Re: Baytrail camera csi / isp support status ?

[Bastien Nocera <hadess@hadess.net>]

On Tue, 2015-01-06 at 09:55 +0200, Laurent Pinchart wrote:
> Hi Bastien,
> 
> On Tuesday 06 January 2015 01:06:54 Bastien Nocera wrote:
> > On Fri, 2014-12-12 at 12:07 +0100, Hans de Goede wrote:
> > > Hi All,
> > > 
> > > A college of mine has a baytrail bases tablet:
> > > 
> > > http://www.onda-tablet.com/onda-v975w-quad-core-win-8-tablet-9-7-inch-reti
> > > na-screen-ram-2gb-wifi-32gb.html
> > > 
> > > And he is trying to get Linux to run on it, he has things mostly
> > > working, but he would also like to get the cameras to work.
> > > 
> > > I've found this:
> > > 
> > > http://sourceforge.net/projects/e3845mipi/files/
> > > 
> > > Which is some not so pretty code, with the usual problems of using
> > > custom ioctls to pass info from the statistics block of the isp
> > > to userspace and then let some userspace thingie (blob?) handle it.
> > > 
> > > So I was wondering if anyone is working on proper support
> > > (targeting upstream) for this ? It would be nice if we could at least
> > > get the csi bits going, using the sensors or software auto-whitebal, etc.
> > > for now.
> > 
> > As I mentioned to Hans in private, I would be ready to provide the
> > hardware for somebody with a track record to keep, to allow testing and
> > hopefully maintaining that code longer term.
> > 
> > I would expect that this sort of hardware is already quite common
> > amongst Windows 8 tablets so it would be very helpful to have working
> > out-of-the-box on a stock Linux.
> 
> $ cat BYT_LSP_3.11_ISP_2013-12-26.patch | diffstat -s
>  302 files changed, 91662 insertions(+), 2 deletions(-)

That's smaller than the Wi-Fi driver I'm cleaning up ;)

> There's no interest in upstreaming the code on Intel's side. As far as I 
> understand their ISP is more like a programmable DSP than a fixed pipeline 
> ISP. I expect the driver they have published to hardcode the pipeline 
> programmed in a particular firmware and thus be specific to a limited number 
> of devices. Given the amount of work required to get the code in shape, and 
> given that reusability would be very limited if my assumptions are correct, I 
> don't really see this happening unless you can find a motivated developer with 
> way too much free time.

Can somebody explain the different parts of this puzzle, or point me to
a document that would? I understand that there's a piece of silicon on
the SoC that would filter the data from the camera sensor, but I would
expect that to be (somewhat) optional to at least get something going.

Is that not the case?