* [PATCH 2/3] sched/cputime: code cleanup in irqtime_account_process_tick
2020-01-02 10:07 [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick Alex Shi
@ 2020-01-02 10:07 ` Alex Shi
2020-01-06 15:53 ` Frederic Weisbecker
2020-01-02 10:07 ` [PATCH 3/3] sched/cputime: cleanup account_process_tick/account_idle_ticks Alex Shi
` (2 subsequent siblings)
3 siblings, 1 reply; 8+ messages in thread
From: Alex Shi @ 2020-01-02 10:07 UTC (permalink / raw)
Cc: Ingo Molnar, Peter Zijlstra, Frederic Weisbecker, Wanpeng Li,
Anna-Maria Gleixner, Thomas Gleixner, linux-kernel
In this func, since account_system_time() considers guest time account
and other system time. we could fold the account_guest_time into
account_system_time() to simply the code.
Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Anna-Maria Gleixner <anna-maria@linutronix.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
---
kernel/sched/cputime.c | 9 +++------
1 file changed, 3 insertions(+), 6 deletions(-)
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index cff3e656566d..46b837e94fce 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -381,13 +381,10 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
account_system_index_time(p, cputime, CPUTIME_SOFTIRQ);
} else if (user_tick) {
account_user_time(p, cputime);
- } else if (p == this_rq()->idle) {
+ } else if ((p != this_rq()->idle) || (irq_count() != HARDIRQ_OFFSET))
+ account_system_time(p, HARDIRQ_OFFSET, cputime);
+ else
account_idle_time(cputime);
- } else if (p->flags & PF_VCPU) { /* System time or guest time */
- account_guest_time(p, cputime);
- } else {
- account_system_index_time(p, cputime, CPUTIME_SYSTEM);
- }
}
static void irqtime_account_idle_ticks(int ticks)
--
1.8.3.1
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH 2/3] sched/cputime: code cleanup in irqtime_account_process_tick
2020-01-02 10:07 ` [PATCH 2/3] sched/cputime: code cleanup " Alex Shi
@ 2020-01-06 15:53 ` Frederic Weisbecker
2020-01-07 9:13 ` Peter Zijlstra
0 siblings, 1 reply; 8+ messages in thread
From: Frederic Weisbecker @ 2020-01-06 15:53 UTC (permalink / raw)
To: Alex Shi
Cc: Ingo Molnar, Peter Zijlstra, Frederic Weisbecker, Wanpeng Li,
Anna-Maria Gleixner, Thomas Gleixner, linux-kernel
On Thu, Jan 02, 2020 at 06:07:53PM +0800, Alex Shi wrote:
> In this func, since account_system_time() considers guest time account
> and other system time. we could fold the account_guest_time into
> account_system_time() to simply the code.
>
> Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Frederic Weisbecker <fweisbec@gmail.com>
> Cc: Wanpeng Li <wanpeng.li@hotmail.com>
> Cc: Anna-Maria Gleixner <anna-maria@linutronix.de>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: linux-kernel@vger.kernel.org
> ---
> kernel/sched/cputime.c | 9 +++------
> 1 file changed, 3 insertions(+), 6 deletions(-)
>
> diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
> index cff3e656566d..46b837e94fce 100644
> --- a/kernel/sched/cputime.c
> +++ b/kernel/sched/cputime.c
> @@ -381,13 +381,10 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
> account_system_index_time(p, cputime, CPUTIME_SOFTIRQ);
> } else if (user_tick) {
> account_user_time(p, cputime);
> - } else if (p == this_rq()->idle) {
> + } else if ((p != this_rq()->idle) || (irq_count() != HARDIRQ_OFFSET))
> + account_system_time(p, HARDIRQ_OFFSET, cputime);
> + else
I fear we can't really play the exact same game as account_process_tick() here.
Since this is irqtime precise accounting, we have already computed the
irqtime delta in account_other_time() (or we will at some point in the future)
and substracted it from the ticks to account. This means that the remaining cputime
to account has to be either utime/stime/gtime/idle-time but not interrupt time, or
we may account interrupt time twice. And account_system_time() tries to account
irq time, for example if we interrupt a softirq.
Thanks.
> account_idle_time(cputime);
> - } else if (p->flags & PF_VCPU) { /* System time or guest time */
> - account_guest_time(p, cputime);
> - } else {
> - account_system_index_time(p, cputime, CPUTIME_SYSTEM);
> - }
> }
>
> static void irqtime_account_idle_ticks(int ticks)
> --
> 1.8.3.1
>
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH 2/3] sched/cputime: code cleanup in irqtime_account_process_tick
2020-01-06 15:53 ` Frederic Weisbecker
@ 2020-01-07 9:13 ` Peter Zijlstra
2020-01-09 13:30 ` Alex Shi
0 siblings, 1 reply; 8+ messages in thread
From: Peter Zijlstra @ 2020-01-07 9:13 UTC (permalink / raw)
To: Frederic Weisbecker
Cc: Alex Shi, Ingo Molnar, Frederic Weisbecker, Wanpeng Li,
Anna-Maria Gleixner, Thomas Gleixner, linux-kernel
On Mon, Jan 06, 2020 at 04:53:51PM +0100, Frederic Weisbecker wrote:
> On Thu, Jan 02, 2020 at 06:07:53PM +0800, Alex Shi wrote:
> > In this func, since account_system_time() considers guest time account
> > and other system time. we could fold the account_guest_time into
> > account_system_time() to simply the code.
> >
> > Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
> > Cc: Ingo Molnar <mingo@redhat.com>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: Frederic Weisbecker <fweisbec@gmail.com>
> > Cc: Wanpeng Li <wanpeng.li@hotmail.com>
> > Cc: Anna-Maria Gleixner <anna-maria@linutronix.de>
> > Cc: Thomas Gleixner <tglx@linutronix.de>
> > Cc: linux-kernel@vger.kernel.org
> > ---
> > kernel/sched/cputime.c | 9 +++------
> > 1 file changed, 3 insertions(+), 6 deletions(-)
> >
> > diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
> > index cff3e656566d..46b837e94fce 100644
> > --- a/kernel/sched/cputime.c
> > +++ b/kernel/sched/cputime.c
> > @@ -381,13 +381,10 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
> > account_system_index_time(p, cputime, CPUTIME_SOFTIRQ);
> > } else if (user_tick) {
> > account_user_time(p, cputime);
> > - } else if (p == this_rq()->idle) {
> > + } else if ((p != this_rq()->idle) || (irq_count() != HARDIRQ_OFFSET))
> > + account_system_time(p, HARDIRQ_OFFSET, cputime);
> > + else
>
> I fear we can't really play the exact same game as account_process_tick() here.
> Since this is irqtime precise accounting, we have already computed the
> irqtime delta in account_other_time() (or we will at some point in the future)
> and substracted it from the ticks to account. This means that the remaining cputime
> to account has to be either utime/stime/gtime/idle-time but not interrupt time, or
> we may account interrupt time twice. And account_system_time() tries to account
> irq time, for example if we interrupt a softirq.
OK, I've dropped 2 and 3. Thanks Frederic!
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH 2/3] sched/cputime: code cleanup in irqtime_account_process_tick
2020-01-07 9:13 ` Peter Zijlstra
@ 2020-01-09 13:30 ` Alex Shi
0 siblings, 0 replies; 8+ messages in thread
From: Alex Shi @ 2020-01-09 13:30 UTC (permalink / raw)
To: Peter Zijlstra, Frederic Weisbecker
Cc: Ingo Molnar, Frederic Weisbecker, Wanpeng Li,
Anna-Maria Gleixner, Thomas Gleixner, linux-kernel
>> I fear we can't really play the exact same game as account_process_tick() here.
>> Since this is irqtime precise accounting, we have already computed the
>> irqtime delta in account_other_time() (or we will at some point in the future)
>> and substracted it from the ticks to account. This means that the remaining cputime
>> to account has to be either utime/stime/gtime/idle-time but not interrupt time, or
>> we may account interrupt time twice. And account_system_time() tries to account
>> irq time, for example if we interrupt a softirq.
>
> OK, I've dropped 2 and 3. Thanks Frederic!
>
Hi Frederic & Peter,
Thanks a lot for the comments and review!
It's my fault to mess up the account_system_time details. And seems there is no easy way to replace irqtime_account_process_tick or account_process_tick with each other.
but on the other side, the account_idle_ticks could be replaced by irqtime_account_process_tick, or at least to remove irqtime_account_idle_ticks function. Any comments?
Thanks
Alex
---
From 7073e60babc3b42a987b4e89f380956887734233 Mon Sep 17 00:00:00 2001
From: Alex Shi <alex.shi@linux.alibaba.com>
Date: Thu, 9 Jan 2020 20:32:55 +0800
Subject: [PATCH] sched/cputime: remove irqtime_account_idle_ticks
irqtime_account_idle_ticks and irqtime_account_process_tick use in same
condition. We don't bother to name and use a irqtime_account_idle_ticks
for only one calling. Remove the function to simply code and reduce a
bit object size of kernel.
And further more, we could replace account_idle_ticks by
irqtime_account_process_tick too. But feed and check 'current' looks weird.
So this is ok.
Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
---
kernel/sched/cputime.c | 7 +------
1 file changed, 1 insertion(+), 6 deletions(-)
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index cff3e656566d..17640d145e44 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -390,12 +390,7 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
}
}
-static void irqtime_account_idle_ticks(int ticks)
-{
- irqtime_account_process_tick(current, 0, ticks);
-}
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
-static inline void irqtime_account_idle_ticks(int ticks) { }
static inline void irqtime_account_process_tick(struct task_struct *p, int user_tick,
int nr_ticks) { }
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
@@ -505,7 +500,7 @@ void account_idle_ticks(unsigned long ticks)
u64 cputime, steal;
if (sched_clock_irqtime) {
- irqtime_account_idle_ticks(ticks);
+ irqtime_account_process_tick(current, 0, ticks);
return;
}
--
1.8.3.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH 3/3] sched/cputime: cleanup account_process_tick/account_idle_ticks
2020-01-02 10:07 [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick Alex Shi
2020-01-02 10:07 ` [PATCH 2/3] sched/cputime: code cleanup " Alex Shi
@ 2020-01-02 10:07 ` Alex Shi
2020-01-06 14:55 ` [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick Peter Zijlstra
2020-01-17 10:08 ` [tip: sched/core] " tip-bot2 for Alex Shi
3 siblings, 0 replies; 8+ messages in thread
From: Alex Shi @ 2020-01-02 10:07 UTC (permalink / raw)
Cc: Ingo Molnar, Peter Zijlstra, Frederic Weisbecker, Wanpeng Li,
Anna-Maria Gleixner, Thomas Gleixner, linux-kernel
The irqtime_account_process_tick() was introduced for precise ns irq
time account from commit abb74cefa9c6 ("sched: Export ns irqtimes
through /proc/stat") which is necessary since account_process_tick()
still use jiffes at that time.
Now account_process_tick/account_idle_ticks functions do the actual
same actions as irqtime_account_process_tick() to account ns precision
cputime when !sched_clock_irqtime. which could be replaced by
irqtime_account_process_tick();
So we move out irqtime_account_process_tick from IRQ_TIME_ACCOUNTING
config, let it work w/o IRQ_TIME_ACCOUNTING. and remove the duplicated
code in account_process_tick/account_idle_ticks.
And furthmore I removed the function account_idle_ticks by a directly
call.
It can simplify the code, also reduce a bit object size.
Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Anna-Maria Gleixner <anna-maria@linutronix.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
---
include/linux/kernel_stat.h | 2 +-
kernel/sched/cputime.c | 58 ++-------------------------------------------
kernel/time/tick-sched.c | 2 +-
3 files changed, 4 insertions(+), 58 deletions(-)
diff --git a/include/linux/kernel_stat.h b/include/linux/kernel_stat.h
index 89f0745c096d..e924bdb8c874 100644
--- a/include/linux/kernel_stat.h
+++ b/include/linux/kernel_stat.h
@@ -113,6 +113,6 @@ static inline void account_process_tick(struct task_struct *tsk, int user)
extern void account_process_tick(struct task_struct *, int user);
#endif
-extern void account_idle_ticks(unsigned long ticks);
+extern void irqtime_account_process_tick(struct task_struct *, int , int);
#endif /* _LINUX_KERNEL_STAT_H */
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index 46b837e94fce..0e0c74c1b3c9 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -332,7 +332,6 @@ void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times)
rcu_read_unlock();
}
-#ifdef CONFIG_IRQ_TIME_ACCOUNTING
/*
* Account a tick to a process and cpustat
* @p: the process that the CPU time gets accounted to
@@ -354,7 +353,7 @@ void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times)
* p->stime and friends are only updated on system time and not on irq
* softirq as those do not count in task exec_runtime any more.
*/
-static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
+void irqtime_account_process_tick(struct task_struct *p, int user_tick,
int ticks)
{
u64 other, cputime = TICK_NSEC * ticks;
@@ -387,16 +386,6 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
account_idle_time(cputime);
}
-static void irqtime_account_idle_ticks(int ticks)
-{
- irqtime_account_process_tick(current, 0, ticks);
-}
-#else /* CONFIG_IRQ_TIME_ACCOUNTING */
-static inline void irqtime_account_idle_ticks(int ticks) { }
-static inline void irqtime_account_process_tick(struct task_struct *p, int user_tick,
- int nr_ticks) { }
-#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
-
/*
* Use precise platform statistics if available:
*/
@@ -467,53 +456,10 @@ void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st)
*/
void account_process_tick(struct task_struct *p, int user_tick)
{
- u64 cputime, steal;
-
if (vtime_accounting_enabled_this_cpu())
return;
- if (sched_clock_irqtime) {
- irqtime_account_process_tick(p, user_tick, 1);
- return;
- }
-
- cputime = TICK_NSEC;
- steal = steal_account_process_time(ULONG_MAX);
-
- if (steal >= cputime)
- return;
-
- cputime -= steal;
-
- if (user_tick)
- account_user_time(p, cputime);
- else if ((p != this_rq()->idle) || (irq_count() != HARDIRQ_OFFSET))
- account_system_time(p, HARDIRQ_OFFSET, cputime);
- else
- account_idle_time(cputime);
-}
-
-/*
- * Account multiple ticks of idle time.
- * @ticks: number of stolen ticks
- */
-void account_idle_ticks(unsigned long ticks)
-{
- u64 cputime, steal;
-
- if (sched_clock_irqtime) {
- irqtime_account_idle_ticks(ticks);
- return;
- }
-
- cputime = ticks * TICK_NSEC;
- steal = steal_account_process_time(ULONG_MAX);
-
- if (steal >= cputime)
- return;
-
- cputime -= steal;
- account_idle_time(cputime);
+ irqtime_account_process_tick(p, user_tick, 1);
}
/*
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 8b192e67aabc..92a0979633e9 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -1142,7 +1142,7 @@ static void tick_nohz_account_idle_ticks(struct tick_sched *ts)
* We might be one off. Do not randomly account a huge number of ticks!
*/
if (ticks && ticks < LONG_MAX)
- account_idle_ticks(ticks);
+ irqtime_account_process_tick(current, 0, ticks);
#endif
}
--
1.8.3.1
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick
2020-01-02 10:07 [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick Alex Shi
2020-01-02 10:07 ` [PATCH 2/3] sched/cputime: code cleanup " Alex Shi
2020-01-02 10:07 ` [PATCH 3/3] sched/cputime: cleanup account_process_tick/account_idle_ticks Alex Shi
@ 2020-01-06 14:55 ` Peter Zijlstra
2020-01-17 10:08 ` [tip: sched/core] " tip-bot2 for Alex Shi
3 siblings, 0 replies; 8+ messages in thread
From: Peter Zijlstra @ 2020-01-06 14:55 UTC (permalink / raw)
To: Alex Shi
Cc: Ingo Molnar, Frederic Weisbecker, Wanpeng Li,
Anna-Maria Gleixner, Thomas Gleixner, linux-kernel
Thanks, I think this looks like.
Frederic, if you have time, a second set of eyes would be appreciated.
^ permalink raw reply [flat|nested] 8+ messages in thread
* [tip: sched/core] sched/cputime: move rq parameter in irqtime_account_process_tick
2020-01-02 10:07 [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick Alex Shi
` (2 preceding siblings ...)
2020-01-06 14:55 ` [PATCH 1/3] sched/cputime: move rq parameter in irqtime_account_process_tick Peter Zijlstra
@ 2020-01-17 10:08 ` tip-bot2 for Alex Shi
3 siblings, 0 replies; 8+ messages in thread
From: tip-bot2 for Alex Shi @ 2020-01-17 10:08 UTC (permalink / raw)
To: linux-tip-commits; +Cc: Alex Shi, Peter Zijlstra (Intel), x86, LKML
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 9dec1b6949ae9509cdc3edb2d75fda39c9db9fa2
Gitweb: https://git.kernel.org/tip/9dec1b6949ae9509cdc3edb2d75fda39c9db9fa2
Author: Alex Shi <alex.shi@linux.alibaba.com>
AuthorDate: Thu, 02 Jan 2020 18:07:52 +08:00
Committer: Peter Zijlstra <peterz@infradead.org>
CommitterDate: Fri, 17 Jan 2020 10:19:21 +01:00
sched/cputime: move rq parameter in irqtime_account_process_tick
Every time we call irqtime_account_process_tick() is in a interrupt,
Every caller will get and assign a parameter rq = this_rq(), This is
unnecessary and increase the code size a little bit. Move the rq getting
action to irqtime_account_process_tick internally is better.
base with this patch
cputime.o 578792 bytes 577888 bytes
Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/1577959674-255537-1-git-send-email-alex.shi@linux.alibaba.com
---
kernel/sched/cputime.c | 15 ++++++---------
1 file changed, 6 insertions(+), 9 deletions(-)
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index d43318a..cff3e65 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -355,7 +355,7 @@ void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times)
* softirq as those do not count in task exec_runtime any more.
*/
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
- struct rq *rq, int ticks)
+ int ticks)
{
u64 other, cputime = TICK_NSEC * ticks;
@@ -381,7 +381,7 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
account_system_index_time(p, cputime, CPUTIME_SOFTIRQ);
} else if (user_tick) {
account_user_time(p, cputime);
- } else if (p == rq->idle) {
+ } else if (p == this_rq()->idle) {
account_idle_time(cputime);
} else if (p->flags & PF_VCPU) { /* System time or guest time */
account_guest_time(p, cputime);
@@ -392,14 +392,12 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
static void irqtime_account_idle_ticks(int ticks)
{
- struct rq *rq = this_rq();
-
- irqtime_account_process_tick(current, 0, rq, ticks);
+ irqtime_account_process_tick(current, 0, ticks);
}
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
static inline void irqtime_account_idle_ticks(int ticks) { }
static inline void irqtime_account_process_tick(struct task_struct *p, int user_tick,
- struct rq *rq, int nr_ticks) { }
+ int nr_ticks) { }
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
/*
@@ -473,13 +471,12 @@ void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st)
void account_process_tick(struct task_struct *p, int user_tick)
{
u64 cputime, steal;
- struct rq *rq = this_rq();
if (vtime_accounting_enabled_this_cpu())
return;
if (sched_clock_irqtime) {
- irqtime_account_process_tick(p, user_tick, rq, 1);
+ irqtime_account_process_tick(p, user_tick, 1);
return;
}
@@ -493,7 +490,7 @@ void account_process_tick(struct task_struct *p, int user_tick)
if (user_tick)
account_user_time(p, cputime);
- else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
+ else if ((p != this_rq()->idle) || (irq_count() != HARDIRQ_OFFSET))
account_system_time(p, HARDIRQ_OFFSET, cputime);
else
account_idle_time(cputime);
^ permalink raw reply related [flat|nested] 8+ messages in thread