From: Frederic Weisbecker <fweisbec@gmail.com>
To: LKML <linux-kernel@vger.kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>,
Benjamin Herrenschmidt <benh@kernel.crashing.org>,
Paul Mackerras <paulus@samba.org>,
Michael Ellerman <mpe@ellerman.id.au>,
Heiko Carstens <heiko.carstens@de.ibm.com>,
Martin Schwidefsky <schwidefsky@de.ibm.com>,
Tony Luck <tony.luck@intel.com>,
Fenghua Yu <fenghua.yu@intel.com>,
Peter Zijlstra <peterz@infradead.org>,
Rik van Riel <riel@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@kernel.org>,
Stanislaw Gruszka <sgruszka@redhat.com>,
Wanpeng Li <wanpeng.li@hotmail.com>,
Christian Borntraeger <borntraeger@de.ibm.com>
Subject: [PATCH 07/10] powerpc/vtime: Accumulate cputime and account only on tick/task switch
Date: Thu, 5 Jan 2017 18:11:47 +0100 [thread overview]
Message-ID: <1483636310-6557-8-git-send-email-fweisbec@gmail.com> (raw)
In-Reply-To: <1483636310-6557-1-git-send-email-fweisbec@gmail.com>
Currently CONFIG_VIRT_CPU_ACCOUNTING_NATIVE accounts the cputime on
any context boundary: irq entry/exit, guest entry/exit, context switch,
etc...
Calling functions such as account_system_time(), account_user_time()
and such can be costly, especially if they are called on many fastpath
such as twice per IRQ. Those functions do more than just accounting to
kcpustat and task cputime. Depending on the config, some subsystems can
perform unpleasant multiplications and divisions, among other things.
So lets accumulate the cputime instead and delay the accounting on ticks
and context switches only.
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
---
arch/powerpc/kernel/time.c | 120 +++++++++++++++++++++++++++++----------------
1 file changed, 77 insertions(+), 43 deletions(-)
diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index 714313e..4255e69 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -280,17 +280,10 @@ void accumulate_stolen_time(void)
static inline u64 calculate_stolen_time(u64 stop_tb)
{
- u64 stolen = 0;
- struct cpu_accounting_data *acct = &local_paca->accounting;
+ if (get_paca()->dtl_ridx != be64_to_cpu(get_lppaca()->dtl_idx))
+ return scan_dispatch_log(stop_tb);
- if (get_paca()->dtl_ridx != be64_to_cpu(get_lppaca()->dtl_idx)) {
- stolen = scan_dispatch_log(stop_tb);
- acct->stime -= stolen;
- }
-
- stolen += acct->steal_time;
- acct->steal_time = 0;
- return stolen;
+ return 0;
}
#else /* CONFIG_PPC_SPLPAR */
@@ -306,27 +299,26 @@ static inline u64 calculate_stolen_time(u64 stop_tb)
* or soft irq state.
*/
static unsigned long vtime_delta(struct task_struct *tsk,
- unsigned long *sys_scaled,
- unsigned long *stolen)
+ unsigned long *stime_scaled,
+ unsigned long *steal_time)
{
unsigned long now, nowscaled, deltascaled;
- unsigned long udelta, delta, user_scaled;
+ unsigned long stime;
+ unsigned long utime, utime_scaled;
struct cpu_accounting_data *acct = get_accounting(tsk);
WARN_ON_ONCE(!irqs_disabled());
now = mftb();
nowscaled = read_spurr(now);
- acct->stime += now - acct->starttime;
+ stime = now - acct->starttime;
acct->starttime = now;
deltascaled = nowscaled - acct->startspurr;
acct->startspurr = nowscaled;
- *stolen = calculate_stolen_time(now);
+ *steal_time = calculate_stolen_time(now);
- delta = acct->stime;
- acct->stime = 0;
- udelta = acct->utime - acct->utime_sspurr;
+ utime = acct->utime - acct->utime_sspurr;
acct->utime_sspurr = acct->utime;
/*
@@ -339,39 +331,54 @@ static unsigned long vtime_delta(struct task_struct *tsk,
* the user ticks get saved up in paca->user_time_scaled to be
* used by account_process_tick.
*/
- *sys_scaled = delta;
- user_scaled = udelta;
- if (deltascaled != delta + udelta) {
- if (udelta) {
- *sys_scaled = deltascaled * delta / (delta + udelta);
- user_scaled = deltascaled - *sys_scaled;
+ *stime_scaled = stime;
+ utime_scaled = utime;
+ if (deltascaled != stime + utime) {
+ if (utime) {
+ *stime_scaled = deltascaled * stime / (stime + utime);
+ utime_scaled = deltascaled - *stime_scaled;
} else {
- *sys_scaled = deltascaled;
+ *stime_scaled = deltascaled;
}
}
- acct->utime_scaled += user_scaled;
+ acct->utime_scaled += utime_scaled;
- return delta;
+ return stime;
}
void vtime_account_system(struct task_struct *tsk)
{
- unsigned long delta, sys_scaled, stolen;
+ unsigned long stime, stime_scaled, steal_time;
+ struct cpu_accounting_data *acct = get_accounting(tsk);
- delta = vtime_delta(tsk, &sys_scaled, &stolen);
- account_system_time(tsk, 0, delta);
- tsk->stimescaled += sys_scaled;
- if (stolen)
- account_steal_time(stolen);
+ stime = vtime_delta(tsk, &stime_scaled, &steal_time);
+
+ stime -= min(stime, steal_time);
+ acct->steal_time += steal_time;
+
+ if ((tsk->flags & PF_VCPU) && !irq_count()) {
+ acct->gtime += stime;
+ acct->utime_scaled += stime_scaled;
+ } else {
+ if (hardirq_count())
+ acct->hardirq_time += stime;
+ else if (in_serving_softirq())
+ acct->softirq_time += stime;
+ else
+ acct->stime += stime;
+
+ acct->stime_scaled += stime_scaled;
+ }
}
EXPORT_SYMBOL_GPL(vtime_account_system);
void vtime_account_idle(struct task_struct *tsk)
{
- unsigned long delta, sys_scaled, stolen;
+ unsigned long stime, stime_scaled, steal_time;
+ struct cpu_accounting_data *acct = get_accounting(tsk);
- delta = vtime_delta(tsk, &sys_scaled, &stolen);
- account_idle_time(delta + stolen);
+ stime = vtime_delta(tsk, &stime_scaled, &steal_time);
+ acct->idle_time += stime + steal_time;
}
/*
@@ -385,16 +392,45 @@ void vtime_account_idle(struct task_struct *tsk)
*/
void vtime_account_user(struct task_struct *tsk)
{
- cputime_t utime, utimescaled;
struct cpu_accounting_data *acct = get_accounting(tsk);
- utime = acct->utime;
- utimescaled = acct->utime_scaled;
+ if (acct->utime)
+ account_user_time(tsk, acct->utime);
+
+ if (acct->utime_scaled)
+ tsk->utimescaled += acct->utime_scaled;
+
+ if (acct->gtime)
+ account_guest_time(tsk, acct->gtime);
+
+ if (acct->steal_time)
+ account_steal_time(acct->steal_time);
+
+ if (acct->idle_time)
+ account_idle_time(acct->idle_time);
+
+ if (acct->stime)
+ account_system_index_time(tsk, acct->stime, CPUTIME_SYSTEM);
+
+ if (acct->stime_scaled)
+ tsk->stimescaled += acct->stime_scaled;
+
+ if (acct->hardirq_time)
+ account_system_index_time(tsk, acct->hardirq_time, CPUTIME_IRQ);
+
+ if (acct->softirq_time)
+ account_system_index_time(tsk, acct->softirq_time, CPUTIME_SOFTIRQ);
+
acct->utime = 0;
acct->utime_scaled = 0;
acct->utime_sspurr = 0;
- account_user_time(tsk, utime);
- tsk->utimescaled += utimescaled;
+ acct->gtime = 0;
+ acct->steal_time = 0;
+ acct->idle_time = 0;
+ acct->stime = 0;
+ acct->stime_scaled = 0;
+ acct->hardirq_time = 0;
+ acct->softirq_time = 0;
}
#ifdef CONFIG_PPC32
@@ -409,8 +445,6 @@ void arch_vtime_task_switch(struct task_struct *prev)
acct->starttime = get_accounting(prev)->starttime;
acct->startspurr = get_accounting(prev)->startspurr;
- acct->stime = 0;
- acct->utime = 0;
}
#endif /* CONFIG_PPC32 */
--
2.7.4
next prev parent reply other threads:[~2017-01-05 17:12 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-01-05 17:11 [PATCH 00/10] vtime: Delay cputime accounting to tick / context switch Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 01/10] powerpc32: Fix stale scaled stime on " Frederic Weisbecker
2017-01-14 10:00 ` [tip:sched/core] sched/cputime, " tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 02/10] ia64: Fix wrong start cputime assignment on task switch Frederic Weisbecker
2017-01-14 10:00 ` [tip:sched/core] sched/cputime, ia64: Fix incorrect " tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 03/10] cputime: Allow accounting system time using cpustat index Frederic Weisbecker
2017-01-14 10:01 ` [tip:sched/core] sched/cputime: " tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 04/10] cputime: Export account_guest_time Frederic Weisbecker
2017-01-14 10:01 ` [tip:sched/core] sched/cputime: Export account_guest_time() tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 05/10] powerpc: Prepare accounting structure for cputime flush on tick Frederic Weisbecker
2017-01-14 10:02 ` [tip:sched/core] sched/cputime, " tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 06/10] powerpc: Migrate stolen_time field to accounting structure Frederic Weisbecker
2017-01-14 10:03 ` [tip:sched/core] sched/cputime, powerpc: Migrate stolen_time field to the " tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` Frederic Weisbecker [this message]
2017-01-14 10:03 ` [tip:sched/core] sched/cputime, powerpc/vtime: Accumulate cputime and account only on tick/task switch tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 08/10] ia64: " Frederic Weisbecker
2017-01-14 10:04 ` [tip:sched/core] sched/cputime, " tip-bot for Frederic Weisbecker
2017-01-05 17:11 ` [PATCH 09/10] s390/cputime: delayed accounting of system time Frederic Weisbecker
2017-01-14 10:04 ` [tip:sched/core] sched/cputime, s390: Implement " tip-bot for Martin Schwidefsky
2017-01-05 17:11 ` [PATCH 10/10] vtime: Rename vtime_account_user() to vtime_flush() Frederic Weisbecker
2017-01-14 10:05 ` [tip:sched/core] sched/cputime: " tip-bot for Frederic Weisbecker
2017-01-09 8:13 ` [PATCH 00/10] vtime: Delay cputime accounting to tick / context switch Martin Schwidefsky
2017-01-10 11:45 ` Thomas Gleixner
2017-01-10 15:21 ` Frederic Weisbecker
-- strict thread matches above, loose matches on Subject: below --
2016-12-06 2:32 [PATCH 00/10] vtime: Delay cputime accounting to tick Frederic Weisbecker
2016-12-06 2:32 ` [PATCH 07/10] powerpc/vtime: Accumulate cputime and account only on tick/task switch Frederic Weisbecker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1483636310-6557-8-git-send-email-fweisbec@gmail.com \
--to=fweisbec@gmail.com \
--cc=benh@kernel.crashing.org \
--cc=borntraeger@de.ibm.com \
--cc=fenghua.yu@intel.com \
--cc=heiko.carstens@de.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=mpe@ellerman.id.au \
--cc=paulus@samba.org \
--cc=peterz@infradead.org \
--cc=riel@redhat.com \
--cc=schwidefsky@de.ibm.com \
--cc=sgruszka@redhat.com \
--cc=tglx@linutronix.de \
--cc=tony.luck@intel.com \
--cc=wanpeng.li@hotmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox