Hi,

On Wed, Apr 27, 2016 at 09:57:51AM +0300, Ivaylo Dimitrov wrote:
> >>https://git.kernel.org/cgit/linux/kernel/git/sre/linux-n900.git/log/?h=n900-camera-ivo
> >
> >Ok, going to diff with my tree to see what I have missed to send in the
> >patchset
> 
> Now, that's getting weird.

[...]

> I you want to try it, zImage and initrd are on
> http://46.249.74.23/linux/camera-n900/

The zImage + initrd works with the steps you described below. I
received a completly black image, but at least there are interrupts
and yavta is happy (=> it does not hang).

For reference I configured the pipeline using media-ctl and used
this to acquire the image:

./yavta --capture=1 --format UYVY --file="/tmp/frame.uyvy" --size 656x488 /dev/video6

Then I copied the file to my notebook using the FTP server coming
with your initrd and displayed it using "display" from imagemagick:

display -size 656x488 -colorspace rgb frame.uyvy

Before analysing why there is only a black image let's start
with the no image at all problem, though.

> I cloned n900-camera-ivo, copied rx51_defconfig from my tree, added:
> 
> CONFIG_VIDEO_SMIAREGS=m
> CONFIG_VIDEO_ET8EK8=m
> CONFIG_VIDEO_BUS_SWITCH=m
> 
> to it, make mrproper, built the kernel using rx51_defconfig and made initrd
> for rescueos, so to be sure that maemo5 did not influence cameras somehow.

Ok, so there is probably a problem when some things are not built as
modules.

> I cloned n900-camera-ivo, copied rx51_defconfig from my tree, added:
> 
> CONFIG_VIDEO_SMIAREGS=m
> CONFIG_VIDEO_ET8EK8=m
> CONFIG_VIDEO_BUS_SWITCH=m
> 
> to it, make mrproper, built the kernel using rx51_defconfig and made initrd
> for rescueos, so to be sure that maemo5 did not influence cameras somehow.

I will test your kernel + your modules with my userspace, so
that I know for sure, that my userland behaves correctly.

Can you try if your config still works if you configure
CONFIG_VIDEO_OMAP3=y, but leaving the sensors configured
as modules? I will try the reverse process (using my config
and moving config options to =m).

> ~$ modprobe smiapp

modprobing smiapp resulted in a kernel message about a missing
symbol btw. I currently don't remember which one and it's no
longer in dmesg due to ISP debug messages.

> Please, Sebastian and Pavel, make sure you're not using some development
> devices, old board versions need VAUX3 enabled as well, and this is not
> supported in the $subject patchset. I guess you may try to make VAUX3
> always-on in board DTS if that's the case, but I've never tested that, my
> device is a production one.

I don't have pre-production N900s. The phone I use for development
is HW revision 2101 with Finish keyboard layout. Apart from that
I have my productive phone, which is rev 2204 with German layout.

-- Sebastian