mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Peter Zijlstra <peterz@infradead.org>
To: frederic@kernel.org, paulmck@kernel.org, rjw@rjwysocki.net,
	x86@kernel.org
Cc: linux-kernel@vger.kernel.org, peterz@infradead.org, jpoimboe@kernel.org
Subject: [RFC][PATCH 4/9] idle: Fix rcu_idle_*() usage
Date: Thu, 19 May 2022 23:27:54 +0200	[thread overview]
Message-ID: <20220519213421.808983977@infradead.org> (raw)
In-Reply-To: <20220519212750.656413111@infradead.org>

The whole disable-RCU, enable-IRQS dance is very intricate since
changing IRQ state is traced, which depends on RCU.

Add two helpers for the cpuidle case that mirror the entry code.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/arm/mach-imx/cpuidle-imx6q.c |    5 ++--
 drivers/acpi/processor_idle.c     |   18 +++++++++------
 drivers/cpuidle/cpuidle.c         |   11 +++++----
 include/linux/sched/idle.h        |   29 ++++++++++++++++++++++++
 kernel/sched/idle.c               |   45 ++++++++++----------------------------
 kernel/time/tick-broadcast.c      |    2 -
 6 files changed, 62 insertions(+), 48 deletions(-)

--- a/arch/arm/mach-imx/cpuidle-imx6q.c
+++ b/arch/arm/mach-imx/cpuidle-imx6q.c
@@ -5,6 +5,7 @@
 
 #include <linux/cpuidle.h>
 #include <linux/module.h>
+#include <linux/sched/idle.h>
 #include <asm/cpuidle.h>
 
 #include <soc/imx/cpuidle.h>
@@ -24,9 +25,9 @@ static int imx6q_enter_wait(struct cpuid
 		imx6_set_lpm(WAIT_UNCLOCKED);
 	raw_spin_unlock(&cpuidle_lock);
 
-	rcu_idle_enter();
+	cpuidle_rcu_enter();
 	cpu_do_idle();
-	rcu_idle_exit();
+	cpuidle_rcu_exit();
 
 	raw_spin_lock(&cpuidle_lock);
 	if (num_idle_cpus-- == num_online_cpus())
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -607,7 +607,7 @@ static DEFINE_RAW_SPINLOCK(c3_lock);
  * @cx: Target state context
  * @index: index of target state
  */
-static int acpi_idle_enter_bm(struct cpuidle_driver *drv,
+static noinstr int acpi_idle_enter_bm(struct cpuidle_driver *drv,
 			       struct acpi_processor *pr,
 			       struct acpi_processor_cx *cx,
 			       int index)
@@ -626,6 +626,8 @@ static int acpi_idle_enter_bm(struct cpu
 	 */
 	bool dis_bm = pr->flags.bm_control;
 
+	instrumentation_begin();
+
 	/* If we can skip BM, demote to a safe state. */
 	if (!cx->bm_sts_skip && acpi_idle_bm_check()) {
 		dis_bm = false;
@@ -647,11 +649,11 @@ static int acpi_idle_enter_bm(struct cpu
 		raw_spin_unlock(&c3_lock);
 	}
 
-	rcu_idle_enter();
+	cpuidle_rcu_enter();
 
 	acpi_idle_do_entry(cx);
 
-	rcu_idle_exit();
+	cpuidle_rcu_exit();
 
 	/* Re-enable bus master arbitration */
 	if (dis_bm) {
@@ -661,11 +663,13 @@ static int acpi_idle_enter_bm(struct cpu
 		raw_spin_unlock(&c3_lock);
 	}
 
+	instrumentation_end();
+
 	return index;
 }
 
-static int acpi_idle_enter(struct cpuidle_device *dev,
-			   struct cpuidle_driver *drv, int index)
+static noinstr int acpi_idle_enter(struct cpuidle_device *dev,
+				   struct cpuidle_driver *drv, int index)
 {
 	struct acpi_processor_cx *cx = per_cpu(acpi_cstate[index], dev->cpu);
 	struct acpi_processor *pr;
@@ -693,8 +697,8 @@ static int acpi_idle_enter(struct cpuidl
 	return index;
 }
 
-static int acpi_idle_enter_s2idle(struct cpuidle_device *dev,
-				  struct cpuidle_driver *drv, int index)
+static noinstr int acpi_idle_enter_s2idle(struct cpuidle_device *dev,
+					  struct cpuidle_driver *drv, int index)
 {
 	struct acpi_processor_cx *cx = per_cpu(acpi_cstate[index], dev->cpu);
 
--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -13,6 +13,7 @@
 #include <linux/mutex.h>
 #include <linux/sched.h>
 #include <linux/sched/clock.h>
+#include <linux/sched/idle.h>
 #include <linux/notifier.h>
 #include <linux/pm_qos.h>
 #include <linux/cpu.h>
@@ -150,12 +151,12 @@ static void enter_s2idle_proper(struct c
 	 */
 	stop_critical_timings();
 	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
-		rcu_idle_enter();
+		cpuidle_rcu_enter();
 	target_state->enter_s2idle(dev, drv, index);
 	if (WARN_ON_ONCE(!irqs_disabled()))
-		local_irq_disable();
+		raw_local_irq_disable();
 	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
-		rcu_idle_exit();
+		cpuidle_rcu_enter();
 	tick_unfreeze();
 	start_critical_timings();
 
@@ -233,14 +234,14 @@ int cpuidle_enter_state(struct cpuidle_d
 
 	stop_critical_timings();
 	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
-		rcu_idle_enter();
+		cpuidle_rcu_enter();
 
 	entered_state = target_state->enter(dev, drv, index);
 	if (WARN_ON_ONCE(!irqs_disabled()))
 		raw_local_irq_disable();
 
 	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
-		rcu_idle_exit();
+		cpuidle_rcu_exit();
 	start_critical_timings();
 
 	sched_clock_idle_wakeup_event();
--- a/include/linux/sched/idle.h
+++ b/include/linux/sched/idle.h
@@ -88,4 +88,33 @@ static inline void current_clr_polling(v
 	preempt_fold_need_resched();
 }
 
+static __always_inline void cpuidle_rcu_enter(void)
+{
+	lockdep_assert_irqs_disabled();
+	/*
+	 * Idle is allowed to (temporary) enable IRQs. It
+	 * will return with IRQs disabled.
+	 *
+	 * Trace IRQs enable here, then switch off RCU, and have
+	 * arch_cpu_idle() use raw_local_irq_enable(). Note that
+	 * rcu_idle_enter() relies on lockdep IRQ state, so switch that
+	 * last -- this is very similar to the entry code.
+	 */
+	trace_hardirqs_on_prepare();
+	lockdep_hardirqs_on_prepare();
+	instrumentation_end();
+	rcu_idle_enter();
+	lockdep_hardirqs_on(_THIS_IP_);
+}
+
+static __always_inline void cpuidle_rcu_exit(void)
+{
+	/*
+	 * Carefully undo the above.
+	 */
+	lockdep_hardirqs_off(_THIS_IP_);
+	rcu_idle_exit();
+	instrumentation_begin();
+}
+
 #endif /* _LINUX_SCHED_IDLE_H */
--- a/kernel/sched/idle.c
+++ b/kernel/sched/idle.c
@@ -51,18 +51,22 @@ __setup("hlt", cpu_idle_nopoll_setup);
 
 static noinline int __cpuidle cpu_idle_poll(void)
 {
+	instrumentation_begin();
 	trace_cpu_idle(0, smp_processor_id());
 	stop_critical_timings();
-	rcu_idle_enter();
-	local_irq_enable();
+	cpuidle_rcu_enter();
 
+	raw_local_irq_enable();
 	while (!tif_need_resched() &&
 	       (cpu_idle_force_poll || tick_check_broadcast_expired()))
 		cpu_relax();
+	raw_local_irq_disable();
 
-	rcu_idle_exit();
+	cpuidle_rcu_exit();
 	start_critical_timings();
 	trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
+	local_irq_enable();
+	instrumentation_end();
 
 	return 1;
 }
@@ -85,44 +87,21 @@ void __weak arch_cpu_idle(void)
  */
 void __cpuidle default_idle_call(void)
 {
-	if (current_clr_polling_and_test()) {
-		local_irq_enable();
-	} else {
-
+	instrumentation_begin();
+	if (!current_clr_polling_and_test()) {
 		trace_cpu_idle(1, smp_processor_id());
 		stop_critical_timings();
 
-		/*
-		 * arch_cpu_idle() is supposed to enable IRQs, however
-		 * we can't do that because of RCU and tracing.
-		 *
-		 * Trace IRQs enable here, then switch off RCU, and have
-		 * arch_cpu_idle() use raw_local_irq_enable(). Note that
-		 * rcu_idle_enter() relies on lockdep IRQ state, so switch that
-		 * last -- this is very similar to the entry code.
-		 */
-		trace_hardirqs_on_prepare();
-		lockdep_hardirqs_on_prepare();
-		rcu_idle_enter();
-		lockdep_hardirqs_on(_THIS_IP_);
-
+		cpuidle_rcu_enter();
 		arch_cpu_idle();
-
-		/*
-		 * OK, so IRQs are enabled here, but RCU needs them disabled to
-		 * turn itself back on.. funny thing is that disabling IRQs
-		 * will cause tracing, which needs RCU. Jump through hoops to
-		 * make it 'work'.
-		 */
 		raw_local_irq_disable();
-		lockdep_hardirqs_off(_THIS_IP_);
-		rcu_idle_exit();
-		lockdep_hardirqs_on(_THIS_IP_);
-		raw_local_irq_enable();
+		cpuidle_rcu_exit();
 
 		start_critical_timings();
 		trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
 	}
+	local_irq_enable();
+	instrumentation_end();
 }
 
 static int call_cpuidle_s2idle(struct cpuidle_driver *drv,
--- a/kernel/time/tick-broadcast.c
+++ b/kernel/time/tick-broadcast.c
@@ -622,7 +622,7 @@ struct cpumask *tick_get_broadcast_onesh
  * to avoid a deep idle transition as we are about to get the
  * broadcast IPI right away.
  */
-int tick_check_broadcast_expired(void)
+int noinstr tick_check_broadcast_expired(void)
 {
 	return cpumask_test_cpu(smp_processor_id(), tick_broadcast_force_mask);
 }



  parent reply	other threads:[~2022-05-19 21:41 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-19 21:27 [RFC][PATCH 0/9] Rework cpuidle vs instrumentation Peter Zijlstra
2022-05-19 21:27 ` [RFC][PATCH 1/9] x86/perf/amd: Remove tracing from perf_lopwr_cb() Peter Zijlstra
2022-05-19 21:27 ` [RFC][PATCH 2/9] x86/idle: Replace x86_idle with a static_call Peter Zijlstra
2022-05-30 11:07   ` Frederic Weisbecker
2022-05-19 21:27 ` [RFC][PATCH 3/9] cpuidle: Move IRQ state validation Peter Zijlstra
2022-05-30 11:36   ` Frederic Weisbecker
2022-05-30 12:00     ` Peter Zijlstra
2022-05-19 21:27 ` Peter Zijlstra [this message]
2022-05-23  9:46   ` [RFC][PATCH 4/9] idle: Fix rcu_idle_*() usage Gautham R. Shenoy
2022-05-31  9:33     ` Peter Zijlstra
2022-05-19 21:27 ` [RFC][PATCH 5/9] rcu: Fix rcu_idle_exit() Peter Zijlstra
2022-05-20 21:00   ` Paul E. McKenney
2022-05-19 21:27 ` [RFC][PATCH 6/9] acpi_idle: Remove tracing Peter Zijlstra
2022-05-19 21:27 ` [RFC][PATCH 7/9] cpuidle: Annotate poll_idle() Peter Zijlstra
2022-05-19 21:27 ` [RFC][PATCH 8/9] objtool/idle: Validate __cpuidle code as noinstr Peter Zijlstra
2022-05-19 21:27 ` [RFC][PATCH 9/9] arch/idle: Change arch_cpu_idle() IRQ behaviour Peter Zijlstra
2022-05-19 22:03   ` Peter Zijlstra
2022-05-20  2:20     ` Kirill A. Shutemov
2022-05-20  7:06       ` Peter Zijlstra
2022-05-20 10:13         ` Kirill A. Shutemov
2022-05-20 12:58           ` Peter Zijlstra
2022-05-24 14:55             ` Isaku Yamahata
2022-05-23 10:02   ` Gautham R. Shenoy

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=20220519213421.808983977@infradead.org \
    --to=peterz@infradead.org \
    --cc=frederic@kernel.org \
    --cc=jpoimboe@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=paulmck@kernel.org \
    --cc=rjw@rjwysocki.net \
    --cc=x86@kernel.org \
    /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

all inboxes | Powered by JetHome®