Re: [PATCH] sched: fix tsk->pi_lock isn't held when do_set_cpus_allowed()
From: Leo Yan
Date: Tue Aug 25 2015 - 04:14:56 EST
Hi Wanpeng,
On Tue, Aug 25, 2015 at 03:59:54PM +0800, Wanpeng Li wrote:
> [ 15.273708] ------------[ cut here ]------------
> [ 15.274097] WARNING: CPU: 0 PID: 13 at kernel/sched/core.c:1156 do_set_cpus_allowed+0x7e/0x80()
> [ 15.274857] Modules linked in:
> [ 15.275101] CPU: 0 PID: 13 Comm: migration/0 Not tainted 4.2.0-rc1-00049-g25834c7 #2
> [ 15.275674] 00000000 00000000 d21f1d24 c19228b2 00000000 d21f1d58 c1056a3b c1ba00e4
> [ 15.276084] 00000000 0000000d c1ba17d8 00000484 c10838be 00000484 c10838be d21e5000
> [ 15.276084] d2121900 d21e5158 d21f1d68 c1056b12 00000009 00000000 d21f1d7c c10838be
> [ 15.276084] Call Trace:
> [ 15.276084] [<c19228b2>] dump_stack+0x4b/0x75
> [ 15.276084] [<c1056a3b>] warn_slowpath_common+0x8b/0xc0
> [ 15.276084] [<c10838be>] ? do_set_cpus_allowed+0x7e/0x80
> [ 15.276084] [<c10838be>] ? do_set_cpus_allowed+0x7e/0x80
> [ 15.276084] [<c1056b12>] warn_slowpath_null+0x22/0x30
> [ 15.276084] [<c10838be>] do_set_cpus_allowed+0x7e/0x80
> [ 15.276084] [<c110154c>] cpuset_cpus_allowed_fallback+0x7c/0x170
> [ 15.276084] [<c11014d0>] ? cpuset_cpus_allowed+0x180/0x180
> [ 15.276084] [<c1083ae1>] select_fallback_rq+0x221/0x280
> [ 15.276084] [<c1085073>] migration_call+0xe3/0x250
> [ 15.276084] [<c1079e23>] notifier_call_chain+0x53/0x70
> [ 15.276084] [<c1079e5e>] __raw_notifier_call_chain+0x1e/0x30
> [ 15.276084] [<c1056cc8>] cpu_notify+0x28/0x50
> [ 15.276084] [<c191e4d2>] take_cpu_down+0x22/0x40
> [ 15.276084] [<c1102895>] multi_cpu_stop+0xd5/0x140
> [ 15.276084] [<c11027c0>] ? __stop_cpus+0x80/0x80
> [ 15.276084] [<c11025cc>] cpu_stopper_thread+0xbc/0x170
> [ 15.276084] [<c1085ec9>] ? preempt_count_sub+0x9/0x50
> [ 15.276084] [<c192b6a7>] ? _raw_spin_unlock_irq+0x37/0x50
> [ 15.276084] [<c192b655>] ? _raw_spin_unlock_irqrestore+0x55/0x70
> [ 15.276084] [<c10a9074>] ? trace_hardirqs_on_caller+0x144/0x1e0
> [ 15.276084] [<c11024a5>] ? cpu_stop_should_run+0x35/0x40
> [ 15.276084] [<c1085ec9>] ? preempt_count_sub+0x9/0x50
> [ 15.276084] [<c192b641>] ? _raw_spin_unlock_irqrestore+0x41/0x70
> [ 15.276084] [<c107c944>] smpboot_thread_fn+0x174/0x2f0
> [ 15.276084] [<c107c7d0>] ? sort_range+0x30/0x30
> [ 15.276084] [<c1078934>] kthread+0xc4/0xe0
> [ 15.276084] [<c192c041>] ret_from_kernel_thread+0x21/0x30
> [ 15.276084] [<c1078870>] ? kthread_create_on_node+0x180/0x180
> [ 15.276084] ---[ end trace 15f4c86d404693b0 ]---
>
> After commit (25834c73f93: sched: Fix a race between __kthread_bind()
> and sched_setaffinity()) do_set_cpus_allowed() should be called w/
> p->pi_lock held, however, it is not true in cpuset path currently.
>
> This patch fix it by holding p->pi_lock in cpuset path.
>
> Signed-off-by: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
> ---
> kernel/cpuset.c | 4 ++++
> 1 files changed, 4 insertions(+), 0 deletions(-)
>
> diff --git a/kernel/cpuset.c b/kernel/cpuset.c
> index e414ae9..605ed66 100644
> --- a/kernel/cpuset.c
> +++ b/kernel/cpuset.c
> @@ -2376,8 +2376,12 @@ void cpuset_cpus_allowed(struct task_struct *tsk, struct cpumask *pmask)
>
> void cpuset_cpus_allowed_fallback(struct task_struct *tsk)
> {
> + unsigned long flags;
> +
> rcu_read_lock();
> + raw_spin_lock_irqsave(&tsk->pi_lock, flags);
> do_set_cpus_allowed(tsk, task_cs(tsk)->effective_cpus);
> + raw_spin_lock_irqsave(&tsk->pi_lock, flags);
Just curious, Will introduce deadlock after acquire lock twice? ;)
Thanks,
Leo Yan
> rcu_read_unlock();
>
> /*
> --
> 1.7.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/