Re: [PATCH] drm/simple-kms: Standardize arguments for callbacks

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Oct 23, 2019 at 08:47:57AM +0200, Thomas Zimmermann wrote:
> Hi
> 
> Am 22.10.19 um 21:03 schrieb Daniel Vetter:
> > On Tue, Oct 22, 2019 at 7:16 PM Thomas Zimmermann <tzimmermann@xxxxxxx> wrote:
> >>
> >> Hi,
> >>
> >> there are two types of callbacks in struct
> >> drm_simple_display_pipe_funcs: functions that are genuine to simple KMS,
> >> and functions that are merely forwarded from another structure (crtc,
> >> plane, etc).
> >>
> >> In the former category are enable(), disable(), check(), and update().
> >> Those should probably receive a simple display pipe as their first argument.
> > 
> > mode_valid _very_ much belongs to this category too, since there's
> > mode_valid hooks also on other objects. But simple pipe helper
> > condenses that down to one mode_valid hook (we could also put the
> > mode_valid onto encoder, wouldn't change anything).
> > 
> >> In the latter category are mode_valid(), prepare_fb(), cleanup_fb(),
> >> enable_vblank(), and disable_vblank(). IMHO those functions should
> >> receive a pointer to the original structure as their first argument.
> >> This type provides the context in which the operations make sense. (Even
> >> their documentation already refers to the original structure.)
> > 
> > Now on those you can maybe make a case that they only exist in one
> > object. But the entire point of simple helpers was to condense the zoo
> > of drm types down to one. Only reason you don't also get a
> > drm_simple_display_pipe_state is that this one would be a bit more
> > work to make work correctly. If we full on leak all the underlying
> > objects, then you might as well set them up yourself and set up all
> > the hooks, it's just a few more lines of code.
> > 
> > Imo for simple pipe we should go more into that direction, not less.
> > 
> >> I admit that not all are as shareable as prepare_fb() and enable_fb().
> >> But what else than boiler-plate wrappers do we get from simple display
> >> pipe here?
> > 
> > Boiler plate wrappers is pretty much the entire point of simple pipe
> > helpers. Anytime you're interested in the things it abstracts away
> > (crtc, plane, encoder) you probably want your own atomic
> > implementation.
> 
> I was speaking of boiler-plate code in drivers and other helpers (e.g.,
> drm_gem_fb_simple_display_pipe_prepare_fb() )
> 
> TBH I don't think it is possible to build and use simple pipe without
> exposing the underlying primitives (crtc, plane, connector). This would
> require a completely separate set of atomic helpers. IMHO the current
> simple pipe is a mid-layer and comes with typical mid-layer problems.

Helpers can be midlayers, as long as their optional. And for simple I
agree it's not a perfect illusion, it's a tradeoff between having a huge
helper library that remaps everything and still enabling useful code and
complexity savings in the tiny drivers.

Maybe another rule of thumb: If your driver has more than one source file,
simple pipe is maybe not what you're looking for. Exceptions apply ofc.
simple pipe was designed for drm/tiny, and it utterly excels at that. But
that doesn't make it a general purpose tool.

> Anyway, given your rational for the current design, I'll update my
> recent patches for prepare_fb() to support simple pipe.
> 
> For this patch
> 
>   Acked-By: Thomas Zimmermann <tzimmermann@xxxxxxx>

Thanks, will apply.
-Daniel

> 
> Best regards
> Thomas
> 
> > conversion is a good fit, it's not meant to be useful for all small
> > drivers. Only for the _really_ simple ones.
> > 
> > Otherwise if we readd all the bells and whistles to simple pipe
> > helpers, then we just end back where we started. That's also why I
> > personally think explicit simple wrappers would fit better, instead of
> > wrestling the prepare/cleanup_fb functions to match full atomic
> > helpers.
> > -Daniel
> > 
> >>
> >> Best regards
> >> Thomas
> >>
> >> Am 22.10.19 um 17:55 schrieb Daniel Vetter:
> >>> Passing the wrong type feels icky, everywhere else we use the pipe as
> >>> the first parameter. Spotted while discussing patches with Thomas
> >>> Zimmermann.
> >>>
> >>> Cc: Thomas Zimmermann <tzimmermann@xxxxxxx>
> >>> Cc: Noralf Trønnes <noralf@xxxxxxxxxxx>
> >>> Cc: Gerd Hoffmann <kraxel@xxxxxxxxxx>
> >>> Cc: Eric Anholt <eric@xxxxxxxxxx>
> >>> Cc: Emil Velikov <emil.velikov@xxxxxxxxxxxxx>
> >>> Cc: virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx
> >>> Cc: Linus Walleij <linus.walleij@xxxxxxxxxx>
> >>> Signed-off-by: Daniel Vetter <daniel.vetter@xxxxxxxxx>
> >>> ---
> >>>  drivers/gpu/drm/cirrus/cirrus.c         | 2 +-
> >>>  drivers/gpu/drm/drm_simple_kms_helper.c | 2 +-
> >>>  drivers/gpu/drm/pl111/pl111_display.c   | 4 ++--
> >>>  include/drm/drm_simple_kms_helper.h     | 2 +-
> >>>  4 files changed, 5 insertions(+), 5 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/cirrus/cirrus.c b/drivers/gpu/drm/cirrus/cirrus.c
> >>> index 7d08d067e1a4..248c9f765c45 100644
> >>> --- a/drivers/gpu/drm/cirrus/cirrus.c
> >>> +++ b/drivers/gpu/drm/cirrus/cirrus.c
> >>> @@ -390,7 +390,7 @@ static int cirrus_conn_init(struct cirrus_device *cirrus)
> >>>  /* ------------------------------------------------------------------ */
> >>>  /* cirrus (simple) display pipe                                            */
> >>>
> >>> -static enum drm_mode_status cirrus_pipe_mode_valid(struct drm_crtc *crtc,
> >>> +static enum drm_mode_status cirrus_pipe_mode_valid(struct drm_simple_display_pipe *pipe,
> >>>                                                  const struct drm_display_mode *mode)
> >>>  {
> >>>       if (cirrus_check_size(mode->hdisplay, mode->vdisplay, NULL) < 0)
> >>> diff --git a/drivers/gpu/drm/drm_simple_kms_helper.c b/drivers/gpu/drm/drm_simple_kms_helper.c
> >>> index 046055719245..15fb516ae2d8 100644
> >>> --- a/drivers/gpu/drm/drm_simple_kms_helper.c
> >>> +++ b/drivers/gpu/drm/drm_simple_kms_helper.c
> >>> @@ -43,7 +43,7 @@ drm_simple_kms_crtc_mode_valid(struct drm_crtc *crtc,
> >>>               /* Anything goes */
> >>>               return MODE_OK;
> >>>
> >>> -     return pipe->funcs->mode_valid(crtc, mode);
> >>> +     return pipe->funcs->mode_valid(pipe, mode);
> >>>  }
> >>>
> >>>  static int drm_simple_kms_crtc_check(struct drm_crtc *crtc,
> >>> diff --git a/drivers/gpu/drm/pl111/pl111_display.c b/drivers/gpu/drm/pl111/pl111_display.c
> >>> index 024771a4083e..703ddc803c55 100644
> >>> --- a/drivers/gpu/drm/pl111/pl111_display.c
> >>> +++ b/drivers/gpu/drm/pl111/pl111_display.c
> >>> @@ -48,10 +48,10 @@ irqreturn_t pl111_irq(int irq, void *data)
> >>>  }
> >>>
> >>>  static enum drm_mode_status
> >>> -pl111_mode_valid(struct drm_crtc *crtc,
> >>> +pl111_mode_valid(struct drm_simple_display_pipe *pipe,
> >>>                const struct drm_display_mode *mode)
> >>>  {
> >>> -     struct drm_device *drm = crtc->dev;
> >>> +     struct drm_device *drm = pipe->crtc.dev;
> >>>       struct pl111_drm_dev_private *priv = drm->dev_private;
> >>>       u32 cpp = priv->variant->fb_bpp / 8;
> >>>       u64 bw;
> >>> diff --git a/include/drm/drm_simple_kms_helper.h b/include/drm/drm_simple_kms_helper.h
> >>> index 4d89cd0a60db..15afee9cf049 100644
> >>> --- a/include/drm/drm_simple_kms_helper.h
> >>> +++ b/include/drm/drm_simple_kms_helper.h
> >>> @@ -49,7 +49,7 @@ struct drm_simple_display_pipe_funcs {
> >>>        *
> >>>        * drm_mode_status Enum
> >>>        */
> >>> -     enum drm_mode_status (*mode_valid)(struct drm_crtc *crtc,
> >>> +     enum drm_mode_status (*mode_valid)(struct drm_simple_display_pipe *pipe,
> >>>                                          const struct drm_display_mode *mode);
> >>>
> >>>       /**
> >>>
> >>
> >> --
> >> Thomas Zimmermann
> >> Graphics Driver Developer
> >> SUSE Software Solutions Germany GmbH
> >> Maxfeldstr. 5, 90409 Nürnberg, Germany
> >> (HRB 36809, AG Nürnberg)
> >> Geschäftsführer: Felix Imendörffer
> >>
> > 
> > 
> 
> -- 
> Thomas Zimmermann
> Graphics Driver Developer
> SUSE Software Solutions Germany GmbH
> Maxfeldstr. 5, 90409 Nürnberg, Germany
> (HRB 36809, AG Nürnberg)
> Geschäftsführer: Felix Imendörffer
> 




-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel




[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux