[00:16] bryce, kms and intel and vbetool: be advised that pm-utils also makes calls to vbetool, which causes some grief. [00:16] bryce, I added a note to #372480 [00:16] * cwillu pokes ubottu with a stick [00:17] https://bugs.edge.launchpad.net/ubuntu/+source/pm-utils/+bug/372480 [00:17] Ubuntu bug 372480 in acpi-support "vbetool should not be called on -intel when KMS is in use" [High,Fix released] [00:22] weird, it should be removing all quirks and adding the nochvt one when it detects KMS is active via /usr/lib/pm-utils/sleep.d/98smart-kernel-video [00:25] not to mention it had a kernel fix to work around it just incase people were still using it, are you on jaunty by any chance? [00:27] because jaunty's pm-utils doesnt have the extra KMS handling is why I ask, sorry [00:30] Sarvatt, no, karmic [00:31] removing the execute perm from vbetool fixed it, so I'm assuming pm-utils is to blame [00:33] I see that file now, wonder why it didn't work [00:39] hmm, the fix was added to 2.6.30-rc6 and should be in ubuntu's 2.6.30-6.7 kernel if you're using that, thats strange [00:40] [ 61.162844] [drm:i915_wait_request] *ERROR* something (likely vbetool) [00:40] disabled interrupts, re-enabling [00:40] yep [00:40] you saw that in the log? [00:40] I'm using a daily out of the mainline ppa, there's another bug that causes a nicely gradual crash after resume [00:40] yep [00:41] under 2.6.30rc7 or our normal kernel, I crash on resume. On the mainline daily, I don't [00:42] jbarnes: ping incase you are interested that vbetool is still causing some problems after the deal from http://bugzilla.freedesktop.org/show_bug.cgi?id=20896 [00:42] Freedesktop bug 20896 in DRM/Intel "[GM965 KMS] X does not draw untill mouse is moved. Probably IRQ problems" [Major,Resolved: notourbug] [00:42] yeah there were some great fixes for intel merged on the 29th that arent in rc7 [00:43] god I hate backing out of my statements [00:43] must have gotten lucky 2-3 times in a row, right after I disabled vbetool [00:44] it's broken again [00:44] mouse cursor moves, but I can't interact with the screen or change vterms [00:44] let me ssh in and check dmesg [00:45] touchpad works, external mouse is dead, even after replug [00:45] no errors in dmesg or xorg.0.log [00:45] machine itself is live [00:45] the classic intel crash :D [00:45] can't interact, change vt's via ctrl-alt-f1 [00:46] Sarvatt, but why would the external mouse not work? [00:47] i have no idea, thats just how all my intel crashes play out, x dying but the machines still alive and the touchpad works [00:54] there have been a *load* of fixes for those types of hangs in the newer libdrm/intel drivers though, might be worth trying newer ones from https://edge.launchpad.net/~xorg-edgers/+archive/ppa if its annoying enough to be unusable but you still want KMS, or trying again in a few days when bryce updates the ones in karmic [00:54] I'm on edgers already [00:54] ah [00:55] * cwillu does that thing where he checks one more thing he forgot about [00:56] hmm, worked that time [00:56] let me suspend 9 more times and then I'll have some confidence in that :p [00:56] nope, failed the next time [00:57] funny, I've actually heard people complaining about suspend/resume hangs alot in the past week since this commit too - https://bugs.freedesktop.org/show_bug.cgi?id=22010 [00:57] Freedesktop bug 22010 in Driver/intel "[945GM] [Bisected] X hang after suspend" [Normal,New] [00:58] the earlier suspend issue I had (the reason why I'm on a daily kernel) is unrelated to X, [01:00] yep, X just hung and mouse pointer still functional when i did a suspend/resume there :D [01:07] * cwillu cheers [01:10] building it now after reverting that commit to see if it makes any difference, i'll put it on edgers if so [01:10] sweet [01:14] well that looks good, first suspend/resume is fine [01:14] * Sarvatt prepares to eat his words [01:15] regurgitated words are so tasty [01:16] 2 3 and 4 are fine \o/ [01:16] gimmegimmegimme! [01:17] already uploaded to edgers but it probably wont get built before the publisher runs at 20 after, just want the deb? [01:17] sure [01:17] :) [01:17] * cwillu beams [01:17] http://sarvatt.com/downloads/xserver-xorg-video-intel_2.7.99.1+git20090601.704771f1-0ubuntu0sarvatt_i386.deb [01:18] no KMS cursor flicker patch in it though :D [01:19] whatever will I do :p [01:20] SHOOT cwilllu, wait a second [01:20] i am using xorg 1.7, built it against that [01:20] and I doubt you are too lol [01:21] heh [01:21] nooo, i386 just finished [01:21] 1 minute after the publisher ran [01:25] 10 suspend/resumes fine here, woohoo [01:25] :) [01:27] i'll hold off updating jaunty until you say it works for you too just incase [01:28] are you uploading another deb, or am I waiting for the edger build? [01:31] edger build, would probably take me the 10 minutes till that gets published to set up pbuilder for it :) it takes less than a minute to compile on my atom cpu if you want to build it yourself though [01:32] nah, I can wait an hour :p [01:32] I get the feeling there's a kernel bisect in my future anyways :p [01:32] also reverting to karmics -intel would tell you if that was the problem too [01:33] well it might have problems in other areas with the same symptom too possibly :D [01:33] I need to run, but I'll let you know when I get back [01:36] oh http://ppa.launchpad.net/xorg-edgers/ubuntu/pool/main/x/xserver-xorg-video-intel/xserver-xorg-video-intel_2.7.99.1+git20090601.704771f1-0ubuntu0sarvatt_i386.deb [01:36] good luck with it [01:37] up already? [01:37] nice [01:43] 25 suspend/resumes, pretty safe to say that fixed it for me [01:43] gotta love these 2 second flicker free suspend and resumes with KMS, thats for sure [03:40] Sarvatt, I know, I squealed the first time I saw the resume work [03:41] I mean, okay, fast reboots are nice [03:41] but a suspend that's so fast that I'm tempted to change "blank screen" to "suspend" is another thing entirely :) [03:41] cant beat 2 seconds though, agreed lol [03:42] talked to 4 people now that reverting that commit fixed suspend/resume for [03:42] I just got back, I'll have another yay or nay for you in a minute [03:42] you have gotta update that drm at least though :D [03:44] if you're still on 2.4.9 from may 2nd like in that bug report [03:45] 2.4.11 [03:45] * Sarvatt crosses fingers for you [04:50] hyperair, check out the mesa i just uploaded to edgers, 30 minutes of running a loop opening/killing firefox and i'm at 211MB gem object bytes with compiz running :) [04:52] that would hit 1GB in 30 minutes before and it was 186MB just enabling compiz after a fresh boot [05:29] Sarvatt: the dri.pc -issue is already fixed in git [05:32] woohoo! [11:01] Sarvatt: my gem memory has hit 900MB [14:52] morning [14:52] hi bryce. feeling better? [14:53] yep, much better [15:36] hi kids [17:15] Sarvatt: do your brightness keys work with KMS on? [17:15] Sarvatt: doesn't work for me [17:16] i built my own devicekit-power and gnome-power-manager because the 2.26 ones in karmic were buggy as hell for me, but yes they do work :D [17:16] i like how gnome-power-manager has powertop built in now too [17:17] yeah i like it too =) [17:17] but i didn't rebuild those two [17:17] Sarvatt: just pushed a fix for the tearing problem [17:18] so gpm only detects my battery level sometimes. [17:18] Sarvatt: wanna package up a fresh xf86-video-intel for edgers? :) [17:18] * cwillu bounces [17:18] thank goodness for the battery charge monitor applet [17:18] nice, sure thing, was just in the middle of updating mesa [17:18] cool [17:18] * hyperair sticks a trampoline underneath cwillu before cwillu lands [17:19] Sarvatt: anyway: 6h of usage and ~702M GEM memory used. this is much better than what it was previously =) [17:20] Sarvatt: but 702M is still excessive. [17:20] is that with the mesa patch? [17:20] i'm at 308MB used after about 14 hours of running an open/close firefox loop that would put me at 1gb in about 30 minutes before [17:21] yeah I put his take3 patch and the other one on edgers last night [17:21] cool [17:21] Sarvatt: what's your pciid? [17:22] are you running the clutter test hyperair? that still leaks to hell for me [17:22] er no i'm not [17:22] i'm lazy to compile it =( [17:22] need a whole lot of dependencies which i'm lazy to track down [17:22] but he even said it didnt fix clutter [17:22] hm [17:22] so basically we can conclude that the mem leak had many sources [17:23] the fact that you're not seeing it any more means that the source that both you and i had was fixed. but there's another source of memory leaks =O [17:23] which i still see [17:23] =( [17:23] 706M [17:27] jbarnes_LHR: uploaded, it'll be published at 40 after [17:27] Sarvatt: cool I'll pull it then [17:32] darn, yeah, suspend/resume is broken unless I revert http://cgit.freedesktop.org/xorg/driver/xf86-video-intel/commit/?id=8e942b70cb9a784b3f1311affd6fc74c4bcf68bb [17:36] Sarvatt: oooh you should file a bug for that [17:37] or raise it on intel-gfx@lists.fdo at least [17:39] offhand though the revert looks ok [17:39] with the possible exception of the actual emit call [17:39] looks like the arg order changed [17:40] * cwillu perks his head up [17:41] Sarvatt, I've still got the hang post resume, although the external mouse will move the cursor once if I'm moving it while resuming [17:41] not sure if this is relevant to that commit revert [17:43] on the other hand, I can suspend dozens of times in a row with kms disabled [17:44] Sarvatt: what's the version number? [17:45] Sarvatt: ah nm, got the git commit [17:47] sorry, stepped away while mesa was building, it was a bug report for the same issue that got me to try it after reverting to see if it fixed it and it did, its https://edge.launchpad.net/%7Exorg-edgers/+archive/ppa/+sourcepub/642605/+listing-archive-extra [17:47] did you want me to apply your cursor flicker patch to it too? [17:47] w00t [17:47] no tearing [17:48] did you mean for me to apply your cursor flicker patch too jbarnes? [17:48] Sarvatt: nah that one isn't as important [17:51] ooh there's a cursor flicker patch! [17:51] O_O [17:51] * hyperair would really love a more substantial mem leak patch, though this will do nicely for now [17:51] until i need to virtualize anything [17:51] Question: When I log out, X crashes with a backlog in Xorg.0.log and shoves me to the terminal. As I'm running the intel driver with KMS enabled, should such a bug be filed against X or xserver-xorg-video-intel? [17:51] the last time i ran virtualbox, i oom-ed within 2 hours. [17:52] backtrace* [17:52] obviously. :3 [17:52] hyperair if you were using edgers you were using the cursor flicker patch for about 2 weeks :D [17:53] Sarvatt: ..no way. [17:53] Sarvatt: i still see a flickering cursor damnit. [17:53] does it go corrupted sometimes too? [17:53] er no [17:53] actually the cursor itself doesn't flicker [17:53] the area around it does [17:53] like one square area [17:53] yeah [17:53] thats what i meant, sorry [17:53] oh, and some of my icons get corrupted [17:54] e.g. my battery charge applet [17:54] the gnome-panel applet [17:54] not gpm [17:56] i'm glad its down to just little things like that compared to being completely able to use uxa with all of the crashes a month ago at this time :D [17:56] *completely unable [17:57] heh yeah [17:57] wait [17:57] i've got oen moer issue [17:57] er [17:57] resuming from hibernation, basically [17:57] it hangs upon resuming from hibernation [17:57] it doesn't appear to happen for suspend [17:57] but hibernation, yes. [17:58] so i've been walking around with my notebook draining power in suspend to ram state [18:05] hmm at 741mb gem objects now, looks like those mesa patches fixed it for opening and closing firefox but its still leaking during actual use (albiet slower) [18:07] but its no increasing no matter what i do now, something just made it jump 300mb in the past hour [18:10] hmm [18:11] interesting eh [18:34] Sarvatt: I see there is a 0602 snapshot of -intel in edgers now; how much testing has that one gotten? Is it sane for me to pull that into karmic, or would an older snapshot be safer? [18:37] it's basically the same as debian-experimental and the one thats been in edgers for a week with a fix for video tearing jbarnes just pushed a few hours ago [18:37] ok, I'll pull it [20:33] heyo tormod, they reverted that gamma commit you had a problem with in xserver master. uploading it now since i wanted to test out changes to the hooks anyway :) [20:34] not that it'll build anytime soon with the firefox build bot of doom running still [20:46] * cwillu points out the existence of bug #382884 to Sarvatt :p [20:46] Launchpad bug 382884 in xorg "[KMS] [GM945]: Xorg hang after resume from suspend" [Undecided,Confirmed] https://launchpad.net/bugs/382884 [21:36] that's odd