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;
Date:   Tue,  8 Oct 2019 10:32:04 -0700
Message-Id: <20191008173204.180879-1-edumazet@google.com>
Mime-Version: 1.0
Subject: [PATCH] hrtimer: annotate lockless access to timer->base
From:   Eric Dumazet <edumazet@google.com>
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     linux-kernel <linux-kernel@vger.kernel.org>,
        Eric Dumazet <edumazet@google.com>,
        Eric Dumazet <eric.dumazet@gmail.com>,
        Julien Grall <julien.grall@arm.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Followup to commit dd2261ed45aa ("hrtimer: Protect lockless access
to timer->base")

lock_hrtimer_base() fetches timer->base without lock exclusion.

Compiler is allowed to read timer->base twice (even if considered dumb)
and we could end up trying to lock migration_base and
return &migration_base.

  base = timer->base;
  if (likely(base != &migration_base)) {

       /* compiler reads timer->base again, and now (base == &migration_base)

       raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
       if (likely(base == timer->base))
            return base; /* == &migration_base ! */

Similarly the write sides should use WRITE_ONCE() to avoid
store tearing.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Julien Grall <julien.grall@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
---
 kernel/time/hrtimer.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
index 0d4dc241c0fb498036c91a571e65cb00f5d19ba6..65605530ee349c9682690c4fccb43aa9284d4287 100644
--- a/kernel/time/hrtimer.c
+++ b/kernel/time/hrtimer.c
@@ -164,7 +164,7 @@ struct hrtimer_clock_base *lock_hrtimer_base(const struct hrtimer *timer,
 	struct hrtimer_clock_base *base;
 
 	for (;;) {
-		base = timer->base;
+		base = READ_ONCE(timer->base);
 		if (likely(base != &migration_base)) {
 			raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
 			if (likely(base == timer->base))
@@ -244,7 +244,7 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
 			return base;
 
 		/* See the comment in lock_hrtimer_base() */
-		timer->base = &migration_base;
+		WRITE_ONCE(timer->base, &migration_base);
 		raw_spin_unlock(&base->cpu_base->lock);
 		raw_spin_lock(&new_base->cpu_base->lock);
 
@@ -253,10 +253,10 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
 			raw_spin_unlock(&new_base->cpu_base->lock);
 			raw_spin_lock(&base->cpu_base->lock);
 			new_cpu_base = this_cpu_base;
-			timer->base = base;
+			WRITE_ONCE(timer->base, base);
 			goto again;
 		}
-		timer->base = new_base;
+		WRITE_ONCE(timer->base, new_base);
 	} else {
 		if (new_cpu_base != this_cpu_base &&
 		    hrtimer_check_target(timer, new_base)) {
-- 
2.23.0.581.g78d2f28ef7-goog