ati driver status?

[Stewart Adam]

On 1/30/09 8:51 AM, Nicolas Chauvet wrote:
> 2009/1/30 Stewart Adam<maillist at diffingo.com>:
>> On 1/29/09 11:43 AM, Neal Becker wrote:
>>> It seems ati drivers are stalled in testing?
>>>
>>> I'm thinking of building a new home media server, and wondering what to
>>> put in
>>> it.  I was thinking amd is now more linux friendly, and would be a good
>>> choice.  The fact that the driver seems stalled in rpmfusion worries me.
>>>
>> It's not stalled, but it won't work without manual user interaction
> Because you have removed the blacklist-radeon scripts !!!
I'm not sure what you mean by scripts... Do you mean the blacklist file + 
udev node config file?

> On hardware where radeon is "working" (or at least ,radeon detects the
> hardware).
> the plymouth boot have made the radeon drm loading within the kernel image.
> For that reason the initrd needs to be regenerated: (without radeon.ko
> and drm.ko in it)
> (using mkinitrd -f /boot/initrd$(uname -r).img $(uname -r)
> That's been said with the 8.12 release, there is a need for a
> workaround for the nodes creation
> /etc/udev/makedev.d/40-fgrlx-dri.nodes:
> dri/card0
> dri/card1
> dri/card2
> dri/card3
All that's been done already on the package in updates-testing, it's been 
available for a while now.

> If we have the files hardcoded we could avoid this very annoying steps
> but it will make the switch fro driver to another more difficult.
> That's not the way we want.
>
> So a more interesting way would be to have another experimental
> package, named fglrx-bootstrap (or else)
> that will bundle theses scripts in an hardcoded way (still they could
> be edited by r-c-d) and have the initrd image regenerated
> accordingly...
> But for now the switch remains messy... (and isn't meant to be fixable easily)
There's no point in duplicating our work - rpmfusion-config-display already 
does all that. I'm going to update the review tomorrow with version 0.3 
which has a bugfix for the blacklist file (I forgot to prefix all module 
names with the "blacklist" keyword) and it also adds --initrd-regen-hack / 
--undo-initrd-regen-hack which will make it extremely easy for users to switch:
#  rpmfusion-config-display --enable
#  rpmfusion-config-display --initrd-regen-hack
and:
#  rpmfusion-config-display --disable
#  rpmfusion-config-display --undo-initrd-regen-hack

>> (regeneration of initrd images) so we're keeping it there for now. I'm going
>> to test 9-1 this weekend and see what's improved (or what hasn't :/)
> I've already uploaded the new ATI 9.1 binary driver, will do some
> investagation for few time right now.
I just finished testing on my machine, here are the results:
* Catalyst 9-1 works much better, both performance and stability-wise.

* fglrx_dri.so is now working with multilib again and both a 32bit and 64bit 
version is required if you want to use i386 programs such as Wine - Seeing 
as such, I've moved it back into -libs.

* udev and blacklist hacks are still required, although vga=0xXYZ can be 
used to get a plymouth boot screen without KMS while the fglrx driver is enabled

I've submitted builds for devel and F-10, I expect that the kmod on devel 
will fail (there were major changes to the ACPI code in 2.6.29 and no 
patches are readily available yet) but F-10 should be working fine as long 
as you're using 2.6.27 or 2.6.28 kernels.

Regards,
Stewart