Skip to contentSkip to conversation

sched/deadline: fixes and improvements

PATCH Superseded Wander Lairson Costa 2 participants 11 messages Jul 22, 2024 – Jul 23, 2024 Scheduler

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover sched/deadline: fixes and improvements

Hello,

This patch series addresses specific issues and improvements within the scheduler's deadline (DL) class. The patches aim to fix warnings, remove redundant checks, and consolidate timer cancellation logic for better consistency and code quality.

Patch 1: Fix warning in migrate_enable for boosted tasks

Fix a warning caused by unnecessary calls to setup_new_dl_entity for boosted tasks during CPU migate_enable calls.

Patch 2: sched/deadline: avoid redundant check for boosted task

Remove redundant WARN_ON check in enqueue_dl_entity, as boosted tasks do not require setup_new_dl_entity.

Patch 3: sched/deadline: Consolidate Timer Cancellation

Consolidate timer cancellation logic into dedicated functions, ensuring consistent behavior and reducing code duplication.

Wander Lairson Costa (3):
sched/deadline: Fix warning in migrate_enable for boosted tasks sched/deadline: avoid redundant check for boosted task sched/deadline: Consolidate Timer Cancellation

 kernel/sched/deadline.c | 56 +++++++++++++++++++++++++++--------------
 1 file changed, 37 insertions(+), 19 deletions(-)
Signature · 1 line
2.45.2
#1 lore raw

PATCH 1/3 sched/deadline: Fix warning in migrate_enable for boosted tasks

When running the following command:

while true; do

    stress-ng --cyclic 30 --timeout 30s --minimize --quiet

done

a warning is eventually triggered:

WARNING: CPU: 43 PID: 2848 at kernel/sched/deadline.c:794
setup_new_dl_entity+0x13e/0x180
...
Call Trace:
 <TASK>
 ? show_trace_log_lvl+0x1c4/0x2df
 ? enqueue_dl_entity+0x631/0x6e0
 ? setup_new_dl_entity+0x13e/0x180
 ? __warn+0x7e/0xd0
 ? report_bug+0x11a/0x1a0
 ? handle_bug+0x3c/0x70
 ? exc_invalid_op+0x14/0x70
 ? asm_exc_invalid_op+0x16/0x20
 enqueue_dl_entity+0x631/0x6e0
 enqueue_task_dl+0x7d/0x120
 __do_set_cpus_allowed+0xe3/0x280
 __set_cpus_allowed_ptr_locked+0x140/0x1d0
 __set_cpus_allowed_ptr+0x54/0xa0
 migrate_enable+0x7e/0x150
 rt_spin_unlock+0x1c/0x90
 group_send_sig_info+0xf7/0x1a0
 ? kill_pid_info+0x1f/0x1d0
 kill_pid_info+0x78/0x1d0
 kill_proc_info+0x5b/0x110
 __x64_sys_kill+0x93/0xc0
 do_syscall_64+0x5c/0xf0
 entry_SYSCALL_64_after_hwframe+0x6e/0x76
 RIP: 0033:0x7f0dab31f92b

This warning occurs because set_cpus_allowed dequeues and enqueues tasks with the ENQUEUE_RESTORE flag set. If the task is boosted, the warning is triggered. A boosted task already had its parameters set by rt_mutex_setprio, and a new call to setup_new_dl_entity is unnecessary, hence the WARN_ON call.

Check if we are requeueing a boosted task and avoid calling setup_new_dl_entity if that's the case.

Fixes 2279f540ea7d ("sched/deadline: Fix priority inheritance with multiple scheduling classes")

---

The initial idea for this fix was to introduce another ENQUEUE flag, ENQUEUE_SET_CPUS_ALLOWED. When this flag was set, the deadline scheduler would bypass the call to setup_new_dl_entity, regardless of whether the task was boosted. However, this idea was abandoned due to the presence of other DEQUEUE_SAVE/ENQUEUE_RESTORE pairs in the code. Ultimately, a simpler approach was chosen, which achieves the same practical effects without the need to create an additional flag for enqueue_task.

---
 kernel/sched/deadline.c | 1 +
 1 file changed, 1 insertion(+)
kernel/sched/deadline.c +1 -0
index f59e5c19d944..312e8fa7ce94 100644
@@ -1753,6 +1753,7 @@ enqueue_dl_entity(struct sched_dl_entity *dl_se, int flags)
 	} else if (flags & ENQUEUE_REPLENISH) {
 		replenish_dl_entity(dl_se);
 	} else if ((flags & ENQUEUE_RESTORE) &&
+		   !is_dl_boosted(dl_se) &&
 		   dl_time_before(dl_se->deadline, rq_clock(rq_of_dl_se(dl_se)))) {
 		setup_new_dl_entity(dl_se);
 	}
Signature · 1 line
2.45.2
maintainer
#2 lore raw

Re: [PATCH 1/3] sched/deadline: Fix warning in migrate_enable for boosted tasks

Hi Wander,

Quoted 43 lines · Wander Lairson Costa

When running the following command:

while true; do

    stress-ng --cyclic 30 --timeout 30s --minimize --quiet

done

a warning is eventually triggered:

WARNING: CPU: 43 PID: 2848 at kernel/sched/deadline.c:794
setup_new_dl_entity+0x13e/0x180
...
Call Trace:
 <TASK>
 ? show_trace_log_lvl+0x1c4/0x2df
 ? enqueue_dl_entity+0x631/0x6e0
 ? setup_new_dl_entity+0x13e/0x180
 ? __warn+0x7e/0xd0
 ? report_bug+0x11a/0x1a0
 ? handle_bug+0x3c/0x70
 ? exc_invalid_op+0x14/0x70
 ? asm_exc_invalid_op+0x16/0x20
 enqueue_dl_entity+0x631/0x6e0
 enqueue_task_dl+0x7d/0x120
 __do_set_cpus_allowed+0xe3/0x280
 __set_cpus_allowed_ptr_locked+0x140/0x1d0
 __set_cpus_allowed_ptr+0x54/0xa0
 migrate_enable+0x7e/0x150
 rt_spin_unlock+0x1c/0x90
 group_send_sig_info+0xf7/0x1a0
 ? kill_pid_info+0x1f/0x1d0
 kill_pid_info+0x78/0x1d0
 kill_proc_info+0x5b/0x110
 __x64_sys_kill+0x93/0xc0
 do_syscall_64+0x5c/0xf0
 entry_SYSCALL_64_after_hwframe+0x6e/0x76
 RIP: 0033:0x7f0dab31f92b

This warning occurs because set_cpus_allowed dequeues and enqueues tasks with the ENQUEUE_RESTORE flag set. If the task is boosted, the warning is triggered. A boosted task already had its parameters set by rt_mutex_setprio, and a new call to setup_new_dl_entity is unnecessary, hence the WARN_ON call.

Quoted 5 lines

Check if we are requeueing a boosted task and avoid calling setup_new_dl_entity if that's the case.

Fixes 2279f540ea7d ("sched/deadline: Fix priority inheritance with multiple scheduling classes")

I believe your fix makes sense to me. I only wonder if however it actually fixes 295d6d5e37360 ("sched/deadline: Fix switching to -deadline") instead of the change you reference above?

Thanks,
Juri

#3 lore raw

Re: [PATCH 1/3] sched/deadline: Fix warning in migrate_enable for boosted tasks

Quoted 50 lines · Juri Lelli

Hi Wander,

Quoted 46 lines · Wander Lairson Costa

When running the following command:

while true; do

    stress-ng --cyclic 30 --timeout 30s --minimize --quiet

done

a warning is eventually triggered:

WARNING: CPU: 43 PID: 2848 at kernel/sched/deadline.c:794
setup_new_dl_entity+0x13e/0x180
...
Call Trace:
 <TASK>
 ? show_trace_log_lvl+0x1c4/0x2df
 ? enqueue_dl_entity+0x631/0x6e0
 ? setup_new_dl_entity+0x13e/0x180
 ? __warn+0x7e/0xd0
 ? report_bug+0x11a/0x1a0
 ? handle_bug+0x3c/0x70
 ? exc_invalid_op+0x14/0x70
 ? asm_exc_invalid_op+0x16/0x20
 enqueue_dl_entity+0x631/0x6e0
 enqueue_task_dl+0x7d/0x120
 __do_set_cpus_allowed+0xe3/0x280
 __set_cpus_allowed_ptr_locked+0x140/0x1d0
 __set_cpus_allowed_ptr+0x54/0xa0
 migrate_enable+0x7e/0x150
 rt_spin_unlock+0x1c/0x90
 group_send_sig_info+0xf7/0x1a0
 ? kill_pid_info+0x1f/0x1d0
 kill_pid_info+0x78/0x1d0
 kill_proc_info+0x5b/0x110
 __x64_sys_kill+0x93/0xc0
 do_syscall_64+0x5c/0xf0
 entry_SYSCALL_64_after_hwframe+0x6e/0x76
 RIP: 0033:0x7f0dab31f92b

This warning occurs because set_cpus_allowed dequeues and enqueues tasks with the ENQUEUE_RESTORE flag set. If the task is boosted, the warning is triggered. A boosted task already had its parameters set by rt_mutex_setprio, and a new call to setup_new_dl_entity is unnecessary, hence the WARN_ON call.

Check if we are requeueing a boosted task and avoid calling setup_new_dl_entity if that's the case.

Quoted 7 lines
Quoted 2 lines
Fixes 2279f540ea7d ("sched/deadline: Fix priority inheritance with multiple scheduling classes")

I believe your fix makes sense to me. I only wonder if however it actually fixes 295d6d5e37360 ("sched/deadline: Fix switching to -deadline") instead of the change you reference above?

That makes more sense, thanks.

Quoted 3 lines

Thanks,
Juri

#4 lore raw

PATCH 2/3 sched/deadline: avoid redundant check for boosted task

In reply to Wander Lairson Costa

enqueue_dl_entity only calls setup_new_dl_entity if the task is not boosted, so the WARN_ON check is unnecessary.

---
 kernel/sched/deadline.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)
kernel/sched/deadline.c +8 -3
index 312e8fa7ce94..908d5ce79425 100644
@@ -785,12 +785,11 @@ static inline void replenish_dl_new_period(struct sched_dl_entity *dl_se,
  * one, and to (try to!) reconcile itself with its own scheduling
  * parameters.
  */
-static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+static inline void __setup_new_dl_entity(struct sched_dl_entity *dl_se)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
 	struct rq *rq = rq_of_dl_rq(dl_rq);
 
-	WARN_ON(is_dl_boosted(dl_se));
 	WARN_ON(dl_time_before(rq_clock(rq), dl_se->deadline));
 
 	/*
@@ -809,6 +808,12 @@ static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
 	replenish_dl_new_period(dl_se, rq);
 }
 
+static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+{
+	WARN_ON(is_dl_boosted(dl_se));
+	__setup_new_dl_entity(dl_se);
+}
+
 /*
  * Pure Earliest Deadline First (EDF) scheduling does not deal with the
  * possibility of a entity lasting more than what it declared, and thus
@@ -1755,7 +1760,7 @@ enqueue_dl_entity(struct sched_dl_entity *dl_se, int flags)
 	} else if ((flags & ENQUEUE_RESTORE) &&
 		   !is_dl_boosted(dl_se) &&
 		   dl_time_before(dl_se->deadline, rq_clock(rq_of_dl_se(dl_se)))) {
-		setup_new_dl_entity(dl_se);
+		__setup_new_dl_entity(dl_se);
 	}
 
 	__enqueue_dl_entity(dl_se);
Signature · 1 line
2.45.2
maintainer
#5 lore raw

Re: [PATCH 2/3] sched/deadline: avoid redundant check for boosted task

Hi Wander,

Quoted 27 lines · Wander Lairson Costa

enqueue_dl_entity only calls setup_new_dl_entity if the task is not boosted, so the WARN_ON check is unnecessary.

---
 kernel/sched/deadline.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)
kernel/sched/deadline.c +1 -2
index 312e8fa7ce94..908d5ce79425 100644
@@ -785,12 +785,11 @@ static inline void replenish_dl_new_period(struct sched_dl_entity *dl_se,
  * one, and to (try to!) reconcile itself with its own scheduling
  * parameters.
  */
-static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+static inline void __setup_new_dl_entity(struct sched_dl_entity *dl_se)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
 	struct rq *rq = rq_of_dl_rq(dl_rq);
 
-	WARN_ON(is_dl_boosted(dl_se));
 	WARN_ON(dl_time_before(rq_clock(rq), dl_se->deadline));
 
 	/*

kernel/sched/deadline.c static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se) · line 816

Quoted 9 lines
kernel/sched/deadline.c +6 -0
@@ -809,6 +808,12 @@ static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
 	replenish_dl_new_period(dl_se, rq);
 }
 
+static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+{
+	WARN_ON(is_dl_boosted(dl_se));
+	__setup_new_dl_entity(dl_se);
+}
+

So, the other call path is from dl_server_start() and for this we know the entity is not boosted either. We could probably just remove the WARN_ON w/o the additional wrapper function. That said, considering it's not fast path, I wonder if we actually want to leave the WARN_ON where it is, so that we can catch potential future erroneous usages?

Thanks,
Juri

#6 lore raw

Re: [PATCH 2/3] sched/deadline: avoid redundant check for boosted task

Quoted 34 lines · Juri Lelli

Hi Wander,

Quoted 30 lines · Wander Lairson Costa

enqueue_dl_entity only calls setup_new_dl_entity if the task is not boosted, so the WARN_ON check is unnecessary.

---
 kernel/sched/deadline.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)
kernel/sched/deadline.c +1 -2
index 312e8fa7ce94..908d5ce79425 100644
@@ -785,12 +785,11 @@ static inline void replenish_dl_new_period(struct sched_dl_entity *dl_se,
  * one, and to (try to!) reconcile itself with its own scheduling
  * parameters.
  */
-static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+static inline void __setup_new_dl_entity(struct sched_dl_entity *dl_se)
 {
      struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
      struct rq *rq = rq_of_dl_rq(dl_rq);
 
-     WARN_ON(is_dl_boosted(dl_se));
      WARN_ON(dl_time_before(rq_clock(rq), dl_se->deadline));
 
      /*
@@ -809,6 +808,12 @@ static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
      replenish_dl_new_period(dl_se, rq);
 }
 
Quoted 13 lines

kernel/sched/deadline.c static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se) · line 816

Quoted 6 lines
kernel/sched/deadline.c +6 -0
+static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+{
+     WARN_ON(is_dl_boosted(dl_se));
+     __setup_new_dl_entity(dl_se);
+}
+

So, the other call path is from dl_server_start() and for this we know the entity is not boosted either. We could probably just remove the WARN_ON w/o the additional wrapper function. That said, considering it's not fast path, I wonder if we actually want to leave the WARN_ON where it is, so that we can catch potential future erroneous usages?

Yeah, if you feel the patch is not worth it, I am more in favor of dropping the patch than removing the WARN_ON.

Quoted 3 lines

Thanks,
Juri

maintainer
#7 lore raw

Re: [PATCH 2/3] sched/deadline: avoid redundant check for boosted task

Quoted 49 lines · Wander Lairson Costa
Quoted 34 lines · Juri Lelli

Hi Wander,

Quoted 30 lines · Wander Lairson Costa

enqueue_dl_entity only calls setup_new_dl_entity if the task is not boosted, so the WARN_ON check is unnecessary.

---
 kernel/sched/deadline.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)
kernel/sched/deadline.c +1 -2
index 312e8fa7ce94..908d5ce79425 100644
@@ -785,12 +785,11 @@ static inline void replenish_dl_new_period(struct sched_dl_entity *dl_se,
  * one, and to (try to!) reconcile itself with its own scheduling
  * parameters.
  */
-static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+static inline void __setup_new_dl_entity(struct sched_dl_entity *dl_se)
 {
      struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
      struct rq *rq = rq_of_dl_rq(dl_rq);
 
-     WARN_ON(is_dl_boosted(dl_se));
      WARN_ON(dl_time_before(rq_clock(rq), dl_se->deadline));
 
      /*
@@ -809,6 +808,12 @@ static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
      replenish_dl_new_period(dl_se, rq);
 }
 
Quoted 13 lines

kernel/sched/deadline.c static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se) · line 816

Quoted 6 lines
kernel/sched/deadline.c +6 -0
+static inline void setup_new_dl_entity(struct sched_dl_entity *dl_se)
+{
+     WARN_ON(is_dl_boosted(dl_se));
+     __setup_new_dl_entity(dl_se);
+}
+

So, the other call path is from dl_server_start() and for this we know the entity is not boosted either. We could probably just remove the WARN_ON w/o the additional wrapper function. That said, considering it's not fast path, I wonder if we actually want to leave the WARN_ON where it is, so that we can catch potential future erroneous usages?

Quoted 2 lines

Yeah, if you feel the patch is not worth it, I am more in favor of dropping the patch than removing the WARN_ON.

Think we can drop it yes.

Thanks,
Juri

#8 lore raw

PATCH 3/3 sched/deadline: Consolidate Timer Cancellation

In reply to Wander Lairson Costa

After commit b58652db66c9 ("sched/deadline: Fix task_struct reference leak"), I identified additional calls to hrtimer_try_to_cancel that might also require a dl_server check. It remains unclear whether this omission was intentional or accidental in those contexts.

This patch consolidates the timer cancellation logic into dedicated functions, ensuring consistent behavior across all calls. Additionally, it reduces code duplication and improves overall code cleanliness.

Note the use of the __always_inline keyword. In some instances, we have a task_struct pointer, dereference the dl member, and then use the container_of macro to retrieve the task_struct pointer again. By inlining the code, the compiler can potentially optimize out this redundant round trip.

---
 kernel/sched/deadline.c | 44 ++++++++++++++++++++++++++---------------
 1 file changed, 28 insertions(+), 16 deletions(-)
kernel/sched/deadline.c +28 -16
index 908d5ce79425..8b0bbade2dcb 100644
@@ -320,6 +320,29 @@ void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
 		__sub_running_bw(dl_se->dl_bw, dl_rq);
 }
 
+static __always_inline
+void cancel_timer(struct sched_dl_entity *dl_se, struct hrtimer *timer)
+{
+	/*
+	 * If the timer callback was running (hrtimer_try_to_cancel == -1),
+	 * it will eventually call put_task_struct().
+	 */
+	if (hrtimer_try_to_cancel(timer) == 1 && !dl_server(dl_se))
+		put_task_struct(dl_task_of(dl_se));
+}
+
+static __always_inline
+void cancel_dl_timer(struct sched_dl_entity *dl_se)
+{
+	cancel_timer(dl_se, &dl_se->dl_timer);
+}
+
+static __always_inline
+void cancel_inactive_timer(struct sched_dl_entity *dl_se)
+{
+	cancel_timer(dl_se, &dl_se->inactive_timer);
+}
+
 static void dl_change_utilization(struct task_struct *p, u64 new_bw)
 {
 	struct rq *rq;
@@ -340,8 +363,7 @@ static void dl_change_utilization(struct task_struct *p, u64 new_bw)
 		 * will not touch the rq's active utilization,
 		 * so we are still safe.
 		 */
-		if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
-			put_task_struct(p);
+		cancel_inactive_timer(&p->dl);
 	}
 	__sub_rq_bw(p->dl.dl_bw, &rq->dl);
 	__add_rq_bw(new_bw, &rq->dl);
@@ -490,10 +512,7 @@ static void task_contending(struct sched_dl_entity *dl_se, int flags)
 		 * will not touch the rq's active utilization,
 		 * so we are still safe.
 		 */
-		if (hrtimer_try_to_cancel(&dl_se->inactive_timer) == 1) {
-			if (!dl_server(dl_se))
-				put_task_struct(dl_task_of(dl_se));
-		}
+		cancel_inactive_timer(dl_se);
 	} else {
 		/*
 		 * Since "dl_non_contending" is not set, the
@@ -1810,13 +1829,8 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 			 * The replenish timer needs to be canceled. No
 			 * problem if it fires concurrently: boosted threads
 			 * are ignored in dl_task_timer().
-			 *
-			 * If the timer callback was running (hrtimer_try_to_cancel == -1),
-			 * it will eventually call put_task_struct().
 			 */
-			if (hrtimer_try_to_cancel(&p->dl.dl_timer) == 1 &&
-			    !dl_server(&p->dl))
-				put_task_struct(p);
+			cancel_dl_timer(&p->dl);
 			p->dl.dl_throttled = 0;
 		}
 	} else if (!dl_prio(p->normal_prio)) {
@@ -1981,8 +1995,7 @@ static void migrate_task_rq_dl(struct task_struct *p, int new_cpu __maybe_unused
 		 * will not touch the rq's active utilization,
 		 * so we are still safe.
 		 */
-		if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
-			put_task_struct(p);
+		cancel_inactive_timer(&p->dl);
 	}
 	sub_rq_bw(&p->dl, &rq->dl);
 	rq_unlock(rq, &rf);
@@ -2737,8 +2750,7 @@ static void switched_from_dl(struct rq *rq, struct task_struct *p)
  */
 static void switched_to_dl(struct rq *rq, struct task_struct *p)
 {
-	if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
-		put_task_struct(p);
+	cancel_inactive_timer(&p->dl);
 
 	/*
 	 * In case a task is setscheduled to SCHED_DEADLINE we need to keep
Signature · 1 line
2.45.2
maintainer
#9 lore raw

Re: [PATCH 3/3] sched/deadline: Consolidate Timer Cancellation

Hi Wander,

Guess this might help yes, only a minor naming suggestion below.

Quoted 32 lines · Wander Lairson Costa

After commit b58652db66c9 ("sched/deadline: Fix task_struct reference leak"), I identified additional calls to hrtimer_try_to_cancel that might also require a dl_server check. It remains unclear whether this omission was intentional or accidental in those contexts.

This patch consolidates the timer cancellation logic into dedicated functions, ensuring consistent behavior across all calls. Additionally, it reduces code duplication and improves overall code cleanliness.

Note the use of the __always_inline keyword. In some instances, we have a task_struct pointer, dereference the dl member, and then use the container_of macro to retrieve the task_struct pointer again. By inlining the code, the compiler can potentially optimize out this redundant round trip.

---
 kernel/sched/deadline.c | 44 ++++++++++++++++++++++++++---------------
 1 file changed, 28 insertions(+), 16 deletions(-)
kernel/sched/deadline.c +3 -0
index 908d5ce79425..8b0bbade2dcb 100644
@@ -320,6 +320,29 @@ void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
 		__sub_running_bw(dl_se->dl_bw, dl_rq);
 }
 
+static __always_inline
+void cancel_timer(struct sched_dl_entity *dl_se, struct hrtimer *timer)
+{

kernel/sched/deadline.c void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq) · line 335

Quoted 10 lines
kernel/sched/deadline.c +10 -0
+	/*
+	 * If the timer callback was running (hrtimer_try_to_cancel == -1),
+	 * it will eventually call put_task_struct().
+	 */
+	if (hrtimer_try_to_cancel(timer) == 1 && !dl_server(dl_se))
+		put_task_struct(dl_task_of(dl_se));
+}
+
+static __always_inline
+void cancel_dl_timer(struct sched_dl_entity *dl_se)

Maybe we could call the above cancel_replenish_timer

kernel/sched/deadline.c void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq) · line 337

Quoted 2 lines
kernel/sched/deadline.c +2 -0
+{
+	cancel_timer(dl_se, &dl_se->dl_timer);

and this one cancel_dl_timer?

Best,
Juri

#10 lore raw

Re: [PATCH 3/3] sched/deadline: Consolidate Timer Cancellation

Quoted 51 lines · Juri Lelli

Hi Wander,

Guess this might help yes, only a minor naming suggestion below.

Quoted 32 lines · Wander Lairson Costa

After commit b58652db66c9 ("sched/deadline: Fix task_struct reference leak"), I identified additional calls to hrtimer_try_to_cancel that might also require a dl_server check. It remains unclear whether this omission was intentional or accidental in those contexts.

This patch consolidates the timer cancellation logic into dedicated functions, ensuring consistent behavior across all calls. Additionally, it reduces code duplication and improves overall code cleanliness.

Note the use of the __always_inline keyword. In some instances, we have a task_struct pointer, dereference the dl member, and then use the container_of macro to retrieve the task_struct pointer again. By inlining the code, the compiler can potentially optimize out this redundant round trip.

---
 kernel/sched/deadline.c | 44 ++++++++++++++++++++++++++---------------
 1 file changed, 28 insertions(+), 16 deletions(-)
kernel/sched/deadline.c +3 -0
index 908d5ce79425..8b0bbade2dcb 100644
@@ -320,6 +320,29 @@ void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
              __sub_running_bw(dl_se->dl_bw, dl_rq);
 }
 
+static __always_inline
+void cancel_timer(struct sched_dl_entity *dl_se, struct hrtimer *timer)
+{

kernel/sched/deadline.c void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq) · line 335

Quoted 10 lines
kernel/sched/deadline.c +10 -0
+     /*
+      * If the timer callback was running (hrtimer_try_to_cancel == -1),
+      * it will eventually call put_task_struct().
+      */
+     if (hrtimer_try_to_cancel(timer) == 1 && !dl_server(dl_se))
+             put_task_struct(dl_task_of(dl_se));
+}
+
+static __always_inline
+void cancel_dl_timer(struct sched_dl_entity *dl_se)

Maybe we could call the above cancel_replenish_timer

Quoted 5 lines

kernel/sched/deadline.c void sub_running_bw(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq) · line 337

Quoted 2 lines
kernel/sched/deadline.c +2 -0
+{
+     cancel_timer(dl_se, &dl_se->dl_timer);

and this one cancel_dl_timer?

That makes sense, I will change it in v2.

Quoted 3 lines

Best,
Juri