Re: [PATCH 3/5] [media] tc358743: support probe from device tree

[Hans Verkuil <hverkuil@xs4all.nl>]

On 07/14/15 12:10, Philipp Zabel wrote:
> Hi Hans,
> 
> Am Montag, den 13.07.2015, 12:57 +0200 schrieb Hans Verkuil:
> [...]
>>> @@ -69,6 +72,7 @@ static const struct v4l2_dv_timings_cap tc358743_timings_cap = {
>>>  
>>>  struct tc358743_state {
>>>  	struct tc358743_platform_data pdata;
>>> +	struct v4l2_of_bus_mipi_csi2 bus;
>>
>> Where is this bus struct set?
> 
> Uh-oh, I have accidentally dropped setting the bus struct when switching
> to v4l2_of_alloc_parse_endpoint.
> 
> [...]
>>> @@ -700,7 +706,8 @@ static void tc358743_set_csi(struct v4l2_subdev *sd)
>>>  			((lanes > 2) ? MASK_D2M_HSTXVREGEN : 0x0) |
>>>  			((lanes > 3) ? MASK_D3M_HSTXVREGEN : 0x0));
>>>  
>>> -	i2c_wr32(sd, TXOPTIONCNTRL, MASK_CONTCLKMODE);
>>> +	i2c_wr32(sd, TXOPTIONCNTRL, (state->bus.flags &
>>> +		 V4L2_MBUS_CSI2_CONTINUOUS_CLOCK) ? MASK_CONTCLKMODE : 0);
>>
>> It's used here.
>>
>> BTW, since I don't see state->bus being set, that means bus.flags == 0 and
>> so this register is now set to 0 instead of MASK_CONTCLKMODE.
>>
>> When using platform data I guess bus.flags should be set to
>> V4L2_MBUS_CSI2_CONTINUOUS_CLOCK to prevent breakage.
> 
> Ok.
> 
> [..]
>>> +	/*
>>> +	 * The CSI bps per lane must be between 62.5 Mbps and 1 Gbps.
>>> +	 * The default is 594 Mbps for 4-lane 1080p60 or 2-lane 720p60.
>>> +	 */
>>> +	bps_pr_lane = 2 * endpoint->link_frequencies[0];
>>> +	if (bps_pr_lane < 62500000U || bps_pr_lane > 1000000000U) {
>>> +		dev_err(dev, "unsupported bps per lane: %u bps\n", bps_pr_lane);
>>> +		goto disable_clk;
>>> +	}
>>> +
>>> +	/* The CSI speed per lane is refclk / pll_prd * pll_fbd */
>>> +	state->pdata.pll_fbd = bps_pr_lane /
>>> +			       state->pdata.refclk_hz * state->pdata.pll_prd;
>>> +
>>> +	/*
>>> +	 * FIXME: These timings are from REF_02 for 594 Mbps per lane (297 MHz
>>> +	 * link frequency). In principle it should be possible to calculate
>>> +	 * them based on link frequency and resolution.
>>> +	 */
>>> +	if (bps_pr_lane != 594000000U)
>>> +		dev_warn(dev, "untested bps per lane: %u bps\n", bps_pr_lane);
>>> +	state->pdata.lineinitcnt = 0xe80;
>>> +	state->pdata.lptxtimecnt = 0x003;
>>> +	/* tclk-preparecnt: 3, tclk-zerocnt: 20 */
>>> +	state->pdata.tclk_headercnt = 0x1403;
>>> +	state->pdata.tclk_trailcnt = 0x00;
>>> +	/* ths-preparecnt: 3, ths-zerocnt: 1 */
>>> +	state->pdata.ths_headercnt = 0x0103;
>>> +	state->pdata.twakeup = 0x4882;
>>> +	state->pdata.tclk_postcnt = 0x008;
>>> +	state->pdata.ths_trailcnt = 0x2;
>>> +	state->pdata.hstxvregcnt = 0;
> 
> Do you have any suggestion how to handle this? AFAIK REF_02 is not
> public, and I do not know the formulas it uses internally to calculate
> these timings. I wouldn't want to add all the timing parameters to the
> device tree just because of that.

As you said, it's not public and without the formulas there is nothing you
can do but hardcode it.

If I understand this correctly these values depend on the link frequency,
so the DT should contain the link frequency and the driver can hardcode the
values based on that. Which means that if someone needs to support a new
link frequency the driver needs to be extended for that frequency.

As long as Toshiba keeps the formulas under NDA there isn't much else you can
do.

Regards,

	Hans

> 
> [...]
>>> @@ -1658,14 +1794,19 @@ static int tc358743_probe(struct i2c_client *client,
>>>  	if (!state)
>>>  		return -ENOMEM;
>>>  
>>> +	state->i2c_client = client;
>>> +
>>>  	/* platform data */
>>> -	if (!pdata) {
>>> -		v4l_err(client, "No platform data!\n");
>>> -		return -ENODEV;
>>> +	if (pdata) {
>>> +		state->pdata = *pdata;
>>> +	} else {
>>> +		err = tc358743_probe_of(state);
>>> +		if (err == -ENODEV)
>>> +			v4l_err(client, "No platform data!\n");
>>
>> I'd replace this with "No device tree data!" or something like that.
> 
> I'll do that, thank you.
> 
> regards
> Philipp
>