@mizo Thanks for the PR. What camera do you use to test the code?

@mizo Thanks for the PR. What camera do you use to test the code?

@asmorkalov Thank you for watching.

I use KBCR-S08MM. The host is Armadillo-IoT G4 (SoC: NXP i.MX 8M Plus).

i.MX 8M Plus only supports multi-planar. Therefore, I made it because it needs multi-planar support.

Hello,

is there anything that can be done to help move forward with this PR?
(disclaimer: I'm a coworker of @mizo and did the initial internal review)

@martinetd Thanks for the contribution and sorry for the delay. We are looking how to test the solution on our side, if it's possible. I try v4l2loopback to emulate your camera and have not reached the goal. Please let me know, if you have some solution.

Thanks for the quick reply!

I took a quick look at other devices in the linux tree that have V4L2_CAP_VIDEO_CAPTURE_MPLANE cap, perhaps you might have any of these available somewhere by chance? Our device's driver isn't even mainline yet so that's out of question, but any driver compatible with these should work:

drivers/media/pci/intel/ipu3/ipu3-cio2-main.c
drivers/media/platform/qcom/camss/camss-video.c
drivers/media/platform/renesas/renesas-ceu.c
drivers/media/platform/renesas/vsp1/vsp1_video.c
drivers/media/platform/rockchip/rkisp1/rkisp1-capture.c
drivers/media/platform/samsung/exynos4-is/fimc-isp-video.c
drivers/media/platform/sunxi/sun4i-csi/sun4i_v4l2.c
drivers/staging/media/ipu3/ipu3-v4l2.c

That sounds easier than trying to emulate multiplanar; I gave v4l2loopback a quick look as well but it only supports V4L2_CAP_VIDEO_CAPTURE and nothing in the code hints as mplane support, so while it could also be added you'd be testing two freshly implemented features together and that doesn't really prove anything (if we misunderstood something, we'll probably add the same bug to both sides...)

Ultimately if you cannot test any mplane-compatible device yet, I think what's most important is that we didn't break the existing: have you tested wtih normal capture support yet?
I think it's fine to treat mplane as experimental with the feature available (perhaps gated by a config flag if you want to play safe) while we gather more user input, how does that sound?

@alalek please take a look.

Thanks!