/srv/irclogs.ubuntu.com/2011/02/22/#ubuntu-kernel.txt

=== solarion_ is now known as Solarion
=== ailo_ is now known as help
beata|lemurHowdy y'all. I'm following the Ubuntu kernel compile guides, checking out natty, running the 'debian/rules binary-$local' method, on AMD64 system. What I'd like to figure out how to do, is build for i386 on that same tree.02:01
AceLanbeata|lemur: you need to setup chroot environment02:02
beata|lemurI presume, then, I can use the same build tree in the chroot?02:03
AceLanbeata|lemur: https://wiki.ubuntu.com/Kernel/Action/BuildChroot02:03
AceLanbeata|lemur: yes02:04
beata|lemurI'll read it up then, thanks.02:04
AceLanbeata|lemur: no problem02:04
psusiI'm trying to chase down a kernel BUG I just ran into on natty in dput().  It does BUG_ON(inode->i_state & I_CLEAR);  this was tripped when removing a dmraid partition device ( linear mapping ).  what is I_CLEAR?02:15
beata|lemur*giggle* The last chroot experience I had was trying to build the natty kernel on hardy. That was..ugh..painful.02:24
psusihrm... maybe a multiple free problem?  hrm...02:25
* psusi sure is glad for make -j402:26
beata|lemurHey. I'm happy for auto-jobbing. We didn't have that back in 1999.02:28
psusiauto jobbing?02:38
psusiyou know what we DID have in 1999 that we don't anymore?  a defrag package ;)02:39
* psusi really needs to get that put back into the main archive, or at least universe02:39
beata|lemur*laugh*02:43
beata|lemurOh dear. The chroot is grabbing over half a gig of archives.02:43
beata|lemurIt wasn't that long ago I wouldn't have had space on the whole disk. ;)02:44
ohsixit's going to be much huger by the time it gets underway02:47
beata|lemurI can believe it.02:49
beata|lemurThis chroot thing is a *lot* harder than I expected, even given the other one.04:31
* apw yawns08:21
apwcking, moin08:59
* smb returns08:59
smbapw, cking morning09:00
apwsmb, and to you09:00
ckingmorning09:02
aboganimorning all09:07
smbabogani, morning09:07
=== smb` is now known as smb
apwsmb, bug 10:30
apwbug #72306610:30
ubot2Launchpad bug 723066 in linux "[STAGING] ENE card reader fails with IO errors on streaming reads" [Medium,Triaged] https://launchpad.net/bugs/72306610:30
smbapw, ok, thanks. I will include that into the reply10:31
apwyeah, its pretty bust, we do not want it backported as it is for sure10:32
=== diwic is now known as diwic_afk
hertonapw, there are two bugs which also will be fixed by latest rebase, bug 722689 and bug 72231012:19
ubot2Launchpad bug 722689 in linux "visual corruption of all screen after upgrading to 2.6.38.4.18" [Undecided,Confirmed] https://launchpad.net/bugs/72268912:19
ubot2Launchpad bug 722310 in linux "Suspend fails due to tpm_tis: error -62 with kernel 2.6.38-4 (2.6.38-3 works)" [High,Triaged] https://launchpad.net/bugs/72231012:19
hertonI asked the reporters to test with the fixes included and they confirmed12:19
=== jjohansen is now known as jj-afk
=== hito_jp0 is now known as hito_jp
=== yofel_ is now known as yofel
apwherton, thanks, will make sure they are listed in the changelog  before i upload it13:09
rsajdokWhere is the list of bugs to bug day?14:57
apwJFo, ^^15:08
JForsajdok, we are looking at all the new status bugs15:09
JFoI'll get you a link :)15:09
JForsajdok, http://goo.gl/f5CAz15:12
JFothat is the list of all kernel bugs in the new status15:12
JFowe'll be working from there15:12
bjf##15:23
bjf## Ubuntu Kernel Team Meeting - Today @ 17:00 UTC - #ubuntu-meeting15:23
bjf##      agenda: https://wiki.ubuntu.com/KernelTeam/Meeting15:23
bjf##15:23
JForsajdok, here is a better list http://goo.gl/2wO4A15:26
tgardnerbjf, how about some CVE reviews so I can get them scraped off my list ?15:27
=== alex_jon1 is now known as alex_joni
bjftgardner, i'll add you to my list15:28
tgardnerapw, your -rc5 kernel seems to be working well on emerald15:31
apwtgardner, excellent ... just finishing up the boot tests here ... will shove it shortly15:31
tgardnerapw, one of the commit messages is wrong. it says -rc5 when it should say -rc6.15:32
apwdoh will correct15:33
tgardnerUBUNTU: v2.6.38-rc5 rebased fixes Bug #716811 ?15:33
ubot2Launchpad bug 716811 in linux "[SandyBridge] kernel BUG at /build/buildd/linux-2.6.38/fs/nfsd/nfs4state.c:3132!" [Undecided,Fix released] https://launchpad.net/bugs/71681115:33
apwtgardner, no that one is actually correct, it was an older bug15:33
tgardneralthough it may be correct15:33
tgardnerok15:33
apwjust added it for completeness as i closed it out15:33
tgardnerapw, the scheduler really seems to work well. even at 322 load average its fairly responsive on another SID15:36
apwtgardner, that is most in-credible15:36
rsajdokJFo: thanks15:50
JForsajdok, thank you :)15:50
tgardnerapw, lemme know when you've pushed natty master so I can start the ti-omap-dev rebase.15:53
apwtgardner, it should be as you have, i have just pushed the tag15:53
tgardnerapw, ack15:54
apwoh its not on master yet sorry15:54
apwwill do that right now15:54
apwtgardner, ok done on master correctly15:54
tgardnerapw, got it15:55
* apw goes do the lucid ones15:55
JFoneed cofee brb16:01
=== herton is now known as herton_lunch
keestgardner: seriously, can't we just remove debugfs?? http://seclists.org/oss-sec/2011/q1/22916:15
keestgardner: it's horrible16:15
tgardnerkees, IMHO its no worse then a regular sysfs. Its just that folks are paying less attention to debug interfaces. Perhaps we can get smb to agree to not mount it by default? We should also fix these holes.16:19
keestgardner: sysfs has a regular interface, debugfs has become a dumping ground16:20
* smb had no issues with not mounting it by default16:21
tgardnerkees, well, maybe we can look at turning off the worst offenders (after fixing some of the permission problems)16:21
keestgardner: we can also mount it somewhere that only root can get to16:21
keestgardner: right, I want to keep the arbitrary memory writing interfaces totally gone (acpi/custom_method)16:22
apwkees, might that not be achieveable by just chanign permissions on the directory to be mounted on16:22
keestgardner: but mostly this is just me pointing out yet more vulnerabilities in debugfs globally.16:22
keesapw: yup.16:22
tgardnerkees, would mounting somewhere else be sufficient ?16:22
tgardnernm16:22
apwwhat mounts them up?  is that upstart16:23
apwi think it needs to use debugs16:23
keestgardner: there are two classes of issue: root gaining arbitrary write (which we've fixed by dropping acpi/custom_method) and non-root messing with the system (for which there are tons of holes it seems)16:23
apwfor its boot tracing16:23
apwso we might need to fix it there16:23
keesafaik, ureadahead and the dri debugging hooks need debugfs16:23
keesboth should be running as root16:23
sforsheeI think debugfs is also needed for perf, isn't it?16:24
keesnone on /sys/kernel/debug type debugfs (rw)16:24
keesnone on /var/lib/ureadahead/debugfs type debugfs (rw,relatime)16:24
keessforshee: not sure.16:25
aboganiftrace16:26
apwkees, i meant more that upstart is likely mounting it early16:27
keesabogani: ftrace is a rootly thing, though, right?16:27
keesapw: likely mountall, but yeah16:28
apwkees, so something you could just try probabally16:29
* kees has 2 days before feature freeze. :)16:30
sforsheekees, looks like debugfs is needed for some perf functionality: https://perf.wiki.kernel.org/index.php/Perf_examples16:31
keessforshee: but that's all rootly stuff.16:31
keessforshee: i.e. not mounting it or making the mount 0600 shouldn't wreck that.16:32
sforsheekees, yep, just an argument for not ripping out debugfs support completely16:32
keessforshee: well... I'm against allowing the root user instrumenting the kernel in general. I'd prefer a "debug" flavor for that kind of thing, but I know it's a losing battle. :)16:33
keessforshee: the middle ground is to just keep from allowing root to manipulate arbitrary memory.16:34
keesapw: setting 0600 is still a kernel change (since the mounted fs determines the mode of the root node)16:35
apwkees, i thought it was the OR of the root in the FS and the directory onto which it was mounted16:38
keesapw: looks like not16:38
apwok, then likely yes a small change kernel wise16:39
apwas its just that top directory16:39
bjfubuntu-meeting16:44
bjf##16:48
bjf## Kernel team meeting in 12 minutes16:48
bjf##16:48
keesapw: I don't see a way to make the root inode 0600 without patching fs/libfs.c simple_fill_super()16:49
=== jj-afk is now known as jjohansen
=== herton_lunch is now known as herton
apwkees, hm16:55
apwprobabally needs a _mode() version16:56
keesapw: well, actually, I could change the mode after the simple_fill_super call16:56
keesapw: in debug_fill_super, test the simple_fill_super return, then set sb->s_root->d_inode->i_mode16:58
keesseems like a psychotic hack, though16:58
apwi think add a new versionw hcih takes a mode17:01
apwand make the existing one pass 0600 to it17:01
keesapw: yeah. there aren't many users of simple_fill_super17:02
JFograbbing a bite bbiab17:13
=== Quintasan_ is now known as Quintasan
jjohansensmb: good news, the hardy -xen kernels do boot on ec217:15
smoserwait17:15
smosermaybe17:16
smoseri don't know that they dont17:16
smoseri can test that fairly quickly17:16
jjohansensmoser: oh, I took it as you had already tested17:16
smoserno. i only know that i did something wrong when they failed.17:16
smoserlets see.17:16
jjohansensmoser: ah17:16
smbjjohansen, partially good news then :)17:18
jjohansensmb: yes17:19
jjohansensmb: smoser and I were discussing whether it was worth while pulling pv-grub back to hardy17:19
smoseroh wait...17:19
smoseri wasn't necessarily going that far...17:19
smoserfirst would be to get to using the archive kernels17:20
jjohansenhehe, /me over enthusistic again17:20
smoserthat would require some changes in image building17:20
jjohansensmoser: right17:20
smbjjohansen, I was just commenting like that. :)17:20
smoserusing pv-grub would be wonderful.17:20
jjohansenI think archive kernels is a given17:20
jjohansenyes it would be17:20
smoserok. 17:20
smoserso17:20
smoserlet me upload a kernel and ramdisk and see if they boot.  this is such a pain for hardy :)17:21
smoserwhich is why we'd want to get to pv-grub17:21
jjohansenyep17:21
keesapw: I'm testing this now, what do you think? http://kernel.ubuntu.com/git?p=kees/ubuntu-natty.git;a=commitdiff;h=eca3653fb5aaa514c734821dc68813bb0fdc49fc http://kernel.ubuntu.com/git?p=kees/ubuntu-natty.git;a=commitdiff;h=f787804f21a3070e14df32ce8098b3f0b4a4064617:23
smbjjohansen, smoser Though I think I noticed a slight pain with pv-grub once an ami is not booting. Just overriding the aki seemes not to work as simple (I guess I would need to publish a ramdisk there)17:23
smosersmb, well, maybe17:24
smbBut I generally agree on the beauty of pv-grub17:24
jjohansensmb: with hardy and lucid you could always just manually specify an aki instead of using the pv-grub aki17:24
smoserwell, they dont run17:25
jjohansenso it should be no worse than now17:25
=== bjf is now known as bjf[afk]
jjohansensmoser: darn17:25
smoseras there is no network driver in the image17:25
smoserunless you get the -modules into the image17:25
jjohansensmoser: okay, /me and smb will persue getting -xen to work17:25
smoseralthough... we could potentially fix that by getting the netowrk driver into the ramdisk17:25
smoseri think that would be very worth it17:25
smbyep. just having to fiddle with the network driver does sound like it could be done without breaking our ppa builders17:26
smosersmb, well, the ramdisks are manually built17:28
smbOk, so that would be an option, too. Probably even better than arguing for a change of the xen image if it can be avoided.17:29
smoseryeah.17:30
=== diwic is now known as diwic_afk
smosersmb, i might have been wrong on network17:31
smoserhttp://paste.ubuntu.com/570675/17:31
smoserthat is output of lsmod in a current instance17:31
smosernone of those look like network module to me 17:32
smbsmoser, neither to me. at least not an nif driver17:33
smoseroff topic, but i'm personally happy to see that kernel clock apparent time travel occurred on this hardy kernel (http://paste.ubuntu.com/570676/) ... ie, htats not a regression in lucid or maverick its been there for quite a while. i had never looked on hardy before17:34
smbIs ifconfig -a show something17:34
smbsmoser, I saw it occasionally when restoring an instance17:35
jjohansensmoser: time travel problems have plagued virtualization for a long time17:35
smosersmb, yes, there is definitely a network device17:37
jjohansenCONFIG_XEN_NETDEV_BACKEND=y17:41
jjohansenCONFIG_XEN_NETDEV_FRONTEND=y17:41
jjohansensmoser, smb ^17:41
smoserso maybe it was disk that was failing 17:41
smoserhold on17:41
smbjjohansen, yup that makes sense with a network interface there but not in modules17:41
smosersmb, you can poke around at ec2-50-17-93-174.compute-1.amazonaws.com if you'd like. please don't modify anhything though17:42
jjohansensmb: yep, /me was just double checking17:42
smosersmb, http://paste.ubuntu.com/570680/ is output of a hardy i386 instance booted with a differnet kernel ramdisk than it is expecting17:43
smoserthats why i had suspected no network driver17:43
smbhm, yes. on a very quick glance that looks like either the network driver is not working, the network not setup or the magic http-meta server missing17:45
smoserwell meta data service isn't missing multiple times in a row17:45
smoseri've only seen that happen maybe 3 times ever17:45
smbsmoser, Yeah, I only came up with it because that is what usually happened to me when trying on my test-system (thanks for the uncloudify script btw, though unfortunately time seems to have warped without me having any to try again)17:47
smoseryeah17:47
smoserright17:47
smoserbut it seems like its something else, and i suspect that it is solved by installation of the apprpriate linux-modules package17:47
smbIt is really hard to say. Disk and network look to be ok from the messages...17:48
smbsmoser, If I have understood jjohansen correctly he was syncing up the patches for the ppa image. So that would only leave some configuration difference to be found17:52
smoserright.17:53
jjohansenhence the question does, the new image boot17:53
smoserwell. it boots.17:53
smoserthat is for sure.17:53
smoseri suspect that re-bundling an image with its modules instlaled will work.17:54
smoserso i guess, we need to test that putting them there would fix this17:55
smoserand then try to work out exactly what is going wrong17:55
=== sforshee is now known as sforshee-lunch
=== bjf[afk] is now known as bjf
=== bjf changed the topic of #ubuntu-kernel to: Home: https://wiki.ubuntu.com/Kernel/ || Natty Kernel Version: 2.6.38 || Ubuntu Kernel Team Meeting - March-01 - 17:00 UTC || If you have a question just ask, and do wait around for an answer!
* tgardner --> lunch19:03
=== sforshee-lunch is now known as sforshee
=== shadeslayer is now known as shadeslayer_
apwJFo, you there?19:14
JFoapw, yep19:15
* apw mumbles at you19:16
* JFo is on there19:16
JFoI don't see you apw19:16
apwand i cannot connect hrm19:17
JFoah, there you are19:17
keestgardner: well there you have it; greg doesn't want to take it.19:25
=== ogra is now known as Guest4712
=== Guest4712 is now known as ogra_
apwkees, what was his objection19:43
JFoapw, bug 709591 is one we only tagged kj-triage19:43
ubot2Launchpad bug 709591 in linux "i915 doesn't detect HDMI display" [Undecided,New] https://launchpad.net/bugs/70959119:43
apwJFo, and weher can i find the script which did this19:45
apwJFo, cranberry?  if so where19:46
tgardnerkees, is he only objecting to 'debugfs: only allow root access to debugging interfaces' ?19:46
JFoit is in the kernel team branch of arsenal19:46
JFoit is the new-process-new-bugs.py under the contrib/linux/ dir19:47
apwand what output did the run produce for that bug19:48
* JFo digs back19:49
keestgardner: I would note that viro just said that debugfs "should not be mounted on production systems".19:49
keesapw: "debugfs has no rules and is for debugging only"19:49
tgardnerkees, I'm just replying in email something to that effect.19:50
keestgardner: okay19:50
tgardnerkees, and it was Alan Cox that said "don't mount debugfs"19:51
JFoapw, can't find that one in the output. but I did find bug 71441519:51
ubot2Launchpad bug 714415 in linux "BUG: unable to handle kernel paging request at 80808080" [Undecided,New] https://launchpad.net/bugs/71441519:51
JFothe output is here https://pastebin.canonical.com/43753/19:51
JFosame deal19:52
apwwhat was the next line of the output19:53
tgardnersforshee, wren't you having tpm suspend issues for awhile ?19:53
JFoapw, it was for the next bug19:53
JFowant me to pull out a few more?19:53
JFoI should note, the output there is with the --verbose option enabled19:54
JFoso there isn't much more for us to get without debugging print statements19:54
JFoor rather debug print statements19:54
sforsheetgardner, not me19:55
sforsheemine was an i915 regression19:55
tgardnersforshee, hmm, ok19:55
=== Quintasan_ is now known as Quintasan
smbtgardner, reading the tales from the unkown stable mailing list?19:56
tgardnersmb, about what?19:56
smbAbout the regression they caused on some tpm hardware19:57
tgardnersmb, oh thatt. no, I saw some patches going by on LKML that fix the tpm patch that was reverted19:57
smbtgardner, Yeah, noted the story on the stable mailing list. 19:58
* jjohansen -> lunch20:02
apwJFo, for it to print Action: <blank> i think something odd happened in skipit20:10
JFoI wondered that too20:10
JFobut I still can't wrap my head around it enough to determine what20:10
apw        if result and not self.reason:20:14
apw            self.reson = "skip it returned true"20:14
apwperhaps add (a correct) version of that to the bottom of skipit20:14
apwand run it in ry run and verbose20:22
apwdry20:24
JFok20:25
JFois it sad that it took me a bit of looking at that to determine what needed to be fixed? ;-P20:32
skaet_JFo, which specific packages do your tools search for, for identifying interesting bugs?  my guess is linux,  linux-ec2, and ???20:40
JFothe tools are looking only at linux for the moment20:40
JFowe are still working out some issues20:40
skaet_ok thanks.20:40
=== sconklin-lunch is now known as sconklin
bdmurrayI've been looking at some kenrel debugging documentation particularly https://wiki.ubuntu.com/Kernel/CrashdumpRecipe - is it current / accurate?21:39
bdmurrayapw: are you still affected by bug 451390?21:44
ubot2Launchpad bug 451390 in launchpad "limited upload rights no longer give series nomination accept/decline rights" [High,Triaged] https://launchpad.net/bugs/45139021:50
=== Nafallo_ is now known as Nafallo
apwbdmurray, has it been fixed?  i cannot trivially tell without asking mr watson to remove the work around22:06
bdmurrayapw: which work around is that?22:07
apwthe bug is that package sets do not give you the nom rights for the packages in, for the linux package mr watson added linux to each of the kernel uploaders direct rights22:08
apwthis works round the issue for the main package for which we use nominations, but does not solve the bug22:08
bdmurrayapw: ah okay I wasn't aware of *that* workaround22:09
apwyeah though it means the other 20 packages on my list i don't have noms for22:10
apwi think i'd need to find a bug on another package to confirm22:10
bdmurrayapw: its okay I'm playing with it on staging now22:10
apwbdmurray, that was the only thing stopping me moaning every single day22:10
bdmurrayapw: actually if you could try one of the other 20 packages on staging.launchpad.net that'd be interesting22:12
apwbdmurray, ok22:14
apwshame everything i do on launchpad also triggers timeouts22:15
apwdammit22:15
apwfileing a bug doesn't work on staging for me ... 22:15
apwbdmurray, it may be working (nomination s for other packages) i need to test it on live i guess22:17
apwhttps://bugs.staging.launchpad.net/ubuntu/+source/linux-meta/+bug/72241422:17
ubot2Launchpad bug 722414 in ubuntuone-android-files "syncable icon too small" [Medium,Confirmed]22:17
brycehapw, upstream is telling me "for all the 915/945 bugs can you please have the reporters test the latest kernel with the enlarged unfenced alignment."22:39
=== sconklin is now known as sconklin-gone
brycehgoogle isn't informing me what an "enlarged unfenced alignment" is... do you recognize that?22:39
ckingapw, do you never sleep?22:45
RAOFbryceh: That sounds like a libdrm commit to me?22:45
brycehRAOF, didn't get much more context than that22:47
brycehBryce, for all the 915/945 bugs can you please have the reporters test the22:47
brycehlatest kernel with the enlarged unfenced alignment. That's the most likely22:47
brycehcause of random writes, though I don't suspect it in this case.22:47
brycehfigure I shouldn't forward all the rest of these bug reports until I've gotten people to test whatever he's referring to22:48
brycehbut I suspect if it doesn't occur in tip, he won't care and we'll be on our own to dig out what patch(es) provide the fix22:48
jjohansencking: of course he doesn't22:49
RAOFbryceh: Sounds about right :)22:50
brycehRAOF, however `git log | grep "unfenced alignment"` in linus' tree returns null22:50
brycehnor found anything in ickle's drm-intel tree22:51
brycehwonder if he is sending me on a snipe hunt ;-)22:51
brycehRAOF, hmm wonder if this is a possible solution to the vesafb bug?  http://git.kernel.org/?p=linux/kernel/git/ickle/drm-intel.git;a=commit;h=86b27d8050b6b2aec31063fa9f40b16fb347afb322:55
RAOFbryceh: I suspect ickle's talking about 5e783323:11
brycehRAOF, ah thanks!23:19
smoserjjohansen, smb i kicked off a hardy-server uec build that is going to use linux-xen from the ubuntu archive.  it should appear in a couple hours on uec-images.  its not using -proposed kernel, but it is loads closer.23:24
jjohansensmoser: nice23:25
keessmoser: do you recognize "2.6.31-302.7-rs" as an Ubuntu kernel version? looks like a fork of an early karmic linux-ec2 version or something.23:31
jjohansenkees: that isn't one of ours23:32
keesjjohansen: I didn't think so.23:32
=== panda is now known as Guest34992
chrismsnzHi Guys, I have a set of Supermicro 2u twins (4 nodes) running Lucid. A couple of them have been having OS trouble under high load23:35
chrismsnzusually what happens is the machine is still reachable via connecting to a port (i.e. you can connect to the http port, memcache port and it will respond to ping), but anything further than that is unresponsive23:36
chrismsnzso you can connect to the HTTP port, but you cannot request a document - it will hang23:37
chrismsnzIt can work pretty hard but as soon as it "crashes" like this the only thing you can do is reboot23:37
chrismsnzI think it's a kernel problem with the version that comes with Lucid - what is the recommended kernel upgrade to try while still sticking with 10.04?23:38
chrismsnzrunning 2.6.32-27-server23:41

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