Re: [PATCH v2 1/3] drm: Add SCDC helpers

[Jose Abreu <Jose.Abreu@synopsys.com>]

Hi,


On 05-12-2016 13:31, Ville Syrjälä wrote:
> On Mon, Dec 05, 2016 at 12:16:52PM +0100, Thierry Reding wrote:
>> On Mon, Dec 05, 2016 at 10:12:39AM +0000, Jose Abreu wrote:
>>> On 02-12-2016 19:24, Thierry Reding wrote:
>> [...]
>>>> +/**
>>>> + * drm_scdc_write - write a block of data to SCDC
>>>> + * @adapter: I2C controller
>>>> + * @offset: start offset of block to write
>>>> + * @buffer: block of data to write
>>>> + * @size: size of the block to write
>>>> + *
>>>> + * Writes a block of data to SCDC, starting at a given offset.
>>>> + *
>>>> + * Returns:
>>>> + * The number of bytes written to SCDC or a negative error code on failure.
>>>> + */
>>>> +ssize_t drm_scdc_write(struct i2c_adapter *adapter, u8 offset,
>>>> +		       const void *buffer, size_t size)
>>>> +{
>>>> +	struct i2c_msg msg = {
>>>> +		.addr = SCDC_I2C_SLAVE_ADDRESS,
>>>> +		.flags = 0,
>>>> +		.len = 1 + size,
>>>> +		.buf = NULL,
>>>> +	};
>>>> +	void *data;
>>>> +	int err;
>>>> +
>>>> +	data = kmalloc(1 + size, GFP_TEMPORARY);
>>>> +	if (!data)
>>>> +		return -ENOMEM;
>>>> +
>>>> +	msg.buf = data;
>>>> +
>>>> +	memcpy(data, &offset, sizeof(offset));
>>>> +	memcpy(data + 1, buffer, size);
>>> Don't you agree it would be better if you use the same scheme as
>>> drm_scdc_read()? Something like:
>>>
>>> struct i2c_msg msgs[] = {
>>>     {
>>>         .addr = SCDC_I2C_SLAVE_ADDRESS,
>>>         .flags = 0,
>>>         .len = 1,
>>>         .buf = &offset,
>>>     }, {
>>>         .addr = SCDC_I2C_SLAVE_ADDRESS,
>>>         .flags = 0,
>>>         .len = size,
>>>         .buf = buffer,
>>>     },
>>> };
>> Ville had a similar comment on a prior iteration. It looks as if the
>> above should work, but it's probably best to test it a little more
>> widely to make sure we're not running into cases where it breaks.
> AFAICS the spec says we shouldn't use a repeated start for writes.
> I guess it might work for some devices, but going against the spec
> sounds questionable to me.
>
>> Have you by any chance verified that it works on your hardware?
>>
>> Thierry

Actually, I do not do two writes. My I2C controller is accessible
through the HDMI controller and has direct mapping in the regbank
of the slave address, reg address and data. I am using a I2C
adapter driver but it was modified to when the slave address is
SCDC and we are doing the first write then store the data and
only send it in the next write. Something like this:
http://lxr.free-electrons.com/source/drivers/gpu/drm/nouveau/nvkm/subdev/i2c/anx9805.c#L43
(but modified to SCDC).

Anyway, I was not remembering this so just disregard my comment.
We should follow the spec as Ville said.

>
>

Best regards,
Jose Miguel Abreu
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel