Re: [git pull] drm v2.6.31 merge (part 1)

["Ozan Çağlayan" <ozan@pardus.org.tr>]

Dave Airlie wrote:
> You don't have an r600, an rs690 isn't an r600 at all.
>
> Can you bisect to where the lockups start or have you always had lockups?
>
> Dave.
>   

Yes I noticed that r600 is a different one, sorry.

First of all I'm maintaining the kernel package for a linux distro[0]
which is currently at 2.6.30. It was plain vanilla in terms of drm/
until I noticed a strange problem on X. X was rendering some pixmaps (In
firefox or other apps) completely black. They were rendered correctly
just after zooming to them 4-6 times.

Then we decided to grab F11 drm patches (for nouveau support and also
radeon patches). The bad rendering problem is disappeared but I started
to have random lockups causing X server to hog the cpu.

Today I reverted the fedora patches and started using vanilla drm/
again, I didn't have a lockup during the day but the rendering problem
is back.

Then I googled a little bit and enabled the debug parameter of the drm
module after the lockup which logs the following 6 lines infinitely:

[ 4653.077514] [drm:radeon_do_wait_for_fifo] wait for fifo failed status
: 0x9003C100 0x00080100
[ 4653.077516] [drm:drm_ioctl] ret = fffffff0
[ 4653.077557] [drm:drm_ioctl] pid=1066, cmd=0x6444, nr=0x44, dev
0xe200, auth=1
[ 4653.077561] [drm:radeon_cp_idle]
[ 4653.077563] [drm:radeon_do_cp_idle]

straceing the X server says infinitely that the ioctl (0x6444) on
/dev/dri/card0 returns -EBUSY.



Thanks,
Ozan Caglayan

[0]: http://www.pardus.org.tr/eng