Survey on high priority Radeons bugs -> 2009-10-29

[oibaf]

Ubuntu 9.10 is scheduled on 2009-10-29 (https://wiki.ubuntu.com/KarmicReleaseSchedule) and currently is shipping kernel 2.6.31 without radeon KMS support (http://packages.ubuntu.com/it/source/karmic/linux), a git snapshot of -ati driver 6.12.99 (http://packages.ubuntu.com/source/karmic/xserver-xorg-video-ati) and a git snapshot of mesa 7.6 (http://packages.ubuntu.com/it/source/karmic/mesa).

What are your high priority bugs you would like to see fixed as soon as possible (and in final release of Ubuntu 9.10 if you are an Ubuntu user)? This includes 2D and 3D, but excludes experimental, disabled by default, features like KMS/DRI2/Power savings/R600+ 3D and so on. I hope that doing this little survey could help developers in knowing which bugs are the most annoying for end users and maybe focus on them.

Actually I am quite satisfied with both 2D and 3D on my RV530, the only bugs I'd like to see fixed are:

r300 crashes inside radeonTexSubImage2D (https://bugs.freedesktop.org/show_bug.cgi?id=22528)
[R500] doom3 lockups when starting a new game (https://bugs.freedesktop.org/show_bug.cgi?id=23710)

[pvautrin]

Actually I am facing a serious "address space collision" bug that prevents me from making a fresh install of karmic (alpha5) and even jaunty. The screen stays black and only the cursor is visible.
From what I can tell, it's due to the 64M AGP aperture not handled well, and may be even a regression since it was ok with hardy.
I'd appreciate if the devs could check out the history and logs I put in the bug report linked below and at least reply if it's a radeon bug or a kernel bug

https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/368187

The bug has been open for some months and I hope it can be solved before 9.10 final.

Thanks again for the great work!

[nhaehnle]

Thanks for the TexSubImage crash reminder. This should really be investigated - I just keep running into issues when trying to reproduce it :/

[adamk]

http://bugs.freedesktop.org/show_bug.cgi?id=23901

:-)

[oibaf]

Thanks for the TexSubImage crash reminder. This should really be investigated - I just keep running into issues when trying to reproduce it :/

Let me know if I can be of help reproducing it (I am the original bug reporter) :).

[BlueJayofEvil]

The screen stays black and only the cursor is visible.

I have the same problem on Mandriva 2010 test releases. I got around it by going to a terminal (Ctrl+Alt+F1), becoming root then typing killall X. X instantly rebooted and everything was fine.
A few people on the Mandriva forums had this or similar issues.

In case it matters I'm running a Radeon X1950 Pro (RV570) PCI-E card.

[bugmenot]

http://www.phoronix.com/forums/showthread.php?t=19179

[Ant P.]

I'm hoping for R700 + working KMS power management soon. Not worried about what happens to Ubuntu 9.10 since I run git versions anyway.

I had to reinstall a bunch of stuff yesterday because of a library messup that left most of my X apps unusable, so I'd left my PC running overnight on VT1. Came back, started X and xfce-sensors was showing me a red bar of death for the case temp :D

[bridgman]

I've been having trouble getting 6xx/7xx 3D running on Karmic, even with fresh radeon/mesa/drm. Does anyone have 6xx/7xx non-KMS 3D running on current Karmic ? Current thinking is that maybe the drm code is built into the kernel which prevents a new drm from working without a kernel rebuild... trying to confirm that now.

[DanL]

I've been having trouble getting 6xx/7xx 3D running on Karmic, even with fresh radeon/mesa/drm. Does anyone have 6xx/7xx non-KMS 3D running on current Karmic ? Current thinking is that maybe the drm code is built into the kernel which prevents a new drm from working without a kernel rebuild... trying to confirm that now.

Yeah, I have it running. See: http://ubuntuforums.org/showthread.php?t=1257453

[bridgman]

Ah yes, looks like the Intel KMS drm is still being loaded automatically even if you're running with non-Intel graphics. Blacklisting the Intel drivers and updating the initramfs then rebooting allowed the drm to start properly and then everything worked. Thanks !

EDIT - just remembered something -- I noticed you mentioned that Google Earth wasn't working well -- it seems to run fine if you run it from the command line with vblank_mode=0.

[d2kx]

For many people includign me, with R6xx+, Ubuntu 9.10 doesn't work at all at the moment.

https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/419126

[agd5f]

For many people includign me, with R6xx+, Ubuntu 9.10 doesn't work at all at the moment.

https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/419126

They shouldn't be shipping the r600 3D driver without the proper drm bits.

[Kano]

You mean mesa?

[bridgman]

Mesa built with "r600" included in the --with-dri-drivers= option string...

... might be easier/better to patch in the r600 drm bits.

[Kano]

Does that maybe mean that U devs do not use R600 hardware to test ;)

[bridgman]

Or that they're all secretly running Kanotix and only pretending to install Ubuntu. I don't know.

The most likely explanation is that the release is still in development. The impact of always loading the Intel KMS drm seems to go a lot further than 6xx/7xx hardware:

https://bugs.launchpad.net/ubuntu/+source/fglrx-installer/+bug/392039

https://bugs.launchpad.net/ubuntu/+source/nouveau-kernel-source/+bug/399702

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/404496

[DanL]

EDIT - just remembered something -- I noticed you mentioned that Google Earth wasn't working well -- it seems to run fine if you run it from the command line with vblank_mode=0.

I just copied that part from the X.org wiki page, but I'll add that tip (thanks). As for Google Earth, it effectively locks up my system because it's a 32-bit program. I need to build a 32-bit mesa, but I haven't got around to it.

... might be easier/better to patch in the r600 drm bits.
That would be really nice. They patched in the kernel bits needed for EXA/Xv in the last build, but I have a feeling that they'll stick with pure software Mesa in Karmic.

They shouldn't be shipping the r600 3D driver without the proper drm bits.
Yeah, it looks like they're patching mesa to not include it at the moment.

Does that maybe mean that U devs do not use R600 hardware to test..or test that hardware with only fglrx?

[monraaf]

Ah yes, looks like the Intel KMS drm is still being loaded automatically even if you're running with non-Intel graphics.

Blacklisting the Intel drivers and updating the initramfs then rebooting allowed the drm to start properly and then everything worked.

This sure sounds like a high priority bug to me, why on earth are they loading Intel graphics drivers when there is no Intel GPU present?

I'm planning to switch to Karmic in about 2 weeks time, I sure hope this is fixed by then.

[Kano]

Because they use absolutely stupid code - did not install U nor did i test kms on intel, but the code which loads em is really easy to find even when you just test the live image:

https://bugs.launchpad.net/ubuntu/+source/initramfs-tools/+bug/392039

Look at my comments.

[nanonyme]

EDIT - just remembered something -- I noticed you mentioned that Google Earth wasn't working well -- it seems to run fine if you run it from the command line with vblank_mode=0.Intriguing. Maybe it works for my rv670 without any params because the vblank sync requires interrupts whose documentation is yet undelivered... ;)

[bridgman]

Actually right now the documentation is unwritten. You can't call it undelivered yet :D

[nanonyme]

Actually right now the documentation is unwritten. You can't call it undelivered yet :DThat's a matter of perception, I guess. I could say a pizza delivery company hasn't yet delivered me a pizza regardless of whether it's still in the oven or not, couldn't I? (I wonder what it is with me and pizzas, I seem to raise them at every occasions; maybe I'm just hungry)

[d2kx]

Alberto Milone (Canonical), hours ago:

We are still looking into the possibility to get a working 3D driver for
r600 in time for Karmic. I can't promise anything though.

Would be pretty cool if you ask me.

[bridgman]

That's a matter of perception, I guess. I could say a pizza delivery company hasn't yet delivered me a pizza regardless of whether it's still in the oven or not, couldn't I? (I wonder what it is with me and pizzas, I seem to raise them at every occasions; maybe I'm just hungry)

I think a better analogy here would be that the pizza company has not delivered a pizza which you have not yet ordered, which is one of those philosophical concepts best discussed over pizza at 3am. Since I live out in the sticks I make my own pizza, thereby avoiding the potential deadlock situation.

Anyways, I don't think interrupts are gating the initial release of 6xx/7xx 3D.

[nanonyme]

I think a better analogy here would be that the pizza company has not delivered a pizza which you have not yet ordered, which is one of those philosophical concepts best discussed over pizza at 3am. Since I live out in the sticks I make my own pizza, thereby avoiding the potential deadlock situation.

Anyways, I don't think interrupts are gating the initial release of 6xx/7xx 3D.I suspect a lot of beer would also help the dicussion. And yeah, I don't claim interrupts mean that much right now. Mostly will probably reduce CPU usage (due to polling not happening anymore) and we getting working vsync with KMS.