[mythtv-users] 0.22 on a Via SP13000 anyone?

[Robin Gilks]

>> Robin Gilks wrote:
>>> I thought I'd dig out my old Via Epia SP13000 mini-ITX board as a spare
>>> frontend for guests etc. I'm seeing the following error in Xorg.0.log
>>> (EE) CHROME(0): [XvMC] ViaXvMCCreateSurface: Too many surfaces !
>>> which makes me think that its a 0.22 specific problem as I'm running
>>> the
>>> 904 openchrome driver, the same as on 0.21 but I have updated the
>>> kernel
>>> from 2.6.25 to 2.6.30 (the versions in between seem to have DRI driver
>>> problems with openchrome). One other oddity - I can't seem to get it to
>>> run realtime.
>>
>> Something to do with the size of the surfaces. I don't remember the
>> details
>> any more, but a quick troll of the archive retrieved something I wrote
>> earlier this year.
>>
>> >> If you are compiling fixes, as distinct to just installing an rpm,
>> you
>> >> could also try re-compiling the source after changing
>> >>
>> >> mythtv/libs/libmythtv/util-xvmc.h:56: max_num_xvmc_surfaces(16),
>> >> to:
>> >> mythtv/libs/libmythtv/util-xvmc.h:56: max_num_xvmc_surfaces(14),
>>
>> This lets xvmc create its surfaces.
>> You also need to, or course, have the libViaXvMC file listed in
>> XvMCConfig,
>> to have explicitly listed all the extra modules in your xorg.conf, and
>> ensured that DRI is set to mode 666.
>>
>> Since you are resurrecting the box, you probably have these already set.
>> The surfaces issue is quite obscure.
>
> Please note that mythtv allocates xvmc surfaces until it can't anymore.
> Trying to allocate the first one that fails gives this message, but other
> than
> that it is quite harmless.
>
> This set me on the wrong foot trying to diagnose problems with an eipa
> m1000,
> which runs quite smoothly with 9.10, by the way.
>
> Ronald

I have a workaround - I've downgraded from a 2-6-30 kernel to my last
known good one of 2.6.25 (and downgraded udev to match) and its OK now.
Looks like the DRM problem that arrived in 2.6.25+ hasn't been patched in
the Gentoo patchset for 2.6.30 :-(

I've not found a definitive answer yet on exactly what the problem is but
I guess its a cleanup for other chipsets and nobody from Via has bothered
to put in the fixes for their systems.

Only difference I can find is that I have a agpgart module in 2.6.25 but
not for 2.6.30

-- 
Robin Gilks