"radeon: error initializing UVD" in kernel 3.10 on hybrid laptop with CEDAR / R600 [Solved]

[Jörg-Volker Peetz]

Alex Deucher wrote, on 07/03/2013 00:49:
> On Tue, Jul 2, 2013 at 4:34 PM, Jörg-Volker Peetz <jvpeetz at web.de> wrote:
>> Alex Deucher wrote, on 07/02/2013 22:17:
>>> On Tue, Jul 2, 2013 at 4:15 PM, Jörg-Volker Peetz <jvpeetz at web.de> wrote:
>>>> Alex Deucher wrote, on 07/02/2013 21:46:
>>>>> On Tue, Jul 2, 2013 at 10:09 AM, Jörg-Volker Peetz <jvpeetz at web.de> wrote:
>>>>>> With self-compiled linux 3.10 on a HP Pavilion dv7 with hybrid graphics (ATI
>>>>>> RS880M [Mobility Radeon HD 4200 Series] / ATI Manhattan [Mobility Radeon HD 5400
>>>>>> Series]) uvd seems to be broken.
>>>>>>
>>>>>> The new firware files are installed in /lib/firmware/radeon:
>>>>>>
>>>>>> sha1 hashes
>>>>>> 3142a64061ade6032c95ed948c85b15dd0ae46be  CEDAR_me.bin
>>>>>> a92856a4fa16926e2451a6335da7e20f01fde210  CEDAR_pfp.bin
>>>>>> 644b29756636687ad31a49da4aa3ed85dcedecdb  CEDAR_rlc.bin
>>>>>> 992d49518d3936986b5ce3ddb0d9bbd75135bb8f  CYPRESS_uvd.bin
>>>>>> 3e04529600d666ddb2f2f83bb0112d4fab516c04  R600_rlc.bin
>>>>>>
>>>>>> The system boots without initial ram disk.
>>>>>
>>>>> Make sure your system is using the latest CEDAR_rlc.bin as well.
>>>>>
>>>>> Alex
>>>>>
>>>> Thanks for the hint, Alex. Actually I took the files today from your repository
>>>> at http://people.freedesktop.org/~agd5f/radeon_ucode/ and checked them against
>>>> the ones downloaded from
>>>> http://git.kernel.org/cgit/linux/kernel/git/firmware/linux-firmware.git .
>>>
>>> Make sure that your kernel is actually using the new ones.
>>>
>>> Alex
>>>
>>
>> The files are located in /lib/firmware/radeon , the kernel configuration contains
>>
>> CONFIG_EXTRA_FIRMWARE="amd-ucode/microcode_amd.bin radeon/R600_rlc.bin
>> radeon/CEDAR_me.bin radeon/CEDAR_pfp.bin radeon/CEDAR_rlc.bin
>> radeon/CYPRESS_uvd.bin"
>> CONFIG_EXTRA_FIRMWARE_DIR="/lib/firmware"
>>
>> I compiled the kernel with the firmware files already in /lib/firmware/radeon .
>> The kernel boots without initial ram disk.
>>
> 
> I've encountered people having all sorts of problems with stale or
> truncated firmware, so I just wanted to double check.  The best test
> would be to build the driver as a module and blacklist the module,
> then, once the system is booted to a non-X runlevel, manually load the
> module so it loads the ucode directly from the filesystem.
> 
> Alex
> 
Thank you Alex, for insisting. The "Denkfehler" was indeed at my side: At first,
I compiled the kernel with the old firmware. Then I noticed the missing firmware
module "CYPRESS_uvd.bin". After an information trip around the internet, I
downloaded the missing and the up-to-date firmware modules and put them into
place as well as adapted the kernel configuration.
Then I just did a new "make" in the kernel directory.

But, it seems the make rules don't recognize changed firmware modules. In the
end I still saw the described error messages in the dmesg-output.
Today, after reading your e-mail I came to this conclusion and recompiled the
kernel completely, i.e., beginning with a "make clean". And, voilà, everything
now works.

Best regards,
Jörg-Volker.

[P.S.: I'm sorry, if you receive this letter twice, I somehow didn't send it to
the list the first time.]