RAOF | Aaah, plymouth. Source of endless fun! | 02:00 |
---|---|---|
* Sarvatt wonders how long maverick will be broken horribly because of it this time | 02:13 | |
Sarvatt | nows about the time where all the brokenness starts if lucid is any indication :) | 02:13 |
RAOF | Me thinks there is a race condition somewhere between X and plymouth on radeon, where drmGetBusId will return “” if called after a certain point, which obviously breaks X. | 02:17 |
Sarvatt | that happens when X loads the actual kernel module | 02:26 |
Sarvatt | and its fixed by vesafb=sucks | 02:27 |
RAOF | Oh, really? | 02:27 |
RAOF | Well, it doesn't surprise me that killing vesafb works - is that the trigger, though? | 02:27 |
Sarvatt | i think udev sees the video device tagged already because of vesafb and doesn't load radeon before X starts, those errors are the same as when X starts and fails because of the agp crap | 02:28 |
RAOF | The race is that X needs to come up before radeon has got loaded? | 02:28 |
Sarvatt | radeon needs to load before X starts or else the ddx tries to load it with xf86LoadKernelModule() which fails like that | 02:29 |
Sarvatt | its happening on nouveau too | 02:29 |
RAOF | I haven't seen it there on my hardware, but I have seen it on radeon. | 02:30 |
Sarvatt | lemme try to dig up a nouveau one, ricotz and someone else were having it | 02:30 |
Sarvatt | the grub change in mid july started all this | 02:31 |
RAOF | Yeah. | 02:31 |
RAOF | To leave the grub-set vesa mode and hand off to vesafb. | 02:33 |
Sarvatt | dropping gfxpayload=keep or vesafb=sucks fixes it for now at least | 02:34 |
Sarvatt | https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/606244 | 02:34 |
ubot4 | Launchpad bug 606244 in linux (Ubuntu) "X doesn't find a screen and is not starting due a race condition (affects: 3) (heat: 434)" [High,Confirmed] | 02:34 |
RAOF | Yeah, that's the bug I got them to file. | 02:36 |
Sarvatt | i had problems on intel the first few boots after updating grub but its fine now without any of those for some strange reason | 02:55 |
Sarvatt | with FRAMEBUFFER=Y plymouth was always racy though | 02:55 |
Sarvatt | sometimes i'd get a text splash at native resolution | 02:55 |
Sarvatt | i'm guessing when it actually uses vesafb it'll be screwed and i just havent gotten into those situations where it would use a text splash before | 02:56 |
Sarvatt | i got the text splash because vga16fb was blacklisted | 02:56 |
RAOF | vga16fb should no longer be loading at all now. | 02:57 |
RAOF | Now… why is DRI2 killing X on radeon. | 02:57 |
Sarvatt | yeah its not, just saying plymouth would start before inteldrmfb was ready and i'd get a text splash that resized to native res once it was ready sometimes and thats probably whats going to be broken now that its using vesafb | 02:58 |
RAOF | I know the thinking was that vesafb *should* handoff nicely to the drmfb. | 02:58 |
Sarvatt | it does without plymouth | 02:59 |
Sarvatt | plymouth segfaults and things are all screwed up at that point though | 02:59 |
RAOF | Right. | 02:59 |
RAOF | Plymouth also seems to trigger a nice kernel mutex trace. | 02:59 |
Sarvatt | display turns off after intel loads until X starts, if i dont hit escape to switch away from the screwed up plymouth before X starts it just hangs | 03:00 |
RAOF | Hurray for race conditions! | 03:00 |
Sarvatt | vesafb > inteldrmfb handoff > plymouth is fine, vesafb > plymouth > inteldrmfb is screwed | 03:01 |
Dr_Jakob | Sarvatt: so if you get plymouth to load the i915 driver you should be okay? | 03:02 |
Sarvatt | it doesn't like using the framebuffer renderer on a drmfb | 03:02 |
Sarvatt | yeah thats a good idea, will look into that | 03:02 |
RAOF | Do we still have plymouth's drm renderer backend? | 03:04 |
Sarvatt | yeah | 03:04 |
RAOF | What happens if you take that away? | 03:04 |
RAOF | That can't be the nouveau problem, because we don't use it there. | 03:04 |
Sarvatt | it picks which backend to use when it starts though and cant change it | 03:04 |
RAOF | You can't change it by just deleting the .so? :) | 03:05 |
Sarvatt | which would you delete? the framebuffer renderer doesn't work on an inteldrmfb :) | 03:06 |
Sarvatt | getting rid of the drm backend would make seeing if thats whats really happening easier though | 03:07 |
RAOF | As I say, that's clearly not what's happening on nouveau, because we don't use the drm renderer on nouveau. | 03:08 |
Sarvatt | yeah I know, X loading the kernel module doesn't load drm and crap in time for nouveau and radeon, entirely different problem | 03:08 |
RAOF | Hm. That's a new one - segfault in radeon_drv due to dereferencing a random pointer. | 03:12 |
Sarvatt | can you reproduce the nouveau or radeon problem? i'd be interested in knowing what device is tagged PRIMARY_DEVICE_FOR_DISPLAY when it fails | 03:12 |
Sarvatt | is it on LP? | 03:12 |
RAOF | Sarvatt: I can probably reproduce on radeon. | 03:13 |
RAOF | Sarvatt: It might be on LP, but that's not where I'm looking. | 03:13 |
RAOF | Flicking through screensavers will kill X in radeon_dri2_copy_region. | 03:14 |
RAOF | However, it's not a problem in (the new) mesa, so it's time to get that uploaded. | 03:15 |
RAOF | Then we can move to server 1.9 | 03:15 |
Sarvatt | guess i shouldn't have filed that x-x-v-sis sync request if we're about to jump to 1.9 and need to rebuild all the drivers :) | 04:02 |
Sarvatt | i have no clue what this guy was trying to do reversioning ppa-purge to 0+bzr46.1 instead of the 0.2.6 it was at and it's already in the archives.. i need to make some changes and i'm not sure what he wants next | 04:16 |
Sarvatt | guess i'll just do (1) UNRELEASED with all these changes, it needs to be higher than 0.2.6 to even upgrade | 04:18 |
RAOF | Good that it's in the archives, though. | 04:24 |
Sarvatt | RAOF: what gen ati was crashing btw? | 04:31 |
Sarvatt | r300? | 04:32 |
Sarvatt | mesa dri driver gen name I mean | 04:32 |
RAOF | r600 | 04:32 |
RAOF | It's an r700 chip, though. | 04:36 |
RAOF | I'll try it against edgers before poking further. | 04:36 |
RAOF | Whoops! There it goes. | 05:05 |
Sarvatt | crashed with edgers too? | 05:19 |
RAOF | Yup. This time in dixLookupPrivate because fo all the privates changes in 1.9 | 05:20 |
RAOF | Edgers builds from master, right? | 05:20 |
Sarvatt | yeah | 05:22 |
=== Amaranth_ is now known as Amaranth | ||
RAOF | Hm. devPrivates = 0x0. That's unlikely to work. | 06:06 |
Sarvatt | i dont suppose you use fglrx on that machine ever? :) | 06:14 |
* Sarvatt is packaging up 10.7 and doesn't know if it works | 06:14 | |
Sarvatt | eh i'll just do the normal upload and wait for whining on phoronix about it being broken | 06:15 |
RAOF | I can test fglrx if you want. | 06:16 |
Sarvatt | thats weird though, i haven't heard anyone complain about r600 being broken with -ati in edgers, i'm guessing you're getting that right after X starts due to the copyfb stuff? | 06:18 |
Sarvatt | nah thats ok, it should be fine | 06:18 |
Sarvatt | dropped all the patches since its supposed to support 2.6.35 out of the box | 06:18 |
Sarvatt | i'll just do a build test manually | 06:18 |
RAOF | No, not at al. | 06:20 |
RAOF | When quitting a DRI2 client - such as flipping wildly through the scrensavers. | 06:21 |
Sarvatt | ahh ok that narrows it down, i think i saw a bug about that on dri-devel | 06:27 |
RAOF | Hm. I hadn't noticed one. | 06:29 |
Sarvatt | problem is it had a weird title i'm sure and i can't find it :( | 06:33 |
RAOF | Ok, so at least one part of the problem is that dixLookupDrawable is returning Success, but filling &drawable with random data. | 06:33 |
Sarvatt | prahal: the way to reproduce is to run gnome-screensaver-preferences and switch back and forth on c-waes . After 2 or three attempts I get the segffault and the gdb print shows the dri2 buffer pixmap devPrivates null | 06:34 |
RAOF | Ah, where's that? | 06:34 |
Sarvatt | some more talk of it here - http://www.radeonhd.org/?page=archive_display&c=radeon&m=7&y=2010&d=2010-7-06 | 06:35 |
RAOF | Ah, thanks. | 06:39 |
RAOF | Yup, that's what I'm seeing. | 06:39 |
RAOF | Pity there doesn't seem to be a bug reference there. | 06:44 |
Sarvatt | i can't find anything else :( | 06:50 |
Sarvatt | wonder if this helps any - https://bugs.freedesktop.org/attachment.cgi?id=37401 | 06:53 |
Sarvatt | + https://bugs.freedesktop.org/attachment.cgi?id=37414 | 06:54 |
RAOF | Hm. I'm not sure. | 07:01 |
alf__ | Hi all! Is there a document that describes the current state of input handling in X in ubuntu? eg I want to connect a second keyboard using a different layout and I am a bit confused on how I should go about it. | 12:18 |
=== yofel_ is now known as yofel | ||
kangarooo | Sarvatt: ping | 22:18 |
kangarooo | im now reinstalled 10.10 to test if no xorg crashes.. im actually having now different crash.. all screen freezes and network also. mouse still moves. with raw mode i can open tty6 (still not showing in screen) and making ctrl+alt+del restart.. what dbg i should install? i have nVidia Corporation NV34 [GeForce FX 5500] (rev a1) but restricted driver i removed couse that makes screen respond slow.. | 22:25 |
kangarooo | happens when in FF opening YT or http://en.wikipedia.org/wiki/Ubuntu_(operating_system)#cite_note-11 if i open without citenote then it doesnt hang. opening cite note makes screen go to cite note and its marked blue till blue fades out.. | 22:27 |
kangarooo | yt flash video i mean.. but not all videos.. but most of them | 22:27 |
kangarooo | what should i install to debug and post report couse i cant get any crash report when crashes happen. | 22:28 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!