Re: [RFC][PATCH] cgroup: fix permanent wait in rmdir

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Tue, 23 Jun 2009 13:13:33 +0900
Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp> wrote:

> On Tue, 23 Jun 2009 09:22:23 +0900, KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> > On Mon, 22 Jun 2009 18:37:07 +0900
> > KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> > 
> > > previous discussion was this => http://marc.info/?t=124478543600001&r=1&w=2
> > > 
> > > I think this is a minimum fix (in code size and behavior) and because
> > > we can take a BIG LOCK, this kind of check is necessary, anyway.
> > > Any comments are welcome.
> > 
> > I'll split this into 2 patches...and I found I should check page-migration, too.
> I'll wait a new version, but can you explain in advance this page-migration case ?
> 

Not far from swap-in case.

Assume cgroup "A" which includes file caches. A task in other group mmap file caches
and do page migration and rmdir against "A" is called at the same time.

In mem_cgroup_prepare_migration(), following check is used.

==
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
		mem = pc->mem_cgroup;
		css_get(&mem->css);
	}
	unlock_page_cgroup(pc);
	<======================================(*)
	if (mem) {
		<==============================(**)
		try_charge();
		...
	}
==

At (*), we grab css refcnt which can be under pre_destroy() and
At (**), pre_destroy may returns 0 but charge may be done after the end of pre_destroy().


> > > +static int mem_cgroup_retry_rmdir(struct cgroup_subsys *ss,
> > > +				  struct cgroup *cont)
> > > +{
> > > +	struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
> > > +
> > > +	if (res_counter_read_u64(&memcg->res, RES_USAGE))
> It should be &mem->res.
> 
yes.
too many typos in my patches in these days..

Thanks,
-Kame

--
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>