[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230330121238.176534-1-chenzhongjin@huawei.com>
Date: Thu, 30 Mar 2023 20:12:37 +0800
From: Chen Zhongjin <chenzhongjin@...wei.com>
To: <linux-kernel@...r.kernel.org>
CC: <chenzhongjin@...wei.com>, <mingo@...hat.com>,
<peterz@...radead.org>, <juri.lelli@...hat.com>,
<vincent.guittot@...aro.org>, <dietmar.eggemann@....com>,
<rostedt@...dmis.org>, <bsegall@...gle.com>, <mgorman@...e.de>,
<bristot@...hat.com>, <vschneid@...hat.com>,
<rmk+kernel@...linux.org.uk>, <geert@...ux-m68k.org>,
<keescook@...omium.org>
Subject: [PATCH] wchan: Fix get_wchan() when task in schedule
get_wchan() check task to unwind is not running or going to run by:
state != TASK_RUNNING && state != TASK_WAKING && !p->on_rq
However this cannot detect task which is going to be scheduled out.
For example, in this path:
__wait_for_common(x, schedule_timeout, timeout, TASK_UNINTERRUPTIBLE)
do_wait_for_common() // state == TASK_UNINTERRUPTIBLE
schedule_timeout()
__schedule()
deactivate_task() // on_rq = 0
After this point get_wchan() can be run on the task but it is still
running actually, and p->pi_lock doesn't work for this case.
It can trigger some warning when running stacktrace on a running task.
Also check p->on_cpu to promise task is really switched out can prevent
this.
Fixes: 42a20f86dc19 ("sched: Add wrapper for get_wchan() to keep task blocked")
Signed-off-by: Chen Zhongjin <chenzhongjin@...wei.com>
---
kernel/sched/core.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 0d18c3969f90..2071d1c0eaca 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2041,7 +2041,8 @@ unsigned long get_wchan(struct task_struct *p)
raw_spin_lock_irq(&p->pi_lock);
state = READ_ONCE(p->__state);
smp_rmb(); /* see try_to_wake_up() */
- if (state != TASK_RUNNING && state != TASK_WAKING && !p->on_rq)
+ if (state != TASK_RUNNING && state != TASK_WAKING &&
+ !p->on_rq && !p->on_cpu)
ip = __get_wchan(p);
raw_spin_unlock_irq(&p->pi_lock);
--
2.17.1
Powered by blists - more mailing lists