Re: [PATCH 2/4 v4] i.MX31: Image Processing Unit DMA and IRQ drivers

[Guennadi Liakhovetski <g.liakhovetski@gmx.de>]

Hi Robert,

On Tue, 23 Dec 2008, Robert Schwebel wrote:

> On Tue, Dec 23, 2008 at 11:50:06AM +0100, Sascha Hauer wrote:
> > > You mean an output v4l device? I think overlays are handled by framebuffer 
> > > drivers... But I'm also not quite sure about it, however, handling overlay 
> > > as another framebuffer seems logical to me.
> >
> > Well the DMA engine seems to suggest that frames should be passed around
> > whereas the framebuffer API only has a single frame. That would fit
> > better into the v4l API. Also the IPU can do things like colourspace
> > conversion and hw scaling which would fit into the V4L API.
> 
> Looks like a candidate for gstreamer on the userspace end. Can it be
> decoupled enough to make proper plugins out of it?

I think, the question is rather "can a driver be written for IPU to 
support some sane hardware-neutral image data manipulation API, like v4l?" 
Then you can start writing any user-space apps on top of that API. So, as 
long as there is such an API, I think, we can put it on IPU, yes. My 
problems ATM is - no use-case. What concerns overlay, it is not used in 
the current application, as for image format conversion from the camera - 
I only have a Bayer camera to test with. I even cannot test monochrome 
sanely, because of a lack of a cable:-) And Bayer and monochrome don't 
make good candidates for such conversions - they are not supported by the 
IPU.

> > BTW is the overlay framebuffer useful in it's current implementation?
> > There seems to be no way to adjust the x/y offset or the blending
> > modes.
> 
> The API Eric Miao just posted for the PXA looks sane to me.

I'll have a look, but as I said: -ENOUSER:-)

> > > If there are no other problems with v5, could we maybe take it as a
> > > basis and then I would submit a patch to reduce the number of IRQs?
> >
> > Please understand my concerns with this driver. It's a quite complex
> > beast and experience shows that once a driver is in the kernel it is
> > far more complicated to change it than to do it right the first way.
> 
> Especially when it comes to userspace visible things.
> 
> > You know that I'm also interested in having a MX31 framebuffer (and
> > camera) driver in kernel but I want to make sure that it works
> > properly and leaves room for feature enhancements without having to
> > refactor the whole driver.
> 
> Yup, looks like it would be better to cook it another round instead of
> trying to bring in a half-tested driver with brute-force.

Ok, no problem. Let's just decide

1. is the drivers/dma the final location.
2. do dmaengine maintainers accept it in present form or require any 
   amendments.
3. which interrupts we make visible by default for irq_desc[] and what 
   granularity we want to enable the rest (we could even just make 
   CONFIG_ALL_IPU_IRQS and be 95% sure noone will ever need them.)
4. throw away every trace of overlay support - if anyone ever needs it 
   they should be able to dig it out in ML archives.

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer