/srv/irclogs.ubuntu.com/2010/03/11/#ubuntu-kernel.txt

pace_t_zuluhey guys02:41
pace_t_zuluis it true there is a kernel freeze tomorrow?02:41
crimsunyes, but in effect it's already in place02:42
pace_t_zulucrimsun, are you aware of https://bugs.launchpad.net/ubuntu/+source/linux/+bug/535297 ?02:50
ubot3Malone bug 535297 in linux "BUG: unable to handle kernel NULL pointer dereference at 0000000000000028" [High,Confirmed] 02:50
pace_t_zuluMalone?02:51
EruditeHermithello, can anyone help me with this error?03:48
EruditeHermithttp://pastebin.com/Hnriax6403:48
EruditeHermitapparently its a bug with kernel package04:09
netritiousHi, is it possible to obtain the linux-headers packages for kernel version 2.6.32-15 ? They appear to be unavailable in the repository, but they are needed to workaround an issue a few are having with 2.6.32-16 and VBox.04:21
apwpace_t_zulu, Malone is the name of the launchpad software13:23
apwpace_t_zulu, ubot3 can translate other bug numbers scemes like CVE numbers and so prefixes its results with where it found things13:24
pace_t_zuluapw: thank you13:39
smbcking, So what other things. beside suspend resume can cause a tsc warp`13:45
smb?13:45
ckingsmb, not sure what triggers it, but I've seen it on Atoms - you sometimes see TSC warp warnings.13:45
* cking tries to think back 12+ months ago when he observed TSC issues on a N270.. hrm.. bit fuzzy13:47
smbAre those the that messages with -<huge number> marking tsc unstable?13:47
ckingno - just small warping13:47
ckingbig warp only observed when coming out of S313:48
smbCause there is also the "Marking TSC unstable due to halts in idle", but probably thats the clocksource framework and not the things you looked at13:48
apwsmb, the "two ack rule" things count as having two acks if they are sent by a core-committer and acked by one other ... right?13:48
smbapw, Usually not later. For now we could go with it13:49
apwrtg i've pushed qcm-msm branch into linux-meta ... you'll have to upload it the first time i guess13:49
smbapw, Thats why I usually pester two of you guys when I got a patch. :)13:50
tgardnerapw, k13:50
* apw pesters smb to look at tgardner's patch13:50
apwtgardner, i'll get the other one done now13:50
ckingsmb, if I recall, it's most probably an unstable TSC as reported by arch/x86/kernel/tsc.c13:50
tgardnerapw, the patch for -virtual ?13:50
smbcking, Oh, so warp not only means warping around to 0 but also going faster forward or even back?13:51
smbapw, tgardner was busy which one13:51
apwtgardner, yeah am getting smb to review that13:51
ckingwarp to zero --> wrapping,  warping where TSC seems to jump back/forwards in a manner that's not expected13:52
tgardnercking, do you suppose that happens when ACPI is reefing on C states as well as S states?13:53
ckingtgardner, can't say for sure, but it may be possible for C states too to affect it13:54
tgardnercking, that would make TSC well nigh unusable13:54
* cking agrees13:55
smbIt seems to be clear that tsc can halt in C states13:55
ckingsmb, have we any evidence of that - perhaps we need to get a script that tests for this13:56
smbcking, So that my problem. I havn't looked for some time but I thought that tsc is usually replaced by hpet for timer events13:56
ckingsmb, maybe so, but TSC does seem to be used for udelay by default if I understand the code correctly13:57
smbRight, probably we look/think of different usages13:58
cking..hopefully not missing any out13:59
smbOne thing is the clockevents framework for the monotonic and also the waking up/timer events (unfortunately its quite some time I looked at that code)13:59
ckingif the only good reason for using the TSC is lower-latency in udelay then maybe an alternative mechanism in udelay can be used and we ditch the TSC14:01
Q-FUNKsmb: it seems that AMD decided to get involved and contact the FIC engineering team themselves to solve bug #396286.  14:11
ubot3Malone bug 396286 in linux "[Geode LX] [ION603] 2.6.31-generic: kernel panic near the end of initramfs" [High,In progress] https://launchpad.net/bugs/39628614:11
Q-FUNKone thing I'm wondering:  what tools do we have to probe the BIOS and print the manufacturer and build date, etc.?14:11
smbQ-FUNK, sudo dmidecode14:12
smbIt would be nice to have someone look at that with some low lever hw information14:12
smbCause it looks like from a sw point of view I don't get farther14:12
Q-FUNKsmb: I suspected as much. Thanks for trying.14:15
smbQ-FUNK, I hope at least the information we have can help there14:16
Q-FUNKat least, the good news is that AMD considers what's left of the Geode market to be wrothy of salvaging enough to contact FIC themselves.14:16
Q-FUNKhopefully, yes14:16
Q-FUNKand it indeed seems that otavio was correct when he said that none of his other Geode hardware showed the symptoms.14:17
Q-FUNKthe program manager for AMD's embedded chips appologized on behalf of AMD, especially for the downfall of laying off Jordan Crouse, back then, and for having momentarily left Geode support in Linux and X.org in limbo.14:19
Q-FUNKthey now have one guy in US and a whole team in Taiwan working on both issues.14:20
smbsweet14:20
Q-FUNKthey're barely coming up to speed on the basics of e.g. using git, but I have a feeling that it's only a question of time before things fall back into place.14:20
Q-FUNKmy only worry is whether any of those issues will be fixed on time for Lucid's release or not.14:21
Q-FUNKI'd rather avoid the situation we had with the X driver on Hardy having to be resucitated after Hardy was released, but well...14:21
smbIt would have to be really fast to hit Lucid release. On the (maybe not  completely satisfying) side at least LTS gets a few point updates which rebuilds cd's14:23
apwsmb, in the first bit of good news of the day ... i plugged and external monitor into my lucid box and it didn't GPU hang on me ... so we do have the fixes :)  you are on there now14:24
smbapw, \o/14:24
Q-FUNK\o/14:24
apwactually i've missed having IRC always visible14:25
apwamitk, in mutt-patched how often do the stats on the left update ?14:25
Q-FUNKapw: you position your IRC app in the second display?14:25
apwQ-FUNK, i have today ... its 'only' 1280x1024 so too small for working on14:26
amitkapw: you mean the unread count?14:26
apwamitk, yeah the numbers over there14:26
amitkapw: only when I move my 'cursor'14:27
amitkor open another mail folder14:27
apwas in C-n C-p in that left panel14:27
amitkright14:27
apwcool better than not seeing them14:27
amitkyou don't see them?14:27
Q-FUNKapw: I wonder if I can do the same on this laptop running -intel.  when I'm at home, I'd love to slide specific apps on a separate display.14:28
apwno i do, i mean them updating only when i take action is better than not seeing them at all as i was without -patched14:28
apwQ-FUNK, works for me, this is a laptop.  added external, ran xrandr, and poop my desktop got wider14:28
Q-FUNKapw: interesting.  do you get separate desktops or just one large virtual desktop area?14:29
apwin the bad old days i used the 'tiny' laptop screen for irc et al and my 'huge' monitor for work ... but now my internal is bigger ...  14:29
apwall one desktop you can have windows spanning the break14:29
Q-FUNKhm.  ok. any way to have separate desktops, like in the old days?14:30
Q-FUNKIIRC with xineramara we had separate desktops.14:30
apwits not the default, i am sure its possible somehow, you'd have to ask on #u-x14:31
apwas this is what i wanted, i am not moaning14:31
Q-FUNK:)14:33
Q-FUNKhm.  due to the timezone difference, my e-mail exchanges with AMD Taiwan are pracically real-time.  pity their firewall prevents them from using IRC.14:33
* apw reboots to test a kernle14:42
tgardnerapw, strange install problem on Karmic when installing a Lucid kernel: 'linux-image-2.6.32-16-server depends on binutils (>= 2.20.1)'15:03
* apw blinks15:04
tgardnerits not like we're still doing dynamic linking15:04
apwtgardner, /me investigates15:05
apwtgardner, is that a binary from the archive or one you made15:06
tgardnerapw, from my archive mirror15:06
smbapw, Could that be because of linux-tools?15:06
tgardnerlinux-image-2.6.32-16-server_2.6.32-16.25_amd64.deb15:06
tgardnersmb, linux-tools is a 'suggests'15:07
apwsmb, linux ... what he said15:07
apwtgardner, i don't have that dep on a local build15:07
smbapw, In one of my past attempts I was unable to install the kernel image without linux-tools15:08
apwtgardner, can you less the package and paste your Depends: line15:08
apwsmb, yeah it changed15:08
tgardnerapw, Depends: binutils (>= 2.20.1), binutils (<< 2.20.2), libc6 (>= 2.8), libelf1 (>= 0.131), initramfs-tools (>= 0.36ubuntu6), coreutils | fileutils (>= 4.0), module-init-tools (>= 3.3-pre11-4ubuntu3), wireless-crda15:08
tgardnerits gotta be from ${misc:Depends}, ${shlibs:Depends}15:09
apwwtf ? _two_ binutil depends15:09
apwtgardner, which is the problem one ?15:09
tgardnerpretty tight constraints :)15:09
* apw wonders why i have none on my local builds ... arrrg15:10
tgardnerbinutils (>= 2.20.1)15:10
tgardnerwe kind of have to figure this out lest it raise hell with the kernel backports15:11
apwwell actually i suspect if you build it in a karmic chroot it'll be fine15:11
tgardnerapw, ok I'll do that15:12
tgardnerwhich is really the right solution anyway15:12
apwyeah we would be building them there .. but why do we have such tight contraints from that15:12
tgardnerapw, I think its coming from the wildcard dependencies. I'll have to read up on them15:13
tgardnerapw, while we're at it, we should reexamine the 'Provides:' boilerplate. most of that stuff is no longer accurate.15:14
apwtgardner, annoyingly for the linu-image package i put both those general defs together so we cannot tell by the position of the deps if some is from one or the other15:16
apwbut i suspect it is cause of the perf binary in there, it is tying us to the current binutils version15:16
apwtgardner, i am starting to think it needs to come out into its own package15:16
tgardnerapw, right, but as you've pointed out the backports will get build in a Karmic environment anyways15:17
apwinto linux-tools flavour package15:17
apwyeah. makes testing kernels on the wrong series a lot harder though15:17
tgardnerindeed15:17
apwto it might make sense to pull it out to its own package15:17
apwi will have a look how horrid that would be15:17
tgardneryeah, its not like there is any real binutils dependency, is there? initrams doesn't use it, right?15:18
apwnope ... its just cause of the binary i am sure ... perf15:18
apwi'll pull it out and see if the deps get better15:19
apwtgardner, what is a bit whack is my local builds have different deps:15:19
apw Depends: libc6 (>= 2.8), libelf1 (>= 0.131), initramfs-tools (>= 0.36ubuntu6), coreutils | fileutils (>= 4.0), module-init-tools (>= 3.3-pre11-4ubuntu3), wireless-crda15:19
apwthey would preclude install on karmic too, but they arn't the same which is most wack15:20
tgardnerwhy is the buildd different?15:20
tgardnerapw, could be we're just tilting at a windmill here15:20
apwyeah i am not sure15:21
apwi guess i need to review that chroot and rebuild it15:21
apwi'll go split it out and we can see how it looks15:22
apwit didn't seem sensible to have a separate package for one bin kernel version locked, but these deps will hurt our testing and so its prolly worth the pain15:23
tgardnerapw, yeah, but linux-tools doesn't have an ABI so how will you keep perf in sync?15:24
apwlinux-tools is the binary independant bits15:24
tgardneroh, you mean a separate package just for perf?15:24
apwwe'd ahve to have a kernel matches binary package for the tool  linnux-tools-15-generic, move linux-tools to linux-tools-common15:24
apwand have a linux-tools meta package15:25
apwyeah thats the proposal.  i thought i'd try it and see how vile it was15:25
tgardnerwhat a pain in the ass15:25
apwyeah... it probabally would have been the 'right' way to do it, we were lazy and it bites us on the ass15:25
manjoapw, smb brb15:44
tgardnerapw, so, a Lucid kernel installs better when built under Karmic16:10
apwgood16:10
tgardnernow lets see if it withstands a stress pass16:10
=== Hedge|Hog is now known as Hedgehog
=== Hedgehog is now known as Guest46804
=== smb is now known as gnarl
cndapw: you remember anything that went in -16 that would cause bug 536858?21:48
ubot3Malone bug 536858 in linux "USB-Audio devices do not show up as input devices" [Low,Invalid] https://launchpad.net/bugs/53685821:48
cndapw: nm, reporter invalidated the bug21:51
=== sconklin is now known as sconklin-gone
Ngare we aware of any virtio problems in lucid atm? I've been trying to install the alpha 3 iso image inside a virtio kvm on my lucid laptop and it gets to the package installation bit and ends up IO erroring itself read-only23:12
daftykinshi all, please excuse the newbie question... will Lucid be sticking with the 2.6.31 kernel through release and further?23:14
RAOFNo; it will be sticking with the 2.6.32 kernel though.  There are plans to pull newer kernels in post-release, but I don't know details of that.23:15
daftykinsah yes my bad, typo/memory failure :) ah ok, thanks very much :)23:16

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