Re: [PATCH 1/4] memcg: simplify page cache charging.

[KAMEZAWA Hiroyuki]

On Tue, 20 Dec 2011 13:58:17 -0800
Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:

> On Mon, 19 Dec 2011 09:01:22 +0900
> KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> 
> > On Fri, 16 Dec 2011 14:28:14 -0800
> > Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> > 
> > > On Wed, 14 Dec 2011 16:49:22 +0900
> > > KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> > > 
> > > > Because of commit ef6a3c6311, FUSE uses replace_page_cache() instead
> > > > of add_to_page_cache(). Then, mem_cgroup_cache_charge() is not
> > > > called against FUSE's pages from splice.
> > > 
> > > Speaking of ef6a3c6311 ("mm: add replace_page_cache_page() function"),
> > > may I pathetically remind people that it's rather inefficient?
> > > 
> > > http://lkml.indiana.edu/hypermail/linux/kernel/1109.1/00375.html
> > > 
> > 
> > IIRC, people says inefficient because it uses memcg codes for page-migration
> > for fixing up accounting. Now, We added replace-page-cache for memcg in
> > memcg-add-mem_cgroup_replace_page_cache-to-fix-lru-issue.patch
> > 
> > So, I think the problem originally mentioned is fixed.
> > 
> 
> No, the inefficiency in replace_page_cache_page() is still there.  Two
> identical walks down the radix tree, a pointless decrement then
> increment of mapping->nrpages, two writes to page->mapping, an often
> pointless decrement then increment of NR_FILE_PAGES, and probably other things.
> 

Hmm, then, replace_page_cache_page() itself has some problem.
I'll look into that.

Thanks,
-Kame




--
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/