exposed in sysfs anymore. [08:42] if the module_param is created with mode 0000, does that mean it doesn't even get exposed in sysfs? [08:47] include/linux/moduleparam.h suggests that that is indeed the case. [08:47] So I lose. :( [08:48] I have a laptop that makes a really annoying noise when on AC in C3 and C4, so I want to set max_cstate=2 when on AC and max_cstate=4 when on battery. I don't see a way to do that now. [08:50] You can use the pm_qos stuff to mimic that [08:50] If you set the dma latency to a number smaller than the C3 latency [08:51] *chuckle* [08:51] Well, yes, that would work. [08:52] Ok, so how do I find the relevant numbers? [08:54] And once I find it, I just write that as a signed 32 bit integer to /dev/cpu_dma_latency and hold it open until I switch to battery? [08:57] mjg59: in /proc/acpi/processor/CPU0/power, the C3 line reads (among other things, of course): "latency[057]", and the same number for C2 is 001... So I just put e.g. 56 in /dev/cpu_dma_latency ? [09:00] soren: Yeah, though I can't remember if the units are the same [09:05] mjg59: I'm not sure how to find out. include/acpi/processor.h just says the latency attribute is a u32. No mention of units. [09:09] soren: I /think/ it's usec [09:10] Lovely. That's what /dev/cpu_dma_latency expects as well, apparantly. [09:10] mjg59: Thanks. I'll tyr. [09:10] try, even. [09:46] mjg59: It seems to work. Thansk! [09:46] Thanks, even! [09:46] (typing is hard) [10:16] evening [10:16] (wrong window) [12:27] hello, did anyone mage to build 2.6.29-rc1? i tried yesterday but build failed... [12:27] sorry typo... mage=manage [13:17] how do you debug the kernel in ubuntu? [13:17] just stare at the source until the truth becomes clear? [13:20] the buddha method? [13:47] TimStarling, some of that yes [13:48] TimStarling: a combination of staring at the source and also the use of printk() :-) [13:48] some bits have debug you can enable, others you add as you stare [13:50] I have a production server with a top that looks like this: http://p.defau.lt/?84j7MTR_v4d7BUgx5E91bw [13:51] it works just fine for a few days, then dies [13:51] I could add printk() all over the whole kernel (one of us thinks the bug could be pretty much anywhere) [13:51] but a few days' worth of output might be a bit ugly [13:52] is that mysql really 30GB ? [13:52] how much ram do you have? [13:52] 32 GB [13:52] can you define dies [13:53] Erk... Those kswapd processes don't look too good. [13:53] connections to mysql time out, attempting to attach with strace hangs forever [13:54] soren: no shit [13:54] how is mysql aquiring all that memory? large shm segments? something else? [13:54] You only have 1GBof swap? [13:54] 30GB is normal [13:54] 31 actually [13:55] i would suspect the machine is trying to find something to release cause it has no swap [13:55] With 32 GB of RAM and a mysql server that wants to eat 31 GB of memory, 1 GB of swap sounds like a disaster waiting to happen. [13:55] and if the memory mysql has is anonymous this can trigger almost infinite list active/inactive list scanning [13:55] it may be that the margin is too small, our DBA likes to set it real small [13:55] in older kernels, what version is this [13:56] Linux db18 2.6.24-22-server #1 SMP Mon Nov 24 20:06:28 UTC 2008 x86_64 GNU/Linux [13:56] stock ubuntu kernel [13:56] yep ... .24 does not have the fixes for that [13:57] What else is running on the machine? [13:58] The run queue is ~22.. [13:59] http://p.defau.lt/?GNY7y_PM7vcQfy4C4bf0xQ [14:01] could be FS-related, with the logrotate processes [14:01] but I've seen OOM conditions crash ubuntu kernels before so adding a bigger margin is an obvious thing to try [14:02] apparently changing the FS was already tried [14:02] could be a block driver [14:02] i would suspect the unevicatable pages problem at first glance, kswapd's running madly and achieving nothing [14:02] note they arn't in D writing stuff, they are running [14:02] the changes to fix that are major and so not likely to appear in a hardy kernel [14:03] _if_ the memory that mysql is at least in theory swappable adding swap that will never be used may prevent the issue (iirc) [14:03] are there fixes in 8.10 that I should know about? [14:04] i believe the fixes are in 2.6.28, so that would be jaunty [14:05] possibly adding swap would tell you if that is the issue, though getting it back will be hard with only 1gb by default [14:06] the easiest thing is to reduce mysql's memory usage back to 28 or so [14:07] if it is reliable there yes [14:08] if the problem is the spinning lists issue, then that tended to get reported by people who had 32GB of ram or more [14:08] and tended to have 0 swap as 'who needs swap with this much ram' [14:08] mysql's memory usage is very stable, it just allocates a big chunk at startup and uses it for most things [14:09] we know enough to not set swap=0, lots of bugs if you do that [14:10] ok, going to reboot it, last chance if you want more data === ubot3` is now known as ubot3 [14:22] presumably there's some page table overhead, what's the effective amount of memory available for mmap() if there's 32GB physical? [14:23] I'm just wondering if that 30.9GB virtual figure for mysql was hitting it [14:24] i tend to work on the rule of thumb that 10% is used by the system but you were clearly 'at the end' as you kswapd was awake [14:42] you are only talking about having 1.1GB free, i have some 180MB missing even before the kernel gets going for the kexec kernel, so you are squeezing the entire system (other than mysql) into 900MB [14:45] mmm [14:48] looks like our (stable) 16GB servers have about 1.7GB free for non-mysql stuff [14:49] so I should probably be thinking about keeping the same proportion free, ~3.4 [14:50] we would like to have graceful OOM behaviour at some point, we've had a lot of trouble with that on our application servers [14:50] because they have a load-dependent memory usage [14:51] maybe we should package our own kernels [15:06] soren: is there any reason we should enable CONFIG_NFSD_V4 in the hardy -virtual flavour? [15:06] s/should/shouldn't/ [15:09] rtg: I can't think of any, no. [15:10] soren: thanks. looks like it must have been an oversight. === TimStarling is now known as Tim-away [16:25] Good day. I'm trying to understand the difference between the "lpia" and "lpiacompat" kernel flavours. A diff -u of the config confuses me, as config.lpia seems to have CONFIG_M586=y and config.lpiacompat seems to have it unset, which is contrary to the information I have previously received. Is there a set of known devices on which the lpia kernel is not expected to work? [16:27] persia: lpiacompat was built for a standard PC, e.g. a non-Atom CPU. [16:27] at least, if I recall correctly. [16:27] rtg, That's the information I had before, which is why I'm confused at the config values I'm seeing. [16:28] Atom doesn't need CONFIG_M586=y and many "standard" PCs (e.g. Via C7M) do. [16:29] where is Amit? He knows all these details. [16:29] You ask the hard questions :) [16:39] soren: where does the JeOS kernel image come from in Intrepid? [16:39] rtg: The server image. [16:39] It's the same image, with a stack of modules ripped out. [16:39] I think BenC calls it subflavours? [16:39] soren: this guy is complaining that NFSv4 doesn't work, but it _is_ enabled. [16:40] ah, perhaps too may debs are ripped out? [16:45] sconklin, ok ... you have filed a bug for the apport changes. acording to sources (#ubuntu-bugs) it seems the appropriate thing to do is to share that bug for all the changes for this overall feature so that they all have the same number and track together [16:46] Well, assuming they are all related to the same issue. [16:46] apw: that makes a lot of sense [16:46] I'm glad I gave it a generic title [16:47] it can be changed anyhow! === asac_ is now known as asac [18:47] hi gusy [18:47] who is in charge of the hardy-updates tree? [18:48] rtg: ? [18:48] fabbione: dude [18:49] fabbione: uh, smb_tp is the Hardy guy [18:49] rtg: hey.. who does pull from linux-2.6.24.x into hary tree? [18:49] something is hutterly wrong in the kernels you are releasing [18:49] smb_tp_: you there? [18:49] how so? [18:49] there are missing commits [18:49] and changelogs are missing entries [18:49] missing from stable? [18:49] fabbione, I am [18:49] yes [18:50] smb_tp_: i am looking at linux (2.6.24-23.46) hardy-proposed; urgency=low [18:50] according to the changelog: [18:50] lots of entry for bug: #301608 [18:50] that is merging 2.6.26.4 into hardy [18:50] but upstream 2.6.24.4 has: [18:50] commit 68b498d251d97de9adda518fda42cfe1451063b7 [18:50] Author: David S. Miller [18:50] Date: Thu Mar 6 14:47:20 2008 -0800 [18:50] SPARC64: Loosen checks in exception table handling. [18:50] [SNIP] [18:50] that is missing from the hardy tree completely [18:51] or at least is not in the released tarball [18:51] I think thats because we haven't incorporated all of the stable commits in Hardy as we have for Intrepid. smb - is that still the case? [18:51] this is an example of a missing commit [18:51] broken changelog: [18:51] * Linux 2.6.24.6, 2.6.24.7 [18:51] - LP: #301634 [18:51] rtg, No the intention was to gte all. [18:51] Let me see [18:52] ^^ points to the same bug as 2.6.24.4 and they have no references to the 2 security bugs fixed in those 2 upstream releases [18:52] smb_tp_: i simply need to know if you are going to pull them all or not.. [18:53] as easy as it, it means for me removing ubuntu from all my machines or keeping it around [18:53] fabbione, They should all get pulled [18:53] smb_tp_: ok then you should check why that one is not there [18:53] 68b498d251d97de9adda518fda42cfe1451063b7 [18:53] i know this is missing because without it my hardy sparc firewall doesn't boot [18:54] fabbione: its in the hardy git repo. [18:54] rtg: it's not in the released tarball [18:54] it may not have been promoted to -updates yet [18:54] rtg: it is according to the changelog... [18:55] i don't use proposed.. that kernel has been copied from proposed to updates [18:55] fabbione: gimme a minute.... [18:56] sure even 20 [18:56] i am honestly worried that something in the release process went wrong [18:56] on the other side i also have the issues to keep my machines running :) [18:56] i am sure you can understand [18:56] anyway you'll sort it out :) [18:57] fabbione: 2.6.24-23.46, right? [18:58] yes [18:59] do you need any more info? [18:59] fabbione, We well do [18:59] my wife is about to yell at me if i don't show up [18:59] fabbione, No I don't think so [18:59] ok [18:59] i'll pass by in about 2 hours.. [19:00] root@vultus5:~/linux-2.6.24# patch -p1 < ../fix_sparc.diff [19:00] patching file arch/sparc64/mm/fault.c [19:00] ^^^fix_sparc is that commit on top of that kernel.. [19:00] so it is missing somehow :) [19:00] later! [19:40] re [19:41] rtg, smb_tp_: found anything? [19:42] fabbione, Yes, found some patches missing (including yours) for 2.6.24.3->2.6.24-4 [19:42] smb_tp_: ok! [19:42] I did not understand however, the thing about 2.6.24.6/7 [19:42] that's something to do with changelog [19:42] not sure if it's automatically generated or manually done [19:42] but: [19:42] automatically [19:42] then it's wrong :) [19:43] look at the same kernel changelog [19:43] * V4L: Fix VIDIOCGAP corruption in ivtv [19:43] - LP: #301634 [19:43] * Linux 2.6.24.6, 2.6.24.7 [19:43] - LP: #301634 [19:43] ^^ last two entries for that release [19:43] .6 and .7 points to the same bug as .4 [19:43] (or .5... doesn't matter) [19:44] 2.6.24.6 fixes one bug, not reported in the changelog [19:44] .7 fixes another bug.. not reported in the changelog [19:44] as long as the fixes are there, this is "minor" but makes the changelog incomplete [19:44] There might be some not in that log because they got fixed before with another name potentially [19:45] worth checking tho [19:45] need to reboot my firewall.. brb [19:45] The updates were only to track importing the patches from -stable that not yet went into hardy [19:49] gotcha [19:49] re [19:49] well at least that fix works :) [19:49] smb_tp_: I think I remember pointing out that there was a typo in the changelog for that bug number. [19:51] ooi, what does SAUCE mean in Ubuntu kernel changelogs? Googling was unfruitful [19:51] maxb, That is a patch that is not (yet) upstream [19:51] maxb: we use it to denote patches that are either not upstream, or patches that will never get upstream [19:54] rtg, Hm, somehow I am blind to that. I posted the missing 6 to the list [19:55] smb_tp_: you are blind to what? [19:55] smb_tp_: did you check that manually or a git pull did show them? [19:55] oh, the typo? [19:55] rtg, yes [19:55] fabbione, I compared the export names and checked all missing from my list [19:56] smb_tp_: maybe it was a different changelog entry. [19:56] smb_tp_: it might be wise to check the whole tree tho.. [19:56] smb_tp_: just to make sure other bits didn't get lost in other pulls [19:57] fabbione, I will double check that [19:57] smb_tp_: is linux_2.6.24.orig.tar.gz a pristine copy from upstream or is it generated from your git trees? [19:57] smb_tp_: if it's a pristine copy from upstream, then it's "simple"... [19:57] if it's generated from your git tree, you want to check the whole thing from scratch [19:58] fabbione: its pristine from upstream [19:58] rtg: ok.. then it's simple.. [19:59] fabbione, don't think that simple the orig is the tree at point of first release [20:00] smb_tp_: yes it's simple in git.. you can just branch out v2.6.24 in git and then compare the commits in your tree with whatever is in 2.6.24.x [20:00] smb_tp_: everything local to ubuntu has proper UBUNTU or SAUCE tags. EVerything else either matching commits (everything in .24.x should be there) or a very detailed commit entry [20:01] but starting from v2.6.24 tag and going, will reduce the load a lot [20:01] anyway.. again... you will sort it out :) [20:02] I guess so :) [20:03] you are all young guys full of energy :) [20:04] who? young? :) [20:15] fabbione, rtg Ok, I went through the remaining commits of 4,5 and 6. Looks like those from 3 were the only missing ones [20:16] "Remote root kernel exploit..." ... [20:16] :) [20:16] ^^doens't exist but can scare the hell out of people ;) [20:18] heh, well you can scare people even by "panic can't find /" as a fortune message. :) [20:25] speaking of which.. do I get a pony for finding this problem? :) [20:25] * fabbione really really wants a pony [20:25] fabbione: I'll buy you a beer the next time we cross paths, how about that? I can't afford a pony. [20:26] Me neither [20:26] But we can make two beers of it [20:26] rtg: ahha works for me.. [20:26] rtg: even if i doubt somebody will ever fly me to a UDS or anything :) [20:28] fabbione, There might be other opportunities ... :) [20:28] smb_tp_: very unlikely tho.. i am located in DK.. [20:28] I'm in DE (just one letter difference) :-P [20:29] smb_tp_: well you could.. rtg might be problematic :) [20:29] fabbione: Its Barcelona in the spring... [20:29] rtg: again? [20:29] we have been close to Barcelona [20:30] rtg: do you have dates already? [20:30] It was Seville last time in Spain. this one is somewhere in the Catelonia district. May 22 I think. [20:30] rtg: there was one before Seville .. a long time before Seville.. in a small city close to Barcelona :) [20:31] rtg: at that time the kernel team was made of me and.. me... [20:31] ah, I wasn't around for that one. [20:31] nope :) [20:31] not even Ben was there [20:31] UDS - 25-29 May [20:31] IIRC it was right after warthy release [20:31] ok.. i'll keep an eye [20:31] can't efford to fly on my own.. maybe i can find a sponsor or two [20:32] send an email to Mark :) [20:32] he is not my only sponsor :) [20:33] night guys [20:33] take care [20:33] and try to get those fixes out :) [20:33] so i can finally stop building my kernels ;) [20:34] they'll percolate in -proposed for awhile. Should be uploaded in the next day or so. [20:35] Right, I am on it [20:35] yeah that's fine... [20:35] day more, day less.... worst case /. will have a nice story of 100's of Ubuntu machines being r00t3d :D [20:36] is there a new root exploit? [22:53] hum, how come linux-libc-dev is only for i386/amd64/armel? since it contains the drm headers now the xserver build fails on other archs (no drm.h) [22:54] although armel fails too.. === Tim-away is now known as TimStarling