Re: [PATCH] Use deltas to replace atomic inc

[Hugh Dickins <hugh@veritas.com>]

On Mon, 22 Aug 2005, Christoph Lameter wrote:
> On Sun, 21 Aug 2005, Linus Torvalds wrote:
> 
> > Why not make the rule be that we only use the delta stuff when we don't 
> > hold the mm lock. Which is pretty seldom, but the big one is obviously 
> > anon page faults.
> 
> Yes I have tried to follow that. But there are some functions that are 
> called from both contexts. Most importantly this is page_add_anon_rmap.
> If we would remove incrementing anon_rss from the function then we can at 
> avoid the problem for the unuse_mm path. But we would still have the issue 
> with handle_mm_fault. So we would still need some flag.

I'm not following this thread closely yet, concentrating on getting my
alternative page fault scalability ready for you (which will need an
rss solution like yours, when locking across the mm is split: the rules
may be a little different, I'll need to see what you end up with).

(Your deltas seem sensible, but hard to place the reaccumulation:
I worry that you may be taking page_table_lock more just for that.)

Just to say, please do remove the anon_rss incrementation from inside
page_add_anon_rmap if it helps you: I put it there as a lazy way of
avoiding a bigger patch, but it would be much nicer to count separate
file_rss and anon_rss, with /proc showing the sum of the two as rss.

Especially since it's anonymous that most concerns you: the present
setup is biased against anonymous since it's counted in two fields -
perhaps no issue once you've rearranged, but not good for atomics.

If you don't get there first, I do intend in due course to replace
rss and anon_rss by file_rss and anon_rss (and try to sort out the
the tlb mmu gathering stuff, which is a little odd, in dealing with
rss but not with anon_rss).

Hugh
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>