Re: [PATCH 01/14] locking/atomic: import atomic_dec_not_zero()

[Peter Zijlstra]

On Tue, Jan 31, 2017 at 06:41:28PM +0100, Fabian Frederick wrote:
> 
> 
> > On 31 January 2017 at 11:41 Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> >
> >
> > On Mon, Jan 30, 2017 at 07:39:38PM +0100, Fabian Frederick wrote:
> > > complementary definition to atomic_inc_not_zero() featured in
> > > lib/fault-inject.c
> >
> > Why?
> 
> Maybe this commit message should be ok ?
> 
> complementary definition to atomic_inc_not_zero() featured in lib/fault-inject.c
> and is more readable than atomic_add_unless((v), -1, 0) used in different
> places.

I still don't see why such a primitive makes sense. Yes there's a few
usage sites, but from them I don't see a sensible pattern.

What sane pattern desires this?