[Intel-gfx] [PATCH 1/3] drm/i915: Convert intel_tv connector properties to atomic, v3.

[Daniel Vetter]

On Thu, Mar 16, 2017 at 04:27:42PM +0100, Maarten Lankhorst wrote:
> Op 16-03-17 om 15:56 schreef Daniel Vetter:
> > On Thu, Mar 16, 2017 at 01:05:06PM +0100, Maarten Lankhorst wrote:
> >> Op 16-03-17 om 11:48 schreef Daniel Vetter:
> >>> On Wed, Mar 15, 2017 at 11:08:56AM +0100, Maarten Lankhorst wrote:
> >>>> As a proof of concept, first try to convert intel_tv, which is a rarely
> >>>> used connector. It has 5 properties, tv format and 4 margins.
> >>>>
> >>>> I'm less certain about the state behavior itself, should we pass a size
> >>>> parameter to intel_connector_alloc instead, so duplicate_state
> >>>> can be done globally if it can be blindly copied?
> >>>>
> >>>> Can we also have a atomic_check function for connectors, so the
> >>>> crtc_state->connectors_changed can be set there? It would be cleaner
> >>>> and more atomic-like.
> >>>>
> >>>> To match the legacy behavior, format can be changed by probing just like
> >>>> in legacy mode.
> >>>>
> >>>> Changes since v1:
> >>>> - Add intel_encoder->swap_state to allow updating connector state.
> >>>> - Add intel_tv->format for detect_mode and mode_valid, updated on atomic commit.
> >>>> Changes since v2:
> >>>> - Fix typo in tv_choose_preferred modes function name.
> >>>> - Assignment of tv properties is done in core, so intel_tv only needs
> >>>>   a atomic_check function. Thanks Ville!
> >>>>
> >>>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> >>>> ---
> >>>>  drivers/gpu/drm/i915/intel_display.c |  15 +++
> >>>>  drivers/gpu/drm/i915/intel_drv.h     |   4 +
> >>>>  drivers/gpu/drm/i915/intel_tv.c      | 200 ++++++++++++++++-------------------
> >>>>  3 files changed, 110 insertions(+), 109 deletions(-)
> >>>>
> >>>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> >>>> index ac25c9bc8b81..18b7e7546ee1 100644
> >>>> --- a/drivers/gpu/drm/i915/intel_display.c
> >>>> +++ b/drivers/gpu/drm/i915/intel_display.c
> >>>> @@ -12954,6 +12954,20 @@ static void intel_atomic_track_fbs(struct drm_atomic_state *state)
> >>>>  				  to_intel_plane(plane)->frontbuffer_bit);
> >>>>  }
> >>>>  
> >>>> +static void swap_connector_state(struct drm_atomic_state *state)
> >>>> +{
> >>>> +	struct drm_connector *connector;
> >>>> +	struct drm_connector_state *new_conn_state;
> >>>> +	int i;
> >>>> +
> >>>> +	for_each_new_connector_in_state(state, connector, new_conn_state, i) {
> >>>> +		struct intel_connector *conn = to_intel_connector(connector);
> >>>> +
> >>>> +		if (conn->swap_state)
> >>>> +			conn->swap_state(conn, new_conn_state);
> >>>> +	}
> >>>> +}
> >>> So yeah I'm late, but this is pretty much exactly what DK's private state
> >>> object stuff was meant for. Have you looked into rebasing on top of that?
> >>> I think hand-rolling private state stuff everywhere isn't really good.
> >>>
> >>> Afaics DK's patch series is ready for merging into drm-misc, so shouldn't
> >>> end up blocking you.
> >> This swap_state hook is used for connector state. We could probably add a swap_state hook that drivers
> >> can use for this in the core, but the connector state is already part of the state, adding it twice is silly.
> >>
> >> Instead, would adding a hook to the core be useful?
> > Why do you need a hook to swap the state? The core already does that for
> > you ... to_tv_state(intel_tv->base.base->state) is probably what you want,
> > instead of deref'ing a new intel_tv->format.
> > -Daniel
> 
> Well in some way I need to access connector_state->tv.mode in mode_valid, which doesn't hold any connector state locks.
> I think setting it from the state is the easiest way to do so. But I'm open to other approaches. :)

Well fundamentally that access is going to be racy, so you better put a
READ_ONCE into mode_valid. But once we accept that it's racy, we might as
well chase connector->state, the only thing we need is to rcu-protected
the freeing of old atomic state.

I've typed such a patch yesterday (for vblank access to crtc->state from
vblank interrupts), but then thrown it away again. Because it didn't fix
my vblank issue correctly.

I'll resurrect and send out.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch