Re: [PATCH 3/3] drm: Add CRTC_GET_SEQUENCE and CRTC_QUEUE_SEQUENCE ioctls [v2]

From: "Keith Packard" <keithp@keithp.com>
To: Daniel Vetter <daniel@ffwll.ch>
Cc: Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Dave Airlie <airlied@redhat.com>,
	dri-devel <dri-devel@lists.freedesktop.org>
Subject: Re: [PATCH 3/3] drm: Add CRTC_GET_SEQUENCE and CRTC_QUEUE_SEQUENCE ioctls [v2]
Date: Mon, 09 Oct 2017 10:18:30 -0700	[thread overview]
Message-ID: <87bmlg44ux.fsf@keithp.com> (raw)
In-Reply-To: <CAKMK7uGuCACJFYoCEPLU9L_AhHMikvTAAm-v_hYeRCpH8-=Lig@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 1279 bytes --]

Daniel Vetter <daniel@ffwll.ch> writes:

> I just figured that -modesetting would be the simplest domenstration
> vehicle, since the vulkan patches don't look ready yet. I need fully
> reviewed&tested userspace before we can land any kernel stuff. Doing
> the quick modesetting conversion would unblock.

I've provided a patch for the modesetting driver and the preliminary
bits are merged, leaving only a fairly straightforward addition of the
new ioctls to that code. I'm not sure how to make more progress there at
this point; that code would need testing, and it requires a hand-patched
kernel to test.

I also posted IGT tests for the new functionality, again, getting those
reviewed and tested depends on someone willing to build a patched
kernel. Dave Airlie has started trying to get that done.

>> (regarding FIRST_PIXEL_OUT):
>>
>> If the timestamp is the only important thing, it sounds like the kernel
>> already satisfies that, which is cool.
>
> Would be good to confirm that. If it's not, we have a problem.

Michel Dänzer says that the timestamp provided is computed to be
first_pixel_out for all hardware. Given that, I suggest we specify that
in the documentation and remove this bit from the API.

Seem reasonable?

-- 
-keith

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]