Re: [RFC][PATCH 4/4] memcg: make oom less frequently

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

On Fri, 9 Jan 2009 11:29:22 +0900
Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp> wrote:

> > > @@ -870,8 +870,13 @@ static int __mem_cgroup_try_charge(struct mm_struct *mm,
> > >  		if (!(gfp_mask & __GFP_WAIT))
> > >  			goto nomem;
> > >  
> > > +		if (signal_pending(current))
> > > +			goto oom;
> > > +
> > 
> > I think it's better to avoid to add this check *now*. and "signal is pending" 
> > doesn't mean oom situation.
> > 
> hmm.. charge is assumed to return 0 or -ENOMEM, what should we return on
> signal_pending case ?
> 
> In case of shmem for example, if charge at shmem_getpage fails by -ENOMEM, 
> shmem_fault returns VM_FAULT_OOM, so pagefault_out_of_memory would be called.
> If memcg had not invoked oom-killer, system wide oom would be invoked.
> 
yes, that's problem.

I think generic -EAGAIN support is appreciated. But it will not be for -rc ;)
(... that will make codes other than memcontrol.c more complicated.)

Thanks,
-Kame

> > Hmm..Maybe we can tell "please retry page fault again, it's too long latency in
> > memory reclaim and you received signal." in future.
> > 
> OK.
> 
> > IMHO, only quick path which we can add here now is
> > ==
> > 	if (test_thread_flag(TIG_MEMDIE)) { /* This thread is killed by OOM */
> > 		*memcg = NULL;
> > 		return 0;
> > 	}
> > ==
> > like this.
> > 
> > Anyway, please discuss this "quick exit path" in other patch and just remove 
> > siginal check.
> > 
> > Other part looks ok to me.
> > 
> Thanks :)
> 
> I'll update this one by removing the signal_pendign check.
> 
> 
> Thanks,
> Daisuke Nishimura.
> 

--
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>