linux-kernel - Re: [patch -mm v2] mm: introduce oom_adj

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.DEB.2.00.0907292356410.5581@chino.kir.corp.google.com>
Date:	Thu, 30 Jul 2009 00:06:38 -0700 (PDT)
From:	David Rientjes <rientjes@...gle.com>
To:	KOSAKI Motohiro <kosaki.motohiro@...fujitsu.com>
cc:	Andrew Morton <akpm@...ux-foundation.org>,
	Rik van Riel <riel@...hat.com>,
	Paul Menage <menage@...gle.com>, linux-kernel@...r.kernel.org,
	linux-mm@...ck.org
Subject: Re: [patch -mm v2] mm: introduce oom_adj_child

On Thu, 30 Jul 2009, KOSAKI Motohiro wrote:

> > diff --git a/kernel/fork.c b/kernel/fork.c
> > --- a/kernel/fork.c
> > +++ b/kernel/fork.c
> > @@ -426,7 +426,7 @@ static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
> >  	init_rwsem(&mm->mmap_sem);
> >  	INIT_LIST_HEAD(&mm->mmlist);
> >  	mm->flags = (current->mm) ? current->mm->flags : default_dump_filter;
> > -	mm->oom_adj = (current->mm) ? current->mm->oom_adj : 0;
> > +	mm->oom_adj = p->oom_adj_child;
> 
> This code doesn't fix anything.
> mm->oom_adj assignment still change vfork() parent process oom_adj value.
> (Again, vfork() parent and child use the same mm)
> 

That's because the oom killer only really considers the highest oom_adj 
value amongst all threads that share the same mm.  Allowing those threads 
to each have different oom_adj values leads (i) to an inconsistency in 
reporting /proc/pid/oom_score for how the oom killer selects a task to 
kill and (ii) the oom killer livelock that it fixes when one thread 
happens to be OOM_DISABLE.

So, yes, changing the oom_adj value for a thread may have side-effects 
on other threads that didn't exist prior to 2.6.31-rc1 because the oom_adj 
value now represents a killable quantity of memory instead of a being a 
characteristic of the task itself.  But we now provide the inheritance 
property in a new way, via /proc/pid/oom_adj_child, that gives you all the 
functionality that the previous way did but without the potential for 
livelock.

> IOW, in vfork case, oom_adj_child parameter doesn't only change child oom_adj,
> but also parent oom_adj value.

Changing oom_adj_child for a task never changes oom_adj for any mm, it 
simply specifies what default value shall be given for a child's newly 
initialized mm.  Chaning oom_adj, on the other hand, will 

> IOW, oom_adj_child is NOT child effective parameter.
> 

It's not meant to be, it's only meant to specify a default value for newly 
initialized mm's of its descendants.  What happens after that is governed 
completely by the child's own /proc/pid/oom_adj.  That's pretty clearly 
explained in Documentation/filesystems/proc.txt.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/