Skip to content

Commit bb7262b

Browse files
committed
timers: Move clearing of base::timer_running under base:: Lock
syzbot reported KCSAN data races vs. timer_base::timer_running being set to NULL without holding base::lock in expire_timers(). This looks innocent and most reads are clearly not problematic, but Frederic identified an issue which is: int data = 0; void timer_func(struct timer_list *t) { data = 1; } CPU 0 CPU 1 ------------------------------ -------------------------- base = lock_timer_base(timer, &flags); raw_spin_unlock(&base->lock); if (base->running_timer != timer) call_timer_fn(timer, fn, baseclk); ret = detach_if_pending(timer, base, true); base->running_timer = NULL; raw_spin_unlock_irqrestore(&base->lock, flags); raw_spin_lock(&base->lock); x = data; If the timer has previously executed on CPU 1 and then CPU 0 can observe base->running_timer == NULL and returns, assuming the timer has completed, but it's not guaranteed on all architectures. The comment for del_timer_sync() makes that guarantee. Moving the assignment under base->lock prevents this. For non-RT kernel it's performance wise completely irrelevant whether the store happens before or after taking the lock. For an RT kernel moving the store under the lock requires an extra unlock/lock pair in the case that there is a waiter for the timer, but that's not the end of the world. Reported-by: [email protected] Reported-by: [email protected] Fixes: 030dcdd ("timers: Prepare support for PREEMPT_RT") Signed-off-by: Thomas Gleixner <[email protected]> Tested-by: Sebastian Andrzej Siewior <[email protected]> Link: https://lore.kernel.org/r/[email protected] Cc: [email protected]
1 parent ff11764 commit bb7262b

File tree

1 file changed

+4
-2
lines changed

1 file changed

+4
-2
lines changed

kernel/time/timer.c

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1265,8 +1265,10 @@ static inline void timer_base_unlock_expiry(struct timer_base *base)
12651265
static void timer_sync_wait_running(struct timer_base *base)
12661266
{
12671267
if (atomic_read(&base->timer_waiters)) {
1268+
raw_spin_unlock_irq(&base->lock);
12681269
spin_unlock(&base->expiry_lock);
12691270
spin_lock(&base->expiry_lock);
1271+
raw_spin_lock_irq(&base->lock);
12701272
}
12711273
}
12721274

@@ -1457,14 +1459,14 @@ static void expire_timers(struct timer_base *base, struct hlist_head *head)
14571459
if (timer->flags & TIMER_IRQSAFE) {
14581460
raw_spin_unlock(&base->lock);
14591461
call_timer_fn(timer, fn, baseclk);
1460-
base->running_timer = NULL;
14611462
raw_spin_lock(&base->lock);
1463+
base->running_timer = NULL;
14621464
} else {
14631465
raw_spin_unlock_irq(&base->lock);
14641466
call_timer_fn(timer, fn, baseclk);
1467+
raw_spin_lock_irq(&base->lock);
14651468
base->running_timer = NULL;
14661469
timer_sync_wait_running(base);
1467-
raw_spin_lock_irq(&base->lock);
14681470
}
14691471
}
14701472
}

0 commit comments

Comments
 (0)