1 /*- 2 * Copyright (c) 1982, 1986, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * From: @(#)kern_clock.c 8.5 (Berkeley) 1/21/94 35 */ 36 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD$"); 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/callout.h> 43 #include <sys/condvar.h> 44 #include <sys/kernel.h> 45 #include <sys/lock.h> 46 #include <sys/mutex.h> 47 #include <sys/sysctl.h> 48 49 static int avg_depth; 50 SYSCTL_INT(_debug, OID_AUTO, to_avg_depth, CTLFLAG_RD, &avg_depth, 0, 51 "Average number of items examined per softclock call. Units = 1/1000"); 52 static int avg_gcalls; 53 SYSCTL_INT(_debug, OID_AUTO, to_avg_gcalls, CTLFLAG_RD, &avg_gcalls, 0, 54 "Average number of Giant callouts made per softclock call. Units = 1/1000"); 55 static int avg_mpcalls; 56 SYSCTL_INT(_debug, OID_AUTO, to_avg_mpcalls, CTLFLAG_RD, &avg_mpcalls, 0, 57 "Average number of MP callouts made per softclock call. Units = 1/1000"); 58 /* 59 * TODO: 60 * allocate more timeout table slots when table overflows. 61 */ 62 63 /* Exported to machdep.c and/or kern_clock.c. */ 64 struct callout *callout; 65 struct callout_list callfree; 66 int callwheelsize, callwheelbits, callwheelmask; 67 struct callout_tailq *callwheel; 68 int softticks; /* Like ticks, but for softclock(). */ 69 struct mtx callout_lock; 70 #ifdef DIAGNOSTIC 71 struct mtx dont_sleep_in_callout; 72 #endif 73 74 static struct callout *nextsoftcheck; /* Next callout to be checked. */ 75 76 /*- 77 * Locked by callout_lock: 78 * curr_callout - If a callout is in progress, it is curr_callout. 79 * If curr_callout is non-NULL, threads waiting on 80 * callout_wait will be woken up as soon as the 81 * relevant callout completes. 82 * wakeup_ctr - Incremented every time a thread wants to wait 83 * for a callout to complete. Modified only when 84 * curr_callout is non-NULL. 85 * wakeup_needed - If a thread is waiting on callout_wait, then 86 * wakeup_needed is nonzero. Increased only when 87 * cutt_callout is non-NULL. 88 */ 89 static struct callout *curr_callout; 90 static int wakeup_ctr; 91 static int wakeup_needed; 92 93 /*- 94 * Locked by callout_wait_lock: 95 * callout_wait - If wakeup_needed is set, callout_wait will be 96 * triggered after the current callout finishes. 97 * wakeup_done_ctr - Set to the current value of wakeup_ctr after 98 * callout_wait is triggered. 99 */ 100 static struct mtx callout_wait_lock; 101 static struct cv callout_wait; 102 static int wakeup_done_ctr; 103 104 /* 105 * kern_timeout_callwheel_alloc() - kernel low level callwheel initialization 106 * 107 * This code is called very early in the kernel initialization sequence, 108 * and may be called more then once. 109 */ 110 caddr_t 111 kern_timeout_callwheel_alloc(caddr_t v) 112 { 113 /* 114 * Calculate callout wheel size 115 */ 116 for (callwheelsize = 1, callwheelbits = 0; 117 callwheelsize < ncallout; 118 callwheelsize <<= 1, ++callwheelbits) 119 ; 120 callwheelmask = callwheelsize - 1; 121 122 callout = (struct callout *)v; 123 v = (caddr_t)(callout + ncallout); 124 callwheel = (struct callout_tailq *)v; 125 v = (caddr_t)(callwheel + callwheelsize); 126 return(v); 127 } 128 129 /* 130 * kern_timeout_callwheel_init() - initialize previously reserved callwheel 131 * space. 132 * 133 * This code is called just once, after the space reserved for the 134 * callout wheel has been finalized. 135 */ 136 void 137 kern_timeout_callwheel_init(void) 138 { 139 int i; 140 141 SLIST_INIT(&callfree); 142 for (i = 0; i < ncallout; i++) { 143 callout_init(&callout[i], 0); 144 callout[i].c_flags = CALLOUT_LOCAL_ALLOC; 145 SLIST_INSERT_HEAD(&callfree, &callout[i], c_links.sle); 146 } 147 for (i = 0; i < callwheelsize; i++) { 148 TAILQ_INIT(&callwheel[i]); 149 } 150 mtx_init(&callout_lock, "callout", NULL, MTX_SPIN | MTX_RECURSE); 151 #ifdef DIAGNOSTIC 152 mtx_init(&dont_sleep_in_callout, "dont_sleep_in_callout", NULL, MTX_DEF); 153 #endif 154 mtx_init(&callout_wait_lock, "callout_wait_lock", NULL, MTX_DEF); 155 cv_init(&callout_wait, "callout_wait"); 156 } 157 158 /* 159 * The callout mechanism is based on the work of Adam M. Costello and 160 * George Varghese, published in a technical report entitled "Redesigning 161 * the BSD Callout and Timer Facilities" and modified slightly for inclusion 162 * in FreeBSD by Justin T. Gibbs. The original work on the data structures 163 * used in this implementation was published by G. Varghese and T. Lauck in 164 * the paper "Hashed and Hierarchical Timing Wheels: Data Structures for 165 * the Efficient Implementation of a Timer Facility" in the Proceedings of 166 * the 11th ACM Annual Symposium on Operating Systems Principles, 167 * Austin, Texas Nov 1987. 168 */ 169 170 /* 171 * Software (low priority) clock interrupt. 172 * Run periodic events from timeout queue. 173 */ 174 void 175 softclock(void *dummy) 176 { 177 struct callout *c; 178 struct callout_tailq *bucket; 179 int curticks; 180 int steps; /* #steps since we last allowed interrupts */ 181 int depth; 182 int mpcalls; 183 int gcalls; 184 int wakeup_cookie; 185 #ifdef DIAGNOSTIC 186 struct bintime bt1, bt2; 187 struct timespec ts2; 188 static uint64_t maxdt = 36893488147419102LL; /* 2 msec */ 189 static timeout_t *lastfunc; 190 #endif 191 192 #ifndef MAX_SOFTCLOCK_STEPS 193 #define MAX_SOFTCLOCK_STEPS 100 /* Maximum allowed value of steps. */ 194 #endif /* MAX_SOFTCLOCK_STEPS */ 195 196 mpcalls = 0; 197 gcalls = 0; 198 depth = 0; 199 steps = 0; 200 mtx_lock_spin(&callout_lock); 201 while (softticks != ticks) { 202 softticks++; 203 /* 204 * softticks may be modified by hard clock, so cache 205 * it while we work on a given bucket. 206 */ 207 curticks = softticks; 208 bucket = &callwheel[curticks & callwheelmask]; 209 c = TAILQ_FIRST(bucket); 210 while (c) { 211 depth++; 212 if (c->c_time != curticks) { 213 c = TAILQ_NEXT(c, c_links.tqe); 214 ++steps; 215 if (steps >= MAX_SOFTCLOCK_STEPS) { 216 nextsoftcheck = c; 217 /* Give interrupts a chance. */ 218 mtx_unlock_spin(&callout_lock); 219 ; /* nothing */ 220 mtx_lock_spin(&callout_lock); 221 c = nextsoftcheck; 222 steps = 0; 223 } 224 } else { 225 void (*c_func)(void *); 226 void *c_arg; 227 int c_flags; 228 229 nextsoftcheck = TAILQ_NEXT(c, c_links.tqe); 230 TAILQ_REMOVE(bucket, c, c_links.tqe); 231 c_func = c->c_func; 232 c_arg = c->c_arg; 233 c_flags = c->c_flags; 234 c->c_func = NULL; 235 if (c->c_flags & CALLOUT_LOCAL_ALLOC) { 236 c->c_flags = CALLOUT_LOCAL_ALLOC; 237 SLIST_INSERT_HEAD(&callfree, c, 238 c_links.sle); 239 } else { 240 c->c_flags = 241 (c->c_flags & ~CALLOUT_PENDING); 242 } 243 curr_callout = c; 244 mtx_unlock_spin(&callout_lock); 245 if (!(c_flags & CALLOUT_MPSAFE)) { 246 mtx_lock(&Giant); 247 gcalls++; 248 } else { 249 mpcalls++; 250 } 251 #ifdef DIAGNOSTIC 252 binuptime(&bt1); 253 mtx_lock(&dont_sleep_in_callout); 254 #endif 255 c_func(c_arg); 256 #ifdef DIAGNOSTIC 257 mtx_unlock(&dont_sleep_in_callout); 258 binuptime(&bt2); 259 bintime_sub(&bt2, &bt1); 260 if (bt2.frac > maxdt) { 261 if (lastfunc != c_func || 262 bt2.frac > maxdt * 2) { 263 bintime2timespec(&bt2, &ts2); 264 printf( 265 "Expensive timeout(9) function: %p(%p) %jd.%09ld s\n", 266 c_func, c_arg, 267 (intmax_t)ts2.tv_sec, 268 ts2.tv_nsec); 269 } 270 maxdt = bt2.frac; 271 lastfunc = c_func; 272 } 273 #endif 274 if (!(c_flags & CALLOUT_MPSAFE)) 275 mtx_unlock(&Giant); 276 mtx_lock_spin(&callout_lock); 277 curr_callout = NULL; 278 if (wakeup_needed) { 279 /* 280 * There might be someone waiting 281 * for the callout to complete. 282 */ 283 wakeup_cookie = wakeup_ctr; 284 mtx_unlock_spin(&callout_lock); 285 mtx_lock(&callout_wait_lock); 286 cv_broadcast(&callout_wait); 287 wakeup_done_ctr = wakeup_cookie; 288 mtx_unlock(&callout_wait_lock); 289 mtx_lock_spin(&callout_lock); 290 wakeup_needed = 0; 291 } 292 steps = 0; 293 c = nextsoftcheck; 294 } 295 } 296 } 297 avg_depth += (depth * 1000 - avg_depth) >> 8; 298 avg_mpcalls += (mpcalls * 1000 - avg_mpcalls) >> 8; 299 avg_gcalls += (gcalls * 1000 - avg_gcalls) >> 8; 300 nextsoftcheck = NULL; 301 mtx_unlock_spin(&callout_lock); 302 } 303 304 /* 305 * timeout -- 306 * Execute a function after a specified length of time. 307 * 308 * untimeout -- 309 * Cancel previous timeout function call. 310 * 311 * callout_handle_init -- 312 * Initialize a handle so that using it with untimeout is benign. 313 * 314 * See AT&T BCI Driver Reference Manual for specification. This 315 * implementation differs from that one in that although an 316 * identification value is returned from timeout, the original 317 * arguments to timeout as well as the identifier are used to 318 * identify entries for untimeout. 319 */ 320 struct callout_handle 321 timeout(ftn, arg, to_ticks) 322 timeout_t *ftn; 323 void *arg; 324 int to_ticks; 325 { 326 struct callout *new; 327 struct callout_handle handle; 328 329 mtx_lock_spin(&callout_lock); 330 331 /* Fill in the next free callout structure. */ 332 new = SLIST_FIRST(&callfree); 333 if (new == NULL) 334 /* XXX Attempt to malloc first */ 335 panic("timeout table full"); 336 SLIST_REMOVE_HEAD(&callfree, c_links.sle); 337 338 callout_reset(new, to_ticks, ftn, arg); 339 340 handle.callout = new; 341 mtx_unlock_spin(&callout_lock); 342 return (handle); 343 } 344 345 void 346 untimeout(ftn, arg, handle) 347 timeout_t *ftn; 348 void *arg; 349 struct callout_handle handle; 350 { 351 352 /* 353 * Check for a handle that was initialized 354 * by callout_handle_init, but never used 355 * for a real timeout. 356 */ 357 if (handle.callout == NULL) 358 return; 359 360 mtx_lock_spin(&callout_lock); 361 if (handle.callout->c_func == ftn && handle.callout->c_arg == arg) 362 callout_stop(handle.callout); 363 mtx_unlock_spin(&callout_lock); 364 } 365 366 void 367 callout_handle_init(struct callout_handle *handle) 368 { 369 handle->callout = NULL; 370 } 371 372 /* 373 * New interface; clients allocate their own callout structures. 374 * 375 * callout_reset() - establish or change a timeout 376 * callout_stop() - disestablish a timeout 377 * callout_init() - initialize a callout structure so that it can 378 * safely be passed to callout_reset() and callout_stop() 379 * 380 * <sys/callout.h> defines three convenience macros: 381 * 382 * callout_active() - returns truth if callout has not been serviced 383 * callout_pending() - returns truth if callout is still waiting for timeout 384 * callout_deactivate() - marks the callout as having been serviced 385 */ 386 void 387 callout_reset(c, to_ticks, ftn, arg) 388 struct callout *c; 389 int to_ticks; 390 void (*ftn)(void *); 391 void *arg; 392 { 393 394 mtx_lock_spin(&callout_lock); 395 if (c == curr_callout && wakeup_needed) { 396 /* 397 * We're being asked to reschedule a callout which is 398 * currently in progress, and someone has called 399 * callout_drain to kill that callout. Don't reschedule. 400 */ 401 mtx_unlock_spin(&callout_lock); 402 return; 403 } 404 if (c->c_flags & CALLOUT_PENDING) 405 callout_stop(c); 406 407 /* 408 * We could unlock callout_lock here and lock it again before the 409 * TAILQ_INSERT_TAIL, but there's no point since doing this setup 410 * doesn't take much time. 411 */ 412 if (to_ticks <= 0) 413 to_ticks = 1; 414 415 c->c_arg = arg; 416 c->c_flags |= (CALLOUT_ACTIVE | CALLOUT_PENDING); 417 c->c_func = ftn; 418 c->c_time = ticks + to_ticks; 419 TAILQ_INSERT_TAIL(&callwheel[c->c_time & callwheelmask], 420 c, c_links.tqe); 421 mtx_unlock_spin(&callout_lock); 422 } 423 424 int 425 _callout_stop_safe(c, safe) 426 struct callout *c; 427 int safe; 428 { 429 int wakeup_cookie; 430 431 mtx_lock_spin(&callout_lock); 432 /* 433 * Don't attempt to delete a callout that's not on the queue. 434 */ 435 if (!(c->c_flags & CALLOUT_PENDING)) { 436 c->c_flags &= ~CALLOUT_ACTIVE; 437 if (c == curr_callout && safe) { 438 /* We need to wait until the callout is finished. */ 439 wakeup_needed = 1; 440 wakeup_cookie = wakeup_ctr++; 441 mtx_unlock_spin(&callout_lock); 442 mtx_lock(&callout_wait_lock); 443 444 /* 445 * Check to make sure that softclock() didn't 446 * do the wakeup in between our dropping 447 * callout_lock and picking up callout_wait_lock 448 */ 449 if (wakeup_cookie - wakeup_done_ctr > 0) 450 cv_wait(&callout_wait, &callout_wait_lock); 451 452 mtx_unlock(&callout_wait_lock); 453 } else 454 mtx_unlock_spin(&callout_lock); 455 return (0); 456 } 457 c->c_flags &= ~(CALLOUT_ACTIVE | CALLOUT_PENDING); 458 459 if (nextsoftcheck == c) { 460 nextsoftcheck = TAILQ_NEXT(c, c_links.tqe); 461 } 462 TAILQ_REMOVE(&callwheel[c->c_time & callwheelmask], c, c_links.tqe); 463 c->c_func = NULL; 464 465 if (c->c_flags & CALLOUT_LOCAL_ALLOC) { 466 SLIST_INSERT_HEAD(&callfree, c, c_links.sle); 467 } 468 mtx_unlock_spin(&callout_lock); 469 return (1); 470 } 471 472 void 473 callout_init(c, mpsafe) 474 struct callout *c; 475 int mpsafe; 476 { 477 bzero(c, sizeof *c); 478 if (mpsafe) 479 c->c_flags |= CALLOUT_MPSAFE; 480 } 481 482 #ifdef APM_FIXUP_CALLTODO 483 /* 484 * Adjust the kernel calltodo timeout list. This routine is used after 485 * an APM resume to recalculate the calltodo timer list values with the 486 * number of hz's we have been sleeping. The next hardclock() will detect 487 * that there are fired timers and run softclock() to execute them. 488 * 489 * Please note, I have not done an exhaustive analysis of what code this 490 * might break. I am motivated to have my select()'s and alarm()'s that 491 * have expired during suspend firing upon resume so that the applications 492 * which set the timer can do the maintanence the timer was for as close 493 * as possible to the originally intended time. Testing this code for a 494 * week showed that resuming from a suspend resulted in 22 to 25 timers 495 * firing, which seemed independant on whether the suspend was 2 hours or 496 * 2 days. Your milage may vary. - Ken Key <key@cs.utk.edu> 497 */ 498 void 499 adjust_timeout_calltodo(time_change) 500 struct timeval *time_change; 501 { 502 register struct callout *p; 503 unsigned long delta_ticks; 504 505 /* 506 * How many ticks were we asleep? 507 * (stolen from tvtohz()). 508 */ 509 510 /* Don't do anything */ 511 if (time_change->tv_sec < 0) 512 return; 513 else if (time_change->tv_sec <= LONG_MAX / 1000000) 514 delta_ticks = (time_change->tv_sec * 1000000 + 515 time_change->tv_usec + (tick - 1)) / tick + 1; 516 else if (time_change->tv_sec <= LONG_MAX / hz) 517 delta_ticks = time_change->tv_sec * hz + 518 (time_change->tv_usec + (tick - 1)) / tick + 1; 519 else 520 delta_ticks = LONG_MAX; 521 522 if (delta_ticks > INT_MAX) 523 delta_ticks = INT_MAX; 524 525 /* 526 * Now rip through the timer calltodo list looking for timers 527 * to expire. 528 */ 529 530 /* don't collide with softclock() */ 531 mtx_lock_spin(&callout_lock); 532 for (p = calltodo.c_next; p != NULL; p = p->c_next) { 533 p->c_time -= delta_ticks; 534 535 /* Break if the timer had more time on it than delta_ticks */ 536 if (p->c_time > 0) 537 break; 538 539 /* take back the ticks the timer didn't use (p->c_time <= 0) */ 540 delta_ticks = -p->c_time; 541 } 542 mtx_unlock_spin(&callout_lock); 543 544 return; 545 } 546 #endif /* APM_FIXUP_CALLTODO */ 547