From 38ff87f77af0b5a93fc8581cff1d6e5692ab8970 Mon Sep 17 00:00:00 2001 From: Stephen Boyd Date: Sat, 1 Jun 2013 23:39:40 -0700 Subject: sched_clock: Make ARM's sched_clock generic for all architectures Nothing about the sched_clock implementation in the ARM port is specific to the architecture. Generalize the code so that other architectures can use it by selecting GENERIC_SCHED_CLOCK. Signed-off-by: Stephen Boyd [jstultz: Merge minor collisions with other patches in my tree] Signed-off-by: John Stultz --- kernel/time/sched_clock.c | 215 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 215 insertions(+) create mode 100644 kernel/time/sched_clock.c (limited to 'kernel/time/sched_clock.c') diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c new file mode 100644 index 000000000000..aad1ae6077ef --- /dev/null +++ b/kernel/time/sched_clock.c @@ -0,0 +1,215 @@ +/* + * sched_clock.c: support for extending counters to full 64-bit ns counter + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +struct clock_data { + u64 epoch_ns; + u32 epoch_cyc; + u32 epoch_cyc_copy; + 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 int irqtime = -1; + +core_param(irqtime, irqtime, int, 0400); + +static struct clock_data cd = { + .mult = NSEC_PER_SEC / HZ, +}; + +static u32 __read_mostly sched_clock_mask = 0xffffffff; + +static u32 notrace jiffy_sched_clock_read(void) +{ + return (u32)(jiffies - INITIAL_JIFFIES); +} + +static u32 __read_mostly (*read_sched_clock)(void) = jiffy_sched_clock_read; + +static inline u64 notrace cyc_to_ns(u64 cyc, u32 mult, u32 shift) +{ + return (cyc * mult) >> shift; +} + +static unsigned long long notrace cyc_to_sched_clock(u32 cyc, u32 mask) +{ + u64 epoch_ns; + u32 epoch_cyc; + + /* + * 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 { + epoch_cyc = cd.epoch_cyc; + smp_rmb(); + epoch_ns = cd.epoch_ns; + smp_rmb(); + } while (epoch_cyc != cd.epoch_cyc_copy); + + return epoch_ns + cyc_to_ns((cyc - epoch_cyc) & mask, cd.mult, cd.shift); +} + +/* + * Atomically update the sched_clock epoch. + */ +static void notrace update_sched_clock(void) +{ + unsigned long flags; + u32 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(); + cd.epoch_ns = ns; + smp_wmb(); + cd.epoch_cyc = cyc; + raw_local_irq_restore(flags); +} + +static void sched_clock_poll(unsigned long wrap_ticks) +{ + mod_timer(&sched_clock_timer, round_jiffies(jiffies + wrap_ticks)); + update_sched_clock(); +} + +void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate) +{ + unsigned long r, w; + 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; + 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); + + r = rate; + if (r >= 4000000) { + r /= 1000000; + r_unit = 'M'; + } else if (r >= 1000) { + r /= 1000; + r_unit = 'k'; + } else + 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; + + /* 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); + + /* + * 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(); + + /* + * Ensure that sched_clock() starts off at 0ns + */ + cd.epoch_ns = 0; + + /* Enable IRQ time accounting if we have a fast enough sched_clock */ + if (irqtime > 0 || (irqtime == -1 && rate >= 1000000)) + enable_sched_clock_irqtime(); + + pr_debug("Registered %pF as sched_clock source\n", read); +} + +static unsigned long long notrace sched_clock_32(void) +{ + u32 cyc = read_sched_clock(); + return cyc_to_sched_clock(cyc, sched_clock_mask); +} + +unsigned long long __read_mostly (*sched_clock_func)(void) = sched_clock_32; + +unsigned long long notrace sched_clock(void) +{ + if (cd.suspended) + return cd.epoch_ns; + + return sched_clock_func(); +} + +void __init sched_clock_postinit(void) +{ + /* + * If no sched_clock function has been provided at that point, + * 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_poll(sched_clock_timer.data); +} + +static int sched_clock_suspend(void) +{ + sched_clock_poll(sched_clock_timer.data); + cd.suspended = true; + return 0; +} + +static void sched_clock_resume(void) +{ + cd.epoch_cyc = read_sched_clock(); + cd.epoch_cyc_copy = cd.epoch_cyc; + cd.suspended = false; +} + +static struct syscore_ops sched_clock_ops = { + .suspend = sched_clock_suspend, + .resume = sched_clock_resume, +}; + +static int __init sched_clock_syscore_init(void) +{ + register_syscore_ops(&sched_clock_ops); + return 0; +} +device_initcall(sched_clock_syscore_init); -- cgit v1.2.3 From 336ae1180df5f69b9e0fb6561bec01c5f64361cf Mon Sep 17 00:00:00 2001 From: Stephen Boyd Date: Mon, 17 Jun 2013 15:40:58 -0700 Subject: ARM: sched_clock: Load cycle count after epoch stabilizes There is a small race between when the cycle count is read from the hardware and when the epoch stabilizes. Consider this scenario: CPU0 CPU1 ---- ---- cyc = read_sched_clock() cyc_to_sched_clock() update_sched_clock() ... cd.epoch_cyc = cyc; epoch_cyc = cd.epoch_cyc; ... epoch_ns + cyc_to_ns((cyc - epoch_cyc) The cyc on cpu0 was read before the epoch changed. But we calculate the nanoseconds based on the new epoch by subtracting the new epoch from the old cycle count. Since epoch is most likely larger than the old cycle count we calculate a large number that will be converted to nanoseconds and added to epoch_ns, causing time to jump forward too much. Fix this problem by reading the hardware after the epoch has stabilized. Cc: Russell King Signed-off-by: Stephen Boyd Signed-off-by: John Stultz --- kernel/time/sched_clock.c | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-) (limited to 'kernel/time/sched_clock.c') diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c index aad1ae6077ef..a326f27d7f09 100644 --- a/kernel/time/sched_clock.c +++ b/kernel/time/sched_clock.c @@ -49,10 +49,14 @@ static inline u64 notrace cyc_to_ns(u64 cyc, u32 mult, u32 shift) return (cyc * mult) >> shift; } -static unsigned long long notrace cyc_to_sched_clock(u32 cyc, u32 mask) +static unsigned long long notrace sched_clock_32(void) { u64 epoch_ns; u32 epoch_cyc; + u32 cyc; + + if (cd.suspended) + return cd.epoch_ns; /* * Load the epoch_cyc and epoch_ns atomically. We do this by @@ -68,7 +72,9 @@ static unsigned long long notrace cyc_to_sched_clock(u32 cyc, u32 mask) smp_rmb(); } while (epoch_cyc != cd.epoch_cyc_copy); - return epoch_ns + cyc_to_ns((cyc - epoch_cyc) & mask, cd.mult, cd.shift); + cyc = read_sched_clock(); + cyc = (cyc - epoch_cyc) & sched_clock_mask; + return epoch_ns + cyc_to_ns(cyc, cd.mult, cd.shift); } /* @@ -160,19 +166,10 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate) pr_debug("Registered %pF as sched_clock source\n", read); } -static unsigned long long notrace sched_clock_32(void) -{ - u32 cyc = read_sched_clock(); - return cyc_to_sched_clock(cyc, sched_clock_mask); -} - unsigned long long __read_mostly (*sched_clock_func)(void) = sched_clock_32; unsigned long long notrace sched_clock(void) { - if (cd.suspended) - return cd.epoch_ns; - return sched_clock_func(); } -- cgit v1.2.3