drivers: media: cfe: Add non-continuous CSI-2 clock mode

[naushir]

Set the DPHY block to manual clock lane termination enable and HS mode i V4L2_MBUS_CSI2_NONCONTINUOUS_CLOCK is set in the bus flags.

[naushir]

@6by9 very speculative change that sets up the DPHY for non-continuous clock.

[6by9]

Thanks - I'll give it a test with OV9281.

[pelwell]

I would enable warnings-as-errors by default, but some don't like it. It's enabled in the auto-builds after the choice of defconfig by this step:

scripts/config --file $ROOT/.config --set-val CONFIG_WERROR y

where $ROOT may be . in your case.

[naushir]

This change will need adjusting so that we can switch between auto/manual termination since it's possible this is mode specific.

[6by9]

That'll do.

I'd have fixed up the active_lanes value too with

if (!ret) {
   cfe->csi2.dphy.noncontinuous_clock = mbus_config.bus.mipi_csi2.flags & V4L2_MBUS_CSI2_NONCONTINUOUS_CLOCK;
   cfe->csi2.dphy.active_lanes = mbus_config.bus.mipi_csi2.num_data_lanes;
	if (cfe->csi2.dphy.active_lanes > cfe->csi2.dphy.max_lanes) {
		cfe_err("Device has requested %u data lanes, which is >%u configured in DT\n",
			cfe->csi2.dphy.active_lanes, cfe->csi2.dphy.max_lanes);
		ret = -EINVAL;
		goto err_disable_cfe;
}} else {
  cfe->csi2.dphy.noncontinuous_clock = cfe->csi2.bus_flags & V4L2_MBUS_CSI2_NONCONTINUOUS_CLOCK;
   cfe->csi2.dphy.active_lanes = cfe->csi2.dphy.max_lanes;
}

but that can wait for another day.

I still haven't managed to get the failure case to happen that we were discussing this for though....

[naushir]

I still haven't managed to get the failure case to happen that we were discussing this for though....

Even with this commit reverted?

[6by9]

I still haven't managed to get the failure case to happen that we were discussing this for though....

Even with this commit reverted?

Reproduced now without this commit.

6.12 doesn't have #6700 to add the option for selecting clock mode. Adding that makes testing far easier.
Continuous clock mode fails. Non-continuous is OK. That's similar to Pi3 which won't stream in continuous clock mode.

With the latest version of this patch (guess who tried testing the version that only looked at the get_mbus_config result version) I get timeouts in continuous mode, and black images non-continuous.

So we look to be in the right sort of area, but not quite right. I'll give it a few more minutes poking.

[naushir]

It's quite likely the register configuration I've used is also incorrect. The datasheet is very convoluted.

[naushir]

After staring a bit more at the databook, I suspect that while the register configuration may be right, the way we (now) write multiple config register through the test interface is likely wrong.