Re: [PATCH v2 1/2] spi: spidev: fix a race condition when accessing spidev->spi

[Bartosz Golaszewski <brgl@xxxxxxxx>]

On Fri, Jan 6, 2023 at 3:13 PM Mark Brown <broonie@xxxxxxxxxx> wrote:
>
> On Fri, Jan 06, 2023 at 11:07:18AM +0100, Bartosz Golaszewski wrote:
>
> > - use a mutex instead of an RW semaphore (but for the record: I believe that
> >   the semaphore is the better solution here)
>
> Why?  Like I said in my original reply I'm not clear what the extra
> complication is buying us.

Typically, we'd want to keep locking as fine-grained as possible.
Logically, there's no reason to exclude concurrent execution of
file_operations callbacks. There's a bunch of code in there that could
run at the same time that we're now covering by the mutex' critical
section. We should only be protecting spidev->spi here so any other
locking should be handled elsewhere.

IMO the complication of using an RW semaphore is insignificant and
maybe a comment next to its declaration in struct spidev would
suffice?

Bart