Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752143AbbFZHro (ORCPT ); Fri, 26 Jun 2015 03:47:44 -0400 Received: from www.linutronix.de ([62.245.132.108]:50828 "EHLO Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751857AbbFZHrh (ORCPT ); Fri, 26 Jun 2015 03:47:37 -0400 Date: Fri, 26 Jun 2015 09:47:34 +0200 (CEST) From: Thomas Gleixner To: Preeti U Murthy cc: Sudeep Holla , "linux-kernel@vger.kernel.org" , Suzuki Poulose , Lorenzo Pieralisi , Catalin Marinas , "Rafael J. Wysocki" Subject: Re: [PATCH] clockevents: return error from tick_broadcast_oneshot_control if !GENERIC_CLOCKEVENTS_BROADCAST In-Reply-To: <558CDE57.3050008@linux.vnet.ibm.com> Message-ID: References: <1435228065-2428-1-git-send-email-sudeep.holla@arm.com> <558C1E97.4020206@arm.com> <558CDE57.3050008@linux.vnet.ibm.com> User-Agent: Alpine 2.11 (DEB 23 2013-08-11) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Linutronix-Spam-Score: -1.0 X-Linutronix-Spam-Level: - X-Linutronix-Spam-Status: No , -1.0 points, 5.0 required, ALL_TRUSTED=-1,SHORTCIRCUIT=-0.0001 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6870 Lines: 202 On Fri, 26 Jun 2015, Preeti U Murthy wrote: > What about the case where GENERIC_CLOCKEVENTS_BROADCAST=y and > TICK_ONESHOT=n (HZ_PERIODIC=y) ? Have you tested this ? > > This will hang the kernel at boot if you are using the hrtimer mode of > broadcast. This is because the local timers of all cpus are shutdown > when the cpuidle driver registers itself, on finding out that there are > idle states where local tick devices stop. The broadcast tick device is > then in charge of waking up the cpus at every period. In hrtimer mode of > broadcast, there is no such real device and we hang. Hmm, no. tick-broadcast-hrtimer.o depends on TICK_ONESHOT=y. So this is covered by the check for the broadcast device, which is NULL. But there is another variant: GENERIC_CLOCKEVENTS_BROADCAST=y TICK_ONESHOT=y and 'highres=off nohz=off' on the kernel command line. So we need to have the broadcast_needs_cpu() check in that case as well. Updated patch below. Thanks, tglx --- diff --git a/include/linux/tick.h b/include/linux/tick.h index 4191b5623a28..8731a58dd747 100644 --- a/include/linux/tick.h +++ b/include/linux/tick.h @@ -63,11 +63,7 @@ extern void tick_broadcast_control(enum tick_broadcast_mode mode); static inline void tick_broadcast_control(enum tick_broadcast_mode mode) { } #endif /* BROADCAST */ -#if defined(CONFIG_GENERIC_CLOCKEVENTS_BROADCAST) && defined(CONFIG_TICK_ONESHOT) extern int tick_broadcast_oneshot_control(enum tick_broadcast_state state); -#else -static inline int tick_broadcast_oneshot_control(enum tick_broadcast_state state) { return 0; } -#endif static inline void tick_broadcast_enable(void) { diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c index d39f32cdd1b5..281ce29d295e 100644 --- a/kernel/time/tick-broadcast.c +++ b/kernel/time/tick-broadcast.c @@ -662,46 +662,39 @@ static void broadcast_shutdown_local(struct clock_event_device *bc, clockevents_switch_state(dev, CLOCK_EVT_STATE_SHUTDOWN); } -/** - * tick_broadcast_oneshot_control - Enter/exit broadcast oneshot mode - * @state: The target state (enter/exit) - * - * The system enters/leaves a state, where affected devices might stop - * Returns 0 on success, -EBUSY if the cpu is used to broadcast wakeups. - * - * Called with interrupts disabled, so clockevents_lock is not - * required here because the local clock event device cannot go away - * under us. - */ -int tick_broadcast_oneshot_control(enum tick_broadcast_state state) +int __tick_broadcast_oneshot_control(enum tick_broadcast_state state) { struct clock_event_device *bc, *dev; - struct tick_device *td; int cpu, ret = 0; ktime_t now; - /* - * Periodic mode does not care about the enter/exit of power - * states - */ - if (tick_broadcast_device.mode == TICKDEV_MODE_PERIODIC) - return 0; + if (!tick_broadcast_device.evtdev) + return -EBUSY; - /* - * We are called with preemtion disabled from the depth of the - * idle code, so we can't be moved away. - */ - td = this_cpu_ptr(&tick_cpu_device); - dev = td->evtdev; - - if (!(dev->features & CLOCK_EVT_FEAT_C3STOP)) - return 0; + dev = this_cpu_ptr(&tick_cpu_device)->evtdev; raw_spin_lock(&tick_broadcast_lock); bc = tick_broadcast_device.evtdev; cpu = smp_processor_id(); if (state == TICK_BROADCAST_ENTER) { + /* + * If the current CPU owns the hrtimer broadcast + * mechanism, it cannot go deep idle and we do not add + * the CPU to the broadcast mask. We don't have to go + * through the EXIT path as the local timer is not + * shutdown. + */ + ret = broadcast_needs_cpu(bc, cpu); + + /* + * If the hrtimer broadcast check tells us that the + * cpu cannot go deep idle, or if the broadcast device + * is in periodic mode, we just return. + */ + if (ret || tick_broadcast_device.mode == TICKDEV_MODE_PERIODIC) + goto out; + if (!cpumask_test_and_set_cpu(cpu, tick_broadcast_oneshot_mask)) { WARN_ON_ONCE(cpumask_test_cpu(cpu, tick_broadcast_pending_mask)); broadcast_shutdown_local(bc, dev); @@ -717,16 +710,6 @@ int tick_broadcast_oneshot_control(enum tick_broadcast_state state) dev->next_event.tv64 < bc->next_event.tv64) tick_broadcast_set_event(bc, cpu, dev->next_event); } - /* - * If the current CPU owns the hrtimer broadcast - * mechanism, it cannot go deep idle and we remove the - * CPU from the broadcast mask. We don't have to go - * through the EXIT path as the local timer is not - * shutdown. - */ - ret = broadcast_needs_cpu(bc, cpu); - if (ret) - cpumask_clear_cpu(cpu, tick_broadcast_oneshot_mask); } else { if (cpumask_test_and_clear_cpu(cpu, tick_broadcast_oneshot_mask)) { clockevents_switch_state(dev, CLOCK_EVT_STATE_ONESHOT); @@ -938,6 +921,13 @@ bool tick_broadcast_oneshot_available(void) return bc ? bc->features & CLOCK_EVT_FEAT_ONESHOT : false; } +#else +int __tick_broadcast_oneshot_control(enum tick_broadcast_state state) +{ + if (!tick_broadcast_device.evtdev) + return -EBUSY; + return 0; +} #endif void __init tick_broadcast_init(void) diff --git a/kernel/time/tick-common.c b/kernel/time/tick-common.c index 17f144450050..f66c11a30348 100644 --- a/kernel/time/tick-common.c +++ b/kernel/time/tick-common.c @@ -342,6 +342,27 @@ out_bc: tick_install_broadcast_device(newdev); } +/** + * tick_broadcast_oneshot_control - Enter/exit broadcast oneshot mode + * @state: The target state (enter/exit) + * + * The system enters/leaves a state, where affected devices might stop + * Returns 0 on success, -EBUSY if the cpu is used to broadcast wakeups. + * + * Called with interrupts disabled, so clockevents_lock is not + * required here because the local clock event device cannot go away + * under us. + */ +int tick_broadcast_oneshot_control(enum tick_broadcast_state state) +{ + struct tick_device *td = this_cpu_ptr(&tick_cpu_device); + + if (!(td->evtdev->features & CLOCK_EVT_FEAT_C3STOP)) + return 0; + + return __tick_broadcast_oneshot_control(state); +} + #ifdef CONFIG_HOTPLUG_CPU /* * Transfer the do_timer job away from a dying cpu. diff --git a/kernel/time/tick-sched.h b/kernel/time/tick-sched.h index 42fdf4958bcc..a4a8d4e9baa1 100644 --- a/kernel/time/tick-sched.h +++ b/kernel/time/tick-sched.h @@ -71,4 +71,14 @@ extern void tick_cancel_sched_timer(int cpu); static inline void tick_cancel_sched_timer(int cpu) { } #endif +#ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST +extern int __tick_broadcast_oneshot_control(enum tick_broadcast_state state); +#else +static inline int +__tick_broadcast_oneshot_control(enum tick_broadcast_state state) +{ + return -EBUSY; +} +#endif + #endif -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/