[PATCH] drm/amd/display: set panel orientation before drm_dev_register

[Hans de Goede]

Hi,

On 8/5/22 19:34, Simon Ser wrote:
> Hi,
> 
> Have you seen [1] and [2]? CC'ing Sean and Hans, it seems like there's
> a disagreement regarding probing early vs. setting the initial value to
> UNKNOWN.

At least for the discussions I've been involved in so far
(see the links in the originak email) it was possible to retreive
the panel orientation before calling drm_dev_register() and IIRC some
changes were merged (I've at least reviewed them) to add a new
callback to the panel drivers to get the orientation from the panel
driver before registering the connector fixing the ordering problem
for those cases.

> If a driver doesn't fetch EDIDs before exposing them to user-space,
> then there is an amount of time where the connectors are exposed as
> CONNECTED but their EDID and mode list isn't properly set. But maybe
> that's what the GETCONNECTOR IOCTLs is supposed to do: probe the
> connector, fetch its EDID, return the proper mode list. So maybe
> drivers shouldn't probe early and should let user-space request probes?
> In which case we could create the panel orientation prop with "Normal",
> and update it accordingly when probing.

I guess that if in some cases it really is not possible to get
the orientation before calling drm_dev_register() then this is
an acceptable solution, as long as the orientation gets set
properly before the first GETCONNECTOR IOCTL finishes then
userspace will never know the difference since it needs to do
the GETCONNECTOR to enumerate the connectors properties anyways.

> At any rate, I've tested v2 on the Deck and it works properly.
> 
> Tested-by: Simon Ser <contact at emersion.fr>

I'm not aware of which patch's v2 you are talking about here, link ?

Regards,

Hans





> 
> Thanks,
> 
> Simon
> 
> [1]: https://lore.kernel.org/dri-devel/CAMavQKJUpYP8jo2JDGMYNBGtbPSSO7z9BAComm5JQoty=HPtJg@mail.gmail.com/
> [2]: https://lore.kernel.org/dri-devel/20220530081910.3947168-1-hsinyi@chromium.org/
>