[06:04] alright. so i'm very lazy and only now upgraded my laptop to karmic [06:05] is 3d rendering now a requirement for rendering? ive got some odd rendering quirks with nouveau [09:00] tseliot, hi, just had another occurance of the corruption and gpu hang.. updated the LP bug [09:01] apw: ah, nice, let me update the upstream report [09:02] there is another GPU dump on there [09:02] * tseliot nods [09:04] done [09:42] apw: on a second thought I think I'll mark 429199 as a duplicate of 429241. After all, all the gpu dumps and all of the details are in the latter [09:43] sure if you want to do that, i'd been keeping the corruption and non-corruption ones apart, but the former are the only ones i've hit since [09:48] apw: also did you manage to reproduce 429191 and get a GPU dump? [09:55] tseliot, not seen another occurance on my 10v yet [09:56] apw: ok, mine was still very stable after 23 suspend/resume cycles... [09:57] it seems much harder to trigger than the isues on the GM45 [09:59] and i use the GM45 based machine more too [10:05] * tseliot has only GM45 chipsets at home [10:05] :-( [10:11] tseliot, then you are in for a world of pain like me :/ [10:12] * tseliot nods [10:51] apw: isn't bug 429241 very similar to what they reported here: https://bugs.edge.launchpad.net/ubuntu/+bug/421736 ? (albeit with different chipsets) [10:51] Launchpad bug 429241 in xserver-xorg-video-intel "[GM45E] i915 graphics corruption and hang" [Undecided,Confirmed] https://launchpad.net/bugs/429241 [10:51] Launchpad bug 421736 in xserver-xorg-video-intel "karmic: Compositing broken on resume from suspend" [Unknown,Confirmed] [10:55] tseliot, there they seem to talk about 'partially transparent things become invisible ' and definatly i've seen that as part of my symtoms at times. [10:56] though they don't seem to talk about it then going on to lokc up solid generally [10:58] apw: it's possible that you were facing two bugs, one of which could be 429241. I think I need to update the upstream bug report [10:59] tseliot, give me a sec, as i think the fix for that was in [10:59] ok [11:01] the fix toted in bug #419264 was the one below: [11:01] drm/i915: Fix CPU-spinning hangs related to fence usage using an LRU [11:01] Launchpad bug 419264 in linux "Uses 100% CPU with latest mesa/libdrm update" [High,Fix released] https://launchpad.net/bugs/419264 [11:01] and that at least is in the karmic kernels i was running [11:03] tseliot, dunno what that means for your upstream report update :) [11:04] apw: talking of G45 hangs, do you happen to know if http://bugs.freedesktop.org/attachment.cgi?id=29294 made it into any trees? [11:04] I seem to have entirely lost the bug that that attachment came from :/ [11:05] it's referenced in https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/424613 at least [11:05] Launchpad bug 424613 in xserver-xorg-video-intel "[i945GM] GPU lockup with intel 2:2.8.1-1ubuntu1" [Undecided,Confirmed] [11:06] Ng, no i do not see that commit anywhere [11:06] I couldn't even begin to comment on its correctness, but it sounds like a good idea ;) [11:07] jbarnes wonders on http://patchwork.kernel.org/patch/45938/ if it's worthy of stable@kernel.org [11:07] apw: sorry, I meant to say that, in addition to the GPU hangs you might be experiencing 421736. Does 419264 have anything to do with it? [11:08] one referenced the other as a dup i thought, hrm perhaps it got unduped [11:08] in which case you are right [11:08] may be there are two issues, that one, and general hanging [11:08] aha got it, it's https://bugs.freedesktop.org/show_bug.cgi?id=22336 [11:08] Freedesktop bug 22336 in Driver/intel "[i965] GPU hang with compiz, active system use" [Critical,New] [11:10] Ng, ok it does appear to be in the drm-intel drm-intel-next branch [11:11] and interestingly its GM45 specific if i am reading the leader correctly [11:12] apw: as regards this patch http://patchwork.kernel.org/patch/45938/ , would it be possible to put it in a PPA for testing (Intel would like to get some feedback about it)? This is for Freedesktop bug 22336 , as Ng said [11:12] Freedesktop bug 22336 in Driver/intel "[i965] GPU hang with compiz, active system use" [Critical,New] http://bugzilla.freedesktop.org/show_bug.cgi?id=22336 [11:12] apw: I have no idea what the process around this is, if upstream like it that much would we naturally cherry pick it? or wait for it to hit a 2.6.31.X? [11:13] well as its has been indicated as likely stable material we would expect to see it come via 31.x when that gets going [11:13] ok :) [11:13] however, sa its gm5 specific and talks about hangs, and i have GM45 with hangs ... i would be temped to pick it up and test it myself to see if it fixes _my_ issues [11:13] hehe [11:13] I too have GM45 and odd random hangs, so I'd be happy to test it too [11:14] but I can't reproduce the hangs in any meaningful way, so the best I could do is "err yeah it hasn't hung today". my suspend seems to be broken too atm, so I'm having to do shutdowns when I go home, which is ruining my long uptime tests of things ;) [11:14] how it could have had this issue for so long and never hit it is a mystery [11:15] Ng, whats your suspend broken symptoms? [11:15] apw: please test it if you can [11:15] tseliot, i'll put together a test kernel against my GM45 curruption and hang bug and see hwo that pans out [11:15] apw: typically my laptop just doesn't suspend at all recently. it worked with 31-9, but around the -10 upgrade I also started testing all of Keybuk's boot stuff, so it's entirely possible it's his fault [11:16] apw: great, thanks a lot [11:16] heheh [11:46] apw: did you have the desktop effects enabled when you experienced the suspend/resume hang with the mini 10v? [11:47] yes always have compiz enabled [11:47] the netbook-remix UI tricked me into thinking that compiz was enabled... [11:47] ok, I'll try with compiz [11:59] apw: can you try what Chris suggested in 429241 when you can, please? [12:02] tseliot, does that imply ... i need that additional file when it next occurs, or is the live data good enough? [12:04] apw: I think you need to get it (together with the GPU dump, etc.) after you reproduce the problem [12:04] sorry for being a pain :-P [12:04] tseliot, these things happen, its a complex problem [12:07] tseliot, i've also asked chris if he can tell if we have a split command in those two, and pointed him to the wrap patch i am about to test [12:09] apw: excellent, thanks [12:18] * tseliot -> lunch [14:34] apw: I sent you an email. It looks like you will have some testers with i965 for the patch [14:40] tseliot, thanks ... [14:40] i also see Albert has an idea what it might be [14:41] apw: Alberto = tseliot ;) [14:41] Alberto == tseliot, Albert == someone else tho. [14:42] apw: who's Albert then? [14:42] * albert23 is Albert [14:42] a good question indeed :) [14:42] hello :) [14:43] albert23, i do indeed have one such render error [14:43] syslog.1:Sep 15 08:14:32 dm kernel: [ 7689.536305] render error detected, EIR: 0x00000010 [14:43] Is that followed by the PBTBL_ER? [14:43] aah, ok ;) [14:43] albert23, no it doesn't seem to be [14:44] doh yes it is [14:44] egrep errror [14:44] hmm happened in the middle of the suspend too [14:44] Sep 15 08:14:32 dm kernel: [ 7689.536305] render error detected, EIR: 0x00000010 [14:44] [...] [14:44] Sep 15 08:14:32 dm kernel: [ 7689.536315] page table error [14:44] Sep 15 08:14:32 dm kernel: [ 7689.536316] PGTBL_ER: 0x00100000 [14:44] Sep 15 08:14:32 dm kernel: [ 7689.536318] [drm:i915_handle_error] *ERROR* EIR stuck: 0x00000010, masking [14:44] Sep 15 08:14:32 dm kernel: [ 7689.552146] PM: suspend devices took 1.648 seconds [14:45] Looks like it [14:45] ok so i'll find and add that patch to my test kerenl [14:46] Might be worth to try the mesa fix then. It points to a freedesktop bug that's pretty similar to our bug [14:46] freedesktop bug 23254 [14:46] Freedesktop bug 23254 in Drivers/DRI/i965 "Compiz doesn't survive suspend/resume cycle" [Major,Resolved: fixed] http://bugzilla.freedesktop.org/show_bug.cgi?id=23254 [14:51] tseliot, ... is an updated mesa with that fix something you can spin for me [14:51] i'll get this kernel made available too [14:52] apw: ok, let me check what's changed in mesa and I'll put the package in a PPA [14:53] sounds gooood to me [14:56] hah, it's just a one line patch [14:57] tseliot, i see it would fit with the patch suggested for the kernel, checking the 'relocation points to within the object' [14:57] apw: let's hope that a combination of the two patches fixes the bug [14:57] at least one [14:59] i am hoping the mesa fix is the whole thing [14:59] that the kernel change just catches it without the mesa fix ... will test that next [14:59] sounds like a plan :-) [15:03] tseliot, though both changes sound like things that should go to stable imo [15:04] apw: I agree. I think we can upload them after the freeze [15:05] if that mesa fix works we might want to see if ti can be uploaded for A6, as its a bit of a mess [15:06] apw: slangasek might disagree ;) [15:06] he might, then he might not [15:06] heh, it's worth trying [15:07] obviously we need to test it first :) [15:10] right. I'm building it now [15:10] tseliot, most excellent [15:28] it built \o/ [15:32] yay [15:33] apw: I'm using this new PPA: https://launchpad.net/~albertomilone/+archive/x-testing [15:34] tseliot, sounds great [15:34] apw: I'll let you know when the upload is complete [15:34] thanks [15:39] apw: finished. It can take a while for it to build though [15:40] yep and to publish, and ... and i need to test without first too to see if this other patch would have found it [15:46] ok [15:48] tseliot, this kernel patch is _well_ unhappy with the default mesa [15:48] 111.176361] [drm:i915_gem_object_pin_and_relocate] *ERROR* Relocation beyond target object bounds: obj ffff88010895da80 target 2 delta 153751552 size 16777216. [15:48] [ 111.176366] [drm:i915_gem_execbuffer] *ERROR* Failed to pin buffers -22 [15:48] [ 111.249110] [drm:i915_gem_object_pin_and_relocate] *ERROR* Relocation beyond target object bounds: obj ffff88010895da80 target 2 delta 153751552 size 16777216. [15:48] [ 111.249115] [drm:i915_gem_execbuffer] *ERROR* Failed to pin buffers -22 [15:48] apw: what is it that fails? [15:48] oh [15:49] tseliot, producing about 5 on login, thats before any apparent problems [15:49] half my screen is missing too :) [15:49] so _if_ that check is right, its well broken :) what does the mesa fix change? [15:50] apw: it fixes the relocation delta [15:50] drm_intel_bo_emit_reloc(brw->wm.surf_bo[unit], [15:50] offsetof(struct brw_surface_state, ss1), [15:50] region_bo, [15:50] - surf.ss1.base_addr, [15:50] + surf.ss1.base_addr - region_bo->offset, [15:50] I915_GEM_DOMAIN_RENDER, [15:50] I915_GEM_DOMAIN_RENDER); [15:50] } [15:50] oh thats pretty seriously off :) [15:51] no wonder you're getting some weird relocation problems [15:51] ;) [15:51] tseliot, would you mind pushing bug 385658 to your x-testing PPA so we can get some testing going on it? [15:51] Launchpad bug 385658 in xorg-server "'nv' is selected when no xorg.conf is present even if it doesn't support the nvidia hardware" [Unknown,Confirmed] https://launchpad.net/bugs/385658 [15:52] superm1: it's on my todo list. I haven't had the time to look at the patch yet [15:52] ok [15:53] it looks like there's a problem in the chroot of the PPA :-/ [15:53] tseliot, ? noooooo i need that .deb :) [15:54] apw: do you use i386? [15:54] if so, I can upload my packages to my webspace [15:54] amd64 on the box in question [15:55] damn damn damn [15:55] :-/ [15:55] apw: do you want to build the source yourself? [15:55] tseliot, can do [15:56] how painful is it deps wise [15:56] https://launchpad.net/~albertomilone/+archive/x-testing/+files/mesa_7.6.0~git20090817.7c422387-0ubuntu4.diff.gz [15:56] you should report that thing too [15:56] https://launchpad.net/~albertomilone/+archive/x-testing/+files/mesa_7.6.0~git20090817.7c422387-0ubuntu4.dsc [15:56] apw: I've just reported that [15:56] https://launchpad.net/~albertomilone/+archive/x-testing/+files/mesa_7.6.0~git20090817.7c422387.orig.tar.gz [15:57] apw: don't you use pbuilder? [15:58] tseliot, pbuilder no [15:58] i build so very few packages i've never needed it [15:58] its on my list to investigate 'when i have a moment' and has been for 4 months [15:58] apw: that would make things easier for you. This way I can build packages for karmic, jaunty, etc. [15:59] i have the chroots etc, so i can build on the right one etc [15:59] apw: to be honest, I use "pbuild" which is a script that makes pbuilder even easier [16:01] * apw adds that tot he list [16:02] * tseliot has a look at superm1 's patch [16:11] tseliot, that didn't build for me [16:11] /usr/include/drm/radeon_cs.h:181: error: 'RADEON_GEM_DOMAIN_VRAM' undeclared (first use in this function) [16:12] apw: how did you build it? [16:12] debuild -b in a chroot [16:14] apw: did you update your chroot and did you make sure that all the dependencies were satisfied? [16:15] i installed the build deps yeah [16:15] * apw updates it more [16:16] weird [16:16] * apw gets grumpy ... this should be easy to test [16:16] tseliot, the range check is looking good [16:16] screen is mush, but very stable [16:16] 10s/r without loss [16:17] apw: nice [16:17] its puking every update of course, mesa is a bad boy [16:17] yes, I know... [16:17] welcome to my world [16:18] as bad as the kernel ... [16:18] heh, right [16:20] and hanging X == the worst of both in the mix [16:21] hehe [16:21] BTW: [16:21] +# Copyright 2007 Red Hat Inc. [16:21] +# This crappy script written by Dave Airlie to avoid hassle of adding [16:21] +# ids in every place. [16:21] LOL [16:22] heh [16:25] tseliot, so what package do i need to install from the output of this build ... i assume not all of it is needed [16:27] apw: I only updated libgl1-mesa-glx and libgl1-mesa-dri [16:28] cool. i'll do the same ... once it finishes building yawn! [16:28] apw: libgl1-mesa-glx libgl1-mesa-dri libglu1-mesa [16:29] tseliot, is there a any docs on what the various layers actually do [16:29] i assume mesa is the hardware abstraction layer for 3d [16:29] apw: also, can you try and see if Option "AccelMethod" "EXA" (in the Device section of xorg.conf) helps with the ATI card? [16:31] isn't exa the default? [16:31] i thought so, but hey, its a heap anyway ... can't make it worse [16:32] jcristau, apw: the log says that xaa is being used (don't ask me why) [16:32] tseliot, it occurs to me that the symptoms on there included the missing borders and stuff [16:33] ignore that thought [16:33] tseliot, ok chaning that and enablinb compiz looks like its hung completly [16:33] apw: as regards mesa, try http://www.mesa3d.org/ and http://dri.freedesktop.org/wiki/ [16:34] apw: ouch, can you add a comment here, please? http://bugs.freedesktop.org/show_bug.cgi?id=23928 [16:34] Freedesktop bug 23928 in Driver/Radeon "ATI graphics corruption with compiz" [Normal,New] [16:35] superm1: I don't see why that shouldn't work. Did you test it? [16:36] tseliot, Yup [16:36] i've got some of this hardware that fails without it [16:37] superm1: ok, let me add a comment (basically my +1) in the bug report [16:37] i'm not really sure more widespread testing is necessary given how straightforward it is, just needed a few pairs of eyes to make sure it looked acceptable (imo) [16:46] superm1: caution is never enough. Not that I see why your patch should fail. A bunch of testers should be more than enough. [16:46] apw: any luck with the build? [16:46] Ok, can you push it to a PPA then? [16:47] tseliot, yep just installing and booting it now [16:47] apw: great [16:47] superm1: would this be ok? https://launchpad.net/~albertomilone/+archive/x-testing [16:47] sure [16:51] tseliot, ok that mesa update gets rid of the kernel bitching about the reloc [16:51] superm1: uploaded. I hope the chroot it's not broken [16:52] apw: and can you see the whole screen now? [16:53] yep normal service is also returned for my screen ... [16:53] :-) [16:54] * apw does some suspend/resume testing [16:55] 5 s/r's and counting [16:57] * tseliot made the dell mini 10v freeze completely (black screen) without the patches after 6 suspend/resume cycles [16:58] * tseliot is downloading the new kernel for testing [16:59] tseliot, _nice_ [16:59] heh [16:59] my machine seems pretty stable so far ... obviously only time will tell but 4 s/r's were typically fatal [17:00] apw: very well, I'll test the fix here too [17:29] superm1: can you upload a package for me, please? (slangasek is ok with this) [17:29] sure [17:32] superm1: http://albertomilone.com/ubuntu/karmic/mesa/toupload.txt Thanks [17:32] tseliot, you'll have to point me at it though.. :) [17:32] Ok. cool [17:33] is that the same one that apw just test built locally? so i can skip a test build right? [17:34] superm1: yes, I only changed a detail in the changelog [17:34] ok thanks [17:35] tseliot, do we know why the PPA builders are sick on this one? [17:35] tseliot, i didn't see the -nv on you x-testing ppa yet, you sure you uploaded it? [17:36] superm1: yes, I did but the PPA is broken :-/ [17:37] but the source didn't even publish? [17:37] apw: I'll bug lamont again [17:37] must be *very* broken :) [17:37] superm1: no, sorry [17:37] is it just your PPA, or the whole PPA system? [17:37] i can just pop it on one of mine instead if 'ts just yours [17:38] superm1: I hope it's just mine [17:39] tseliot, is it something to do with upstart and mountall and things not installing? [17:39] if so its affecting the whole build system ... someone broke it [17:40] awesome [17:40] apw: The following packages have unmet dependencies: sysv-rc: Breaks: initscripts (< 2.86.ds1-63) but 2.86.ds1-61ubuntu16 is to be installed E: Error, pkgProblemResolver::Resolve generated breaks, this may be caused by held packages. [17:40] d'oh! [17:41] yeha thats the overall mess. the archive admins are applying hammers to it [17:42] it sounds like something which has been laying in wait for a while, a dep circle which is only hit if you upload both packages at the same time [17:42] tseliot, hows the 10v, how many s/r did it survive? [17:43] apw: 6 or 7 without problems [17:43] most unexpected [17:43] now my 945 is fixed i am going to test on here too [17:43] (another 10v) [17:43] good :-) [17:44] if i add the kernel only and see the same vomit from Reloc tests i think we could [17:44] call my other bug a dup of the main one [17:44] will do the testing now [17:46] :-) [17:50] 12 suspend/resume cycles and still no problem whatsoever [17:53] tseliot, yay [17:58] apw: ok, I guess that 20 suspend/resume cycles are enough ;) [17:58] I should stop before I break the netbook lid :-P [17:59] heh yeah ... you can wake it with the power button [17:59] right [18:07] apw, superm1: it looks like it will take a while before the PPAs are fixed [18:08] just FYI [18:08] archive is broke too it looks like [18:08] that mesa upload FTBFS all across the board [18:11] :-( [18:16] * tseliot -> dinner === michael__ is now known as michaellarabel