Skip to contentSkip to conversation

sched: Clean up newidle_balance() and pick_next_task()

PATCH Superseded Chen Yu 7 participants 12 messages Apr 19, 2020 – Apr 22, 2020 Scheduler

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Chen Yu

author
#0 lore raw

cover sched: Clean up newidle_balance() and pick_next_task()

As Peter suggested, here are two minor code clean up in the scheduler:
1. turn newidle_balance() into a static function.
2. code extraction in pick_next_task() for future re-use.

Chen Yu (2):
sched: Make newidle_balance() static again
sched: Extract the code to put previous task in pick_next_task()

 kernel/sched/core.c  | 39 +++++++++++++++++++++++----------------
 kernel/sched/fair.c  |  4 +++-
 kernel/sched/sched.h |  4 ----
 3 files changed, 26 insertions(+), 21 deletions(-)
Signature · 1 line
2.20.1
author
#1 lore raw

PATCH 1/2 sched: Make newidle_balance() static again

After Commit 6e2df0581f56 ("sched: Fix pick_next_task() vs 'change' pattern race"), there is no need to expose newidle_balance() as it is only used within fair.c file. Change this function back to static again.

No functional change.

---
 kernel/sched/fair.c  | 4 +++-
 kernel/sched/sched.h | 4 ----
 2 files changed, 3 insertions(+), 5 deletions(-)
kernel/sched/fair.c +3 -1
index 02f323b85b6d..da7a505c28a6 100644
@@ -86,6 +86,8 @@ static unsigned int normalized_sysctl_sched_wakeup_granularity	= 1000000UL;
 
 const_debug unsigned int sysctl_sched_migration_cost	= 500000UL;
 
+static int newidle_balance(struct rq *this_rq, struct rq_flags *rf);
+
 int sched_thermal_decay_shift;
 static int __init setup_sched_thermal_decay_shift(char *str)
 {
@@ -10425,7 +10427,7 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }
  *     0 - failed, no new tasks
  *   > 0 - success, new (fair) tasks present
  */
-int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
+static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 {
 	unsigned long next_balance = jiffies + HZ;
 	int this_cpu = this_rq->cpu;
kernel/sched/sched.h +0 -4
index db3a57675ccf..be83f88495fb 100644
@@ -1504,14 +1504,10 @@ static inline void unregister_sched_domain_sysctl(void)
 }
 #endif
 
-extern int newidle_balance(struct rq *this_rq, struct rq_flags *rf);
-
 #else
 
 static inline void sched_ttwu_pending(void) { }
 
-static inline int newidle_balance(struct rq *this_rq, struct rq_flags *rf) { return 0; }
-
 #endif /* CONFIG_SMP */
 
 #include "stats.h"
Signature · 1 line
2.20.1
#2 lore raw

Re: [PATCH 1/2] sched: Make newidle_balance() static again

I love your patch! Yet something to improve:

Hi Chen,

I love your patch! Yet something to improve:

[auto build test ERROR on tip/sched/core]
[also build test ERROR on tip/auto-latest linus/master v5.7-rc2 next-20200421] [cannot apply to linux/master]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Chen-Yu/sched-Clean-up-newidle_balance-and-pick_next_task/20200421-161412
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 8f3d9f354286745c751374f5f1fcafee6b3f3136
config: nds32-defconfig (attached as .config)
compiler: nds32le-linux-gcc (GCC) 9.3.0

reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day GCC_VERSION=9.3.0 make.cross ARCH=nds32

If you fix the issue, kindly add following tag as appropriate

All errors (new ones prefixed by >>):

nds32le-linux-ld: kernel/sched/fair.o: in function `pick_next_task_fair':
fair.c:(.text+0x1384): undefined reference to `newidle_balance'

>> nds32le-linux-ld: fair.c:(.text+0x1388): undefined reference to `newidle_balance'

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

#3 lore raw

Re: [PATCH 1/2] sched: Make newidle_balance() static again

Hi LKP,
This has been fixed according to your previous report in Peter's branch.
https://lore.kernel.org/patchwork/patch/1228070/

Thanks,
Chenyu

-----Original Message-----
From: lkp <[email protected]>
Sent: Wednesday, April 22, 2020 3:23 AM
To: Chen, Yu C <[email protected]>
Cc: kbuild-all(a)lists.01.org
Subject: Re: [PATCH 1/2] sched: Make newidle_balance() static again

Hi Chen,

I love your patch! Yet something to improve:

[auto build test ERROR on tip/sched/core] [also build test ERROR on tip/auto-latest linus/master v5.7-rc2 next-20200421] [cannot apply to linux/master] [if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Chen-Yu/sched-Clean-up-newidle_balance-and-pick_next_task/20200421-161412
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 8f3d9f354286745c751374f5f1fcafee6b3f3136
config: nds32-defconfig (attached as .config)
compiler: nds32le-linux-gcc (GCC) 9.3.0

reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day GCC_VERSION=9.3.0 make.cross ARCH=nds32

If you fix the issue, kindly add following tag as appropriate

All errors (new ones prefixed by >>):

nds32le-linux-ld: kernel/sched/fair.o: in function `pick_next_task_fair':
fair.c:(.text+0x1384): undefined reference to `newidle_balance'

Quoted 1 line
Quoted 1 line

nds32le-linux-ld: fair.c:(.text+0x1388): undefined reference to `newidle_balance'

---
0-DAY CI Kernel Test Service, Intel Corporation https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

author
#4 lore raw

PATCH 2/2 sched: Extract the task putting code from pick_next_task()

In reply to Chen Yu

Introduce a new function finish_prev_task() to do the balance when necessary, and then put previous task back to the run queue. This function is extracted from pick_next_task() to prepare for future usage by other type of task picking logic.

No functional change.

---
 kernel/sched/core.c | 39 +++++++++++++++++++++++----------------
 1 file changed, 23 insertions(+), 16 deletions(-)
kernel/sched/core.c +23 -16
index 3a61a3b8eaa9..bf59a5cf030c 100644
@@ -3904,6 +3904,28 @@ static inline void schedule_debug(struct task_struct *prev, bool preempt)
 	schedstat_inc(this_rq()->sched_count);
 }
 
+static void finish_prev_task(struct rq *rq, struct task_struct *prev,
+			     struct rq_flags *rf)
+{
+	const struct sched_class *class;
+#ifdef CONFIG_SMP
+	/*
+	 * We must do the balancing pass before put_next_task(), such
+	 * that when we release the rq->lock the task is in the same
+	 * state as before we took rq->lock.
+	 *
+	 * We can terminate the balance pass as soon as we know there is
+	 * a runnable task of @class priority or higher.
+	 */
+	for_class_range(class, prev->sched_class, &idle_sched_class) {
+		if (class->balance(rq, prev, rf))
+			break;
+	}
+#endif
+
+	put_prev_task(rq, prev);
+}
+
 /*
  * Pick up the highest-prio task:
  */
@@ -3937,22 +3959,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 	}
 
 restart:
-#ifdef CONFIG_SMP
-	/*
-	 * We must do the balancing pass before put_next_task(), such
-	 * that when we release the rq->lock the task is in the same
-	 * state as before we took rq->lock.
-	 *
-	 * We can terminate the balance pass as soon as we know there is
-	 * a runnable task of @class priority or higher.
-	 */
-	for_class_range(class, prev->sched_class, &idle_sched_class) {
-		if (class->balance(rq, prev, rf))
-			break;
-	}
-#endif
-
-	put_prev_task(rq, prev);
+	finish_prev_task(rq, prev, rf);
 
 	for_each_class(class) {
 		p = class->pick_next_task(rq);
Signature · 1 line
2.20.1
#5 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

Quoted 18 lines · On Mon, 20 Apr 2020 00:31:52 +0800 Chen Yu <[email protected]> wrote:

Introduce a new function finish_prev_task() to do the balance when necessary, and then put previous task back to the run queue. This function is extracted from pick_next_task() to prepare for future usage by other type of task picking logic.

No functional change.

---
 kernel/sched/core.c | 39 +++++++++++++++++++++++----------------
 1 file changed, 23 insertions(+), 16 deletions(-)
kernel/sched/core.c +0 -0
index 3a61a3b8eaa9..bf59a5cf030c 100644

kernel/sched/core.c static inline void schedule_debug(struct task_struct *prev, bool preempt) · line 3913

Quoted 10 lines
kernel/sched/core.c +7 -0
@@ -3904,6 +3904,28 @@ static inline void schedule_debug(struct task_struct *prev, bool preempt)
 	schedstat_inc(this_rq()->sched_count);
 }
 
+static void finish_prev_task(struct rq *rq, struct task_struct *prev,
+			     struct rq_flags *rf)
+{
+	const struct sched_class *class;
+#ifdef CONFIG_SMP
+	/*
+	 * We must do the balancing pass before put_next_task(), such

I know this is just a cut and paste move, but I'm thinking that this comment is wrong. Shouldn't this be "put_prev_task()" as we have no "put_next_task()" function.

Quoted 29 lines
kernel/sched/core.c +15 -7
+	 * that when we release the rq->lock the task is in the same
+	 * state as before we took rq->lock.
+	 *
+	 * We can terminate the balance pass as soon as we know there is
+	 * a runnable task of @class priority or higher.
+	 */
+	for_class_range(class, prev->sched_class, &idle_sched_class) {
+		if (class->balance(rq, prev, rf))
+			break;
+	}
+#endif
+
+	put_prev_task(rq, prev);
+}
+
 /*
  * Pick up the highest-prio task:
  */
@@ -3937,22 +3959,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 	}
 
 restart:
-#ifdef CONFIG_SMP
-	/*
-	 * We must do the balancing pass before put_next_task(), such
-	 * that when we release the rq->lock the task is in the same
-	 * state as before we took rq->lock.
-	 *
-	 * We can terminate the balance pass as soon as we know there is

kernel/sched/core.c pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) · line 3962

Quoted 10 lines
kernel/sched/core.c +1 -9
-	 * a runnable task of @class priority or higher.
-	 */
-	for_class_range(class, prev->sched_class, &idle_sched_class) {
-		if (class->balance(rq, prev, rf))
-			break;
-	}
-#endif
-
-	put_prev_task(rq, prev);
+	finish_prev_task(rq, prev, rf);

I'm not sure I like the name of this function. Perhaps "balance_and_put_prev_task()"? Something more in kind to what the function does.

-- Steve

Quoted 3 lines
kernel/sched/core.c +0 -0
 	for_each_class(class) {
 		p = class->pick_next_task(rq);
#6 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

(There's a v2 at [email protected] but I think this still applies)

Quoted 4 lines · Steven Rostedt
Quoted 4 lines
kernel/sched/core.c +0 -0
@@ -3904,6 +3904,28 @@ static inline void schedule_debug(struct task_struct *prev, bool preempt)
     schedstat_inc(this_rq()->sched_count);
 }
 
Quoted 12 lines

kernel/sched/core.c static inline void schedule_debug(struct task_struct *prev, bool preempt) · line 3913

Quoted 7 lines
kernel/sched/core.c +7 -0
+static void finish_prev_task(struct rq *rq, struct task_struct *prev,
+			     struct rq_flags *rf)
+{
+	const struct sched_class *class;
+#ifdef CONFIG_SMP
+	/*
+	 * We must do the balancing pass before put_next_task(), such

I know this is just a cut and paste move, but I'm thinking that this comment is wrong. Shouldn't this be "put_prev_task()" as we have no "put_next_task()" function.

Oh, I think you're right.

Quoted 36 lines
Quoted 36 lines
kernel/sched/core.c +15 -13
+	 * that when we release the rq->lock the task is in the same
+	 * state as before we took rq->lock.
+	 *
+	 * We can terminate the balance pass as soon as we know there is
+	 * a runnable task of @class priority or higher.
+	 */
+	for_class_range(class, prev->sched_class, &idle_sched_class) {
+		if (class->balance(rq, prev, rf))
+			break;
+	}
+#endif
+
+	put_prev_task(rq, prev);
+}
+
 /*
  * Pick up the highest-prio task:
  */
@@ -3937,22 +3959,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
     }
 
 restart:
-#ifdef CONFIG_SMP
-	/*
-	 * We must do the balancing pass before put_next_task(), such
-	 * that when we release the rq->lock the task is in the same
-	 * state as before we took rq->lock.
-	 *
-	 * We can terminate the balance pass as soon as we know there is
-	 * a runnable task of @class priority or higher.
-	 */
-	for_class_range(class, prev->sched_class, &idle_sched_class) {
-		if (class->balance(rq, prev, rf))
-			break;
-	}
Quoted 9 lines

kernel/sched/core.c pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) · line 3962

Quoted 4 lines
kernel/sched/core.c +1 -3
-#endif
-
-	put_prev_task(rq, prev);
+	finish_prev_task(rq, prev, rf);

I'm not sure I like the name of this function. Perhaps "balance_and_put_prev_task()"? Something more in kind to what the function does.

The 'finish' thing isn't too far from the truth; it's the last thing we need to do with the prev task (in terms of sched bookkeeping, I mean) - and in Chen's defence ISTR Peter suggested that name.

Seeing as it's a "supercharged" put_prev_task(), I could live with the marginally shorter "put_prev_task_balance()".

Quoted 5 lines

-- Steve

Quoted 3 lines
kernel/sched/core.c +0 -0
     for_each_class(class) {
             p = class->pick_next_task(rq);
maintainer
#7 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

Quoted 7 lines · Valentin Schneider
Quoted 6 lines

kernel/sched/core.c pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) · line 3962

Quoted 1 line
kernel/sched/core.c +1 -0
+	finish_prev_task(rq, prev, rf);

I'm not sure I like the name of this function. Perhaps "balance_and_put_prev_task()"? Something more in kind to what the function does.

Quoted 6 lines

The 'finish' thing isn't too far from the truth; it's the last thing we need to do with the prev task (in terms of sched bookkeeping, I mean) - and in Chen's defence ISTR Peter suggested that name.

Seeing as it's a "supercharged" put_prev_task(), I could live with the marginally shorter "put_prev_task_balance()".

What Valentin said; it's the last put we do before picking a new task. Also, I don't like long names. That said, I'm open to short and appropriate suggestions.

#8 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

Quoted 10 lines · On Tue, 21 Apr 2020 01:13:55 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 6 lines

The 'finish' thing isn't too far from the truth; it's the last thing we need to do with the prev task (in terms of sched bookkeeping, I mean) - and in Chen's defence ISTR Peter suggested that name.

Seeing as it's a "supercharged" put_prev_task(), I could live with the marginally shorter "put_prev_task_balance()".

What Valentin said; it's the last put we do before picking a new task. Also, I don't like long names. That said, I'm open to short and appropriate suggestions.

I wont bikeshed this too much.

Is the "finish" more appropriate with the other use cases that are coming. I do like that "put_prev_task_balance()" too.

-- Steve

maintainer
#9 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

Quoted 11 lines · Steven Rostedt
Quoted 7 lines · On Tue, 21 Apr 2020 01:13:55 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 6 lines

The 'finish' thing isn't too far from the truth; it's the last thing we need to do with the prev task (in terms of sched bookkeeping, I mean) - and in Chen's defence ISTR Peter suggested that name.

Seeing as it's a "supercharged" put_prev_task(), I could live with the marginally shorter "put_prev_task_balance()".

Quoted 8 lines
Quoted 3 lines

What Valentin said; it's the last put we do before picking a new task. Also, I don't like long names. That said, I'm open to short and appropriate suggestions.

I wont bikeshed this too much.

Is the "finish" more appropriate with the other use cases that are coming. I do like that "put_prev_task_balance()" too.

This name looks reasonnable

Quoted 2 lines

-- Steve

author
#10 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

Quoted 16 lines · Vincent Guittot
Quoted 15 lines · Steven Rostedt
Quoted 10 lines · On Tue, 21 Apr 2020 01:13:55 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 6 lines

The 'finish' thing isn't too far from the truth; it's the last thing we need to do with the prev task (in terms of sched bookkeeping, I mean) - and in Chen's defence ISTR Peter suggested that name.

Seeing as it's a "supercharged" put_prev_task(), I could live with the marginally shorter "put_prev_task_balance()".

What Valentin said; it's the last put we do before picking a new task. Also, I don't like long names. That said, I'm open to short and appropriate suggestions.

Quoted 7 lines
Quoted 4 lines

I wont bikeshed this too much.

Is the "finish" more appropriate with the other use cases that are coming. I do like that "put_prev_task_balance()" too.

This name looks reasonnable

Okay, I'll change it to this name.

Thanks,
Chenyu

Quoted 2 lines
Quoted 2 lines

-- Steve

author
#11 lore raw

Re: [PATCH 2/2] sched: Extract the task putting code from pick_next_task()

In reply to Steven Rostedt
Quoted 24 lines · Steven Rostedt
Quoted 21 lines · On Mon, 20 Apr 2020 00:31:52 +0800 Chen Yu <[email protected]> wrote:

Introduce a new function finish_prev_task() to do the balance when necessary, and then put previous task back to the run queue. This function is extracted from pick_next_task() to prepare for future usage by other type of task picking logic.

No functional change.

---
 kernel/sched/core.c | 39 +++++++++++++++++++++++----------------
 1 file changed, 23 insertions(+), 16 deletions(-)
kernel/sched/core.c +0 -0
index 3a61a3b8eaa9..bf59a5cf030c 100644
@@ -3904,6 +3904,28 @@ static inline void schedule_debug(struct task_struct *prev, bool preempt)
 	schedstat_inc(this_rq()->sched_count);
 }
 
Quoted 13 lines

kernel/sched/core.c static inline void schedule_debug(struct task_struct *prev, bool preempt) · line 3913

Quoted 7 lines
kernel/sched/core.c +7 -0
+static void finish_prev_task(struct rq *rq, struct task_struct *prev,
+			     struct rq_flags *rf)
+{
+	const struct sched_class *class;
+#ifdef CONFIG_SMP
+	/*
+	 * We must do the balancing pass before put_next_task(), such

I know this is just a cut and paste move, but I'm thinking that this comment is wrong. Shouldn't this be "put_prev_task()" as we have no "put_next_task()" function.

Okay, I'll fix it in v3.

Quoted 6 lines

kernel/sched/core.c pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) · line 3962

Quoted 1 line
kernel/sched/core.c +1 -0
+	finish_prev_task(rq, prev, rf);

I'm not sure I like the name of this function. Perhaps "balance_and_put_prev_task()"? Something more in kind to what the function does.

Per the discussion, I think put_prev_task_balance() might be an appropriate one.

Thanks,
Chenyu

Quoted 6 lines

-- Steve

Quoted 3 lines
kernel/sched/core.c +0 -0
 	for_each_class(class) {
 		p = class->pick_next_task(rq);