From: Vladimir Davydov <vdavydov.dev@gmail.com>
To: Shakeel Butt <shakeelb@google.com>
Cc: Michal Hocko <mhocko@kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
Greg Thelen <gthelen@google.com>,
Christoph Lameter <cl@linux.com>,
Pekka Enberg <penberg@kernel.org>,
David Rientjes <rientjes@google.com>,
Joonsoo Kim <iamjoonsoo.kim@lge.com>,
Johannes Weiner <hannes@cmpxchg.org>, Tejun Heo <tj@kernel.org>,
Linux MM <linux-mm@kvack.org>,
cgroups@vger.kernel.org, LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2] mm: fix race between kmem_cache destroy, create and deactivate
Date: Sat, 26 May 2018 21:58:37 +0300 [thread overview]
Message-ID: <20180526185837.k5ztrillokpi65qj@esperanza> (raw)
In-Reply-To: <20180522201336.196994-1-shakeelb@google.com>
On Tue, May 22, 2018 at 01:13:36PM -0700, Shakeel Butt wrote:
> The memcg kmem cache creation and deactivation (SLUB only) is
> asynchronous. If a root kmem cache is destroyed whose memcg cache is in
> the process of creation or deactivation, the kernel may crash.
>
> Example of one such crash:
> general protection fault: 0000 [#1] SMP PTI
> CPU: 1 PID: 1721 Comm: kworker/14:1 Not tainted 4.17.0-smp
> ...
> Workqueue: memcg_kmem_cache kmemcg_deactivate_workfn
> RIP: 0010:has_cpu_slab
> ...
> Call Trace:
> ? on_each_cpu_cond
> __kmem_cache_shrink
> kmemcg_cache_deact_after_rcu
> kmemcg_deactivate_workfn
> process_one_work
> worker_thread
> kthread
> ret_from_fork+0x35/0x40
>
> This issue is due to the lack of real reference counting for the root
> kmem_caches. Currently kmem_cache does have a field named refcount which
> has been used for multiple purposes i.e. shared count, reference count
> and noshare flag. Due to its conflated nature, it can not be used for
> reference counting by other subsystems.
>
> This patch decoupled the reference counting from shared count and
> noshare flag. The new field 'shared_count' represents the shared count
> and noshare flag while 'refcount' is converted into a real reference
> counter.
>
> The reference counting is only implemented for root kmem_caches for
> simplicity. The reference of a root kmem_cache is elevated on sharing or
> while its memcg kmem_cache creation or deactivation request is in the
> fly and thus it is made sure that the root kmem_cache is not destroyed
> in the middle. As the reference of kmem_cache is elevated on sharing,
> the 'shared_count' does not need any locking protection as at worst it
> can be out-dated for a small window which is tolerable.
I wonder if we could fix this problem without introducing reference
counting for kmem caches (which seems a bit of an overkill to me TBO),
e.g. by flushing memcg_kmem_cache_wq before root cache destruction?
next prev parent reply other threads:[~2018-05-26 18:58 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-05-22 20:13 Shakeel Butt
2018-05-26 18:58 ` Vladimir Davydov [this message]
2018-05-26 22:43 ` Shakeel Butt
2018-05-29 13:58 ` Christopher Lameter
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=20180526185837.k5ztrillokpi65qj@esperanza \
--to=vdavydov.dev@gmail.com \
--cc=akpm@linux-foundation.org \
--cc=cgroups@vger.kernel.org \
--cc=cl@linux.com \
--cc=gthelen@google.com \
--cc=hannes@cmpxchg.org \
--cc=iamjoonsoo.kim@lge.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@kernel.org \
--cc=penberg@kernel.org \
--cc=rientjes@google.com \
--cc=shakeelb@google.com \
--cc=tj@kernel.org \
/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