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_clock.c

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1982, 1986, 1991, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  * (c) UNIX System Laboratories, Inc.
    5  * All or some portions of this file are derived from material licensed
    6  * to the University of California by American Telephone and Telegraph
    7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
    8  * the permission of UNIX System Laboratories, Inc.
    9  *
   10  * Redistribution and use in source and binary forms, with or without
   11  * modification, are permitted provided that the following conditions
   12  * are met:
   13  * 1. Redistributions of source code must retain the above copyright
   14  *    notice, this list of conditions and the following disclaimer.
   15  * 2. Redistributions in binary form must reproduce the above copyright
   16  *    notice, this list of conditions and the following disclaimer in the
   17  *    documentation and/or other materials provided with the distribution.
   18  * 4. Neither the name of the University nor the names of its contributors
   19  *    may be used to endorse or promote products derived from this software
   20  *    without specific prior written permission.
   21  *
   22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   32  * SUCH DAMAGE.
   33  *
   34  *      @(#)kern_clock.c        8.5 (Berkeley) 1/21/94
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD: releng/8.1/sys/kern/kern_clock.c 207138 2010-04-24 00:53:41Z attilio $");
   39 
   40 #include "opt_kdb.h"
   41 #include "opt_device_polling.h"
   42 #include "opt_hwpmc_hooks.h"
   43 #include "opt_ntp.h"
   44 #include "opt_watchdog.h"
   45 
   46 #include <sys/param.h>
   47 #include <sys/systm.h>
   48 #include <sys/callout.h>
   49 #include <sys/kdb.h>
   50 #include <sys/kernel.h>
   51 #include <sys/kthread.h>
   52 #include <sys/ktr.h>
   53 #include <sys/lock.h>
   54 #include <sys/mutex.h>
   55 #include <sys/proc.h>
   56 #include <sys/resource.h>
   57 #include <sys/resourcevar.h>
   58 #include <sys/sched.h>
   59 #include <sys/signalvar.h>
   60 #include <sys/sleepqueue.h>
   61 #include <sys/smp.h>
   62 #include <vm/vm.h>
   63 #include <vm/pmap.h>
   64 #include <vm/vm_map.h>
   65 #include <sys/sysctl.h>
   66 #include <sys/bus.h>
   67 #include <sys/interrupt.h>
   68 #include <sys/limits.h>
   69 #include <sys/timetc.h>
   70 
   71 #ifdef GPROF
   72 #include <sys/gmon.h>
   73 #endif
   74 
   75 #ifdef HWPMC_HOOKS
   76 #include <sys/pmckern.h>
   77 #endif
   78 
   79 #ifdef DEVICE_POLLING
   80 extern void hardclock_device_poll(void);
   81 #endif /* DEVICE_POLLING */
   82 
   83 static void initclocks(void *dummy);
   84 SYSINIT(clocks, SI_SUB_CLOCKS, SI_ORDER_FIRST, initclocks, NULL);
   85 
   86 /* Spin-lock protecting profiling statistics. */
   87 static struct mtx time_lock;
   88 
   89 static int
   90 sysctl_kern_cp_time(SYSCTL_HANDLER_ARGS)
   91 {
   92         int error;
   93         long cp_time[CPUSTATES];
   94 #ifdef SCTL_MASK32
   95         int i;
   96         unsigned int cp_time32[CPUSTATES];
   97 #endif
   98 
   99         read_cpu_time(cp_time);
  100 #ifdef SCTL_MASK32
  101         if (req->flags & SCTL_MASK32) {
  102                 if (!req->oldptr)
  103                         return SYSCTL_OUT(req, 0, sizeof(cp_time32));
  104                 for (i = 0; i < CPUSTATES; i++)
  105                         cp_time32[i] = (unsigned int)cp_time[i];
  106                 error = SYSCTL_OUT(req, cp_time32, sizeof(cp_time32));
  107         } else
  108 #endif
  109         {
  110                 if (!req->oldptr)
  111                         return SYSCTL_OUT(req, 0, sizeof(cp_time));
  112                 error = SYSCTL_OUT(req, cp_time, sizeof(cp_time));
  113         }
  114         return error;
  115 }
  116 
  117 SYSCTL_PROC(_kern, OID_AUTO, cp_time, CTLTYPE_LONG|CTLFLAG_RD|CTLFLAG_MPSAFE,
  118     0,0, sysctl_kern_cp_time, "LU", "CPU time statistics");
  119 
  120 static long empty[CPUSTATES];
  121 
  122 static int
  123 sysctl_kern_cp_times(SYSCTL_HANDLER_ARGS)
  124 {
  125         struct pcpu *pcpu;
  126         int error;
  127         int c;
  128         long *cp_time;
  129 #ifdef SCTL_MASK32
  130         unsigned int cp_time32[CPUSTATES];
  131         int i;
  132 #endif
  133 
  134         if (!req->oldptr) {
  135 #ifdef SCTL_MASK32
  136                 if (req->flags & SCTL_MASK32)
  137                         return SYSCTL_OUT(req, 0, sizeof(cp_time32) * (mp_maxid + 1));
  138                 else
  139 #endif
  140                         return SYSCTL_OUT(req, 0, sizeof(long) * CPUSTATES * (mp_maxid + 1));
  141         }
  142         for (error = 0, c = 0; error == 0 && c <= mp_maxid; c++) {
  143                 if (!CPU_ABSENT(c)) {
  144                         pcpu = pcpu_find(c);
  145                         cp_time = pcpu->pc_cp_time;
  146                 } else {
  147                         cp_time = empty;
  148                 }
  149 #ifdef SCTL_MASK32
  150                 if (req->flags & SCTL_MASK32) {
  151                         for (i = 0; i < CPUSTATES; i++)
  152                                 cp_time32[i] = (unsigned int)cp_time[i];
  153                         error = SYSCTL_OUT(req, cp_time32, sizeof(cp_time32));
  154                 } else
  155 #endif
  156                         error = SYSCTL_OUT(req, cp_time, sizeof(long) * CPUSTATES);
  157         }
  158         return error;
  159 }
  160 
  161 SYSCTL_PROC(_kern, OID_AUTO, cp_times, CTLTYPE_LONG|CTLFLAG_RD|CTLFLAG_MPSAFE,
  162     0,0, sysctl_kern_cp_times, "LU", "per-CPU time statistics");
  163 
  164 #ifdef DEADLKRES
  165 static const char *blessed[] = {
  166         "getblk",
  167         "so_snd_sx",
  168         "so_rcv_sx",
  169         NULL
  170 };
  171 static int slptime_threshold = 1800;
  172 static int blktime_threshold = 900;
  173 static int sleepfreq = 3;
  174 
  175 static void
  176 deadlkres(void)
  177 {
  178         struct proc *p;
  179         struct thread *td;
  180         void *wchan;
  181         int blkticks, i, slpticks, slptype, tryl, tticks;
  182 
  183         tryl = 0;
  184         for (;;) {
  185                 blkticks = blktime_threshold * hz;
  186                 slpticks = slptime_threshold * hz;
  187 
  188                 /*
  189                  * Avoid to sleep on the sx_lock in order to avoid a possible
  190                  * priority inversion problem leading to starvation.
  191                  * If the lock can't be held after 100 tries, panic.
  192                  */
  193                 if (!sx_try_slock(&allproc_lock)) {
  194                         if (tryl > 100)
  195                 panic("%s: possible deadlock detected on allproc_lock\n",
  196                                     __func__);
  197                         tryl++;
  198                         pause("allproc_lock deadlkres", sleepfreq * hz);
  199                         continue;
  200                 }
  201                 tryl = 0;
  202                 FOREACH_PROC_IN_SYSTEM(p) {
  203                         PROC_LOCK(p);
  204                         FOREACH_THREAD_IN_PROC(p, td) {
  205                                 thread_lock(td);
  206                                 if (TD_ON_LOCK(td)) {
  207 
  208                                         /*
  209                                          * The thread should be blocked on a
  210                                          * turnstile, simply check if the
  211                                          * turnstile channel is in good state.
  212                                          */
  213                                         MPASS(td->td_blocked != NULL);
  214 
  215                                         /* Handle ticks wrap-up. */
  216                                         if (ticks < td->td_blktick)
  217                                                 continue;
  218                                         tticks = ticks - td->td_blktick;
  219                                         thread_unlock(td);
  220                                         if (tticks > blkticks) {
  221 
  222                                                 /*
  223                                                  * Accordingly with provided
  224                                                  * thresholds, this thread is
  225                                                  * stuck for too long on a
  226                                                  * turnstile.
  227                                                  */
  228                                                 PROC_UNLOCK(p);
  229                                                 sx_sunlock(&allproc_lock);
  230         panic("%s: possible deadlock detected for %p, blocked for %d ticks\n",
  231                                                     __func__, td, tticks);
  232                                         }
  233                                 } else if (TD_IS_SLEEPING(td)) {
  234 
  235                                         /* Handle ticks wrap-up. */
  236                                         if (ticks < td->td_blktick)
  237                                                 continue;
  238 
  239                                         /*
  240                                          * Check if the thread is sleeping on a
  241                                          * lock, otherwise skip the check.
  242                                          * Drop the thread lock in order to
  243                                          * avoid a LOR with the sleepqueue
  244                                          * spinlock.
  245                                          */
  246                                         wchan = td->td_wchan;
  247                                         tticks = ticks - td->td_slptick;
  248                                         thread_unlock(td);
  249                                         slptype = sleepq_type(wchan);
  250                                         if ((slptype == SLEEPQ_SX ||
  251                                             slptype == SLEEPQ_LK) &&
  252                                             tticks > slpticks) {
  253 
  254                                                 /*
  255                                                  * Accordingly with provided
  256                                                  * thresholds, this thread is
  257                                                  * stuck for too long on a
  258                                                  * sleepqueue.
  259                                                  * However, being on a
  260                                                  * sleepqueue, we might still
  261                                                  * check for the blessed
  262                                                  * list.
  263                                                  */
  264                                                 tryl = 0;
  265                                                 for (i = 0; blessed[i] != NULL;
  266                                                     i++) {
  267                                                         if (!strcmp(blessed[i],
  268                                                             td->td_wmesg)) {
  269                                                                 tryl = 1;
  270                                                                 break;
  271                                                         }
  272                                                 }
  273                                                 if (tryl != 0) {
  274                                                         tryl = 0;
  275                                                         continue;
  276                                                 }
  277                                                 PROC_UNLOCK(p);
  278                                                 sx_sunlock(&allproc_lock);
  279         panic("%s: possible deadlock detected for %p, blocked for %d ticks\n",
  280                                                     __func__, td, tticks);
  281                                         }
  282                                 } else
  283                                         thread_unlock(td);
  284                         }
  285                         PROC_UNLOCK(p);
  286                 }
  287                 sx_sunlock(&allproc_lock);
  288 
  289                 /* Sleep for sleepfreq seconds. */
  290                 pause("deadlkres", sleepfreq * hz);
  291         }
  292 }
  293 
  294 static struct kthread_desc deadlkres_kd = {
  295         "deadlkres",
  296         deadlkres,
  297         (struct thread **)NULL
  298 };
  299 
  300 SYSINIT(deadlkres, SI_SUB_CLOCKS, SI_ORDER_ANY, kthread_start, &deadlkres_kd);
  301 
  302 SYSCTL_NODE(_debug, OID_AUTO, deadlkres, CTLFLAG_RW, 0, "Deadlock resolver");
  303 SYSCTL_INT(_debug_deadlkres, OID_AUTO, slptime_threshold, CTLFLAG_RW,
  304     &slptime_threshold, 0,
  305     "Number of seconds within is valid to sleep on a sleepqueue");
  306 SYSCTL_INT(_debug_deadlkres, OID_AUTO, blktime_threshold, CTLFLAG_RW,
  307     &blktime_threshold, 0,
  308     "Number of seconds within is valid to block on a turnstile");
  309 SYSCTL_INT(_debug_deadlkres, OID_AUTO, sleepfreq, CTLFLAG_RW, &sleepfreq, 0,
  310     "Number of seconds between any deadlock resolver thread run");
  311 #endif  /* DEADLKRES */
  312 
  313 void
  314 read_cpu_time(long *cp_time)
  315 {
  316         struct pcpu *pc;
  317         int i, j;
  318 
  319         /* Sum up global cp_time[]. */
  320         bzero(cp_time, sizeof(long) * CPUSTATES);
  321         for (i = 0; i <= mp_maxid; i++) {
  322                 if (CPU_ABSENT(i))
  323                         continue;
  324                 pc = pcpu_find(i);
  325                 for (j = 0; j < CPUSTATES; j++)
  326                         cp_time[j] += pc->pc_cp_time[j];
  327         }
  328 }
  329 
  330 #ifdef SW_WATCHDOG
  331 #include <sys/watchdog.h>
  332 
  333 static int watchdog_ticks;
  334 static int watchdog_enabled;
  335 static void watchdog_fire(void);
  336 static void watchdog_config(void *, u_int, int *);
  337 #endif /* SW_WATCHDOG */
  338 
  339 /*
  340  * Clock handling routines.
  341  *
  342  * This code is written to operate with two timers that run independently of
  343  * each other.
  344  *
  345  * The main timer, running hz times per second, is used to trigger interval
  346  * timers, timeouts and rescheduling as needed.
  347  *
  348  * The second timer handles kernel and user profiling,
  349  * and does resource use estimation.  If the second timer is programmable,
  350  * it is randomized to avoid aliasing between the two clocks.  For example,
  351  * the randomization prevents an adversary from always giving up the cpu
  352  * just before its quantum expires.  Otherwise, it would never accumulate
  353  * cpu ticks.  The mean frequency of the second timer is stathz.
  354  *
  355  * If no second timer exists, stathz will be zero; in this case we drive
  356  * profiling and statistics off the main clock.  This WILL NOT be accurate;
  357  * do not do it unless absolutely necessary.
  358  *
  359  * The statistics clock may (or may not) be run at a higher rate while
  360  * profiling.  This profile clock runs at profhz.  We require that profhz
  361  * be an integral multiple of stathz.
  362  *
  363  * If the statistics clock is running fast, it must be divided by the ratio
  364  * profhz/stathz for statistics.  (For profiling, every tick counts.)
  365  *
  366  * Time-of-day is maintained using a "timecounter", which may or may
  367  * not be related to the hardware generating the above mentioned
  368  * interrupts.
  369  */
  370 
  371 int     stathz;
  372 int     profhz;
  373 int     profprocs;
  374 int     ticks;
  375 int     psratio;
  376 
  377 /*
  378  * Initialize clock frequencies and start both clocks running.
  379  */
  380 /* ARGSUSED*/
  381 static void
  382 initclocks(dummy)
  383         void *dummy;
  384 {
  385         register int i;
  386 
  387         /*
  388          * Set divisors to 1 (normal case) and let the machine-specific
  389          * code do its bit.
  390          */
  391         mtx_init(&time_lock, "time lock", NULL, MTX_SPIN);
  392         cpu_initclocks();
  393 
  394         /*
  395          * Compute profhz/stathz, and fix profhz if needed.
  396          */
  397         i = stathz ? stathz : hz;
  398         if (profhz == 0)
  399                 profhz = i;
  400         psratio = profhz / i;
  401 #ifdef SW_WATCHDOG
  402         EVENTHANDLER_REGISTER(watchdog_list, watchdog_config, NULL, 0);
  403 #endif
  404 }
  405 
  406 /*
  407  * Each time the real-time timer fires, this function is called on all CPUs.
  408  * Note that hardclock() calls hardclock_cpu() for the boot CPU, so only
  409  * the other CPUs in the system need to call this function.
  410  */
  411 void
  412 hardclock_cpu(int usermode)
  413 {
  414         struct pstats *pstats;
  415         struct thread *td = curthread;
  416         struct proc *p = td->td_proc;
  417         int flags;
  418 
  419         /*
  420          * Run current process's virtual and profile time, as needed.
  421          */
  422         pstats = p->p_stats;
  423         flags = 0;
  424         if (usermode &&
  425             timevalisset(&pstats->p_timer[ITIMER_VIRTUAL].it_value)) {
  426                 PROC_SLOCK(p);
  427                 if (itimerdecr(&pstats->p_timer[ITIMER_VIRTUAL], tick) == 0)
  428                         flags |= TDF_ALRMPEND | TDF_ASTPENDING;
  429                 PROC_SUNLOCK(p);
  430         }
  431         if (timevalisset(&pstats->p_timer[ITIMER_PROF].it_value)) {
  432                 PROC_SLOCK(p);
  433                 if (itimerdecr(&pstats->p_timer[ITIMER_PROF], tick) == 0)
  434                         flags |= TDF_PROFPEND | TDF_ASTPENDING;
  435                 PROC_SUNLOCK(p);
  436         }
  437         thread_lock(td);
  438         sched_tick();
  439         td->td_flags |= flags;
  440         thread_unlock(td);
  441 
  442 #ifdef  HWPMC_HOOKS
  443         if (PMC_CPU_HAS_SAMPLES(PCPU_GET(cpuid)))
  444                 PMC_CALL_HOOK_UNLOCKED(curthread, PMC_FN_DO_SAMPLES, NULL);
  445 #endif
  446         callout_tick();
  447 }
  448 
  449 /*
  450  * The real-time timer, interrupting hz times per second.
  451  */
  452 void
  453 hardclock(int usermode, uintfptr_t pc)
  454 {
  455 
  456         atomic_add_int((volatile int *)&ticks, 1);
  457         hardclock_cpu(usermode);
  458         tc_ticktock();
  459         /*
  460          * If no separate statistics clock is available, run it from here.
  461          *
  462          * XXX: this only works for UP
  463          */
  464         if (stathz == 0) {
  465                 profclock(usermode, pc);
  466                 statclock(usermode);
  467         }
  468 #ifdef DEVICE_POLLING
  469         hardclock_device_poll();        /* this is very short and quick */
  470 #endif /* DEVICE_POLLING */
  471 #ifdef SW_WATCHDOG
  472         if (watchdog_enabled > 0 && --watchdog_ticks <= 0)
  473                 watchdog_fire();
  474 #endif /* SW_WATCHDOG */
  475 }
  476 
  477 /*
  478  * Compute number of ticks in the specified amount of time.
  479  */
  480 int
  481 tvtohz(tv)
  482         struct timeval *tv;
  483 {
  484         register unsigned long ticks;
  485         register long sec, usec;
  486 
  487         /*
  488          * If the number of usecs in the whole seconds part of the time
  489          * difference fits in a long, then the total number of usecs will
  490          * fit in an unsigned long.  Compute the total and convert it to
  491          * ticks, rounding up and adding 1 to allow for the current tick
  492          * to expire.  Rounding also depends on unsigned long arithmetic
  493          * to avoid overflow.
  494          *
  495          * Otherwise, if the number of ticks in the whole seconds part of
  496          * the time difference fits in a long, then convert the parts to
  497          * ticks separately and add, using similar rounding methods and
  498          * overflow avoidance.  This method would work in the previous
  499          * case but it is slightly slower and assumes that hz is integral.
  500          *
  501          * Otherwise, round the time difference down to the maximum
  502          * representable value.
  503          *
  504          * If ints have 32 bits, then the maximum value for any timeout in
  505          * 10ms ticks is 248 days.
  506          */
  507         sec = tv->tv_sec;
  508         usec = tv->tv_usec;
  509         if (usec < 0) {
  510                 sec--;
  511                 usec += 1000000;
  512         }
  513         if (sec < 0) {
  514 #ifdef DIAGNOSTIC
  515                 if (usec > 0) {
  516                         sec++;
  517                         usec -= 1000000;
  518                 }
  519                 printf("tvotohz: negative time difference %ld sec %ld usec\n",
  520                        sec, usec);
  521 #endif
  522                 ticks = 1;
  523         } else if (sec <= LONG_MAX / 1000000)
  524                 ticks = (sec * 1000000 + (unsigned long)usec + (tick - 1))
  525                         / tick + 1;
  526         else if (sec <= LONG_MAX / hz)
  527                 ticks = sec * hz
  528                         + ((unsigned long)usec + (tick - 1)) / tick + 1;
  529         else
  530                 ticks = LONG_MAX;
  531         if (ticks > INT_MAX)
  532                 ticks = INT_MAX;
  533         return ((int)ticks);
  534 }
  535 
  536 /*
  537  * Start profiling on a process.
  538  *
  539  * Kernel profiling passes proc0 which never exits and hence
  540  * keeps the profile clock running constantly.
  541  */
  542 void
  543 startprofclock(p)
  544         register struct proc *p;
  545 {
  546 
  547         PROC_LOCK_ASSERT(p, MA_OWNED);
  548         if (p->p_flag & P_STOPPROF)
  549                 return;
  550         if ((p->p_flag & P_PROFIL) == 0) {
  551                 p->p_flag |= P_PROFIL;
  552                 mtx_lock_spin(&time_lock);
  553                 if (++profprocs == 1)
  554                         cpu_startprofclock();
  555                 mtx_unlock_spin(&time_lock);
  556         }
  557 }
  558 
  559 /*
  560  * Stop profiling on a process.
  561  */
  562 void
  563 stopprofclock(p)
  564         register struct proc *p;
  565 {
  566 
  567         PROC_LOCK_ASSERT(p, MA_OWNED);
  568         if (p->p_flag & P_PROFIL) {
  569                 if (p->p_profthreads != 0) {
  570                         p->p_flag |= P_STOPPROF;
  571                         while (p->p_profthreads != 0)
  572                                 msleep(&p->p_profthreads, &p->p_mtx, PPAUSE,
  573                                     "stopprof", 0);
  574                         p->p_flag &= ~P_STOPPROF;
  575                 }
  576                 if ((p->p_flag & P_PROFIL) == 0)
  577                         return;
  578                 p->p_flag &= ~P_PROFIL;
  579                 mtx_lock_spin(&time_lock);
  580                 if (--profprocs == 0)
  581                         cpu_stopprofclock();
  582                 mtx_unlock_spin(&time_lock);
  583         }
  584 }
  585 
  586 /*
  587  * Statistics clock.  Updates rusage information and calls the scheduler
  588  * to adjust priorities of the active thread.
  589  *
  590  * This should be called by all active processors.
  591  */
  592 void
  593 statclock(int usermode)
  594 {
  595         struct rusage *ru;
  596         struct vmspace *vm;
  597         struct thread *td;
  598         struct proc *p;
  599         long rss;
  600         long *cp_time;
  601 
  602         td = curthread;
  603         p = td->td_proc;
  604 
  605         cp_time = (long *)PCPU_PTR(cp_time);
  606         if (usermode) {
  607                 /*
  608                  * Charge the time as appropriate.
  609                  */
  610                 td->td_uticks++;
  611                 if (p->p_nice > NZERO)
  612                         cp_time[CP_NICE]++;
  613                 else
  614                         cp_time[CP_USER]++;
  615         } else {
  616                 /*
  617                  * Came from kernel mode, so we were:
  618                  * - handling an interrupt,
  619                  * - doing syscall or trap work on behalf of the current
  620                  *   user process, or
  621                  * - spinning in the idle loop.
  622                  * Whichever it is, charge the time as appropriate.
  623                  * Note that we charge interrupts to the current process,
  624                  * regardless of whether they are ``for'' that process,
  625                  * so that we know how much of its real time was spent
  626                  * in ``non-process'' (i.e., interrupt) work.
  627                  */
  628                 if ((td->td_pflags & TDP_ITHREAD) ||
  629                     td->td_intr_nesting_level >= 2) {
  630                         td->td_iticks++;
  631                         cp_time[CP_INTR]++;
  632                 } else {
  633                         td->td_pticks++;
  634                         td->td_sticks++;
  635                         if (!TD_IS_IDLETHREAD(td))
  636                                 cp_time[CP_SYS]++;
  637                         else
  638                                 cp_time[CP_IDLE]++;
  639                 }
  640         }
  641 
  642         /* Update resource usage integrals and maximums. */
  643         MPASS(p->p_vmspace != NULL);
  644         vm = p->p_vmspace;
  645         ru = &td->td_ru;
  646         ru->ru_ixrss += pgtok(vm->vm_tsize);
  647         ru->ru_idrss += pgtok(vm->vm_dsize);
  648         ru->ru_isrss += pgtok(vm->vm_ssize);
  649         rss = pgtok(vmspace_resident_count(vm));
  650         if (ru->ru_maxrss < rss)
  651                 ru->ru_maxrss = rss;
  652         KTR_POINT2(KTR_SCHED, "thread", sched_tdname(td), "statclock",
  653             "prio:%d", td->td_priority, "stathz:%d", (stathz)?stathz:hz);
  654         thread_lock_flags(td, MTX_QUIET);
  655         sched_clock(td);
  656         thread_unlock(td);
  657 }
  658 
  659 void
  660 profclock(int usermode, uintfptr_t pc)
  661 {
  662         struct thread *td;
  663 #ifdef GPROF
  664         struct gmonparam *g;
  665         uintfptr_t i;
  666 #endif
  667 
  668         td = curthread;
  669         if (usermode) {
  670                 /*
  671                  * Came from user mode; CPU was in user state.
  672                  * If this process is being profiled, record the tick.
  673                  * if there is no related user location yet, don't
  674                  * bother trying to count it.
  675                  */
  676                 if (td->td_proc->p_flag & P_PROFIL)
  677                         addupc_intr(td, pc, 1);
  678         }
  679 #ifdef GPROF
  680         else {
  681                 /*
  682                  * Kernel statistics are just like addupc_intr, only easier.
  683                  */
  684                 g = &_gmonparam;
  685                 if (g->state == GMON_PROF_ON && pc >= g->lowpc) {
  686                         i = PC_TO_I(g, pc);
  687                         if (i < g->textsize) {
  688                                 KCOUNT(g, i)++;
  689                         }
  690                 }
  691         }
  692 #endif
  693 }
  694 
  695 /*
  696  * Return information about system clocks.
  697  */
  698 static int
  699 sysctl_kern_clockrate(SYSCTL_HANDLER_ARGS)
  700 {
  701         struct clockinfo clkinfo;
  702         /*
  703          * Construct clockinfo structure.
  704          */
  705         bzero(&clkinfo, sizeof(clkinfo));
  706         clkinfo.hz = hz;
  707         clkinfo.tick = tick;
  708         clkinfo.profhz = profhz;
  709         clkinfo.stathz = stathz ? stathz : hz;
  710         return (sysctl_handle_opaque(oidp, &clkinfo, sizeof clkinfo, req));
  711 }
  712 
  713 SYSCTL_PROC(_kern, KERN_CLOCKRATE, clockrate,
  714         CTLTYPE_STRUCT|CTLFLAG_RD|CTLFLAG_MPSAFE,
  715         0, 0, sysctl_kern_clockrate, "S,clockinfo",
  716         "Rate and period of various kernel clocks");
  717 
  718 #ifdef SW_WATCHDOG
  719 
  720 static void
  721 watchdog_config(void *unused __unused, u_int cmd, int *error)
  722 {
  723         u_int u;
  724 
  725         u = cmd & WD_INTERVAL;
  726         if (u >= WD_TO_1SEC) {
  727                 watchdog_ticks = (1 << (u - WD_TO_1SEC)) * hz;
  728                 watchdog_enabled = 1;
  729                 *error = 0;
  730         } else {
  731                 watchdog_enabled = 0;
  732         }
  733 }
  734 
  735 /*
  736  * Handle a watchdog timeout by dumping interrupt information and
  737  * then either dropping to DDB or panicking.
  738  */
  739 static void
  740 watchdog_fire(void)
  741 {
  742         int nintr;
  743         u_int64_t inttotal;
  744         u_long *curintr;
  745         char *curname;
  746 
  747         curintr = intrcnt;
  748         curname = intrnames;
  749         inttotal = 0;
  750         nintr = eintrcnt - intrcnt;
  751 
  752         printf("interrupt                   total\n");
  753         while (--nintr >= 0) {
  754                 if (*curintr)
  755                         printf("%-12s %20lu\n", curname, *curintr);
  756                 curname += strlen(curname) + 1;
  757                 inttotal += *curintr++;
  758         }
  759         printf("Total        %20ju\n", (uintmax_t)inttotal);
  760 
  761 #if defined(KDB) && !defined(KDB_UNATTENDED)
  762         kdb_backtrace();
  763         kdb_enter(KDB_WHY_WATCHDOG, "watchdog timeout");
  764 #else
  765         panic("watchdog timeout");
  766 #endif
  767 }
  768 
  769 #endif /* SW_WATCHDOG */

Cache object: eae4fc4fb97a90a11be7e98d44070609


[ 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.