From: Peter Zijlstra <peterz@infradead.org>
To: Prateek Sood <prsood@codeaurora.org>
Cc: tj@kernel.org, lizefan@huawei.com, cgroups@vger.kernel.org,
mingo@kernel.org, longman@redhat.com, boqun.feng@gmail.com,
tglx@linutronix.de, linux-kernel@vger.kernel.org,
sramana@codeaurora.org
Subject: Re: [PATCH] cgroup/cpuset: remove circular dependency deadlock
Date: Thu, 7 Sep 2017 19:45:05 +0200 [thread overview]
Message-ID: <20170907174505.GF17526@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <1504792583-10424-1-git-send-email-prsood@codeaurora.org>
On Thu, Sep 07, 2017 at 07:26:23PM +0530, Prateek Sood wrote:
> Remove circular dependency deadlock in a scenario where hotplug of CPU is
> being done while there is updation in cgroup and cpuset triggered from
> userspace.
>
> Process A => kthreadd => Process B => Process C => Process A
> Process A
> cpu_subsys_offline();
> cpu_down();
> _cpu_down();
> percpu_down_write(&cpu_hotplug_lock); //held
> cpuhp_invoke_callback();
> workqueue_offline_cpu();
> wq_update_unbound_numa();
> kthread_create_on_node();
> wake_up_process(); //wakeup kthreadd
TJ, I'm puzzled, why would we need to spawn new threads to update NUMA
affinity when taking a CPU out? That doesn't make sense to me, we can
either shrink the affinity of an existing thread or completely kill of a
thread if the mask becomes empty. But why spawn a new thread?
> flush_work();
> wait_for_completion();
>
> kthreadd
> kthreadd();
> kernel_thread();
> do_fork();
> copy_process();
> percpu_down_read(&cgroup_threadgroup_rwsem);
> __rwsem_down_read_failed_common(); //waiting
So this will eventually do our:
complete() to make A go.
> Process B
> kernfs_fop_write();
> cgroup_file_write();
> cgroup_procs_write();
> percpu_down_write(&cgroup_threadgroup_rwsem); //held
> cgroup_attach_task();
> cgroup_migrate();
> cgroup_migrate_execute();
> cpuset_can_attach();
> mutex_lock(&cpuset_mutex); //waiting
>
> Process C
> kernfs_fop_write();
> cgroup_file_write();
> cpuset_write_resmask();
> mutex_lock(&cpuset_mutex); //held
> update_cpumask();
> update_cpumasks_hier();
> rebuild_sched_domains_locked();
> get_online_cpus();
> percpu_down_read(&cpu_hotplug_lock); //waiting
So the whole thing looks like:
A B C D
L(hotplug)
L(threadgroup)
L(cpuset)
L(threadgroup)
WFC(c)
L(cpuset)
L(hotplug)
C(c)
Yes, inverting cpuset and hotplug would break that chain, but I'm still
wondering why workqueue needs to spawn threads on CPU down.
next prev parent reply other threads:[~2017-09-07 17:45 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-09-07 13:56 [PATCH] cgroup/cpuset: remove circular dependency deadlock Prateek Sood
2017-09-07 17:45 ` Peter Zijlstra [this message]
2017-09-08 2:13 ` Prateek Sood
2017-09-07 17:51 ` Peter Zijlstra
2017-10-09 13:27 ` Prateek Sood
2017-10-11 9:48 ` Peter Zijlstra
2017-10-25 8:39 ` Prateek Sood
2017-10-25 9:30 ` Peter Zijlstra
2017-10-26 11:52 ` Prateek Sood
2017-10-26 14:05 ` Waiman Long
2017-10-27 8:03 ` Prateek Sood
2017-10-30 7:16 ` [PATCH v2] " Prateek Sood
2017-11-06 4:01 ` Prateek Sood
2017-11-15 10:26 ` Prateek Sood
2017-11-15 10:37 ` Peter Zijlstra
2017-11-15 14:20 ` [PATCH v3 0/2] Invert cpu_hotplug_lock and cpuset_mutex locking order Prateek Sood
2017-11-15 14:20 ` [PATCH v3 1/2] cgroup/cpuset: remove circular dependency deadlock Prateek Sood
2017-11-15 14:20 ` [PATCH v3 2/2] cpuset: Make cpuset hotplug synchronous Prateek Sood
2017-11-27 16:48 ` [PATCH v3 0/2] Invert cpu_hotplug_lock and cpuset_mutex locking order Tejun Heo
2017-11-15 17:05 ` [PATCH v2] cgroup/cpuset: remove circular dependency deadlock Tejun Heo
2017-11-15 17:18 ` Prateek Sood
-- strict thread matches above, loose matches on Subject: below --
2017-09-07 6:04 [PATCH] " Prateek Sood
2017-09-07 7:28 ` Peter Zijlstra
2017-09-07 8:56 ` Boqun Feng
2017-09-07 9:07 ` Prateek Sood
2017-09-07 9:05 ` Prateek Sood
2017-09-06 11:48 Prateek Sood
2017-09-06 12:56 ` Waiman Long
2017-09-06 14:23 ` Prateek Sood
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=20170907174505.GF17526@worktop.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=boqun.feng@gmail.com \
--cc=cgroups@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=lizefan@huawei.com \
--cc=longman@redhat.com \
--cc=mingo@kernel.org \
--cc=prsood@codeaurora.org \
--cc=sramana@codeaurora.org \
--cc=tglx@linutronix.de \
--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;
as well as URLs for NNTP newsgroup(s).