Re: [PATCH v5] HID: corsair-void: Add Corsair Void headset family driver

[Jiri Slaby <jirislaby@kernel.org>]

On 10. 02. 25, 17:15, Stuart wrote:
>> This effectively kills the system. We came here via:
>> corsair_void_raw_event (hid_driver::raw_event)
>>     -> corsair_void_process_receiver
>>       -> scoped_guard(mutex, &drvdata->battery_mutex)
> 
>> And hid_driver::raw_event can be called from the interrupt context. This
>> happened at:
> 
> Damn it, I was suspicious that it was too easy. It's been running for a year
> without an issue on my system, I'll clearly have to read a bit more.
> 
> I'm not sure what the best solution is for this.
> 'corsair_void_process_receiver()' shouldn't be called frequently, and
> the locks shouldn't be waiting on each other very often, so I suppose
> swapping the mutex for a spinlock is probably the best option, instead of
> increasing code complexity?

Not that easy, as AFAIR power unregister cannot sleep.

-- 
js
suse labs