From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: Vlastimil Babka <vbabka@suse.cz>
Cc: Shakeel Butt <shakeel.butt@linux.dev>,
Andrew Morton <akpm@linux-foundation.org>,
Johannes Weiner <hannes@cmpxchg.org>,
Michal Hocko <mhocko@kernel.org>,
Roman Gushchin <roman.gushchin@linux.dev>,
Muchun Song <muchun.song@linux.dev>,
linux-mm@kvack.org, cgroups@vger.kernel.org,
linux-kernel@vger.kernel.org,
Meta kernel team <kernel-team@meta.com>
Subject: Re: [RFC PATCH 10/10] memcg: no more irq disabling for stock locks
Date: Fri, 14 Mar 2025 12:58:02 +0100 [thread overview]
Message-ID: <20250314115802.DESa-C1z@linutronix.de> (raw)
In-Reply-To: <9e1e3877-55ae-4546-a5c1-08ea730ea638@suse.cz>
On 2025-03-14 11:54:34 [+0100], Vlastimil Babka wrote:
> On 3/14/25 07:15, Shakeel Butt wrote:
> > Let's switch all memcg_stock locks acquire and release places to not
> > disable and enable irqs. There are two still functions (i.e.
> > mod_objcg_state() and drain_obj_stock) which needs to disable irqs to
> > update the stats on non-RT kernels. For now add a simple wrapper for
> > that functionality.
>
> BTW, which part of __mod_objcg_mlstate() really needs disabled irqs and not
> just preemption? I see it does rcu_read_lock() anyway, which disables
> preemption. Then in __mod_memcg_lruvec_state() we do some __this_cpu_add()
> updates. I think these also are fine with just disabled preemption as they
> are atomic vs irqs (but don't need LOCK prefix to be atomic vs other cpus
> updates).
__this_cpu_add() is not safe if also used in interrupt context. Some
architectures (not x86) implemented as read, add, write.
this_cpu_add()() does the same but disables interrupts during the
operation.
So __this_cpu_add() should not be used if interrupts are not disabled
and a modification can happen from interrupt context.
> Is it just memcg_rstat_updated() which does READ_ONCE/WRITE_ONCE? Could we
> perhaps just change it to operations where disabled preemption is enough?
>
> > Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
…
> > @@ -2757,6 +2745,28 @@ static void replace_stock_objcg(struct memcg_stock_pcp *stock,
> > WRITE_ONCE(stock->cached_objcg, objcg);
> > }
> >
> > +static unsigned long rt_lock(void)
> > +{
No, we don't name it rt_lock(). We have local_lock() for this exact
reason. And migrate_disable() does not protect vs re-enter of the
function on the CPU while local_irq_save() does.
> > +#ifdef CONFIG_PREEMPT_RT
> > + migrate_disable();
> > + return 0;
> > +#else
> > + unsigned long flags = 0;
> > +
> > + local_irq_save(flags);
> > + return flags;
> > +#endif
> > +}
> > +
> > +static void rt_unlock(unsigned long flags)
> > +{
> > +#ifdef CONFIG_PREEMPT_RT
> > + migrate_enable();
> > +#else
> > + local_irq_restore(flags);
> > +#endif
> > +}
> > +
> > static void mod_objcg_state(struct obj_cgroup *objcg, struct pglist_data *pgdat,
> > enum node_stat_item idx, int nr)
> > {
Sebastian
next prev parent reply other threads:[~2025-03-14 11:58 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-14 6:15 [RFC PATCH 00/10] memcg: stock code cleanups Shakeel Butt
2025-03-14 6:15 ` [RFC PATCH 01/10] memcg: remove root memcg check from refill_stock Shakeel Butt
2025-03-14 6:15 ` [RFC PATCH 02/10] memcg: decouple drain_obj_stock from local stock Shakeel Butt
2025-03-14 9:57 ` Vlastimil Babka
2025-03-14 6:15 ` [RFC PATCH 03/10] memcg: introduce memcg_uncharge Shakeel Butt
2025-03-14 10:01 ` Vlastimil Babka
2025-03-14 6:15 ` [RFC PATCH 04/10] memcg: manually inline __refill_stock Shakeel Butt
2025-03-14 10:05 ` Vlastimil Babka
2025-03-14 6:15 ` [RFC PATCH 05/10] memcg: no refilling stock from obj_cgroup_release Shakeel Butt
2025-03-14 10:09 ` Vlastimil Babka
2025-03-14 11:26 ` Sebastian Andrzej Siewior
2025-03-14 15:25 ` Shakeel Butt
2025-03-14 6:15 ` [RFC PATCH 06/10] memcg: do obj_cgroup_put inside drain_obj_stock Shakeel Butt
2025-03-14 10:17 ` Vlastimil Babka
2025-03-14 11:35 ` Sebastian Andrzej Siewior
2025-03-14 15:29 ` Shakeel Butt
2025-03-14 6:15 ` [RFC PATCH 07/10] memcg: use __mod_memcg_state in drain_obj_stock Shakeel Butt
2025-03-14 10:27 ` Vlastimil Babka
2025-03-14 11:44 ` Sebastian Andrzej Siewior
2025-03-14 11:38 ` Sebastian Andrzej Siewior
2025-03-14 6:15 ` [RFC PATCH 08/10] memcg: assert in_task for couple of local_lock holders Shakeel Butt
2025-03-14 6:15 ` [RFC PATCH 09/10] memcg: trylock stock for objcg Shakeel Butt
2025-03-14 11:47 ` Sebastian Andrzej Siewior
2025-03-14 15:33 ` Shakeel Butt
2025-03-14 6:15 ` [RFC PATCH 10/10] memcg: no more irq disabling for stock locks Shakeel Butt
2025-03-14 10:54 ` Vlastimil Babka
2025-03-14 11:58 ` Sebastian Andrzej Siewior [this message]
2025-03-14 15:55 ` Shakeel Butt
2025-03-14 16:42 ` Sebastian Andrzej Siewior
2025-03-14 17:02 ` Shakeel Butt
2025-03-14 17:38 ` Shakeel Butt
2025-03-14 18:19 ` Sebastian Andrzej Siewior
2025-03-14 13:33 ` [RFC PATCH 00/10] memcg: stock code cleanups Vlastimil Babka
2025-03-14 16:03 ` Shakeel Butt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250314115802.DESa-C1z@linutronix.de \
--to=bigeasy@linutronix.de \
--cc=akpm@linux-foundation.org \
--cc=cgroups@vger.kernel.org \
--cc=hannes@cmpxchg.org \
--cc=kernel-team@meta.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@kernel.org \
--cc=muchun.song@linux.dev \
--cc=roman.gushchin@linux.dev \
--cc=shakeel.butt@linux.dev \
--cc=vbabka@suse.cz \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox