Messages in this thread | | | Subject | Re: [RFC][PATCH] Dynamic Tick: Allow 32-bit machines to sleep formorethan2.15 seconds | From | john stultz <> | Date | Fri, 22 May 2009 12:23:12 -0700 |
| |
On Fri, 2009-05-22 at 13:21 -0500, Jon Hunter wrote: > John Stultz wrote: > > So cyc2ns is a very very hot path, so I don't think we want to muck with > > that much. > > > > Instead of catching the overflows, and then trying to handle them, we > > really want to prevent overflows from happening. That is the main point > > of the timekeeping_max_deferrment() interface after all ;) > > > > So thinking about it a bit more, what we really want from > > timekeeping_max_deferrment() is roughly: > > > > /* find the max cycle value that would overflow the mult */ > > max_cycles = -1UUL/clocksource->mult; > > > > /* pick the smaller of max_cycles or the mask value */ > > max_cycles = min(max_cycles, clocksource->mask); > > > > /* convert max_cycles into ns */ > > max_time = cyc2ns(clocksource, max_cycles); > > > > /* take off 5% of the max to make sure we don't show up late */ > > max_time = max_time - max_time/20; > > > > > > We should be able to make this reasonably fast via: > > max_cycles = 1<<(64 - ilog2(clocksource->mult) - 1); > > max_cycles = min(max_cycles, clocksource->mask); > > max_time = cyc2ns(clocksource, max_cycles); > > max_time = max_time - max_time >> 4; > > > > Does that seem reasonable? > > Yes, good idea. Sorry for the delay in getting back to this. I have > implemented the above and incorporated into the patch below. > > Please note that the only modification I made to the above was to add 1 > to result of the log2. I found that the result of the log2 was rounded > down and so this was still making max_cycles too big. Therefore to be on > the safe side I add 1 to the result of log2 which will ensure max_cycles > is not too big.
Yep you're right. I was trying to handle that with the -1, but my mistake was using 1<<64 not 1<<63. Thanks for catching that!
> > Signed-off-by: Jon Hunter <jon-hunter@ti.com>
Acked-by: John Stultz <johnstul@us.ibm.com>
This looks great to me!
Thanks for putting up with my nit-picky review and revised suggestions!
Thomas, you ok to pick this up? -john
> --- > include/linux/time.h | 1 + > kernel/time/tick-sched.c | 36 +++++++++++++++++++++++---------- > kernel/time/timekeeping.c | 47 > +++++++++++++++++++++++++++++++++++++++++++++ > 3 files changed, 73 insertions(+), 11 deletions(-) > > diff --git a/include/linux/time.h b/include/linux/time.h > index 242f624..090be07 100644 > --- a/include/linux/time.h > +++ b/include/linux/time.h > @@ -130,6 +130,7 @@ extern void monotonic_to_bootbased(struct timespec *ts); > > extern struct timespec timespec_trunc(struct timespec t, unsigned gran); > extern int timekeeping_valid_for_hres(void); > +extern s64 timekeeping_max_deferment(void); > extern void update_wall_time(void); > extern void update_xtime_cache(u64 nsec); > > diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c > index d3f1ef4..f0155ae 100644 > --- a/kernel/time/tick-sched.c > +++ b/kernel/time/tick-sched.c > @@ -217,6 +217,7 @@ void tick_nohz_stop_sched_tick(int inidle) > ktime_t last_update, expires, now; > struct clock_event_device *dev = __get_cpu_var(tick_cpu_device).evtdev; > int cpu; > + s64 time_delta, max_time_delta; > > local_irq_save(flags); > > @@ -264,6 +265,7 @@ void tick_nohz_stop_sched_tick(int inidle) > seq = read_seqbegin(&xtime_lock); > last_update = last_jiffies_update; > last_jiffies = jiffies; > + max_time_delta = timekeeping_max_deferment(); > } while (read_seqretry(&xtime_lock, seq)); > > /* Get the next timer wheel timer */ > @@ -283,11 +285,22 @@ void tick_nohz_stop_sched_tick(int inidle) > if ((long)delta_jiffies >= 1) { > > /* > - * calculate the expiry time for the next timer wheel > - * timer > - */ > - expires = ktime_add_ns(last_update, tick_period.tv64 * > - delta_jiffies); > + * Calculate the time delta for the next timer event. > + * If the time delta exceeds the maximum time delta > + * permitted by the current clocksource then adjust > + * the time delta accordingly to ensure the > + * clocksource does not wrap. > + */ > + time_delta = tick_period.tv64 * delta_jiffies; > + > + if (time_delta > max_time_delta) > + time_delta = max_time_delta; > + > + /* > + * calculate the expiry time for the next timer wheel > + * timer > + */ > + expires = ktime_add_ns(last_update, time_delta); > > /* > * If this cpu is the one which updates jiffies, then > @@ -300,7 +313,7 @@ void tick_nohz_stop_sched_tick(int inidle) > if (cpu == tick_do_timer_cpu) > tick_do_timer_cpu = TICK_DO_TIMER_NONE; > > - if (delta_jiffies > 1) > + if (time_delta > tick_period.tv64) > cpumask_set_cpu(cpu, nohz_cpu_mask); > > /* Skip reprogram of event if its not changed */ > @@ -332,12 +345,13 @@ void tick_nohz_stop_sched_tick(int inidle) > ts->idle_sleeps++; > > /* > - * delta_jiffies >= NEXT_TIMER_MAX_DELTA signals that > - * there is no timer pending or at least extremly far > - * into the future (12 days for HZ=1000). In this case > - * we simply stop the tick timer: > + * time_delta >= (tick_period.tv64 * NEXT_TIMER_MAX_DELTA) > + * signals that there is no timer pending or at least > + * extremely far into the future (12 days for HZ=1000). > + * In this case we simply stop the tick timer: > */ > - if (unlikely(delta_jiffies >= NEXT_TIMER_MAX_DELTA)) { > + if (unlikely(time_delta >= > + (tick_period.tv64 * NEXT_TIMER_MAX_DELTA))) { > ts->idle_expires.tv64 = KTIME_MAX; > if (ts->nohz_mode == NOHZ_MODE_HIGHRES) > hrtimer_cancel(&ts->sched_timer); > diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c > index 687dff4..608fc6f 100644 > --- a/kernel/time/timekeeping.c > +++ b/kernel/time/timekeeping.c > @@ -271,6 +271,53 @@ int timekeeping_valid_for_hres(void) > } > > /** > + * timekeeping_max_deferment - Returns max time the clocksource can be > deferred > + * > + * IMPORTANT: Must be called with xtime_lock held! > + */ > +s64 timekeeping_max_deferment(void) > +{ > + s64 max_nsecs; > + u64 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/clock->mult which > + * is equivalent to the below. > + * max_cycles < (2^63)/clock->mult > + * max_cycles < 2^(log2((2^63)/clock->mult)) > + * max_cycles < 2^(log2(2^63) - log2(clock->mult)) > + * max_cycles < 2^(63 - log2(clock->mult)) > + * max_cycles < 1 << (63 - log2(clock->mult)) > + * 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(clock->mult) + 1)); > + > + /* > + * The actual maximum number of cycles we can defer the clocksource is > + * determined by the minimum of max_cycles and clock->mask. > + */ > + max_cycles = min(max_cycles, clock->mask); > + max_nsecs = cyc2ns(clock, max_cycles); > + > + /* > + * To ensure that the clocksource does not wrap whilst we are idle, > + * limit the time the clocksource can be deferred by 6.25%. Please > + * note a margin of 6.25% is used because this can be computed with > + * a shift, versus say 5% which would require division. > + */ > + max_nsecs = max_nsecs - (max_nsecs >> 4); > + > + if (max_nsecs < 0) > + max_nsecs = 0; > + > + return max_nsecs; > +} > + > +/** > * read_persistent_clock - Return time in seconds from the persistent > clock. > * > * Weak dummy function for arches that do not yet support it.
| |