Operating KMS UAPI (Re: RFC: Drm-connector properties managed by another driver / privacy screen support)

[Simon Ser]

On Monday, April 20, 2020 2:22 PM, Pekka Paalanen <ppaalanen at gmail.com> wrote:

> On Mon, 20 Apr 2020 10:15:39 +0000
> Simon Ser contact at emersion.fr wrote:
>
> > On Monday, April 20, 2020 10:27 AM, Pekka Paalanen ppaalanen at gmail.com wrote:
> >
> > > The only "random" KMS state is the properties the userspace KMS
> > > program does not know that are set on start-up. I have been assuming
> > > that as long as you had fbdev active before the KMS program started,
> > > the unknown properties have "harmless" default values. And maybe even at
> > > driver device init if fbdev does not exist?
>
> I meant fbcon, not fbdev above.
>
> > Note, this is not the case when using e.g. a display manager. In the
> > past there have been cases of a display manager setting a hw cursor
> > and launching a compositor not supporting hw cursors. This results in
> > a stuck hw cursor.
>
> Indeed. So the display manager might get sensible defaults, but the
> session compositor might not. Or maybe boot splash uses KMS already, so
> even display manager doesn't get all-defaults state.
>
> It seems we really do need "sane defaults" from the kernel explicitly.
> Writing a userspace tool to save it at boot time before any KMS program
> runs would be awkward.

Agreed.

> > > Btw. I searched for all occurrences of link_status in
> > > https://dri.freedesktop.org/docs/drm/gpu/drm-kms.html and it seems it
> > > only has two possible values, good and bad, and no mention whether it
> > > is writable. Looks like it's writable. There does not seem to be a) an
> > > explanation how exactly it needs to the handled (writing it does
> > > something? what can you write?) or b) any way discern between kernel
> > > and userspace set values like HDCP "Content Protection" has.
> >
> > User-space needs to reset the value to GOOD when recovering from a BAD
> > value.
>
> What if userspace writes BAD?
>
> BAD cannot be default state, so getting default state from somewhere
> would solve this property's restoring as well. Reading back the true
> current value could accidentally return BAD.


Interestingly, setting it to BAD is a no-op (BAD is silently
discarded):

	/* Never downgrade from GOOD to BAD on userspace's request here,
	 * only hw issues can do that.
	 *
	 * For an atomic property the userspace doesn't need to be able
	 * to understand all the properties, but needs to be able to
	 * restore the state it wants on VT switch. So if the userspace
	 * tries to change the link_status from GOOD to BAD, driver
	 * silently rejects it and returns a 0. This prevents userspace
	 * from accidently breaking  the display when it restores the
	 * state.
	 */
	if (state->link_status != DRM_LINK_STATUS_GOOD)
		state->link_status = val;

So restoring the "sane default" would be work, even if the link happens
to be BAD when saving said "sane defaults".

> Just to reiterate for everyone, the important thing here is to figure
> out how userspace is supposed to reset unknown properties to sensible
> defaults. Once we know how that should work, we can review whether new
> properties support or break that.

Yes, that's a good description of the problem.

I see two main solutions here: either the kernel provides the default
values in its property descriptions (e.g. drmModeGetProperty), either
user-space can ask the kernel to reset properties to their default
values.