Re: [PATCH 03/12] Export unusable free space index via/proc/pagetypeinfo

[Mel Gorman]

On Tue, Feb 16, 2010 at 05:41:39PM +0900, KOSAKI Motohiro wrote:
> > On Tue, Feb 16, 2010 at 04:03:29PM +0900, KOSAKI Motohiro wrote:
> > > > Unusuable free space index is a measure of external fragmentation that
> > > > takes the allocation size into account. For the most part, the huge page
> > > > size will be the size of interest but not necessarily so it is exported
> > > > on a per-order and per-zone basis via /proc/pagetypeinfo.
> > > 
> > > Hmmm..
> > > /proc/pagetype have a machine unfriendly format. perhaps, some user have own ugly
> > > /proc/pagetype parser. It have a little risk to break userland ABI.
> > > 
> > 
> > It's very low risk. I doubt there are machine parsers of
> > /proc/pagetypeinfo because there are very few machine-orientated actions
> > that can be taken based on the information. It's more informational for
> > a user if they were investigating fragmentation problems.
> > 
> > > I have dumb question. Why can't we use another file?
> > 
> > I could. What do you suggest?
> 
> I agree it's low risk. but personally I hope fragmentation ABI keep very stable because
> I expect some person makes userland compaction daemon. (read fragmentation index
> from /proc and write /proc/compact_memory if necessary).
> then, if possible, I hope fragmentation info have individual /proc file.
> 

I'd be somewhat surprised if there was an active userland compaction daemon
because I'd expect them to be depending on direct compaction.  Userspace
compaction is more likely to be an all-or-nothing affair and confined to
NUMA nodes if they are being used as containers. If a compaction daemon was
to exist, I'd have expected it to be in-kernel because the triggers from
userspace are so coarse.

Still, I can break out the indices into separate files to cover all the
bases.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/