[RFC,04/10] drm/panel_helper: Introduce drm_panel_helper

  The goal of this file is to contain helper functions for panel drivers
to use. To start off with, let's add drm_panel_helper_shutdown() for
use by panels that want to make sure they're powered off at
shutdown/remove time if they happen to be powered on.

The main goal of introducting this function is so that panel drivers
don't need to track the enabled/prepared state themselves.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
---
If I've misunderstood and the drm_panel_helper_shutdown() should
belong in some other file and we don't need to introduce a "helper"
for this then please le me know.

 drivers/gpu/drm/Makefile           |  1 +
 drivers/gpu/drm/drm_panel_helper.c | 37 ++++++++++++++++++++++++++++++
 include/drm/drm_panel_helper.h     | 13 +++++++++++
 3 files changed, 51 insertions(+)
 create mode 100644 drivers/gpu/drm/drm_panel_helper.c
 create mode 100644 include/drm/drm_panel_helper.h
  

Hi Doug,

Thanks for working on this :)

On Fri, Aug 04, 2023 at 02:06:07PM -0700, Douglas Anderson wrote:
> The goal of this file is to contain helper functions for panel drivers
> to use. To start off with, let's add drm_panel_helper_shutdown() for
> use by panels that want to make sure they're powered off at
> shutdown/remove time if they happen to be powered on.
> 
> The main goal of introducting this function is so that panel drivers
> don't need to track the enabled/prepared state themselves.
> 
> Signed-off-by: Douglas Anderson <dianders@chromium.org>

It shouldn't be necessary at all: drivers should call
drm_atomic_helper_shutdown at removal time which will disable the
connector (which in turn should unprepare/disable its panel).

If either the driver is missing drm_atomic_helper_shutdown, or if the
connector doesn't properly disable the panel, then I would consider that
a bug.

Maxime
  

On Fri, Aug 25, 2023 at 02:58:02PM -0700, Doug Anderson wrote:
> Maxime,
> 
> On Sun, Aug 6, 2023 at 11:41 PM Maxime Ripard <mripard@kernel.org> wrote:
> >
> > Hi Doug,
> >
> > Thanks for working on this :)
> >
> > On Fri, Aug 04, 2023 at 02:06:07PM -0700, Douglas Anderson wrote:
> > > The goal of this file is to contain helper functions for panel drivers
> > > to use. To start off with, let's add drm_panel_helper_shutdown() for
> > > use by panels that want to make sure they're powered off at
> > > shutdown/remove time if they happen to be powered on.
> > >
> > > The main goal of introducting this function is so that panel drivers
> > > don't need to track the enabled/prepared state themselves.
> > >
> > > Signed-off-by: Douglas Anderson <dianders@chromium.org>
> >
> > It shouldn't be necessary at all: drivers should call
> > drm_atomic_helper_shutdown at removal time which will disable the
> > connector (which in turn should unprepare/disable its panel).
> >
> > If either the driver is missing drm_atomic_helper_shutdown, or if the
> > connector doesn't properly disable the panel, then I would consider that
> > a bug.
> 
> Hmmm. I'm a bit hesitant here. I guess I'm less worried about the
> removal time and more worried about the shutdown time.
> 
> For removal I'd be fine with just dropping the call and saying it's
> the responsibility of the driver to call drm_atomic_helper_shutdown(),
> as you suggest. I'd tend to believe that removal of DRM drivers is not
> used anywhere in "production" code (or at least not common) and I
> think it's super hard to get it right, to unregister and unbind all
> the DRM components in the right order.

It depends on the kind of devices. USB devices are very likely to be
removed, platform devices very unlikely, and PCIe cards somewhere in the
middle :)

I'm not sure the likelyhood of the device getting removed has much to do
with it though, and likely or not, it's definitely something we should
address and fix if an issue is to be found.

> Presumably anyone trying to remove a DRM panel in a generic case
> supporting lots of different hardware is used to it being a bit
> broken...

It's not. Most drivers might be broken, but it's totally something we
support and should strive for.

> Not that it's a super great situation to be in for remove() not to
> work reliably, but that's how I think it is right now.
> 
> For shutdown, however, I'm not really OK with just blindly removing
> the code that tries to power off the panel.

I disagree with that statement. It's not "blindly removing the code",
that code is still there, in the disable hook.

> Shutdown is called any time you reboot a device. That means that if a
> DRM driver is _not_ calling drm_atomic_helper_shutdown() on the
> panel's behalf at shutdown time then the panel won't be powered off
> properly. This feels to me like something that might actually matter.

It does matter. What I disagree on is that you suggest working around
that brokenness in the core framework. What I'm saying is driver is
broken, we should keep the core framework sane and fix it in the driver.

It should be fairly easy with a coccinelle script to figure out which
panels are affected, and to add that call in remove.

> Panels tend to be one of those things that really care about their
> power sequencing and can even get damaged (or not turn on properly
> next time) if sequencing is not done properly, so just removing this
> code and putting the blame on the DRM driver seems scary to me.

Sure, it's bad. But there's no difference compared to the approach you
suggest in that patch: you created a helper, yes, but every driver will
still have to call that helper and if they don't, the panel will still
be called and it's a bug. And we would have to convert everything to
that new helper.

It's fundamentally the same discussion than what you were opposed to
above.

> Sure enough, a quick survey of DRM drivers shows that many don't call
> drm_atomic_helper_shutdown() at .shutdown time. A _very_ quick skim of
> callers to drm_atomic_helper_shutdown():
> 
> * omapdrm/omap_drv.c - calls at remove, not shutdown
> * arm/hdlcd_drv.c - calls at unbind, not shutdown
> * arm/malidp_drv.c - calls at unbind, not shutdown
> * armada/armada_drv.c - calls at unbind, not shutdown
> 
> ...huh, actually, there are probably too many to list that don't call
> it at shutdown. There are some that do, but also quite a few that
> don't. I'm not sure I really want to blindly add
> drm_atomic_helper_shutdown() to all those DRM driver's shutdown
> callbacks... That feels like asking for someone to flame me...

No one will flame you, and if they do, we'll take care of it. And yes,
those are bugs, so let's fix them instead of working around them?

> ...but then, what's the way forward? I think normally the panel's
> shutdown() callback would happen _before_ the DRM driver's shutdown()
> callback,

Is there such a guarantee?

> so we can't easily write logic in the panel's shutdown like "if the
> DRM panel didn't shut the panel down then print a warning and shut
> down the panel". We'd have to somehow invent and register for a "late
> shutdown" callback and have the panel use that to shut itself down if
> the DRM driver didn't. That seems like a bad idea...
> 
> Do you have any brilliant ideas here? I could keep the function as-is
> but only have panels only call it at shutdown time if you want. I
> could add to the comments and/or the commit message some summary of
> the above and that the call is important for panels that absolutely
> need to be powered off at shutdown time even if the DRM driver doesn't
> do anything special at shutdown... Any other ideas?

Brilliant ideas to do what exactly?

I disagree that the solution to our problem is to disable the panel
resources at shutdown time and we should only do it at disable time.
Your helper does exactly that though, so I don't think the helper is a
good idea.

Maxime
  

Hi,

On Thu, Sep 7, 2023 at 7:15 AM Maxime Ripard <mripard@kernel.org> wrote:
>
> > a) Block landing the change to "panel-tdo-tl070wsh30.c" until after
> > all drivers properly call drm_atomic_helper_shutdown().
>
> I don't think we care about all drivers here. Only the driver it's
> enabled with would be a blocker. Like, let's say sun4i hasn't been
> converted but that panel is only used with rockchip anyway, we don't
> really care that sun4i shutdown patch hasn't been merged (yet).

Yeah, I suppose that would work, though it would require a bit of
research. Some other things have popped onto my plate recently, but
for now I'm going to focus on seeing how much of the drivers can get
their shutdown fixed. When that stalls out then we can see if we can
unblock some of the panels by digging into which DRM drivers they're
used with.

Also, as my proposal in the cover letter [1], I'm leaving a breadcrumb
here that I landed the first 3 patches in this series just to get them
out of the way. Those 3 patches didn't depend on the resolution of the
issues discussed here.

[1] https://lore.kernel.org/lkml/CAD=FV=UFuUsrrZmkL8_RL5WLvkJryDwRSAy_PWTa-hX_p0dF+Q@mail.gmail.com/