diff options
Diffstat (limited to 'kernel/time')
| -rw-r--r-- | kernel/time/Kconfig | 51 | ||||
| -rw-r--r-- | kernel/time/clocksource.c | 45 | ||||
| -rw-r--r-- | kernel/time/sched_clock.c | 111 | ||||
| -rw-r--r-- | kernel/time/tick-sched.c | 71 | ||||
| -rw-r--r-- | kernel/time/timer_list.c | 41 |
5 files changed, 196 insertions, 123 deletions
diff --git a/kernel/time/Kconfig b/kernel/time/Kconfig index 70f27e89012b..2b62fe86f9ec 100644 --- a/kernel/time/Kconfig +++ b/kernel/time/Kconfig @@ -105,7 +105,6 @@ config NO_HZ_FULL select RCU_USER_QS select RCU_NOCB_CPU select VIRT_CPU_ACCOUNTING_GEN - select CONTEXT_TRACKING_FORCE select IRQ_WORK help Adaptively try to shutdown the tick whenever possible, even when @@ -134,6 +133,56 @@ config NO_HZ_FULL_ALL Note the boot CPU will still be kept outside the range to handle the timekeeping duty. +config NO_HZ_FULL_SYSIDLE + bool "Detect full-system idle state for full dynticks system" + depends on NO_HZ_FULL + default n + help + At least one CPU must keep the scheduling-clock tick running for + timekeeping purposes whenever there is a non-idle CPU, where + "non-idle" also includes dynticks CPUs as long as they are + running non-idle tasks. Because the underlying adaptive-tick + support cannot distinguish between all CPUs being idle and + all CPUs each running a single task in dynticks mode, the + underlying support simply ensures that there is always a CPU + handling the scheduling-clock tick, whether or not all CPUs + are idle. This Kconfig option enables scalable detection of + the all-CPUs-idle state, thus allowing the scheduling-clock + tick to be disabled when all CPUs are idle. Note that scalable + detection of the all-CPUs-idle state means that larger systems + will be slower to declare the all-CPUs-idle state. + + Say Y if you would like to help debug all-CPUs-idle detection. + + Say N if you are unsure. + +config NO_HZ_FULL_SYSIDLE_SMALL + int "Number of CPUs above which large-system approach is used" + depends on NO_HZ_FULL_SYSIDLE + range 1 NR_CPUS + default 8 + help + The full-system idle detection mechanism takes a lazy approach + on large systems, as is required to attain decent scalability. + However, on smaller systems, scalability is not anywhere near as + large a concern as is energy efficiency. The sysidle subsystem + therefore uses a fast but non-scalable algorithm for small + systems and a lazier but scalable algorithm for large systems. + This Kconfig parameter defines the number of CPUs in the largest + system that will be considered to be "small". + + The default value will be fine in most cases. Battery-powered + systems that (1) enable NO_HZ_FULL_SYSIDLE, (2) have larger + numbers of CPUs, and (3) are suffering from battery-lifetime + problems due to long sysidle latencies might wish to experiment + with larger values for this Kconfig parameter. On the other + hand, they might be even better served by disabling NO_HZ_FULL + entirely, given that NO_HZ_FULL is intended for HPC and + real-time workloads that at present do not tend to be run on + battery-powered systems. + + Take the default if you are unsure. + config NO_HZ bool "Old Idle dynticks config" depends on !ARCH_USES_GETTIMEOFFSET && GENERIC_CLOCKEVENTS diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index a2e72b8d28d0..64cf63ca09cc 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -538,40 +538,55 @@ static u32 clocksource_max_adjustment(struct clocksource *cs) } /** - * clocksource_max_deferment - Returns max time the clocksource can be deferred - * @cs: Pointer to clocksource - * + * clocks_calc_max_nsecs - Returns maximum nanoseconds that can be converted + * @mult: cycle to nanosecond multiplier + * @shift: cycle to nanosecond divisor (power of two) + * @maxadj: maximum adjustment value to mult (~11%) + * @mask: bitmask for two's complement subtraction of non 64 bit counters */ -static u64 clocksource_max_deferment(struct clocksource *cs) +u64 clocks_calc_max_nsecs(u32 mult, u32 shift, u32 maxadj, u64 mask) { u64 max_nsecs, max_cycles; /* * Calculate the maximum number of cycles that we can pass to the * cyc2ns function without overflowing a 64-bit signed result. The - * maximum number of cycles is equal to ULLONG_MAX/(cs->mult+cs->maxadj) + * maximum number of cycles is equal to ULLONG_MAX/(mult+maxadj) * which is equivalent to the below. - * max_cycles < (2^63)/(cs->mult + cs->maxadj) - * max_cycles < 2^(log2((2^63)/(cs->mult + cs->maxadj))) - * max_cycles < 2^(log2(2^63) - log2(cs->mult + cs->maxadj)) - * max_cycles < 2^(63 - log2(cs->mult + cs->maxadj)) - * max_cycles < 1 << (63 - log2(cs->mult + cs->maxadj)) + * max_cycles < (2^63)/(mult + maxadj) + * max_cycles < 2^(log2((2^63)/(mult + maxadj))) + * max_cycles < 2^(log2(2^63) - log2(mult + maxadj)) + * max_cycles < 2^(63 - log2(mult + maxadj)) + * max_cycles < 1 << (63 - log2(mult + maxadj)) * Please note that we add 1 to the result of the log2 to account for * any rounding errors, ensure the above inequality is satisfied and * no overflow will occur. */ - max_cycles = 1ULL << (63 - (ilog2(cs->mult + cs->maxadj) + 1)); + max_cycles = 1ULL << (63 - (ilog2(mult + maxadj) + 1)); /* * The actual maximum number of cycles we can defer the clocksource is - * determined by the minimum of max_cycles and cs->mask. + * determined by the minimum of max_cycles and mask. * Note: Here we subtract the maxadj to make sure we don't sleep for * too long if there's a large negative adjustment. */ - max_cycles = min_t(u64, max_cycles, (u64) cs->mask); - max_nsecs = clocksource_cyc2ns(max_cycles, cs->mult - cs->maxadj, - cs->shift); + max_cycles = min(max_cycles, mask); + max_nsecs = clocksource_cyc2ns(max_cycles, mult - maxadj, shift); + + return max_nsecs; +} + +/** + * clocksource_max_deferment - Returns max time the clocksource can be deferred + * @cs: Pointer to clocksource + * + */ +static u64 clocksource_max_deferment(struct clocksource *cs) +{ + u64 max_nsecs; + max_nsecs = clocks_calc_max_nsecs(cs->mult, cs->shift, cs->maxadj, + cs->mask); /* * To ensure that the clocksource does not wrap whilst we are idle, * limit the time the clocksource can be deferred by 12.5%. Please diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c index a326f27d7f09..f388baeaf2b6 100644 --- a/kernel/time/sched_clock.c +++ b/kernel/time/sched_clock.c @@ -8,25 +8,28 @@ #include <linux/clocksource.h> #include <linux/init.h> #include <linux/jiffies.h> +#include <linux/ktime.h> #include <linux/kernel.h> #include <linux/moduleparam.h> #include <linux/sched.h> #include <linux/syscore_ops.h> -#include <linux/timer.h> +#include <linux/hrtimer.h> #include <linux/sched_clock.h> +#include <linux/seqlock.h> +#include <linux/bitops.h> struct clock_data { + ktime_t wrap_kt; u64 epoch_ns; - u32 epoch_cyc; - u32 epoch_cyc_copy; + u64 epoch_cyc; + seqcount_t seq; unsigned long rate; u32 mult; u32 shift; bool suspended; }; -static void sched_clock_poll(unsigned long wrap_ticks); -static DEFINE_TIMER(sched_clock_timer, sched_clock_poll, 0, 0); +static struct hrtimer sched_clock_timer; static int irqtime = -1; core_param(irqtime, irqtime, int, 0400); @@ -35,14 +38,25 @@ static struct clock_data cd = { .mult = NSEC_PER_SEC / HZ, }; -static u32 __read_mostly sched_clock_mask = 0xffffffff; +static u64 __read_mostly sched_clock_mask; -static u32 notrace jiffy_sched_clock_read(void) +static u64 notrace jiffy_sched_clock_read(void) { - return (u32)(jiffies - INITIAL_JIFFIES); + /* + * We don't need to use get_jiffies_64 on 32-bit arches here + * because we register with BITS_PER_LONG + */ + return (u64)(jiffies - INITIAL_JIFFIES); } -static u32 __read_mostly (*read_sched_clock)(void) = jiffy_sched_clock_read; +static u32 __read_mostly (*read_sched_clock_32)(void); + +static u64 notrace read_sched_clock_32_wrapper(void) +{ + return read_sched_clock_32(); +} + +static u64 __read_mostly (*read_sched_clock)(void) = jiffy_sched_clock_read; static inline u64 notrace cyc_to_ns(u64 cyc, u32 mult, u32 shift) { @@ -52,25 +66,18 @@ static inline u64 notrace cyc_to_ns(u64 cyc, u32 mult, u32 shift) static unsigned long long notrace sched_clock_32(void) { u64 epoch_ns; - u32 epoch_cyc; - u32 cyc; + u64 epoch_cyc; + u64 cyc; + unsigned long seq; if (cd.suspended) return cd.epoch_ns; - /* - * Load the epoch_cyc and epoch_ns atomically. We do this by - * ensuring that we always write epoch_cyc, epoch_ns and - * epoch_cyc_copy in strict order, and read them in strict order. - * If epoch_cyc and epoch_cyc_copy are not equal, then we're in - * the middle of an update, and we should repeat the load. - */ do { + seq = read_seqcount_begin(&cd.seq); epoch_cyc = cd.epoch_cyc; - smp_rmb(); epoch_ns = cd.epoch_ns; - smp_rmb(); - } while (epoch_cyc != cd.epoch_cyc_copy); + } while (read_seqcount_retry(&cd.seq, seq)); cyc = read_sched_clock(); cyc = (cyc - epoch_cyc) & sched_clock_mask; @@ -83,49 +90,46 @@ static unsigned long long notrace sched_clock_32(void) static void notrace update_sched_clock(void) { unsigned long flags; - u32 cyc; + u64 cyc; u64 ns; cyc = read_sched_clock(); ns = cd.epoch_ns + cyc_to_ns((cyc - cd.epoch_cyc) & sched_clock_mask, cd.mult, cd.shift); - /* - * Write epoch_cyc and epoch_ns in a way that the update is - * detectable in cyc_to_fixed_sched_clock(). - */ + raw_local_irq_save(flags); - cd.epoch_cyc_copy = cyc; - smp_wmb(); + write_seqcount_begin(&cd.seq); cd.epoch_ns = ns; - smp_wmb(); cd.epoch_cyc = cyc; + write_seqcount_end(&cd.seq); raw_local_irq_restore(flags); } -static void sched_clock_poll(unsigned long wrap_ticks) +static enum hrtimer_restart sched_clock_poll(struct hrtimer *hrt) { - mod_timer(&sched_clock_timer, round_jiffies(jiffies + wrap_ticks)); update_sched_clock(); + hrtimer_forward_now(hrt, cd.wrap_kt); + return HRTIMER_RESTART; } -void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate) +void __init sched_clock_register(u64 (*read)(void), int bits, + unsigned long rate) { - unsigned long r, w; + unsigned long r; u64 res, wrap; char r_unit; if (cd.rate > rate) return; - BUG_ON(bits > 32); WARN_ON(!irqs_disabled()); read_sched_clock = read; - sched_clock_mask = (1 << bits) - 1; + sched_clock_mask = CLOCKSOURCE_MASK(bits); cd.rate = rate; /* calculate the mult/shift to convert counter ticks to ns. */ - clocks_calc_mult_shift(&cd.mult, &cd.shift, rate, NSEC_PER_SEC, 0); + clocks_calc_mult_shift(&cd.mult, &cd.shift, rate, NSEC_PER_SEC, 3600); r = rate; if (r >= 4000000) { @@ -138,20 +142,14 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate) r_unit = ' '; /* calculate how many ns until we wrap */ - wrap = cyc_to_ns((1ULL << bits) - 1, cd.mult, cd.shift); - do_div(wrap, NSEC_PER_MSEC); - w = wrap; + wrap = clocks_calc_max_nsecs(cd.mult, cd.shift, 0, sched_clock_mask); + cd.wrap_kt = ns_to_ktime(wrap - (wrap >> 3)); /* calculate the ns resolution of this counter */ res = cyc_to_ns(1ULL, cd.mult, cd.shift); - pr_info("sched_clock: %u bits at %lu%cHz, resolution %lluns, wraps every %lums\n", - bits, r, r_unit, res, w); + pr_info("sched_clock: %u bits at %lu%cHz, resolution %lluns, wraps every %lluns\n", + bits, r, r_unit, res, wrap); - /* - * Start the timer to keep sched_clock() properly updated and - * sets the initial epoch. - */ - sched_clock_timer.data = msecs_to_jiffies(w - (w / 10)); update_sched_clock(); /* @@ -166,6 +164,12 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate) pr_debug("Registered %pF as sched_clock source\n", read); } +void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate) +{ + read_sched_clock_32 = read; + sched_clock_register(read_sched_clock_32_wrapper, bits, rate); +} + unsigned long long __read_mostly (*sched_clock_func)(void) = sched_clock_32; unsigned long long notrace sched_clock(void) @@ -180,14 +184,22 @@ void __init sched_clock_postinit(void) * make it the final one one. */ if (read_sched_clock == jiffy_sched_clock_read) - setup_sched_clock(jiffy_sched_clock_read, 32, HZ); + sched_clock_register(jiffy_sched_clock_read, BITS_PER_LONG, HZ); - sched_clock_poll(sched_clock_timer.data); + update_sched_clock(); + + /* + * Start the timer to keep sched_clock() properly updated and + * sets the initial epoch. + */ + hrtimer_init(&sched_clock_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + sched_clock_timer.function = sched_clock_poll; + hrtimer_start(&sched_clock_timer, cd.wrap_kt, HRTIMER_MODE_REL); } static int sched_clock_suspend(void) { - sched_clock_poll(sched_clock_timer.data); + sched_clock_poll(&sched_clock_timer); cd.suspended = true; return 0; } @@ -195,7 +207,6 @@ static int sched_clock_suspend(void) static void sched_clock_resume(void) { cd.epoch_cyc = read_sched_clock(); - cd.epoch_cyc_copy = cd.epoch_cyc; cd.suspended = false; } diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 69601726a745..3612fc77f834 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -23,6 +23,7 @@ #include <linux/irq_work.h> #include <linux/posix-timers.h> #include <linux/perf_event.h> +#include <linux/context_tracking.h> #include <asm/irq_regs.h> @@ -148,8 +149,8 @@ static void tick_sched_handle(struct tick_sched *ts, struct pt_regs *regs) } #ifdef CONFIG_NO_HZ_FULL -static cpumask_var_t nohz_full_mask; -bool have_nohz_full_mask; +cpumask_var_t tick_nohz_full_mask; +bool tick_nohz_full_running; static bool can_stop_full_tick(void) { @@ -182,7 +183,8 @@ static bool can_stop_full_tick(void) * Don't allow the user to think they can get * full NO_HZ with this machine. */ - WARN_ONCE(1, "NO_HZ FULL will not work with unstable sched clock"); + WARN_ONCE(tick_nohz_full_running, + "NO_HZ FULL will not work with unstable sched clock"); return false; } #endif @@ -196,7 +198,7 @@ static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now); * Re-evaluate the need for the tick on the current CPU * and restart it if necessary. */ -void tick_nohz_full_check(void) +void __tick_nohz_full_check(void) { struct tick_sched *ts = &__get_cpu_var(tick_cpu_sched); @@ -210,7 +212,7 @@ void tick_nohz_full_check(void) static void nohz_full_kick_work_func(struct irq_work *work) { - tick_nohz_full_check(); + __tick_nohz_full_check(); } static DEFINE_PER_CPU(struct irq_work, nohz_full_kick_work) = { @@ -229,7 +231,7 @@ void tick_nohz_full_kick(void) static void nohz_full_kick_ipi(void *info) { - tick_nohz_full_check(); + __tick_nohz_full_check(); } /* @@ -238,12 +240,13 @@ static void nohz_full_kick_ipi(void *info) */ void tick_nohz_full_kick_all(void) { - if (!have_nohz_full_mask) + if (!tick_nohz_full_running) return; preempt_disable(); - smp_call_function_many(nohz_full_mask, + smp_call_function_many(tick_nohz_full_mask, nohz_full_kick_ipi, NULL, false); + tick_nohz_full_kick(); preempt_enable(); } @@ -252,7 +255,7 @@ void tick_nohz_full_kick_all(void) * It might need the tick due to per task/process properties: * perf events, posix cpu timers, ... */ -void tick_nohz_task_switch(struct task_struct *tsk) +void __tick_nohz_task_switch(struct task_struct *tsk) { unsigned long flags; @@ -268,37 +271,29 @@ out: local_irq_restore(flags); } -int tick_nohz_full_cpu(int cpu) -{ - if (!have_nohz_full_mask) - return 0; - - return cpumask_test_cpu(cpu, nohz_full_mask); -} - /* Parse the boot-time nohz CPU list from the kernel parameters. */ static int __init tick_nohz_full_setup(char *str) { int cpu; - alloc_bootmem_cpumask_var(&nohz_full_mask); - if (cpulist_parse(str, nohz_full_mask) < 0) { + alloc_bootmem_cpumask_var(&tick_nohz_full_mask); + if (cpulist_parse(str, tick_nohz_full_mask) < 0) { pr_warning("NOHZ: Incorrect nohz_full cpumask\n"); return 1; } cpu = smp_processor_id(); - if (cpumask_test_cpu(cpu, nohz_full_mask)) { + if (cpumask_test_cpu(cpu, tick_nohz_full_mask)) { pr_warning("NO_HZ: Clearing %d from nohz_full range for timekeeping\n", cpu); - cpumask_clear_cpu(cpu, nohz_full_mask); + cpumask_clear_cpu(cpu, tick_nohz_full_mask); } - have_nohz_full_mask = true; + tick_nohz_full_running = true; return 1; } __setup("nohz_full=", tick_nohz_full_setup); -static int __cpuinit tick_nohz_cpu_down_callback(struct notifier_block *nfb, +static int tick_nohz_cpu_down_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) { @@ -310,7 +305,7 @@ static int __cpuinit tick_nohz_cpu_down_callback(struct notifier_block *nfb, * If we handle the timekeeping duty for full dynticks CPUs, * we can't safely shutdown that CPU. */ - if (have_nohz_full_mask && tick_do_timer_cpu == cpu) + if (tick_nohz_full_running && tick_do_timer_cpu == cpu) return NOTIFY_BAD; break; } @@ -329,14 +324,14 @@ static int tick_nohz_init_all(void) int err = -1; #ifdef CONFIG_NO_HZ_FULL_ALL - if (!alloc_cpumask_var(&nohz_full_mask, GFP_KERNEL)) { + if (!alloc_cpumask_var(&tick_nohz_full_mask, GFP_KERNEL)) { pr_err("NO_HZ: Can't allocate full dynticks cpumask\n"); return err; } err = 0; - cpumask_setall(nohz_full_mask); - cpumask_clear_cpu(smp_processor_id(), nohz_full_mask); - have_nohz_full_mask = true; + cpumask_setall(tick_nohz_full_mask); + cpumask_clear_cpu(smp_processor_id(), tick_nohz_full_mask); + tick_nohz_full_running = true; #endif return err; } @@ -345,17 +340,18 @@ void __init tick_nohz_init(void) { int cpu; - if (!have_nohz_full_mask) { + if (!tick_nohz_full_running) { if (tick_nohz_init_all() < 0) return; } + for_each_cpu(cpu, tick_nohz_full_mask) + context_tracking_cpu_set(cpu); + cpu_notifier(tick_nohz_cpu_down_callback, 0); - cpulist_scnprintf(nohz_full_buf, sizeof(nohz_full_buf), nohz_full_mask); + cpulist_scnprintf(nohz_full_buf, sizeof(nohz_full_buf), tick_nohz_full_mask); pr_info("NO_HZ: Full dynticks CPUs: %s.\n", nohz_full_buf); } -#else -#define have_nohz_full_mask (0) #endif /* @@ -733,7 +729,7 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts) return false; } - if (have_nohz_full_mask) { + if (tick_nohz_full_enabled()) { /* * Keep the tick alive to guarantee timekeeping progression * if there are full dynticks CPUs around @@ -827,13 +823,10 @@ void tick_nohz_irq_exit(void) { struct tick_sched *ts = &__get_cpu_var(tick_cpu_sched); - if (ts->inidle) { - /* Cancel the timer because CPU already waken up from the C-states*/ - menu_hrtimer_cancel(); + if (ts->inidle) __tick_nohz_idle_enter(ts); - } else { + else tick_nohz_full_stop_tick(ts); - } } /** @@ -931,8 +924,6 @@ void tick_nohz_idle_exit(void) ts->inidle = 0; - /* Cancel the timer because CPU already waken up from the C-states*/ - menu_hrtimer_cancel(); if (ts->idle_active || ts->tick_stopped) now = ktime_get(); diff --git a/kernel/time/timer_list.c b/kernel/time/timer_list.c index 3bdf28323012..61ed862cdd37 100644 --- a/kernel/time/timer_list.c +++ b/kernel/time/timer_list.c @@ -265,10 +265,9 @@ static inline void timer_list_header(struct seq_file *m, u64 now) static int timer_list_show(struct seq_file *m, void *v) { struct timer_list_iter *iter = v; - u64 now = ktime_to_ns(ktime_get()); if (iter->cpu == -1 && !iter->second_pass) - timer_list_header(m, now); + timer_list_header(m, iter->now); else if (!iter->second_pass) print_cpu(m, iter->cpu, iter->now); #ifdef CONFIG_GENERIC_CLOCKEVENTS @@ -298,33 +297,41 @@ void sysrq_timer_list_show(void) return; } -static void *timer_list_start(struct seq_file *file, loff_t *offset) +static void *move_iter(struct timer_list_iter *iter, loff_t offset) { - struct timer_list_iter *iter = file->private; - - if (!*offset) { - iter->cpu = -1; - iter->now = ktime_to_ns(ktime_get()); - } else if (iter->cpu >= nr_cpu_ids) { + for (; offset; offset--) { + iter->cpu = cpumask_next(iter->cpu, cpu_online_mask); + if (iter->cpu >= nr_cpu_ids) { #ifdef CONFIG_GENERIC_CLOCKEVENTS - if (!iter->second_pass) { - iter->cpu = -1; - iter->second_pass = true; - } else - return NULL; + if (!iter->second_pass) { + iter->cpu = -1; + iter->second_pass = true; + } else + return NULL; #else - return NULL; + return NULL; #endif + } } return iter; } +static void *timer_list_start(struct seq_file *file, loff_t *offset) +{ + struct timer_list_iter *iter = file->private; + + if (!*offset) + iter->now = ktime_to_ns(ktime_get()); + iter->cpu = -1; + iter->second_pass = false; + return move_iter(iter, *offset); +} + static void *timer_list_next(struct seq_file *file, void *v, loff_t *offset) { struct timer_list_iter *iter = file->private; - iter->cpu = cpumask_next(iter->cpu, cpu_online_mask); ++*offset; - return timer_list_start(file, offset); + return move_iter(iter, 1); } static void timer_list_stop(struct seq_file *seq, void *v) |