Re: [gentoo-user] Issues with >=x11-drivers/xf86-video-ati-6.14.4: driver issue or hardware issue?

[Andrey Moshbear <andrey.vul@gmail.com>]

On Tue, Jun 5, 2012 at 11:41 AM, Michael Scherer
<a6702894@unet.univie.ac.at> wrote:
> On Fri, 1 Jun 2012 04:11:35 -0400
> Andrey Moshbear <andrey.vul@gmail.com> wrote:
>
>> On Wed, May 23, 2012 at 11:12 AM, Michael Scherer
>> <a6702894@unet.univie.ac.at> wrote:
>> > ----- Original Message ----- From: "Andrey Moshbear"
>> > <andrey.vul@gmail.com> To: "gentoo-user"
>> > <gentoo-user@lists.gentoo.org> Sent: Tuesday, 22 May, 2012 08:02
>> > Subject: [gentoo-user] Issues with
>> > >=x11-drivers/xf86-video-ati-6.14.4: driver issue or hardware issue?
>> >
>> >
>> >
>> >> Lately, I've been having some issues with segfaults when running
>> >> startx and it's been pretty persistent.
>> >>
>> >> Xorg.0.log and emerge --info are available at
>> >> https://gist.github.com/2766926 .
>> >> Kernel config is available at https://gist.github.com/276943 .
>> >>
>> >> I've tried downgrading, but <=x11-drivers/xf86-video-ati-6.14.2
>> >> fails to compile due to incomplete structs.
>> >>
>> >> Is this more a driver or a hardware issue?
>> >>
>> >
>> > first thing thing is your usage of ACCEPT_KEYWORDS="amd64 ~amd64".
>> > with ~amd64 you globally
>> > allow all packages masked for amd64. unless you are a
>> > developer/tester for gentoo you should remove
>> > this keyword, because gentoo usually has good reasons to mask some
>> > packages. if for some reason you
>> > really need a masked package, you can do this easily only for that
>> > package. global unmasking alone might be the reason for half of
>> > your troubles.
>> >
>> > second, it is advisable to use kernel modesetting, which is
>> > obviously not enabled. gentoo has a detailed
>> > howto for this under http://www.gentoo.org/doc/en/xorg-config.xml.
>> > this gives you all necessary details.
>> >
>> > just a quick shot for the moment. your kernel config doesn't under
>> > the link you give, I'd like to see that
>> > too, and maybe /etc/X11/xorg.conf or the contents
>> > of /etc/X11/xorg.conf.d, if you have any of these,
>> >
>> >
>> >
>>
>> Well, I added radeon drm & modesetting and the kernel is crashing
>> right after the microcode is loaded, but without any signs of panic,
>> be it flashing keyboard lights or kernel backtrace with register
>> values.
>>
>> Also, the segfault was caused by the DRI code interpreting a DRI
>> opcode as a pointer, hence 0xa4 or similar in the back trace.

>
> in any case, I'd need some more information:
> the link to your .config is broken, I only get an empty page.
> if the kernel is involved, output of dmesg and rc.log would
> be needed to.

Here's the .config of the kernel which crashed when loading the radeon
microcode: http://servbox.moshbear.net/r9k-3.3.6-pf-.config.txt

On 3.3.6, dmesg did tend to fill up with "uvesafb: mode switch failed
(eax=0x34f, err=0). Trying again with default settings."

Also, I'm unsure how use of userspace modeseting contributes to X's
drm code confusing an opcode for a pointer.

xorg.conf is blank since X's hardware detection is sufficiently
functional to not require a xorg.conf.

>
> by the way, did you anything with your ACCEPT_KEYWORDS?
> if not, the best thing would be set it simply ="" (and
> afterwards --update world, --depclean and revdep-rebuild),
> but that's up to you.

I use ~amd64 because I like having everything be recent. Otherwise, it
tends to feel a bit like a fast version of debian. That, and 99.9%
functionality is enough. I've been able to fix almost all issues that
arise by means of google - this one is simply proving to be tricky.