Kernel Bug Tracker – Bug 12359
WARNING and other intel modesetting problems
Last modified: 2010-07-24 09:29:38 UTC
Distribution: Ubuntu 8.10
Hardware Environment: Acer Extensa 5220
Software Environment: Xorg 7.4~5ubuntu3
Intel modesetting works fine on the console (as long as vesafb is not used).
[ 51.675136] allocated 1280x1024 fb: 0x007df000, bo ffff88007f1fd600
[ 51.681325] Console: switching to colour frame buffer device 160x50
[ 51.742152] [drm] DAC-6: set mode 1280x1024 24
[ 51.762350] [drm:i915_get_vblank_counter] *ERROR* trying to get vblank count for disabled pipe 1
[ 51.802988] [drm] LVDS-8: set mode 1280x800 20
[ 51.806349] fb0: inteldrmfb frame buffer device
[ 51.806351] registered panic notifier
[ 51.806356] [drm] Initialized i915 1.6.0 20080730 on minor 0
But I get a
WARNING: at drivers/gpu/drm/drm_crtc.c:213 drm_mode_object_get+0x49/0xaf()
And I can't start X with it activated. I use intel xorg module v. 2.5.0.
[ 155.265134] Xorg:8838 map pfn expected mapping type uncached-minus for d0000000-e0000000, got write-combining
Fatal server error:
AddScreen/ScreenInit failed for driver 0
I will attach log files.
Created attachment 19643 [details]
This seems related btw. http://www.gossamer-threads.com/lists/linux/kernel/917056
Created attachment 19644 [details]
Booting with nopat let me start X, but the screens were partly garbled. I could move the mouse, but when I tried to exit (Ctrl-Alt-F1 or Backspace), the display hung up completely filled with noise.
Created attachment 19645 [details]
btw. intel module version 126.96.36.199 makes no difference. xorg.conf attached.
Maybe try to build intel module with --enable-kms?
Oh. that might well be it. intel driver 2.5.99 doesn't compile with --enable-kms though... 2.5.0 doesn't have the option. I'll try 2.5.1.
You first have to figure out how to build 2.5.99 with enable-kms (mainly, I needed to update x11proto-core-dev to 7.0.14.
I tried 2.5.1 with --enable-kms and it was worse than 2.5.99 without. My Ubuntu doesn't have that package version available(only 7.0.12). The compile error of intel 2.5.99 --enable-kms is:
drmmode_display.c:149: error: 'struct _xf86Crtc' has no member named 'transformPresent'
drmmode_display.c:170: error: too few arguments to function 'xf86CrtcRotate'
I'll try again sometime in a week.
> And I can't start X with it activated. I use intel xorg module v. 2.5.0.
Choose this option if you want kernel modesetting enabled by default,
and you have a new enough userspace to support this. Running old
userspaces with this enabled will cause pain.
You can't run non-kms aware X with kms enabled. This should be marked invalid.
Well at least the kernel warning on loading the i915 module is valid, right? And the other part is yet to be confirmed.
btw. intel v. 2.5.1 was compiled with --enable-kms and still didn't work.(see #8)
I'm not really sure 2.5.1 even supports KMS?
2.5.1's got a configure option --enable-kms. I assumed it's there for a reason. But I haven't found any compatibility lists anywhere.
If it doesn't build by default, it's probably not meant for real use.
(No, don't use old kms code, please)
If you are talking about the --enable-kms - it doesn't build by default even in 2.5.99(which doesn't build even with jaunty's xorg packages). So there's just no released "new" code for it yet?
As the new drivers don't build with any current ubuntu xorg (and jaunty's intel drivers don't work either), I won't be able to test them until ubuntu upgrades it xorg packages. So for now there's just the warning I guess.
Somehow the warning now comes along with several seconds of waiting (and blocking the modprobe util) until the resolutions are set.
(In reply to comment #15)
> If you are talking about the --enable-kms - it doesn't build by default even in
> 2.5.99(which doesn't build even with jaunty's xorg packages). So there's just
> no released "new" code for it yet?
They *do* build. I had problem with build-deps (as I said, with x11proto-core for example), but I eventually managed to do that. I don't know jaunty packages, but basically you'll need a lot of .99.x -dev package.
Ah. Silly me. I forgot some packages: not only xserver-xorg but also x11proto-* xserver-xorg-dev. I will test it again now.
Okay, now I've tried 2.5.99, too. (Thanks, Yves-Alexis!) Didn't work either. It hung in the console with a blinking cursor (mirrored on both screens). I could Sysrq-REISU but not B. (It hung up completely during the emergency reset. The cursor even stopped blinking. ;) )
Maybe we can now start to find out where the problem lies?
Ah, great. Now I can't remove the NEEDINFO. Someone please set it back to NEW.
You'll probably want to use the git version of xf86-video-intel along with setting the "AccelMethod" option to "UXA" in your xorg.conf. Eric's kernel tree has a few fixes as well, git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel (drm-intel-next branch).
Created attachment 19735 [details]
dualhead xorg log w 188.8.131.52
I've tried again with the just-released 184.108.40.206. And I've gotten it to work for a while. I used UXA as you suggested. I could start xterm and switch back and forth between console and X. The external monitor has the wrong resolution and the display was corrupted. The notebooks internal monitor seemed fine. The background was pitch black. mplayer worked, I think with the x11 driver.
Then I tried starting konsole. Everything hung when I tried to "ls". I'll attach fresh xorg and dmesg logs:
FatalError re-entered, aborting
Failure to wait for IRQ: Device or resource busy
[ 683.442040] [drm:i915_wait_irq] *ERROR* EBUSY -- rec: 7 emitted: 8
[ 686.442040] [drm:i915_wait_irq] *ERROR* EBUSY -- rec: 7 emitted: 9
I still couldn't start a single-head setup.
Created attachment 19736 [details]
dualhead dmesg w. 220.127.116.11
Does this still exist in kernel 2.6.31 + xf86-video-intel 2.8.x? Lots of KMS fixes have been added in the past 8 months.
Yeah, well I currently have no dual-head setup. But the problems in here do not exist for me with my single head setup. On the other hand I have no console at all right now when I boot, only X - but that's another bug...
Timeout, appear to be very old userspace bugs.