Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Jia-Ju Bai <baijiaju1990@gmail.com>
To:     peterz@infradead.org, mingo@redhat.com, will.deacon@arm.com
Cc:     linux-kernel@vger.kernel.org, Jia-Ju Bai <baijiaju1990@gmail.com>
Subject: [PATCH V2] kernel: locking: rtmutex: Fix a possible sleep-in-atomic-context bug in rt_mutex_handle_deadlock()
Date:   Sat, 11 Aug 2018 11:00:37 +0800
Message-Id: <20180811030037.14885-1-baijiaju1990@gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

The driver may sleep with holding a spinlock.

The function call paths (from bottom to top) in Linux-4.16 are:

[FUNC] schedule
kernel/locking/rtmutex.c, 1223:
	schedule in rt_mutex_handle_deadlock
kernel/locking/rtmutex.c, 1273:
	rt_mutex_handle_deadlock in rt_mutex_slowlock
kernel/locking/rtmutex.c, 1249:
	_raw_spin_lock_irqsave in rt_mutex_slowlock

To fix the bug, the spinlock is released before the loop of schedule()
This is found by my static analysis tool (DSAC).

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
---
v2:
* Release the spinlock before the loop, instead of v1 that releases the
  spinlock before schedule() and then acquires the spinlock again.
  Thank Steven for good advice.
---
 kernel/locking/rtmutex.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 2823d4163a37..8f25a289abe8 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -1205,7 +1205,7 @@ __rt_mutex_slowlock(struct rt_mutex *lock, int state,
 }
 
 static void rt_mutex_handle_deadlock(int res, int detect_deadlock,
-				     struct rt_mutex_waiter *w)
+				     struct rt_mutex_waiter *w, struct rt_mutex *lock)
 {
 	/*
 	 * If the result is not -EDEADLOCK or the caller requested
@@ -1218,6 +1218,8 @@ static void rt_mutex_handle_deadlock(int res, int detect_deadlock,
 	 * Yell lowdly and stop the task right here.
 	 */
 	rt_mutex_print_deadlock(w);
+	/* We're not going anywhere, release the wait_lock */
+	raw_spin_unlock_irq(&lock->wait_lock);
 	while (1) {
 		set_current_state(TASK_INTERRUPTIBLE);
 		schedule();
@@ -1269,7 +1271,7 @@ rt_mutex_slowlock(struct rt_mutex *lock, int state,
 	if (unlikely(ret)) {
 		__set_current_state(TASK_RUNNING);
 		remove_waiter(lock, &waiter);
-		rt_mutex_handle_deadlock(ret, chwalk, &waiter);
+		rt_mutex_handle_deadlock(ret, chwalk, &waiter, lock);
 	}
 
 	/*
-- 
2.17.0