Messages in this thread | | | From | KOSAKI Motohiro <> | Subject | Re: [PATCH 6/9] readahead: clean up and simplify the code for filemap page fault readahead | Date | Mon, 13 Apr 2009 08:16:54 +0900 (JST) |
| |
> On Sat, 11 Apr 2009 22:58:31 +0900 KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> wrote: > > > > Ah. __But I have a note here that I didn't like it, because it adds lots > > > of new spinlocking to fastpaths. __So I'll leave things as they stand > > > until we have had a little talk about that. > > > > add? > > > > old code: grab zone->lru_lock via mark_page_accessed() > > new code: grab zone->lru_lock via update_reclaim_stat() > > > > one remove, one add. > > > > mm-update_page_reclaim_stat-is-called-from-page-fault-path.patch adds > new calls to update_page_reclaim_stat() into do_swap_page() and > filemap_fault(). update_page_reclaim_stat() does spin_lock_irq() and > spin_unlock_irq(). It looks like a net slowdown to me.
Ah, I compared with the code before Nick's mark_page_accessed() removing. but you don't.
but I agree this patch reduce the worth of nick's work. I have to improve more.
> > But I agree its lock can be removed maybe.. > > It would be nice to try to do something about it - every little bit > counts.
Yes. I can't opoose it.
| |