Re: [PATCH] spi: Use non-atomic xxx_bit() functions

[Christophe JAILLET]

Le 30/04/2023 à 17:49, Mark Brown a écrit :
> On Sun, Apr 30, 2023 at 11:35:35AM +0200, Christophe JAILLET wrote:
>
> > Accesses to 'minors' are guarded by the 'device_list_lock' mutex. So, it is
> > safe to use the non-atomic version of (set|clear)_bit() in the
> > corresponding sections.
>
> Is it a problem to use the atomic version?

Not at all. It just wastes a few cycles (in a place where it doesn't 
matter).

I spotted it while looking for some other patterns, so I sent a patch 
for it.

> >   	if (status == 0) {
> > -		set_bit(minor, minors);
> > +		__set_bit(minor, minors);
> >   		list_add(&spidev->device_entry, &device_list);
>
> The __ usually means something is the more complicated and less
> preferred API.

Ok, let keep things as-is and simple then.
Performance doesn't matter here, anyway.

CJ