[RFC v2 0/1] drm/connector: Add support for privacy-screen properties

[Pekka Paalanen]

On Tue, 12 May 2020 18:09:15 +0200
Hans de Goede <hdegoede at redhat.com> wrote:

> Hi,
> 
> On 5/12/20 4:20 PM, Pekka Paalanen wrote:
> > On Tue, 12 May 2020 10:18:31 +0200
> > Hans de Goede <hdegoede at redhat.com> wrote:
> >   
> >> Hi,
> >>
> >> On 5/11/20 9:55 PM, Rajat Jain wrote:  
> >>> Hi Hans,
> >>>
> >>> On Mon, May 11, 2020 at 10:47 AM Hans de Goede <hdegoede at redhat.com <mailto:hdegoede at redhat.com>> wrote:
> >>>
> >>>      Hi All,
> >>>
> >>>      This RFC takes Rajat's earlier patch for adding privacy-screen properties
> >>>      infra to drm_connector.c and then adds the results of the discussion from
> >>>      the "RFC: Drm-connector properties managed by another driver / privacy
> >>>      screen support" mail thread on top, hence the v2.
> >>>
> >>>
> >>> Thank you so much for doing this. I was following the said discussion and eventually it became quite complex for me to understand and follow :-)  
> >>
> >> I hope the new doc text makes things clear again?
> >>
> >>  
> >>>      The most important thing here is big kernel-doc comment which gets added in
> >>>      the first patch-chunk modifying drm_connector.c, this summarizes, or at
> >>>      least tries to summarize, the conclusions of our previous discussion on
> >>>      the userspace API and lays down the ground rules for how the 2 new
> >>>      "privacy-screen sw-state" and  "privacy-screen hw-state" properties are
> >>>      to be used both from the driver side as well as from the userspace side.
> >>>
> >>>      Other then that this modifies Rajat's patch to add 2 properties instead
> >>>      of one, without much other changes.
> >>>
> >>>      Rajat, perhaps you can do a new version of your patch-set integration /
> >>>      using this version of the properties and then if everyone is ok with
> >>>      the proposed userspace API Jani can hopefully merge the whole set
> >>>      through the i915 tree sometime during the 5.9 cycle.
> >>>
> >>>
> >>> SGTM. I have actually moved to working on something else now, so I will most likely wait for this patch to get merged, before rebasing my other / remaining patches on top of that.  
> >>
> >> We have the rule that code like this will not be merged until it has at least
> >> one in kernel user. I plan to eventually use this too, but that is still
> >> a while away as I first need to write a lcdshadow subsystem which the
> >> thinkpad_acpi code can then use to register a lcdshadow device; and when
> >> that all is in place, then I can hook it up on the drm code.  
> > 
> > Hi,
> > 
> > I believe this falls under "new UAPI" rules, because this is adding new
> > KMS properties. Hence an in-kernel user is not enough:
> > 
> > https://www.kernel.org/doc/html/latest/gpu/drm-uapi.html#open-source-userspace-requirements  
> 
> Hmm, I believe that that mostly applies to new DRI (ioclt) interfaces for
> submitting rendering commands to new GPUs and other complex new APIs and
> not necessarily to introducing new properties.    Also note that all
> properties are exported under X through Xrandr, at least reading them,
> not sure about setting them.

Please check with Daniel Vetter.

My belief is that all new KMS properties that were never exposed by any
driver before are new UAPI.

My personal opinion is that Xorg/RandR exposing a KMS property does
*not* count as real userspace *alone*. Simply plumbing a KMS property
through RandR and then nothing actually using it does not prove
anything about the property's design or usability.

IMO, if you use Xorg/RandR as your userspace, you also need something
that uses RandR and really pokes at the new property to prove it's
viable.

But that's just me.

> Anyways I do plan to write some mutter code to test my lcdshadow subsys <->
> DRM driver integration when that is all more then just vaporware. But I
> would prefer for Rajat's series to land before that so that I can build
> on top of it.

The DRM maintainers make that call.


On Tue, 12 May 2020 10:38:11 -0700
Rajat Jain <rajatja at google.com> wrote:

> The chrome browser currently uses the API exposed by my (previous)
> patchset to control privacy screen.
> https://source.chromium.org/chromium/chromium/src/+/master:ui/ozone/platform/drm/common/drm_util.cc;l=180?q=%22privacy-screen%22%20-f:third_party%2Fkernel%2Fv&originalUrl=https:%2F%2Fcs.chromium.org%2F
> 
> I know this doesn't help directly, but just to say that there are
> users waiting to use an API that we release. If these changes are
> accepted, I expect to see the change in browser again, to match the
> new API,  although that will be not until we decide to uprev our
> kernel again.

Chromium counts as userspace, I think many new features have landed
with it as the userspace.

Is that from some development branch, not actually merged or released
yet? If yes, very good. When you submit kernel patches with new UAPI,
it would be nice to point to the userspace review discussion where the
userspace patches have been reviewed and accepted but not merged.


Thanks,
pq
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.freedesktop.org/archives/dri-devel/attachments/20200513/531f6804/attachment-0001.sig>