Re: [PATCH 2/4] soc: qcom: icc-bwmon: Allow for interrupts to be shared across instances

[Sibi Sankar <quic_sibis@quicinc.com>]

On 6/15/24 03:12, Dmitry Baryshkov wrote:
> On Sat, Jun 15, 2024 at 01:49:34AM GMT, Sibi Sankar wrote:
>>
>>
>> On 6/14/24 13:54, Krzysztof Kozlowski wrote:
>>> On 13/06/2024 19:02, Sibi Sankar wrote:
>>>>
>>>>
>>>> On 6/4/24 12:16, Krzysztof Kozlowski wrote:
>>>>> On 04/06/2024 03:11, Sibi Sankar wrote:
>>>>>> The multiple BWMONv4 instances available on the X1E80100 SoC use the
>>>>>> same interrupt number. Mark them are shared to allow for re-use across
>>>>>> instances.
>>>>
>>>> Hey Krzysztof,
>>>>
>>>> Thanks for taking time to review the series :)
>>>>
>>>>>
>>>>> Would be nice if you also mention you checked that it is safe to have
>>>>> both devm and shared interrupts (so you investigated possibility of race
>>>>> on exit path).
>>>>
>>>> I didn't see any problems with devm being used with SHARED when I posted
>>>> it out. After your review comments I went back again to vett the exit
>>>> path for races and ran into an pre-existing splat [1] but the bwmon
>>>> instances work as expected on module removal/re-insertion.
>>>
>>> Using devm and shared interrupts is in general sign of possible race
>>> issues and should be avoided. Just "not seeing problems" is not an
>>> argument for me, to be honest.
>>
>> Didn't I go further and say I got it tested though? Also can you
>> elaborate on what race do you think the bwmon will hit rather than
>> being too generic about it?
> 
> devm_request_threaded_irq means that the IRQ is freed after the
> bwmon_remove() function returns. Having IRQF_SHARED means that the IRQ
> can still be triggered even though IRQ for this device has been disabled
> in bwmon_disable().
> 
> In this particular case such IRQ probably won't cause issues, but at
> least it needs to be validated and probably commented in bwmon_remove().
> Just stating that "you tested and had no problems" usually isn't enough
> for the expected race condition issues.

Cool, thanks for the info. I'll get this fixed in the next re-spin.

-Sibi

>