Re: [PATCHv2] mtd: spi-nor: cadence-quadspi: add reset control

[Dinh Nguyen <dinguyen@kernel.org>]

On 4/17/19 3:48 AM, Tudor.Ambarus@microchip.com wrote:
> Hi, Dinh,
> 
> On 04/09/2019 06:38 PM, Dinh Nguyen wrote:
>> Get the reset control for the QSPI controller and bring it out of reset.
> 
> Is there a public datasheet where I can check this?

You can look at the reset manager bits here:

https://www.intel.com/content/www/us/en/programmable/hps/arria-10/hps.html#reg_soc_top/sfo1429890575955.html

Otherwise, the QSPI block is just the standard IP from Cadence.

>>
>> Suggested-by: Tien-Fong Chee <tien.fong.chee@intel.com>
>> Signed-off-by: Dinh Nguyen <dinguyen@kernel.org>
>> ---
>> v2: use devm_reset_control_get_optional_exclusive
>>     print an error message
>>     return -EPROBE_DEFER
>> ---
>>  drivers/mtd/spi-nor/cadence-quadspi.c | 14 ++++++++++++++
> 
> would you update the bindings as well?

Yes, I will do that.


> 
>>  1 file changed, 14 insertions(+)
>>
>> diff --git a/drivers/mtd/spi-nor/cadence-quadspi.c b/drivers/mtd/spi-nor/cadence-quadspi.c
>> index 792628750eec..c548567adcf0 100644
>> --- a/drivers/mtd/spi-nor/cadence-quadspi.c
>> +++ b/drivers/mtd/spi-nor/cadence-quadspi.c
>> @@ -34,6 +34,7 @@
>>  #include <linux/of.h>
>>  #include <linux/platform_device.h>
>>  #include <linux/pm_runtime.h>
>> +#include <linux/reset.h>
>>  #include <linux/sched.h>
>>  #include <linux/spi/spi.h>
>>  #include <linux/timer.h>
>> @@ -1336,6 +1337,7 @@ static int cqspi_probe(struct platform_device *pdev)
>>  	struct cqspi_st *cqspi;
>>  	struct resource *res;
>>  	struct resource *res_ahb;
>> +	struct reset_control *rstc;
>>  	const struct cqspi_driver_platdata *ddata;
>>  	int ret;
>>  	int irq;
>> @@ -1362,6 +1364,18 @@ static int cqspi_probe(struct platform_device *pdev)
>>  		return PTR_ERR(cqspi->clk);
>>  	}
>>  
>> +	/* Obtain QSPI reset control */
> 
> OSPI version of the IP has been added recently. Is this available just for QSPI?
> 
>> +	rstc = devm_reset_control_get_optional_exclusive(dev, NULL);
>> +	if (IS_ERR(rstc)) {
>> +		dev_err(dev, "Cannot get QSPI reset.\n");
>> +		if (PTR_ERR(rstc) == -EPROBE_DEFER)
>> +			return -EPROBE_DEFER;
> 
> why ignoring all the other possible errors? Maybe return PTR_ERR(rstc); ?

Ok..

> 
>> +	} else {
>> +		reset_control_assert(rstc);
> 
> Shouldn't the clock be enabled before asserting reset?
> 

Yes, it should. Thanks for catching that!

> Does it matter if reset_control_assert() is returning any error?
> 
>> +		udelay(1);
> 
> Is there a range or 1 us is the maximum time that we should wait?
> 
>> +		reset_control_deassert(rstc);

I'll add a 1-2 us range.

> 
> Should we check if reset_control_deassert() is returning any error?
> 

I don't think there is a need for error checking here. We've already
made sure that the reset property is there.

Thanks for the review...

Dinh