From e05b2efb82596905ebfe88e8612ee81dec9b6592 Mon Sep 17 00:00:00 2001 From: john stultz Date: Wed, 4 May 2011 18:16:50 -0700 Subject: clocksource: Install completely before selecting Christian Hoffmann reported that the command line clocksource override with acpi_pm timer fails: Kernel command line: clocksource=acpi_pm hpet clockevent registered Switching to clocksource hpet Override clocksource acpi_pm is not HRT compatible. Cannot switch while in HRT/NOHZ mode. The watchdog code is what enables CLOCK_SOURCE_VALID_FOR_HRES, but we actually end up selecting the clocksource before we enqueue it into the watchdog list, so that's why we see the warning and fail to switch to acpi_pm timer as requested. That's particularly bad when we want to debug timekeeping related problems in early boot. Put the selection call last. Reported-by: Christian Hoffmann Signed-off-by: John Stultz Cc: stable@kernel.org # 32... Link: http://lkml.kernel.org/r/%3C1304558210.2943.24.camel%40work-vm%3E Signed-off-by: Thomas Gleixner --- kernel/time/clocksource.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel/time') diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index 6519cf62d9cd..0e17c10f8a9d 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -685,8 +685,8 @@ int __clocksource_register_scale(struct clocksource *cs, u32 scale, u32 freq) /* Add clocksource to the clcoksource list */ mutex_lock(&clocksource_mutex); clocksource_enqueue(cs); - clocksource_select(); clocksource_enqueue_watchdog(cs); + clocksource_select(); mutex_unlock(&clocksource_mutex); return 0; } @@ -706,8 +706,8 @@ int clocksource_register(struct clocksource *cs) mutex_lock(&clocksource_mutex); clocksource_enqueue(cs); - clocksource_select(); clocksource_enqueue_watchdog(cs); + clocksource_select(); mutex_unlock(&clocksource_mutex); return 0; } -- cgit v1.2.3-58-ga151 From 07f4beb0b5bbfaf36a64aa00d59e670ec578a95a Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Mon, 16 May 2011 11:07:48 +0200 Subject: tick: Clear broadcast active bit when switching to oneshot The first cpu which switches from periodic to oneshot mode switches also the broadcast device into oneshot mode. The broadcast device serves as a backup for per cpu timers which stop in deeper C-states. To avoid starvation of the cpus which might be in idle and depend on broadcast mode it marks the other cpus as broadcast active and sets the brodcast expiry value of those cpus to the next tick. The oneshot mode broadcast bit for the other cpus is sticky and gets only cleared when those cpus exit idle. If a cpu was not idle while the bit got set in consequence the bit prevents that the broadcast device is armed on behalf of that cpu when it enters idle for the first time after it switched to oneshot mode. In most cases that goes unnoticed as one of the other cpus has usually a timer pending which keeps the broadcast device armed with a short timeout. Now if the only cpu which has a short timer active has the bit set then the broadcast device will not be armed on behalf of that cpu and will fire way after the expected timer expiry. In the case of Christians bug report it took ~145 seconds which is about half of the wrap around time of HPET (the limit for that device) due to the fact that all other cpus had no timers armed which expired before the 145 seconds timeframe. The solution is simply to clear the broadcast active bit unconditionally when a cpu switches to oneshot mode after the first cpu switched the broadcast device over. It's not idle at that point otherwise it would not be executing that code. [ I fundamentally hate that broadcast crap. Why the heck thought some folks that when going into deep idle it's a brilliant concept to switch off the last device which brings the cpu back from that state? ] Thanks to Christian for providing all the valuable debug information! Reported-and-tested-by: Christian Hoffmann Cc: John Stultz Link: http://lkml.kernel.org/r/%3Calpine.LFD.2.02.1105161105170.3078%40ionos%3E Cc: stable@kernel.org Signed-off-by: Thomas Gleixner --- kernel/time/tick-broadcast.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'kernel/time') diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c index da800ffa810c..723c7637e55a 100644 --- a/kernel/time/tick-broadcast.c +++ b/kernel/time/tick-broadcast.c @@ -522,10 +522,11 @@ static void tick_broadcast_init_next_event(struct cpumask *mask, */ void tick_broadcast_setup_oneshot(struct clock_event_device *bc) { + int cpu = smp_processor_id(); + /* Set it up only once ! */ if (bc->event_handler != tick_handle_oneshot_broadcast) { int was_periodic = bc->mode == CLOCK_EVT_MODE_PERIODIC; - int cpu = smp_processor_id(); bc->event_handler = tick_handle_oneshot_broadcast; clockevents_set_mode(bc, CLOCK_EVT_MODE_ONESHOT); @@ -551,6 +552,15 @@ void tick_broadcast_setup_oneshot(struct clock_event_device *bc) tick_broadcast_set_event(tick_next_period, 1); } else bc->next_event.tv64 = KTIME_MAX; + } else { + /* + * The first cpu which switches to oneshot mode sets + * the bit for all other cpus which are in the general + * (periodic) broadcast mask. So the bit is set and + * would prevent the first broadcast enter after this + * to program the bc device. + */ + tick_broadcast_clear_oneshot(cpu); } } -- cgit v1.2.3-58-ga151 From 724ed53e8ac2c5278af8955673049714c1073464 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 18 May 2011 21:33:40 +0000 Subject: clocksource: Get rid of the hardcoded 5 seconds sleep time limit Slow clocksources can have a way longer sleep time than 5 seconds and even fast ones can easily cope with 600 seconds and still maintain proper accuracy. Signed-off-by: Thomas Gleixner Cc: John Stultz Reviewed-by: Ingo Molnar Link: http://lkml.kernel.org/r/%3C20110518210136.109811585%40linutronix.de%3E --- kernel/time/clocksource.c | 38 +++++++++++++++++++------------------- 1 file changed, 19 insertions(+), 19 deletions(-) (limited to 'kernel/time') diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index 6519cf62d9cd..6dbbbb1ae6ba 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -626,19 +626,6 @@ static void clocksource_enqueue(struct clocksource *cs) list_add(&cs->list, entry); } - -/* - * Maximum time we expect to go between ticks. This includes idle - * tickless time. It provides the trade off between selecting a - * mult/shift pair that is very precise but can only handle a short - * period of time, vs. a mult/shift pair that can handle long periods - * of time but isn't as precise. - * - * This is a subsystem constant, and actual hardware limitations - * may override it (ie: clocksources that wrap every 3 seconds). - */ -#define MAX_UPDATE_LENGTH 5 /* Seconds */ - /** * __clocksource_updatefreq_scale - Used update clocksource with new freq * @t: clocksource to be registered @@ -652,15 +639,28 @@ static void clocksource_enqueue(struct clocksource *cs) */ void __clocksource_updatefreq_scale(struct clocksource *cs, u32 scale, u32 freq) { + unsigned long sec; + /* - * Ideally we want to use some of the limits used in - * clocksource_max_deferment, to provide a more informed - * MAX_UPDATE_LENGTH. But for now this just gets the - * register interface working properly. + * Calc the maximum number of seconds which we can run before + * wrapping around. For clocksources which have a mask > 32bit + * we need to limit the max sleep time to have a good + * conversion precision. 10 minutes is still a reasonable + * amount. That results in a shift value of 24 for a + * clocksource with mask >= 40bit and f >= 4GHz. That maps to + * ~ 0.06ppm granularity for NTP. We apply the same 12.5% + * margin as we do in clocksource_max_deferment() */ + sec = (cs->mask - (cs->mask >> 5)); + do_div(sec, freq); + do_div(sec, scale); + if (!sec) + sec = 1; + else if (sec > 600 && cs->mask > UINT_MAX) + sec = 600; + clocks_calc_mult_shift(&cs->mult, &cs->shift, freq, - NSEC_PER_SEC/scale, - MAX_UPDATE_LENGTH*scale); + NSEC_PER_SEC / scale, sec * scale); cs->max_idle_ns = clocksource_max_deferment(cs); } EXPORT_SYMBOL_GPL(__clocksource_updatefreq_scale); -- cgit v1.2.3-58-ga151 From 57f0fcbe1dea8a36c9d1673086326059991c5f81 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 18 May 2011 21:33:41 +0000 Subject: clockevents: Provide combined configure and register function All clockevent devices have the same open coded initialization functions. Provide an interface which does all necessary initialization in the core code. Signed-off-by: Thomas Gleixner Cc: John Stultz Reviewed-by: Ingo Molnar Link: http://lkml.kernel.org/r/%3C20110518210136.331975870%40linutronix.de%3E --- include/linux/clockchips.h | 9 +++++++++ kernel/time/clockevents.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 53 insertions(+) (limited to 'kernel/time') diff --git a/include/linux/clockchips.h b/include/linux/clockchips.h index 9466eebc8e19..80acc79e0dc5 100644 --- a/include/linux/clockchips.h +++ b/include/linux/clockchips.h @@ -69,6 +69,8 @@ enum clock_event_nofitiers { * @retries: number of forced programming retries * @set_mode: set mode function * @broadcast: function to broadcast events + * @min_delta_ticks: minimum delta value in ticks stored for reconfiguration + * @max_delta_ticks: maximum delta value in ticks stored for reconfiguration * @name: ptr to clock event name * @rating: variable to rate clock event devices * @irq: IRQ number (only for non CPU local devices) @@ -91,6 +93,9 @@ struct clock_event_device { void (*broadcast)(const struct cpumask *mask); void (*set_mode)(enum clock_event_mode mode, struct clock_event_device *); + unsigned long min_delta_ticks; + unsigned long max_delta_ticks; + const char *name; int rating; int irq; @@ -123,6 +128,10 @@ extern u64 clockevent_delta2ns(unsigned long latch, struct clock_event_device *evt); extern void clockevents_register_device(struct clock_event_device *dev); +extern void clockevents_config_and_register(struct clock_event_device *dev, + u32 freq, unsigned long min_delta, + unsigned long max_delta); + extern void clockevents_exchange_device(struct clock_event_device *old, struct clock_event_device *new); extern void clockevents_set_mode(struct clock_event_device *dev, diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c index 0d74b9ba90c8..c69e88c94446 100644 --- a/kernel/time/clockevents.c +++ b/kernel/time/clockevents.c @@ -194,6 +194,50 @@ void clockevents_register_device(struct clock_event_device *dev) } EXPORT_SYMBOL_GPL(clockevents_register_device); +static void clockevents_config(struct clock_event_device *dev, + u32 freq) +{ + unsigned long sec; + + if (!(dev->features & CLOCK_EVT_FEAT_ONESHOT)) + return; + + /* + * Calculate the maximum number of seconds we can sleep. Limit + * to 10 minutes for hardware which can program more than + * 32bit ticks so we still get reasonable conversion values. + */ + sec = dev->max_delta_ticks; + do_div(sec, freq); + if (!sec) + sec = 1; + else if (sec > 600 && dev->max_delta_ticks > UINT_MAX) + sec = 600; + + clockevents_calc_mult_shift(dev, freq, sec); + dev->min_delta_ns = clockevent_delta2ns(dev->min_delta_ticks, dev); + dev->max_delta_ns = clockevent_delta2ns(dev->max_delta_ticks, dev); +} + +/** + * clockevents_config_and_register - Configure and register a clock event device + * @dev: device to register + * @freq: The clock frequency + * @min_delta: The minimum clock ticks to program in oneshot mode + * @max_delta: The maximum clock ticks to program in oneshot mode + * + * min/max_delta can be 0 for devices which do not support oneshot mode. + */ +void clockevents_config_and_register(struct clock_event_device *dev, + u32 freq, unsigned long min_delta, + unsigned long max_delta) +{ + dev->min_delta_ticks = min_delta; + dev->max_delta_ticks = max_delta; + clockevents_config(dev, freq); + clockevents_register_device(dev); +} + /* * Noop handler when we shut down an event device */ -- cgit v1.2.3-58-ga151 From 80b816b736cfa5b9582279127099b20a479ab7d9 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 18 May 2011 21:33:42 +0000 Subject: clockevents: Provide interface to reconfigure an active clock event device Some ARM SoCs have clock event devices which have their frequency modified due to frequency scaling. Provide an interface which allows to reconfigure an active device. After reconfiguration reprogram the current pending event. Signed-off-by: Thomas Gleixner Cc: LAK Cc: John Stultz Acked-by: Linus Walleij Reviewed-by: Ingo Molnar Link: http://lkml.kernel.org/r/%3C20110518210136.437459958%40linutronix.de%3E --- include/linux/clockchips.h | 2 ++ kernel/time/clockevents.c | 20 ++++++++++++++++++++ 2 files changed, 22 insertions(+) (limited to 'kernel/time') diff --git a/include/linux/clockchips.h b/include/linux/clockchips.h index 80acc79e0dc5..d6733e27af34 100644 --- a/include/linux/clockchips.h +++ b/include/linux/clockchips.h @@ -132,6 +132,8 @@ extern void clockevents_config_and_register(struct clock_event_device *dev, u32 freq, unsigned long min_delta, unsigned long max_delta); +extern int clockevents_update_freq(struct clock_event_device *ce, u32 freq); + extern void clockevents_exchange_device(struct clock_event_device *old, struct clock_event_device *new); extern void clockevents_set_mode(struct clock_event_device *dev, diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c index c69e88c94446..22a9da9a9c96 100644 --- a/kernel/time/clockevents.c +++ b/kernel/time/clockevents.c @@ -238,6 +238,26 @@ void clockevents_config_and_register(struct clock_event_device *dev, clockevents_register_device(dev); } +/** + * clockevents_update_freq - Update frequency and reprogram a clock event device. + * @dev: device to modify + * @freq: new device frequency + * + * Reconfigure and reprogram a clock event device in oneshot + * mode. Must be called on the cpu for which the device delivers per + * cpu timer events with interrupts disabled! Returns 0 on success, + * -ETIME when the event is in the past. + */ +int clockevents_update_freq(struct clock_event_device *dev, u32 freq) +{ + clockevents_config(dev, freq); + + if (dev->mode != CLOCK_EVT_MODE_ONESHOT) + return 0; + + return clockevents_program_event(dev, dev->next_event, ktime_get()); +} + /* * Noop handler when we shut down an event device */ -- cgit v1.2.3-58-ga151 From c0e299b1a91cbdb21ae08e382a4176200398bc36 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Fri, 20 May 2011 10:50:52 +0200 Subject: clockevents/source: Use u64 to make 32bit happy unsigned long is not 64bit on 32bit machine. Signed-off-by: Thomas Gleixner --- kernel/time/clockevents.c | 2 +- kernel/time/clocksource.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel/time') diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c index 22a9da9a9c96..c027d4f602f1 100644 --- a/kernel/time/clockevents.c +++ b/kernel/time/clockevents.c @@ -197,7 +197,7 @@ EXPORT_SYMBOL_GPL(clockevents_register_device); static void clockevents_config(struct clock_event_device *dev, u32 freq) { - unsigned long sec; + u64 sec; if (!(dev->features & CLOCK_EVT_FEAT_ONESHOT)) return; diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index d9d5f8c885f6..1c95fd677328 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -639,7 +639,7 @@ static void clocksource_enqueue(struct clocksource *cs) */ void __clocksource_updatefreq_scale(struct clocksource *cs, u32 scale, u32 freq) { - unsigned long sec; + u64 sec; /* * Calc the maximum number of seconds which we can run before -- cgit v1.2.3-58-ga151