/srv/irclogs.ubuntu.com/2008/03/26/#ubuntu-kernel.txt

JanCadinc: I have a laptop that uses iwl3945 fine, so I wonder what your issue is exactly...00:10
JanC(what your issue is exactly caused by)00:13
adincJanC: it is not only my issue00:20
JanCyeah, I saw the bug00:20
adincthe problem is described in the bug 18547000:20
ubotuLaunchpad bug 185470 in linux "iwl3945 not functioning : microcode error" [Unknown,In progress] https://launchpad.net/bugs/18547000:20
adincJanC: so i wonder how this device works for you00:20
JanCbut it works fine for me, so I guess the driver can work properly00:20
adincthat would be great00:20
JanCmaybe it's just not working fine on some variations or combinatiosn with other hardware?00:21
adincbut unfortunately it doesn't here, i'm just downlaoding the kernel sources in order to make use of the bugfix from mohamed abbas00:21
adincJanC: possible, but how should we figure this out, i'm not very well involved into driver programming for wireless devices. we wouldneed to do a lot of work, theonly ones who could do this actually are the developers00:22
adincof this driver00:22
adinccool the kernel now supports hsdpa00:23
JanCI guess the Intel developers, with some help of the Ubuntu kernel developers, would be the best people to investigate  ;)00:23
adincyes, you are right, but as i can see there is already work done.00:24
JanCmight be, looking at that comment00:25
adincJanC: which kernel version are you running?00:25
adinccomment 2500:25
JanClatest hardy kernel available this afternoon (8 hours ago or so)00:26
JanCmaybe 10h00:26
adincthere hasn't beeen any kernel changes in hardy, was there?00:26
adincif you do a uname -a, what do you get?00:26
JanCI haven't checked in the last 8-12h  ;)00:26
JanClet's check on my laptop00:26
JanCkernel build based on linux-meta 2.6.24.12.13, which seems to be the last (according to packages.ubuntu.com)00:30
adinci've 2.6.24-12-generic00:31
JanCyeah, that's what uname says00:32
adincso your kernel is 2.6.24-12-generic?00:32
JanCuhu00:32
adinci suppose uhu means yes00:33
JanCyes ツ00:33
adinci don't know the numbering you showed above, there must be a mistake in your kernel versioning scheme00:34
JanCmy laptop seems to have a 3945 "rev 2" wireless with PCI-ID 8086:4222 00:38
JanCthe version number I showed is for the kernel meta package--it shows the internal build number used by the Ubuntu kernel developers I suppose00:39
adinchttp://packages.ubuntu.com/hardy/allpackages00:40
adinchere are the kernel packages listed00:40
JanClinux-meta is the source package (it's called "meta" because it builds several different kernals)00:42
JanCsee http://packages.ubuntu.com/hardy/linux-generic and then in the bar at the right00:42
JanCanyway, it's just an internal Ubuntu kernel version number00:43
adinci see00:43
adincis there a way to see when the latest kernel was published for ubuntu hardy?00:44
JanChttp://packages.ubuntu.com/source/hardy/linux-meta should be pretty up-to-date00:45
JanCotherwise, if you update hardy from the main servers, it should be obvious  :-)00:45
adinca dist-upgrade i suppose would download the latest kernel build, wouldn't it00:47
JanCyes, it should00:47
JanCwhat device ID does 'lspci -nn' give for your wireless?00:49
JanCor 'lspci -n' maybe00:49
adinc8086:422200:50
JanChm, so it's supposed to be exactly the same hardware00:50
adincyes00:52
adincJanC: have you got a smp kernel?00:54
adincsomehow a dual core or more than one cpu?00:54
JanCthe -generic kernel is SMP  ツ00:55
JanCand I have a Core Duo, so it's actually using SMP too00:55
JanCfro mthe upstream bug report: """Guess what... I enabled the iwl3945 debugging (and disabled CGROUPS to get system back to usable state), and problem is no longer reproducible easily."""00:56
JanCsounds like a fun bug to "debug" that way  ;-)00:56
adinci'm compiling the kernel now.00:57
adincdoes he mean debugging enabled by adding something like debug=0x44444 for modprobe iwl394500:57
JanCsomething like that00:58
adincwhat about cgroups?00:59
JanCif I understand things correctly (but I am not a kernel devloper in any way), it seems like the firmware crashes or gives errors when it's under load (sometimes)01:03
adinci think that is a different bug01:03
adincwhen the module loads a firmware, which is the case when the device is configured, then the problem appears01:04
Nghas the behaviour of cpu frequency scaling intentionally changed since <=gutsy?07:28
Ng /sys/devices/system/cpu/cpu1/cpufreq is a symlink to cpu0/cpufreq07:29
Ng(which gets lost on suspend/resume because we save the value of each and then reset it to performance for the actual suspend)07:31
Ngbug #16265207:31
ubotuLaunchpad bug 162652 in pm-utils "pm-utils changes default cpu policy after resuming from suspend-to-ram" [Undecided,Confirmed] https://launchpad.net/bugs/16265207:31
krautmoin08:14
adinccan someone have a look into the kernel config file of hardies kernel if it has a driver compiled in called IWL3945, for config-2.6.24-12-generic08:39
sorensoren@butch:~$ grep IWL /boot/config-2.6.24-12-generic 08:56
soren# CONFIG_IWLWIFI is not set08:56
sorensoren@butch:~$08:56
sorenadinc: ^^08:56
pwnguinyour server names are strange ;)08:59
adincsoren: sorry?09:01
adinci see09:01
adincwell but what about this bug 185470 then09:02
ubotuLaunchpad bug 185470 in linux "iwl3945 not functioning : microcode error" [Unknown,In progress] https://launchpad.net/bugs/18547009:02
adincso this module is not even enabled09:02
adincsoren: could you please have a look for me if this module exists in your /lib/modules/2.6.24-12-generic/ubuntu/wireless/iwlwifi/iwlwifi/compatible/iwl3945.ko09:04
sorenadinc: It's probably in linux-ubuntu-modules.09:04
adincsoren: but still it would need to be enabled in the kernel09:04
adincthe .config file should show that this module is enabled or not, valid for the whole kernel09:05
sorenadinc: Erm.. no?09:05
adincno?09:05
sorenlum has it's own config.09:05
adinclum?09:05
sorens/it's/its/09:05
adincwhat is lum?09:05
sorenlinux-ubuntu-modules09:05
pwnguinisn't ubuntu kernel packaging fun?09:05
adincsoren: can you please tell me where this config is?09:05
sorenI'm laughing on the inside.09:05
adinci'm new to ubuntu09:06
sorenIt's in git.09:06
adincpwnguin: looks like...09:06
sorenI don't know if it's installed anywhere.09:06
adincsoren: ok, how can i get it09:06
adincwith git09:06
adinci mean where from09:06
sorenWEll, you can use look at it through gitweb?09:06
sorenhttp://kernel.ubuntu.com09:06
adinci don't know why this is done this way, but makes look through more complicated09:07
adincso we won't be able to compile the same kernel here withouth this config file09:07
sorenThe main repo tracks the upstream kernel + a few patches here and there. Extra drivers that we add are in lum.09:08
pwnguinapt-get source should help you09:08
adincwhat does lum mean09:08
pwnguinLinux09:09
pwnguinUbuntu09:09
pwnguinModules09:09
pwnguinits a package09:09
adincohh09:09
* soren chuckles09:09
soren09:05:37 < adinc> what is lum?09:09
soren09:05:40 < soren> linux-ubuntu-modules09:09
adincyes, but i didn't realize this was the shortened way for linux-ubuntu-modules09:09
adincmy mistake09:10
adinc;)09:10
sorenNo worries.09:10
alex_joniadinc: as an exercise you'll have to figure out lrm by yourself :P09:11
pwnguinheh09:11
pwnguinIs there a published guide to rebuilding the ubuntu kernel packages, for sysadmins and beginning hackers?09:12
pwnguini used to build my own kernel years ago on debian, but its become so much more complicated09:12
alex_jonithere is a guide around wiki.ubuntu.com09:12
adinchttps://wiki.ubuntu.com/KernelTeam/GitKernelBuild09:12
alex_jonipwnguin: actually it's a bit easier once you know the steps to do it :)09:12
alex_joniI never had a thing for make-kpkg :)09:12
pwnguini liked make kpkg09:13
pwnguinmuch easier to revert when i screw up things09:13
alex_jonipwnguin: oh, it still builds debs which you install09:13
alex_joniand revert easily09:13
alex_jonibut not using make-kpkg, using debian/rules from the ubuntu linux source09:14
pwnguinthe wiki page adinc pointed out uses make-pkg =/09:14
aboganihttps://wiki.ubuntu.com/KernelMaintenanceStarter 09:15
aboganimake-pkg isn't support at all09:15
adincno, he means make-kpkg09:16
alex_jonilook at https://help.ubuntu.com/community/Kernel/Compile09:16
chaosmasterHi09:32
chaosmasterI've done some pm-utils and suspend testing over the holidays and found an issue with resuming on my Samsung P35 notebook and radeon 9700 (r300) based card.09:34
chaosmasterIn contrast to current ubuntu practice radeonfb is *absolutely* needed for working resume09:35
chaosmasteranyone interested?09:35
alex_jonichaosmaster: there was a bug related to this reported at launchpad.net, I'd suggest you search there first09:35
=== \sh_away is now known as \sh
chaosmaster"[Bug 201591] atyfb regression - screen blank except for blinking cursor after fbcon vtswitch"  is another issue and I already commented it09:41
ubotuLaunchpad bug 201591 in linux "atyfb regression - screen blank except for blinking cursor after fbcon vtswitch " [Medium,In progress] https://launchpad.net/bugs/20159109:41
alex_jonidid you try the recent fix?09:42
chaosmaster"81722 Laptop sometimes fails to resume from ACPI S3 suspend"  is the only one mentioning radeonfb with resume/suspend and mentions another issue09:43
alex_jonithe one from mjg on the 23.03.2008 ?09:43
alex_jonichaosmaster: then I suggest you file a new bug report for the issue you're having :)09:43
chaosmasteryes, I patched my kernel - without radeonfb breaks completely09:43
chaosmasterI wanted to ask about the ubuntu policy in not allowing any framebuffer drivers, as they are blacklisted09:44
chaosmasteralex_joni: for which component - the kernel?09:45
alex_jonichaosmaster: sorry, not in a real position to be more precise.. :(09:45
chaosmasteralex_joni: OK, I'll file a new bug09:47
sorenI have a laptop that since upgrading to hardy doesn't suspend nor hibernate anymore. Both end up with a blank vt with a blinking cursor in the top left corner. AFAICS, all the quirks listed on the pm-{suspend,hibernate,suspend-hybrid} man page deal with the resuming process..09:51
soren..so I'm at a bit of a loss. What to do?09:51
chaosmastersoren: what laptop do you use and what is your graphics card?09:52
sorenI believe it's called a Uniwill something-something. It's got Intel graphics.09:53
sorenSuspend and hibernate have both worked flawlessly since breezy on it, afair.09:54
sorenHm... I could try booting the gutsy kernel.09:54
chaosmastersoren: please paste the output of: lspci | grep VGA09:54
sorenThat might provide a hint as to whether it's a pm-utils issue or a kernel on.e09:55
chaosmastersoren: not really :)09:55
sorenWhy?09:55
sorenIt's an intel 855.09:57
sorenchaosmaster: Why couldn't booting an older kernel provide a hint as to whether it's a pm-utils issue or a kernel one?09:57
chaosmastersoren: the suspend/resume process is very delicate and complex with a lot of parts beeing integrated. in Hardy a lot of new technologies (kerne, xorg, pm-utils) are updated so it's not that easy to pinpoint it to a single part09:58
sorenEr... Sure it is.09:58
sorenDon't start X, boot an older kernel, and presto! All you've got left is the new pm-utils.09:58
adincdoes someone know how to get the processes below the window with top ? sorry for asking here, but i suppose not many people are using it09:58
alex_joniassuming it works with the older kernel09:59
sorenalex_joni: 09:54:20 < soren> Suspend and hibernate have both worked flawlessly since breezy on it, afair.09:59
sorenadinc: I don't understand the question?10:00
elmargolWhere can I ask iwlwifi specific questions?10:00
chaosmastersoren: are you ready to test some quirks?10:00
adincsoren: if you call top, the process listing application, it shows the processes listed till the bottom of the current window. what do you do in order to see the rest of the processes10:01
adincelmargol: have you got a problem with iwl3945?10:01
sorenadinc: Stretch the window? :)10:01
sorenchaosmaster: Which ones, for instance?10:01
adincsoren: cool, but i can't find a 24 inch monitor for the time beeing10:01
elmargoladinc, yes  I have bug #17627110:02
ubotuLaunchpad bug 176271 in linux-ubuntu-modules-2.6.24 "major throughput difference (between upload and download) when using iwl3945" [Undecided,Confirmed] https://launchpad.net/bugs/17627110:02
elmargolhttp://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=159210:02
sorenadinc: You can make the window larger than your screen, you know..10:02
ubotubughost.org bug 1592 in data transfer "Speed drops when downloading at high speeds" [Normal,Assigned] 10:02
adincelmargol: do you use hardy10:02
elmargoladinc, yes10:02
chaosmastersoren: I don't have intel GMA specific knowledge, but vbestate-restore works for me on my ATI radeon 9700 M10:02
adincsoren: heheh, but thats not the solution10:02
adincelmargol: so you are one of those lucky guys, you at least get a connection10:02
sorenchaosmaster: This is hibernation and suspension. Not resuming.10:03
elmargolMy connection is fast 2-3MB/s after about 100-200MB traffic I get stuck at 1Mbit/s10:03
chaosmastersoren: best way is to find the most similar notebook / card and try the quirks listed there10:04
adincdo you have kernel 2.6.22 or 2.6.24?10:04
elmargol2.6.24-12-generic10:04
chaosmastersoren: suspend is (almost) never the problem - it's always about the resuming10:04
sorenchaosmaster: Look... vbestate-restore is about loading the vbestate on *resume*.10:04
sorenchaosmaster: It doesn't suspend!10:05
chaosmastersoren: oh, then I misunterstood your problem10:05
elmargolI'm using Ben Collins git tree10:05
adincstrange, here i can't get it running at all with a samsung q45, but lets wait i'm compiling a fresh git tree10:06
adincsee bug 18547010:07
ubotuLaunchpad bug 185470 in linux "iwl3945 not functioning : microcode error" [Unknown,In progress] https://launchpad.net/bugs/18547010:07
sorenJust tested with 2.6.22. It suspends just fine.10:07
sorenIt doesn't resume properly, though, but that's just quirks, so that's easy.10:07
adinchow is ubuntu managing firmware loading, is it hotplug doing?10:08
sorenadinc: I belive it's udev.10:09
adincsoren: i see10:09
elmargolI try iwlwifi HEAD now...10:09
adincthose changes should bit in kernel git10:09
chaosmastersoren: do you use only ubuntu kernels?10:10
sorenchaosmaster: Yes.10:10
adincelmargol: what device are you using it10:10
elmargol0c:00.0 Network controller: Intel Corporation PRO/Wireless 3945ABG Network Connection (rev 02)10:10
adincelmargol: same here, whichnotebook is it?10:11
elmargolDell inspiron 940010:11
adincohh, my new kernel hangs when loading acpi modules with init10:12
elmargolI'm using v1.2.2510:13
adincof what?10:13
elmargoliwl394510:13
elmargolthats from 02-04-0810:14
chaosmastersoren: if the same system suspends fine just with an older kernel, the you have to try other kernels between 2.6.22-working and 2.6.24-nonworking - the problem is I don't know of any mirror with all older ubuntu versions10:14
sorenchaosmaster: I know how to use git.10:17
sorenchaosmaster: I'm just not unambiguously a fan of bisecting as a debugging tool.10:18
chaosmastersoren: but then you have to compile them and I tried it yesterday - it was not funny, as I didn't find a way to compile only one flavour. By default debuild compiles -generic -386 -server ....10:19
sorenI also know my way around Ubuntu's kernel build system by now :)10:20
sorenBuilding just a single flavour is even documented on the wiki.10:20
chaosmastersoren: so you know more than me :)10:21
elmargolIs there a way to have the card in g only mode?10:23
alex_jonichaosmaster: iirc it's something like debian/rules binary-image-i38610:25
=== asac_ is now known as asac
BenCGood morning everyone11:17
chaosmastergood morning (even if it's past 12 here in DE) :)11:18
=== \sh is now known as \sh_away
=== \sh_away is now known as \sh
elmargoladinc, switching my accesspoint do mode b seems to fix the issue11:48
elmargolat least it takes longer to happen11:49
AnAntHello, will bug #201591 be fixed in next kernel upload ? will it use that patch suggested by Garrett ?12:04
ubotuLaunchpad bug 201591 in linux "atyfb regression - screen blank except for blinking cursor after fbcon vtswitch " [Medium,In progress] https://launchpad.net/bugs/20159112:04
sorenmjg59: Maybe you could help me out a bit? I have a laptop where 2.6.22 suspends and hibernates just fine, but on 2.6.24, it fails. I get a blank vt (blinking cursor in the top-left corner) and then it just sits there, never actually suspending or hibernating. I'm happy to fiddle around with the kernel to try to fix it, but do you have a few pointers about where to look?12:04
mjg59soren: Did any of the 2.6.24 kernels work?12:05
=== \sh is now known as \sh_away
sorenmjg59: It's my wife's laptop which I only just upgraded yesterday.12:05
sorenmjg59: So I've only tried the most recent one, I'm afraid.12:05
mjg59soren: Hm. Can you boot with the kernel argument no_console_suspend and without quiet?12:06
mjg59Then give it a go12:06
sorenSure.12:06
=== \sh_away is now known as \sh
sorenmjg59: Just to be sure: I call pm-suspend to suspend. That's the way to do it, right?12:09
mjg59soren: No.12:09
sorenAh.12:09
mjg59soren: What desktop are you running?12:09
sorenI've killed X.12:10
mjg59Don't do that12:10
sorenTo keep that out of the equation.12:10
sorenHm... Ok. Is it likely to work better with X running? That's surprising.12:10
mjg59I suspect it's unrelated to the problem you're having, but please try from X12:10
sorenWEll, if it matters, I discovered the problem while trying to hibernate from GNOME. Since then, I've used the terminal to make reduce the number of places it could fail.12:11
soren...but ok, I'll try from GNOME.12:11
sorenmjg59: Oh.12:15
sorenmjg59: pm-hibernate oopsed.12:15
soreni915_suspend12:15
mjg59soren: Hm. How about suspend?12:17
sorenmjg59: I'll try.12:18
sorenmjg59: ba8bbcf6ff4650712f64c0ef61139c73898e2165 says: "i915: add suspend/resume support". If it didn't have suspend/resume support before, how is it that this laptop used to suspend just fine?12:19
sorenIt has an i855 graphics adapter, by the way.12:20
mjg59We attempted to restore graphics state from userspace12:20
mjg59Hm. Ok, interesting.12:20
mjg59If it oopses on suspend, that's something I'll have to look at12:20
sorenThe i915 driver *is* supposed to handle i855, I suppose?12:20
mjg59Yes12:21
adincelmargol, i could get it the connect and request a dns with a new kernel but still i get errors12:21
adincelmargol: is it running without any problems at all now?12:22
elmargolno12:22
elmargolIts the same... this was just random (12:22
adinc`so same here, i'm sure this is nothing we can solve12:22
adinconly the developer of this driver can do this12:23
elmargolAnd i was thinking intel wireless cards are well supported :(12:23
adincor someone who is involved in developing this module12:23
tjaaltonsoren: I debugged that issue yesterday (hibernate fails to power down) :)12:23
adincelmargol: yes, my intention was the same actually when looking to this device12:23
sorentjaalton: Any luck?12:23
tjaaltonsoren: yes12:23
sorentjaalton: Do tell.12:23
elmargolI'm wondering if I can change my wireless card12:24
tjaaltonsoren: see bug 19706412:24
ubotuLaunchpad bug 197064 in linux "Hibernate fails on Centrino-laptop" [Medium,Triaged] https://launchpad.net/bugs/19706412:24
tjaaltonit's probably the same12:24
sorenMight be.12:25
sorenmjg59: When suspending, I just get a VC with "[  120.351344] drm_sysfs_suspend" and nothing else.12:25
tjaaltonsuspend works here12:26
sorenmjg59: I can't help but wonder if I mistyped the kernel option..12:27
mjg59soren: That's with no_console_suspend?12:27
sorenmjg59: I think so, but I may have mistyped it. I'll try again.12:27
mjg59Ok12:27
mjg59And make sure you don't have quiet on the command line12:27
sorenSure.12:29
Ngsoren: is this on your x40?12:30
soren/proc/cmdline only lists "root=UUID=blahblahb ro no_console_suspend", so I'll try again.12:31
sorenNg: Nope.12:31
sorenNg: It might do the same, though. I haven't tried.12:31
sorenmjg59: Hm... No, I don't get anything but the drm_sysfs_suspend line. After that, it's quiet.12:31
mjg59soren: Does alt+sysrq+t print anything at that point?12:32
sorenmjg59: Nope.12:32
sorensysrq+o does power it off, though.12:32
sorenalt+sysrq+o, that is.12:33
mjg59Hm.12:33
mjg59I'll see if I can reproduce12:33
sorenSo the kernel's somewhat alive, but I get no VC love.12:33
mjg59My X40 is still around somewhere12:33
sorenmjg59: This is not an X40, though. I realise it's the same graphics card, I'm just saying.12:34
Ngsoren: purely out of interest, does it work if you call acpi-support's sleep.sh?12:34
sorenNg: I doubt it will, but I can try.12:34
NgI get different suspend/resume behaviour between acpi-support and pm-utils. Reading through their scripts last night they seem to do fairly different things12:34
mjg59Ng: What sort of different behaviour?12:34
Ngmjg59: well specifically for me I need to unload e1000 for suspend to work, which acpi-support did because it seems to unload all network modules, where pm-utils doesn't unload any modules by default (if I was reading the scripts properly)12:35
mjg59Ok. Have you got a bug filed about that?12:36
Ngmjg59: I filed one originally about suspend not working, which I've kept updated, bug 201037. It may need to be moved to a different package or so12:37
ubotuLaunchpad bug 201037 in pm-utils "suspend fails on Thinkpad X300" [Undecided,New] https://launchpad.net/bugs/20103712:37
Ngand I guess the description should be more generic12:37
sorenI've never been able to really figure out if more or less generic is preferred in these cases.12:37
sorenI'm leaning towards less.12:38
sorenIt's easier to mark a bug a as a dupe than to split it into two.12:38
NgI can't answer that, but I would guess this affects more than just the x300 and it's a fairly significant behaviour change that's very undiscoverable. the price of unloading/reloading networking modules can't be very high, but it may make some weird problems go away.12:39
Nghaving said that, it just seems like its masking driver bugs12:39
Nganyway, I'll stop hijacking soren's bugtime ;)12:40
sorenFWIW, I think don't-unload-by-default would be a good idea early in the dev cycle and then switch to unload-by-default at around beta time.12:41
NgI'd buy that for a dollar12:42
sorenIt gives us a chance to spot the issues and perhaps fix them early on, and then switch to the safer approach for the release.12:42
mjg59Ng: Ok, I've poked pitti about that. I'll try to hack up a patch later on if I get a chance.12:43
sorenMy not-up-to-date X40 suspends just fine. I'll try getting it current and see if it changes.12:43
mjg59soren: Just pull the kernel12:44
Ngmjg59: nice, thanks12:44
sorenmjg59: Alright.12:46
sorenmjg59: Oh, kernel seems to be up-to-date.12:47
mjg59soren: -12?12:47
sorenuname -a is identical.12:47
mjg59Ok12:47
mjg59Hrmph.12:47
sorenpm-utils is not up-to-date, though.12:47
mjg59/shouldn't/ matter, but yeah, give it a go12:47
sorenNo, the X40 still suspends just fine.12:48
mjg59Bother12:48
mjg59Makes it harder for me to try to reproduce, then :)12:48
sorenQuite.12:48
mjg59soren: On the broken machine, can you try moving i915.ko out of the way so it doesn't get loaded?12:48
sorenmjg59: I could. 12:48
sorenThe bug tjaalton linked to suggested that it'd end up in an infinite loop. I've left the machine alone for a while and the fan is rather quiet, so I have a hunch that's not it.12:52
sorenmjg59: Without i915, it suspends just fine.12:55
mjg59soren: Yeah, the failure we've seen with hibernate shouldn't happen on suspend12:59
sorenmjg59: Oh?12:59
mjg59It's specific to the suspend call being made twice13:00
mjg59Which only happens in hibernate13:00
sorenOh, ok.13:01
soren(hibernate works just fine now, too)13:01
sorenWhy does hibernation cause two calls to the suspend code?13:02
mjg59Because it's somewhat broken13:02
mjg59soren: Hm. Ok. Anyway, on the off-chance that it's actually the same problem - any chance you can grab the oops on hibernation?13:03
mjg59Camera or something would be good13:03
sorenSure.13:04
sorenJust modprobe i915 and recycling X should do the trick, I guess.13:05
=== \sh is now known as \sh_away
smbBenC: Morging Ben. One question: I was looking over the milestone bugs and there is one (bug 176090) assigned to Tim which is about iwlwifi. Does it make sense to add LED support (kernel driver would have that but is different version)?13:07
ubotuLaunchpad bug 176090 in linux-ubuntu-modules-2.6.24 "WiFi / WLAN LED not working on notebooks with Intel iwl4965 | iwl3945" [Medium,Confirmed] https://launchpad.net/bugs/17609013:07
BenCsmb: Is it just a matter of enabling the right config option?13:07
sorenmjg59: Hm... Guess not. It looks suspiciously like a font-less vc.13:08
smbBenC: No, there is more code change involved13:08
BenCsmb: looks to me like enabling MAC80211_LEDS would do the trick...what else is there?13:08
mjg59soren: Heh. Well, if you can get it again, that would be helpful :)13:09
sorenmjg59: I'm on it, I'm on it. I think my phone takes decent enough pictures to be useful. Do you happen to know of a relevant bug to attach it to, or should I just throw them somewhere for you to see.13:09
soren?13:10
smbBenC: If I am not wrong, it requires *-led.c files in both specific drivers and at least one other file that isn't there right now (something with hbus)13:10
smbBenC: if you want a quick look. I just moved the patch that enabled leds to zinc (my home)13:11
mjg59soren: Just file a new one for now13:12
smbBenC: It might be doable. I just wanted to be sure it is agreed because there there certainly is a reason we use the lum driver and not the kernel version13:15
chaosmasterHi again13:17
chaosmasteras mjg59 is here I'm asking again about my suspend issue13:18
chaosmasterI've done some pm-utils and suspend testing over the holidays and found an issue with resuming on my Samsung P35 notebook and radeon 9700 (r300) based card. In contrast to current ubuntu practice radeonfb is *absolutely* needed for working resume13:18
chaosmastershall I file a bug for the kernel or another component?13:19
mjg59There's not really anything we can do about that case13:19
chaosmasterI can't believe I'm the only one with this problem13:20
sorenmjg59: Will do.13:22
sorenmjg59: https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/20709813:27
ubotuLaunchpad bug 207098 in linux "i915 drm driver fails to hibernate with certain i855's" [Undecided,New] 13:27
BenCsmb: Hold off a sec and let me test one thing real quick13:27
smbBenC: sure13:28
BenCchaosmaster: does it work with fglrx?13:33
BenCbrb13:34
=== \sh_away is now known as \sh
=== setmora is now known as eradicus
BenCsmb: Just pushed that patch out for iwlwifi14:49
BenCsmb: it was gruesome getting it to apply, but it's there14:49
chaosmasterBenC: I never used fglrx14:50
smbBenC: Ah, ok. Great15:02
elmargolBenC, I'm using your kernel from git. And still have this http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1570 bug. Any chances to get this fixed for hardy?15:03
ubotubughost.org bug 1570 in data transfer "iwl3945 speed capped at 100 kB/sec" [Normal,Assigned] 15:03
BenCelmargol: first off, using the kernel wont help anything...you need to use latest lum in git15:07
BenCelmargol: second off, I'm using iwl3945, and I don't see any cap...I was just transferring at 1.5M/s15:08
elmargolBenC, sorry I mean I'm using your lum tree15:09
BenCelmargol: Ah, and I am using 11g, so that explains that part15:09
elmargolBenC, works for me too i transfer at 2-3Megabyte/sec after 200M speed drops to 100kb/s15:09
BenCelmargol: well, I transferred 1.2gigs, so...15:10
elmargolI'm using 11g15:10
elmargolBenC, do you use wpa?15:10
BenCelmargol: yes15:11
elmargolaes or psk?15:11
elmargolwpa1 or wpa215:12
BenCwpa tkip15:14
sorenBenC: You work on the kernel, which is one of the most widely available pieces of software on the planet, and AIUI the only living creatures within miles of you are cows... and you're using WPA.15:14
BenCsoren: a) I like to test the latest and greatest (for testing sake) and b) I'm paranoid over my bandwidth :)15:16
BenCnever know what those mad cows will do15:16
sorenYou must be. :)15:16
mjg59soren: 2.6.24-12-generic-12.22, right?15:16
sorenmjg59: Yup.15:16
BenCsoren: but mainly I worry about my financial transactions over-the-air15:17
sorenBenC: Yeah, I figured. That spoils the joke, though.15:17
BenCelmargol: at 400Megs transfer and still holding at 1.4M/sec15:17
* lamont` wonders if vmware 6.0.2 works with the current hardy kernel...15:19
=== lamont` is now known as lamont
BenCelmargol: anyway, if latest doesn't fix it, it's not going to get fixed for release15:19
BenClamont: you'll want the vmmon.tar I have on chinstrap15:19
lamontthanks15:20
lamonthrm... vmmon.tar or vmmon-161/vmmon.tar?15:21
BenCtry one, if you get an error about mismatched API version, try the other15:22
elmargol_BenC, is there a way to set the card to g only mode?15:30
BenCelmargol: My AP is set to G-only15:32
BenCso that has the same affect, I would think15:32
AnAntHello, will bug #201591 be fixed in next kernel upload ? will it use that patch suggested by Garrett ?15:35
ubotuLaunchpad bug 201591 in linux "atyfb regression - screen blank except for blinking cursor after fbcon vtswitch " [Medium,In progress] https://launchpad.net/bugs/20159115:35
elmargol_[  131.230353] wmaster0: TKIP decrypt failed for RX frame from 00:16:b6:19:5e:0a (res=-3) :/15:35
mjg59BenC: Do you have an M1330 with Intel graphics?15:35
BenCmjg59: No, nvidia15:35
mjg59Ah, ok15:36
BenCI think rtg's is Intel15:36
BenCmjg59: might also ask jono, not sure what graphics his has15:36
* BenC is off to lunch15:37
* lamont wonders if he should care that the hardy upgrade is gonna nuke restricted-manager15:42
mjg59soren: Can you do objdump -d /lib/modules/2.6.24-12-generic/kernel/drivers/char/drm/i915.ko and stick the output somewhere?15:43
sorenmjg59: Sure.15:43
Nglamont: I think it's called jockey these days15:44
sorenmjg59: http://people.ubuntu.com/~soren/i915.disassembled.txt15:46
lamontah, of course.15:50
mjg59soren: Did you start X, or was this just from the console?15:53
sorenmjg59: From X as you requested.15:53
mjg59soren: Ok, tanks15:53
mjg59Thanks15:53
mjg59Urgh. h key failure.15:53
sorenmjg59: i915_suspend+0x51d is the I915_READ(I915REG_INT_IDENTITY_R) call isn't it?15:58
mjg59soren: We think it's the FBC reads15:58
sorenmjg59: Hm.. Ok.15:59
mjg59soren: Can you try http://www.codon.org.uk/~mjg59/tmp/915.diff ?15:59
sorenIS_MOBILE is true for which devices?16:00
soren(just curious)16:00
sorenmjg59: I'm on my way out the door right now, I'm afraid. I'll try it when I get back sometime later this evening.16:05
mjg59Should be any of the mobile varients of the chipsets16:05
mjg59soren: Cool, thanks16:05
sorenmjg59: Oh, thank *you* for looking into it. Much appreciated.16:05
nowheremanhi everybody16:17
nowheremanwhy are make-kpkg builds so much bigger than their ubuntu own git counterparts?16:21
nowheremanI've built 2.6.25-rc7 and it's around 233M16:22
nowheremannvm, found out16:32
=== \sh is now known as \sh_away
BenCnowhereman: find /lib/modules/`uname -r` -name \*.ko | sudo xargs strip --strip-debug17:44
nowheremanBenC: yes thanks found that out already :)17:50
=== \sh_away is now known as \sh
=== thegodfather is now known as fabbione
sectechI have an  AMD Athlon(tm) 64 X2 Dual Core Processor 4200+ and athcool doesn't pick up that I have an Athlon... I am running 2.6.24-12-generic... Is there something special that I have to do?20:19
=== smb is now known as smb_away
adincis someone her who is involved into bug 185470? i've used a new kernel dig and iwl3945 makes still problems21:30
ubotuLaunchpad bug 185470 in linux "iwl3945 not functioning : microcode error" [Unknown,In progress] https://launchpad.net/bugs/18547021:31
sohois the new kernel 2.6.24-4 with all it's new features (intel 945 stuff etc.) going into hardy final?21:48
=== \sh is now known as \sh_away
sohoi mean 2.6.24.421:48
tjaaltonduh, I don't see any i945 goodness in 24.421:55
mjg59soren: Had a chance to test that patch? :)22:08
sorenmjg59: I've gotten as far as setting up an i386 chroot to compile it in. Give me half an hour. :)22:08
mjg59Ha22:08
mjg59Col22:08
sorenmjg59: Ok, kernel module built. I just need to get my hands on the laptop and try it.22:33
=== jepler_ is now known as jepler
sorenmjg59: Hm... Different oops.22:58
sorenmjg59: This time at i915_suspend+0x60.22:58
sorenmjg59: Actually..22:58
sorenmjg59: I think that's what it said the first time I saw it, too.22:58
sorenmjg59: I just thought I was on crack, so didn't think much of it when I saw the oops on the photos.22:59
mjg59soren: Hm. Can you try adding similar blocks around some of the other registers?23:23
sorenmjg59: Any particular ones?23:23
sorenmjg59: Or just a random (sub)set?23:23
mjg59Whichever ones you think it might be oopsing on access to :)23:23
sorenmjg59: Hm.. In any case, I should be adding similar blocks around the resume code, I suppose.23:25
sorenbut let's deal with that, when we've found the offending ones.23:25
sorenmjg59: kees and I are looking at the assembler code. We're having trouble making sense of some of it. Well, *I* am, that's for sure.23:27
mjg59soren: Heh23:27
sorenmjg59: A bit before i915_suspend+0x60, there's a "call 1b6", which is one byte into the call instruction.23:27
sorenwtf?23:27
mjg59soren: Hrm. Can you stick the disassembly up somewhere again?23:28
sorenhttp://people.ubuntu.com/~soren/i915.disassembled.txt23:28
mjg59soren: That's the current module, not the old one?23:28
sorenkeescook: Welcome.23:28
sorenmjg59: Old one, but that bit's the same.23:28
keescookwhee23:28
mjg59soren: Sure? Ok23:28
sorenmjg59: Yes.23:28
sorenI can toss up the new one instead, just to be sure.23:29
mjg59soren: Passing it on now23:29
sorenhttp://people.ubuntu.com/~soren/i915-new.disassembled.txt23:29
keescookfun.  i386 compiler appears to have calculated the relative offset wrong.23:30
sorenYeah. Something doesn't smell right.23:30
mjg59soren: Hm. You're definitely starting X, right?23:32
mjg59And triggering the hibernate with X still running?23:32
sorenmjg59: Hibernating from X, yes.23:32
mjg59Ok, cool23:32
mjg59soren: jbarnes is the guy to speak to about this23:32
* mjg59 goes to play Portal for a bit23:33
jbarnessoren: lspci -v somewhere?23:33
keescookthe disassembly is showing a busted compiler -- a -3 relative jump target is nonsense.23:33
sorenkeescook: But if that was really the problem, I wouldn't be getting invalid page requests, but invalid instructions and shit.23:33
sorenjbarnes: Hang on, I'll reboot it.23:33
jbarnessoren: no it probably wouldn't generate bad opcodes, rather bad memory requests23:34
jbarnesand/or might overwrite important stuff23:34
sorenjbarnes: Did mjg59 mention the weird looking disassembly?23:34
jbarnesno, I haven't seen it23:34
keescook     192:   e8 fc ff ff ff          call   193 <i915_suspend+0x33>23:34
keescook     197:   8d 8e e0 00 00 00       lea    0xe0(%esi),%ecx23:34
keescookit's calling into it's own call instruction.23:35
keescookamd64 compile of the same code shows a regular self-call thunk (e8 00 00 00 00)23:35
jbarnesit's a rel32 call23:37
keescookI was about to say -- I can't possibly be right -- the modules are filled with it.23:38
jbarnesthe loader should fix it up23:38
sorenI'll probably figure out what that means when I grow up.23:38
soren:)23:38
jbarnesit's probably calling the palette save routine23:39
keescookooohhhh23:39
keescookit's what the prelink targets look like.23:39
keescooksorry, I haven't disasm'd .o's before.  I'm used to looking at executables23:39
sorenhttp://pastebin.ca/95886623:41
sorenthe lscpi -v (finally)23:41
jbarnesok well not the palette routine (that's befcffff) but something23:41
sleepsterhow do I become a kernel hacker like you guys?23:42
sleepsterfor Ubuntu I mean23:42
sleepsterwould you guys offer any advice on how to learn?  I've read the linux kernel books by O'Reilly23:43
jbarnessleepster: that's a good start23:43
sorensleepster: Then you're way ahead of me :)23:43
jbarneskernelnewbies has some good tips too23:43
jbarnesbeyond that, I'd recommend finding bugs and fixing them23:43
jbarnesand don't get too caught up in the janitorial stuff23:44
sorenjbarnes: I tried  http://www.codon.org.uk/~mjg59/tmp/915.diff, but that seems insufficient.23:44
sleepsteryeah. I tend to do that.  I wrote my own OS from scratch, so I know a lot of concepts.23:44
jbarnessoren: yeah I don't think that's the real issue here23:44
jbarnessleepster: well if you wrote your own OS there's not much I can tell you :)23:45
sleepsterI just see you guys talk about the Ubuntu kernel stuff in here .. and it seems interesting.. I want to hop on the band wagon23:45
sorenjbarnes: The new oops is at i915_suspend+0x60, which makes it just before  I915_READ(PIPEACONF);23:45
jbarnesis that the pci config space read in your case?23:45
sorenpci_read_config_byte(dev->pdev, LBB, &dev_priv->saveLBB);23:46
sorenAFAIUI23:46
keescookah, readelf -r.23:46
sorenIf the oops says it's at i915_suspend+0x60, that's the address immediately *After* whatever failed, right?23:46
sleepstercan you guys recommend maybe.. an easy device driver I could implement to learn?23:46
jbarnessoren: where are the original oopses again?  do you have a complete dmesg from the machine possibly?23:47
sorenjbarnes: The oops happens as I hibernate, so it's unresponsive after it.23:48
jbarnessleepster: you should look for the linux driver project, gregkh is running that, they have all sorts of hw that needs linux drivers23:48
sorenjbarnes: https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/20709823:48
ubotuLaunchpad bug 207098 in linux "i915 drm driver fails to hibernate with certain i855's" [High,Triaged] 23:48
jbarnessoren: you could try echo test > /sys/power/disk23:48
jbarnesthen echo disk > /sys/power/state23:48
jbarnesbut that may not help...23:48
sorenjbarnes: I've seen two different oops'es.23:49
sorenjbarnes: The one in the bug and one at i915_suspend+0x60 (but otherwise identical, AFAICS)23:49
jbarnesok, so the actual oops is from the access of dfcef0a423:50
sleepsterthanks guys for the info23:50
sleepsterit's much appreciated23:50
sorenjbarnes: The what now?23:50
jbarnessoren: I think the cause is the "unable to handle paging request at ..."23:51
jbarnesand 0xdfcef0a4 is the addr23:51
sorenjbarnes: How do I translate that into something useful?23:52
keescooksoren: where is the +0x60 coming from?23:52
jbarnessoren: well, looking at the registers, eax has 0xdfced00023:53
sorenkeescook: The monitor on the laptop behing me.23:53
sorenbehind me, even.23:53
keescookah, the photos from 207098 are 0x51d, then?23:53
jbarnesand 0xdfcef0a4 - 0xdfced000 = 0x20a423:53
jbarneswhich corresponds to 72a:8b 80 a4 20 00 00    mov    0x20a4(%eax),%eax23:54
sorenkeescook: Yes.23:54
jbarneswhich is why I suspected the fbc regs at first, since they're at 0x2xxx23:54
jbarnesbut now I think your whole register window is getting lost somehow23:54
sorenI'll see if I can get the +0x60 oops again and take new photos. Gimme a few minutes.23:55
mjg59jbarnes: Hm. This is without the diff that ensures suspend only gets usefully called once23:55
mjg59(for hibernation)23:55
mjg59You don't unmap anything during suspend, do you?23:55
jbarnesmjg59: we'll disable the device in the suspend path23:56
mjg59jbarnes: Ah. Perhaps that's the issue.23:56
jbarnesbut it could also be X somehow unmapping the regs23:56
sorenjbarnes: I did the "echo disk > /sys/power/state ; echo disk > /sys/power/state" thing and then it hanged.23:56
mjg59soren: test, rater than disk?23:56
sorenmjg59: YEs, sorry.23:56
jbarnesyeah echo test > /sys/power/disk23:56
jbarnesthen echo disk > /sys/power/state23:56
sorenThat's what I did.23:56
jbarnesbut yeah the oops might still cause you to hang23:56
sorenI did it from X. Perhaps that was optimistic.23:56
jbarnessoren: no, that *should* work fine23:57
jbarnesbut there may be a specific interaction with the versions you have on your system23:57
sorenOk.23:57
sorentest > /sys/power/disk does what?23:58
jbarnesit'll make your system hibernate, wait 5 seconds, then resume23:58
sorenIncrese debugging or something?23:58
jbarnesw/o actually powering off23:58
sorenOh.23:58
jbarnesso you can debug things a lot faster23:58
sorenAh... Clever.23:58
jbarnesbut as you discovered, it's only useful in some cases :)23:59
sorenI got the system back up now... What was I about to try?23:59
sorenOh, right. Getting the +0x60 oops again.23:59

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