Re: [PATCH 2/4] memcg: separate local_trylock for memcg and obj

[Shakeel Butt <shakeel.butt@linux.dev>]

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@linux.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.