[tip: sched/core] sched/cputime: Spare a seqcount lock/unlock cycle on context switch

From: tip-bot2 for Frederic Weisbecker
Date: Wed Oct 09 2019 - 08:59:52 EST


The following commit has been merged into the sched/core branch of tip:

Commit-ID: 8d495477d62e4397207f22a432fcaa86d9f2bc2d
Gitweb: https://git.kernel.org/tip/8d495477d62e4397207f22a432fcaa86d9f2bc2d
Author: Frederic Weisbecker <frederic@xxxxxxxxxx>
AuthorDate: Thu, 03 Oct 2019 18:17:45 +02:00
Committer: Ingo Molnar <mingo@xxxxxxxxxx>
CommitterDate: Wed, 09 Oct 2019 12:39:26 +02:00

sched/cputime: Spare a seqcount lock/unlock cycle on context switch

On context switch we are locking the vtime seqcount of the scheduling-out
task twice:

* On vtime_task_switch_common(), when we flush the pending vtime through
vtime_account_system()

* On arch_vtime_task_switch() to reset the vtime state.

This is pointless as these actions can be performed without the need
to unlock/lock in the middle. The reason these steps are separated is to
consolidate a very small amount of common code between
CONFIG_VIRT_CPU_ACCOUNTING_GEN and CONFIG_VIRT_CPU_ACCOUNTING_NATIVE.

Performance in this fast path is definitely a priority over artificial
code factorization so split the task switch code between GEN and
NATIVE and mutualize the parts than can run under a single seqcount
locked block.

As a side effect, vtime_account_idle() becomes included in the seqcount
protection. This happens to be a welcome preparation in order to
properly support kcpustat under vtime in the future and fetch
CPUTIME_IDLE without race.

Signed-off-by: Frederic Weisbecker <frederic@xxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Rik van Riel <riel@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Wanpeng Li <wanpengli@xxxxxxxxxxx>
Cc: Yauheni Kaliuta <yauheni.kaliuta@xxxxxxxxxx>
Link: https://lkml.kernel.org/r/20191003161745.28464-3-frederic@xxxxxxxxxx
Signed-off-by: Ingo Molnar <mingo@xxxxxxxxxx>
---
include/linux/vtime.h | 32 ++++++++++++++++----------------
kernel/sched/cputime.c | 30 +++++++++++++++++++-----------
2 files changed, 35 insertions(+), 27 deletions(-)

diff --git a/include/linux/vtime.h b/include/linux/vtime.h
index 2fd247f..d9160ab 100644
--- a/include/linux/vtime.h
+++ b/include/linux/vtime.h
@@ -14,8 +14,12 @@ struct task_struct;
* vtime_accounting_cpu_enabled() definitions/declarations
*/
#if defined(CONFIG_VIRT_CPU_ACCOUNTING_NATIVE)
+
static inline bool vtime_accounting_cpu_enabled(void) { return true; }
+extern void vtime_task_switch(struct task_struct *prev);
+
#elif defined(CONFIG_VIRT_CPU_ACCOUNTING_GEN)
+
/*
* Checks if vtime is enabled on some CPU. Cputime readers want to be careful
* in that case and compute the tickless cputime.
@@ -36,33 +40,29 @@ static inline bool vtime_accounting_cpu_enabled(void)

return false;
}
+
+extern void vtime_task_switch_generic(struct task_struct *prev);
+
+static inline void vtime_task_switch(struct task_struct *prev)
+{
+ if (vtime_accounting_cpu_enabled())
+ vtime_task_switch_generic(prev);
+}
+
#else /* !CONFIG_VIRT_CPU_ACCOUNTING */
+
static inline bool vtime_accounting_cpu_enabled(void) { return false; }
-#endif
+static inline void vtime_task_switch(struct task_struct *prev) { }

+#endif

/*
* Common vtime APIs
*/
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
-
-#ifdef __ARCH_HAS_VTIME_TASK_SWITCH
-extern void vtime_task_switch(struct task_struct *prev);
-#else
-extern void vtime_common_task_switch(struct task_struct *prev);
-static inline void vtime_task_switch(struct task_struct *prev)
-{
- if (vtime_accounting_cpu_enabled())
- vtime_common_task_switch(prev);
-}
-#endif /* __ARCH_HAS_VTIME_TASK_SWITCH */
-
extern void vtime_account_kernel(struct task_struct *tsk);
extern void vtime_account_idle(struct task_struct *tsk);
-
#else /* !CONFIG_VIRT_CPU_ACCOUNTING */
-
-static inline void vtime_task_switch(struct task_struct *prev) { }
static inline void vtime_account_kernel(struct task_struct *tsk) { }
#endif /* !CONFIG_VIRT_CPU_ACCOUNTING */

diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index b45932e..cef23c2 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -405,9 +405,10 @@ static inline void irqtime_account_process_tick(struct task_struct *p, int user_
/*
* Use precise platform statistics if available:
*/
-#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
+
# ifndef __ARCH_HAS_VTIME_TASK_SWITCH
-void vtime_common_task_switch(struct task_struct *prev)
+void vtime_task_switch(struct task_struct *prev)
{
if (is_idle_task(prev))
vtime_account_idle(prev);
@@ -418,10 +419,7 @@ void vtime_common_task_switch(struct task_struct *prev)
arch_vtime_task_switch(prev);
}
# endif
-#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
-

-#ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
/*
* Archs that account the whole time spent in the idle task
* (outside irq) as idle time can rely on this and just implement
@@ -731,6 +729,16 @@ static void vtime_account_guest(struct task_struct *tsk,
}
}

+static void __vtime_account_kernel(struct task_struct *tsk,
+ struct vtime *vtime)
+{
+ /* We might have scheduled out from guest path */
+ if (tsk->flags & PF_VCPU)
+ vtime_account_guest(tsk, vtime);
+ else
+ vtime_account_system(tsk, vtime);
+}
+
void vtime_account_kernel(struct task_struct *tsk)
{
struct vtime *vtime = &tsk->vtime;
@@ -739,11 +747,7 @@ void vtime_account_kernel(struct task_struct *tsk)
return;

write_seqcount_begin(&vtime->seqcount);
- /* We might have scheduled out from guest path */
- if (tsk->flags & PF_VCPU)
- vtime_account_guest(tsk, vtime);
- else
- vtime_account_system(tsk, vtime);
+ __vtime_account_kernel(tsk, vtime);
write_seqcount_end(&vtime->seqcount);
}

@@ -804,11 +808,15 @@ void vtime_account_idle(struct task_struct *tsk)
account_idle_time(get_vtime_delta(&tsk->vtime));
}

-void arch_vtime_task_switch(struct task_struct *prev)
+void vtime_task_switch_generic(struct task_struct *prev)
{
struct vtime *vtime = &prev->vtime;

write_seqcount_begin(&vtime->seqcount);
+ if (is_idle_task(prev))
+ vtime_account_idle(prev);
+ else
+ __vtime_account_kernel(prev, vtime);
vtime->state = VTIME_INACTIVE;
write_seqcount_end(&vtime->seqcount);