Re: [V3, 4/4] media: platform: dwc: Add MIPI CSI-2 controller driver

[Luis de Oliveira]

On 11-Jan-19 7:25, Eugen.Hristev@xxxxxxxxxxxxx wrote:
> 
> 
> On 10.01.2019 18:18, Luis de Oliveira wrote:
>>
>>
>> On 09-Jan-19 13:07, Eugen.Hristev@xxxxxxxxxxxxx wrote:
>>>
>>>
>>> On 19.10.2018 15:52, Luis Oliveira wrote:
>>>> Add the Synopsys MIPI CSI-2 controller driver. This
>>>> controller driver is divided in platform dependent functions
>>>> and core functions. It also includes a platform for future
>>>> DesignWare drivers.
>>>>
>>>> Signed-off-by: Luis Oliveira <lolivei@xxxxxxxxxxxx>
>>>> ---
>>>> Changelog
>>>> v2-V3
>>>> - exposed IPI settings to userspace
>>>> - fixed headers
>>>>
>>>>    MAINTAINERS                              |  11 +
>>>>    drivers/media/platform/dwc/Kconfig       |  30 +-
>>>>    drivers/media/platform/dwc/Makefile      |   2 +
>>>>    drivers/media/platform/dwc/dw-csi-plat.c | 699 +++++++++++++++++++++++++++++++
>>>>    drivers/media/platform/dwc/dw-csi-plat.h |  77 ++++
>>>>    drivers/media/platform/dwc/dw-mipi-csi.c | 494 ++++++++++++++++++++++
>>>>    drivers/media/platform/dwc/dw-mipi-csi.h | 202 +++++++++
>>>>    include/media/dwc/dw-mipi-csi-pltfrm.h   | 102 +++++
>>>>    8 files changed, 1616 insertions(+), 1 deletion(-)
>>>>    create mode 100644 drivers/media/platform/dwc/dw-csi-plat.c
>>>>    create mode 100644 drivers/media/platform/dwc/dw-csi-plat.h
>>>>    create mode 100644 drivers/media/platform/dwc/dw-mipi-csi.c
>>>>    create mode 100644 drivers/media/platform/dwc/dw-mipi-csi.h
>>>>    create mode 100644 include/media/dwc/dw-mipi-csi-pltfrm.h
>>>>
>>>
>>> [snip]
>>>
>>>> +/* Video formats supported by the MIPI CSI-2 */
>>>> +const struct mipi_fmt dw_mipi_csi_formats[] = {
>>>> +	{
>>>> +		/* RAW 8 */
>>>> +		.code = MEDIA_BUS_FMT_SBGGR8_1X8,
>>>> +		.depth = 8,
>>>> +	},
>>>> +	{
>>>> +		/* RAW 10 */
>>>> +		.code = MEDIA_BUS_FMT_SBGGR10_2X8_PADHI_BE,
>>>> +		.depth = 10,
>>>> +	},
>>>
>>> Hi Luis,
>>>
>>> Any reason why RAW12 format is not handled here ?
>>>
>>> Here, namely MEDIA_BUS_FMT_SBGGR12_1X12 etc.
>>>
>> Hi Eugen,
>>
>> My Hw testing setup currently does not support it, so I didn't add it.
>>
>>>> +	{
>>>> +		/* RGB 565 */
>>>> +		.code = MEDIA_BUS_FMT_RGB565_2X8_BE,
>>>> +		.depth = 16,
>>>> +	},
>>>> +	{
>>>> +		/* BGR 565 */
>>>> +		.code = MEDIA_BUS_FMT_RGB565_2X8_LE,
>>>> +		.depth = 16,
>>>> +	},
>>>> +	{
>>>> +		/* RGB 888 */
>>>> +		.code = MEDIA_BUS_FMT_RGB888_2X12_LE,
>>>> +		.depth = 24,
>>>> +	},
>>>> +	{
>>>> +		/* BGR 888 */
>>>> +		.code = MEDIA_BUS_FMT_RGB888_2X12_BE,
>>>> +		.depth = 24,
>>>> +	},
>>>> +};
>>>
>>> [snip]
>>>
>>>> +
>>>> +void dw_mipi_csi_set_ipi_fmt(struct mipi_csi_dev *csi_dev)
>>>> +{
>>>> +	struct device *dev = csi_dev->dev;
>>>> +
>>>> +	if (csi_dev->ipi_dt)
>>>> +		dw_mipi_csi_write(csi_dev, reg.IPI_DATA_TYPE, csi_dev->ipi_dt);
>>>> +	else {
>>>> +		switch (csi_dev->fmt->code) {
>>>> +		case MEDIA_BUS_FMT_RGB565_2X8_BE:
>>>> +		case MEDIA_BUS_FMT_RGB565_2X8_LE:
>>>> +			dw_mipi_csi_write(csi_dev,
>>>> +					reg.IPI_DATA_TYPE, CSI_2_RGB565);
>>>> +			dev_dbg(dev, "DT: RGB 565");
>>>> +			break;
>>>> +
>>>> +		case MEDIA_BUS_FMT_RGB888_2X12_LE:
>>>> +		case MEDIA_BUS_FMT_RGB888_2X12_BE:
>>>> +			dw_mipi_csi_write(csi_dev,
>>>> +					reg.IPI_DATA_TYPE, CSI_2_RGB888);
>>>> +			dev_dbg(dev, "DT: RGB 888");
>>>> +			break;
>>>> +		case MEDIA_BUS_FMT_SBGGR10_2X8_PADHI_BE:
>>>> +			dw_mipi_csi_write(csi_dev,
>>>> +					reg.IPI_DATA_TYPE, CSI_2_RAW10);
>>>> +			dev_dbg(dev, "DT: RAW 10");
>>>> +			break;
>>>> +		case MEDIA_BUS_FMT_SBGGR8_1X8:
>>>> +			dw_mipi_csi_write(csi_dev,
>>>> +					reg.IPI_DATA_TYPE, CSI_2_RAW8);
>>>> +			dev_dbg(dev, "DT: RAW 8");
>>>> +			break;
>>>
>>> Same here, in CSI_2_RAW12 case it will default to a RGB565 case.
>>>
>>> Thanks,
>>>
>>> Eugen
>>>
>>>
>> I will try to add the support for this data type in my next patchset if not I
>> will flag it as unsupported for now in the commit message and code.
> 
> Hi Luis,
> 
> I am currently trying to see if your driver works , and I need the RAW12 
> type, that's why I am asking about it.
> 
> One question related to the subdevice you create here, how do you 
> register this subdev into the media subsystem ? sync or async, or not at 
> all ?
> After the driver probes, there is no call to the set format functions, I 
> added a node inside the Device tree, I see you are registering media 
> pads, but the other endpoint needs to be an async waiting for completion 
> node or your subdev registers in some other way ? (maybe some link 
> validation required ?)
> 
> Thanks for your help,
> 
> Eugen
> 
Hi Eugen,

On top of this dphy+controller solution I use a wrapper driver that binds this
two together and create the links.
I use V4L2_ASYNC_MATCH_FWNODE and v4l2_async_notifier_operations to match and
bound all my sensors until completion.

My dt looks like this:

camera wrapper {
 	
	video_dev {
		dma
	}
	vif_1 @reg {
		...
	}
	..
	vif_n @reg {
		...
	}
	csi_1 @reg {
		...
		phy
		port 1..2 {}
	}
	csi_n @reg {
		...
		phy
		port 1..2 {}
	}

Fundamentally It is the same principle as this
drivers/media/platform/exynos4-is/media-dev.c but my solution has more entities
for testing purposes.
Check the exynos4-is because is very similar to my top solution.

When I started doing this patchset I was thinking of sending the wrapper also
but I then decided to not do it because it is very narrow and focused in my
tests. But I can include it in v4 if everyone think its best.

Thank you,
Luis
>>
>> Thanks for your review,
>> Luis
>>
>>>
>>>> +		default:
>>>> +			dw_mipi_csi_write(csi_dev,
>>>> +					reg.IPI_DATA_TYPE, CSI_2_RGB565);
>>>> +			dev_dbg(dev, "Error");
>>>> +			break;
>>>> +		}
>>>> +	}
>>>> +}
>>>> +
>>>
>>> [snip]
>>>
>>