[PATCH] kmemleak: don't hang if user disables scanning early

[PATCH] kmemleak: don't hang if user disables scanning early

[Vegard Nossum]

If the user tries to disable automatic scanning early in the boot
process using e.g.:

  echo scan=off > /sys/kernel/debug/kmemleak

then this command will hang until SECS_FIRST_SCAN (= 60) seconds have
elapsed, even though the system is fully initialised.

We can fix this using interruptible sleep and checking if we're
supposed to stop whenever we wake up (like the rest of the code does).

Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
---
 mm/kmemleak.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/mm/kmemleak.c b/mm/kmemleak.c
index 04320d3..086292f 100644
--- a/mm/kmemleak.c
+++ b/mm/kmemleak.c
@@ -1485,8 +1485,10 @@ static int kmemleak_scan_thread(void *arg)
 	 * Wait before the first scan to allow the system to fully initialize.
 	 */
 	if (first_run) {
+		signed long timeout = msecs_to_jiffies(SECS_FIRST_SCAN * 1000);
 		first_run = 0;
-		ssleep(SECS_FIRST_SCAN);
+		while (timeout && !kthread_should_stop())
+			timeout = schedule_timeout_interruptible(timeout);
 	}
 
 	while (!kthread_should_stop()) {
-- 
1.9.1

Re: [PATCH] kmemleak: don't hang if user disables scanning early

[Catalin Marinas]

On Mon, Jul 18, 2016 at 11:43:25AM +0200, Vegard Nossum wrote:
> If the user tries to disable automatic scanning early in the boot
> process using e.g.:
> 
>   echo scan=off > /sys/kernel/debug/kmemleak
> 
> then this command will hang until SECS_FIRST_SCAN (= 60) seconds have
> elapsed, even though the system is fully initialised.
> 
> We can fix this using interruptible sleep and checking if we're
> supposed to stop whenever we wake up (like the rest of the code does).
> 
> Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>

Acked-by: Catalin Marinas <catalin.marinas@arm.com>