sys/compat/linuxkpi/common/src/linux_schedule.c

   1 /*-
   2  * Copyright (c) 2017 Mark Johnston <markj@FreeBSD.org>
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conds
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice unmodified, this list of conds, and the following
  10  *    disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conds and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  *
  15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  16  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  17  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  18  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  20  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  21  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  22  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  23  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  24  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  25  */
  26
  27 #include <sys/cdefs.h>
  28 __FBSDID("$FreeBSD$");
  29
  30 #include <sys/param.h>
  31 #include <sys/systm.h>
  32 #include <sys/proc.h>
  33 #include <sys/signalvar.h>
  34 #include <sys/sleepqueue.h>
  35
  36 #include <linux/delay.h>
  37 #include <linux/errno.h>
  38 #include <linux/kernel.h>
  39 #include <linux/list.h>
  40 #include <linux/sched.h>
  41 #include <linux/spinlock.h>
  42 #include <linux/wait.h>
  43
  44 static int
  45 linux_add_to_sleepqueue(void *wchan, struct task_struct *task,
  46     const char *wmesg, int timeout, int state)
  47 {
  48         int flags, ret;
  49
  50         MPASS((state & ~(TASK_PARKED | TASK_NORMAL)) == 0);
  51
  52         flags = SLEEPQ_SLEEP | ((state & TASK_INTERRUPTIBLE) != 0 ?
  53             SLEEPQ_INTERRUPTIBLE : 0);
  54
  55         sleepq_add(wchan, NULL, wmesg, flags, 0);
  56         if (timeout != 0)
  57                 sleepq_set_timeout(wchan, timeout);
  58
  59         DROP_GIANT();
  60         if ((state & TASK_INTERRUPTIBLE) != 0) {
  61                 if (timeout == 0)
  62                         ret = -sleepq_wait_sig(wchan, 0);
  63                 else
  64                         ret = -sleepq_timedwait_sig(wchan, 0);
  65         } else {
  66                 if (timeout == 0) {
  67                         sleepq_wait(wchan, 0);
  68                         ret = 0;
  69                 } else
  70                         ret = -sleepq_timedwait(wchan, 0);
  71         }
  72         PICKUP_GIANT();
  73
  74         /* filter return value */
  75         if (ret != 0 && ret != -EWOULDBLOCK) {
  76                 linux_schedule_save_interrupt_value(task, ret);
  77                 ret = -ERESTARTSYS;
  78         }
  79         return (ret);
  80 }
  81
  82 unsigned int
  83 linux_msleep_interruptible(unsigned int ms)
  84 {
  85         int ret;
  86
  87         /* guard against invalid values */
  88         if (ms == 0)
  89                 ms = 1;
  90         ret = -pause_sbt("lnxsleep", mstosbt(ms), 0, C_HARDCLOCK | C_CATCH);
  91
  92         switch (ret) {
  93         case -EWOULDBLOCK:
  94                 return (0);
  95         default:
  96                 linux_schedule_save_interrupt_value(current, ret);
  97                 return (ms);
  98         }
  99 }
 100
 101 static int
 102 wake_up_task(struct task_struct *task, unsigned int state)
 103 {
 104         int ret, wakeup_swapper;
 105
 106         ret = wakeup_swapper = 0;
 107         sleepq_lock(task);
 108         if ((atomic_read(&task->state) & state) != 0) {
 109                 set_task_state(task, TASK_WAKING);
 110                 wakeup_swapper = sleepq_signal(task, SLEEPQ_SLEEP, 0, 0);
 111                 ret = 1;
 112         }
 113         sleepq_release(task);
 114         if (wakeup_swapper)
 115                 kick_proc0();
 116         return (ret);
 117 }
 118
 119 bool
 120 linux_signal_pending(struct task_struct *task)
 121 {
 122         struct thread *td;
 123         sigset_t pending;
 124
 125         td = task->task_thread;
 126         PROC_LOCK(td->td_proc);
 127         pending = td->td_siglist;
 128         SIGSETOR(pending, td->td_proc->p_siglist);
 129         SIGSETNAND(pending, td->td_sigmask);
 130         PROC_UNLOCK(td->td_proc);
 131         return (!SIGISEMPTY(pending));
 132 }
 133
 134 bool
 135 linux_fatal_signal_pending(struct task_struct *task)
 136 {
 137         struct thread *td;
 138         bool ret;
 139
 140         td = task->task_thread;
 141         PROC_LOCK(td->td_proc);
 142         ret = SIGISMEMBER(td->td_siglist, SIGKILL) ||
 143             SIGISMEMBER(td->td_proc->p_siglist, SIGKILL);
 144         PROC_UNLOCK(td->td_proc);
 145         return (ret);
 146 }
 147
 148 bool
 149 linux_signal_pending_state(long state, struct task_struct *task)
 150 {
 151
 152         MPASS((state & ~TASK_NORMAL) == 0);
 153
 154         if ((state & TASK_INTERRUPTIBLE) == 0)
 155                 return (false);
 156         return (linux_signal_pending(task));
 157 }
 158
 159 void
 160 linux_send_sig(int signo, struct task_struct *task)
 161 {
 162         struct thread *td;
 163
 164         td = task->task_thread;
 165         PROC_LOCK(td->td_proc);
 166         tdsignal(td, signo);
 167         PROC_UNLOCK(td->td_proc);
 168 }
 169
 170 int
 171 autoremove_wake_function(wait_queue_t *wq, unsigned int state, int flags,
 172     void *key __unused)
 173 {
 174         struct task_struct *task;
 175         int ret;
 176
 177         task = wq->private;
 178         if ((ret = wake_up_task(task, state)) != 0)
 179                 list_del_init(&wq->task_list);
 180         return (ret);
 181 }
 182
 183 int
 184 default_wake_function(wait_queue_t *wq, unsigned int state, int flags,
 185     void *key __unused)
 186 {
 187         return (wake_up_task(wq->private, state));
 188 }
 189
 190 void
 191 linux_init_wait_entry(wait_queue_t *wq, int flags)
 192 {
 193
 194         memset(wq, 0, sizeof(*wq));
 195         wq->flags = flags;
 196         wq->private = current;
 197         wq->func = autoremove_wake_function;
 198         INIT_LIST_HEAD(&wq->task_list);
 199 }
 200
 201 void
 202 linux_wake_up(wait_queue_head_t *wqh, unsigned int state, int nr, bool locked)
 203 {
 204         wait_queue_t *pos, *next;
 205
 206         if (!locked)
 207                 spin_lock(&wqh->lock);
 208         list_for_each_entry_safe(pos, next, &wqh->task_list, task_list) {
 209                 if (pos->func == NULL) {
 210                         if (wake_up_task(pos->private, state) != 0 && --nr == 0)
 211                                 break;
 212                 } else {
 213                         if (pos->func(pos, state, 0, NULL) != 0 && --nr == 0)
 214                                 break;
 215                 }
 216         }
 217         if (!locked)
 218                 spin_unlock(&wqh->lock);
 219 }
 220
 221 void
 222 linux_prepare_to_wait(wait_queue_head_t *wqh, wait_queue_t *wq, int state)
 223 {
 224
 225         spin_lock(&wqh->lock);
 226         if (list_empty(&wq->task_list))
 227                 __add_wait_queue(wqh, wq);
 228         set_task_state(current, state);
 229         spin_unlock(&wqh->lock);
 230 }
 231
 232 void
 233 linux_finish_wait(wait_queue_head_t *wqh, wait_queue_t *wq)
 234 {
 235
 236         spin_lock(&wqh->lock);
 237         set_task_state(current, TASK_RUNNING);
 238         if (!list_empty(&wq->task_list)) {
 239                 __remove_wait_queue(wqh, wq);
 240                 INIT_LIST_HEAD(&wq->task_list);
 241         }
 242         spin_unlock(&wqh->lock);
 243 }
 244
 245 bool
 246 linux_waitqueue_active(wait_queue_head_t *wqh)
 247 {
 248         bool ret;
 249
 250         spin_lock(&wqh->lock);
 251         ret = !list_empty(&wqh->task_list);
 252         spin_unlock(&wqh->lock);
 253         return (ret);
 254 }
 255
 256 int
 257 linux_wait_event_common(wait_queue_head_t *wqh, wait_queue_t *wq, int timeout,
 258     unsigned int state, spinlock_t *lock)
 259 {
 260         struct task_struct *task;
 261         int ret;
 262
 263         if (lock != NULL)
 264                 spin_unlock_irq(lock);
 265
 266         /* range check timeout */
 267         if (timeout < 1)
 268                 timeout = 1;
 269         else if (timeout == MAX_SCHEDULE_TIMEOUT)
 270                 timeout = 0;
 271
 272         task = current;
 273
 274         /*
 275          * Our wait queue entry is on the stack - make sure it doesn't
 276          * get swapped out while we sleep.
 277          */
 278         PHOLD(task->task_thread->td_proc);
 279         sleepq_lock(task);
 280         if (atomic_read(&task->state) != TASK_WAKING) {
 281                 ret = linux_add_to_sleepqueue(task, task, "wevent", timeout,
 282                     state);
 283         } else {
 284                 sleepq_release(task);
 285                 ret = 0;
 286         }
 287         PRELE(task->task_thread->td_proc);
 288
 289         if (lock != NULL)
 290                 spin_lock_irq(lock);
 291         return (ret);
 292 }
 293
 294 int
 295 linux_schedule_timeout(int timeout)
 296 {
 297         struct task_struct *task;
 298         int ret;
 299         int state;
 300         int remainder;
 301
 302         task = current;
 303
 304         /* range check timeout */
 305         if (timeout < 1)
 306                 timeout = 1;
 307         else if (timeout == MAX_SCHEDULE_TIMEOUT)
 308                 timeout = 0;
 309
 310         remainder = ticks + timeout;
 311
 312         sleepq_lock(task);
 313         state = atomic_read(&task->state);
 314         if (state != TASK_WAKING) {
 315                 ret = linux_add_to_sleepqueue(task, task, "sched", timeout,
 316                     state);
 317         } else {
 318                 sleepq_release(task);
 319                 ret = 0;
 320         }
 321         set_task_state(task, TASK_RUNNING);
 322
 323         if (timeout == 0)
 324                 return (MAX_SCHEDULE_TIMEOUT);
 325
 326         /* range check return value */
 327         remainder -= ticks;
 328
 329         /* range check return value */
 330         if (ret == -ERESTARTSYS && remainder < 1)
 331                 remainder = 1;
 332         else if (remainder < 0)
 333                 remainder = 0;
 334         else if (remainder > timeout)
 335                 remainder = timeout;
 336         return (remainder);
 337 }
 338
 339 static void
 340 wake_up_sleepers(void *wchan)
 341 {
 342         int wakeup_swapper;
 343
 344         sleepq_lock(wchan);
 345         wakeup_swapper = sleepq_signal(wchan, SLEEPQ_SLEEP, 0, 0);
 346         sleepq_release(wchan);
 347         if (wakeup_swapper)
 348                 kick_proc0();
 349 }
 350
 351 #define bit_to_wchan(word, bit) ((void *)(((uintptr_t)(word) << 6) | (bit)))
 352
 353 void
 354 linux_wake_up_bit(void *word, int bit)
 355 {
 356
 357         wake_up_sleepers(bit_to_wchan(word, bit));
 358 }
 359
 360 int
 361 linux_wait_on_bit_timeout(unsigned long *word, int bit, unsigned int state,
 362     int timeout)
 363 {
 364         struct task_struct *task;
 365         void *wchan;
 366         int ret;
 367
 368         /* range check timeout */
 369         if (timeout < 1)
 370                 timeout = 1;
 371         else if (timeout == MAX_SCHEDULE_TIMEOUT)
 372                 timeout = 0;
 373
 374         task = current;
 375         wchan = bit_to_wchan(word, bit);
 376         for (;;) {
 377                 sleepq_lock(wchan);
 378                 if ((*word & (1 << bit)) == 0) {
 379                         sleepq_release(wchan);
 380                         ret = 0;
 381                         break;
 382                 }
 383                 set_task_state(task, state);
 384                 ret = linux_add_to_sleepqueue(wchan, task, "wbit", timeout,
 385                     state);
 386                 if (ret != 0)
 387                         break;
 388         }
 389         set_task_state(task, TASK_RUNNING);
 390
 391         return (ret);
 392 }
 393
 394 void
 395 linux_wake_up_atomic_t(atomic_t *a)
 396 {
 397
 398         wake_up_sleepers(a);
 399 }
 400
 401 int
 402 linux_wait_on_atomic_t(atomic_t *a, unsigned int state)
 403 {
 404         struct task_struct *task;
 405         void *wchan;
 406         int ret;
 407
 408         task = current;
 409         wchan = a;
 410         for (;;) {
 411                 sleepq_lock(wchan);
 412                 if (atomic_read(a) == 0) {
 413                         sleepq_release(wchan);
 414                         ret = 0;
 415                         break;
 416                 }
 417                 set_task_state(task, state);
 418                 ret = linux_add_to_sleepqueue(wchan, task, "watomic", 0, state);
 419                 if (ret != 0)
 420                         break;
 421         }
 422         set_task_state(task, TASK_RUNNING);
 423
 424         return (ret);
 425 }
 426
 427 bool
 428 linux_wake_up_state(struct task_struct *task, unsigned int state)
 429 {
 430
 431         return (wake_up_task(task, state) != 0);
 432 }