The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_time.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (c) 1982, 1986, 1989, 1993
    5  *      The Regents of the University of California.  All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. Neither the name of the University nor the names of its contributors
   16  *    may be used to endorse or promote products derived from this software
   17  *    without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  *
   31  *      @(#)kern_time.c 8.1 (Berkeley) 6/10/93
   32  */
   33 
   34 #include <sys/cdefs.h>
   35 __FBSDID("$FreeBSD$");
   36 
   37 #include "opt_ktrace.h"
   38 
   39 #include <sys/param.h>
   40 #include <sys/systm.h>
   41 #include <sys/limits.h>
   42 #include <sys/clock.h>
   43 #include <sys/lock.h>
   44 #include <sys/mutex.h>
   45 #include <sys/sysproto.h>
   46 #include <sys/resourcevar.h>
   47 #include <sys/signalvar.h>
   48 #include <sys/kernel.h>
   49 #include <sys/sleepqueue.h>
   50 #include <sys/syscallsubr.h>
   51 #include <sys/sysctl.h>
   52 #include <sys/sysent.h>
   53 #include <sys/priv.h>
   54 #include <sys/proc.h>
   55 #include <sys/posix4.h>
   56 #include <sys/time.h>
   57 #include <sys/timers.h>
   58 #include <sys/timetc.h>
   59 #include <sys/vnode.h>
   60 #ifdef KTRACE
   61 #include <sys/ktrace.h>
   62 #endif
   63 
   64 #include <vm/vm.h>
   65 #include <vm/vm_extern.h>
   66 
   67 #define MAX_CLOCKS      (CLOCK_MONOTONIC+1)
   68 #define CPUCLOCK_BIT            0x80000000
   69 #define CPUCLOCK_PROCESS_BIT    0x40000000
   70 #define CPUCLOCK_ID_MASK        (~(CPUCLOCK_BIT|CPUCLOCK_PROCESS_BIT))
   71 #define MAKE_THREAD_CPUCLOCK(tid)       (CPUCLOCK_BIT|(tid))
   72 #define MAKE_PROCESS_CPUCLOCK(pid)      \
   73         (CPUCLOCK_BIT|CPUCLOCK_PROCESS_BIT|(pid))
   74 
   75 static struct kclock    posix_clocks[MAX_CLOCKS];
   76 static uma_zone_t       itimer_zone = NULL;
   77 
   78 /*
   79  * Time of day and interval timer support.
   80  *
   81  * These routines provide the kernel entry points to get and set
   82  * the time-of-day and per-process interval timers.  Subroutines
   83  * here provide support for adding and subtracting timeval structures
   84  * and decrementing interval timers, optionally reloading the interval
   85  * timers when they expire.
   86  */
   87 
   88 static int      settime(struct thread *, struct timeval *);
   89 static void     timevalfix(struct timeval *);
   90 static int      user_clock_nanosleep(struct thread *td, clockid_t clock_id,
   91                     int flags, const struct timespec *ua_rqtp,
   92                     struct timespec *ua_rmtp);
   93 
   94 static void     itimer_start(void);
   95 static int      itimer_init(void *, int, int);
   96 static void     itimer_fini(void *, int);
   97 static void     itimer_enter(struct itimer *);
   98 static void     itimer_leave(struct itimer *);
   99 static struct itimer *itimer_find(struct proc *, int);
  100 static void     itimers_alloc(struct proc *);
  101 static int      realtimer_create(struct itimer *);
  102 static int      realtimer_gettime(struct itimer *, struct itimerspec *);
  103 static int      realtimer_settime(struct itimer *, int,
  104                         struct itimerspec *, struct itimerspec *);
  105 static int      realtimer_delete(struct itimer *);
  106 static void     realtimer_clocktime(clockid_t, struct timespec *);
  107 static void     realtimer_expire(void *);
  108 
  109 static int      register_posix_clock(int, const struct kclock *);
  110 void            itimer_fire(struct itimer *it);
  111 int             itimespecfix(struct timespec *ts);
  112 
  113 #define CLOCK_CALL(clock, call, arglist)                \
  114         ((*posix_clocks[clock].call) arglist)
  115 
  116 SYSINIT(posix_timer, SI_SUB_P1003_1B, SI_ORDER_FIRST+4, itimer_start, NULL);
  117 
  118 
  119 static int
  120 settime(struct thread *td, struct timeval *tv)
  121 {
  122         struct timeval delta, tv1, tv2;
  123         static struct timeval maxtime, laststep;
  124         struct timespec ts;
  125 
  126         microtime(&tv1);
  127         delta = *tv;
  128         timevalsub(&delta, &tv1);
  129 
  130         /*
  131          * If the system is secure, we do not allow the time to be 
  132          * set to a value earlier than 1 second less than the highest
  133          * time we have yet seen. The worst a miscreant can do in
  134          * this circumstance is "freeze" time. He couldn't go
  135          * back to the past.
  136          *
  137          * We similarly do not allow the clock to be stepped more
  138          * than one second, nor more than once per second. This allows
  139          * a miscreant to make the clock march double-time, but no worse.
  140          */
  141         if (securelevel_gt(td->td_ucred, 1) != 0) {
  142                 if (delta.tv_sec < 0 || delta.tv_usec < 0) {
  143                         /*
  144                          * Update maxtime to latest time we've seen.
  145                          */
  146                         if (tv1.tv_sec > maxtime.tv_sec)
  147                                 maxtime = tv1;
  148                         tv2 = *tv;
  149                         timevalsub(&tv2, &maxtime);
  150                         if (tv2.tv_sec < -1) {
  151                                 tv->tv_sec = maxtime.tv_sec - 1;
  152                                 printf("Time adjustment clamped to -1 second\n");
  153                         }
  154                 } else {
  155                         if (tv1.tv_sec == laststep.tv_sec)
  156                                 return (EPERM);
  157                         if (delta.tv_sec > 1) {
  158                                 tv->tv_sec = tv1.tv_sec + 1;
  159                                 printf("Time adjustment clamped to +1 second\n");
  160                         }
  161                         laststep = *tv;
  162                 }
  163         }
  164 
  165         ts.tv_sec = tv->tv_sec;
  166         ts.tv_nsec = tv->tv_usec * 1000;
  167         tc_setclock(&ts);
  168         resettodr();
  169         return (0);
  170 }
  171 
  172 #ifndef _SYS_SYSPROTO_H_
  173 struct clock_getcpuclockid2_args {
  174         id_t id;
  175         int which,
  176         clockid_t *clock_id;
  177 };
  178 #endif
  179 /* ARGSUSED */
  180 int
  181 sys_clock_getcpuclockid2(struct thread *td, struct clock_getcpuclockid2_args *uap)
  182 {
  183         clockid_t clk_id;
  184         int error;
  185 
  186         error = kern_clock_getcpuclockid2(td, uap->id, uap->which, &clk_id);
  187         if (error == 0)
  188                 error = copyout(&clk_id, uap->clock_id, sizeof(clockid_t));
  189         return (error);
  190 }
  191 
  192 int
  193 kern_clock_getcpuclockid2(struct thread *td, id_t id, int which,
  194     clockid_t *clk_id)
  195 {
  196         struct proc *p;
  197         pid_t pid;
  198         lwpid_t tid;
  199         int error;
  200 
  201         switch (which) {
  202         case CPUCLOCK_WHICH_PID:
  203                 if (id != 0) {
  204                         error = pget(id, PGET_CANSEE | PGET_NOTID, &p);
  205                         if (error != 0)
  206                                 return (error);
  207                         PROC_UNLOCK(p);
  208                         pid = id;
  209                 } else {
  210                         pid = td->td_proc->p_pid;
  211                 }
  212                 *clk_id = MAKE_PROCESS_CPUCLOCK(pid);
  213                 return (0);
  214         case CPUCLOCK_WHICH_TID:
  215                 tid = id == 0 ? td->td_tid : id;
  216                 *clk_id = MAKE_THREAD_CPUCLOCK(tid);
  217                 return (0);
  218         default:
  219                 return (EINVAL);
  220         }
  221 }
  222 
  223 #ifndef _SYS_SYSPROTO_H_
  224 struct clock_gettime_args {
  225         clockid_t clock_id;
  226         struct  timespec *tp;
  227 };
  228 #endif
  229 /* ARGSUSED */
  230 int
  231 sys_clock_gettime(struct thread *td, struct clock_gettime_args *uap)
  232 {
  233         struct timespec ats;
  234         int error;
  235 
  236         error = kern_clock_gettime(td, uap->clock_id, &ats);
  237         if (error == 0)
  238                 error = copyout(&ats, uap->tp, sizeof(ats));
  239 
  240         return (error);
  241 }
  242 
  243 static inline void
  244 cputick2timespec(uint64_t runtime, struct timespec *ats)
  245 {
  246         runtime = cputick2usec(runtime);
  247         ats->tv_sec = runtime / 1000000;
  248         ats->tv_nsec = runtime % 1000000 * 1000;
  249 }
  250 
  251 void
  252 kern_thread_cputime(struct thread *targettd, struct timespec *ats)
  253 {
  254         uint64_t runtime, curtime, switchtime;
  255 
  256         if (targettd == NULL) { /* current thread */
  257                 critical_enter();
  258                 switchtime = PCPU_GET(switchtime);
  259                 curtime = cpu_ticks();
  260                 runtime = curthread->td_runtime;
  261                 critical_exit();
  262                 runtime += curtime - switchtime;
  263         } else {
  264                 PROC_LOCK_ASSERT(targettd->td_proc, MA_OWNED);
  265                 thread_lock(targettd);
  266                 runtime = targettd->td_runtime;
  267                 thread_unlock(targettd);
  268         }
  269         cputick2timespec(runtime, ats);
  270 }
  271 
  272 void
  273 kern_process_cputime(struct proc *targetp, struct timespec *ats)
  274 {
  275         uint64_t runtime;
  276         struct rusage ru;
  277 
  278         PROC_LOCK_ASSERT(targetp, MA_OWNED);
  279         PROC_STATLOCK(targetp);
  280         rufetch(targetp, &ru);
  281         runtime = targetp->p_rux.rux_runtime;
  282         if (curthread->td_proc == targetp)
  283                 runtime += cpu_ticks() - PCPU_GET(switchtime);
  284         PROC_STATUNLOCK(targetp);
  285         cputick2timespec(runtime, ats);
  286 }
  287 
  288 static int
  289 get_cputime(struct thread *td, clockid_t clock_id, struct timespec *ats)
  290 {
  291         struct proc *p, *p2;
  292         struct thread *td2;
  293         lwpid_t tid;
  294         pid_t pid;
  295         int error;
  296 
  297         p = td->td_proc;
  298         if ((clock_id & CPUCLOCK_PROCESS_BIT) == 0) {
  299                 tid = clock_id & CPUCLOCK_ID_MASK;
  300                 td2 = tdfind(tid, p->p_pid);
  301                 if (td2 == NULL)
  302                         return (EINVAL);
  303                 kern_thread_cputime(td2, ats);
  304                 PROC_UNLOCK(td2->td_proc);
  305         } else {
  306                 pid = clock_id & CPUCLOCK_ID_MASK;
  307                 error = pget(pid, PGET_CANSEE, &p2);
  308                 if (error != 0)
  309                         return (EINVAL);
  310                 kern_process_cputime(p2, ats);
  311                 PROC_UNLOCK(p2);
  312         }
  313         return (0);
  314 }
  315 
  316 int
  317 kern_clock_gettime(struct thread *td, clockid_t clock_id, struct timespec *ats)
  318 {
  319         struct timeval sys, user;
  320         struct proc *p;
  321 
  322         p = td->td_proc;
  323         switch (clock_id) {
  324         case CLOCK_REALTIME:            /* Default to precise. */
  325         case CLOCK_REALTIME_PRECISE:
  326                 nanotime(ats);
  327                 break;
  328         case CLOCK_REALTIME_FAST:
  329                 getnanotime(ats);
  330                 break;
  331         case CLOCK_VIRTUAL:
  332                 PROC_LOCK(p);
  333                 PROC_STATLOCK(p);
  334                 calcru(p, &user, &sys);
  335                 PROC_STATUNLOCK(p);
  336                 PROC_UNLOCK(p);
  337                 TIMEVAL_TO_TIMESPEC(&user, ats);
  338                 break;
  339         case CLOCK_PROF:
  340                 PROC_LOCK(p);
  341                 PROC_STATLOCK(p);
  342                 calcru(p, &user, &sys);
  343                 PROC_STATUNLOCK(p);
  344                 PROC_UNLOCK(p);
  345                 timevaladd(&user, &sys);
  346                 TIMEVAL_TO_TIMESPEC(&user, ats);
  347                 break;
  348         case CLOCK_MONOTONIC:           /* Default to precise. */
  349         case CLOCK_MONOTONIC_PRECISE:
  350         case CLOCK_UPTIME:
  351         case CLOCK_UPTIME_PRECISE:
  352                 nanouptime(ats);
  353                 break;
  354         case CLOCK_UPTIME_FAST:
  355         case CLOCK_MONOTONIC_FAST:
  356                 getnanouptime(ats);
  357                 break;
  358         case CLOCK_SECOND:
  359                 ats->tv_sec = time_second;
  360                 ats->tv_nsec = 0;
  361                 break;
  362         case CLOCK_THREAD_CPUTIME_ID:
  363                 kern_thread_cputime(NULL, ats);
  364                 break;
  365         case CLOCK_PROCESS_CPUTIME_ID:
  366                 PROC_LOCK(p);
  367                 kern_process_cputime(p, ats);
  368                 PROC_UNLOCK(p);
  369                 break;
  370         default:
  371                 if ((int)clock_id >= 0)
  372                         return (EINVAL);
  373                 return (get_cputime(td, clock_id, ats));
  374         }
  375         return (0);
  376 }
  377 
  378 #ifndef _SYS_SYSPROTO_H_
  379 struct clock_settime_args {
  380         clockid_t clock_id;
  381         const struct    timespec *tp;
  382 };
  383 #endif
  384 /* ARGSUSED */
  385 int
  386 sys_clock_settime(struct thread *td, struct clock_settime_args *uap)
  387 {
  388         struct timespec ats;
  389         int error;
  390 
  391         if ((error = copyin(uap->tp, &ats, sizeof(ats))) != 0)
  392                 return (error);
  393         return (kern_clock_settime(td, uap->clock_id, &ats));
  394 }
  395 
  396 static int allow_insane_settime = 0;
  397 SYSCTL_INT(_debug, OID_AUTO, allow_insane_settime, CTLFLAG_RWTUN,
  398     &allow_insane_settime, 0,
  399     "do not perform possibly restrictive checks on settime(2) args");
  400 
  401 int
  402 kern_clock_settime(struct thread *td, clockid_t clock_id, struct timespec *ats)
  403 {
  404         struct timeval atv;
  405         int error;
  406 
  407         if ((error = priv_check(td, PRIV_CLOCK_SETTIME)) != 0)
  408                 return (error);
  409         if (clock_id != CLOCK_REALTIME)
  410                 return (EINVAL);
  411         if (ats->tv_nsec < 0 || ats->tv_nsec >= 1000000000 ||
  412             ats->tv_sec < 0)
  413                 return (EINVAL);
  414         if (!allow_insane_settime && ats->tv_sec > 8000ULL * 365 * 24 * 60 * 60)
  415                 return (EINVAL);
  416         /* XXX Don't convert nsec->usec and back */
  417         TIMESPEC_TO_TIMEVAL(&atv, ats);
  418         error = settime(td, &atv);
  419         return (error);
  420 }
  421 
  422 #ifndef _SYS_SYSPROTO_H_
  423 struct clock_getres_args {
  424         clockid_t clock_id;
  425         struct  timespec *tp;
  426 };
  427 #endif
  428 int
  429 sys_clock_getres(struct thread *td, struct clock_getres_args *uap)
  430 {
  431         struct timespec ts;
  432         int error;
  433 
  434         if (uap->tp == NULL)
  435                 return (0);
  436 
  437         error = kern_clock_getres(td, uap->clock_id, &ts);
  438         if (error == 0)
  439                 error = copyout(&ts, uap->tp, sizeof(ts));
  440         return (error);
  441 }
  442 
  443 int
  444 kern_clock_getres(struct thread *td, clockid_t clock_id, struct timespec *ts)
  445 {
  446 
  447         ts->tv_sec = 0;
  448         switch (clock_id) {
  449         case CLOCK_REALTIME:
  450         case CLOCK_REALTIME_FAST:
  451         case CLOCK_REALTIME_PRECISE:
  452         case CLOCK_MONOTONIC:
  453         case CLOCK_MONOTONIC_FAST:
  454         case CLOCK_MONOTONIC_PRECISE:
  455         case CLOCK_UPTIME:
  456         case CLOCK_UPTIME_FAST:
  457         case CLOCK_UPTIME_PRECISE:
  458                 /*
  459                  * Round up the result of the division cheaply by adding 1.
  460                  * Rounding up is especially important if rounding down
  461                  * would give 0.  Perfect rounding is unimportant.
  462                  */
  463                 ts->tv_nsec = 1000000000 / tc_getfrequency() + 1;
  464                 break;
  465         case CLOCK_VIRTUAL:
  466         case CLOCK_PROF:
  467                 /* Accurately round up here because we can do so cheaply. */
  468                 ts->tv_nsec = howmany(1000000000, hz);
  469                 break;
  470         case CLOCK_SECOND:
  471                 ts->tv_sec = 1;
  472                 ts->tv_nsec = 0;
  473                 break;
  474         case CLOCK_THREAD_CPUTIME_ID:
  475         case CLOCK_PROCESS_CPUTIME_ID:
  476         cputime:
  477                 /* sync with cputick2usec */
  478                 ts->tv_nsec = 1000000 / cpu_tickrate();
  479                 if (ts->tv_nsec == 0)
  480                         ts->tv_nsec = 1000;
  481                 break;
  482         default:
  483                 if ((int)clock_id < 0)
  484                         goto cputime;
  485                 return (EINVAL);
  486         }
  487         return (0);
  488 }
  489 
  490 int
  491 kern_nanosleep(struct thread *td, struct timespec *rqt, struct timespec *rmt)
  492 {
  493 
  494         return (kern_clock_nanosleep(td, CLOCK_REALTIME, TIMER_RELTIME, rqt,
  495             rmt));
  496 }
  497 
  498 static uint8_t nanowait[MAXCPU];
  499 
  500 int
  501 kern_clock_nanosleep(struct thread *td, clockid_t clock_id, int flags,
  502     const struct timespec *rqt, struct timespec *rmt)
  503 {
  504         struct timespec ts, now;
  505         sbintime_t sbt, sbtt, prec, tmp;
  506         time_t over;
  507         int error;
  508         bool is_abs_real;
  509 
  510         if (rqt->tv_nsec < 0 || rqt->tv_nsec >= 1000000000)
  511                 return (EINVAL);
  512         if ((flags & ~TIMER_ABSTIME) != 0)
  513                 return (EINVAL);
  514         switch (clock_id) {
  515         case CLOCK_REALTIME:
  516         case CLOCK_REALTIME_PRECISE:
  517         case CLOCK_REALTIME_FAST:
  518         case CLOCK_SECOND:
  519                 is_abs_real = (flags & TIMER_ABSTIME) != 0;
  520                 break;
  521         case CLOCK_MONOTONIC:
  522         case CLOCK_MONOTONIC_PRECISE:
  523         case CLOCK_MONOTONIC_FAST:
  524         case CLOCK_UPTIME:
  525         case CLOCK_UPTIME_PRECISE:
  526         case CLOCK_UPTIME_FAST:
  527                 is_abs_real = false;
  528                 break;
  529         case CLOCK_VIRTUAL:
  530         case CLOCK_PROF:
  531         case CLOCK_PROCESS_CPUTIME_ID:
  532                 return (ENOTSUP);
  533         case CLOCK_THREAD_CPUTIME_ID:
  534         default:
  535                 return (EINVAL);
  536         }
  537         do {
  538                 ts = *rqt;
  539                 if ((flags & TIMER_ABSTIME) != 0) {
  540                         if (is_abs_real)
  541                                 td->td_rtcgen =
  542                                     atomic_load_acq_int(&rtc_generation);
  543                         error = kern_clock_gettime(td, clock_id, &now);
  544                         KASSERT(error == 0, ("kern_clock_gettime: %d", error));
  545                         timespecsub(&ts, &now, &ts);
  546                 }
  547                 if (ts.tv_sec < 0 || (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
  548                         error = EWOULDBLOCK;
  549                         break;
  550                 }
  551                 if (ts.tv_sec > INT32_MAX / 2) {
  552                         over = ts.tv_sec - INT32_MAX / 2;
  553                         ts.tv_sec -= over;
  554                 } else
  555                         over = 0;
  556                 tmp = tstosbt(ts);
  557                 prec = tmp;
  558                 prec >>= tc_precexp;
  559                 if (TIMESEL(&sbt, tmp))
  560                         sbt += tc_tick_sbt;
  561                 sbt += tmp;
  562                 error = tsleep_sbt(&nanowait[curcpu], PWAIT | PCATCH, "nanslp",
  563                     sbt, prec, C_ABSOLUTE);
  564         } while (error == 0 && is_abs_real && td->td_rtcgen == 0);
  565         td->td_rtcgen = 0;
  566         if (error != EWOULDBLOCK) {
  567                 if (TIMESEL(&sbtt, tmp))
  568                         sbtt += tc_tick_sbt;
  569                 if (sbtt >= sbt)
  570                         return (0);
  571                 if (error == ERESTART)
  572                         error = EINTR;
  573                 if ((flags & TIMER_ABSTIME) == 0 && rmt != NULL) {
  574                         ts = sbttots(sbt - sbtt);
  575                         ts.tv_sec += over;
  576                         if (ts.tv_sec < 0)
  577                                 timespecclear(&ts);
  578                         *rmt = ts;
  579                 }
  580                 return (error);
  581         }
  582         return (0);
  583 }
  584 
  585 #ifndef _SYS_SYSPROTO_H_
  586 struct nanosleep_args {
  587         struct  timespec *rqtp;
  588         struct  timespec *rmtp;
  589 };
  590 #endif
  591 /* ARGSUSED */
  592 int
  593 sys_nanosleep(struct thread *td, struct nanosleep_args *uap)
  594 {
  595 
  596         return (user_clock_nanosleep(td, CLOCK_REALTIME, TIMER_RELTIME,
  597             uap->rqtp, uap->rmtp));
  598 }
  599 
  600 #ifndef _SYS_SYSPROTO_H_
  601 struct clock_nanosleep_args {
  602         clockid_t clock_id;
  603         int       flags;
  604         struct  timespec *rqtp;
  605         struct  timespec *rmtp;
  606 };
  607 #endif
  608 /* ARGSUSED */
  609 int
  610 sys_clock_nanosleep(struct thread *td, struct clock_nanosleep_args *uap)
  611 {
  612         int error;
  613 
  614         error = user_clock_nanosleep(td, uap->clock_id, uap->flags, uap->rqtp,
  615             uap->rmtp);
  616         return (kern_posix_error(td, error));
  617 }
  618 
  619 static int
  620 user_clock_nanosleep(struct thread *td, clockid_t clock_id, int flags,
  621     const struct timespec *ua_rqtp, struct timespec *ua_rmtp)
  622 {
  623         struct timespec rmt, rqt;
  624         int error, error2;
  625 
  626         error = copyin(ua_rqtp, &rqt, sizeof(rqt));
  627         if (error)
  628                 return (error);
  629         error = kern_clock_nanosleep(td, clock_id, flags, &rqt, &rmt);
  630         if (error == EINTR && ua_rmtp != NULL && (flags & TIMER_ABSTIME) == 0) {
  631                 error2 = copyout(&rmt, ua_rmtp, sizeof(rmt));
  632                 if (error2 != 0)
  633                         error = error2;
  634         }
  635         return (error);
  636 }
  637 
  638 #ifndef _SYS_SYSPROTO_H_
  639 struct gettimeofday_args {
  640         struct  timeval *tp;
  641         struct  timezone *tzp;
  642 };
  643 #endif
  644 /* ARGSUSED */
  645 int
  646 sys_gettimeofday(struct thread *td, struct gettimeofday_args *uap)
  647 {
  648         struct timeval atv;
  649         struct timezone rtz;
  650         int error = 0;
  651 
  652         if (uap->tp) {
  653                 microtime(&atv);
  654                 error = copyout(&atv, uap->tp, sizeof (atv));
  655         }
  656         if (error == 0 && uap->tzp != NULL) {
  657                 rtz.tz_minuteswest = tz_minuteswest;
  658                 rtz.tz_dsttime = tz_dsttime;
  659                 error = copyout(&rtz, uap->tzp, sizeof (rtz));
  660         }
  661         return (error);
  662 }
  663 
  664 #ifndef _SYS_SYSPROTO_H_
  665 struct settimeofday_args {
  666         struct  timeval *tv;
  667         struct  timezone *tzp;
  668 };
  669 #endif
  670 /* ARGSUSED */
  671 int
  672 sys_settimeofday(struct thread *td, struct settimeofday_args *uap)
  673 {
  674         struct timeval atv, *tvp;
  675         struct timezone atz, *tzp;
  676         int error;
  677 
  678         if (uap->tv) {
  679                 error = copyin(uap->tv, &atv, sizeof(atv));
  680                 if (error)
  681                         return (error);
  682                 tvp = &atv;
  683         } else
  684                 tvp = NULL;
  685         if (uap->tzp) {
  686                 error = copyin(uap->tzp, &atz, sizeof(atz));
  687                 if (error)
  688                         return (error);
  689                 tzp = &atz;
  690         } else
  691                 tzp = NULL;
  692         return (kern_settimeofday(td, tvp, tzp));
  693 }
  694 
  695 int
  696 kern_settimeofday(struct thread *td, struct timeval *tv, struct timezone *tzp)
  697 {
  698         int error;
  699 
  700         error = priv_check(td, PRIV_SETTIMEOFDAY);
  701         if (error)
  702                 return (error);
  703         /* Verify all parameters before changing time. */
  704         if (tv) {
  705                 if (tv->tv_usec < 0 || tv->tv_usec >= 1000000 ||
  706                     tv->tv_sec < 0)
  707                         return (EINVAL);
  708                 error = settime(td, tv);
  709         }
  710         if (tzp && error == 0) {
  711                 tz_minuteswest = tzp->tz_minuteswest;
  712                 tz_dsttime = tzp->tz_dsttime;
  713         }
  714         return (error);
  715 }
  716 
  717 /*
  718  * Get value of an interval timer.  The process virtual and profiling virtual
  719  * time timers are kept in the p_stats area, since they can be swapped out.
  720  * These are kept internally in the way they are specified externally: in
  721  * time until they expire.
  722  *
  723  * The real time interval timer is kept in the process table slot for the
  724  * process, and its value (it_value) is kept as an absolute time rather than
  725  * as a delta, so that it is easy to keep periodic real-time signals from
  726  * drifting.
  727  *
  728  * Virtual time timers are processed in the hardclock() routine of
  729  * kern_clock.c.  The real time timer is processed by a timeout routine,
  730  * called from the softclock() routine.  Since a callout may be delayed in
  731  * real time due to interrupt processing in the system, it is possible for
  732  * the real time timeout routine (realitexpire, given below), to be delayed
  733  * in real time past when it is supposed to occur.  It does not suffice,
  734  * therefore, to reload the real timer .it_value from the real time timers
  735  * .it_interval.  Rather, we compute the next time in absolute time the timer
  736  * should go off.
  737  */
  738 #ifndef _SYS_SYSPROTO_H_
  739 struct getitimer_args {
  740         u_int   which;
  741         struct  itimerval *itv;
  742 };
  743 #endif
  744 int
  745 sys_getitimer(struct thread *td, struct getitimer_args *uap)
  746 {
  747         struct itimerval aitv;
  748         int error;
  749 
  750         error = kern_getitimer(td, uap->which, &aitv);
  751         if (error != 0)
  752                 return (error);
  753         return (copyout(&aitv, uap->itv, sizeof (struct itimerval)));
  754 }
  755 
  756 int
  757 kern_getitimer(struct thread *td, u_int which, struct itimerval *aitv)
  758 {
  759         struct proc *p = td->td_proc;
  760         struct timeval ctv;
  761 
  762         if (which > ITIMER_PROF)
  763                 return (EINVAL);
  764 
  765         if (which == ITIMER_REAL) {
  766                 /*
  767                  * Convert from absolute to relative time in .it_value
  768                  * part of real time timer.  If time for real time timer
  769                  * has passed return 0, else return difference between
  770                  * current time and time for the timer to go off.
  771                  */
  772                 PROC_LOCK(p);
  773                 *aitv = p->p_realtimer;
  774                 PROC_UNLOCK(p);
  775                 if (timevalisset(&aitv->it_value)) {
  776                         microuptime(&ctv);
  777                         if (timevalcmp(&aitv->it_value, &ctv, <))
  778                                 timevalclear(&aitv->it_value);
  779                         else
  780                                 timevalsub(&aitv->it_value, &ctv);
  781                 }
  782         } else {
  783                 PROC_ITIMLOCK(p);
  784                 *aitv = p->p_stats->p_timer[which];
  785                 PROC_ITIMUNLOCK(p);
  786         }
  787 #ifdef KTRACE
  788         if (KTRPOINT(td, KTR_STRUCT))
  789                 ktritimerval(aitv);
  790 #endif
  791         return (0);
  792 }
  793 
  794 #ifndef _SYS_SYSPROTO_H_
  795 struct setitimer_args {
  796         u_int   which;
  797         struct  itimerval *itv, *oitv;
  798 };
  799 #endif
  800 int
  801 sys_setitimer(struct thread *td, struct setitimer_args *uap)
  802 {
  803         struct itimerval aitv, oitv;
  804         int error;
  805 
  806         if (uap->itv == NULL) {
  807                 uap->itv = uap->oitv;
  808                 return (sys_getitimer(td, (struct getitimer_args *)uap));
  809         }
  810 
  811         if ((error = copyin(uap->itv, &aitv, sizeof(struct itimerval))))
  812                 return (error);
  813         error = kern_setitimer(td, uap->which, &aitv, &oitv);
  814         if (error != 0 || uap->oitv == NULL)
  815                 return (error);
  816         return (copyout(&oitv, uap->oitv, sizeof(struct itimerval)));
  817 }
  818 
  819 int
  820 kern_setitimer(struct thread *td, u_int which, struct itimerval *aitv,
  821     struct itimerval *oitv)
  822 {
  823         struct proc *p = td->td_proc;
  824         struct timeval ctv;
  825         sbintime_t sbt, pr;
  826 
  827         if (aitv == NULL)
  828                 return (kern_getitimer(td, which, oitv));
  829 
  830         if (which > ITIMER_PROF)
  831                 return (EINVAL);
  832 #ifdef KTRACE
  833         if (KTRPOINT(td, KTR_STRUCT))
  834                 ktritimerval(aitv);
  835 #endif
  836         if (itimerfix(&aitv->it_value) ||
  837             aitv->it_value.tv_sec > INT32_MAX / 2)
  838                 return (EINVAL);
  839         if (!timevalisset(&aitv->it_value))
  840                 timevalclear(&aitv->it_interval);
  841         else if (itimerfix(&aitv->it_interval) ||
  842             aitv->it_interval.tv_sec > INT32_MAX / 2)
  843                 return (EINVAL);
  844 
  845         if (which == ITIMER_REAL) {
  846                 PROC_LOCK(p);
  847                 if (timevalisset(&p->p_realtimer.it_value))
  848                         callout_stop(&p->p_itcallout);
  849                 microuptime(&ctv);
  850                 if (timevalisset(&aitv->it_value)) {
  851                         pr = tvtosbt(aitv->it_value) >> tc_precexp;
  852                         timevaladd(&aitv->it_value, &ctv);
  853                         sbt = tvtosbt(aitv->it_value);
  854                         callout_reset_sbt(&p->p_itcallout, sbt, pr,
  855                             realitexpire, p, C_ABSOLUTE);
  856                 }
  857                 *oitv = p->p_realtimer;
  858                 p->p_realtimer = *aitv;
  859                 PROC_UNLOCK(p);
  860                 if (timevalisset(&oitv->it_value)) {
  861                         if (timevalcmp(&oitv->it_value, &ctv, <))
  862                                 timevalclear(&oitv->it_value);
  863                         else
  864                                 timevalsub(&oitv->it_value, &ctv);
  865                 }
  866         } else {
  867                 if (aitv->it_interval.tv_sec == 0 &&
  868                     aitv->it_interval.tv_usec != 0 &&
  869                     aitv->it_interval.tv_usec < tick)
  870                         aitv->it_interval.tv_usec = tick;
  871                 if (aitv->it_value.tv_sec == 0 &&
  872                     aitv->it_value.tv_usec != 0 &&
  873                     aitv->it_value.tv_usec < tick)
  874                         aitv->it_value.tv_usec = tick;
  875                 PROC_ITIMLOCK(p);
  876                 *oitv = p->p_stats->p_timer[which];
  877                 p->p_stats->p_timer[which] = *aitv;
  878                 PROC_ITIMUNLOCK(p);
  879         }
  880 #ifdef KTRACE
  881         if (KTRPOINT(td, KTR_STRUCT))
  882                 ktritimerval(oitv);
  883 #endif
  884         return (0);
  885 }
  886 
  887 /*
  888  * Real interval timer expired:
  889  * send process whose timer expired an alarm signal.
  890  * If time is not set up to reload, then just return.
  891  * Else compute next time timer should go off which is > current time.
  892  * This is where delay in processing this timeout causes multiple
  893  * SIGALRM calls to be compressed into one.
  894  * tvtohz() always adds 1 to allow for the time until the next clock
  895  * interrupt being strictly less than 1 clock tick, but we don't want
  896  * that here since we want to appear to be in sync with the clock
  897  * interrupt even when we're delayed.
  898  */
  899 void
  900 realitexpire(void *arg)
  901 {
  902         struct proc *p;
  903         struct timeval ctv;
  904         sbintime_t isbt;
  905 
  906         p = (struct proc *)arg;
  907         kern_psignal(p, SIGALRM);
  908         if (!timevalisset(&p->p_realtimer.it_interval)) {
  909                 timevalclear(&p->p_realtimer.it_value);
  910                 if (p->p_flag & P_WEXIT)
  911                         wakeup(&p->p_itcallout);
  912                 return;
  913         }
  914         isbt = tvtosbt(p->p_realtimer.it_interval);
  915         if (isbt >= sbt_timethreshold)
  916                 getmicrouptime(&ctv);
  917         else
  918                 microuptime(&ctv);
  919         do {
  920                 timevaladd(&p->p_realtimer.it_value,
  921                     &p->p_realtimer.it_interval);
  922         } while (timevalcmp(&p->p_realtimer.it_value, &ctv, <=));
  923         callout_reset_sbt(&p->p_itcallout, tvtosbt(p->p_realtimer.it_value),
  924             isbt >> tc_precexp, realitexpire, p, C_ABSOLUTE);
  925 }
  926 
  927 /*
  928  * Check that a proposed value to load into the .it_value or
  929  * .it_interval part of an interval timer is acceptable, and
  930  * fix it to have at least minimal value (i.e. if it is less
  931  * than the resolution of the clock, round it up.)
  932  */
  933 int
  934 itimerfix(struct timeval *tv)
  935 {
  936 
  937         if (tv->tv_sec < 0 || tv->tv_usec < 0 || tv->tv_usec >= 1000000)
  938                 return (EINVAL);
  939         if (tv->tv_sec == 0 && tv->tv_usec != 0 &&
  940             tv->tv_usec < (u_int)tick / 16)
  941                 tv->tv_usec = (u_int)tick / 16;
  942         return (0);
  943 }
  944 
  945 /*
  946  * Decrement an interval timer by a specified number
  947  * of microseconds, which must be less than a second,
  948  * i.e. < 1000000.  If the timer expires, then reload
  949  * it.  In this case, carry over (usec - old value) to
  950  * reduce the value reloaded into the timer so that
  951  * the timer does not drift.  This routine assumes
  952  * that it is called in a context where the timers
  953  * on which it is operating cannot change in value.
  954  */
  955 int
  956 itimerdecr(struct itimerval *itp, int usec)
  957 {
  958 
  959         if (itp->it_value.tv_usec < usec) {
  960                 if (itp->it_value.tv_sec == 0) {
  961                         /* expired, and already in next interval */
  962                         usec -= itp->it_value.tv_usec;
  963                         goto expire;
  964                 }
  965                 itp->it_value.tv_usec += 1000000;
  966                 itp->it_value.tv_sec--;
  967         }
  968         itp->it_value.tv_usec -= usec;
  969         usec = 0;
  970         if (timevalisset(&itp->it_value))
  971                 return (1);
  972         /* expired, exactly at end of interval */
  973 expire:
  974         if (timevalisset(&itp->it_interval)) {
  975                 itp->it_value = itp->it_interval;
  976                 itp->it_value.tv_usec -= usec;
  977                 if (itp->it_value.tv_usec < 0) {
  978                         itp->it_value.tv_usec += 1000000;
  979                         itp->it_value.tv_sec--;
  980                 }
  981         } else
  982                 itp->it_value.tv_usec = 0;              /* sec is already 0 */
  983         return (0);
  984 }
  985 
  986 /*
  987  * Add and subtract routines for timevals.
  988  * N.B.: subtract routine doesn't deal with
  989  * results which are before the beginning,
  990  * it just gets very confused in this case.
  991  * Caveat emptor.
  992  */
  993 void
  994 timevaladd(struct timeval *t1, const struct timeval *t2)
  995 {
  996 
  997         t1->tv_sec += t2->tv_sec;
  998         t1->tv_usec += t2->tv_usec;
  999         timevalfix(t1);
 1000 }
 1001 
 1002 void
 1003 timevalsub(struct timeval *t1, const struct timeval *t2)
 1004 {
 1005 
 1006         t1->tv_sec -= t2->tv_sec;
 1007         t1->tv_usec -= t2->tv_usec;
 1008         timevalfix(t1);
 1009 }
 1010 
 1011 static void
 1012 timevalfix(struct timeval *t1)
 1013 {
 1014 
 1015         if (t1->tv_usec < 0) {
 1016                 t1->tv_sec--;
 1017                 t1->tv_usec += 1000000;
 1018         }
 1019         if (t1->tv_usec >= 1000000) {
 1020                 t1->tv_sec++;
 1021                 t1->tv_usec -= 1000000;
 1022         }
 1023 }
 1024 
 1025 /*
 1026  * ratecheck(): simple time-based rate-limit checking.
 1027  */
 1028 int
 1029 ratecheck(struct timeval *lasttime, const struct timeval *mininterval)
 1030 {
 1031         struct timeval tv, delta;
 1032         int rv = 0;
 1033 
 1034         getmicrouptime(&tv);            /* NB: 10ms precision */
 1035         delta = tv;
 1036         timevalsub(&delta, lasttime);
 1037 
 1038         /*
 1039          * check for 0,0 is so that the message will be seen at least once,
 1040          * even if interval is huge.
 1041          */
 1042         if (timevalcmp(&delta, mininterval, >=) ||
 1043             (lasttime->tv_sec == 0 && lasttime->tv_usec == 0)) {
 1044                 *lasttime = tv;
 1045                 rv = 1;
 1046         }
 1047 
 1048         return (rv);
 1049 }
 1050 
 1051 /*
 1052  * ppsratecheck(): packets (or events) per second limitation.
 1053  *
 1054  * Return 0 if the limit is to be enforced (e.g. the caller
 1055  * should drop a packet because of the rate limitation).
 1056  *
 1057  * maxpps of 0 always causes zero to be returned.  maxpps of -1
 1058  * always causes 1 to be returned; this effectively defeats rate
 1059  * limiting.
 1060  *
 1061  * Note that we maintain the struct timeval for compatibility
 1062  * with other bsd systems.  We reuse the storage and just monitor
 1063  * clock ticks for minimal overhead.  
 1064  */
 1065 int
 1066 ppsratecheck(struct timeval *lasttime, int *curpps, int maxpps)
 1067 {
 1068         int now;
 1069 
 1070         /*
 1071          * Reset the last time and counter if this is the first call
 1072          * or more than a second has passed since the last update of
 1073          * lasttime.
 1074          */
 1075         now = ticks;
 1076         if (lasttime->tv_sec == 0 || (u_int)(now - lasttime->tv_sec) >= hz) {
 1077                 lasttime->tv_sec = now;
 1078                 *curpps = 1;
 1079                 return (maxpps != 0);
 1080         } else {
 1081                 (*curpps)++;            /* NB: ignore potential overflow */
 1082                 return (maxpps < 0 || *curpps <= maxpps);
 1083         }
 1084 }
 1085 
 1086 static void
 1087 itimer_start(void)
 1088 {
 1089         static const struct kclock rt_clock = {
 1090                 .timer_create  = realtimer_create,
 1091                 .timer_delete  = realtimer_delete,
 1092                 .timer_settime = realtimer_settime,
 1093                 .timer_gettime = realtimer_gettime,
 1094         };
 1095 
 1096         itimer_zone = uma_zcreate("itimer", sizeof(struct itimer),
 1097                 NULL, NULL, itimer_init, itimer_fini, UMA_ALIGN_PTR, 0);
 1098         register_posix_clock(CLOCK_REALTIME,  &rt_clock);
 1099         register_posix_clock(CLOCK_MONOTONIC, &rt_clock);
 1100         p31b_setcfg(CTL_P1003_1B_TIMERS, 200112L);
 1101         p31b_setcfg(CTL_P1003_1B_DELAYTIMER_MAX, INT_MAX);
 1102         p31b_setcfg(CTL_P1003_1B_TIMER_MAX, TIMER_MAX);
 1103 }
 1104 
 1105 static int
 1106 register_posix_clock(int clockid, const struct kclock *clk)
 1107 {
 1108         if ((unsigned)clockid >= MAX_CLOCKS) {
 1109                 printf("%s: invalid clockid\n", __func__);
 1110                 return (0);
 1111         }
 1112         posix_clocks[clockid] = *clk;
 1113         return (1);
 1114 }
 1115 
 1116 static int
 1117 itimer_init(void *mem, int size, int flags)
 1118 {
 1119         struct itimer *it;
 1120 
 1121         it = (struct itimer *)mem;
 1122         mtx_init(&it->it_mtx, "itimer lock", NULL, MTX_DEF);
 1123         return (0);
 1124 }
 1125 
 1126 static void
 1127 itimer_fini(void *mem, int size)
 1128 {
 1129         struct itimer *it;
 1130 
 1131         it = (struct itimer *)mem;
 1132         mtx_destroy(&it->it_mtx);
 1133 }
 1134 
 1135 static void
 1136 itimer_enter(struct itimer *it)
 1137 {
 1138 
 1139         mtx_assert(&it->it_mtx, MA_OWNED);
 1140         it->it_usecount++;
 1141 }
 1142 
 1143 static void
 1144 itimer_leave(struct itimer *it)
 1145 {
 1146 
 1147         mtx_assert(&it->it_mtx, MA_OWNED);
 1148         KASSERT(it->it_usecount > 0, ("invalid it_usecount"));
 1149 
 1150         if (--it->it_usecount == 0 && (it->it_flags & ITF_WANTED) != 0)
 1151                 wakeup(it);
 1152 }
 1153 
 1154 #ifndef _SYS_SYSPROTO_H_
 1155 struct ktimer_create_args {
 1156         clockid_t clock_id;
 1157         struct sigevent * evp;
 1158         int * timerid;
 1159 };
 1160 #endif
 1161 int
 1162 sys_ktimer_create(struct thread *td, struct ktimer_create_args *uap)
 1163 {
 1164         struct sigevent *evp, ev;
 1165         int id;
 1166         int error;
 1167 
 1168         if (uap->evp == NULL) {
 1169                 evp = NULL;
 1170         } else {
 1171                 error = copyin(uap->evp, &ev, sizeof(ev));
 1172                 if (error != 0)
 1173                         return (error);
 1174                 evp = &ev;
 1175         }
 1176         error = kern_ktimer_create(td, uap->clock_id, evp, &id, -1);
 1177         if (error == 0) {
 1178                 error = copyout(&id, uap->timerid, sizeof(int));
 1179                 if (error != 0)
 1180                         kern_ktimer_delete(td, id);
 1181         }
 1182         return (error);
 1183 }
 1184 
 1185 int
 1186 kern_ktimer_create(struct thread *td, clockid_t clock_id, struct sigevent *evp,
 1187     int *timerid, int preset_id)
 1188 {
 1189         struct proc *p = td->td_proc;
 1190         struct itimer *it;
 1191         int id;
 1192         int error;
 1193 
 1194         if (clock_id < 0 || clock_id >= MAX_CLOCKS)
 1195                 return (EINVAL);
 1196 
 1197         if (posix_clocks[clock_id].timer_create == NULL)
 1198                 return (EINVAL);
 1199 
 1200         if (evp != NULL) {
 1201                 if (evp->sigev_notify != SIGEV_NONE &&
 1202                     evp->sigev_notify != SIGEV_SIGNAL &&
 1203                     evp->sigev_notify != SIGEV_THREAD_ID)
 1204                         return (EINVAL);
 1205                 if ((evp->sigev_notify == SIGEV_SIGNAL ||
 1206                      evp->sigev_notify == SIGEV_THREAD_ID) &&
 1207                         !_SIG_VALID(evp->sigev_signo))
 1208                         return (EINVAL);
 1209         }
 1210         
 1211         if (p->p_itimers == NULL)
 1212                 itimers_alloc(p);
 1213         
 1214         it = uma_zalloc(itimer_zone, M_WAITOK);
 1215         it->it_flags = 0;
 1216         it->it_usecount = 0;
 1217         it->it_active = 0;
 1218         timespecclear(&it->it_time.it_value);
 1219         timespecclear(&it->it_time.it_interval);
 1220         it->it_overrun = 0;
 1221         it->it_overrun_last = 0;
 1222         it->it_clockid = clock_id;
 1223         it->it_timerid = -1;
 1224         it->it_proc = p;
 1225         ksiginfo_init(&it->it_ksi);
 1226         it->it_ksi.ksi_flags |= KSI_INS | KSI_EXT;
 1227         error = CLOCK_CALL(clock_id, timer_create, (it));
 1228         if (error != 0)
 1229                 goto out;
 1230 
 1231         PROC_LOCK(p);
 1232         if (preset_id != -1) {
 1233                 KASSERT(preset_id >= 0 && preset_id < 3, ("invalid preset_id"));
 1234                 id = preset_id;
 1235                 if (p->p_itimers->its_timers[id] != NULL) {
 1236                         PROC_UNLOCK(p);
 1237                         error = 0;
 1238                         goto out;
 1239                 }
 1240         } else {
 1241                 /*
 1242                  * Find a free timer slot, skipping those reserved
 1243                  * for setitimer().
 1244                  */
 1245                 for (id = 3; id < TIMER_MAX; id++)
 1246                         if (p->p_itimers->its_timers[id] == NULL)
 1247                                 break;
 1248                 if (id == TIMER_MAX) {
 1249                         PROC_UNLOCK(p);
 1250                         error = EAGAIN;
 1251                         goto out;
 1252                 }
 1253         }
 1254         it->it_timerid = id;
 1255         p->p_itimers->its_timers[id] = it;
 1256         if (evp != NULL)
 1257                 it->it_sigev = *evp;
 1258         else {
 1259                 it->it_sigev.sigev_notify = SIGEV_SIGNAL;
 1260                 switch (clock_id) {
 1261                 default:
 1262                 case CLOCK_REALTIME:
 1263                         it->it_sigev.sigev_signo = SIGALRM;
 1264                         break;
 1265                 case CLOCK_VIRTUAL:
 1266                         it->it_sigev.sigev_signo = SIGVTALRM;
 1267                         break;
 1268                 case CLOCK_PROF:
 1269                         it->it_sigev.sigev_signo = SIGPROF;
 1270                         break;
 1271                 }
 1272                 it->it_sigev.sigev_value.sival_int = id;
 1273         }
 1274 
 1275         if (it->it_sigev.sigev_notify == SIGEV_SIGNAL ||
 1276             it->it_sigev.sigev_notify == SIGEV_THREAD_ID) {
 1277                 it->it_ksi.ksi_signo = it->it_sigev.sigev_signo;
 1278                 it->it_ksi.ksi_code = SI_TIMER;
 1279                 it->it_ksi.ksi_value = it->it_sigev.sigev_value;
 1280                 it->it_ksi.ksi_timerid = id;
 1281         }
 1282         PROC_UNLOCK(p);
 1283         *timerid = id;
 1284         return (0);
 1285 
 1286 out:
 1287         ITIMER_LOCK(it);
 1288         CLOCK_CALL(it->it_clockid, timer_delete, (it));
 1289         ITIMER_UNLOCK(it);
 1290         uma_zfree(itimer_zone, it);
 1291         return (error);
 1292 }
 1293 
 1294 #ifndef _SYS_SYSPROTO_H_
 1295 struct ktimer_delete_args {
 1296         int timerid;
 1297 };
 1298 #endif
 1299 int
 1300 sys_ktimer_delete(struct thread *td, struct ktimer_delete_args *uap)
 1301 {
 1302 
 1303         return (kern_ktimer_delete(td, uap->timerid));
 1304 }
 1305 
 1306 static struct itimer *
 1307 itimer_find(struct proc *p, int timerid)
 1308 {
 1309         struct itimer *it;
 1310 
 1311         PROC_LOCK_ASSERT(p, MA_OWNED);
 1312         if ((p->p_itimers == NULL) ||
 1313             (timerid < 0) || (timerid >= TIMER_MAX) ||
 1314             (it = p->p_itimers->its_timers[timerid]) == NULL) {
 1315                 return (NULL);
 1316         }
 1317         ITIMER_LOCK(it);
 1318         if ((it->it_flags & ITF_DELETING) != 0) {
 1319                 ITIMER_UNLOCK(it);
 1320                 it = NULL;
 1321         }
 1322         return (it);
 1323 }
 1324 
 1325 int
 1326 kern_ktimer_delete(struct thread *td, int timerid)
 1327 {
 1328         struct proc *p = td->td_proc;
 1329         struct itimer *it;
 1330 
 1331         PROC_LOCK(p);
 1332         it = itimer_find(p, timerid);
 1333         if (it == NULL) {
 1334                 PROC_UNLOCK(p);
 1335                 return (EINVAL);
 1336         }
 1337         PROC_UNLOCK(p);
 1338 
 1339         it->it_flags |= ITF_DELETING;
 1340         while (it->it_usecount > 0) {
 1341                 it->it_flags |= ITF_WANTED;
 1342                 msleep(it, &it->it_mtx, PPAUSE, "itimer", 0);
 1343         }
 1344         it->it_flags &= ~ITF_WANTED;
 1345         CLOCK_CALL(it->it_clockid, timer_delete, (it));
 1346         ITIMER_UNLOCK(it);
 1347 
 1348         PROC_LOCK(p);
 1349         if (KSI_ONQ(&it->it_ksi))
 1350                 sigqueue_take(&it->it_ksi);
 1351         p->p_itimers->its_timers[timerid] = NULL;
 1352         PROC_UNLOCK(p);
 1353         uma_zfree(itimer_zone, it);
 1354         return (0);
 1355 }
 1356 
 1357 #ifndef _SYS_SYSPROTO_H_
 1358 struct ktimer_settime_args {
 1359         int timerid;
 1360         int flags;
 1361         const struct itimerspec * value;
 1362         struct itimerspec * ovalue;
 1363 };
 1364 #endif
 1365 int
 1366 sys_ktimer_settime(struct thread *td, struct ktimer_settime_args *uap)
 1367 {
 1368         struct itimerspec val, oval, *ovalp;
 1369         int error;
 1370 
 1371         error = copyin(uap->value, &val, sizeof(val));
 1372         if (error != 0)
 1373                 return (error);
 1374         ovalp = uap->ovalue != NULL ? &oval : NULL;
 1375         error = kern_ktimer_settime(td, uap->timerid, uap->flags, &val, ovalp);
 1376         if (error == 0 && uap->ovalue != NULL)
 1377                 error = copyout(ovalp, uap->ovalue, sizeof(*ovalp));
 1378         return (error);
 1379 }
 1380 
 1381 int
 1382 kern_ktimer_settime(struct thread *td, int timer_id, int flags,
 1383     struct itimerspec *val, struct itimerspec *oval)
 1384 {
 1385         struct proc *p;
 1386         struct itimer *it;
 1387         int error;
 1388 
 1389         p = td->td_proc;
 1390         PROC_LOCK(p);
 1391         if (timer_id < 3 || (it = itimer_find(p, timer_id)) == NULL) {
 1392                 PROC_UNLOCK(p);
 1393                 error = EINVAL;
 1394         } else {
 1395                 PROC_UNLOCK(p);
 1396                 itimer_enter(it);
 1397                 error = CLOCK_CALL(it->it_clockid, timer_settime, (it,
 1398                     flags, val, oval));
 1399                 itimer_leave(it);
 1400                 ITIMER_UNLOCK(it);
 1401         }
 1402         return (error);
 1403 }
 1404 
 1405 #ifndef _SYS_SYSPROTO_H_
 1406 struct ktimer_gettime_args {
 1407         int timerid;
 1408         struct itimerspec * value;
 1409 };
 1410 #endif
 1411 int
 1412 sys_ktimer_gettime(struct thread *td, struct ktimer_gettime_args *uap)
 1413 {
 1414         struct itimerspec val;
 1415         int error;
 1416 
 1417         error = kern_ktimer_gettime(td, uap->timerid, &val);
 1418         if (error == 0)
 1419                 error = copyout(&val, uap->value, sizeof(val));
 1420         return (error);
 1421 }
 1422 
 1423 int
 1424 kern_ktimer_gettime(struct thread *td, int timer_id, struct itimerspec *val)
 1425 {
 1426         struct proc *p;
 1427         struct itimer *it;
 1428         int error;
 1429 
 1430         p = td->td_proc;
 1431         PROC_LOCK(p);
 1432         if (timer_id < 3 || (it = itimer_find(p, timer_id)) == NULL) {
 1433                 PROC_UNLOCK(p);
 1434                 error = EINVAL;
 1435         } else {
 1436                 PROC_UNLOCK(p);
 1437                 itimer_enter(it);
 1438                 error = CLOCK_CALL(it->it_clockid, timer_gettime, (it, val));
 1439                 itimer_leave(it);
 1440                 ITIMER_UNLOCK(it);
 1441         }
 1442         return (error);
 1443 }
 1444 
 1445 #ifndef _SYS_SYSPROTO_H_
 1446 struct timer_getoverrun_args {
 1447         int timerid;
 1448 };
 1449 #endif
 1450 int
 1451 sys_ktimer_getoverrun(struct thread *td, struct ktimer_getoverrun_args *uap)
 1452 {
 1453 
 1454         return (kern_ktimer_getoverrun(td, uap->timerid));
 1455 }
 1456 
 1457 int
 1458 kern_ktimer_getoverrun(struct thread *td, int timer_id)
 1459 {
 1460         struct proc *p = td->td_proc;
 1461         struct itimer *it;
 1462         int error ;
 1463 
 1464         PROC_LOCK(p);
 1465         if (timer_id < 3 ||
 1466             (it = itimer_find(p, timer_id)) == NULL) {
 1467                 PROC_UNLOCK(p);
 1468                 error = EINVAL;
 1469         } else {
 1470                 td->td_retval[0] = it->it_overrun_last;
 1471                 ITIMER_UNLOCK(it);
 1472                 PROC_UNLOCK(p);
 1473                 error = 0;
 1474         }
 1475         return (error);
 1476 }
 1477 
 1478 static int
 1479 realtimer_create(struct itimer *it)
 1480 {
 1481         callout_init_mtx(&it->it_callout, &it->it_mtx, 0);
 1482         return (0);
 1483 }
 1484 
 1485 static int
 1486 realtimer_delete(struct itimer *it)
 1487 {
 1488         mtx_assert(&it->it_mtx, MA_OWNED);
 1489         
 1490         /*
 1491          * clear timer's value and interval to tell realtimer_expire
 1492          * to not rearm the timer.
 1493          */
 1494         timespecclear(&it->it_time.it_value);
 1495         timespecclear(&it->it_time.it_interval);
 1496         ITIMER_UNLOCK(it);
 1497         callout_drain(&it->it_callout);
 1498         ITIMER_LOCK(it);
 1499         return (0);
 1500 }
 1501 
 1502 static int
 1503 realtimer_gettime(struct itimer *it, struct itimerspec *ovalue)
 1504 {
 1505         struct timespec cts;
 1506 
 1507         mtx_assert(&it->it_mtx, MA_OWNED);
 1508 
 1509         realtimer_clocktime(it->it_clockid, &cts);
 1510         *ovalue = it->it_time;
 1511         if (ovalue->it_value.tv_sec != 0 || ovalue->it_value.tv_nsec != 0) {
 1512                 timespecsub(&ovalue->it_value, &cts, &ovalue->it_value);
 1513                 if (ovalue->it_value.tv_sec < 0 ||
 1514                     (ovalue->it_value.tv_sec == 0 &&
 1515                      ovalue->it_value.tv_nsec == 0)) {
 1516                         ovalue->it_value.tv_sec  = 0;
 1517                         ovalue->it_value.tv_nsec = 1;
 1518                 }
 1519         }
 1520         return (0);
 1521 }
 1522 
 1523 static int
 1524 realtimer_settime(struct itimer *it, int flags,
 1525         struct itimerspec *value, struct itimerspec *ovalue)
 1526 {
 1527         struct timespec cts, ts;
 1528         struct timeval tv;
 1529         struct itimerspec val;
 1530 
 1531         mtx_assert(&it->it_mtx, MA_OWNED);
 1532 
 1533         val = *value;
 1534         if (itimespecfix(&val.it_value))
 1535                 return (EINVAL);
 1536 
 1537         if (timespecisset(&val.it_value)) {
 1538                 if (itimespecfix(&val.it_interval))
 1539                         return (EINVAL);
 1540         } else {
 1541                 timespecclear(&val.it_interval);
 1542         }
 1543         
 1544         if (ovalue != NULL)
 1545                 realtimer_gettime(it, ovalue);
 1546 
 1547         it->it_time = val;
 1548         if (timespecisset(&val.it_value)) {
 1549                 realtimer_clocktime(it->it_clockid, &cts);
 1550                 ts = val.it_value;
 1551                 if ((flags & TIMER_ABSTIME) == 0) {
 1552                         /* Convert to absolute time. */
 1553                         timespecadd(&it->it_time.it_value, &cts,
 1554                                 &it->it_time.it_value);
 1555                 } else {
 1556                         timespecsub(&ts, &cts, &ts);
 1557                         /*
 1558                          * We don't care if ts is negative, tztohz will
 1559                          * fix it.
 1560                          */
 1561                 }
 1562                 TIMESPEC_TO_TIMEVAL(&tv, &ts);
 1563                 callout_reset(&it->it_callout, tvtohz(&tv),
 1564                         realtimer_expire, it);
 1565         } else {
 1566                 callout_stop(&it->it_callout);
 1567         }
 1568 
 1569         return (0);
 1570 }
 1571 
 1572 static void
 1573 realtimer_clocktime(clockid_t id, struct timespec *ts)
 1574 {
 1575         if (id == CLOCK_REALTIME)
 1576                 getnanotime(ts);
 1577         else    /* CLOCK_MONOTONIC */
 1578                 getnanouptime(ts);
 1579 }
 1580 
 1581 int
 1582 itimer_accept(struct proc *p, int timerid, ksiginfo_t *ksi)
 1583 {
 1584         struct itimer *it;
 1585 
 1586         PROC_LOCK_ASSERT(p, MA_OWNED);
 1587         it = itimer_find(p, timerid);
 1588         if (it != NULL) {
 1589                 ksi->ksi_overrun = it->it_overrun;
 1590                 it->it_overrun_last = it->it_overrun;
 1591                 it->it_overrun = 0;
 1592                 ITIMER_UNLOCK(it);
 1593                 return (0);
 1594         }
 1595         return (EINVAL);
 1596 }
 1597 
 1598 int
 1599 itimespecfix(struct timespec *ts)
 1600 {
 1601 
 1602         if (ts->tv_sec < 0 || ts->tv_nsec < 0 || ts->tv_nsec >= 1000000000)
 1603                 return (EINVAL);
 1604         if (ts->tv_sec == 0 && ts->tv_nsec != 0 && ts->tv_nsec < tick * 1000)
 1605                 ts->tv_nsec = tick * 1000;
 1606         return (0);
 1607 }
 1608 
 1609 /* Timeout callback for realtime timer */
 1610 static void
 1611 realtimer_expire(void *arg)
 1612 {
 1613         struct timespec cts, ts;
 1614         struct timeval tv;
 1615         struct itimer *it;
 1616 
 1617         it = (struct itimer *)arg;
 1618 
 1619         realtimer_clocktime(it->it_clockid, &cts);
 1620         /* Only fire if time is reached. */
 1621         if (timespeccmp(&cts, &it->it_time.it_value, >=)) {
 1622                 if (timespecisset(&it->it_time.it_interval)) {
 1623                         timespecadd(&it->it_time.it_value,
 1624                                     &it->it_time.it_interval,
 1625                                     &it->it_time.it_value);
 1626                         while (timespeccmp(&cts, &it->it_time.it_value, >=)) {
 1627                                 if (it->it_overrun < INT_MAX)
 1628                                         it->it_overrun++;
 1629                                 else
 1630                                         it->it_ksi.ksi_errno = ERANGE;
 1631                                 timespecadd(&it->it_time.it_value,
 1632                                             &it->it_time.it_interval,
 1633                                             &it->it_time.it_value);
 1634                         }
 1635                 } else {
 1636                         /* single shot timer ? */
 1637                         timespecclear(&it->it_time.it_value);
 1638                 }
 1639                 if (timespecisset(&it->it_time.it_value)) {
 1640                         timespecsub(&it->it_time.it_value, &cts, &ts);
 1641                         TIMESPEC_TO_TIMEVAL(&tv, &ts);
 1642                         callout_reset(&it->it_callout, tvtohz(&tv),
 1643                                  realtimer_expire, it);
 1644                 }
 1645                 itimer_enter(it);
 1646                 ITIMER_UNLOCK(it);
 1647                 itimer_fire(it);
 1648                 ITIMER_LOCK(it);
 1649                 itimer_leave(it);
 1650         } else if (timespecisset(&it->it_time.it_value)) {
 1651                 ts = it->it_time.it_value;
 1652                 timespecsub(&ts, &cts, &ts);
 1653                 TIMESPEC_TO_TIMEVAL(&tv, &ts);
 1654                 callout_reset(&it->it_callout, tvtohz(&tv), realtimer_expire,
 1655                         it);
 1656         }
 1657 }
 1658 
 1659 void
 1660 itimer_fire(struct itimer *it)
 1661 {
 1662         struct proc *p = it->it_proc;
 1663         struct thread *td;
 1664 
 1665         if (it->it_sigev.sigev_notify == SIGEV_SIGNAL ||
 1666             it->it_sigev.sigev_notify == SIGEV_THREAD_ID) {
 1667                 if (sigev_findtd(p, &it->it_sigev, &td) != 0) {
 1668                         ITIMER_LOCK(it);
 1669                         timespecclear(&it->it_time.it_value);
 1670                         timespecclear(&it->it_time.it_interval);
 1671                         callout_stop(&it->it_callout);
 1672                         ITIMER_UNLOCK(it);
 1673                         return;
 1674                 }
 1675                 if (!KSI_ONQ(&it->it_ksi)) {
 1676                         it->it_ksi.ksi_errno = 0;
 1677                         ksiginfo_set_sigev(&it->it_ksi, &it->it_sigev);
 1678                         tdsendsignal(p, td, it->it_ksi.ksi_signo, &it->it_ksi);
 1679                 } else {
 1680                         if (it->it_overrun < INT_MAX)
 1681                                 it->it_overrun++;
 1682                         else
 1683                                 it->it_ksi.ksi_errno = ERANGE;
 1684                 }
 1685                 PROC_UNLOCK(p);
 1686         }
 1687 }
 1688 
 1689 static void
 1690 itimers_alloc(struct proc *p)
 1691 {
 1692         struct itimers *its;
 1693         int i;
 1694 
 1695         its = malloc(sizeof (struct itimers), M_SUBPROC, M_WAITOK | M_ZERO);
 1696         LIST_INIT(&its->its_virtual);
 1697         LIST_INIT(&its->its_prof);
 1698         TAILQ_INIT(&its->its_worklist);
 1699         for (i = 0; i < TIMER_MAX; i++)
 1700                 its->its_timers[i] = NULL;
 1701         PROC_LOCK(p);
 1702         if (p->p_itimers == NULL) {
 1703                 p->p_itimers = its;
 1704                 PROC_UNLOCK(p);
 1705         }
 1706         else {
 1707                 PROC_UNLOCK(p);
 1708                 free(its, M_SUBPROC);
 1709         }
 1710 }
 1711 
 1712 /* Clean up timers when some process events are being triggered. */
 1713 static void
 1714 itimers_event_exit_exec(int start_idx, struct proc *p)
 1715 {
 1716         struct itimers *its;
 1717         struct itimer *it;
 1718         int i;
 1719 
 1720         its = p->p_itimers;
 1721         if (its == NULL)
 1722                 return;
 1723 
 1724         for (i = start_idx; i < TIMER_MAX; ++i) {
 1725                 if ((it = its->its_timers[i]) != NULL)
 1726                         kern_ktimer_delete(curthread, i);
 1727         }
 1728         if (its->its_timers[0] == NULL && its->its_timers[1] == NULL &&
 1729             its->its_timers[2] == NULL) {
 1730                 free(its, M_SUBPROC);
 1731                 p->p_itimers = NULL;
 1732         }
 1733 }
 1734 
 1735 void
 1736 itimers_exec(struct proc *p)
 1737 {
 1738         /*
 1739          * According to susv3, XSI interval timers should be inherited
 1740          * by new image.
 1741          */
 1742         itimers_event_exit_exec(3, p);
 1743 }
 1744 
 1745 void
 1746 itimers_exit(struct proc *p)
 1747 {
 1748         itimers_event_exit_exec(0, p);
 1749 }

Cache object: 615aaa7a42f919f6cdac295d80b6280e


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.