kern_time.c revision 1.86
1/* $NetBSD: kern_time.c,v 1.86 2005/01/06 19:26:41 mycroft Exp $ */ 2 3/*- 4 * Copyright (c) 2000 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Christopher G. Demetriou. 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 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39/* 40 * Copyright (c) 1982, 1986, 1989, 1993 41 * The Regents of the University of California. All rights reserved. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. Neither the name of the University nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 65 * SUCH DAMAGE. 66 * 67 * @(#)kern_time.c 8.4 (Berkeley) 5/26/95 68 */ 69 70#include <sys/cdefs.h> 71__KERNEL_RCSID(0, "$NetBSD: kern_time.c,v 1.86 2005/01/06 19:26:41 mycroft Exp $"); 72 73#include "fs_nfs.h" 74#include "opt_nfs.h" 75#include "opt_nfsserver.h" 76 77#include <sys/param.h> 78#include <sys/resourcevar.h> 79#include <sys/kernel.h> 80#include <sys/systm.h> 81#include <sys/malloc.h> 82#include <sys/proc.h> 83#include <sys/sa.h> 84#include <sys/savar.h> 85#include <sys/vnode.h> 86#include <sys/signalvar.h> 87#include <sys/syslog.h> 88 89#include <sys/mount.h> 90#include <sys/syscallargs.h> 91 92#include <uvm/uvm_extern.h> 93 94#if defined(NFS) || defined(NFSSERVER) 95#include <nfs/rpcv2.h> 96#include <nfs/nfsproto.h> 97#include <nfs/nfs_var.h> 98#endif 99 100#include <machine/cpu.h> 101 102static void timerupcall(struct lwp *, void *); 103 104 105/* Time of day and interval timer support. 106 * 107 * These routines provide the kernel entry points to get and set 108 * the time-of-day and per-process interval timers. Subroutines 109 * here provide support for adding and subtracting timeval structures 110 * and decrementing interval timers, optionally reloading the interval 111 * timers when they expire. 112 */ 113 114/* This function is used by clock_settime and settimeofday */ 115int 116settime(struct timeval *tv) 117{ 118 struct timeval delta; 119 struct cpu_info *ci; 120 int s; 121 122 /* WHAT DO WE DO ABOUT PENDING REAL-TIME TIMEOUTS??? */ 123 s = splclock(); 124 timersub(tv, &time, &delta); 125 if ((delta.tv_sec < 0 || delta.tv_usec < 0) && securelevel > 1) { 126 splx(s); 127 return (EPERM); 128 } 129#ifdef notyet 130 if ((delta.tv_sec < 86400) && securelevel > 0) { 131 splx(s); 132 return (EPERM); 133 } 134#endif 135 time = *tv; 136 (void) spllowersoftclock(); 137 timeradd(&boottime, &delta, &boottime); 138 /* 139 * XXXSMP 140 * This is wrong. We should traverse a list of all 141 * CPUs and add the delta to the runtime of those 142 * CPUs which have a process on them. 143 */ 144 ci = curcpu(); 145 timeradd(&ci->ci_schedstate.spc_runtime, &delta, 146 &ci->ci_schedstate.spc_runtime); 147# if (defined(NFS) && !defined (NFS_V2_ONLY)) || defined(NFSSERVER) 148 nqnfs_lease_updatetime(delta.tv_sec); 149# endif 150 splx(s); 151 resettodr(); 152 return (0); 153} 154 155/* ARGSUSED */ 156int 157sys_clock_gettime(struct lwp *l, void *v, register_t *retval) 158{ 159 struct sys_clock_gettime_args /* { 160 syscallarg(clockid_t) clock_id; 161 syscallarg(struct timespec *) tp; 162 } */ *uap = v; 163 clockid_t clock_id; 164 struct timeval atv; 165 struct timespec ats; 166 int s; 167 168 clock_id = SCARG(uap, clock_id); 169 switch (clock_id) { 170 case CLOCK_REALTIME: 171 microtime(&atv); 172 TIMEVAL_TO_TIMESPEC(&atv,&ats); 173 break; 174 case CLOCK_MONOTONIC: 175 /* XXX "hz" granularity */ 176 s = splclock(); 177 atv = mono_time; 178 splx(s); 179 TIMEVAL_TO_TIMESPEC(&atv,&ats); 180 break; 181 default: 182 return (EINVAL); 183 } 184 185 return copyout(&ats, SCARG(uap, tp), sizeof(ats)); 186} 187 188/* ARGSUSED */ 189int 190sys_clock_settime(l, v, retval) 191 struct lwp *l; 192 void *v; 193 register_t *retval; 194{ 195 struct sys_clock_settime_args /* { 196 syscallarg(clockid_t) clock_id; 197 syscallarg(const struct timespec *) tp; 198 } */ *uap = v; 199 struct proc *p = l->l_proc; 200 int error; 201 202 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 203 return (error); 204 205 return (clock_settime1(SCARG(uap, clock_id), SCARG(uap, tp))); 206} 207 208 209int 210clock_settime1(clock_id, tp) 211 clockid_t clock_id; 212 const struct timespec *tp; 213{ 214 struct timespec ats; 215 struct timeval atv; 216 int error; 217 218 if ((error = copyin(tp, &ats, sizeof(ats))) != 0) 219 return (error); 220 221 switch (clock_id) { 222 case CLOCK_REALTIME: 223 TIMESPEC_TO_TIMEVAL(&atv, &ats); 224 if ((error = settime(&atv)) != 0) 225 return (error); 226 break; 227 case CLOCK_MONOTONIC: 228 return (EINVAL); /* read-only clock */ 229 default: 230 return (EINVAL); 231 } 232 233 return 0; 234} 235 236int 237sys_clock_getres(struct lwp *l, void *v, register_t *retval) 238{ 239 struct sys_clock_getres_args /* { 240 syscallarg(clockid_t) clock_id; 241 syscallarg(struct timespec *) tp; 242 } */ *uap = v; 243 clockid_t clock_id; 244 struct timespec ts; 245 int error = 0; 246 247 clock_id = SCARG(uap, clock_id); 248 switch (clock_id) { 249 case CLOCK_REALTIME: 250 case CLOCK_MONOTONIC: 251 ts.tv_sec = 0; 252 ts.tv_nsec = 1000000000 / hz; 253 break; 254 default: 255 return (EINVAL); 256 } 257 258 if (SCARG(uap, tp)) 259 error = copyout(&ts, SCARG(uap, tp), sizeof(ts)); 260 261 return error; 262} 263 264/* ARGSUSED */ 265int 266sys_nanosleep(struct lwp *l, void *v, register_t *retval) 267{ 268 static int nanowait; 269 struct sys_nanosleep_args/* { 270 syscallarg(struct timespec *) rqtp; 271 syscallarg(struct timespec *) rmtp; 272 } */ *uap = v; 273 struct timespec rqt; 274 struct timespec rmt; 275 struct timeval atv, utv; 276 int error, s, timo; 277 278 error = copyin((caddr_t)SCARG(uap, rqtp), (caddr_t)&rqt, 279 sizeof(struct timespec)); 280 if (error) 281 return (error); 282 283 TIMESPEC_TO_TIMEVAL(&atv,&rqt); 284 if (itimerfix(&atv)) 285 return (EINVAL); 286 287 s = splclock(); 288 timeradd(&atv,&time,&atv); 289 timo = hzto(&atv); 290 /* 291 * Avoid inadvertantly sleeping forever 292 */ 293 if (timo == 0) 294 timo = 1; 295 splx(s); 296 297 error = tsleep(&nanowait, PWAIT | PCATCH, "nanosleep", timo); 298 if (error == ERESTART) 299 error = EINTR; 300 if (error == EWOULDBLOCK) 301 error = 0; 302 303 if (SCARG(uap, rmtp)) { 304 int error; 305 306 s = splclock(); 307 utv = time; 308 splx(s); 309 310 timersub(&atv, &utv, &utv); 311 if (utv.tv_sec < 0) 312 timerclear(&utv); 313 314 TIMEVAL_TO_TIMESPEC(&utv,&rmt); 315 error = copyout((caddr_t)&rmt, (caddr_t)SCARG(uap,rmtp), 316 sizeof(rmt)); 317 if (error) 318 return (error); 319 } 320 321 return error; 322} 323 324/* ARGSUSED */ 325int 326sys_gettimeofday(struct lwp *l, void *v, register_t *retval) 327{ 328 struct sys_gettimeofday_args /* { 329 syscallarg(struct timeval *) tp; 330 syscallarg(void *) tzp; really "struct timezone *" 331 } */ *uap = v; 332 struct timeval atv; 333 int error = 0; 334 struct timezone tzfake; 335 336 if (SCARG(uap, tp)) { 337 microtime(&atv); 338 error = copyout(&atv, SCARG(uap, tp), sizeof(atv)); 339 if (error) 340 return (error); 341 } 342 if (SCARG(uap, tzp)) { 343 /* 344 * NetBSD has no kernel notion of time zone, so we just 345 * fake up a timezone struct and return it if demanded. 346 */ 347 tzfake.tz_minuteswest = 0; 348 tzfake.tz_dsttime = 0; 349 error = copyout(&tzfake, SCARG(uap, tzp), sizeof(tzfake)); 350 } 351 return (error); 352} 353 354/* ARGSUSED */ 355int 356sys_settimeofday(struct lwp *l, void *v, register_t *retval) 357{ 358 struct sys_settimeofday_args /* { 359 syscallarg(const struct timeval *) tv; 360 syscallarg(const void *) tzp; really "const struct timezone *" 361 } */ *uap = v; 362 struct proc *p = l->l_proc; 363 int error; 364 365 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 366 return (error); 367 368 return settimeofday1(SCARG(uap, tv), SCARG(uap, tzp), p); 369} 370 371int 372settimeofday1(utv, utzp, p) 373 const struct timeval *utv; 374 const struct timezone *utzp; 375 struct proc *p; 376{ 377 struct timeval atv; 378 struct timezone atz; 379 struct timeval *tv = NULL; 380 struct timezone *tzp = NULL; 381 int error; 382 383 /* Verify all parameters before changing time. */ 384 if (utv) { 385 if ((error = copyin(utv, &atv, sizeof(atv))) != 0) 386 return (error); 387 tv = &atv; 388 } 389 /* XXX since we don't use tz, probably no point in doing copyin. */ 390 if (utzp) { 391 if ((error = copyin(utzp, &atz, sizeof(atz))) != 0) 392 return (error); 393 tzp = &atz; 394 } 395 396 if (tv) 397 if ((error = settime(tv)) != 0) 398 return (error); 399 /* 400 * NetBSD has no kernel notion of time zone, and only an 401 * obsolete program would try to set it, so we log a warning. 402 */ 403 if (tzp) 404 log(LOG_WARNING, "pid %d attempted to set the " 405 "(obsolete) kernel time zone\n", p->p_pid); 406 return (0); 407} 408 409int tickdelta; /* current clock skew, us. per tick */ 410long timedelta; /* unapplied time correction, us. */ 411long bigadj = 1000000; /* use 10x skew above bigadj us. */ 412int time_adjusted; /* set if an adjustment is made */ 413 414/* ARGSUSED */ 415int 416sys_adjtime(struct lwp *l, void *v, register_t *retval) 417{ 418 struct sys_adjtime_args /* { 419 syscallarg(const struct timeval *) delta; 420 syscallarg(struct timeval *) olddelta; 421 } */ *uap = v; 422 struct proc *p = l->l_proc; 423 int error; 424 425 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 426 return (error); 427 428 return adjtime1(SCARG(uap, delta), SCARG(uap, olddelta), p); 429} 430 431int 432adjtime1(delta, olddelta, p) 433 const struct timeval *delta; 434 struct timeval *olddelta; 435 struct proc *p; 436{ 437 struct timeval atv; 438 long ndelta, ntickdelta, odelta; 439 int error; 440 int s; 441 442 error = copyin(delta, &atv, sizeof(struct timeval)); 443 if (error) 444 return (error); 445 446 /* 447 * Compute the total correction and the rate at which to apply it. 448 * Round the adjustment down to a whole multiple of the per-tick 449 * delta, so that after some number of incremental changes in 450 * hardclock(), tickdelta will become zero, lest the correction 451 * overshoot and start taking us away from the desired final time. 452 */ 453 ndelta = atv.tv_sec * 1000000 + atv.tv_usec; 454 if (ndelta > bigadj || ndelta < -bigadj) 455 ntickdelta = 10 * tickadj; 456 else 457 ntickdelta = tickadj; 458 if (ndelta % ntickdelta) 459 ndelta = ndelta / ntickdelta * ntickdelta; 460 461 /* 462 * To make hardclock()'s job easier, make the per-tick delta negative 463 * if we want time to run slower; then hardclock can simply compute 464 * tick + tickdelta, and subtract tickdelta from timedelta. 465 */ 466 if (ndelta < 0) 467 ntickdelta = -ntickdelta; 468 if (ndelta != 0) 469 /* We need to save the system clock time during shutdown */ 470 time_adjusted |= 1; 471 s = splclock(); 472 odelta = timedelta; 473 timedelta = ndelta; 474 tickdelta = ntickdelta; 475 splx(s); 476 477 if (olddelta) { 478 atv.tv_sec = odelta / 1000000; 479 atv.tv_usec = odelta % 1000000; 480 error = copyout(&atv, olddelta, sizeof(struct timeval)); 481 } 482 return error; 483} 484 485/* 486 * Interval timer support. Both the BSD getitimer() family and the POSIX 487 * timer_*() family of routines are supported. 488 * 489 * All timers are kept in an array pointed to by p_timers, which is 490 * allocated on demand - many processes don't use timers at all. The 491 * first three elements in this array are reserved for the BSD timers: 492 * element 0 is ITIMER_REAL, element 1 is ITIMER_VIRTUAL, and element 493 * 2 is ITIMER_PROF. The rest may be allocated by the timer_create() 494 * syscall. 495 * 496 * Realtime timers are kept in the ptimer structure as an absolute 497 * time; virtual time timers are kept as a linked list of deltas. 498 * Virtual time timers are processed in the hardclock() routine of 499 * kern_clock.c. The real time timer is processed by a callout 500 * routine, called from the softclock() routine. Since a callout may 501 * be delayed in real time due to interrupt processing in the system, 502 * it is possible for the real time timeout routine (realtimeexpire, 503 * given below), to be delayed in real time past when it is supposed 504 * to occur. It does not suffice, therefore, to reload the real timer 505 * .it_value from the real time timers .it_interval. Rather, we 506 * compute the next time in absolute time the timer should go off. */ 507 508/* Allocate a POSIX realtime timer. */ 509int 510sys_timer_create(struct lwp *l, void *v, register_t *retval) 511{ 512 struct sys_timer_create_args /* { 513 syscallarg(clockid_t) clock_id; 514 syscallarg(struct sigevent *) evp; 515 syscallarg(timer_t *) timerid; 516 } */ *uap = v; 517 struct proc *p = l->l_proc; 518 clockid_t id; 519 struct sigevent *evp; 520 struct ptimer *pt; 521 timer_t timerid; 522 int error; 523 524 id = SCARG(uap, clock_id); 525 if (id < CLOCK_REALTIME || 526 id > CLOCK_PROF) 527 return (EINVAL); 528 529 if (p->p_timers == NULL) 530 timers_alloc(p); 531 532 /* Find a free timer slot, skipping those reserved for setitimer(). */ 533 for (timerid = 3; timerid < TIMER_MAX; timerid++) 534 if (p->p_timers->pts_timers[timerid] == NULL) 535 break; 536 537 if (timerid == TIMER_MAX) 538 return EAGAIN; 539 540 pt = pool_get(&ptimer_pool, PR_WAITOK); 541 evp = SCARG(uap, evp); 542 if (evp) { 543 if (((error = 544 copyin(evp, &pt->pt_ev, sizeof (pt->pt_ev))) != 0) || 545 ((pt->pt_ev.sigev_notify < SIGEV_NONE) || 546 (pt->pt_ev.sigev_notify > SIGEV_SA))) { 547 pool_put(&ptimer_pool, pt); 548 return (error ? error : EINVAL); 549 } 550 } else { 551 pt->pt_ev.sigev_notify = SIGEV_SIGNAL; 552 switch (id) { 553 case CLOCK_REALTIME: 554 pt->pt_ev.sigev_signo = SIGALRM; 555 break; 556 case CLOCK_VIRTUAL: 557 pt->pt_ev.sigev_signo = SIGVTALRM; 558 break; 559 case CLOCK_PROF: 560 pt->pt_ev.sigev_signo = SIGPROF; 561 break; 562 } 563 pt->pt_ev.sigev_value.sival_int = timerid; 564 } 565 pt->pt_info.ksi_signo = pt->pt_ev.sigev_signo; 566 pt->pt_info.ksi_errno = 0; 567 pt->pt_info.ksi_code = 0; 568 pt->pt_info.ksi_pid = p->p_pid; 569 pt->pt_info.ksi_uid = p->p_cred->p_ruid; 570 pt->pt_info.ksi_sigval = pt->pt_ev.sigev_value; 571 572 pt->pt_type = id; 573 pt->pt_proc = p; 574 pt->pt_overruns = 0; 575 pt->pt_poverruns = 0; 576 pt->pt_entry = timerid; 577 timerclear(&pt->pt_time.it_value); 578 if (id == CLOCK_REALTIME) 579 callout_init(&pt->pt_ch); 580 else 581 pt->pt_active = 0; 582 583 p->p_timers->pts_timers[timerid] = pt; 584 585 return copyout(&timerid, SCARG(uap, timerid), sizeof(timerid)); 586} 587 588 589/* Delete a POSIX realtime timer */ 590int 591sys_timer_delete(struct lwp *l, void *v, register_t *retval) 592{ 593 struct sys_timer_delete_args /* { 594 syscallarg(timer_t) timerid; 595 } */ *uap = v; 596 struct proc *p = l->l_proc; 597 timer_t timerid; 598 struct ptimer *pt, *ptn; 599 int s; 600 601 timerid = SCARG(uap, timerid); 602 603 if ((p->p_timers == NULL) || 604 (timerid < 2) || (timerid >= TIMER_MAX) || 605 ((pt = p->p_timers->pts_timers[timerid]) == NULL)) 606 return (EINVAL); 607 608 if (pt->pt_type == CLOCK_REALTIME) 609 callout_stop(&pt->pt_ch); 610 else if (pt->pt_active) { 611 s = splclock(); 612 ptn = LIST_NEXT(pt, pt_list); 613 LIST_REMOVE(pt, pt_list); 614 for ( ; ptn; ptn = LIST_NEXT(ptn, pt_list)) 615 timeradd(&pt->pt_time.it_value, &ptn->pt_time.it_value, 616 &ptn->pt_time.it_value); 617 splx(s); 618 } 619 620 p->p_timers->pts_timers[timerid] = NULL; 621 pool_put(&ptimer_pool, pt); 622 623 return (0); 624} 625 626/* 627 * Set up the given timer. The value in pt->pt_time.it_value is taken 628 * to be an absolute time for CLOCK_REALTIME timers and a relative 629 * time for virtual timers. 630 * Must be called at splclock(). 631 */ 632void 633timer_settime(struct ptimer *pt) 634{ 635 struct ptimer *ptn, *pptn; 636 struct ptlist *ptl; 637 638 if (pt->pt_type == CLOCK_REALTIME) { 639 callout_stop(&pt->pt_ch); 640 if (timerisset(&pt->pt_time.it_value)) { 641 /* 642 * Don't need to check hzto() return value, here. 643 * callout_reset() does it for us. 644 */ 645 callout_reset(&pt->pt_ch, hzto(&pt->pt_time.it_value), 646 realtimerexpire, pt); 647 } 648 } else { 649 if (pt->pt_active) { 650 ptn = LIST_NEXT(pt, pt_list); 651 LIST_REMOVE(pt, pt_list); 652 for ( ; ptn; ptn = LIST_NEXT(ptn, pt_list)) 653 timeradd(&pt->pt_time.it_value, 654 &ptn->pt_time.it_value, 655 &ptn->pt_time.it_value); 656 } 657 if (timerisset(&pt->pt_time.it_value)) { 658 if (pt->pt_type == CLOCK_VIRTUAL) 659 ptl = &pt->pt_proc->p_timers->pts_virtual; 660 else 661 ptl = &pt->pt_proc->p_timers->pts_prof; 662 663 for (ptn = LIST_FIRST(ptl), pptn = NULL; 664 ptn && timercmp(&pt->pt_time.it_value, 665 &ptn->pt_time.it_value, >); 666 pptn = ptn, ptn = LIST_NEXT(ptn, pt_list)) 667 timersub(&pt->pt_time.it_value, 668 &ptn->pt_time.it_value, 669 &pt->pt_time.it_value); 670 671 if (pptn) 672 LIST_INSERT_AFTER(pptn, pt, pt_list); 673 else 674 LIST_INSERT_HEAD(ptl, pt, pt_list); 675 676 for ( ; ptn ; ptn = LIST_NEXT(ptn, pt_list)) 677 timersub(&ptn->pt_time.it_value, 678 &pt->pt_time.it_value, 679 &ptn->pt_time.it_value); 680 681 pt->pt_active = 1; 682 } else 683 pt->pt_active = 0; 684 } 685} 686 687void 688timer_gettime(struct ptimer *pt, struct itimerval *aitv) 689{ 690 struct ptimer *ptn; 691 692 *aitv = pt->pt_time; 693 if (pt->pt_type == CLOCK_REALTIME) { 694 /* 695 * Convert from absolute to relative time in .it_value 696 * part of real time timer. If time for real time 697 * timer has passed return 0, else return difference 698 * between current time and time for the timer to go 699 * off. 700 */ 701 if (timerisset(&aitv->it_value)) { 702 if (timercmp(&aitv->it_value, &time, <)) 703 timerclear(&aitv->it_value); 704 else 705 timersub(&aitv->it_value, &time, 706 &aitv->it_value); 707 } 708 } else if (pt->pt_active) { 709 if (pt->pt_type == CLOCK_VIRTUAL) 710 ptn = LIST_FIRST(&pt->pt_proc->p_timers->pts_virtual); 711 else 712 ptn = LIST_FIRST(&pt->pt_proc->p_timers->pts_prof); 713 for ( ; ptn && ptn != pt; ptn = LIST_NEXT(ptn, pt_list)) 714 timeradd(&aitv->it_value, 715 &ptn->pt_time.it_value, &aitv->it_value); 716 KASSERT(ptn != NULL); /* pt should be findable on the list */ 717 } else 718 timerclear(&aitv->it_value); 719} 720 721 722 723/* Set and arm a POSIX realtime timer */ 724int 725sys_timer_settime(struct lwp *l, void *v, register_t *retval) 726{ 727 struct sys_timer_settime_args /* { 728 syscallarg(timer_t) timerid; 729 syscallarg(int) flags; 730 syscallarg(const struct itimerspec *) value; 731 syscallarg(struct itimerspec *) ovalue; 732 } */ *uap = v; 733 struct proc *p = l->l_proc; 734 int error, s, timerid; 735 struct itimerval val, oval; 736 struct itimerspec value, ovalue; 737 struct ptimer *pt; 738 739 timerid = SCARG(uap, timerid); 740 741 if ((p->p_timers == NULL) || 742 (timerid < 2) || (timerid >= TIMER_MAX) || 743 ((pt = p->p_timers->pts_timers[timerid]) == NULL)) 744 return (EINVAL); 745 746 if ((error = copyin(SCARG(uap, value), &value, 747 sizeof(struct itimerspec))) != 0) 748 return (error); 749 750 TIMESPEC_TO_TIMEVAL(&val.it_value, &value.it_value); 751 TIMESPEC_TO_TIMEVAL(&val.it_interval, &value.it_interval); 752 if (itimerfix(&val.it_value) || itimerfix(&val.it_interval)) 753 return (EINVAL); 754 755 oval = pt->pt_time; 756 pt->pt_time = val; 757 758 s = splclock(); 759 /* 760 * If we've been passed a relative time for a realtime timer, 761 * convert it to absolute; if an absolute time for a virtual 762 * timer, convert it to relative and make sure we don't set it 763 * to zero, which would cancel the timer, or let it go 764 * negative, which would confuse the comparison tests. 765 */ 766 if (timerisset(&pt->pt_time.it_value)) { 767 if (pt->pt_type == CLOCK_REALTIME) { 768 if ((SCARG(uap, flags) & TIMER_ABSTIME) == 0) 769 timeradd(&pt->pt_time.it_value, &time, 770 &pt->pt_time.it_value); 771 } else { 772 if ((SCARG(uap, flags) & TIMER_ABSTIME) != 0) { 773 timersub(&pt->pt_time.it_value, &time, 774 &pt->pt_time.it_value); 775 if (!timerisset(&pt->pt_time.it_value) || 776 pt->pt_time.it_value.tv_sec < 0) { 777 pt->pt_time.it_value.tv_sec = 0; 778 pt->pt_time.it_value.tv_usec = 1; 779 } 780 } 781 } 782 } 783 784 timer_settime(pt); 785 splx(s); 786 787 if (SCARG(uap, ovalue)) { 788 TIMEVAL_TO_TIMESPEC(&oval.it_value, &ovalue.it_value); 789 TIMEVAL_TO_TIMESPEC(&oval.it_interval, &ovalue.it_interval); 790 return copyout(&ovalue, SCARG(uap, ovalue), 791 sizeof(struct itimerspec)); 792 } 793 794 return (0); 795} 796 797/* Return the time remaining until a POSIX timer fires. */ 798int 799sys_timer_gettime(struct lwp *l, void *v, register_t *retval) 800{ 801 struct sys_timer_gettime_args /* { 802 syscallarg(timer_t) timerid; 803 syscallarg(struct itimerspec *) value; 804 } */ *uap = v; 805 struct itimerval aitv; 806 struct itimerspec its; 807 struct proc *p = l->l_proc; 808 int s, timerid; 809 struct ptimer *pt; 810 811 timerid = SCARG(uap, timerid); 812 813 if ((p->p_timers == NULL) || 814 (timerid < 2) || (timerid >= TIMER_MAX) || 815 ((pt = p->p_timers->pts_timers[timerid]) == NULL)) 816 return (EINVAL); 817 818 s = splclock(); 819 timer_gettime(pt, &aitv); 820 splx(s); 821 822 TIMEVAL_TO_TIMESPEC(&aitv.it_interval, &its.it_interval); 823 TIMEVAL_TO_TIMESPEC(&aitv.it_value, &its.it_value); 824 825 return copyout(&its, SCARG(uap, value), sizeof(its)); 826} 827 828/* 829 * Return the count of the number of times a periodic timer expired 830 * while a notification was already pending. The counter is reset when 831 * a timer expires and a notification can be posted. 832 */ 833int 834sys_timer_getoverrun(struct lwp *l, void *v, register_t *retval) 835{ 836 struct sys_timer_getoverrun_args /* { 837 syscallarg(timer_t) timerid; 838 } */ *uap = v; 839 struct proc *p = l->l_proc; 840 int timerid; 841 struct ptimer *pt; 842 843 timerid = SCARG(uap, timerid); 844 845 if ((p->p_timers == NULL) || 846 (timerid < 2) || (timerid >= TIMER_MAX) || 847 ((pt = p->p_timers->pts_timers[timerid]) == NULL)) 848 return (EINVAL); 849 850 *retval = pt->pt_poverruns; 851 852 return (0); 853} 854 855/* Glue function that triggers an upcall; called from userret(). */ 856static void 857timerupcall(struct lwp *l, void *arg) 858{ 859 struct ptimers *pt = (struct ptimers *)arg; 860 unsigned int i, fired, done; 861 extern struct pool siginfo_pool; /* XXX Ew. */ 862 863 KDASSERT(l->l_proc->p_sa); 864 /* Bail out if we do not own the virtual processor */ 865 if (l->l_savp->savp_lwp != l) 866 return ; 867 868 KERNEL_PROC_LOCK(l); 869 870 fired = pt->pts_fired; 871 done = 0; 872 while ((i = ffs(fired)) != 0) { 873 siginfo_t *si; 874 int mask = 1 << --i; 875 int f; 876 877 f = l->l_flag & L_SA; 878 l->l_flag &= ~L_SA; 879 si = pool_get(&siginfo_pool, PR_WAITOK); 880 si->_info = pt->pts_timers[i]->pt_info.ksi_info; 881 if (sa_upcall(l, SA_UPCALL_SIGEV | SA_UPCALL_DEFER, NULL, l, 882 sizeof(*si), si) != 0) { 883 pool_put(&siginfo_pool, si); 884 /* XXX What do we do here?? */ 885 } else 886 done |= mask; 887 fired &= ~mask; 888 l->l_flag |= f; 889 } 890 pt->pts_fired &= ~done; 891 if (pt->pts_fired == 0) 892 l->l_proc->p_userret = NULL; 893 894 KERNEL_PROC_UNLOCK(l); 895} 896 897 898/* 899 * Real interval timer expired: 900 * send process whose timer expired an alarm signal. 901 * If time is not set up to reload, then just return. 902 * Else compute next time timer should go off which is > current time. 903 * This is where delay in processing this timeout causes multiple 904 * SIGALRM calls to be compressed into one. 905 */ 906void 907realtimerexpire(void *arg) 908{ 909 struct ptimer *pt; 910 int s; 911 912 pt = (struct ptimer *)arg; 913 914 itimerfire(pt); 915 916 if (!timerisset(&pt->pt_time.it_interval)) { 917 timerclear(&pt->pt_time.it_value); 918 return; 919 } 920 for (;;) { 921 s = splclock(); 922 timeradd(&pt->pt_time.it_value, 923 &pt->pt_time.it_interval, &pt->pt_time.it_value); 924 if (timercmp(&pt->pt_time.it_value, &time, >)) { 925 /* 926 * Don't need to check hzto() return value, here. 927 * callout_reset() does it for us. 928 */ 929 callout_reset(&pt->pt_ch, hzto(&pt->pt_time.it_value), 930 realtimerexpire, pt); 931 splx(s); 932 return; 933 } 934 splx(s); 935 pt->pt_overruns++; 936 } 937} 938 939/* BSD routine to get the value of an interval timer. */ 940/* ARGSUSED */ 941int 942sys_getitimer(struct lwp *l, void *v, register_t *retval) 943{ 944 struct sys_getitimer_args /* { 945 syscallarg(int) which; 946 syscallarg(struct itimerval *) itv; 947 } */ *uap = v; 948 struct proc *p = l->l_proc; 949 struct itimerval aitv; 950 int s, which; 951 952 which = SCARG(uap, which); 953 954 if ((u_int)which > ITIMER_PROF) 955 return (EINVAL); 956 957 if ((p->p_timers == NULL) || (p->p_timers->pts_timers[which] == NULL)){ 958 timerclear(&aitv.it_value); 959 timerclear(&aitv.it_interval); 960 } else { 961 s = splclock(); 962 timer_gettime(p->p_timers->pts_timers[which], &aitv); 963 splx(s); 964 } 965 966 return (copyout(&aitv, SCARG(uap, itv), sizeof(struct itimerval))); 967 968} 969 970/* BSD routine to set/arm an interval timer. */ 971/* ARGSUSED */ 972int 973sys_setitimer(struct lwp *l, void *v, register_t *retval) 974{ 975 struct sys_setitimer_args /* { 976 syscallarg(int) which; 977 syscallarg(const struct itimerval *) itv; 978 syscallarg(struct itimerval *) oitv; 979 } */ *uap = v; 980 struct proc *p = l->l_proc; 981 int which = SCARG(uap, which); 982 struct sys_getitimer_args getargs; 983 struct itimerval aitv; 984 const struct itimerval *itvp; 985 struct ptimer *pt; 986 int s, error; 987 988 if ((u_int)which > ITIMER_PROF) 989 return (EINVAL); 990 itvp = SCARG(uap, itv); 991 if (itvp && 992 (error = copyin(itvp, &aitv, sizeof(struct itimerval)) != 0)) 993 return (error); 994 if (SCARG(uap, oitv) != NULL) { 995 SCARG(&getargs, which) = which; 996 SCARG(&getargs, itv) = SCARG(uap, oitv); 997 if ((error = sys_getitimer(l, &getargs, retval)) != 0) 998 return (error); 999 } 1000 if (itvp == 0) 1001 return (0); 1002 if (itimerfix(&aitv.it_value) || itimerfix(&aitv.it_interval)) 1003 return (EINVAL); 1004 1005 /* 1006 * Don't bother allocating data structures if the process just 1007 * wants to clear the timer. 1008 */ 1009 if (!timerisset(&aitv.it_value) && 1010 ((p->p_timers == NULL) ||(p->p_timers->pts_timers[which] == NULL))) 1011 return (0); 1012 1013 if (p->p_timers == NULL) 1014 timers_alloc(p); 1015 if (p->p_timers->pts_timers[which] == NULL) { 1016 pt = pool_get(&ptimer_pool, PR_WAITOK); 1017 pt->pt_ev.sigev_notify = SIGEV_SIGNAL; 1018 pt->pt_ev.sigev_value.sival_int = which; 1019 pt->pt_overruns = 0; 1020 pt->pt_proc = p; 1021 pt->pt_type = which; 1022 pt->pt_entry = which; 1023 switch (which) { 1024 case ITIMER_REAL: 1025 callout_init(&pt->pt_ch); 1026 pt->pt_ev.sigev_signo = SIGALRM; 1027 break; 1028 case ITIMER_VIRTUAL: 1029 pt->pt_active = 0; 1030 pt->pt_ev.sigev_signo = SIGVTALRM; 1031 break; 1032 case ITIMER_PROF: 1033 pt->pt_active = 0; 1034 pt->pt_ev.sigev_signo = SIGPROF; 1035 break; 1036 } 1037 } else 1038 pt = p->p_timers->pts_timers[which]; 1039 1040 pt->pt_time = aitv; 1041 p->p_timers->pts_timers[which] = pt; 1042 1043 s = splclock(); 1044 if ((which == ITIMER_REAL) && timerisset(&pt->pt_time.it_value)) { 1045 /* Convert to absolute time */ 1046 timeradd(&pt->pt_time.it_value, &time, &pt->pt_time.it_value); 1047 } 1048 timer_settime(pt); 1049 splx(s); 1050 1051 return (0); 1052} 1053 1054/* Utility routines to manage the array of pointers to timers. */ 1055void 1056timers_alloc(struct proc *p) 1057{ 1058 int i; 1059 struct ptimers *pts; 1060 1061 pts = malloc(sizeof (struct ptimers), M_SUBPROC, 0); 1062 LIST_INIT(&pts->pts_virtual); 1063 LIST_INIT(&pts->pts_prof); 1064 for (i = 0; i < TIMER_MAX; i++) 1065 pts->pts_timers[i] = NULL; 1066 pts->pts_fired = 0; 1067 p->p_timers = pts; 1068} 1069 1070/* 1071 * Clean up the per-process timers. If "which" is set to TIMERS_ALL, 1072 * then clean up all timers and free all the data structures. If 1073 * "which" is set to TIMERS_POSIX, only clean up the timers allocated 1074 * by timer_create(), not the BSD setitimer() timers, and only free the 1075 * structure if none of those remain. 1076 */ 1077void 1078timers_free(struct proc *p, int which) 1079{ 1080 int i, s; 1081 struct ptimers *pts; 1082 struct ptimer *pt, *ptn; 1083 struct timeval tv; 1084 1085 if (p->p_timers) { 1086 pts = p->p_timers; 1087 if (which == TIMERS_ALL) 1088 i = 0; 1089 else { 1090 s = splclock(); 1091 timerclear(&tv); 1092 for (ptn = LIST_FIRST(&p->p_timers->pts_virtual); 1093 ptn && ptn != pts->pts_timers[ITIMER_VIRTUAL]; 1094 ptn = LIST_NEXT(ptn, pt_list)) 1095 timeradd(&tv, &ptn->pt_time.it_value, &tv); 1096 LIST_FIRST(&p->p_timers->pts_virtual) = NULL; 1097 if (ptn) { 1098 timeradd(&tv, &ptn->pt_time.it_value, 1099 &ptn->pt_time.it_value); 1100 LIST_INSERT_HEAD(&p->p_timers->pts_virtual, 1101 ptn, pt_list); 1102 } 1103 1104 timerclear(&tv); 1105 for (ptn = LIST_FIRST(&p->p_timers->pts_prof); 1106 ptn && ptn != pts->pts_timers[ITIMER_PROF]; 1107 ptn = LIST_NEXT(ptn, pt_list)) 1108 timeradd(&tv, &ptn->pt_time.it_value, &tv); 1109 LIST_FIRST(&p->p_timers->pts_prof) = NULL; 1110 if (ptn) { 1111 timeradd(&tv, &ptn->pt_time.it_value, 1112 &ptn->pt_time.it_value); 1113 LIST_INSERT_HEAD(&p->p_timers->pts_prof, ptn, 1114 pt_list); 1115 } 1116 splx(s); 1117 i = 3; 1118 } 1119 for ( ; i < TIMER_MAX; i++) 1120 if ((pt = pts->pts_timers[i]) != NULL) { 1121 if (pt->pt_type == CLOCK_REALTIME) 1122 callout_stop(&pt->pt_ch); 1123 pts->pts_timers[i] = NULL; 1124 pool_put(&ptimer_pool, pt); 1125 } 1126 if ((pts->pts_timers[0] == NULL) && 1127 (pts->pts_timers[1] == NULL) && 1128 (pts->pts_timers[2] == NULL)) { 1129 p->p_timers = NULL; 1130 free(pts, M_SUBPROC); 1131 } 1132 } 1133} 1134 1135/* 1136 * Check that a proposed value to load into the .it_value or 1137 * .it_interval part of an interval timer is acceptable, and 1138 * fix it to have at least minimal value (i.e. if it is less 1139 * than the resolution of the clock, round it up.) 1140 */ 1141int 1142itimerfix(struct timeval *tv) 1143{ 1144 1145 if (tv->tv_sec < 0 || tv->tv_usec < 0 || tv->tv_usec >= 1000000) 1146 return (EINVAL); 1147 if (tv->tv_sec == 0 && tv->tv_usec != 0 && tv->tv_usec < tick) 1148 tv->tv_usec = tick; 1149 return (0); 1150} 1151 1152/* 1153 * Decrement an interval timer by a specified number 1154 * of microseconds, which must be less than a second, 1155 * i.e. < 1000000. If the timer expires, then reload 1156 * it. In this case, carry over (usec - old value) to 1157 * reduce the value reloaded into the timer so that 1158 * the timer does not drift. This routine assumes 1159 * that it is called in a context where the timers 1160 * on which it is operating cannot change in value. 1161 */ 1162int 1163itimerdecr(struct ptimer *pt, int usec) 1164{ 1165 struct itimerval *itp; 1166 1167 itp = &pt->pt_time; 1168 if (itp->it_value.tv_usec < usec) { 1169 if (itp->it_value.tv_sec == 0) { 1170 /* expired, and already in next interval */ 1171 usec -= itp->it_value.tv_usec; 1172 goto expire; 1173 } 1174 itp->it_value.tv_usec += 1000000; 1175 itp->it_value.tv_sec--; 1176 } 1177 itp->it_value.tv_usec -= usec; 1178 usec = 0; 1179 if (timerisset(&itp->it_value)) 1180 return (1); 1181 /* expired, exactly at end of interval */ 1182expire: 1183 if (timerisset(&itp->it_interval)) { 1184 itp->it_value = itp->it_interval; 1185 itp->it_value.tv_usec -= usec; 1186 if (itp->it_value.tv_usec < 0) { 1187 itp->it_value.tv_usec += 1000000; 1188 itp->it_value.tv_sec--; 1189 } 1190 timer_settime(pt); 1191 } else 1192 itp->it_value.tv_usec = 0; /* sec is already 0 */ 1193 return (0); 1194} 1195 1196void 1197itimerfire(struct ptimer *pt) 1198{ 1199 struct proc *p = pt->pt_proc; 1200 struct sadata_vp *vp; 1201 int s; 1202 unsigned int i; 1203 1204 if (pt->pt_ev.sigev_notify == SIGEV_SIGNAL) { 1205 /* 1206 * No RT signal infrastructure exists at this time; 1207 * just post the signal number and throw away the 1208 * value. 1209 */ 1210 if (sigismember(&p->p_sigctx.ps_siglist, pt->pt_ev.sigev_signo)) 1211 pt->pt_overruns++; 1212 else { 1213 ksiginfo_t ksi; 1214 (void)memset(&ksi, 0, sizeof(ksi)); 1215 ksi.ksi_signo = pt->pt_ev.sigev_signo; 1216 ksi.ksi_code = SI_TIMER; 1217 ksi.ksi_sigval = pt->pt_ev.sigev_value; 1218 pt->pt_poverruns = pt->pt_overruns; 1219 pt->pt_overruns = 0; 1220 kpsignal(p, &ksi, NULL); 1221 } 1222 } else if (pt->pt_ev.sigev_notify == SIGEV_SA && (p->p_flag & P_SA)) { 1223 /* Cause the process to generate an upcall when it returns. */ 1224 1225 if (p->p_userret == NULL) { 1226 /* 1227 * XXX stop signals can be processed inside tsleep, 1228 * which can be inside sa_yield's inner loop, which 1229 * makes testing for sa_idle alone insuffucent to 1230 * determine if we really should call setrunnable. 1231 */ 1232 pt->pt_poverruns = pt->pt_overruns; 1233 pt->pt_overruns = 0; 1234 i = 1 << pt->pt_entry; 1235 p->p_timers->pts_fired = i; 1236 p->p_userret = timerupcall; 1237 p->p_userret_arg = p->p_timers; 1238 1239 SCHED_LOCK(s); 1240 SLIST_FOREACH(vp, &p->p_sa->sa_vps, savp_next) { 1241 if (vp->savp_lwp->l_flag & L_SA_IDLE) { 1242 vp->savp_lwp->l_flag &= ~L_SA_IDLE; 1243 sched_wakeup(vp->savp_lwp); 1244 break; 1245 } 1246 } 1247 SCHED_UNLOCK(s); 1248 } else if (p->p_userret == timerupcall) { 1249 i = 1 << pt->pt_entry; 1250 if ((p->p_timers->pts_fired & i) == 0) { 1251 pt->pt_poverruns = pt->pt_overruns; 1252 pt->pt_overruns = 0; 1253 p->p_timers->pts_fired |= i; 1254 } else 1255 pt->pt_overruns++; 1256 } else { 1257 pt->pt_overruns++; 1258 if ((p->p_flag & P_WEXIT) == 0) 1259 printf("itimerfire(%d): overrun %d on timer %x (userret is %p)\n", 1260 p->p_pid, pt->pt_overruns, 1261 pt->pt_ev.sigev_value.sival_int, 1262 p->p_userret); 1263 } 1264 } 1265 1266} 1267 1268/* 1269 * ratecheck(): simple time-based rate-limit checking. see ratecheck(9) 1270 * for usage and rationale. 1271 */ 1272int 1273ratecheck(struct timeval *lasttime, const struct timeval *mininterval) 1274{ 1275 struct timeval tv, delta; 1276 int s, rv = 0; 1277 1278 s = splclock(); 1279 tv = mono_time; 1280 splx(s); 1281 1282 timersub(&tv, lasttime, &delta); 1283 1284 /* 1285 * check for 0,0 is so that the message will be seen at least once, 1286 * even if interval is huge. 1287 */ 1288 if (timercmp(&delta, mininterval, >=) || 1289 (lasttime->tv_sec == 0 && lasttime->tv_usec == 0)) { 1290 *lasttime = tv; 1291 rv = 1; 1292 } 1293 1294 return (rv); 1295} 1296 1297/* 1298 * ppsratecheck(): packets (or events) per second limitation. 1299 */ 1300int 1301ppsratecheck(struct timeval *lasttime, int *curpps, int maxpps) 1302{ 1303 struct timeval tv, delta; 1304 int s, rv; 1305 1306 s = splclock(); 1307 tv = mono_time; 1308 splx(s); 1309 1310 timersub(&tv, lasttime, &delta); 1311 1312 /* 1313 * check for 0,0 is so that the message will be seen at least once. 1314 * if more than one second have passed since the last update of 1315 * lasttime, reset the counter. 1316 * 1317 * we do increment *curpps even in *curpps < maxpps case, as some may 1318 * try to use *curpps for stat purposes as well. 1319 */ 1320 if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) || 1321 delta.tv_sec >= 1) { 1322 *lasttime = tv; 1323 *curpps = 0; 1324 } 1325 if (maxpps < 0) 1326 rv = 1; 1327 else if (*curpps < maxpps) 1328 rv = 1; 1329 else 1330 rv = 0; 1331 1332#if 1 /*DIAGNOSTIC?*/ 1333 /* be careful about wrap-around */ 1334 if (*curpps + 1 > *curpps) 1335 *curpps = *curpps + 1; 1336#else 1337 /* 1338 * assume that there's not too many calls to this function. 1339 * not sure if the assumption holds, as it depends on *caller's* 1340 * behavior, not the behavior of this function. 1341 * IMHO it is wrong to make assumption on the caller's behavior, 1342 * so the above #if is #if 1, not #ifdef DIAGNOSTIC. 1343 */ 1344 *curpps = *curpps + 1; 1345#endif 1346 1347 return (rv); 1348} 1349