[PATCH 0/5] drm/tinydrm: Add tinydrm_panel abstraction

[Daniel Vetter]

On Sun, Mar 12, 2017 at 06:50:17PM +0100, Daniel Vetter wrote:
> Hi Noralf,
> 
> On Sat, Mar 11, 2017 at 10:35:31PM +0100, Noralf Trønnes wrote:
> > Add support for displays that have a register interface and can be
> > operated using a simple vtable.
> > 
> > I have looked through the staging/fbtft drivers and it seems that,
> > except the MIPI controllers, most if not all controllers are operated
> > through a register. And since most controllers have more than one bus
> > interface option, regmap seems like a good choice to describe the
> > interface (tested[1,2]).
> > MIPI DCS can't be represented using regmap since some commands doesn't
> > have a parameter. That would be like a register without a value, which
> > doesn't make sense.
> > 
> > In my second RFC of tinydrm I used drm_panel to decribe the panels
> > since it was a good match to the fbtft displays. I was then told that
> > drm_panel wasn't supposed to used like that, so I dropped it and have
> > tried to use the drm_simple_display_pipe_funcs vtable directly. This
> > hasn't been all successful, since I ended up using devm_add_action() to
> > power down the controller at the right time. Thierry Reding wasn't
> > happy with this and suggested "to add an explicit callback somewhere".
> > My solution has been to copy the drm_panel_funcs vtable.
> > Since I now have a vtable, I also added a callback to flush the
> > framebuffer. So presumably all the fbtft drivers can now be operated
> > through the tinydrm_panel_funcs vtable.
> 
> Ehrm, what? I admit I didn't follow the discussion in-depth, but if
> drm_panel can't be used to describe a panel, it's not fit for the job and
> needs to be extended. Adding even more abstraction, matroschka-style,
> isn't a solution.
> 
> Personally I think tinydrm itself is already a bit much wrapping, but I
> see that for really simple drivers it has it's uses. But more layers feels
> like going in the wrong direction.
> 
> For the callback you're looking for (i.e. the regulator_disable call) I
> think the correct place is to enable/disable the regulator in the
> enable/disable hooks of the drm_simple_display_pipe functions. Or maybe in
> their equivalent in drm_panel (well, probably pre_enable and post_disable,
> since I guess you need that regulator to driver anything). Same for _init,
> if the display is completely off there's no point in keeping the hw
> running. Enabling/disabling the entire hw is pretty much what ->enable and
> ->disable are for. This also gives you proper runtime pm for almost for
> free ...
> 
> Also, since the regulator is actually stored in struct mipi_dbi, it's
> probably best to handle it in the mipi_dbi helpers too. You do that
> already with the backlight anyway.
> 
> I noticed that the version of tinydrm that landed doesn't use drm_panel
> anymore, I thought that was the case once, and for the version I acked?
> 
> > After having done this the question arises:
> > Why not extend tinydrm_device instead of subclassing it?
> > 
> > The benefit of subclassing is that it keeps the door open for drivers
> > that can use tinydrm_device, but not tinydrm_panel. But I don't know of
> > such a driver now, then again who knows what the future brings.
> > Something that might or might not happen isn't a good reason, so it
> > seems that I want it this way because I just like it. And it's easy to
> > merge the two should it be that no one uses tinydrm_device directly
> > three years down the line. But I'm actually not sure what's best.
> 
> As a rule of thumb, never design code for future use that you don't know
> yet will happen. No one can reliably predict the future, which means
> you'll get it wrong, and in the future we'll have to do 2x the work: Once
> do unto the code resulting from the wrong prediction, then redoing it the
> way we need to. Not including the on-going burden of maintaining unused
> functionally.
> 
> So let's pls merge first, split later when there's a clean need.
> 
> > To recap:
> > 
> > tinydrm_device
> > - Combines drm_simple_display_pipe with CMA backed framebuffer and fbdev.
> > - Optional pipe setup with a connector with one mode, but the driver
> >   can do it's own.
> > 
> > tinydrm_panel
> > - All drm operations are distilled down to tinydrm_panel_funcs.
> > - Some common driver properties
> 
> So overal sorry I'm shredding this a bit, but I don't see the point. Imo
> much more useful would be:
> 
> 1. Extract the non-tinydrm specific helpers from tinydrm and put them into
> the right places. I think from our discussions this was:
> 
> - backlight helpers, probably best to put them into a new drm_backlight.c.
>   This is because drivers/video is defactor unmaintained. We could also
>   move drivers/video/backlight to drivers/gpu/backlight and take it all
>   over within drm-misc, but that's more work.
> 
> - spi helpers, probably best put into spi core/helper code. Thierry said
>   the spi maintainer is fast&reactive, so shouldn't be a big issue.
> 
> - extract the mipi-dbi helper (well, the non-tinydrm specific parts at
>   least) into a separate helper, like we have for mipi-dsi already.

A large chunk of the tinydrm functions should probably be moved into
relevant existin drm helpers, e.g.

- tinydrm_lastclose could be drm_fb_helper_lastclose. Only thing we need
  for that is to store the drm_fb_helper pointer somewhere in
  drm_device->mode_config. And thenc we could roll that out to all the
  drivers.

- tinydrm_gem_cma_prime_import_sg_table should probably go into the cma
  helpers, as a _vmapped variant (since not every driver needs the vmap).
  And tinydrm_gem_cma_free_object could the be merged into
  drm_gem_cma_free_object().

- tinydrm_fb_create we could move into drm_simple_pipe, only need to add
  the fb_create hook there, which would again simplify a bunch of things
  (since it gives you a one-stop vfunc for simple drivers).

- Quick aside: The unregister devm stuff is kinda getting the lifetimes of
  a drm_device wrong. Doesn't matter, since everyone else gets it wrong
  too :-)
 
- With the fbdev pointer in dev->mode_config we could also make
  suspend/resume helpers entirely generic, at least if we add a
  dev->mode_config.suspend_state.

Just a bunch of ideas. If you don't feel like doing those, ok if I add
them to todo.rst as tinydrm refactorings?
-Daniel

> 
> 2. Add tons more panel drivers. Personally I'd do that first :-)
> 
> Cheers, Daniel
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch