kbingham: It seems to work for the default configuration at least. It does require CCS static data though. We should probably discuss how to distribute that actually.
Where is the (working) static data currently? 
https://www.retiisi.eu/~sailus/v4l2/tmp/ccs-static-data/
You'll also need these: https://git.retiisi.eu/?p=~sailus/linux.git;a=shortlog;h=refs/heads/ccs-ident .
As well as this: https://git.retiisi.eu/?p=~sailus/linux.git;a=shortlog;h=refs/heads/rpi/v6.13-rc1/overlays .
I have patches for the RPi 4B in a few other branches there if you use that (rpi-metadata and rpi/v6.13-rc1/isp/v2).
I have my imx519 hooked up to an I.mx8mp currently, so I shouldn't need RPI patches.
Thanks, this is on my to-do list to try. I want to understand what we need to add next to libcamera for CCS devices
Does CCS let us know what the actual camera is underneath? I.e. can we still identify it for the correct tuning files?
KieranBingham[m]: That's going to need some work. The CCS driver generally doesn't know the sensor name and without full compatibility there's no guarantee of vendor / device IDs either so another naming scheme (see the CCS patches) is required.
We'll also need to discuss how to convey this to the user space, it's currently not exported. Probably take it into account in device names or expose controls? The entity and sub-device names could well become too long.
KieranBingham[m]: CCS sensors are supposed to report the vendor and model IDs in CCS registers. IIRC, the imx519 leaves those register blank
sailus: is that right ?
pinchartl: At least we can't rely on their contents being allocated from a namespace administered by the vendor, i.e. they may not be unique. AFAIR they're 0's.
Most Sony sensors have the numerical ID of the sensor there but there may be more fine grained model information than that, too.
It was the user space question I was most interested in from libcamera perspective. 
I see you already have DT overrides to specifics which device it is, but it doesn't seem to pass that up to user space yet
Then we'll have to handle any differences in gain model or expose that somehow
Essentially all the information we have in helpers currently in userspace needs to be identifiable or still constructed explicitly on the sensor type
KieranBingham[m]: we'll probably need a CCS camera class
Yes, but that will still need to know about the sensor it's managing.
KieranBingham[m]: I don't disagree. I can post patches for that, most likely next week. I can think of controls exposing module and sensor related information in format that happens to be available, be that either vendor and product IDs or a string.
sailus: i see the v6 of my imx219 series is already picked by media-pending.git. i have already sent a v7, and was planning to send a (hopefully final) v8 as well. 
do you force push to that tree? or should i rebase on top before sending v8?
jluthra: Please rebase further patches on the media commiters' tree.