Nouveau - Dec 2016 - NVAC "No Signal"

On 10/20/2016 03:58 AM, Pierre Moreau wrote:
> Hello,
> 
> On 07:37 pm - Oct 19 2016, poma wrote:
>> On 19.10.2016 17:03, Karol Herbst wrote:
>>
>>> You don't get why I try to say. We have to actually find out
when to
>>> apply this workaround, not to create some silly
whitelist/blacklist.
>>> It's the last option, we never want to actually use.
>>>
>>
>> Well if you do not say, who can understand!? :)
>> Besides, you can mock with "silly" whitelist/blacklist",
however there is nothing wrong with the method as such, it is used practically
everywhere.
>>
>>> And even if we would have to create such lists, who tells us, that
if
>>> affects every GPU with your device id? Usually quirks are applied
>>> depending on the sub-vendor-id and sub-device-id if actually
required.
>>>
>>> In the end we need something like this: If byte X in table Y is set
in
>>> the vbios or if bits A-B in reg Z in the MMIO space are set to
>>> whatever, we have to apply that workaround.
>>>
>>> In the end we should also wait until Ben replies, because he might
>>> know the exact reasons why this workaround was actually needed.
I'd like to see a mmiotrace of the NVIDIA binary driver on a system
where this WAR breaks things.  I applied it to all the GPUs that NVIDIA
told me required it.

Ben.
>>>
>>
>> If you eager to leave it broken even more than three months that have
already been passed since the original commit ...
>>
>>> We might have a GPU with the same chipset like yours and we might
be
>>> able to verify the issue
>>>
>>
>> Ah, I see.
>> You do not have confidence in my test results, good to know.
> 
> Testing on another GPU with the same chipset does not mean he does not
trust
> your results. For example, my laptop (which also has an NVAC) has been
> triggering the no-signal message on external monitors way before Ben’s
patch
> landed, but only for some adapters. I haven’t tried Ben’s patch yet, nor
yours,
> but I will certainly do it, and see what effect each of them has.
> 
> Regards,
> Pierre
> 
>>
>> Oh! Carol
>>
>>
>> _______________________________________________
>> Nouveau mailing list
>> Nouveau at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/nouveau
>>
>>
>> _______________________________________________
>> Nouveau mailing list
>> Nouveau at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/nouveau
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: OpenPGP digital signature
URL:
<https://lists.freedesktop.org/archives/nouveau/attachments/20161020/9767ad62/attachment.sig>

On 20.10.2016 00:46, Ben Skeggs wrote:
[...]
> I'd like to see a mmiotrace of the NVIDIA binary driver on a system
> where this WAR breaks things.  I applied it to all the GPUs that NVIDIA
> told me required it.
> 
> Ben.
> 
Still broken, more than four months,
tested with mainline 4.9-rc6.

According to Ben Skeggs,
since this is "told" by NVIDIA i.e. by you, or perhaps some other of
your colleagues,
what's more that I don't see progress in this regard,
would you guys from NVIDIA R&D mind help him to finally solve this?

Thanks.


Ref.
https://lists.freedesktop.org/archives/nouveau/2016-October/026242.html
http://goo.gl/Gm4ffO
mmiotrace-nouveau/

I've observed this regression on my NVAC board; a 1920x1080 TV on HDMI 
(single monitor set-up) gets no signal with Fedora kernels from 4.8. 
Trace sent to the mmio dumps mailbox. A brief scan already revealed that 
register 0xe840 is never touched, so it appears that NVIDIA does 
something different.
VBIOS for this board is in the usual place. Commenting out 0xac from the 
workaround seems to solve the problem, as tested on a Fedora 4.9 kernel. 
I hope that helps you get a little further with this issue. Cheers, and 
happy holidays!

Roy.

Op 22-11-16 om 15:04 schreef poma:
> On 20.10.2016 00:46, Ben Skeggs wrote:
> [...]
>> I'd like to see a mmiotrace of the NVIDIA binary driver on a system
>> where this WAR breaks things.  I applied it to all the GPUs that NVIDIA
>> told me required it.
>>
>> Ben.
>>
> Still broken, more than four months,
> tested with mainline 4.9-rc6.
>
> According to Ben Skeggs,
> since this is "told" by NVIDIA i.e. by you, or perhaps some other
of your colleagues,
> what's more that I don't see progress in this regard,
> would you guys from NVIDIA R&D mind help him to finally solve this?
>
> Thanks.
>
>
> Ref.
> https://lists.freedesktop.org/archives/nouveau/2016-October/026242.html
> http://goo.gl/Gm4ffO
> mmiotrace-nouveau/
> _______________________________________________
> Nouveau mailing list
> Nouveau at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://lists.freedesktop.org/archives/nouveau/attachments/20161224/20b6d4e6/attachment.html>