linux-kernel - Re: [PATCH 2/4] memcg: separate local

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: <s636pqlp5tfg6p2vt3argmqyysx5d72jtwjpekk5nj7yerbolf@vco5fbwicubb>
Date: Wed, 30 Apr 2025 08:03:09 -0700
From: Shakeel Butt <shakeel.butt@...ux.dev>
To: Vlastimil Babka <vbabka@...e.cz>
Cc: Andrew Morton <akpm@...ux-foundation.org>, 
	Johannes Weiner <hannes@...xchg.org>, Michal Hocko <mhocko@...nel.org>, 
	Roman Gushchin <roman.gushchin@...ux.dev>, Muchun Song <muchun.song@...ux.dev>, 
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>, linux-mm@...ck.org, cgroups@...r.kernel.org, 
	linux-kernel@...r.kernel.org, Meta kernel team <kernel-team@...a.com>, 
	bpf <bpf@...r.kernel.org>
Subject: Re: [PATCH 2/4] memcg: separate local_trylock for memcg and obj

On Wed, Apr 30, 2025 at 01:42:47PM +0200, Vlastimil Babka wrote:
> On 4/30/25 01:04, Shakeel Butt wrote:
> > The per-cpu stock_lock protects cached memcg and cached objcg and their
> > respective fields. However there is no dependency between these fields
> > and it is better to have fine grained separate locks for cached memcg
> > and cached objcg. This decoupling of locks allows us to make the memcg
> > charge cache and objcg charge cache to be nmi safe independently.
> > 
> > At the moment, memcg charge cache is already nmi safe and this
> > decoupling will allow to make memcg charge cache work without disabling
> > irqs.
> > 
> > Signed-off-by: Shakeel Butt <shakeel.butt@...ux.dev>
> > ---
> >  mm/memcontrol.c | 52 +++++++++++++++++++++++++++----------------------
> >  1 file changed, 29 insertions(+), 23 deletions(-)
> 
> > @@ -1883,19 +1885,22 @@ static void drain_local_stock(struct work_struct *dummy)
> >  	struct memcg_stock_pcp *stock;
> >  	unsigned long flags;
> >  
> > -	/*
> > -	 * The only protection from cpu hotplug (memcg_hotplug_cpu_dead) vs.
> > -	 * drain_stock races is that we always operate on local CPU stock
> > -	 * here with IRQ disabled
> > -	 */
> > -	local_lock_irqsave(&memcg_stock.stock_lock, flags);
> > +	if (WARN_ONCE(!in_task(), "drain in non-task context"))
> > +		return;
> >  
> > +	preempt_disable();
> >  	stock = this_cpu_ptr(&memcg_stock);
> > +
> > +	local_lock_irqsave(&memcg_stock.obj_lock, flags);
> >  	drain_obj_stock(stock);
> > +	local_unlock_irqrestore(&memcg_stock.obj_lock, flags);
> > +
> > +	local_lock_irqsave(&memcg_stock.memcg_lock, flags);
> >  	drain_stock_fully(stock);
> > -	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
> > +	local_unlock_irqrestore(&memcg_stock.memcg_lock, flags);
> >  
> > -	local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
> > +	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
> > +	preempt_enable();
> 
> This usage of preempt_disable() looks rather weird and makes RT unhappy as
> the local lock is a mutex, so it gives you this:
> 
> BUG: sleeping function called from invalid context at
> kernel/locking/spinlock_rt.c:48
> 
> I know the next patch removes it again but for bisectability purposes it
> should be avoided. Instead of preempt_disable() we can extend the local lock
> scope here?
> 

Indeed and thanks for the suggestion, will fix in v2.