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
  22 #include <asm/irq_regs.h>
  23
  24 #include "tick-internal.h"
  25
  26 /*
  27  * Tick devices
  28  */
  29 DEFINE_PER_CPU(struct tick_device, tick_cpu_device);
  30 /*
  31  * Tick next event: keeps track of the tick time
  32  */
  33 ktime_t tick_next_period;
  34 ktime_t tick_period;
  35 int tick_do_timer_cpu __read_mostly = TICK_DO_TIMER_BOOT;
  36 static DEFINE_RAW_SPINLOCK(tick_device_lock);
  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.
 198          */
 199         if (tick_device_uses_broadcast(newdev, cpu))
 200                 return;
 201
 202         if (td->mode == TICKDEV_MODE_PERIODIC)
 203                 tick_setup_periodic(newdev, 0);
 204         else
 205                 tick_setup_oneshot(newdev, handler, next_event);
 206 }
 207
 208 /*
 209  * Check, if the new registered device should be used.
 210  */
 211 static int tick_check_new_device(struct clock_event_device *newdev)
 212 {
 213         struct clock_event_device *curdev;
 214         struct tick_device *td;
 215         int cpu, ret = NOTIFY_OK;
 216         unsigned long flags;
 217
 218         raw_spin_lock_irqsave(&tick_device_lock, flags);
 219
 220         cpu = smp_processor_id();
 221         if (!cpumask_test_cpu(cpu, newdev->cpumask))
 222                 goto out_bc;
 223
 224         td = &per_cpu(tick_cpu_device, cpu);
 225         curdev = td->evtdev;
 226
 227         /* cpu local device ? */
 228         if (!cpumask_equal(newdev->cpumask, cpumask_of(cpu))) {
 229
 230                 /*
 231                  * If the cpu affinity of the device interrupt can not
 232                  * be set, ignore it.
 233                  */
 234                 if (!irq_can_set_affinity(newdev->irq))
 235                         goto out_bc;
 236
 237                 /*
 238                  * If we have a cpu local device already, do not replace it
 239                  * by a non cpu local device
 240                  */
 241                 if (curdev && cpumask_equal(curdev->cpumask, cpumask_of(cpu)))
 242                         goto out_bc;
 243         }
 244
 245         /*
 246          * If we have an active device, then check the rating and the oneshot
 247          * feature.
 248          */
 249         if (curdev) {
 250                 /*
 251                  * Prefer one shot capable devices !
 252                  */
 253                 if ((curdev->features & CLOCK_EVT_FEAT_ONESHOT) &&
 254                     !(newdev->features & CLOCK_EVT_FEAT_ONESHOT))
 255                         goto out_bc;
 256                 /*
 257                  * Check the rating
 258                  */
 259                 if (curdev->rating >= newdev->rating)
 260                         goto out_bc;
 261         }
 262
 263         /*
 264          * Replace the eventually existing device by the new
 265          * device. If the current device is the broadcast device, do
 266          * not give it back to the clockevents layer !
 267          */
 268         if (tick_is_broadcast_device(curdev)) {
 269                 clockevents_shutdown(curdev);
 270                 curdev = NULL;
 271         }
 272         clockevents_exchange_device(curdev, newdev);
 273         tick_setup_device(td, newdev, cpu, cpumask_of(cpu));
 274         if (newdev->features & CLOCK_EVT_FEAT_ONESHOT)
 275                 tick_oneshot_notify();
 276
 277         raw_spin_unlock_irqrestore(&tick_device_lock, flags);
 278         return NOTIFY_STOP;
 279
 280 out_bc:
 281         /*
 282          * Can the new device be used as a broadcast device ?
 283          */
 284         if (tick_check_broadcast_device(newdev))
 285                 ret = NOTIFY_STOP;
 286
 287         raw_spin_unlock_irqrestore(&tick_device_lock, flags);
 288
 289         return ret;
 290 }
 291
 292 /*
 293  * Transfer the do_timer job away from a dying cpu.
 294  *
 295  * Called with interrupts disabled.
 296  */
 297 static void tick_handover_do_timer(int *cpup)
 298 {
 299         if (*cpup == tick_do_timer_cpu) {
 300                 int cpu = cpumask_first(cpu_online_mask);
 301
 302                 tick_do_timer_cpu = (cpu < nr_cpu_ids) ? cpu :
 303                         TICK_DO_TIMER_NONE;
 304         }
 305 }
 306
 307 /*
 308  * Shutdown an event device on a given cpu:
 309  *
 310  * This is called on a life CPU, when a CPU is dead. So we cannot
 311  * access the hardware device itself.
 312  * We just set the mode and remove it from the lists.
 313  */
 314 static void tick_shutdown(unsigned int *cpup)
 315 {
 316         struct tick_device *td = &per_cpu(tick_cpu_device, *cpup);
 317         struct clock_event_device *dev = td->evtdev;
 318         unsigned long flags;
 319
 320         raw_spin_lock_irqsave(&tick_device_lock, flags);
 321         td->mode = TICKDEV_MODE_PERIODIC;
 322         if (dev) {
 323                 /*
 324                  * Prevent that the clock events layer tries to call
 325                  * the set mode function!
 326                  */
 327                 dev->mode = CLOCK_EVT_MODE_UNUSED;
 328                 clockevents_exchange_device(dev, NULL);
 329                 td->evtdev = NULL;
 330         }
 331         raw_spin_unlock_irqrestore(&tick_device_lock, flags);
 332 }
 333
 334 static void tick_suspend(void)
 335 {
 336         struct tick_device *td = &__get_cpu_var(tick_cpu_device);
 337         unsigned long flags;
 338
 339         raw_spin_lock_irqsave(&tick_device_lock, flags);
 340         clockevents_shutdown(td->evtdev);
 341         raw_spin_unlock_irqrestore(&tick_device_lock, flags);
 342 }
 343
 344 static void tick_resume(void)
 345 {
 346         struct tick_device *td = &__get_cpu_var(tick_cpu_device);
 347         unsigned long flags;
 348         int broadcast = tick_resume_broadcast();
 349
 350         raw_spin_lock_irqsave(&tick_device_lock, flags);
 351         clockevents_set_mode(td->evtdev, CLOCK_EVT_MODE_RESUME);
 352
 353         if (!broadcast) {
 354                 if (td->mode == TICKDEV_MODE_PERIODIC)
 355                         tick_setup_periodic(td->evtdev, 0);
 356                 else
 357                         tick_resume_oneshot();
 358         }
 359         raw_spin_unlock_irqrestore(&tick_device_lock, flags);
 360 }
 361
 362 /*
 363  * Notification about clock event devices
 364  */
 365 static int tick_notify(struct notifier_block *nb, unsigned long reason,
 366                                void *dev)
 367 {
 368         switch (reason) {
 369
 370         case CLOCK_EVT_NOTIFY_ADD:
 371                 return tick_check_new_device(dev);
 372
 373         case CLOCK_EVT_NOTIFY_BROADCAST_ON:
 374         case CLOCK_EVT_NOTIFY_BROADCAST_OFF:
 375         case CLOCK_EVT_NOTIFY_BROADCAST_FORCE:
 376                 tick_broadcast_on_off(reason, dev);
 377                 break;
 378
 379         case CLOCK_EVT_NOTIFY_BROADCAST_ENTER:
 380         case CLOCK_EVT_NOTIFY_BROADCAST_EXIT:
 381                 tick_broadcast_oneshot_control(reason);
 382                 break;
 383
 384         case CLOCK_EVT_NOTIFY_CPU_DYING:
 385                 tick_handover_do_timer(dev);
 386                 break;
 387
 388         case CLOCK_EVT_NOTIFY_CPU_DEAD:
 389                 tick_shutdown_broadcast_oneshot(dev);
 390                 tick_shutdown_broadcast(dev);
 391                 tick_shutdown(dev);
 392                 break;
 393
 394         case CLOCK_EVT_NOTIFY_SUSPEND:
 395                 tick_suspend();
 396                 tick_suspend_broadcast();
 397                 break;
 398
 399         case CLOCK_EVT_NOTIFY_RESUME:
 400                 tick_resume();
 401                 break;
 402
 403         default:
 404                 break;
 405         }
 406
 407         return NOTIFY_OK;
 408 }
 409
 410 static struct notifier_block tick_notifier = {
 411         .notifier_call = tick_notify,
 412 };
 413
 414 /**
 415  * tick_init - initialize the tick control
 416  *
 417  * Register the notifier with the clockevents framework
 418  */
 419 void __init tick_init(void)
 420 {
 421         clockevents_register_notifier(&tick_notifier);
 422 }