linux-kernel - [PATCH 2/9] sched,idle,rcu: Push rcu

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200820074640.401636694@infradead.org>
Date:   Thu, 20 Aug 2020 09:30:33 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     linux-kernel@...r.kernel.org, mingo@...nel.org, will@...nel.org
Cc:     npiggin@...il.com, elver@...gle.com, jgross@...e.com,
        paulmck@...nel.org, rostedt@...dmis.org, rjw@...ysocki.net,
        joel@...lfernandes.org, svens@...ux.ibm.com, tglx@...utronix.de,
        peterz@...radead.org
Subject: [PATCH 2/9] sched,idle,rcu: Push rcu_idle deeper into the idle path

Lots of things take locks, due to a wee bug, rcu_lockdep didn't notice
that the locking tracepoints were using RCU.

Push rcu_idle_{enter,exit}() as deep as possible into the idle paths,
this also resolves a lot of _rcuidle()/RCU_NONIDLE() usage.

Specifically, sched_clock_idle_wakeup_event() will use ktime which
will use seqlocks which will tickle lockdep.

Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
---
 drivers/cpuidle/cpuidle.c            |   15 +++++++++++----
 kernel/sched/idle.c                  |   18 ++++++------------
 kernel/time/tick-broadcast-hrtimer.c |   29 ++++++++++++-----------------
 3 files changed, 29 insertions(+), 33 deletions(-)

--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -145,22 +145,27 @@ static void enter_s2idle_proper(struct c
 	 * executing it contains RCU usage regarded as invalid in the idle
 	 * context, so tell RCU about that.
 	 */
-	RCU_NONIDLE(tick_freeze());
+	tick_freeze();
 	/*
 	 * The state used here cannot be a "coupled" one, because the "coupled"
 	 * cpuidle mechanism enables interrupts and doing that with timekeeping
 	 * suspended is generally unsafe.
 	 */
+	rcu_idle_enter();
 	stop_critical_timings();
 	drv->states[index].enter_s2idle(dev, drv, index);
-	WARN_ON(!irqs_disabled());
+	if (WARN_ON(!irqs_disabled()))
+		local_irq_disable();
 	/*
 	 * timekeeping_resume() that will be called by tick_unfreeze() for the
 	 * first CPU executing it calls functions containing RCU read-side
 	 * critical sections, so tell RCU about that.
+	 *
+	 * FIXME, order against rcu_idle_exit ?
 	 */
 	RCU_NONIDLE(tick_unfreeze());
 	start_critical_timings();
+	rcu_idle_exit();
 
 	time_end = ns_to_ktime(local_clock());
 
@@ -228,16 +233,18 @@ int cpuidle_enter_state(struct cpuidle_d
 	/* Take note of the planned idle state. */
 	sched_idle_set_state(target_state);
 
-	trace_cpu_idle_rcuidle(index, dev->cpu);
+	trace_cpu_idle(index, dev->cpu);
 	time_start = ns_to_ktime(local_clock());
 
+	rcu_idle_enter();
 	stop_critical_timings();
 	entered_state = target_state->enter(dev, drv, index);
 	start_critical_timings();
+	rcu_idle_exit();
 
 	sched_clock_idle_wakeup_event();
 	time_end = ns_to_ktime(local_clock());
-	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
+	trace_cpu_idle(PWR_EVENT_EXIT, dev->cpu);
 
 	/* The cpu is no longer idle or about to enter idle. */
 	sched_idle_set_state(NULL);
--- a/kernel/sched/idle.c
+++ b/kernel/sched/idle.c
@@ -54,17 +54,18 @@ __setup("hlt", cpu_idle_nopoll_setup);
 
 static noinline int __cpuidle cpu_idle_poll(void)
 {
+	trace_cpu_idle(0, smp_processor_id());
 	rcu_idle_enter();
-	trace_cpu_idle_rcuidle(0, smp_processor_id());
 	local_irq_enable();
 	stop_critical_timings();
 
 	while (!tif_need_resched() &&
-		(cpu_idle_force_poll || tick_check_broadcast_expired()))
+	       (cpu_idle_force_poll || tick_check_broadcast_expired()))
 		cpu_relax();
+
 	start_critical_timings();
-	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
 	rcu_idle_exit();
+	trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
 
 	return 1;
 }
@@ -90,9 +91,11 @@ void __cpuidle default_idle_call(void)
 	if (current_clr_polling_and_test()) {
 		local_irq_enable();
 	} else {
+		rcu_idle_enter();
 		stop_critical_timings();
 		arch_cpu_idle();
 		start_critical_timings();
+		rcu_idle_exit();
 	}
 }
 
@@ -158,7 +161,6 @@ static void cpuidle_idle_call(void)
 
 	if (cpuidle_not_available(drv, dev)) {
 		tick_nohz_idle_stop_tick();
-		rcu_idle_enter();
 
 		default_idle_call();
 		goto exit_idle;
@@ -178,21 +180,17 @@ static void cpuidle_idle_call(void)
 		u64 max_latency_ns;
 
 		if (idle_should_enter_s2idle()) {
-			rcu_idle_enter();
 
 			entered_state = call_cpuidle_s2idle(drv, dev);
 			if (entered_state > 0)
 				goto exit_idle;
 
-			rcu_idle_exit();
-
 			max_latency_ns = U64_MAX;
 		} else {
 			max_latency_ns = dev->forced_idle_latency_limit_ns;
 		}
 
 		tick_nohz_idle_stop_tick();
-		rcu_idle_enter();
 
 		next_state = cpuidle_find_deepest_state(drv, dev, max_latency_ns);
 		call_cpuidle(drv, dev, next_state);
@@ -209,8 +207,6 @@ static void cpuidle_idle_call(void)
 		else
 			tick_nohz_idle_retain_tick();
 
-		rcu_idle_enter();
-
 		entered_state = call_cpuidle(drv, dev, next_state);
 		/*
 		 * Give the governor an opportunity to reflect on the outcome
@@ -226,8 +222,6 @@ static void cpuidle_idle_call(void)
 	 */
 	if (WARN_ON_ONCE(irqs_disabled()))
 		local_irq_enable();
-
-	rcu_idle_exit();
 }
 
 /*