Re: [PATCH 2/3] spi: s3c64xx: fix runtime PM handling

[Heiner Kallweit <hkallweit1@xxxxxxxxx>]

Am 20.08.2015 um 01:41 schrieb Krzysztof Kozlowski:
> On 20.08.2015 04:36, Heiner Kallweit wrote:
>> Am 19.08.2015 um 07:54 schrieb Heiner Kallweit:
>>> Am 19.08.2015 um 02:35 schrieb Krzysztof Kozlowski:
>>>> On 19.08.2015 06:06, Heiner Kallweit wrote:
>>>>> The patch fixes / extends the runtime PM handling.
>>>>> - Currently the driver instructs the SPI core to use auto-suspend
>>>>>   but doesn't configure it. Add the necessary calls and let it
>>>>>   autosuspend after probe.
>>>>
>>>> After quick look at the driver I think it does not use autosuspend
>>>> fully. It just sets master->auto_runtime_pm (so the core code will use
>>>> autosuspend) but its own calls to pm_runtime_put() are regular ones.
>>>> Shouldn't this be enhanced as well?
>>> Right, the pm_runtime_put calls in s3c64xx_spi_setup should be
>>> replaced as well. Will provide an updated patch.
>>>>
>>>>> - In case the driver registration fails runtime PM needs to be
>>>>>   properly disabled.
>>>>>
>>>>> Signed-off-by: Heiner Kallweit <hkallweit1@xxxxxxxxx>
>>>>> ---
>>>>>  drivers/spi/spi-s3c64xx.c | 11 ++++++++++-
>>>>>  1 file changed, 10 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/spi/spi-s3c64xx.c b/drivers/spi/spi-s3c64xx.c
>>>>> index c4d3b06..15a6de1 100644
>>>>> --- a/drivers/spi/spi-s3c64xx.c
>>>>> +++ b/drivers/spi/spi-s3c64xx.c
>>>>> @@ -32,6 +32,7 @@
>>>>>  #define MAX_SPI_PORTS		6
>>>>>  #define S3C64XX_SPI_QUIRK_POLL		(1 << 0)
>>>>>  #define S3C64XX_SPI_QUIRK_CS_AUTO	(1 << 1)
>>>>> +#define AUTOSUSPEND_TIMEOUT     2000
>>>>>  
>>>>>  /* Registers and bit-fields */
>>>>>  
>>>>> @@ -1180,13 +1181,15 @@ static int s3c64xx_spi_probe(struct platform_device *pdev)
>>>>>  	       S3C64XX_SPI_INT_TX_OVERRUN_EN | S3C64XX_SPI_INT_TX_UNDERRUN_EN,
>>>>>  	       sdd->regs + S3C64XX_SPI_INT_EN);
>>>>>  
>>>>> +	pm_runtime_set_autosuspend_delay(dev, AUTOSUSPEND_TIMEOUT);
>>>>> +	pm_runtime_use_autosuspend(dev);
>>>>>  	pm_runtime_set_active(dev);
>>>>>  	pm_runtime_enable(dev);
>>>>>  
>>>>>  	ret = devm_spi_register_master(dev, master);
>>>>>  	if (ret != 0) {
>>>>>  		dev_err(dev, "cannot register SPI master: %d\n", ret);
>>>>> -		goto err3;
>>>>> +		goto err4;
>>>>>  	}
>>>>>  
>>>>>  	dev_dbg(dev, "Samsung SoC SPI Driver loaded for Bus SPI-%d with %d Slaves attached\n",
>>>>> @@ -1195,8 +1198,14 @@ static int s3c64xx_spi_probe(struct platform_device *pdev)
>>>>>  					mem_res, (FIFO_LVL_MASK(sdd) >> 1) + 1,
>>>>>  					sdd->rx_dma.dmach, sdd->tx_dma.dmach);
>>>>>  
>>>>> +	pm_runtime_mark_last_busy(dev);
>>>>> +	pm_runtime_autosuspend(dev);
>>>>
>>>> There is a high chance that device will be suspended already at this
>>>> point because pm_runtime_put() in setup. In most cases (when SPI is
>>>> started) this won't do anything so we are back at first of my questions
>>>> - shouldn't you update pm_runtime_put() calls?
>>> Yes, these calls should be updated.
>>>>
>>>> Additionally (which is not related to your patch) shouldn't we
>>>> pm_runtime_get() at some point in probe() to be sure that it won't be
>>>> suspended to early?
>>> Not 100% sure at this moment, I will check.
>> Kernel documentation says:
>> "Note, if the device may execute pm_runtime calls during the probe (such as
>> if it is registers with a subsystem that may call back in) then the
>> pm_runtime_get_sync() call paired with a pm_runtime_put() call will be
>> appropriate to ensure that the device is not put back to sleep during the
>> probe. This can happen with systems such as the network device layer."
>>
>> I don't think this applies to SPI but I'm not expert enough to guarantee.
>> At least no other SPI driver using runtime PM protects the probe function this way.
>> Therefore I would dare to leave it as it is, especially as there don't seem
>> to be any related known problems with SPI drivers.
> 
> But you know that these runtime callbacks will be executed anyway during
> probe because what I mentioned earlier - setup calls get and put.
> Additionally pm_runtime_enable() is called in the middle of probe, not
> at the end. This means that already it is assumed probe in certain time
> can handle runtime callbacks.
> 
> I think it is good opportunity to fix any remaining issues with runtime
> PM in this driver.
> 
> Best regards,
> Krzysztof
OK. Will include this and resend. Thanks for the review.

Regards, Heiner

--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html