Re: [PATCH 2/4] media: i2c: Add ST VD55G0 global shutter sensor driver

[George Neacsu]

On Mon, Oct 19, 2026 at 04:27:50PM +0000, Sergey Lebedev wrote:
> The clock tree is computed from the input clock and the one link frequency
> the endpoint gives, as in vd55g1. The MIPI rate must be an integer multiple
> of the input clock, or the sensor would run at another one, so any other
> link frequency fails probe.

Thank you for posting this. I tested v1 on the Snapdragon Surface Pro 11
(X1E80100), where the endpoint is 420 MHz link frequency (840 Mbps) with
a 19.2 MHz xclk.

As posted, patch 2/4 fails probe at the integer-multiple check because
840 MHz is not an integer multiple of 19.2 MHz.

For a test only, I removed that rejection and left the rest of the driver
unchanged. The sensor then probes, the CAMSS media graph completes, and
capture delivers one complete 644x604 Y10 frame. A continuous five-frame
capture then times out waiting for frame 2.

The 1200-pixel line length below 900 Mbps is already present in this
standalone driver, so this is not the earlier line-length issue. I also
tried the ST one-lane OIF_CTRL value 0x81; the behaviour was unchanged.

For comparison, Peter's generalized vd55g1 series, with the 1200-pixel
line-length fix, streams continuously on this same machine with the same
19.2 MHz xclk and nominal 840 Mbps endpoint setting. Whether the sensor
internally rounds that rate is a separate question, but rejecting the
configuration outright prevents this Snapdragon model from probing.

There also appears to be a second difference to narrow down after the
first frame. I'm continuing to bisect that and will report the exact
cause/delta once I have it, so I can't give this a Tested-by yet.

Best regards,
George