[Regression, post-2.6.37-rc2] Unable to start KDE with 2.6.37-rc3 on Acer Ferrari One

[Rafael J. Wysocki]

On Wednesday, November 24, 2010, Alex Deucher wrote:
> On Tue, Nov 23, 2010 at 8:45 PM, Rafael J. Wysocki <rjw at sisk.pl> wrote:
> > On Wednesday, November 24, 2010, Alex Deucher wrote:
> >> 2010/11/23 Rafael J. Wysocki <rjw at sisk.pl>:
> >> > On Wednesday, November 24, 2010, Rafael J. Wysocki wrote:
> >> >> Hi,
> >> >>
> >> >> I cannot start KDE with the 2.6.37-rc3 kernel.  The X server and display
> >> >> manager start normally, but when I attempt to log into KDE, it attempts to
> >> >> start, but then the X apparently crashes and this appears in the dmesg output:
> >> >>
> >> >> [   24.342287] radeon 0000:01:05.0: r600_cs_track_validate_cb offset[0] 0 46097920 4325376 too big
> >> >> [   24.342298] radeon 0000:01:05.0: r600_packet3_check:1331 invalid cmd stream 502
> >> >> [   24.342304] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
> >> >> [   24.694106] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
> >> >> [   93.938946] radeon 0000:01:05.0: r600_cs_track_validate_cb offset[0] 0 46097920 4325376 too big
> >> >> [   93.938956] radeon 0000:01:05.0: r600_packet3_check:1331 invalid cmd stream 502
> >> >> [   93.938962] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
> >> >> [   94.278050] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
> >> >> [  114.898043] radeon 0000:01:05.0: r600_cs_track_validate_cb offset[0] 0 46097920 4325376 too big
> >> >> [  114.898053] radeon 0000:01:05.0: r600_packet3_check:1331 invalid cmd stream 502
> >> >> [  114.898058] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
> >> >> [  115.266113] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
> >> >> [  129.722472] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
> >> >> [  142.984314] radeon 0000:01:05.0: r600_cs_track_validate_cb offset[0] 0 46097920 4325376 too big
> >> >> [  142.984324] radeon 0000:01:05.0: r600_packet3_check:1331 invalid cmd stream 502
> >> >> [  142.984330] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
> >> >> [  143.370083] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
> >> >>
> >> >> This doesn't happen with 2.6.37-rc2.
> >> >
> >> > Reverting the following commit:
> >> >
> >> > commit 16790569eddf4d406a16a65f4750f405fa669c52
> >> > Author: Alex Deucher <alexdeucher at gmail.com>
> >> > Date:   Sun Nov 14 20:24:35 2010 -0500
> >> >
> >> >    drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx
> >> >
> >> > fixes the issue for me (so everyone involved is already in the Cc list ;-)).
> >> >
> >>
> >> Are you sure you've got the right commit?
> >
> > Well, as I said, I reverted that particular commit and things started to work.
> >
> >> That shouldn't change the calculation that's failing at all.  I'd guess it's
> >> probably one of these:
> >> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=d33ef52d9db8a36900dd53f2e32db9a521ace259
> >> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=f30df2fad0c901e74ac9a52a488a54c69a373a41
> >
> > I don't know.  Perhaps there are some interactions between the two and the
> > reverted one.  I can try reverting these tomorrow if you want me to.
> 
> Can you try just reverting this patch:
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=d33ef52d9db8a36900dd53f2e32db9a521ace259

That helps too.

> I think the real issue is a bug in the 3D driver (probably a flush at
> the wrong time) that one of these commits exposing.

Well, that's possible, but unfortunately this is a regression in the kernel still.

Thanks,
Rafael