[PATCH] media: imx-mipi-csis: avoid logging while holding spinlock

[Matthias Fend]

Hi Laurent,

Am 26.07.2024 um 12:42 schrieb Laurent Pinchart:
> Hi Matthias,
> 
> Thank you for the patch.
> 
> On Tue, Jul 23, 2024 at 08:09:08AM +0200, Matthias Fend wrote:
>> Refactor mipi_csis_log_counters() to prevent calling dev_info() while
>> IRQs are disabled. This reduces crucial IRQs off time to a bare minimum.
> 
> Should we add
> 
> Cc: stable at vger.kernel.org
> 
> to get this backported to stable kernels ?


Hard for me to tell. Since this problem only occurs when you explicitly 
trigger the debug output from this driver, it probably only affects very 
few end users. So I'm not sure if a backport would be worth it.

> 
>> Signed-off-by: Matthias Fend <matthias.fend at emfend.at>
>> ---
>>   drivers/media/platform/nxp/imx-mipi-csis.c | 9 ++++++---
>>   1 file changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/media/platform/nxp/imx-mipi-csis.c b/drivers/media/platform/nxp/imx-mipi-csis.c
>> index f49b06978f14..0c34d316ed29 100644
>> --- a/drivers/media/platform/nxp/imx-mipi-csis.c
>> +++ b/drivers/media/platform/nxp/imx-mipi-csis.c
>> @@ -857,18 +857,21 @@ static void mipi_csis_log_counters(struct mipi_csis_device *csis, bool non_error
>>   {
>>   	unsigned int num_events = non_errors ? MIPI_CSIS_NUM_EVENTS
>>   				: MIPI_CSIS_NUM_EVENTS - 8;
>> +	unsigned int counters[MIPI_CSIS_NUM_EVENTS];
>>   	unsigned long flags;
>>   	unsigned int i;
>>   
>>   	spin_lock_irqsave(&csis->slock, flags);
>> +	for (i = 0; i < num_events; ++i)
>> +		counters[i] =  csis->events[i].counter;
>> +	spin_unlock_irqrestore(&csis->slock, flags);
>>   
>>   	for (i = 0; i < num_events; ++i) {
>> -		if (csis->events[i].counter > 0 || csis->debug.enable)
>> +		if (counters[i] > 0 || csis->debug.enable)
>>   			dev_info(csis->dev, "%s events: %d\n",
>>   				 csis->events[i].name,
>> -				 csis->events[i].counter);
>> +				 counters[i]);
> 
> The last two lines now fit in a single line. No need for a v2, I'll
> rewrap the code myself, and add the Cc line if you think that's a good
> idea.

Thanks! Regarding Cc, I would like to leave it up to you; you have the 
necessary experience and overview to judge that.

~Matthias

> 
> Reviewed-by: Laurent Pinchart <laurent.pinchart at ideasonboard.com>
> 
>>   	}
>> -	spin_unlock_irqrestore(&csis->slock, flags);
>>   }
>>   
>>   static int mipi_csis_dump_regs(struct mipi_csis_device *csis)
>