| 1 | /* |
| 2 | * Copyright (c) 2000-2020 Apple Inc. All rights reserved. |
| 3 | * |
| 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
| 5 | * |
| 6 | * This file contains Original Code and/or Modifications of Original Code |
| 7 | * as defined in and that are subject to the Apple Public Source License |
| 8 | * Version 2.0 (the 'License'). You may not use this file except in |
| 9 | * compliance with the License. The rights granted to you under the License |
| 10 | * may not be used to create, or enable the creation or redistribution of, |
| 11 | * unlawful or unlicensed copies of an Apple operating system, or to |
| 12 | * circumvent, violate, or enable the circumvention or violation of, any |
| 13 | * terms of an Apple operating system software license agreement. |
| 14 | * |
| 15 | * Please obtain a copy of the License at |
| 16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
| 17 | * |
| 18 | * The Original Code and all software distributed under the License are |
| 19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
| 20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
| 21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
| 22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
| 23 | * Please see the License for the specific language governing rights and |
| 24 | * limitations under the License. |
| 25 | * |
| 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
| 27 | */ |
| 28 | /* |
| 29 | * Mach Operating System |
| 30 | * Copyright (c) 1987 Carnegie-Mellon University |
| 31 | * All rights reserved. The CMU software License Agreement specifies |
| 32 | * the terms and conditions for use and redistribution. |
| 33 | */ |
| 34 | |
| 35 | #include <sys/param.h> |
| 36 | #include <sys/systm.h> |
| 37 | #include <sys/proc_internal.h> |
| 38 | #include <sys/user.h> |
| 39 | #include <sys/file_internal.h> |
| 40 | #include <sys/vnode.h> |
| 41 | #include <sys/kernel.h> |
| 42 | |
| 43 | #include <kern/queue.h> |
| 44 | #include <sys/lock.h> |
| 45 | #include <kern/thread.h> |
| 46 | #include <kern/sched_prim.h> |
| 47 | #include <kern/ast.h> |
| 48 | |
| 49 | #include <kern/cpu_number.h> |
| 50 | #include <vm/vm_kern.h> |
| 51 | |
| 52 | #include <kern/task.h> |
| 53 | #include <mach/time_value.h> |
| 54 | #include <kern/locks.h> |
| 55 | #include <kern/policy_internal.h> |
| 56 | |
| 57 | #include <sys/systm.h> /* for unix_syscall_return() */ |
| 58 | #include <libkern/OSAtomic.h> |
| 59 | |
| 60 | extern void compute_averunnable(void *); /* XXX */ |
| 61 | |
| 62 | __attribute__((noreturn)) |
| 63 | static void |
| 64 | _sleep_continue( __unused void *parameter, wait_result_t wresult) |
| 65 | { |
| 66 | struct proc *p = current_proc(); |
| 67 | thread_t self = current_thread(); |
| 68 | struct uthread * ut; |
| 69 | int sig, catch; |
| 70 | int error = 0; |
| 71 | int dropmutex, spinmutex; |
| 72 | |
| 73 | ut = get_bsdthread_info(self); |
| 74 | catch = ut->uu_pri & PCATCH; |
| 75 | dropmutex = ut->uu_pri & PDROP; |
| 76 | spinmutex = ut->uu_pri & PSPIN; |
| 77 | |
| 78 | switch (wresult) { |
| 79 | case THREAD_TIMED_OUT: |
| 80 | error = EWOULDBLOCK; |
| 81 | break; |
| 82 | case THREAD_AWAKENED: |
| 83 | /* |
| 84 | * Posix implies any signal should be delivered |
| 85 | * first, regardless of whether awakened due |
| 86 | * to receiving event. |
| 87 | */ |
| 88 | if (!catch) { |
| 89 | break; |
| 90 | } |
| 91 | OS_FALLTHROUGH; |
| 92 | case THREAD_INTERRUPTED: |
| 93 | if (catch) { |
| 94 | if (thread_should_abort(self)) { |
| 95 | error = EINTR; |
| 96 | } else if (SHOULDissignal(p, ut)) { |
| 97 | if ((sig = CURSIG(p)) != 0) { |
| 98 | if (p->p_sigacts.ps_sigintr & sigmask(sig)) { |
| 99 | error = EINTR; |
| 100 | } else { |
| 101 | error = ERESTART; |
| 102 | } |
| 103 | } |
| 104 | if (thread_should_abort(self)) { |
| 105 | error = EINTR; |
| 106 | } |
| 107 | } else if ((ut->uu_flag & (UT_CANCELDISABLE | UT_CANCEL | UT_CANCELED)) == UT_CANCEL) { |
| 108 | /* due to thread cancel */ |
| 109 | error = EINTR; |
| 110 | } |
| 111 | } else { |
| 112 | error = EINTR; |
| 113 | } |
| 114 | break; |
| 115 | } |
| 116 | |
| 117 | if (error == EINTR || error == ERESTART) { |
| 118 | act_set_astbsd(self); |
| 119 | } |
| 120 | |
| 121 | if (ut->uu_mtx && !dropmutex) { |
| 122 | if (spinmutex) { |
| 123 | lck_mtx_lock_spin(lck: ut->uu_mtx); |
| 124 | } else { |
| 125 | lck_mtx_lock(lck: ut->uu_mtx); |
| 126 | } |
| 127 | } |
| 128 | ut->uu_wchan = NULL; |
| 129 | ut->uu_wmesg = NULL; |
| 130 | |
| 131 | unix_syscall_return((*ut->uu_continuation)(error)); |
| 132 | } |
| 133 | |
| 134 | /* |
| 135 | * Give up the processor till a wakeup occurs |
| 136 | * on chan, at which time the process |
| 137 | * enters the scheduling queue at priority pri. |
| 138 | * The most important effect of pri is that when |
| 139 | * pri<=PZERO a signal cannot disturb the sleep; |
| 140 | * if pri>PZERO signals will be processed. |
| 141 | * If pri&PCATCH is set, signals will cause sleep |
| 142 | * to return 1, rather than longjmp. |
| 143 | * Callers of this routine must be prepared for |
| 144 | * premature return, and check that the reason for |
| 145 | * sleeping has gone away. |
| 146 | * |
| 147 | * if msleep was the entry point, than we have a mutex to deal with |
| 148 | * |
| 149 | * The mutex is unlocked before the caller is blocked, and |
| 150 | * relocked before msleep returns unless the priority includes the PDROP |
| 151 | * flag... if PDROP is specified, _sleep returns with the mutex unlocked |
| 152 | * regardless of whether it actually blocked or not. |
| 153 | */ |
| 154 | |
| 155 | static int |
| 156 | _sleep( |
| 157 | caddr_t chan, |
| 158 | int pri, |
| 159 | const char *wmsg, |
| 160 | u_int64_t abstime, |
| 161 | int (*continuation)(int), |
| 162 | lck_mtx_t *mtx) |
| 163 | { |
| 164 | struct proc *p; |
| 165 | thread_t self = current_thread(); |
| 166 | struct uthread * ut; |
| 167 | int sig, catch; |
| 168 | int dropmutex = pri & PDROP; |
| 169 | int spinmutex = pri & PSPIN; |
| 170 | int wait_result; |
| 171 | int error = 0; |
| 172 | |
| 173 | ut = get_bsdthread_info(self); |
| 174 | |
| 175 | p = current_proc(); |
| 176 | p->p_priority = pri & PRIMASK; |
| 177 | /* It can still block in proc_exit() after the teardown. */ |
| 178 | if (p->p_stats != NULL) { |
| 179 | OSIncrementAtomicLong(address: &p->p_stats->p_ru.ru_nvcsw); |
| 180 | } |
| 181 | |
| 182 | if (pri & PCATCH) { |
| 183 | catch = THREAD_ABORTSAFE; |
| 184 | } else { |
| 185 | catch = THREAD_UNINT; |
| 186 | } |
| 187 | |
| 188 | /* set wait message & channel */ |
| 189 | ut->uu_wchan = chan; |
| 190 | ut->uu_wmesg = wmsg ? wmsg : "unknown" ; |
| 191 | |
| 192 | if (mtx != NULL && chan != NULL && continuation == NULL) { |
| 193 | int flags; |
| 194 | |
| 195 | if (dropmutex) { |
| 196 | flags = LCK_SLEEP_UNLOCK; |
| 197 | } else { |
| 198 | flags = LCK_SLEEP_DEFAULT; |
| 199 | } |
| 200 | |
| 201 | if (spinmutex) { |
| 202 | flags |= LCK_SLEEP_SPIN; |
| 203 | } |
| 204 | |
| 205 | if (abstime) { |
| 206 | wait_result = lck_mtx_sleep_deadline(lck: mtx, lck_sleep_action: flags, event: chan, interruptible: catch, deadline: abstime); |
| 207 | } else { |
| 208 | wait_result = lck_mtx_sleep(lck: mtx, lck_sleep_action: flags, event: chan, interruptible: catch); |
| 209 | } |
| 210 | } else { |
| 211 | if (chan != NULL) { |
| 212 | assert_wait_deadline(event: chan, interruptible: catch, deadline: abstime); |
| 213 | } |
| 214 | if (mtx) { |
| 215 | lck_mtx_unlock(lck: mtx); |
| 216 | } |
| 217 | |
| 218 | if (catch == THREAD_ABORTSAFE) { |
| 219 | if (SHOULDissignal(p, ut)) { |
| 220 | if ((sig = CURSIG(p)) != 0) { |
| 221 | if (clear_wait(thread: self, THREAD_INTERRUPTED) == KERN_FAILURE) { |
| 222 | goto block; |
| 223 | } |
| 224 | if (p->p_sigacts.ps_sigintr & sigmask(sig)) { |
| 225 | error = EINTR; |
| 226 | } else { |
| 227 | error = ERESTART; |
| 228 | } |
| 229 | if (mtx && !dropmutex) { |
| 230 | if (spinmutex) { |
| 231 | lck_mtx_lock_spin(lck: mtx); |
| 232 | } else { |
| 233 | lck_mtx_lock(lck: mtx); |
| 234 | } |
| 235 | } |
| 236 | goto out; |
| 237 | } |
| 238 | } |
| 239 | if (thread_should_abort(self)) { |
| 240 | if (clear_wait(thread: self, THREAD_INTERRUPTED) == KERN_FAILURE) { |
| 241 | goto block; |
| 242 | } |
| 243 | error = EINTR; |
| 244 | |
| 245 | if (mtx && !dropmutex) { |
| 246 | if (spinmutex) { |
| 247 | lck_mtx_lock_spin(lck: mtx); |
| 248 | } else { |
| 249 | lck_mtx_lock(lck: mtx); |
| 250 | } |
| 251 | } |
| 252 | goto out; |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | |
| 257 | block: |
| 258 | if (continuation != NULL) { |
| 259 | ut->uu_continuation = continuation; |
| 260 | ut->uu_pri = (uint16_t)pri; |
| 261 | ut->uu_mtx = mtx; |
| 262 | (void) thread_block(continuation: _sleep_continue); |
| 263 | /* NOTREACHED */ |
| 264 | } |
| 265 | |
| 266 | wait_result = thread_block(THREAD_CONTINUE_NULL); |
| 267 | |
| 268 | if (mtx && !dropmutex) { |
| 269 | if (spinmutex) { |
| 270 | lck_mtx_lock_spin(lck: mtx); |
| 271 | } else { |
| 272 | lck_mtx_lock(lck: mtx); |
| 273 | } |
| 274 | } |
| 275 | } |
| 276 | |
| 277 | switch (wait_result) { |
| 278 | case THREAD_TIMED_OUT: |
| 279 | error = EWOULDBLOCK; |
| 280 | break; |
| 281 | case THREAD_AWAKENED: |
| 282 | case THREAD_RESTART: |
| 283 | /* |
| 284 | * Posix implies any signal should be delivered |
| 285 | * first, regardless of whether awakened due |
| 286 | * to receiving event. |
| 287 | */ |
| 288 | if (catch != THREAD_ABORTSAFE) { |
| 289 | break; |
| 290 | } |
| 291 | OS_FALLTHROUGH; |
| 292 | case THREAD_INTERRUPTED: |
| 293 | if (catch == THREAD_ABORTSAFE) { |
| 294 | if (thread_should_abort(self)) { |
| 295 | error = EINTR; |
| 296 | } else if (SHOULDissignal(p, ut)) { |
| 297 | if ((sig = CURSIG(p)) != 0) { |
| 298 | if (p->p_sigacts.ps_sigintr & sigmask(sig)) { |
| 299 | error = EINTR; |
| 300 | } else { |
| 301 | error = ERESTART; |
| 302 | } |
| 303 | } |
| 304 | if (thread_should_abort(self)) { |
| 305 | error = EINTR; |
| 306 | } |
| 307 | } else if ((ut->uu_flag & (UT_CANCELDISABLE | UT_CANCEL | UT_CANCELED)) == UT_CANCEL) { |
| 308 | /* due to thread cancel */ |
| 309 | error = EINTR; |
| 310 | } |
| 311 | } else { |
| 312 | error = EINTR; |
| 313 | } |
| 314 | break; |
| 315 | } |
| 316 | out: |
| 317 | if (error == EINTR || error == ERESTART) { |
| 318 | act_set_astbsd(self); |
| 319 | } |
| 320 | ut->uu_wchan = NULL; |
| 321 | ut->uu_wmesg = NULL; |
| 322 | |
| 323 | return error; |
| 324 | } |
| 325 | |
| 326 | int |
| 327 | sleep( |
| 328 | void *chan, |
| 329 | int pri) |
| 330 | { |
| 331 | return _sleep(chan: (caddr_t)chan, pri, wmsg: (char *)NULL, abstime: 0, continuation: (int (*)(int))0, mtx: (lck_mtx_t *)0); |
| 332 | } |
| 333 | |
| 334 | int |
| 335 | msleep0( |
| 336 | void *chan, |
| 337 | lck_mtx_t *mtx, |
| 338 | int pri, |
| 339 | const char *wmsg, |
| 340 | int timo, |
| 341 | int (*continuation)(int)) |
| 342 | { |
| 343 | u_int64_t abstime = 0; |
| 344 | |
| 345 | if (timo) { |
| 346 | clock_interval_to_deadline(interval: timo, NSEC_PER_SEC / hz, result: &abstime); |
| 347 | } |
| 348 | |
| 349 | return _sleep(chan: (caddr_t)chan, pri, wmsg, abstime, continuation, mtx); |
| 350 | } |
| 351 | |
| 352 | int |
| 353 | msleep( |
| 354 | void *chan, |
| 355 | lck_mtx_t *mtx, |
| 356 | int pri, |
| 357 | const char *wmsg, |
| 358 | struct timespec *ts) |
| 359 | { |
| 360 | u_int64_t abstime = 0; |
| 361 | |
| 362 | if (ts && (ts->tv_sec || ts->tv_nsec)) { |
| 363 | nanoseconds_to_absolutetime(nanoseconds: (uint64_t)ts->tv_sec * NSEC_PER_SEC + ts->tv_nsec, result: &abstime ); |
| 364 | clock_absolutetime_interval_to_deadline( abstime, result: &abstime ); |
| 365 | } |
| 366 | |
| 367 | return _sleep(chan: (caddr_t)chan, pri, wmsg, abstime, continuation: (int (*)(int))0, mtx); |
| 368 | } |
| 369 | |
| 370 | int |
| 371 | msleep1( |
| 372 | void *chan, |
| 373 | lck_mtx_t *mtx, |
| 374 | int pri, |
| 375 | const char *wmsg, |
| 376 | u_int64_t abstime) |
| 377 | { |
| 378 | return _sleep(chan: (caddr_t)chan, pri, wmsg, abstime, continuation: (int (*)(int))0, mtx); |
| 379 | } |
| 380 | |
| 381 | int |
| 382 | tsleep( |
| 383 | void *chan, |
| 384 | int pri, |
| 385 | const char *wmsg, |
| 386 | int timo) |
| 387 | { |
| 388 | u_int64_t abstime = 0; |
| 389 | |
| 390 | if (timo) { |
| 391 | clock_interval_to_deadline(interval: timo, NSEC_PER_SEC / hz, result: &abstime); |
| 392 | } |
| 393 | return _sleep(chan: (caddr_t)chan, pri, wmsg, abstime, continuation: (int (*)(int))0, mtx: (lck_mtx_t *)0); |
| 394 | } |
| 395 | |
| 396 | int |
| 397 | tsleep0( |
| 398 | void *chan, |
| 399 | int pri, |
| 400 | const char *wmsg, |
| 401 | int timo, |
| 402 | int (*continuation)(int)) |
| 403 | { |
| 404 | u_int64_t abstime = 0; |
| 405 | |
| 406 | if (timo) { |
| 407 | clock_interval_to_deadline(interval: timo, NSEC_PER_SEC / hz, result: &abstime); |
| 408 | } |
| 409 | return _sleep(chan: (caddr_t)chan, pri, wmsg, abstime, continuation, mtx: (lck_mtx_t *)0); |
| 410 | } |
| 411 | |
| 412 | int |
| 413 | tsleep1( |
| 414 | void *chan, |
| 415 | int pri, |
| 416 | const char *wmsg, |
| 417 | u_int64_t abstime, |
| 418 | int (*continuation)(int)) |
| 419 | { |
| 420 | return _sleep(chan: (caddr_t)chan, pri, wmsg, abstime, continuation, mtx: (lck_mtx_t *)0); |
| 421 | } |
| 422 | |
| 423 | /* |
| 424 | * Wake up all processes sleeping on chan. |
| 425 | */ |
| 426 | void |
| 427 | wakeup(void *chan) |
| 428 | { |
| 429 | thread_wakeup((caddr_t)chan); |
| 430 | } |
| 431 | |
| 432 | /* |
| 433 | * Wake up the first process sleeping on chan. |
| 434 | * |
| 435 | * Be very sure that the first process is really |
| 436 | * the right one to wakeup. |
| 437 | */ |
| 438 | void |
| 439 | wakeup_one(caddr_t chan) |
| 440 | { |
| 441 | thread_wakeup_one((caddr_t)chan); |
| 442 | } |
| 443 | |
| 444 | /* |
| 445 | * Compute the priority of a process when running in user mode. |
| 446 | * Arrange to reschedule if the resulting priority is better |
| 447 | * than that of the current process. |
| 448 | */ |
| 449 | void |
| 450 | resetpriority(struct proc *p) |
| 451 | { |
| 452 | (void)task_importance(task: proc_task(p), importance: -p->p_nice); |
| 453 | } |
| 454 | |
| 455 | struct loadavg averunnable = |
| 456 | { {0, 0, 0}, FSCALE }; /* load average, of runnable procs */ |
| 457 | /* |
| 458 | * Constants for averages over 1, 5, and 15 minutes |
| 459 | * when sampling at 5 second intervals. |
| 460 | */ |
| 461 | static fixpt_t cexp[3] = { |
| 462 | (fixpt_t)(0.9200444146293232 * FSCALE), /* exp(-1/12) */ |
| 463 | (fixpt_t)(0.9834714538216174 * FSCALE), /* exp(-1/60) */ |
| 464 | (fixpt_t)(0.9944598480048967 * FSCALE), /* exp(-1/180) */ |
| 465 | }; |
| 466 | |
| 467 | void |
| 468 | compute_averunnable(void *arg) |
| 469 | { |
| 470 | unsigned int nrun = *(unsigned int *)arg; |
| 471 | struct loadavg *avg = &averunnable; |
| 472 | int i; |
| 473 | |
| 474 | for (i = 0; i < 3; i++) { |
| 475 | avg->ldavg[i] = (cexp[i] * avg->ldavg[i] + |
| 476 | nrun * FSCALE * (FSCALE - cexp[i])) >> FSHIFT; |
| 477 | } |
| 478 | } |
| 479 | |