kernel/time/tick-common.c

   1 /*
   2  * linux/kernel/time/tick-common.c
   3  *
   4  * This file contains the base functions to manage periodic tick
   5  * related events.
   6  *
   7  * Copyright(C) 2005-2006, Thomas Gleixner <tglx@linutronix.de>
   8  * Copyright(C) 2005-2007, Red Hat, Inc., Ingo Molnar
   9  * Copyright(C) 2006-2007, Timesys Corp., Thomas Gleixner
  10  *
  11  * This code is licenced under the GPL version 2. For details see
  12  * kernel-base/COPYING.
  13  */
  14 #include <linux/cpu.h>
  15 #include <linux/err.h>
  16 #include <linux/hrtimer.h>
  17 #include <linux/interrupt.h>
  18 #include <linux/percpu.h>
  19 #include <linux/profile.h>
  20 #include <linux/sched.h>
  21 #include <linux/module.h>
  22
  23 #include <asm/irq_regs.h>
  24
  25 #include "tick-internal.h"
  26
  27 /*
  28  * Tick devices
  29  */
  30 DEFINE_PER_CPU(struct tick_device, tick_cpu_device);
  31 /*
  32  * Tick next event: keeps track of the tick time
  33  */
  34 ktime_t tick_next_period;
  35 ktime_t tick_period;
  36 int tick_do_timer_cpu __read_mostly = TICK_DO_TIMER_BOOT;
  37
  38 /*
  39  * Debugging: see timer_list.c
  40  */
  41 struct tick_device *tick_get_device(int cpu)
  42 {
  43         return &per_cpu(tick_cpu_device, cpu);
  44 }
  45
  46 /**
  47  * tick_is_oneshot_available - check for a oneshot capable event device
  48  */
  49 int tick_is_oneshot_available(void)
  50 {
  51         struct clock_event_device *dev = __this_cpu_read(tick_cpu_device.evtdev);
  52
  53         if (!dev || !(dev->features & CLOCK_EVT_FEAT_ONESHOT))
  54                 return 0;
  55         if (!(dev->features & CLOCK_EVT_FEAT_C3STOP))
  56                 return 1;
  57         return tick_broadcast_oneshot_available();
  58 }
  59
  60 /*
  61  * Periodic tick
  62  */
  63 static void tick_periodic(int cpu)
  64 {
  65         if (tick_do_timer_cpu == cpu) {
  66                 write_seqlock(&jiffies_lock);
  67
  68                 /* Keep track of the next tick event */
  69                 tick_next_period = ktime_add(tick_next_period, tick_period);
  70
  71                 do_timer(1);
  72                 write_sequnlock(&jiffies_lock);
  73         }
  74
  75         update_process_times(user_mode(get_irq_regs()));
  76         profile_tick(CPU_PROFILING);
  77 }
  78
  79 /*
  80  * Event handler for periodic ticks
  81  */
  82 void tick_handle_periodic(struct clock_event_device *dev)
  83 {
  84         int cpu = smp_processor_id();
  85         ktime_t next;
  86
  87         tick_periodic(cpu);
  88
  89         if (dev->mode != CLOCK_EVT_MODE_ONESHOT)
  90                 return;
  91         /*
  92          * Setup the next period for devices, which do not have
  93          * periodic mode:
  94          */
  95         next = ktime_add(dev->next_event, tick_period);
  96         for (;;) {
  97                 if (!clockevents_program_event(dev, next, false))
  98                         return;
  99                 /*
 100                  * Have to be careful here. If we're in oneshot mode,
 101                  * before we call tick_periodic() in a loop, we need
 102                  * to be sure we're using a real hardware clocksource.
 103                  * Otherwise we could get trapped in an infinite
 104                  * loop, as the tick_periodic() increments jiffies,
 105                  * when then will increment time, posibly causing
 106                  * the loop to trigger again and again.
 107                  */
 108                 if (timekeeping_valid_for_hres())
 109                         tick_periodic(cpu);
 110                 next = ktime_add(next, tick_period);
 111         }
 112 }
 113
 114 /*
 115  * Setup the device for a periodic tick
 116  */
 117 void tick_setup_periodic(struct clock_event_device *dev, int broadcast)
 118 {
 119         tick_set_periodic_handler(dev, broadcast);
 120
 121         /* Broadcast setup ? */
 122         if (!tick_device_is_functional(dev))
 123                 return;
 124
 125         if ((dev->features & CLOCK_EVT_FEAT_PERIODIC) &&
 126             !tick_broadcast_oneshot_active()) {
 127                 clockevents_set_mode(dev, CLOCK_EVT_MODE_PERIODIC);
 128         } else {
 129                 unsigned long seq;
 130                 ktime_t next;
 131
 132                 do {
 133                         seq = read_seqbegin(&jiffies_lock);
 134                         next = tick_next_period;
 135                 } while (read_seqretry(&jiffies_lock, seq));
 136
 137                 clockevents_set_mode(dev, CLOCK_EVT_MODE_ONESHOT);
 138
 139                 for (;;) {
 140                         if (!clockevents_program_event(dev, next, false))
 141                                 return;
 142                         next = ktime_add(next, tick_period);
 143                 }
 144         }
 145 }
 146
 147 /*
 148  * Setup the tick device
 149  */
 150 static void tick_setup_device(struct tick_device *td,
 151                               struct clock_event_device *newdev, int cpu,
 152                               const struct cpumask *cpumask)
 153 {
 154         ktime_t next_event;
 155         void (*handler)(struct clock_event_device *) = NULL;
 156
 157         /*
 158          * First device setup ?
 159          */
 160         if (!td->evtdev) {
 161                 /*
 162                  * If no cpu took the do_timer update, assign it to
 163                  * this cpu:
 164                  */
 165                 if (tick_do_timer_cpu == TICK_DO_TIMER_BOOT) {
 166                         if (!tick_nohz_full_cpu(cpu))
 167                                 tick_do_timer_cpu = cpu;
 168                         else
 169                                 tick_do_timer_cpu = TICK_DO_TIMER_NONE;
 170                         tick_next_period = ktime_get();
 171                         tick_period = ktime_set(0, NSEC_PER_SEC / HZ);
 172                 }
 173
 174                 /*
 175                  * Startup in periodic mode first.
 176                  */
 177                 td->mode = TICKDEV_MODE_PERIODIC;
 178         } else {
 179                 handler = td->evtdev->event_handler;
 180                 next_event = td->evtdev->next_event;
 181                 td->evtdev->event_handler = clockevents_handle_noop;
 182         }
 183
 184         td->evtdev = newdev;
 185
 186         /*
 187          * When the device is not per cpu, pin the interrupt to the
 188          * current cpu:
 189          */
 190         if (!cpumask_equal(newdev->cpumask, cpumask))
 191                 irq_set_affinity(newdev->irq, cpumask);
 192
 193         /*
 194          * When global broadcasting is active, check if the current
 195          * device is registered as a placeholder for broadcast mode.
 196          * This allows us to handle this x86 misfeature in a generic
 197          * way. This function also returns !=0 when we keep the
 198          * current active broadcast state for this CPU.
 199          */
 200         if (tick_device_uses_broadcast(newdev, cpu))
 201                 return;
 202
 203         if (td->mode == TICKDEV_MODE_PERIODIC)
 204                 tick_setup_periodic(newdev, 0);
 205         else
 206                 tick_setup_oneshot(newdev, handler, next_event);
 207 }
 208
 209 void tick_install_replacement(struct clock_event_device *newdev)
 210 {
 211         struct tick_device *td = &__get_cpu_var(tick_cpu_device);
 212         int cpu = smp_processor_id();
 213
 214         clockevents_exchange_device(td->evtdev, newdev);
 215         tick_setup_device(td, newdev, cpu, cpumask_of(cpu));
 216         if (newdev->features & CLOCK_EVT_FEAT_ONESHOT)
 217                 tick_oneshot_notify();
 218 }
 219
 220 static bool tick_check_percpu(struct clock_event_device *curdev,
 221                               struct clock_event_device *newdev, int cpu)
 222 {
 223         if (!cpumask_test_cpu(cpu, newdev->cpumask))
 224                 return false;
 225         if (cpumask_equal(newdev->cpumask, cpumask_of(cpu)))
 226                 return true;
 227         /* Check if irq affinity can be set */
 228         if (newdev->irq >= 0 && !irq_can_set_affinity(newdev->irq))
 229                 return false;
 230         /* Prefer an existing cpu local device */
 231         if (curdev && cpumask_equal(curdev->cpumask, cpumask_of(cpu)))
 232                 return false;
 233         return true;
 234 }
 235
 236 static bool tick_check_preferred(struct clock_event_device *curdev,
 237                                  struct clock_event_device *newdev)
 238 {
 239         /* Prefer oneshot capable device */
 240         if (!(newdev->features & CLOCK_EVT_FEAT_ONESHOT)) {
 241                 if (curdev && (curdev->features & CLOCK_EVT_FEAT_ONESHOT))
 242                         return false;
 243                 if (tick_oneshot_mode_active())
 244                         return false;
 245         }
 246
 247         /*
 248          * Use the higher rated one, but prefer a CPU local device with a lower
 249          * rating than a non-CPU local device
 250          */
 251         return !curdev ||
 252                 newdev->rating > curdev->rating ||
 253                !cpumask_equal(curdev->cpumask, newdev->cpumask);
 254 }
 255
 256 /*
 257  * Check whether the new device is a better fit than curdev. curdev
 258  * can be NULL !
 259  */
 260 bool tick_check_replacement(struct clock_event_device *curdev,
 261                             struct clock_event_device *newdev)
 262 {
 263         if (!tick_check_percpu(curdev, newdev, smp_processor_id()))
 264                 return false;
 265
 266         return tick_check_preferred(curdev, newdev);
 267 }
 268
 269 /*
 270  * Check, if the new registered device should be used. Called with
 271  * clockevents_lock held and interrupts disabled.
 272  */
 273 void tick_check_new_device(struct clock_event_device *newdev)
 274 {
 275         struct clock_event_device *curdev;
 276         struct tick_device *td;
 277         int cpu;
 278
 279         cpu = smp_processor_id();
 280         if (!cpumask_test_cpu(cpu, newdev->cpumask))
 281                 goto out_bc;
 282
 283         td = &per_cpu(tick_cpu_device, cpu);
 284         curdev = td->evtdev;
 285
 286         /* cpu local device ? */
 287         if (!tick_check_percpu(curdev, newdev, cpu))
 288                 goto out_bc;
 289
 290         /* Preference decision */
 291         if (!tick_check_preferred(curdev, newdev))
 292                 goto out_bc;
 293
 294         if (!try_module_get(newdev->owner))
 295                 return;
 296
 297         /*
 298          * Replace the eventually existing device by the new
 299          * device. If the current device is the broadcast device, do
 300          * not give it back to the clockevents layer !
 301          */
 302         if (tick_is_broadcast_device(curdev)) {
 303                 clockevents_shutdown(curdev);
 304                 curdev = NULL;
 305         }
 306         clockevents_exchange_device(curdev, newdev);
 307         tick_setup_device(td, newdev, cpu, cpumask_of(cpu));
 308         if (newdev->features & CLOCK_EVT_FEAT_ONESHOT)
 309                 tick_oneshot_notify();
 310         return;
 311
 312 out_bc:
 313         /*
 314          * Can the new device be used as a broadcast device ?
 315          */
 316         tick_install_broadcast_device(newdev);
 317 }
 318
 319 /*
 320  * Transfer the do_timer job away from a dying cpu.
 321  *
 322  * Called with interrupts disabled.
 323  */
 324 void tick_handover_do_timer(int *cpup)
 325 {
 326         if (*cpup == tick_do_timer_cpu) {
 327                 int cpu = cpumask_first(cpu_online_mask);
 328
 329                 tick_do_timer_cpu = (cpu < nr_cpu_ids) ? cpu :
 330                         TICK_DO_TIMER_NONE;
 331         }
 332 }
 333
 334 /*
 335  * Shutdown an event device on a given cpu:
 336  *
 337  * This is called on a life CPU, when a CPU is dead. So we cannot
 338  * access the hardware device itself.
 339  * We just set the mode and remove it from the lists.
 340  */
 341 void tick_shutdown(unsigned int *cpup)
 342 {
 343         struct tick_device *td = &per_cpu(tick_cpu_device, *cpup);
 344         struct clock_event_device *dev = td->evtdev;
 345
 346         td->mode = TICKDEV_MODE_PERIODIC;
 347         if (dev) {
 348                 /*
 349                  * Prevent that the clock events layer tries to call
 350                  * the set mode function!
 351                  */
 352                 dev->mode = CLOCK_EVT_MODE_UNUSED;
 353                 clockevents_exchange_device(dev, NULL);
 354                 dev->event_handler = clockevents_handle_noop;
 355                 td->evtdev = NULL;
 356         }
 357 }
 358
 359 void tick_suspend(void)
 360 {
 361         struct tick_device *td = &__get_cpu_var(tick_cpu_device);
 362
 363         clockevents_shutdown(td->evtdev);
 364 }
 365
 366 void tick_resume(void)
 367 {
 368         struct tick_device *td = &__get_cpu_var(tick_cpu_device);
 369         int broadcast = tick_resume_broadcast();
 370
 371         clockevents_set_mode(td->evtdev, CLOCK_EVT_MODE_RESUME);
 372
 373         if (!broadcast) {
 374                 if (td->mode == TICKDEV_MODE_PERIODIC)
 375                         tick_setup_periodic(td->evtdev, 0);
 376                 else
 377                         tick_resume_oneshot();
 378         }
 379 }
 380
 381 /**
 382  * tick_init - initialize the tick control
 383  */
 384 void __init tick_init(void)
 385 {
 386         tick_broadcast_init();
 387 }