[PATCH v2 1/3] imx-drm: Add mx6 hdmi transmitter support

[Sascha Hauer]

On Wed, Oct 16, 2013 at 08:02:11PM +0100, Russell King - ARM Linux wrote:
> On Wed, Oct 16, 2013 at 11:31:07AM -0700, Greg Kroah-Hartman wrote:
> > On Wed, Oct 16, 2013 at 07:07:35PM +0100, Russell King - ARM Linux wrote:
> > > Sorry, but I don't think imx-drm is driving the hardware correctly, and
> > > I know that Greg wants it moved out of drivers/staging, but frankly it
> > > seems to be far from ready for that.  Certainly the HDMI parts seems to
> > > be especially problematical.
> > 
> > I want it out of staging if it's working properly.  Yours is the first
> > report of it not working properly, and in fact, probably one of the
> > first users of the driver, as I haven't gotten any reports of it working
> > or not at all over the years.
> 
> Well, part of that is because I have this other thing called Armada DRM,
> which is a similar thing to imx-drm, except for the Marvell Dove SoCs,
> so it's been really quite easy to get a full Ubuntu 12.04 up and running
> on the IMX SoC I have here.
> 
> As part of that effort, I'm now using my Armada DRM Xorg driver with
> imx-drm to test this stuff out.  (Bearing in mind that IMX and Dove use
> the same Vivante GPU, there's some sense in getting my Xorg Armada DRM
> driver working on both.)
> 
> I'm not aware of there being any X drivers for imx-drm (google turns
> up nothing), which might be a reason why it hasn't been well tested
> and has also languished in drivers/staging for soo long.  Alternatively,
> maybe my google searching sucks, or it is out there somewhere but
> hidden from googlebot?

There is no X driver for imx-drm. I once tested it with the
xf86-modesetting driver which worked with a few patches back then. These
patches are now part of the modesetting driver.

We have a wrapper tool here which we use to configure the KMS part and
pass the framebuffers to QT.

> 
> To be fair, so far most of the problems I'm finding are with the HDMI
> addition to imx-drm rather than imx-drm itself: there's been the lockdep
> problem and the clock polarity problem which the HDMI stuff discovered.
> 
> Looking at the todo list for moving it out of staging, we have:
> 
> - get DRM Maintainer review for this code
> - Wait for common display framework to hit mainline and update the IPU
>   driver to use it. This will most probably make changes to the devicetree
>   bindings necessary.
> - Factor out more code to common helper functions
> - decide where to put the base driver. It is not specific to a subsystem
>   and would be used by DRM/KMS and media/V4L2
> 
> (1) is quite a difficult task: I'm waiting for a review of the Armada DRM
> stuff, which I'm hoping will make the next merge window.  Rob Clark is
> looking at that but has been distracted recently from completing it.

We have waited for quite a long time aswell before we decided to push
the imx-drm driver to staging to at least get more exposure for the
driver. The situation is, well... unsatisfying.

> 
> (2) CDF... has been around in RFC according to LWN.net but doesn't seem to
> make much in the way of progress from what I can see, despite there
> allegedly being "many developers show[ing] interest in the first RFC".
> CDF is over a year old now - first RFC 17 Aug 2012, second 22 Nov 2012,
> third 9 Aug 2013.

>From what I heard Laurent is still committed to mainline CDF.

> 
> (3) is an on-going effort and really isn't a reason for it to stay in
> staging.
> 
> (4) is probably the biggest issue.  The problem there is that the IMX
> display subsystem is very flexible: it's basically a set of DMA engines
> on the front, and behind that a fair number of modules implementing
> facilities like image rotation, display driving and image capture.
> Display driving alone involves setting up waveforms and writing some
> microcode to tell the thing what to do on certain events (like end
> of frame etc.)  Hence it doesn't fit well into any particular "Linux"
> subsystem.  In this regard, it's a victim of its own flexibility.
> 
> I think the biggest problem though is its complexity.  It doesn't fit
> into the "single device for a card" model which DRM has.  It's made
> up from several separate devices, which is all very well, but it leads
> to it having its own private "framework" to glue all the different
> devices together - which adds a layer of indirection and makes the code
> harder to understand.

I tried adding generic stuff to DRM to support this and failed badly due
to resistance of the maintainers. I didn't get much input what I should
do better or different (besides a general "make it a helper library").
Also nobody seemed to understand the problems I had with the multi device
nature of the i.MX IPU.

I think we made ourselves comfortable with being in the kernel, even if
it's only in staging. We should move forward instead and see how we get
the driver into the kernel.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |