commit e3ec2821b2c489119f230d82e82c99724b93104c
author Gaurav Kohli <gkohli@codeaurora.org> Tue Apr 24 12:13:34 2018 +0530
committer Gaurav Kohli <gkohli@codeaurora.org> Sun May 13 20:53:46 2018 +0530
tree 801845d407f32a933954c2fb4eb90278b9072f89
parent ea8e42153ec76f8fa3ec4158c57384902989f2d6

kthread/smpboot:  Serialize kthread parking against wakeup

The control cpu thread which initiates hotplug calls kthread_park()
for hotplug thread and sets KTHREAD_SHOULD_PARK. After this control
thread wakes up the hotplug thread. There is a chance that wakeup
code sees the hotplug thread (running on AP core) in INTERRUPTIBLE
state, but sets its state to RUNNING after hotplug thread has entered
kthread_parkme() and changed its state to TASK_PARKED. This can result
in panic later on in kthread_unpark(), as it sees KTHREAD_IS_PARKED
flag set but fails to rebind the kthread, due to it being not in
TASK_PARKED state. Fix this, by serializing wakeup state change,
against state change before parking the kthread.

Below is the possible race:

Control thread				      Hotplug Thread

kthread_park()
set KTHREAD_SHOULD_PARK
					      smpboot_thread_fn
					      set_current_state(TASK_INTERRUPTIBLE);
					      kthread_parkme

wake_up_process()

raw_spin_lock_irqsave(&p->pi_lock, flags);
if (!(p->state & state)) -> this will fail
            goto out;

					      __kthread_parkme
					       __set_current_state(TASK_PARKED);

if (p->on_rq && ttwu_remote(p, wake_flags))
    ttwu_remote()
        p->state = TASK_RUNNING;
						schedule();

So to avoid this race, take pi_lock to serial state changes.

Change-Id: Ie71645d37046f7ee74df880dbead29efbaad199a
Suggested-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Neeraj Upadhyay <neeraju@codeaurora.org>
Signed-off-by: Gaurav Kohli <gkohli@codeaurora.org>

kernel/smpboot.c

diff --git a/kernel/smpboot.c b/kernel/smpboot.c
index 1650578..10c5a3a 100644
--- a/kernel/smpboot.c
+++ b/kernel/smpboot.c
@@ -121,7 +121,45 @@
 		}
 
 		if (kthread_should_park()) {
+			/*
+			 * Serialize against wakeup. If we take the lock first,
+			 * wakeup is skipped. If we run later, we observe,
+			 * TASK_RUNNING update from wakeup path, before moving
+			 * forward. This helps avoid the race, where wakeup
+			 * observes TASK_INTERRUPTIBLE, and also observes
+			 * the TASK_PARKED in kthread_parkme() before updating
+			 * task state to TASK_RUNNING. In this case, kthread
+			 * gets parked in TASK_RUNNING state. This results
+			 * in panic later on in kthread_unpark(), as it sees
+			 * KTHREAD_IS_PARKED flag set but fails to rebind the
+			 * kthread, due to it being not in TASK_PARKED state.
+			 *
+			 * Control thread                      Hotplug Thread
+			 *
+			 * kthread_park()
+			 *   set KTHREAD_SHOULD_PARK
+			 *                                smpboot_thread_fn()
+			 *                                  set_current_state(
+			 *                                  TASK_INTERRUPTIBLE);
+			 *                                  kthread_parkme()
+			 *
+			 *   wake_up_process()
+			 *
+			 * raw_spin_lock_irqsave(&p->pi_lock, flags);
+			 * if (!(p->state & state))
+			 *            goto out;
+			 *
+			 *                                  __set_current_state(
+			 *                                  TASK_PARKED);
+			 *
+			 * if (p->on_rq && ttwu_remote(p, wake_flags))
+			 *   ttwu_remote()
+			 *     p->state = TASK_RUNNING;
+			 *                                   schedule();
+			 */
+			raw_spin_lock(&current->pi_lock);
 			__set_current_state(TASK_RUNNING);
+			raw_spin_unlock(&current->pi_lock);
 			preempt_enable();
 			if (ht->park && td->status == HP_THREAD_ACTIVE) {
 				BUG_ON(td->cpu != smp_processor_id());