Re: [PATCH v3 00/11] CSI2RX support on J721E

From: Tomi Valkeinen
Date: Thu Jul 08 2021 - 04:43:33 EST


On 08/07/2021 11:19, Jacopo Mondi wrote:
Hi Tomi, Pratyush,

On Thu, Jul 08, 2021 at 12:26:38AM +0530, Pratyush Yadav wrote:
Hi Tomi,

Thanks for looking into the patches.

On 01/07/21 10:56AM, Tomi Valkeinen wrote:
Hi Pratyush,

On 24/06/2021 22:21, Pratyush Yadav wrote:
Hi,

This series adds support for CSI2 capture on J721E. It includes some
fixes to the Cadence CSI2RX driver, adds runtime PM support to OV5640
driver, and finally adds the TI CSI2RX wrapper driver.

This series used to include the DPHY and DMA engine patches as well, but
they have been split off to facilitate easier merging. Patch 3 is
build-dependent on the DPHY series [0].

The DMA engine patch [1] can go in any order since that is only a run
time dependency. Things probably won't work without it but it will still
build fine.

Tested on TI's J721E with OV5640 sensor.

I applied these (csi-2 rx, phy, dma-engine) to linux-media/master, and added dts changes to add the csi2-rx. When sending the series, can you also push the branch you use for testing, as the posted patches do not include everything needed?

Please use https://github.com/prati0100/linux-next/ branch "capture"

I will include the link in the cover letter from next version onwards.


Here are some notes from quick tests:

Capture works, but the fps is ~28.98. I would expect it to be closer to 30. Are the clocks configured correctly?

I see this as well. I figured this had something to do with the sensor.

Tomi you might remember your patch to change the h/vtot values which I
collected in a series which I never managed to bring to v1, as Hugues
reported it was broken for JPEG capture.

I'll leave it here just for reference, I admit I dropped the ball
rather quickly there:
https://patchwork.linuxtv.org/project/linux-media/cover/20201028225706.110078-1-jacopo+renesas@xxxxxxxxxx/

I wish I could re-test but seems I've lost the powering cable of the
device I used to test ov5640 :(

Yes, I'm still using my hack patch when working with OV5640. With that hack, on TI platforms with CAL IP, I get ~30fps. With this series on J7, I get the above mentioned 28.98.

It's possible my hack patch is wrong, and CAL driver is buggy, but together they make things right. I guess I should also try J7 without my hack patch.

If I recall right, I tested your changes but I couldn't get them to work on my HW.

I haven't worked on that since then, as I decided that debugging blind is pointless. We need someone to analyze the signals to see what OV5640 is sending. Or some new understanding about the OV5640 HW.

Tomi