[arch-general] Kernel 2.6.32 and Radeon KMS

[Tom K]

Tom K wrote:
> Thomas Bächler wrote:
>> Gabriel Morrison Lima Dantas schrieb:
>>> Removing radeon from initramfs and putting it in MODULES section of 
>>> rc.conf
>>> solves the problem.
>>
>> Hm, I hope you are happy this way until we know what's going on ... 
>> the problem is certainly not that the firmware is not put in 
>> initramfs, but something else ... something weird.
>>
> FYI, not-so-happy radeon kms user here:
> http://bbs.archlinux.org/viewtopic.php?pid=667665#p667665
> he'd like it earlier for plymouth.
>
> T.
>
Interim fix available in the forum thread referenced above. In the 
longer term, I think a different approach is required, as this could 
potentially affect any driver that requires firmware and is used in the 
init image. It doesn't make sense to create individual hooks for all of 
them.

The requirement here is that firmware-dependent modules are loaded after 
the udev hook has been run i.e. after udevd starts. A simple solution, 
based on the one in the forum thread, would be a hook called e.g. 
fwmodules, where modules with this requirement are specified. 
Alternatively, maybe it would work as a second array in mkinitcpio.conf 
e.g. MODULES= and FWMODULES=. This would, of course, require a patch for 
mkinitcpio and related files. I might even write it. :)

T.