[G45] i915.modeset=1 causes black screen instead of gdm (UXA bug) (KMS)

[Martin Olsson]

If I add "i915.modeset=1" as a kernel boot parameter on a machine with this
config:

00:02.0 VGA compatible controller [0300]: Intel Corporation 4 Series Chipset
Integrated Graphics Controller [8086:2e22] (rev 03)
Linux kingfish 2.6.30-020630rc4-generic #020630rc4 SMP Fri May 1 08:25:59 UTC
2009 x86_64 GNU/Linux
libdrm2 2.4.9-1ubuntu1~xup~1
xserver-xorg-video-intel 2:2.7.0-1ubuntu1~xup~1
libgl1-mesa-dri 7.4-0ubuntu3
libgl1-mesa-glx 7.4-0ubuntu3
(**) intel(0): Using UXA for acceleration

Then basically I never get to gdm, instead I just get a black screen. If I ssh
into the box and gdb into X then I see it stuck indefinitely in an ioctl(). I
added INTEL_DEBUG=batch at the top of the gdm script and then rebooted, same
thing happened so I captured a batch buffer dump for this freeze.

[Martin Olsson (mnemo)]

Created an attachment (id=25396)
dmesg captured using ssh while frozen

[Martin Olsson (mnemo)]

Created an attachment (id=25397)
xorg.conf captured using ssh while frozen (nothing special here, just UXA)

[Martin Olsson (mnemo)]

Created an attachment (id=25398)
xorg.log captured using ssh while frozen

[Martin Olsson (mnemo)]

Created an attachment (id=25399)
xorg.log.old captured using ssh while frozen

[Martin Olsson (mnemo)]

Created an attachment (id=25400)
.xsession-errors captured using ssh while frozen

[Martin Olsson (mnemo)]

Created an attachment (id=25401)
intel GPU registers captured using ssh while frozen

[Martin Olsson (mnemo)]

Created an attachment (id=25402)
versions of various packages on this machine

[Martin Olsson (mnemo)]

Created an attachment (id=25403)
/proc/dri/0/gem_objects captured using ssh while frozen

[Martin Olsson (mnemo)]

FYI, the "xup" in the package version "2:2.7.0-1ubuntu1~xup~1" refers to the Ubuntu X-Updates PPA available here:
https://edge.launchpad.net/~ubuntu-x-swat/+archive/x-updates/
(it's basically just DEBs for upstream stable releases of drivers that some people use for testing etc)

Overall this is an Ubuntu 9.04 system with upgraded drivers and the kernel is a vanilla mainline 2.6.30-rc4 built with Ubuntu's kernel config.

[Martin Olsson (mnemo)]

I found some wiki page that had some hints on how to get KMS to work so I tried this.

# inside the file "/etc/initramfs-tools/modules" I added these lines:
intel_agp
drm
i915 modeset=1

# and then I ran this command:
update-initramfs -k `uname -r` -u

---

The result was that X boots, however it comes up in 1152x864 (4:3) on my widescreen LCD which looks really awkward. When I press CTRL-ALT-F1 I get graphics corruption but X doesn't freeze or so and if I press CTRL-ALT-F7 again the corruption disappears.

Next, I'm gonna see what happens if I use a more recent intel ddx driver.

Anyway, so basically, maybe this bug should be closed as INVALID.

[Jesse Barnes (jbarnes-virtuousgeek)]

Ok, marking NOTABUG until you sort out your config.

[Martin Olsson (mnemo)]

If I add "i915.modeset=1" as a kernel boot parameter on a machine with this
config:

00:02.0 VGA compatible controller [0300]: Intel Corporation 4 Series Chipset
Integrated Graphics Controller [8086:2e22] (rev 03)
Linux kingfish 2.6.30-020630rc4-generic #020630rc4 SMP Fri May 1 08:25:59 UTC
2009 x86_64 GNU/Linux
libdrm2 2.4.9-1ubuntu1~xup~1
xserver-xorg-video-intel 2:2.7.0-1ubuntu1~xup~1
libgl1-mesa-dri 7.4-0ubuntu3
libgl1-mesa-glx 7.4-0ubuntu3
(**) intel(0): Using UXA for acceleration

Then basically I never get to gdm, instead I just get a black screen. If I ssh
into the box and gdb into X then I see it stuck indefinitely in an ioctl(). I
added INTEL_DEBUG=batch at the top of the gdm script and then rebooted, same
thing happened so I captured a batch buffer dump for this freeze.

[Martin Olsson (mnemo)]

batch buffer dump:
https://bugs.freedesktop.org/attachment.cgi?id=25395

dmesg while frozen:
https://bugs.freedesktop.org/attachment.cgi?id=25396

xorg.log while frozen:
https://bugs.freedesktop.org/attachment.cgi?id=25398

[Bryce Harrington (bryce)]

I notice the upstream bug got marked invalid due to inconsistent configuration; should this be closed as well?

Also, fwiw, KMS issues that occur prior to gdm probably should be filed against linux.

[Martin Olsson (mnemo)]

This was unfortunately a config issue. I've updated the Ubuntu Kernel Mode Setting wikipage under "More information" so that it explains why mainline kernels can't be used out of the box to test KMS: https://wiki.ubuntu.com/X/KernelModeSetting#preview