Skip to contentSkip to conversation

time/alarmtimer: Add bounds checking for the timer interval and expiration

PATCH In review Prarit Bhargava 2 participants 2 messages Jun 12, 2017 Timekeeping, clocksource core, NTP, alarmtimer

lore
Share & tools
lore

Prarit Bhargava

author
#0 lore raw

PATCH 1/1 time/alarmtimer: Add bounds checking for the timer interval and expiration

Running the syzkaller timer_hang_poc.c
(https://groups.google.com/forum/#!topic/syzkaller/355tWdc8oHY) causes a kernel hang by passing in a time that results in the timer being started with an expiration that exceeds KTIME_MAX.

Check for overflow values of expiration (exp) and interval.

Also cleanup the code. Set and check the values, and then modify the timer.

Signed-off-by Prarit Bhargava <[email protected]>
Cc John Stultz <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Stephen Boyd <[email protected]>
---
 kernel/time/alarmtimer.c |   20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)
kernel/time/alarmtimer.c +13 -7
index 5cb5b0008d97..19400dee83c9 100644
@@ -643,7 +643,7 @@ static int alarm_timer_set(struct k_itimer *timr, int flags,
 			   struct itimerspec64 *new_setting,
 			   struct itimerspec64 *old_setting)
 {
-	ktime_t exp;
+	ktime_t exp, interval;
 
 	if (!rtcdev)
 		return -ENOTSUPP;

@@ -654,12 +654,6 @@ static int alarm_timer_set(struct k_itimer *timr, int flags,
 	if (old_setting)
 		alarm_timer_get(timr, old_setting);
 
-	/* If the timer was already set, cancel it */
-	if (alarm_try_to_cancel(&timr->it.alarm.alarmtimer) < 0)
-		return TIMER_RETRY;
-
-	/* start the timer */
-	timr->it.alarm.interval = timespec64_to_ktime(new_setting->it_interval);
 	exp = timespec64_to_ktime(new_setting->it_value);
 	/* Convert (if necessary) to absolute time */
 	if (flags != TIMER_ABSTIME) {

@@ -668,7 +662,19 @@ static int alarm_timer_set(struct k_itimer *timr, int flags,
 		now = alarm_bases[timr->it.alarm.alarmtimer.type].gettime();
 		exp = ktime_add(now, exp);
 	}
+	if (exp < 0)
+		exp = KTIME_MAX;
 
+	interval = timespec64_to_ktime(new_setting->it_interval);
+	if (interval < 0)
+		interval = KTIME_MAX;
+
+	/* If the timer was already set, cancel it */
+	if (alarm_try_to_cancel(&timr->it.alarm.alarmtimer) < 0)
+		return TIMER_RETRY;
+
+	/* start the timer */
+	timr->it.alarm.interval = interval;
 	alarm_start(&timr->it.alarm.alarmtimer, exp);
 	return 0;
 }
Signature · 1 line
1.7.9.3
#1 lore raw
Quoted 9 lines · Prarit Bhargava

Running the syzkaller timer_hang_poc.c
(https://groups.google.com/forum/#!topic/syzkaller/355tWdc8oHY) causes a kernel hang by passing in a time that results in the timer being started with an expiration that exceeds KTIME_MAX.

Check for overflow values of expiration (exp) and interval.

Also cleanup the code. Set and check the values, and then modify the timer.

This is already addressed in

git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git timers/urgent

I just did not come around to send it to Linus last week.

Thanks,

tglx