sched/deadline: fix task_struct reference leak
PATCH 1/1 sched/deadline: fix task_struct reference leak
During the execution of the following stress test with linux-rt:
stress-ng --cyclic 30 --timeout 30 --minimize --quiet
kmemleak frequently reported a memory leak concerning the task_struct:
unreferenced object 0xffff8881305b8000 (size 16136):
comm "stress-ng", pid 614, jiffies 4294883961 (age 286.412s) object hex dump (first 32 bytes):
02 40 00 00 00 00 00 00 00 00 00 00 00 00 00 00 .@..............
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
debug hex dump (first 16 bytes):
53 09 00 00 00 00 00 00 00 00 00 00 00 00 00 00 S...............
backtrace:
[<00000000046b6790>] dup_task_struct+0x30/0x540
[<00000000c5ca0f0b>] copy_process+0x3d9/0x50e0
[<00000000ced59777>] kernel_clone+0xb0/0x770
[<00000000a50befdc>] __do_sys_clone+0xb6/0xf0
[<000000001dbf2008>] do_syscall_64+0x5d/0xf0
[<00000000552900ff>] entry_SYSCALL_64_after_hwframe+0x6e/0x76The issue occurs in start_dl_timer(), which increments the task_struct reference count and sets a timer. The timer callback, dl_task_timer, is supposed to decrement the reference count upon expiration. However, if enqueue_task_dl() is called before the timer expires and cancels it, the reference count is not decremented, leading to the leak.
This patch fixes the reference leak by ensuring the task_struct reference count is properly decremented when the timer is canceled.
--- kernel/sched/deadline.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/sched/deadline.c +3 -1
@@ -1805,7 +1805,9 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)* problem if it fires concurrently: boosted threads * are ignored in dl_task_timer(). */ -hrtimer_try_to_cancel (&p->dl.dl_timer); +if (hrtimer_try_to_cancel (&p->dl.dl_timer) ==1 && + !dl_server (&p->dl)) +put_task_struct (p); p->dl.dl_throttled =0 ; } }else if (!dl_prio (p->normal_prio)) {
Signature · 1 line
2.45.1
Juri Lelli
maintainerHi Wander,
Quoted 40 lines · Wander Lairson Costa
During the execution of the following stress test with linux-rt:
stress-ng --cyclic 30 --timeout 30 --minimize --quiet
kmemleak frequently reported a memory leak concerning the
task_struct:unreferenced object
0xffff8881305b8000(size 16136):
comm "stress-ng", pid 614, jiffies 4294883961 (age 286.412s) object hex dump (first 32 bytes):02 40 00 00 00 00 00 00 00 00 00 00 00 00 00 00 .@.............. 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ debug hex dump (first 16 bytes): 53 09 00 00 00 00 00 00 00 00 00 00 00 00 00 00 S............... backtrace: [<00000000046b6790>] dup_task_struct+0x30/0x540 [<00000000c5ca0f0b>] copy_process+0x3d9/0x50e0 [<00000000ced59777>] kernel_clone+0xb0/0x770 [<00000000a50befdc>] __do_sys_clone+0xb6/0xf0 [<000000001dbf2008>] do_syscall_64+0x5d/0xf0 [<00000000552900ff>] entry_SYSCALL_64_after_hwframe+0x6e/0x76The issue occurs in
start_dl_timer(), which increments thetask_structreference count and sets a timer. The timer callback,dl_task_timer, is supposed to decrement the reference count upon expiration. However, ifenqueue_task_dl()is called before the timer expires and cancels it, the reference count is not decremented, leading to the leak.This patch fixes the reference leak by ensuring the
task_structreference count is properly decremented when the timer is canceled.Signed-off-by Wander Lairson Costa <[email protected]>--- kernel/sched/deadline.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)kernel/sched/deadline.c +0
-0
kernel/sched/deadline.c static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags) · line 1810
Quoted 7 lines
kernel/sched/deadline.c +3
-1@@ -1805,7 +1805,9 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)* problem if it fires concurrently: boosted threads * are ignored in dl_task_timer(). */ -hrtimer_try_to_cancel (&p->dl.dl_timer); +if (hrtimer_try_to_cancel (&p->dl.dl_timer) ==1 && + !dl_server (&p->dl)) +put_task_struct (p);
Wonder if this deserves a little comment saying that if the timer callback was running (hrtimer_try_to_cancel() returns -1) it will eventually call put_task_struct(). Other than that, the fix makes sense to me.
Oh, another thing is that I believe we should add 'Fixes: feff2e65efd8d ("sched/deadline: Unthrottle PI boosted threads while enqueuing")' in the changelog, shouldn't we?
Thanks,
Juri