[tip:smp/urgent] smp/hotplug: Move unparking of percpu threads to the control CPU

From: tip-bot for Thomas Gleixner
Date: Thu Jul 06 2017 - 05:01:43 EST

Commit-ID: 9cd4f1a4e7a858849e889a081a99adff83e08e4c
Gitweb: http://git.kernel.org/tip/9cd4f1a4e7a858849e889a081a99adff83e08e4c
Author: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
AuthorDate: Tue, 4 Jul 2017 22:20:23 +0200
Committer: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
CommitDate: Thu, 6 Jul 2017 10:55:10 +0200

smp/hotplug: Move unparking of percpu threads to the control CPU

Vikram reported the following backtrace:

BUG: scheduling while atomic: swapper/7/0/0x00000002
CPU: 7 PID: 0 Comm: swapper/7 Not tainted 4.9.32-perf+ #680

He analyzed correctly that a parked cpu hotplug thread of an offlined CPU
was still on the runqueue when the CPU came back online and tried to unpark
it. This causes the thread which invoked kthread_unpark() to call
wait_task_inactive() and subsequently schedule() with preemption disabled.
His proposed workaround was to "make sure" that a parked thread has
scheduled out when the CPU goes offline, so the situation cannot happen.

But that's still wrong because the root cause is not the fact that the
percpu thread is still on the runqueue and neither that preemption is
disabled, which could be simply solved by enabling preemption before
calling kthread_unpark().

The real issue is that the calling thread is the idle task of the upcoming
CPU, which is not supposed to call anything which might sleep. The moron,
who wrote that code, missed completely that kthread_unpark() might end up
in schedule().

The solution is simpler than expected. The thread which controls the
hotplug operation is waiting for the CPU to call complete() on the hotplug
state completion. So the idle task of the upcoming CPU can set its state to
CPUHP_AP_ONLINE_IDLE and invoke complete(). This in turn wakes the control
task on a different CPU, which then can safely do the unpark and kick the
now unparked hotplug thread of the upcoming CPU to complete the bringup to
the final target state.

Control CPU AP

__cpu_up() ------------>
<------------ complete();
wait_for_completion(); hotplug_thread()

Fixes: 8df3e07e7f21 ("cpu/hotplug: Let upcoming cpu bring itself fully up")
Reported-by: Vikram Mulukutla <markivx@xxxxxxxxxxxxxx>
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Acked-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Sebastian Sewior <bigeasy@xxxxxxxxxxxxx>
Cc: Rusty Russell <rusty@xxxxxxxxxxxxxxx>
Cc: Tejun Heo <tj@xxxxxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Link: alpine.DEB.2.20.1707042218020.2131@nanos">http://lkml.kernel.org/r/alpine.DEB.2.20.1707042218020.2131@nanos
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>

kernel/cpu.c | 37 +++++++++++++++++++------------------
1 file changed, 19 insertions(+), 18 deletions(-)

diff --git a/kernel/cpu.c b/kernel/cpu.c
index b03a325..ab86045 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -271,11 +271,25 @@ void cpu_hotplug_enable(void)

+static void __cpuhp_kick_ap_work(struct cpuhp_cpu_state *st);
static int bringup_wait_for_ap(unsigned int cpu)
struct cpuhp_cpu_state *st = per_cpu_ptr(&cpuhp_state, cpu);

+ /* Wait for the CPU to reach CPUHP_AP_ONLINE_IDLE */
+ BUG_ON(!cpu_online(cpu));
+ /* Unpark the stopper thread and the hotplug thread of the target cpu */
+ stop_machine_unpark(cpu);
+ kthread_unpark(st->thread);
+ /* Should we go further up ? */
+ if (st->target > CPUHP_AP_ONLINE_IDLE) {
+ __cpuhp_kick_ap_work(st);
+ wait_for_completion(&st->done);
+ }
return st->result;

@@ -296,9 +310,7 @@ static int bringup_cpu(unsigned int cpu)
if (ret)
return ret;
- ret = bringup_wait_for_ap(cpu);
- BUG_ON(!cpu_online(cpu));
- return ret;
+ return bringup_wait_for_ap(cpu);

@@ -767,31 +779,20 @@ void notify_cpu_starting(unsigned int cpu)

- * Called from the idle task. We need to set active here, so we can kick off
- * the stopper thread and unpark the smpboot threads. If the target state is
- * beyond CPUHP_AP_ONLINE_IDLE we kick cpuhp thread and let it bring up the
- * cpu further.
+ * Called from the idle task. Wake up the controlling task which brings the
+ * stopper and the hotplug thread of the upcoming CPU up and then delegates
+ * the rest of the online bringup to the hotplug thread.
void cpuhp_online_idle(enum cpuhp_state state)
struct cpuhp_cpu_state *st = this_cpu_ptr(&cpuhp_state);
- unsigned int cpu = smp_processor_id();

/* Happens for the boot cpu */
if (state != CPUHP_AP_ONLINE_IDLE)

- /* Unpark the stopper thread and the hotplug thread of this cpu */
- stop_machine_unpark(cpu);
- kthread_unpark(st->thread);
- /* Should we go further up ? */
- if (st->target > CPUHP_AP_ONLINE_IDLE)
- __cpuhp_kick_ap_work(st);
- else
- complete(&st->done);
+ complete(&st->done);

/* Requires cpu_add_remove_lock to be held */