Re: [PATCH-v3 1/4] idr: Percpu ida

[Kent Overstreet <kmo@daterainc.com>]

On Wed, Aug 21, 2013 at 06:25:58PM +0000, Christoph Lameter wrote:
> On Fri, 16 Aug 2013, Nicholas A. Bellinger wrote:
> 
> > +	spinlock_t			lock;
> 
> Remove the spinlock.

As Andrew noted, the spinlock is needed because of tag stealing. (You
don't think I'd stick a spinlock on a percpu data structure without a
real reason, would you?)

> > +	unsigned			nr_free;
> > +	unsigned			freelist[];
> > +};
> > +
> > +static inline void move_tags(unsigned *dst, unsigned *dst_nr,
> > +			     unsigned *src, unsigned *src_nr,
> > +			     unsigned nr)
> > +{
> > +	*src_nr -= nr;
> > +	memcpy(dst + *dst_nr, src + *src_nr, sizeof(unsigned) * nr);
> > +	*dst_nr += nr;
> > +}
> > +
> 
> > +static inline unsigned alloc_local_tag(struct percpu_ida *pool,
> > +				       struct percpu_ida_cpu *tags)
> 
> Pass the __percpu offset and not the tags pointer.

Why? It just changes where the this_cpu_ptr

> 
> > +{
> > +	int tag = -ENOSPC;
> > +
> > +	spin_lock(&tags->lock);
> 
> Interupts are already disabled. Drop the spinlock.
> 
> > +	if (tags->nr_free)
> > +		tag = tags->freelist[--tags->nr_free];
> 
> You can keep this or avoid address calculation through segment prefixes.
> F.e.
> 
> if (__this_cpu_read(tags->nrfree) {
> 	int n = __this_cpu_dec_return(tags->nr_free);
> 	tag =  __this_cpu_read(tags->freelist[n]);
> }

Can you explain what the point of that change would be? It sounds like
it's preferable to do it that way and avoid this_cpu_ptr() for some
reason, but you're not explaining why.