Re: [PATCH 1/4]percpu_counter: make API return consistent value

[Shaohua Li <shaohua.li@intel.com>]

On Wed, 2011-04-13 at 02:49 +0800, Tejun Heo wrote:
> Hello,
> 
> First of all, please somehow link patches of the same series.  Either
> write a head message and make all the patches replies to it
> (preferred) or chain reply the patches (only when the number of
> patches is small).
> 
> On Tue, Apr 12, 2011 at 04:03:57PM +0800, Shaohua Li wrote:
> > the percpu_counter_*_positive() API SMP and !SMP aren't consistent. From
> > the API name, we should return a non-negative value for them.
> > Also if count < 0, returns 0 instead of 1 for *read_positive().
> 
> Ummm, on UP, the counters cannot be positive.
s/positive/negative?

>   The _positive interface
> is there to make it easier to cope with deviations introduced by
> unsynchronized modifications by different CPUs.  On UP, such
> deviations don't happen at all so _positive interface is the same as
> the counterpart without the postfix.
I'm confused. the counter could be negative, we have *_dec, *_sub.

Thanks,
Shaohua