Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1429407 > unrolled thread

[PATCH v2 0/5] sched,time: fix irq time accounting with nohz_idle

Back to article view | Back to linux.kernel

Contents

  [PATCH v2 0/5] sched,time: fix irq time accounting with nohz_idle riel@redhat.com - 2016-06-23 04:30 +0200
    [PATCH 5/5] irqtime: drop local_irq_save/restore from irqtime_account_irq riel@redhat.com - 2016-06-23 04:30 +0200
    [PATCH 4/5] irqtime: add irq type parameter to irqtime_account_irq riel@redhat.com - 2016-06-23 04:30 +0200
    [PATCH 3/5] cputime: allow irq time accounting to be selected as an option riel@redhat.com - 2016-06-23 04:30 +0200

#1429407 — [PATCH v2 0/5] sched,time: fix irq time accounting with nohz_idle

Currently irq time accounting only works in these cases:
1) purely ticke based accounting
2) nohz_full accounting, but only on housekeeping & nohz_full CPUs
3) architectures with native vtime accounting

On nohz_idle CPUs, which are probably the majority nowadays,
irq time accounting is currently broken. This leads to systems
reporting a dramatically lower amount of irq & softirq time than
is actually spent handling them, with all the time spent while the
system is in the idle task being accounted as idle.

This patch set seems to bring the amount of irq time reported by
top (and /proc/stat) roughly in line with that measured when I do
a "perf record -g -a" run to see what is using all that time.

The amount of irq time used, especially softirq, is shockingly high,
to the point of me thinking this patch set may be wrong, but the
numbers seem to match what perf is giving me...

These patches apply on top of Wanpeng Li's steal time patches.

CONFIG_IRQ_TIME_ACCOUNTING is now a config option that is available
as a separate choice from tick based / nohz_idle / nohz_full mode,
a suggested by Frederic Weisbecker.

Next up: look at the things that are using CPU time on an otherwise
idle system, and see if I can make those a little faster :)

v2: address Peterz's concerns, some more cleanups

[toc] | [next] | [standalone]

#1429408 — [PATCH 5/5] irqtime: drop local_irq_save/restore from irqtime_account_irq

From: Rik van Riel <riel@redhat.com>

Drop local_irq_save/restore from irqtime_account_irq.
Instead, have softirq and hardirq track their time spent
independently, with the softirq code subtracting hardirq
time that happened during the duration of the softirq run.

The softirq code can be interrupted by hardirq code at
any point in time, but it can check whether it got a
consistent snapshot of the timekeeping variables it wants,
and loop around in the unlikely case that it did not.

Signed-off-by: Rik van Riel <riel@redhat.com>
---
 kernel/sched/cputime.c | 72 +++++++++++++++++++++++++++++++++++++++++---------
 kernel/sched/sched.h   | 38 +++++++++++++++++++++-----
 2 files changed, 90 insertions(+), 20 deletions(-)

diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index fc4122afc022..e4c7b2d17141 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -26,7 +26,9 @@
 DEFINE_PER_CPU(u64, cpu_hardirq_time);
 DEFINE_PER_CPU(u64, cpu_softirq_time);
 
-static DEFINE_PER_CPU(u64, irq_start_time);
+static DEFINE_PER_CPU(u64, hardirq_start_time);
+static DEFINE_PER_CPU(u64, softirq_start_time);
+static DEFINE_PER_CPU(u64, prev_hardirq_time);
 static int sched_clock_irqtime;
 
 void enable_sched_clock_irqtime(void)
@@ -41,6 +43,7 @@ void disable_sched_clock_irqtime(void)
 
 #ifndef CONFIG_64BIT
 DEFINE_PER_CPU(seqcount_t, irq_time_seq);
+DEFINE_PER_CPU(seqcount_t, softirq_time_seq);
 #endif /* CONFIG_64BIT */
 
 /*
@@ -53,36 +56,79 @@ DEFINE_PER_CPU(seqcount_t, irq_time_seq);
  * softirq -> hardirq, hardirq -> softirq
  *
  * When exiting hardirq or softirq time, account the elapsed time.
+ *
+ * When exiting softirq time, subtract the amount of hardirq time that
+ * interrupted this softirq run, to avoid double accounting of that time.
  */
 void irqtime_account_irq(struct task_struct *curr, int irqtype)
 {
-	unsigned long flags;
+	u64 prev_softirq_start;
+	bool leaving_softirq;
+	u64 prev_hardirq;
+	u64 hardirq_time;
 	s64 delta;
 	int cpu;
 
 	if (!sched_clock_irqtime)
 		return;
 
-	local_irq_save(flags);
-
 	cpu = smp_processor_id();
-	delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
-	__this_cpu_add(irq_start_time, delta);
 
-	irq_time_write_begin();
+	/*
+	 * Hardirq time accounting is pretty straightforward. If not in
+	 * hardirq context yet (entering hardirq), set the start time.
+	 * If already in hardirq context (leaving), account the elapsed time.
+	 */
+	if (irqtype == HARDIRQ_OFFSET) {
+		bool leaving_hardirq = hardirq_count();
+		delta = sched_clock_cpu(cpu) - __this_cpu_read(hardirq_start_time);
+		__this_cpu_add(hardirq_start_time, delta);
+		if (leaving_hardirq) {
+			hardirq_time_write_begin();
+			__this_cpu_add(cpu_hardirq_time, delta);
+			hardirq_time_write_end();
+		}
+		return;
+	}
+
+	/*
+	 * Softirq context may get interrupted by hardirq context, on the
+	 * same CPU. At softirq entry time the amount of time this CPU spent
+	 * in hardirq context is stored. At softirq exit time, the time spent
+	 * in hardirq context during the softirq is subtracted.
+	 */
+	prev_softirq_start = __this_cpu_read(softirq_start_time);
+	prev_hardirq = __this_cpu_read(prev_hardirq_time);
+	leaving_softirq = in_serving_softirq();
+
+	do {
+		u64 now = sched_clock_cpu(cpu);
+
+		hardirq_time = READ_ONCE(per_cpu(cpu_hardirq_time, cpu));
+		__this_cpu_write(softirq_start_time, now);
+		__this_cpu_write(prev_hardirq_time, hardirq_time);
+
+		if (leaving_softirq) {
+			/*
+			 * Subtract hardirq time that happened during this
+			 * softirq.
+			 */
+			s64 hi_delta = hardirq_time - prev_hardirq;
+			delta = now - prev_softirq_start - hi_delta;
+		}
+		/* Loop around if interrupted by a hardirq. */
+	} while (hardirq_time != READ_ONCE(per_cpu(cpu_hardirq_time, cpu)));
+
 	/*
 	 * We do not account for softirq time from ksoftirqd here.
 	 * We want to continue accounting softirq time to ksoftirqd thread
 	 * in that case, so as not to confuse scheduler with a special task
 	 * that do not consume any time, but still wants to run.
 	 */
-	if (hardirq_count())
-		__this_cpu_add(cpu_hardirq_time, delta);
-	else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
+	softirq_time_write_begin();
+	if (leaving_softirq && curr != this_cpu_ksoftirqd())
 		__this_cpu_add(cpu_softirq_time, delta);
-
-	irq_time_write_end();
-	local_irq_restore(flags);
+	softirq_time_write_end();
 }
 EXPORT_SYMBOL_GPL(irqtime_account_irq);
 
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index ec2e8d23527e..cad4df9835f7 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1752,38 +1752,62 @@ DECLARE_PER_CPU(u64, cpu_softirq_time);
 
 #ifndef CONFIG_64BIT
 DECLARE_PER_CPU(seqcount_t, irq_time_seq);
+DECLARE_PER_CPU(seqcount_t, softirq_time_seq);
 
-static inline void irq_time_write_begin(void)
+static inline void hardirq_time_write_begin(void)
 {
 	__this_cpu_inc(irq_time_seq.sequence);
 	smp_wmb();
 }
 
-static inline void irq_time_write_end(void)
+static inline void hardirq_time_write_end(void)
 {
 	smp_wmb();
 	__this_cpu_inc(irq_time_seq.sequence);
 }
 
+static inline void softirq_time_write_begin(void)
+{
+	__this_cpu_inc(softirq_time_seq.sequence);
+	smp_wmb();
+}
+
+static inline void softirq_time_write_end(void)
+{
+	smp_wmb();
+	__this_cpu_inc(softirq_time_seq.sequence);
+}
+
 static inline u64 irq_time_read(int cpu)
 {
 	u64 irq_time;
-	unsigned seq;
+	unsigned hi_seq;
+	unsigned si_seq;
 
 	do {
-		seq = read_seqcount_begin(&per_cpu(irq_time_seq, cpu));
+		hi_seq = read_seqcount_begin(&per_cpu(irq_time_seq, cpu));
+		si_seq = read_seqcount_begin(&per_cpu(softirq_time_seq, cpu));
 		irq_time = per_cpu(cpu_softirq_time, cpu) +
 			   per_cpu(cpu_hardirq_time, cpu);
-	} while (read_seqcount_retry(&per_cpu(irq_time_seq, cpu), seq));
+	} while (read_seqcount_retry(&per_cpu(irq_time_seq, cpu), hi_seq) ||
+		 read_seqcount_retry(&per_cpu(softirq_time_seq, cpu), si_seq));
 
 	return irq_time;
 }
 #else /* CONFIG_64BIT */
-static inline void irq_time_write_begin(void)
+static inline void hardirq_time_write_begin(void)
+{
+}
+
+static inline void hardirq_time_write_end(void)
+{
+}
+
+static inline void softirq_time_write_begin(void)
 {
 }
 
-static inline void irq_time_write_end(void)
+static inline void softirq_time_write_end(void)
 {
 }
 
-- 
2.5.5

[toc] | [prev] | [next] | [standalone]

#1429409 — [PATCH 4/5] irqtime: add irq type parameter to irqtime_account_irq

From: Rik van Riel <riel@redhat.com>

Add an irq type parameter and documentation to irqtime_account_irq,
this can be used to distinguish between transitioning from process
context to hardirq time, and from process context to softirq time.

This is necessary to be able to remove the local_irq_disable from
irqtime_account_irq.

Signed-off-by: Rik van Riel <riel@redhat.com>
---
 include/linux/hardirq.h | 20 ++++++++++----------
 include/linux/vtime.h   | 12 ++++++------
 kernel/sched/cputime.c  |  9 ++++++++-
 kernel/softirq.c        |  6 +++---
 4 files changed, 27 insertions(+), 20 deletions(-)

diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
index dfd59d6bc6f0..1ebb31f56285 100644
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -32,11 +32,11 @@ extern void rcu_nmi_exit(void);
  * always balanced, so the interrupted value of ->hardirq_context
  * will always be restored.
  */
-#define __irq_enter()					\
-	do {						\
-		account_irq_enter_time(current);	\
-		preempt_count_add(HARDIRQ_OFFSET);	\
-		trace_hardirq_enter();			\
+#define __irq_enter()							\
+	do {								\
+		account_irq_enter_time(current, HARDIRQ_OFFSET);	\
+		preempt_count_add(HARDIRQ_OFFSET);			\
+		trace_hardirq_enter();					\
 	} while (0)
 
 /*
@@ -47,11 +47,11 @@ extern void irq_enter(void);
 /*
  * Exit irq context without processing softirqs:
  */
-#define __irq_exit()					\
-	do {						\
-		trace_hardirq_exit();			\
-		account_irq_exit_time(current);		\
-		preempt_count_sub(HARDIRQ_OFFSET);	\
+#define __irq_exit()							\
+	do {								\
+		trace_hardirq_exit();					\
+		account_irq_exit_time(current, HARDIRQ_OFFSET);		\
+		preempt_count_sub(HARDIRQ_OFFSET);			\
 	} while (0)
 
 /*
diff --git a/include/linux/vtime.h b/include/linux/vtime.h
index d1977d84ebdf..31e486d741b0 100644
--- a/include/linux/vtime.h
+++ b/include/linux/vtime.h
@@ -112,21 +112,21 @@ static inline void vtime_init_idle(struct task_struct *tsk, int cpu) { }
 #endif
 
 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
-extern void irqtime_account_irq(struct task_struct *tsk);
+extern void irqtime_account_irq(struct task_struct *tsk, int irqtype);
 #else
-static inline void irqtime_account_irq(struct task_struct *tsk) { }
+static inline void irqtime_account_irq(struct task_struct *tsk, int irqtype) { }
 #endif
 
-static inline void account_irq_enter_time(struct task_struct *tsk)
+static inline void account_irq_enter_time(struct task_struct *tsk, int irqtype)
 {
 	vtime_account_irq_enter(tsk);
-	irqtime_account_irq(tsk);
+	irqtime_account_irq(tsk, irqtype);
 }
 
-static inline void account_irq_exit_time(struct task_struct *tsk)
+static inline void account_irq_exit_time(struct task_struct *tsk, int irqtype)
 {
 	vtime_account_irq_exit(tsk);
-	irqtime_account_irq(tsk);
+	irqtime_account_irq(tsk, irqtype);
 }
 
 #endif /* _LINUX_KERNEL_VTIME_H */
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index 0280528800f2..fc4122afc022 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -46,8 +46,15 @@ DEFINE_PER_CPU(seqcount_t, irq_time_seq);
 /*
  * Called before incrementing preempt_count on {soft,}irq_enter
  * and before decrementing preempt_count on {soft,}irq_exit.
+ *
+ * There are six possible transitions:
+ * process -> softirq, softirq -> process
+ * process -> hardirq, hardirq -> process
+ * softirq -> hardirq, hardirq -> softirq
+ *
+ * When exiting hardirq or softirq time, account the elapsed time.
  */
-void irqtime_account_irq(struct task_struct *curr)
+void irqtime_account_irq(struct task_struct *curr, int irqtype)
 {
 	unsigned long flags;
 	s64 delta;
diff --git a/kernel/softirq.c b/kernel/softirq.c
index 17caf4b63342..a311c9622c86 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -245,7 +245,7 @@ asmlinkage __visible void __softirq_entry __do_softirq(void)
 	current->flags &= ~PF_MEMALLOC;
 
 	pending = local_softirq_pending();
-	account_irq_enter_time(current);
+	account_irq_enter_time(current, SOFTIRQ_OFFSET);
 
 	__local_bh_disable_ip(_RET_IP_, SOFTIRQ_OFFSET);
 	in_hardirq = lockdep_softirq_start();
@@ -295,7 +295,7 @@ asmlinkage __visible void __softirq_entry __do_softirq(void)
 	}
 
 	lockdep_softirq_end(in_hardirq);
-	account_irq_exit_time(current);
+	account_irq_exit_time(current, SOFTIRQ_OFFSET);
 	__local_bh_enable(SOFTIRQ_OFFSET);
 	WARN_ON_ONCE(in_interrupt());
 	tsk_restore_flags(current, old_flags, PF_MEMALLOC);
@@ -385,7 +385,7 @@ void irq_exit(void)
 	WARN_ON_ONCE(!irqs_disabled());
 #endif
 
-	account_irq_exit_time(current);
+	account_irq_exit_time(current, HARDIRQ_OFFSET);
 	preempt_count_sub(HARDIRQ_OFFSET);
 	if (!in_interrupt() && local_softirq_pending())
 		invoke_softirq();
-- 
2.5.5

[toc] | [prev] | [next] | [standalone]

#1429410 — [PATCH 3/5] cputime: allow irq time accounting to be selected as an option

From: Rik van Riel <riel@redhat.com>

Allow CONFIG_IRQ_TIME_ACCOUNTING to be selected as an option, on top
of CONFIG_VIRT_CPU_ACCOUNTING_GEN (and potentially others?).

This allows for the irq time accounting code to be used with nohz_idle
CPUs, which is how several distributions ship their kernels. Using the
same code for several timer modes also allows us to drop duplicate code.

Signed-off-by: Rik van Riel <riel@redhat.com>
---
 init/Kconfig | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/init/Kconfig b/init/Kconfig
index 0dfd09d54c65..4c7ee4f136cf 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -375,9 +375,11 @@ config VIRT_CPU_ACCOUNTING_GEN
 
 	  If unsure, say N.
 
+endchoice
+
 config IRQ_TIME_ACCOUNTING
 	bool "Fine granularity task level IRQ time accounting"
-	depends on HAVE_IRQ_TIME_ACCOUNTING && !NO_HZ_FULL
+	depends on HAVE_IRQ_TIME_ACCOUNTING && !VIRT_CPU_ACCOUNTING_NATIVE
 	help
 	  Select this option to enable fine granularity task irq time
 	  accounting. This is done by reading a timestamp on each
@@ -386,8 +388,6 @@ config IRQ_TIME_ACCOUNTING
 
 	  If in doubt, say N here.
 
-endchoice
-
 config BSD_PROCESS_ACCT
 	bool "BSD Process Accounting"
 	depends on MULTIUSER
-- 
2.5.5

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web