drm/ast regression, when upgrading from v6.3 to v6.6
Jocelyn Falempe
jfalempe at redhat.com
Thu Feb 13 17:25:15 UTC 2025
On 13/02/2025 15:01, Thomas Zimmermann wrote:
> Hi
>
> Am 13.02.25 um 14:33 schrieb Jocelyn Falempe:
>> Hi Thomas and Jammy,
>>
>> I've got a regression report, and it's caused by https://
>> patchwork.freedesktop.org/patch/540015/
>> On this ast2600 server, the firmware says it has a DP output, but
>> physically, there is only one VGA output.
>> Before this commit, "ast->tx_chip_types |= BIT(AST_TX_ASTDP)" means it
>> will have both VGA and DP configured, but after this changed to "ast-
>> >tx_chip_types = AST_TX_ASTDP_BIT" and the VGA connector disappear.
>
> Having multiple outputs (besides the virtual BMC) was never really
> supported. DRM userspace also doesn't really support it. You likely
> remember that we had problems when the BMC was a separate output.
>
>> The EDID of the VGA monitor is not read anymore, and a wrong
>> resolution is setup causing a blank screen.
>>
>> Of course fixing the firmware would be the best solution, it's not
>> always possible.
>>
>> I'm considering different options:
>> - Add a kernel module parameter to override the tx_chip_types,
>> something like "ast.output=vga"
>
> No parameters please. If we give people parameters, they will configure
> the hell out of them and then complain if it doesn't work. And we never
> get rid of parameters, because "there's users".
>
>> - Add a DMI quirk, to force vga on this particular motherboard.
>
> That could work.
>
>>
>> What do you think?
>
> I think I saw another bit in the HW docs that hints towards a VGA
> output, but I cannot find it anymore. Still, I have VGA and DP on my
> test board. There's gotta be a way of switching between them.
>
> Unless something else shows up, let's to the DMI quirk. It should also
> be easy to backport.
Ok, I will try to make something. DMI quirk is less flexible, but the
regression is a bit old, so I hope there won't be much other hardware
affected.
Best regards,
--
Jocelyn
>
> Best regards
> Thomas
>
>>
>> Best regards,
>>
>
More information about the dri-devel
mailing list