[v1,2/3] spi: cadence-quadspi: Add clock configuration for StarFive JH7110 QSPI

  Add QSPI clock operation in device probe.

Signed-off-by: William Qiu <william.qiu@starfivetech.com>
Reviewed-by: Hal Feng <hal.feng@starfivetech.com>
---
 drivers/spi/spi-cadence-quadspi.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
  

On Fri, May 26, 2023 at 02:25:28PM +0800, William Qiu wrote:

>  	if (of_device_is_compatible(pdev->dev.of_node, "starfive,jh7110-qspi")) {
> +		qspi_ahb = devm_clk_get(dev, "qspi-ahb");
> +		if (IS_ERR(qspi_ahb)) {
> +			dev_err(dev, "Cannot claim QSPI_AHB clock.\n");
> +			ret = PTR_ERR(qspi_ahb);
> +			return ret;
> +		}
> +
> +		ret = clk_prepare_enable(qspi_ahb);
> +		if (ret) {
> +			dev_err(dev, "Cannot enable QSPI AHB clock.\n");
> +			goto probe_clk_failed;
> +		}

Nothing ever disables or unprepares this clock as far as I can tell?
Perhaps also consider using the clk_bulk_ APIs.
  

On 2023/5/26 23:36, Mark Brown wrote:
> On Fri, May 26, 2023 at 02:25:28PM +0800, William Qiu wrote:
> 
>>  	if (of_device_is_compatible(pdev->dev.of_node, "starfive,jh7110-qspi")) {
>> +		qspi_ahb = devm_clk_get(dev, "qspi-ahb");
>> +		if (IS_ERR(qspi_ahb)) {
>> +			dev_err(dev, "Cannot claim QSPI_AHB clock.\n");
>> +			ret = PTR_ERR(qspi_ahb);
>> +			return ret;
>> +		}
>> +
>> +		ret = clk_prepare_enable(qspi_ahb);
>> +		if (ret) {
>> +			dev_err(dev, "Cannot enable QSPI AHB clock.\n");
>> +			goto probe_clk_failed;
>> +		}
> 
> Nothing ever disables or unprepares this clock as far as I can tell?
> Perhaps also consider using the clk_bulk_ APIs.

I will add in next version.

Thanks for taking time to review this patch series and give useful
suggestions.

Best regards,
William
  

On 2023/5/29 14:44, William Qiu wrote:
> 
> 
> On 2023/5/26 23:36, Mark Brown wrote:
>> On Fri, May 26, 2023 at 02:25:28PM +0800, William Qiu wrote:
>> 
>>>  	if (of_device_is_compatible(pdev->dev.of_node, "starfive,jh7110-qspi")) {
>>> +		qspi_ahb = devm_clk_get(dev, "qspi-ahb");
>>> +		if (IS_ERR(qspi_ahb)) {
>>> +			dev_err(dev, "Cannot claim QSPI_AHB clock.\n");
>>> +			ret = PTR_ERR(qspi_ahb);
>>> +			return ret;
>>> +		}
>>> +
>>> +		ret = clk_prepare_enable(qspi_ahb);
>>> +		if (ret) {
>>> +			dev_err(dev, "Cannot enable QSPI AHB clock.\n");
>>> +			goto probe_clk_failed;
>>> +		}
>> 
>> Nothing ever disables or unprepares this clock as far as I can tell?
>> Perhaps also consider using the clk_bulk_ APIs.
> 
> I will add in next version.
> 
> Thanks for taking time to review this patch series and give useful
> suggestions.
> 
> Best regards,
> William

Hi Mark,

	Now I want to replace the original devm_clk_get API in the
driver with devm_clk_bulk_get_all API, which can achieve compatibility,
but it seems that it is not good for other ip with only one clock, so I
want to ask about that can I replace it? Or define that inside jh7110?

Best regards,
William
  

On Tue, May 30, 2023 at 10:05:38AM +0800, William Qiu wrote:
> On 2023/5/29 14:44, William Qiu wrote:
> > On 2023/5/26 23:36, Mark Brown wrote:

> >> Nothing ever disables or unprepares this clock as far as I can tell?
> >> Perhaps also consider using the clk_bulk_ APIs.

> > I will add in next version.

> 	Now I want to replace the original devm_clk_get API in the
> driver with devm_clk_bulk_get_all API, which can achieve compatibility,
> but it seems that it is not good for other ip with only one clock, so I
> want to ask about that can I replace it? Or define that inside jh7110?

You could always specify a different array of clocks depending on which
compatible the driver sees, just like you'd conditionally request clocks
individually.
  

On 2023/5/30 18:33, Mark Brown wrote:
> On Tue, May 30, 2023 at 10:05:38AM +0800, William Qiu wrote:
>> On 2023/5/29 14:44, William Qiu wrote:
>> > On 2023/5/26 23:36, Mark Brown wrote:
> 
>> >> Nothing ever disables or unprepares this clock as far as I can tell?
>> >> Perhaps also consider using the clk_bulk_ APIs.
> 
>> > I will add in next version.
> 
>> 	Now I want to replace the original devm_clk_get API in the
>> driver with devm_clk_bulk_get_all API, which can achieve compatibility,
>> but it seems that it is not good for other ip with only one clock, so I
>> want to ask about that can I replace it? Or define that inside jh7110?
> 
> You could always specify a different array of clocks depending on which
> compatible the driver sees, just like you'd conditionally request clocks
> individually.
Hi Mark,

	If specify a different array of clocks depending on which compatible
the driver sees, since there will also be clock operations in the suspend
and resume interfaces, this can make the code look complicated.
	My thoughts are as follows:
	Modify the following code

1658	/* Obtain QSPI clock. */
1659	cqspi->clk = devm_clk_get(dev, NULL);
1660	if (IS_ERR(cqspi->clk)) {
1661		dev_err(dev, "Cannot claim QSPI clock.\n");
1662		ret = PTR_ERR(cqspi->clk);
1663		return ret;
1664	}

	as following:

	/* Obtain QSPI clock. */
	cqspi->num_clks = devm_clk_bulk_get_all(dev, &cqspi->clks);
	if (cqspi->num_clks < 0) {
		dev_err(dev, "Cannot claim QSPI clock: %u\n", cqspi->num_clks);
		return -EINVAL;
	}

	This way, the code will look simpler and clearer. How do you think
about it.

Best Regards,
William
  

On Wed, May 31, 2023 at 02:19:16PM +0800, William Qiu wrote:
> On 2023/5/30 18:33, Mark Brown wrote:

> > You could always specify a different array of clocks depending on which
> > compatible the driver sees, just like you'd conditionally request clocks
> > individually.

> 	If specify a different array of clocks depending on which compatible
> the driver sees, since there will also be clock operations in the suspend
> and resume interfaces, this can make the code look complicated.

If you store the clock count and array in the driver data that should be
fairly simple I think.

> 	as following:

> 	/* Obtain QSPI clock. */
> 	cqspi->num_clks = devm_clk_bulk_get_all(dev, &cqspi->clks);
> 	if (cqspi->num_clks < 0) {
> 		dev_err(dev, "Cannot claim QSPI clock: %u\n", cqspi->num_clks);
> 		return -EINVAL;
> 	}

> 	This way, the code will look simpler and clearer. How do you think
> about it.

I'm not clear how enable and disable would then work?
  

On 2023/5/31 21:20, Mark Brown wrote:
> On Wed, May 31, 2023 at 02:19:16PM +0800, William Qiu wrote:
>> On 2023/5/30 18:33, Mark Brown wrote:
> 
>> > You could always specify a different array of clocks depending on which
>> > compatible the driver sees, just like you'd conditionally request clocks
>> > individually.
> 
>> 	If specify a different array of clocks depending on which compatible
>> the driver sees, since there will also be clock operations in the suspend
>> and resume interfaces, this can make the code look complicated.
> 
> If you store the clock count and array in the driver data that should be
> fairly simple I think.
> 
>> 	as following:
> 
>> 	/* Obtain QSPI clock. */
>> 	cqspi->num_clks = devm_clk_bulk_get_all(dev, &cqspi->clks);
>> 	if (cqspi->num_clks < 0) {
>> 		dev_err(dev, "Cannot claim QSPI clock: %u\n", cqspi->num_clks);
>> 		return -EINVAL;
>> 	}
> 
>> 	This way, the code will look simpler and clearer. How do you think
>> about it.
> 
> I'm not clear how enable and disable would then work?

enable use this API:
clk_bulk_prepare_enable(dev->num_clks, dev->clks);

then disable:
clk_bulk_disable_unprepare(dev->num_clks, dev->clks);

But I'll first try specify a different array of clocks depending on which
compatible the driver sees first.

Best regards,
William