Conversation

Is anybody still working on getting camera to work on ?

1
0
0

@ciopor I'll have another go at it eventually. more or less ran out of things to try

0
0
0

@ciopor @cas I think the drivers for these cameras are hundreds of thousands of lines of code that need to be ported to mainline linux to get anything out of them. And the stack of programs outside of the kernel (libcamera, pipewire) is also pretty incomplete which makes it even harder.

2
0
0

@fell @ciopor the camss driver is complicated sure, but not that bad :P

There's two issues

The first: for the op6 is that the communication between the main sensor and the soc happens using the C-PHY protocol on the wire. This is a much more complicated tri-state logic, rather than the relatively simple inverted pairs of the more typical (but lower bandwidth) D-PHY.

It requires a bunch of changes to the CSIPHY and CSID which aren't supported in upstream.

The SHIFT6mq has the same soc, the same sensors, but uses D-PHY mode. Everything seeems fine, it looks like data is coming in, but we never get the interrupt telling us that the buffer is full, sooo it just hangs

anyways, it's been weeks since i looked into this. just don't have the energy. The pocophone f1 and pixel 3a are looking great

0
1
1
@fell @ciopor @cas Drivers for cameras are normally not _that_ huge. Camera chip is simple device connected over i2c... and we don't have real drivers for most of them, we have just lists "write this to this register". Yes, then there are SoC pieces and ISP chips and ... but parts can be bypassed and done in software. If you are experimenting with this, take a look at megapixels, first. It is simpler than libcamera.
0
0
0