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  * Copyright (c) 1982, 1986, 1989, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 4. Neither the name of the University nor the names of its contributors
   14  *    may be used to endorse or promote products derived from this software
   15  *    without specific prior written permission.
   16  *
   17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  *
   29  *      @(#)kern_time.c 8.1 (Berkeley) 6/10/93
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD: releng/11.2/sys/kern/kern_time.c 331722 2018-03-29 02:50:57Z eadler $");
   34 
   35 #include "opt_ktrace.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/systm.h>
   39 #include <sys/limits.h>
   40 #include <sys/clock.h>
   41 #include <sys/lock.h>
   42 #include <sys/mutex.h>
   43 #include <sys/sysproto.h>
   44 #include <sys/eventhandler.h>
   45 #include <sys/resourcevar.h>
   46 #include <sys/signalvar.h>
   47 #include <sys/kernel.h>
   48 #include <sys/sleepqueue.h>
   49 #include <sys/syscallsubr.h>
   50 #include <sys/sysctl.h>
   51 #include <sys/sysent.h>
   52 #include <sys/priv.h>
   53 #include <sys/proc.h>
   54 #include <sys/posix4.h>
   55 #include <sys/time.h>
   56 #include <sys/timers.h>
   57 #include <sys/timetc.h>
   58 #include <sys/vnode.h>
   59 #ifdef KTRACE
   60 #include <sys/ktrace.h>
   61 #endif
   62 
   63 #include <vm/vm.h>
   64 #include <vm/vm_extern.h>
   65 
   66 #define MAX_CLOCKS      (CLOCK_MONOTONIC+1)
   67 #define CPUCLOCK_BIT            0x80000000
   68 #define CPUCLOCK_PROCESS_BIT    0x40000000
   69 #define CPUCLOCK_ID_MASK        (~(CPUCLOCK_BIT|CPUCLOCK_PROCESS_BIT))
   70 #define MAKE_THREAD_CPUCLOCK(tid)       (CPUCLOCK_BIT|(tid))
   71 #define MAKE_PROCESS_CPUCLOCK(pid)      \
   72         (CPUCLOCK_BIT|CPUCLOCK_PROCESS_BIT|(pid))
   73 
   74 static struct kclock    posix_clocks[MAX_CLOCKS];
   75 static uma_zone_t       itimer_zone = NULL;
   76 
   77 /*
   78  * Time of day and interval timer support.
   79  *
   80  * These routines provide the kernel entry points to get and set
   81  * the time-of-day and per-process interval timers.  Subroutines
   82  * here provide support for adding and subtracting timeval structures
   83  * and decrementing interval timers, optionally reloading the interval
   84  * timers when they expire.
   85  */
   86 
   87 static int      settime(struct thread *, struct timeval *);
   88 static void     timevalfix(struct timeval *);
   89 static int      user_clock_nanosleep(struct thread *td, clockid_t clock_id,
   90                     int flags, const struct timespec *ua_rqtp,
   91                     struct timespec *ua_rmtp);
   92 
   93 static void     itimer_start(void);
   94 static int      itimer_init(void *, int, int);
   95 static void     itimer_fini(void *, int);
   96 static void     itimer_enter(struct itimer *);
   97 static void     itimer_leave(struct itimer *);
   98 static struct itimer *itimer_find(struct proc *, int);
   99 static void     itimers_alloc(struct proc *);
  100 static void     itimers_event_hook_exec(void *arg, struct proc *p, struct image_params *imgp);
  101 static void     itimers_event_hook_exit(void *arg, struct proc *p);
  102 static int      realtimer_create(struct itimer *);
  103 static int      realtimer_gettime(struct itimer *, struct itimerspec *);
  104 static int      realtimer_settime(struct itimer *, int,
  105                         struct itimerspec *, struct itimerspec *);
  106 static int      realtimer_delete(struct itimer *);
  107 static void     realtimer_clocktime(clockid_t, struct timespec *);
  108 static void     realtimer_expire(void *);
  109 
  110 int             register_posix_clock(int, struct kclock *);
  111 void            itimer_fire(struct itimer *it);
  112 int             itimespecfix(struct timespec *ts);
  113 
  114 #define CLOCK_CALL(clock, call, arglist)                \
  115         ((*posix_clocks[clock].call) arglist)
  116 
  117 SYSINIT(posix_timer, SI_SUB_P1003_1B, SI_ORDER_FIRST+4, itimer_start, NULL);
  118 
  119 
  120 static int
  121 settime(struct thread *td, struct timeval *tv)
  122 {
  123         struct timeval delta, tv1, tv2;
  124         static struct timeval maxtime, laststep;
  125         struct timespec ts;
  126 
  127         microtime(&tv1);
  128         delta = *tv;
  129         timevalsub(&delta, &tv1);
  130 
  131         /*
  132          * If the system is secure, we do not allow the time to be 
  133          * set to a value earlier than 1 second less than the highest
  134          * time we have yet seen. The worst a miscreant can do in
  135          * this circumstance is "freeze" time. He couldn't go
  136          * back to the past.
  137          *
  138          * We similarly do not allow the clock to be stepped more
  139          * than one second, nor more than once per second. This allows
  140          * a miscreant to make the clock march double-time, but no worse.
  141          */
  142         if (securelevel_gt(td->td_ucred, 1) != 0) {
  143                 if (delta.tv_sec < 0 || delta.tv_usec < 0) {
  144                         /*
  145                          * Update maxtime to latest time we've seen.
  146                          */
  147                         if (tv1.tv_sec > maxtime.tv_sec)
  148                                 maxtime = tv1;
  149                         tv2 = *tv;
  150                         timevalsub(&tv2, &maxtime);
  151                         if (tv2.tv_sec < -1) {
  152                                 tv->tv_sec = maxtime.tv_sec - 1;
  153                                 printf("Time adjustment clamped to -1 second\n");
  154                         }
  155                 } else {
  156                         if (tv1.tv_sec == laststep.tv_sec)
  157                                 return (EPERM);
  158                         if (delta.tv_sec > 1) {
  159                                 tv->tv_sec = tv1.tv_sec + 1;
  160                                 printf("Time adjustment clamped to +1 second\n");
  161                         }
  162                         laststep = *tv;
  163                 }
  164         }
  165 
  166         ts.tv_sec = tv->tv_sec;
  167         ts.tv_nsec = tv->tv_usec * 1000;
  168         tc_setclock(&ts);
  169         resettodr();
  170         return (0);
  171 }
  172 
  173 #ifndef _SYS_SYSPROTO_H_
  174 struct clock_getcpuclockid2_args {
  175         id_t id;
  176         int which,
  177         clockid_t *clock_id;
  178 };
  179 #endif
  180 /* ARGSUSED */
  181 int
  182 sys_clock_getcpuclockid2(struct thread *td, struct clock_getcpuclockid2_args *uap)
  183 {
  184         clockid_t clk_id;
  185         int error;
  186 
  187         error = kern_clock_getcpuclockid2(td, uap->id, uap->which, &clk_id);
  188         if (error == 0)
  189                 error = copyout(&clk_id, uap->clock_id, sizeof(clockid_t));
  190         return (error);
  191 }
  192 
  193 int
  194 kern_clock_getcpuclockid2(struct thread *td, id_t id, int which,
  195     clockid_t *clk_id)
  196 {
  197         struct proc *p;
  198         pid_t pid;
  199         lwpid_t tid;
  200         int error;
  201 
  202         switch (which) {
  203         case CPUCLOCK_WHICH_PID:
  204                 if (id != 0) {
  205                         error = pget(id, PGET_CANSEE | PGET_NOTID, &p);
  206                         if (error != 0)
  207                                 return (error);
  208                         PROC_UNLOCK(p);
  209                         pid = id;
  210                 } else {
  211                         pid = td->td_proc->p_pid;
  212                 }
  213                 *clk_id = MAKE_PROCESS_CPUCLOCK(pid);
  214                 return (0);
  215         case CPUCLOCK_WHICH_TID:
  216                 tid = id == 0 ? td->td_tid : id;
  217                 *clk_id = MAKE_THREAD_CPUCLOCK(tid);
  218                 return (0);
  219         default:
  220                 return (EINVAL);
  221         }
  222 }
  223 
  224 #ifndef _SYS_SYSPROTO_H_
  225 struct clock_gettime_args {
  226         clockid_t clock_id;
  227         struct  timespec *tp;
  228 };
  229 #endif
  230 /* ARGSUSED */
  231 int
  232 sys_clock_gettime(struct thread *td, struct clock_gettime_args *uap)
  233 {
  234         struct timespec ats;
  235         int error;
  236 
  237         error = kern_clock_gettime(td, uap->clock_id, &ats);
  238         if (error == 0)
  239                 error = copyout(&ats, uap->tp, sizeof(ats));
  240 
  241         return (error);
  242 }
  243 
  244 static inline void 
  245 cputick2timespec(uint64_t runtime, struct timespec *ats)
  246 {
  247         runtime = cputick2usec(runtime);
  248         ats->tv_sec = runtime / 1000000;
  249         ats->tv_nsec = runtime % 1000000 * 1000;
  250 }
  251 
  252 static void
  253 get_thread_cputime(struct thread *targettd, struct timespec *ats)
  254 {
  255         uint64_t runtime, curtime, switchtime;
  256 
  257         if (targettd == NULL) { /* current thread */
  258                 critical_enter();
  259                 switchtime = PCPU_GET(switchtime);
  260                 curtime = cpu_ticks();
  261                 runtime = curthread->td_runtime;
  262                 critical_exit();
  263                 runtime += curtime - switchtime;
  264         } else {
  265                 thread_lock(targettd);
  266                 runtime = targettd->td_runtime;
  267                 thread_unlock(targettd);
  268         }
  269         cputick2timespec(runtime, ats);
  270 }
  271 
  272 static void
  273 get_process_cputime(struct proc *targetp, struct timespec *ats)
  274 {
  275         uint64_t runtime;
  276         struct rusage ru;
  277 
  278         PROC_STATLOCK(targetp);
  279         rufetch(targetp, &ru);
  280         runtime = targetp->p_rux.rux_runtime;
  281         PROC_STATUNLOCK(targetp);
  282         cputick2timespec(runtime, ats);
  283 }
  284 
  285 static int
  286 get_cputime(struct thread *td, clockid_t clock_id, struct timespec *ats)
  287 {
  288         struct proc *p, *p2;
  289         struct thread *td2;
  290         lwpid_t tid;
  291         pid_t pid;
  292         int error;
  293 
  294         p = td->td_proc;
  295         if ((clock_id & CPUCLOCK_PROCESS_BIT) == 0) {
  296                 tid = clock_id & CPUCLOCK_ID_MASK;
  297                 td2 = tdfind(tid, p->p_pid);
  298                 if (td2 == NULL)
  299                         return (EINVAL);
  300                 get_thread_cputime(td2, ats);
  301                 PROC_UNLOCK(td2->td_proc);
  302         } else {
  303                 pid = clock_id & CPUCLOCK_ID_MASK;
  304                 error = pget(pid, PGET_CANSEE, &p2);
  305                 if (error != 0)
  306                         return (EINVAL);
  307                 get_process_cputime(p2, ats);
  308                 PROC_UNLOCK(p2);
  309         }
  310         return (0);
  311 }
  312 
  313 int
  314 kern_clock_gettime(struct thread *td, clockid_t clock_id, struct timespec *ats)
  315 {
  316         struct timeval sys, user;
  317         struct proc *p;
  318 
  319         p = td->td_proc;
  320         switch (clock_id) {
  321         case CLOCK_REALTIME:            /* Default to precise. */
  322         case CLOCK_REALTIME_PRECISE:
  323                 nanotime(ats);
  324                 break;
  325         case CLOCK_REALTIME_FAST:
  326                 getnanotime(ats);
  327                 break;
  328         case CLOCK_VIRTUAL:
  329                 PROC_LOCK(p);
  330                 PROC_STATLOCK(p);
  331                 calcru(p, &user, &sys);
  332                 PROC_STATUNLOCK(p);
  333                 PROC_UNLOCK(p);
  334                 TIMEVAL_TO_TIMESPEC(&user, ats);
  335                 break;
  336         case CLOCK_PROF:
  337                 PROC_LOCK(p);
  338                 PROC_STATLOCK(p);
  339                 calcru(p, &user, &sys);
  340                 PROC_STATUNLOCK(p);
  341                 PROC_UNLOCK(p);
  342                 timevaladd(&user, &sys);
  343                 TIMEVAL_TO_TIMESPEC(&user, ats);
  344                 break;
  345         case CLOCK_MONOTONIC:           /* Default to precise. */
  346         case CLOCK_MONOTONIC_PRECISE:
  347         case CLOCK_UPTIME:
  348         case CLOCK_UPTIME_PRECISE:
  349                 nanouptime(ats);
  350                 break;
  351         case CLOCK_UPTIME_FAST:
  352         case CLOCK_MONOTONIC_FAST:
  353                 getnanouptime(ats);
  354                 break;
  355         case CLOCK_SECOND:
  356                 ats->tv_sec = time_second;
  357                 ats->tv_nsec = 0;
  358                 break;
  359         case CLOCK_THREAD_CPUTIME_ID:
  360                 get_thread_cputime(NULL, ats);
  361                 break;
  362         case CLOCK_PROCESS_CPUTIME_ID:
  363                 PROC_LOCK(p);
  364                 get_process_cputime(p, ats);
  365                 PROC_UNLOCK(p);
  366                 break;
  367         default:
  368                 if ((int)clock_id >= 0)
  369                         return (EINVAL);
  370                 return (get_cputime(td, clock_id, ats));
  371         }
  372         return (0);
  373 }
  374 
  375 #ifndef _SYS_SYSPROTO_H_
  376 struct clock_settime_args {
  377         clockid_t clock_id;
  378         const struct    timespec *tp;
  379 };
  380 #endif
  381 /* ARGSUSED */
  382 int
  383 sys_clock_settime(struct thread *td, struct clock_settime_args *uap)
  384 {
  385         struct timespec ats;
  386         int error;
  387 
  388         if ((error = copyin(uap->tp, &ats, sizeof(ats))) != 0)
  389                 return (error);
  390         return (kern_clock_settime(td, uap->clock_id, &ats));
  391 }
  392 
  393 int
  394 kern_clock_settime(struct thread *td, clockid_t clock_id, struct timespec *ats)
  395 {
  396         struct timeval atv;
  397         int error;
  398 
  399         if ((error = priv_check(td, PRIV_CLOCK_SETTIME)) != 0)
  400                 return (error);
  401         if (clock_id != CLOCK_REALTIME)
  402                 return (EINVAL);
  403         if (ats->tv_nsec < 0 || ats->tv_nsec >= 1000000000 ||
  404             ats->tv_sec < 0)
  405                 return (EINVAL);
  406         /* XXX Don't convert nsec->usec and back */
  407         TIMESPEC_TO_TIMEVAL(&atv, ats);
  408         error = settime(td, &atv);
  409         return (error);
  410 }
  411 
  412 #ifndef _SYS_SYSPROTO_H_
  413 struct clock_getres_args {
  414         clockid_t clock_id;
  415         struct  timespec *tp;
  416 };
  417 #endif
  418 int
  419 sys_clock_getres(struct thread *td, struct clock_getres_args *uap)
  420 {
  421         struct timespec ts;
  422         int error;
  423 
  424         if (uap->tp == NULL)
  425                 return (0);
  426 
  427         error = kern_clock_getres(td, uap->clock_id, &ts);
  428         if (error == 0)
  429                 error = copyout(&ts, uap->tp, sizeof(ts));
  430         return (error);
  431 }
  432 
  433 int
  434 kern_clock_getres(struct thread *td, clockid_t clock_id, struct timespec *ts)
  435 {
  436 
  437         ts->tv_sec = 0;
  438         switch (clock_id) {
  439         case CLOCK_REALTIME:
  440         case CLOCK_REALTIME_FAST:
  441         case CLOCK_REALTIME_PRECISE:
  442         case CLOCK_MONOTONIC:
  443         case CLOCK_MONOTONIC_FAST:
  444         case CLOCK_MONOTONIC_PRECISE:
  445         case CLOCK_UPTIME:
  446         case CLOCK_UPTIME_FAST:
  447         case CLOCK_UPTIME_PRECISE:
  448                 /*
  449                  * Round up the result of the division cheaply by adding 1.
  450                  * Rounding up is especially important if rounding down
  451                  * would give 0.  Perfect rounding is unimportant.
  452                  */
  453                 ts->tv_nsec = 1000000000 / tc_getfrequency() + 1;
  454                 break;
  455         case CLOCK_VIRTUAL:
  456         case CLOCK_PROF:
  457                 /* Accurately round up here because we can do so cheaply. */
  458                 ts->tv_nsec = howmany(1000000000, hz);
  459                 break;
  460         case CLOCK_SECOND:
  461                 ts->tv_sec = 1;
  462                 ts->tv_nsec = 0;
  463                 break;
  464         case CLOCK_THREAD_CPUTIME_ID:
  465         case CLOCK_PROCESS_CPUTIME_ID:
  466         cputime:
  467                 /* sync with cputick2usec */
  468                 ts->tv_nsec = 1000000 / cpu_tickrate();
  469                 if (ts->tv_nsec == 0)
  470                         ts->tv_nsec = 1000;
  471                 break;
  472         default:
  473                 if ((int)clock_id < 0)
  474                         goto cputime;
  475                 return (EINVAL);
  476         }
  477         return (0);
  478 }
  479 
  480 int
  481 kern_nanosleep(struct thread *td, struct timespec *rqt, struct timespec *rmt)
  482 {
  483 
  484         return (kern_clock_nanosleep(td, CLOCK_REALTIME, TIMER_RELTIME, rqt,
  485             rmt));
  486 }
  487 
  488 static uint8_t nanowait[MAXCPU];
  489 
  490 int
  491 kern_clock_nanosleep(struct thread *td, clockid_t clock_id, int flags,
  492     const struct timespec *rqt, struct timespec *rmt)
  493 {
  494         struct timespec ts, now;
  495         sbintime_t sbt, sbtt, prec, tmp;
  496         time_t over;
  497         int error;
  498         bool is_abs_real;
  499 
  500         if (rqt->tv_nsec < 0 || rqt->tv_nsec >= 1000000000)
  501                 return (EINVAL);
  502         if ((flags & ~TIMER_ABSTIME) != 0)
  503                 return (EINVAL);
  504         switch (clock_id) {
  505         case CLOCK_REALTIME:
  506         case CLOCK_REALTIME_PRECISE:
  507         case CLOCK_REALTIME_FAST:
  508         case CLOCK_SECOND:
  509                 is_abs_real = (flags & TIMER_ABSTIME) != 0;
  510                 break;
  511         case CLOCK_MONOTONIC:
  512         case CLOCK_MONOTONIC_PRECISE:
  513         case CLOCK_MONOTONIC_FAST:
  514         case CLOCK_UPTIME:
  515         case CLOCK_UPTIME_PRECISE:
  516         case CLOCK_UPTIME_FAST:
  517                 is_abs_real = false;
  518                 break;
  519         case CLOCK_VIRTUAL:
  520         case CLOCK_PROF:
  521         case CLOCK_PROCESS_CPUTIME_ID:
  522                 return (ENOTSUP);
  523         case CLOCK_THREAD_CPUTIME_ID:
  524         default:
  525                 return (EINVAL);
  526         }
  527         do {
  528                 ts = *rqt;
  529                 if ((flags & TIMER_ABSTIME) != 0) {
  530                         if (is_abs_real)
  531                                 td->td_rtcgen =
  532                                     atomic_load_acq_int(&rtc_generation);
  533                         error = kern_clock_gettime(td, clock_id, &now);
  534                         KASSERT(error == 0, ("kern_clock_gettime: %d", error));
  535                         timespecsub(&ts, &now);
  536                 }
  537                 if (ts.tv_sec < 0 || (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
  538                         error = EWOULDBLOCK;
  539                         break;
  540                 }
  541                 if (ts.tv_sec > INT32_MAX / 2) {
  542                         over = ts.tv_sec - INT32_MAX / 2;
  543                         ts.tv_sec -= over;
  544                 } else
  545                         over = 0;
  546                 tmp = tstosbt(ts);
  547                 prec = tmp;
  548                 prec >>= tc_precexp;
  549                 if (TIMESEL(&sbt, tmp))
  550                         sbt += tc_tick_sbt;
  551                 sbt += tmp;
  552                 error = tsleep_sbt(&nanowait[curcpu], PWAIT | PCATCH, "nanslp",
  553                     sbt, prec, C_ABSOLUTE);
  554         } while (error == 0 && is_abs_real && td->td_rtcgen == 0);
  555         td->td_rtcgen = 0;
  556         if (error != EWOULDBLOCK) {
  557                 if (TIMESEL(&sbtt, tmp))
  558                         sbtt += tc_tick_sbt;
  559                 if (sbtt >= sbt)
  560                         return (0);
  561                 if (error == ERESTART)
  562                         error = EINTR;
  563                 if ((flags & TIMER_ABSTIME) == 0 && rmt != NULL) {
  564                         ts = sbttots(sbt - sbtt);
  565                         ts.tv_sec += over;
  566                         if (ts.tv_sec < 0)
  567                                 timespecclear(&ts);
  568                         *rmt = ts;
  569                 }
  570                 return (error);
  571         }
  572         return (0);
  573 }
  574 
  575 #ifndef _SYS_SYSPROTO_H_
  576 struct nanosleep_args {
  577         struct  timespec *rqtp;
  578         struct  timespec *rmtp;
  579 };
  580 #endif
  581 /* ARGSUSED */
  582 int
  583 sys_nanosleep(struct thread *td, struct nanosleep_args *uap)
  584 {
  585 
  586         return (user_clock_nanosleep(td, CLOCK_REALTIME, TIMER_RELTIME,
  587             uap->rqtp, uap->rmtp));
  588 }
  589 
  590 #ifndef _SYS_SYSPROTO_H_
  591 struct clock_nanosleep_args {
  592         clockid_t clock_id;
  593         int       flags;
  594         struct  timespec *rqtp;
  595         struct  timespec *rmtp;
  596 };
  597 #endif
  598 /* ARGSUSED */
  599 int
  600 sys_clock_nanosleep(struct thread *td, struct clock_nanosleep_args *uap)
  601 {
  602         int error;
  603 
  604         error = user_clock_nanosleep(td, uap->clock_id, uap->flags, uap->rqtp,
  605             uap->rmtp);
  606         return (kern_posix_error(td, error));
  607 }
  608 
  609 static int
  610 user_clock_nanosleep(struct thread *td, clockid_t clock_id, int flags,
  611     const struct timespec *ua_rqtp, struct timespec *ua_rmtp)
  612 {
  613         struct timespec rmt, rqt;
  614         int error;
  615 
  616         error = copyin(ua_rqtp, &rqt, sizeof(rqt));
  617         if (error)
  618                 return (error);
  619         if (ua_rmtp != NULL && (flags & TIMER_ABSTIME) == 0 &&
  620             !useracc(ua_rmtp, sizeof(rmt), VM_PROT_WRITE))
  621                 return (EFAULT);
  622         error = kern_clock_nanosleep(td, clock_id, flags, &rqt, &rmt);
  623         if (error == EINTR && ua_rmtp != NULL && (flags & TIMER_ABSTIME) == 0) {
  624                 int error2;
  625 
  626                 error2 = copyout(&rmt, ua_rmtp, sizeof(rmt));
  627                 if (error2)
  628                         error = error2;
  629         }
  630         return (error);
  631 }
  632 
  633 #ifndef _SYS_SYSPROTO_H_
  634 struct gettimeofday_args {
  635         struct  timeval *tp;
  636         struct  timezone *tzp;
  637 };
  638 #endif
  639 /* ARGSUSED */
  640 int
  641 sys_gettimeofday(struct thread *td, struct gettimeofday_args *uap)
  642 {
  643         struct timeval atv;
  644         struct timezone rtz;
  645         int error = 0;
  646 
  647         if (uap->tp) {
  648                 microtime(&atv);
  649                 error = copyout(&atv, uap->tp, sizeof (atv));
  650         }
  651         if (error == 0 && uap->tzp != NULL) {
  652                 rtz.tz_minuteswest = tz_minuteswest;
  653                 rtz.tz_dsttime = tz_dsttime;
  654                 error = copyout(&rtz, uap->tzp, sizeof (rtz));
  655         }
  656         return (error);
  657 }
  658 
  659 #ifndef _SYS_SYSPROTO_H_
  660 struct settimeofday_args {
  661         struct  timeval *tv;
  662         struct  timezone *tzp;
  663 };
  664 #endif
  665 /* ARGSUSED */
  666 int
  667 sys_settimeofday(struct thread *td, struct settimeofday_args *uap)
  668 {
  669         struct timeval atv, *tvp;
  670         struct timezone atz, *tzp;
  671         int error;
  672 
  673         if (uap->tv) {
  674                 error = copyin(uap->tv, &atv, sizeof(atv));
  675                 if (error)
  676                         return (error);
  677                 tvp = &atv;
  678         } else
  679                 tvp = NULL;
  680         if (uap->tzp) {
  681                 error = copyin(uap->tzp, &atz, sizeof(atz));
  682                 if (error)
  683                         return (error);
  684                 tzp = &atz;
  685         } else
  686                 tzp = NULL;
  687         return (kern_settimeofday(td, tvp, tzp));
  688 }
  689 
  690 int
  691 kern_settimeofday(struct thread *td, struct timeval *tv, struct timezone *tzp)
  692 {
  693         int error;
  694 
  695         error = priv_check(td, PRIV_SETTIMEOFDAY);
  696         if (error)
  697                 return (error);
  698         /* Verify all parameters before changing time. */
  699         if (tv) {
  700                 if (tv->tv_usec < 0 || tv->tv_usec >= 1000000 ||
  701                     tv->tv_sec < 0)
  702                         return (EINVAL);
  703                 error = settime(td, tv);
  704         }
  705         if (tzp && error == 0) {
  706                 tz_minuteswest = tzp->tz_minuteswest;
  707                 tz_dsttime = tzp->tz_dsttime;
  708         }
  709         return (error);
  710 }
  711 
  712 /*
  713  * Get value of an interval timer.  The process virtual and profiling virtual
  714  * time timers are kept in the p_stats area, since they can be swapped out.
  715  * These are kept internally in the way they are specified externally: in
  716  * time until they expire.
  717  *
  718  * The real time interval timer is kept in the process table slot for the
  719  * process, and its value (it_value) is kept as an absolute time rather than
  720  * as a delta, so that it is easy to keep periodic real-time signals from
  721  * drifting.
  722  *
  723  * Virtual time timers are processed in the hardclock() routine of
  724  * kern_clock.c.  The real time timer is processed by a timeout routine,
  725  * called from the softclock() routine.  Since a callout may be delayed in
  726  * real time due to interrupt processing in the system, it is possible for
  727  * the real time timeout routine (realitexpire, given below), to be delayed
  728  * in real time past when it is supposed to occur.  It does not suffice,
  729  * therefore, to reload the real timer .it_value from the real time timers
  730  * .it_interval.  Rather, we compute the next time in absolute time the timer
  731  * should go off.
  732  */
  733 #ifndef _SYS_SYSPROTO_H_
  734 struct getitimer_args {
  735         u_int   which;
  736         struct  itimerval *itv;
  737 };
  738 #endif
  739 int
  740 sys_getitimer(struct thread *td, struct getitimer_args *uap)
  741 {
  742         struct itimerval aitv;
  743         int error;
  744 
  745         error = kern_getitimer(td, uap->which, &aitv);
  746         if (error != 0)
  747                 return (error);
  748         return (copyout(&aitv, uap->itv, sizeof (struct itimerval)));
  749 }
  750 
  751 int
  752 kern_getitimer(struct thread *td, u_int which, struct itimerval *aitv)
  753 {
  754         struct proc *p = td->td_proc;
  755         struct timeval ctv;
  756 
  757         if (which > ITIMER_PROF)
  758                 return (EINVAL);
  759 
  760         if (which == ITIMER_REAL) {
  761                 /*
  762                  * Convert from absolute to relative time in .it_value
  763                  * part of real time timer.  If time for real time timer
  764                  * has passed return 0, else return difference between
  765                  * current time and time for the timer to go off.
  766                  */
  767                 PROC_LOCK(p);
  768                 *aitv = p->p_realtimer;
  769                 PROC_UNLOCK(p);
  770                 if (timevalisset(&aitv->it_value)) {
  771                         microuptime(&ctv);
  772                         if (timevalcmp(&aitv->it_value, &ctv, <))
  773                                 timevalclear(&aitv->it_value);
  774                         else
  775                                 timevalsub(&aitv->it_value, &ctv);
  776                 }
  777         } else {
  778                 PROC_ITIMLOCK(p);
  779                 *aitv = p->p_stats->p_timer[which];
  780                 PROC_ITIMUNLOCK(p);
  781         }
  782 #ifdef KTRACE
  783         if (KTRPOINT(td, KTR_STRUCT))
  784                 ktritimerval(aitv);
  785 #endif
  786         return (0);
  787 }
  788 
  789 #ifndef _SYS_SYSPROTO_H_
  790 struct setitimer_args {
  791         u_int   which;
  792         struct  itimerval *itv, *oitv;
  793 };
  794 #endif
  795 int
  796 sys_setitimer(struct thread *td, struct setitimer_args *uap)
  797 {
  798         struct itimerval aitv, oitv;
  799         int error;
  800 
  801         if (uap->itv == NULL) {
  802                 uap->itv = uap->oitv;
  803                 return (sys_getitimer(td, (struct getitimer_args *)uap));
  804         }
  805 
  806         if ((error = copyin(uap->itv, &aitv, sizeof(struct itimerval))))
  807                 return (error);
  808         error = kern_setitimer(td, uap->which, &aitv, &oitv);
  809         if (error != 0 || uap->oitv == NULL)
  810                 return (error);
  811         return (copyout(&oitv, uap->oitv, sizeof(struct itimerval)));
  812 }
  813 
  814 int
  815 kern_setitimer(struct thread *td, u_int which, struct itimerval *aitv,
  816     struct itimerval *oitv)
  817 {
  818         struct proc *p = td->td_proc;
  819         struct timeval ctv;
  820         sbintime_t sbt, pr;
  821 
  822         if (aitv == NULL)
  823                 return (kern_getitimer(td, which, oitv));
  824 
  825         if (which > ITIMER_PROF)
  826                 return (EINVAL);
  827 #ifdef KTRACE
  828         if (KTRPOINT(td, KTR_STRUCT))
  829                 ktritimerval(aitv);
  830 #endif
  831         if (itimerfix(&aitv->it_value) ||
  832             aitv->it_value.tv_sec > INT32_MAX / 2)
  833                 return (EINVAL);
  834         if (!timevalisset(&aitv->it_value))
  835                 timevalclear(&aitv->it_interval);
  836         else if (itimerfix(&aitv->it_interval) ||
  837             aitv->it_interval.tv_sec > INT32_MAX / 2)
  838                 return (EINVAL);
  839 
  840         if (which == ITIMER_REAL) {
  841                 PROC_LOCK(p);
  842                 if (timevalisset(&p->p_realtimer.it_value))
  843                         callout_stop(&p->p_itcallout);
  844                 microuptime(&ctv);
  845                 if (timevalisset(&aitv->it_value)) {
  846                         pr = tvtosbt(aitv->it_value) >> tc_precexp;
  847                         timevaladd(&aitv->it_value, &ctv);
  848                         sbt = tvtosbt(aitv->it_value);
  849                         callout_reset_sbt(&p->p_itcallout, sbt, pr,
  850                             realitexpire, p, C_ABSOLUTE);
  851                 }
  852                 *oitv = p->p_realtimer;
  853                 p->p_realtimer = *aitv;
  854                 PROC_UNLOCK(p);
  855                 if (timevalisset(&oitv->it_value)) {
  856                         if (timevalcmp(&oitv->it_value, &ctv, <))
  857                                 timevalclear(&oitv->it_value);
  858                         else
  859                                 timevalsub(&oitv->it_value, &ctv);
  860                 }
  861         } else {
  862                 if (aitv->it_interval.tv_sec == 0 &&
  863                     aitv->it_interval.tv_usec != 0 &&
  864                     aitv->it_interval.tv_usec < tick)
  865                         aitv->it_interval.tv_usec = tick;
  866                 if (aitv->it_value.tv_sec == 0 &&
  867                     aitv->it_value.tv_usec != 0 &&
  868                     aitv->it_value.tv_usec < tick)
  869                         aitv->it_value.tv_usec = tick;
  870                 PROC_ITIMLOCK(p);
  871                 *oitv = p->p_stats->p_timer[which];
  872                 p->p_stats->p_timer[which] = *aitv;
  873                 PROC_ITIMUNLOCK(p);
  874         }
  875 #ifdef KTRACE
  876         if (KTRPOINT(td, KTR_STRUCT))
  877                 ktritimerval(oitv);
  878 #endif
  879         return (0);
  880 }
  881 
  882 /*
  883  * Real interval timer expired:
  884  * send process whose timer expired an alarm signal.
  885  * If time is not set up to reload, then just return.
  886  * Else compute next time timer should go off which is > current time.
  887  * This is where delay in processing this timeout causes multiple
  888  * SIGALRM calls to be compressed into one.
  889  * tvtohz() always adds 1 to allow for the time until the next clock
  890  * interrupt being strictly less than 1 clock tick, but we don't want
  891  * that here since we want to appear to be in sync with the clock
  892  * interrupt even when we're delayed.
  893  */
  894 void
  895 realitexpire(void *arg)
  896 {
  897         struct proc *p;
  898         struct timeval ctv;
  899         sbintime_t isbt;
  900 
  901         p = (struct proc *)arg;
  902         kern_psignal(p, SIGALRM);
  903         if (!timevalisset(&p->p_realtimer.it_interval)) {
  904                 timevalclear(&p->p_realtimer.it_value);
  905                 if (p->p_flag & P_WEXIT)
  906                         wakeup(&p->p_itcallout);
  907                 return;
  908         }
  909         isbt = tvtosbt(p->p_realtimer.it_interval);
  910         if (isbt >= sbt_timethreshold)
  911                 getmicrouptime(&ctv);
  912         else
  913                 microuptime(&ctv);
  914         do {
  915                 timevaladd(&p->p_realtimer.it_value,
  916                     &p->p_realtimer.it_interval);
  917         } while (timevalcmp(&p->p_realtimer.it_value, &ctv, <=));
  918         callout_reset_sbt(&p->p_itcallout, tvtosbt(p->p_realtimer.it_value),
  919             isbt >> tc_precexp, realitexpire, p, C_ABSOLUTE);
  920 }
  921 
  922 /*
  923  * Check that a proposed value to load into the .it_value or
  924  * .it_interval part of an interval timer is acceptable, and
  925  * fix it to have at least minimal value (i.e. if it is less
  926  * than the resolution of the clock, round it up.)
  927  */
  928 int
  929 itimerfix(struct timeval *tv)
  930 {
  931 
  932         if (tv->tv_sec < 0 || tv->tv_usec < 0 || tv->tv_usec >= 1000000)
  933                 return (EINVAL);
  934         if (tv->tv_sec == 0 && tv->tv_usec != 0 &&
  935             tv->tv_usec < (u_int)tick / 16)
  936                 tv->tv_usec = (u_int)tick / 16;
  937         return (0);
  938 }
  939 
  940 /*
  941  * Decrement an interval timer by a specified number
  942  * of microseconds, which must be less than a second,
  943  * i.e. < 1000000.  If the timer expires, then reload
  944  * it.  In this case, carry over (usec - old value) to
  945  * reduce the value reloaded into the timer so that
  946  * the timer does not drift.  This routine assumes
  947  * that it is called in a context where the timers
  948  * on which it is operating cannot change in value.
  949  */
  950 int
  951 itimerdecr(struct itimerval *itp, int usec)
  952 {
  953 
  954         if (itp->it_value.tv_usec < usec) {
  955                 if (itp->it_value.tv_sec == 0) {
  956                         /* expired, and already in next interval */
  957                         usec -= itp->it_value.tv_usec;
  958                         goto expire;
  959                 }
  960                 itp->it_value.tv_usec += 1000000;
  961                 itp->it_value.tv_sec--;
  962         }
  963         itp->it_value.tv_usec -= usec;
  964         usec = 0;
  965         if (timevalisset(&itp->it_value))
  966                 return (1);
  967         /* expired, exactly at end of interval */
  968 expire:
  969         if (timevalisset(&itp->it_interval)) {
  970                 itp->it_value = itp->it_interval;
  971                 itp->it_value.tv_usec -= usec;
  972                 if (itp->it_value.tv_usec < 0) {
  973                         itp->it_value.tv_usec += 1000000;
  974                         itp->it_value.tv_sec--;
  975                 }
  976         } else
  977                 itp->it_value.tv_usec = 0;              /* sec is already 0 */
  978         return (0);
  979 }
  980 
  981 /*
  982  * Add and subtract routines for timevals.
  983  * N.B.: subtract routine doesn't deal with
  984  * results which are before the beginning,
  985  * it just gets very confused in this case.
  986  * Caveat emptor.
  987  */
  988 void
  989 timevaladd(struct timeval *t1, const struct timeval *t2)
  990 {
  991 
  992         t1->tv_sec += t2->tv_sec;
  993         t1->tv_usec += t2->tv_usec;
  994         timevalfix(t1);
  995 }
  996 
  997 void
  998 timevalsub(struct timeval *t1, const struct timeval *t2)
  999 {
 1000 
 1001         t1->tv_sec -= t2->tv_sec;
 1002         t1->tv_usec -= t2->tv_usec;
 1003         timevalfix(t1);
 1004 }
 1005 
 1006 static void
 1007 timevalfix(struct timeval *t1)
 1008 {
 1009 
 1010         if (t1->tv_usec < 0) {
 1011                 t1->tv_sec--;
 1012                 t1->tv_usec += 1000000;
 1013         }
 1014         if (t1->tv_usec >= 1000000) {
 1015                 t1->tv_sec++;
 1016                 t1->tv_usec -= 1000000;
 1017         }
 1018 }
 1019 
 1020 /*
 1021  * ratecheck(): simple time-based rate-limit checking.
 1022  */
 1023 int
 1024 ratecheck(struct timeval *lasttime, const struct timeval *mininterval)
 1025 {
 1026         struct timeval tv, delta;
 1027         int rv = 0;
 1028 
 1029         getmicrouptime(&tv);            /* NB: 10ms precision */
 1030         delta = tv;
 1031         timevalsub(&delta, lasttime);
 1032 
 1033         /*
 1034          * check for 0,0 is so that the message will be seen at least once,
 1035          * even if interval is huge.
 1036          */
 1037         if (timevalcmp(&delta, mininterval, >=) ||
 1038             (lasttime->tv_sec == 0 && lasttime->tv_usec == 0)) {
 1039                 *lasttime = tv;
 1040                 rv = 1;
 1041         }
 1042 
 1043         return (rv);
 1044 }
 1045 
 1046 /*
 1047  * ppsratecheck(): packets (or events) per second limitation.
 1048  *
 1049  * Return 0 if the limit is to be enforced (e.g. the caller
 1050  * should drop a packet because of the rate limitation).
 1051  *
 1052  * maxpps of 0 always causes zero to be returned.  maxpps of -1
 1053  * always causes 1 to be returned; this effectively defeats rate
 1054  * limiting.
 1055  *
 1056  * Note that we maintain the struct timeval for compatibility
 1057  * with other bsd systems.  We reuse the storage and just monitor
 1058  * clock ticks for minimal overhead.  
 1059  */
 1060 int
 1061 ppsratecheck(struct timeval *lasttime, int *curpps, int maxpps)
 1062 {
 1063         int now;
 1064 
 1065         /*
 1066          * Reset the last time and counter if this is the first call
 1067          * or more than a second has passed since the last update of
 1068          * lasttime.
 1069          */
 1070         now = ticks;
 1071         if (lasttime->tv_sec == 0 || (u_int)(now - lasttime->tv_sec) >= hz) {
 1072                 lasttime->tv_sec = now;
 1073                 *curpps = 1;
 1074                 return (maxpps != 0);
 1075         } else {
 1076                 (*curpps)++;            /* NB: ignore potential overflow */
 1077                 return (maxpps < 0 || *curpps <= maxpps);
 1078         }
 1079 }
 1080 
 1081 static void
 1082 itimer_start(void)
 1083 {
 1084         struct kclock rt_clock = {
 1085                 .timer_create  = realtimer_create,
 1086                 .timer_delete  = realtimer_delete,
 1087                 .timer_settime = realtimer_settime,
 1088                 .timer_gettime = realtimer_gettime,
 1089                 .event_hook    = NULL
 1090         };
 1091 
 1092         itimer_zone = uma_zcreate("itimer", sizeof(struct itimer),
 1093                 NULL, NULL, itimer_init, itimer_fini, UMA_ALIGN_PTR, 0);
 1094         register_posix_clock(CLOCK_REALTIME,  &rt_clock);
 1095         register_posix_clock(CLOCK_MONOTONIC, &rt_clock);
 1096         p31b_setcfg(CTL_P1003_1B_TIMERS, 200112L);
 1097         p31b_setcfg(CTL_P1003_1B_DELAYTIMER_MAX, INT_MAX);
 1098         p31b_setcfg(CTL_P1003_1B_TIMER_MAX, TIMER_MAX);
 1099         EVENTHANDLER_REGISTER(process_exit, itimers_event_hook_exit,
 1100                 (void *)ITIMER_EV_EXIT, EVENTHANDLER_PRI_ANY);
 1101         EVENTHANDLER_REGISTER(process_exec, itimers_event_hook_exec,
 1102                 (void *)ITIMER_EV_EXEC, EVENTHANDLER_PRI_ANY);
 1103 }
 1104 
 1105 int
 1106 register_posix_clock(int clockid, 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);
 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                 } else {
 1555                         timespecsub(&ts, &cts);
 1556                         /*
 1557                          * We don't care if ts is negative, tztohz will
 1558                          * fix it.
 1559                          */
 1560                 }
 1561                 TIMESPEC_TO_TIMEVAL(&tv, &ts);
 1562                 callout_reset(&it->it_callout, tvtohz(&tv),
 1563                         realtimer_expire, it);
 1564         } else {
 1565                 callout_stop(&it->it_callout);
 1566         }
 1567 
 1568         return (0);
 1569 }
 1570 
 1571 static void
 1572 realtimer_clocktime(clockid_t id, struct timespec *ts)
 1573 {
 1574         if (id == CLOCK_REALTIME)
 1575                 getnanotime(ts);
 1576         else    /* CLOCK_MONOTONIC */
 1577                 getnanouptime(ts);
 1578 }
 1579 
 1580 int
 1581 itimer_accept(struct proc *p, int timerid, ksiginfo_t *ksi)
 1582 {
 1583         struct itimer *it;
 1584 
 1585         PROC_LOCK_ASSERT(p, MA_OWNED);
 1586         it = itimer_find(p, timerid);
 1587         if (it != NULL) {
 1588                 ksi->ksi_overrun = it->it_overrun;
 1589                 it->it_overrun_last = it->it_overrun;
 1590                 it->it_overrun = 0;
 1591                 ITIMER_UNLOCK(it);
 1592                 return (0);
 1593         }
 1594         return (EINVAL);
 1595 }
 1596 
 1597 int
 1598 itimespecfix(struct timespec *ts)
 1599 {
 1600 
 1601         if (ts->tv_sec < 0 || ts->tv_nsec < 0 || ts->tv_nsec >= 1000000000)
 1602                 return (EINVAL);
 1603         if (ts->tv_sec == 0 && ts->tv_nsec != 0 && ts->tv_nsec < tick * 1000)
 1604                 ts->tv_nsec = tick * 1000;
 1605         return (0);
 1606 }
 1607 
 1608 /* Timeout callback for realtime timer */
 1609 static void
 1610 realtimer_expire(void *arg)
 1611 {
 1612         struct timespec cts, ts;
 1613         struct timeval tv;
 1614         struct itimer *it;
 1615 
 1616         it = (struct itimer *)arg;
 1617 
 1618         realtimer_clocktime(it->it_clockid, &cts);
 1619         /* Only fire if time is reached. */
 1620         if (timespeccmp(&cts, &it->it_time.it_value, >=)) {
 1621                 if (timespecisset(&it->it_time.it_interval)) {
 1622                         timespecadd(&it->it_time.it_value,
 1623                                     &it->it_time.it_interval);
 1624                         while (timespeccmp(&cts, &it->it_time.it_value, >=)) {
 1625                                 if (it->it_overrun < INT_MAX)
 1626                                         it->it_overrun++;
 1627                                 else
 1628                                         it->it_ksi.ksi_errno = ERANGE;
 1629                                 timespecadd(&it->it_time.it_value,
 1630                                             &it->it_time.it_interval);
 1631                         }
 1632                 } else {
 1633                         /* single shot timer ? */
 1634                         timespecclear(&it->it_time.it_value);
 1635                 }
 1636                 if (timespecisset(&it->it_time.it_value)) {
 1637                         ts = it->it_time.it_value;
 1638                         timespecsub(&ts, &cts);
 1639                         TIMESPEC_TO_TIMEVAL(&tv, &ts);
 1640                         callout_reset(&it->it_callout, tvtohz(&tv),
 1641                                  realtimer_expire, it);
 1642                 }
 1643                 itimer_enter(it);
 1644                 ITIMER_UNLOCK(it);
 1645                 itimer_fire(it);
 1646                 ITIMER_LOCK(it);
 1647                 itimer_leave(it);
 1648         } else if (timespecisset(&it->it_time.it_value)) {
 1649                 ts = it->it_time.it_value;
 1650                 timespecsub(&ts, &cts);
 1651                 TIMESPEC_TO_TIMEVAL(&tv, &ts);
 1652                 callout_reset(&it->it_callout, tvtohz(&tv), realtimer_expire,
 1653                         it);
 1654         }
 1655 }
 1656 
 1657 void
 1658 itimer_fire(struct itimer *it)
 1659 {
 1660         struct proc *p = it->it_proc;
 1661         struct thread *td;
 1662 
 1663         if (it->it_sigev.sigev_notify == SIGEV_SIGNAL ||
 1664             it->it_sigev.sigev_notify == SIGEV_THREAD_ID) {
 1665                 if (sigev_findtd(p, &it->it_sigev, &td) != 0) {
 1666                         ITIMER_LOCK(it);
 1667                         timespecclear(&it->it_time.it_value);
 1668                         timespecclear(&it->it_time.it_interval);
 1669                         callout_stop(&it->it_callout);
 1670                         ITIMER_UNLOCK(it);
 1671                         return;
 1672                 }
 1673                 if (!KSI_ONQ(&it->it_ksi)) {
 1674                         it->it_ksi.ksi_errno = 0;
 1675                         ksiginfo_set_sigev(&it->it_ksi, &it->it_sigev);
 1676                         tdsendsignal(p, td, it->it_ksi.ksi_signo, &it->it_ksi);
 1677                 } else {
 1678                         if (it->it_overrun < INT_MAX)
 1679                                 it->it_overrun++;
 1680                         else
 1681                                 it->it_ksi.ksi_errno = ERANGE;
 1682                 }
 1683                 PROC_UNLOCK(p);
 1684         }
 1685 }
 1686 
 1687 static void
 1688 itimers_alloc(struct proc *p)
 1689 {
 1690         struct itimers *its;
 1691         int i;
 1692 
 1693         its = malloc(sizeof (struct itimers), M_SUBPROC, M_WAITOK | M_ZERO);
 1694         LIST_INIT(&its->its_virtual);
 1695         LIST_INIT(&its->its_prof);
 1696         TAILQ_INIT(&its->its_worklist);
 1697         for (i = 0; i < TIMER_MAX; i++)
 1698                 its->its_timers[i] = NULL;
 1699         PROC_LOCK(p);
 1700         if (p->p_itimers == NULL) {
 1701                 p->p_itimers = its;
 1702                 PROC_UNLOCK(p);
 1703         }
 1704         else {
 1705                 PROC_UNLOCK(p);
 1706                 free(its, M_SUBPROC);
 1707         }
 1708 }
 1709 
 1710 static void
 1711 itimers_event_hook_exec(void *arg, struct proc *p, struct image_params *imgp __unused)
 1712 {
 1713         itimers_event_hook_exit(arg, p);
 1714 }
 1715 
 1716 /* Clean up timers when some process events are being triggered. */
 1717 static void
 1718 itimers_event_hook_exit(void *arg, struct proc *p)
 1719 {
 1720         struct itimers *its;
 1721         struct itimer *it;
 1722         int event = (int)(intptr_t)arg;
 1723         int i;
 1724 
 1725         if (p->p_itimers != NULL) {
 1726                 its = p->p_itimers;
 1727                 for (i = 0; i < MAX_CLOCKS; ++i) {
 1728                         if (posix_clocks[i].event_hook != NULL)
 1729                                 CLOCK_CALL(i, event_hook, (p, i, event));
 1730                 }
 1731                 /*
 1732                  * According to susv3, XSI interval timers should be inherited
 1733                  * by new image.
 1734                  */
 1735                 if (event == ITIMER_EV_EXEC)
 1736                         i = 3;
 1737                 else if (event == ITIMER_EV_EXIT)
 1738                         i = 0;
 1739                 else
 1740                         panic("unhandled event");
 1741                 for (; i < TIMER_MAX; ++i) {
 1742                         if ((it = its->its_timers[i]) != NULL)
 1743                                 kern_ktimer_delete(curthread, i);
 1744                 }
 1745                 if (its->its_timers[0] == NULL &&
 1746                     its->its_timers[1] == NULL &&
 1747                     its->its_timers[2] == NULL) {
 1748                         free(its, M_SUBPROC);
 1749                         p->p_itimers = NULL;
 1750                 }
 1751         }
 1752 }

Cache object: f146df81447811b5f814803d5083c9db


[ 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.