Re: [PATCH 4/5] hugetlb: add per node hstate attributes

[Mel Gorman <mel@csn.ul.ie>]

On Thu, Aug 27, 2009 at 12:52:10PM -0400, Lee Schermerhorn wrote:
> <snip>
> 
> > > @@ -1253,7 +1255,21 @@ static unsigned long set_max_huge_pages(
> > >  	if (h->order >= MAX_ORDER)
> > >  		return h->max_huge_pages;
> > >  
> > > -	nodes_allowed = huge_mpol_nodes_allowed();
> > > +	if (nid == NO_NODEID_SPECIFIED)
> > > +		nodes_allowed = huge_mpol_nodes_allowed();
> > > +	else {
> > > +		/*
> > > +		 * incoming 'count' is for node 'nid' only, so
> > > +		 * adjust count to global, but restrict alloc/free
> > > +		 * to the specified node.
> > > +		 */
> > > +		count += h->nr_huge_pages - h->nr_huge_pages_node[nid];
> > > +		nodes_allowed = alloc_nodemask_of_node(nid);
> > 
> > alloc_nodemask_of_node() isn't defined anywhere.
> 
> 
> Well, that's because the patch that defines it is in a message that I
> meant to send before this one.  I see it's in my Drafts folder.  I'll
> attach that patch below.  I'm rebasing against the 0827 mmotm, and I'll
> resend the rebased series.  However, I wanted to get your opinion of the
> nodemask patch below.
> 

It looks very reasonable to my eye. The caller must know that kfree() is
used to free it instead of free_nodemask_of_node() but it's not worth
getting into a twist over.

> <SNIP>

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab