Re: [PATCH v5 2/7] drm/panel: simple: Add ability to override typical timing

[Doug Anderson]

Hi,

On Mon, Jul 8, 2019 at 10:56 AM Sam Ravnborg <sam@xxxxxxxxxxxx> wrote:
>
> On Mon, Jul 01, 2019 at 09:39:06AM -0700, Doug Anderson wrote:
> > Hi,
> >
> > On Sun, Jun 30, 2019 at 1:55 PM Sam Ravnborg <sam@xxxxxxxxxxxx> wrote:
> > >
> > > Hi Douglas.
> > >
> > > > > +
> > > > > +   /* Only add timings if override was not there or failed to validate */
> > > > > +   if (num == 0 && panel->desc->num_timings)
> > > > > +           num = panel_simple_get_timings_modes(panel);
> > > > > +
> > > > > +   /*
> > > > > +    * Only add fixed modes if timings/override added no mode.
> > > >
> > > > This part I fail to understand.
> > > > If we have a panel where we in panel-simple have specified the timings,
> > > > and done so using display_timing so with proper {min, typ, max} then it
> > > > should be perfectly legal to specify a more precise variant in the DT
> > > > file.
> > > > Or what did I miss here?
> > >
> > > Got it now.
> > > If display_mode is used for timings this is what you call "fixed mode".
> > > Hmm, if I got confused someone else may also be confused by this naming.
> >
> > The name "fixed mode" comes from the old code, though I guess in the
> > old code it used to refer to a mode that came from either the
> > display_timing or the display_mode.
> >
> > How about if I call it "panel_simple_get_from_fixed_display_mode"?
> > ...or if you have another suggestion feel free to chime in.
> What we really want to distingush here is the use of display_mode
> and display_timings (if I got the names right).
> That display_mode specify a fixed timing and display_timing specify
> a valid range is something in the semantics of the two types.
> So naming that refer to display_mode versus display_timing will make the
> code simpler to understand. and then a nice comment that when
> display_mode
> is used one looses the possibility to use override_mode.
> That would be fine to have in the struct in the driver.

OK, I can change the names here and try to find a good place to add a comment.


> > NOTE: Since this feedback is minor and this patch has been outstanding
> > for a while (and is blocking other work), I am assuming that the best
> > path forward is for Heiko to land this patch with Thierry's Ack and
> > I'll send a follow-up.  Please yell if you disagree.
> Let's give the patches a spin more as we have passed the possibility for
> the current merge window.

Any way I can convince you to change your mind here?  There are no
functional changes requested so far in your feedback and no bugs--it's
just a few variable names and comments.  By landing the existing
patches as-is:

1. We stop spamming all the people CCed on this whole series (which
includes device tree patches) that might be interested in the series
as a whole but aren't interested in details.

2. We can debate the bikeshed-type issues on their own merit and I
don't have to debate removing existing Acks / Reviewed-by / Tested-by
tags as I make changes.

3. Even if it's not a good time to land the patches right now we know
that these patches will be ready to land as soon as the window opens.
As I mentioned earlier these patches are blocking other work [1] and
landing that patch is actually preventing Matthias from submitting
another series of patches to add support for rk3288-veyron-tiger and
rk3288-veyron-fievel.  Certainly I know that upstream doesn't make a
policy of landing things just to suit the timelines of a downstream
project, but in this case there seems very little downsides to landing
the existing patches and taking a later cleanup patch.


> I am on vacation at the moment and thus slow in responses, but will be back
> at the home office next week and will be more responsive again.
>
>         Sam - who is enjoying the alps in Austria

Hope you have had a great vacation!

[1] https://lkml.kernel.org/r/20190625222629.154619-1-mka@xxxxxxxxxxxx

-Doug