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 PHC | |
Open Source and information security mailing list archives
| ||
|
Date: Wed, 26 Aug 2020 21:18:26 -0400 From: Joel Fernandes <joel@...lfernandes.org> To: Peter Zijlstra <peterz@...radead.org> Cc: linux-kernel@...r.kernel.org, mingo@...nel.org, will@...nel.org, npiggin@...il.com, elver@...gle.com, jgross@...e.com, paulmck@...nel.org, rostedt@...dmis.org, rjw@...ysocki.net, svens@...ux.ibm.com, tglx@...utronix.de Subject: Re: [PATCH v2 03/11] sched,idle,rcu: Push rcu_idle deeper into the idle path On Fri, Aug 21, 2020 at 10:47:41AM +0200, Peter Zijlstra wrote: > 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, and > stop_critical_timings() uses lock. I was wondering if those tracepoints should just use _rcuidle variant of the trace call. But that's a terrible idea considering that would add unwanted overhead I think. Reviewed-by: Joel Fernandes (Google) <joel@...lfernandes.org> thanks, - Joel > > Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org> > Reviewed-by: Steven Rostedt (VMware) <rostedt@...dmis.org> > Tested-by: Marco Elver <elver@...gle.com> > --- > drivers/cpuidle/cpuidle.c | 12 ++++++++---- > kernel/sched/idle.c | 22 ++++++++-------------- > 2 files changed, 16 insertions(+), 18 deletions(-) > > --- a/drivers/cpuidle/cpuidle.c > +++ b/drivers/cpuidle/cpuidle.c > @@ -145,13 +145,14 @@ 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. > */ > stop_critical_timings(); > + rcu_idle_enter(); > drv->states[index].enter_s2idle(dev, drv, index); > if (WARN_ON_ONCE(!irqs_disabled())) > local_irq_disable(); > @@ -160,7 +161,8 @@ static void enter_s2idle_proper(struct c > * first CPU executing it calls functions containing RCU read-side > * critical sections, so tell RCU about that. > */ > - RCU_NONIDLE(tick_unfreeze()); > + rcu_idle_exit(); > + tick_unfreeze(); > start_critical_timings(); > > time_end = ns_to_ktime(local_clock()); > @@ -229,16 +231,18 @@ 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()); > > stop_critical_timings(); > + rcu_idle_enter(); > entered_state = target_state->enter(dev, drv, index); > + rcu_idle_exit(); > start_critical_timings(); > > 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()); > + stop_critical_timings(); > 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(); > + start_critical_timings(); > + trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id()); > > return 1; > } > @@ -91,7 +92,9 @@ void __cpuidle default_idle_call(void) > local_irq_enable(); > } else { > stop_critical_timings(); > + rcu_idle_enter(); > arch_cpu_idle(); > + rcu_idle_exit(); > start_critical_timings(); > } > } > @@ -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(); > } > > /* > >
Powered by blists - more mailing lists