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 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)kern_clock.c 8.5 (Berkeley) 1/21/94 39 * $FreeBSD$ 40 */ 41 42 #include "opt_ntp.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/callout.h> 47 #include <sys/kernel.h> 48 #include <sys/lock.h> 49 #include <sys/ktr.h> 50 #include <sys/mutex.h> 51 #include <sys/proc.h> 52 #include <sys/resource.h> 53 #include <sys/resourcevar.h> 54 #include <sys/sched.h> 55 #include <sys/signalvar.h> 56 #include <sys/smp.h> 57 #include <vm/vm.h> 58 #include <vm/pmap.h> 59 #include <vm/vm_map.h> 60 #include <sys/sysctl.h> 61 #include <sys/bus.h> 62 #include <sys/interrupt.h> 63 #include <sys/limits.h> 64 #include <sys/timetc.h> 65 66 #include <machine/cpu.h> 67 68 #ifdef GPROF 69 #include <sys/gmon.h> 70 #endif 71 72 #ifdef DEVICE_POLLING 73 extern void hardclock_device_poll(void); 74 #endif /* DEVICE_POLLING */ 75 76 static void initclocks(void *dummy); 77 SYSINIT(clocks, SI_SUB_CLOCKS, SI_ORDER_FIRST, initclocks, NULL) 78 79 /* Some of these don't belong here, but it's easiest to concentrate them. */ 80 long cp_time[CPUSTATES]; 81 82 SYSCTL_OPAQUE(_kern, OID_AUTO, cp_time, CTLFLAG_RD, &cp_time, sizeof(cp_time), 83 "LU", "CPU time statistics"); 84 85 /* 86 * Clock handling routines. 87 * 88 * This code is written to operate with two timers that run independently of 89 * each other. 90 * 91 * The main timer, running hz times per second, is used to trigger interval 92 * timers, timeouts and rescheduling as needed. 93 * 94 * The second timer handles kernel and user profiling, 95 * and does resource use estimation. If the second timer is programmable, 96 * it is randomized to avoid aliasing between the two clocks. For example, 97 * the randomization prevents an adversary from always giving up the cpu 98 * just before its quantum expires. Otherwise, it would never accumulate 99 * cpu ticks. The mean frequency of the second timer is stathz. 100 * 101 * If no second timer exists, stathz will be zero; in this case we drive 102 * profiling and statistics off the main clock. This WILL NOT be accurate; 103 * do not do it unless absolutely necessary. 104 * 105 * The statistics clock may (or may not) be run at a higher rate while 106 * profiling. This profile clock runs at profhz. We require that profhz 107 * be an integral multiple of stathz. 108 * 109 * If the statistics clock is running fast, it must be divided by the ratio 110 * profhz/stathz for statistics. (For profiling, every tick counts.) 111 * 112 * Time-of-day is maintained using a "timecounter", which may or may 113 * not be related to the hardware generating the above mentioned 114 * interrupts. 115 */ 116 117 int stathz; 118 int profhz; 119 int profprocs; 120 int ticks; 121 int psratio; 122 123 /* 124 * Initialize clock frequencies and start both clocks running. 125 */ 126 /* ARGSUSED*/ 127 static void 128 initclocks(dummy) 129 void *dummy; 130 { 131 register int i; 132 133 /* 134 * Set divisors to 1 (normal case) and let the machine-specific 135 * code do its bit. 136 */ 137 cpu_initclocks(); 138 139 /* 140 * Compute profhz/stathz, and fix profhz if needed. 141 */ 142 i = stathz ? stathz : hz; 143 if (profhz == 0) 144 profhz = i; 145 psratio = profhz / i; 146 } 147 148 /* 149 * Each time the real-time timer fires, this function is called on all CPUs. 150 * Note that hardclock() calls hardclock_process() for the boot CPU, so only 151 * the other CPUs in the system need to call this function. 152 */ 153 void 154 hardclock_process(frame) 155 register struct clockframe *frame; 156 { 157 struct pstats *pstats; 158 struct thread *td = curthread; 159 struct proc *p = td->td_proc; 160 161 /* 162 * Run current process's virtual and profile time, as needed. 163 */ 164 mtx_lock_spin_flags(&sched_lock, MTX_QUIET); 165 if (p->p_flag & P_THREADED) { 166 /* XXXKSE What to do? */ 167 } else { 168 pstats = p->p_stats; 169 if (CLKF_USERMODE(frame) && 170 timevalisset(&pstats->p_timer[ITIMER_VIRTUAL].it_value) && 171 itimerdecr(&pstats->p_timer[ITIMER_VIRTUAL], tick) == 0) { 172 p->p_sflag |= PS_ALRMPEND; 173 td->td_flags |= TDF_ASTPENDING; 174 } 175 if (timevalisset(&pstats->p_timer[ITIMER_PROF].it_value) && 176 itimerdecr(&pstats->p_timer[ITIMER_PROF], tick) == 0) { 177 p->p_sflag |= PS_PROFPEND; 178 td->td_flags |= TDF_ASTPENDING; 179 } 180 } 181 mtx_unlock_spin_flags(&sched_lock, MTX_QUIET); 182 } 183 184 /* 185 * The real-time timer, interrupting hz times per second. 186 */ 187 void 188 hardclock(frame) 189 register struct clockframe *frame; 190 { 191 int need_softclock = 0; 192 193 CTR0(KTR_CLK, "hardclock fired"); 194 hardclock_process(frame); 195 196 tc_ticktock(); 197 /* 198 * If no separate statistics clock is available, run it from here. 199 * 200 * XXX: this only works for UP 201 */ 202 if (stathz == 0) { 203 profclock(frame); 204 statclock(frame); 205 } 206 207 #ifdef DEVICE_POLLING 208 hardclock_device_poll(); /* this is very short and quick */ 209 #endif /* DEVICE_POLLING */ 210 211 /* 212 * Process callouts at a very low cpu priority, so we don't keep the 213 * relatively high clock interrupt priority any longer than necessary. 214 */ 215 mtx_lock_spin_flags(&callout_lock, MTX_QUIET); 216 ticks++; 217 if (TAILQ_FIRST(&callwheel[ticks & callwheelmask]) != NULL) { 218 need_softclock = 1; 219 } else if (softticks + 1 == ticks) 220 ++softticks; 221 mtx_unlock_spin_flags(&callout_lock, MTX_QUIET); 222 223 /* 224 * swi_sched acquires sched_lock, so we don't want to call it with 225 * callout_lock held; incorrect locking order. 226 */ 227 if (need_softclock) 228 swi_sched(softclock_ih, 0); 229 } 230 231 /* 232 * Compute number of ticks in the specified amount of time. 233 */ 234 int 235 tvtohz(tv) 236 struct timeval *tv; 237 { 238 register unsigned long ticks; 239 register long sec, usec; 240 241 /* 242 * If the number of usecs in the whole seconds part of the time 243 * difference fits in a long, then the total number of usecs will 244 * fit in an unsigned long. Compute the total and convert it to 245 * ticks, rounding up and adding 1 to allow for the current tick 246 * to expire. Rounding also depends on unsigned long arithmetic 247 * to avoid overflow. 248 * 249 * Otherwise, if the number of ticks in the whole seconds part of 250 * the time difference fits in a long, then convert the parts to 251 * ticks separately and add, using similar rounding methods and 252 * overflow avoidance. This method would work in the previous 253 * case but it is slightly slower and assumes that hz is integral. 254 * 255 * Otherwise, round the time difference down to the maximum 256 * representable value. 257 * 258 * If ints have 32 bits, then the maximum value for any timeout in 259 * 10ms ticks is 248 days. 260 */ 261 sec = tv->tv_sec; 262 usec = tv->tv_usec; 263 if (usec < 0) { 264 sec--; 265 usec += 1000000; 266 } 267 if (sec < 0) { 268 #ifdef DIAGNOSTIC 269 if (usec > 0) { 270 sec++; 271 usec -= 1000000; 272 } 273 printf("tvotohz: negative time difference %ld sec %ld usec\n", 274 sec, usec); 275 #endif 276 ticks = 1; 277 } else if (sec <= LONG_MAX / 1000000) 278 ticks = (sec * 1000000 + (unsigned long)usec + (tick - 1)) 279 / tick + 1; 280 else if (sec <= LONG_MAX / hz) 281 ticks = sec * hz 282 + ((unsigned long)usec + (tick - 1)) / tick + 1; 283 else 284 ticks = LONG_MAX; 285 if (ticks > INT_MAX) 286 ticks = INT_MAX; 287 return ((int)ticks); 288 } 289 290 /* 291 * Start profiling on a process. 292 * 293 * Kernel profiling passes proc0 which never exits and hence 294 * keeps the profile clock running constantly. 295 */ 296 void 297 startprofclock(p) 298 register struct proc *p; 299 { 300 301 /* 302 * XXX; Right now sched_lock protects statclock(), but perhaps 303 * it should be protected later on by a time_lock, which would 304 * cover psdiv, etc. as well. 305 */ 306 PROC_LOCK_ASSERT(p, MA_OWNED); 307 if (p->p_flag & P_STOPPROF) 308 return; 309 if ((p->p_flag & P_PROFIL) == 0) { 310 mtx_lock_spin(&sched_lock); 311 p->p_flag |= P_PROFIL; 312 if (++profprocs == 1) 313 cpu_startprofclock(); 314 mtx_unlock_spin(&sched_lock); 315 } 316 } 317 318 /* 319 * Stop profiling on a process. 320 */ 321 void 322 stopprofclock(p) 323 register struct proc *p; 324 { 325 326 PROC_LOCK_ASSERT(p, MA_OWNED); 327 if (p->p_flag & P_PROFIL) { 328 if (p->p_profthreads != 0) { 329 p->p_flag |= P_STOPPROF; 330 while (p->p_profthreads != 0) 331 msleep(&p->p_profthreads, &p->p_mtx, PPAUSE, 332 "stopprof", NULL); 333 p->p_flag &= ~P_STOPPROF; 334 } 335 mtx_lock_spin(&sched_lock); 336 p->p_flag &= ~P_PROFIL; 337 if (--profprocs == 0) 338 cpu_stopprofclock(); 339 mtx_unlock_spin(&sched_lock); 340 } 341 } 342 343 /* 344 * Statistics clock. Grab profile sample, and if divider reaches 0, 345 * do process and kernel statistics. Most of the statistics are only 346 * used by user-level statistics programs. The main exceptions are 347 * ke->ke_uticks, p->p_sticks, p->p_iticks, and p->p_estcpu. 348 * This should be called by all active processors. 349 */ 350 void 351 statclock(frame) 352 register struct clockframe *frame; 353 { 354 struct pstats *pstats; 355 struct rusage *ru; 356 struct vmspace *vm; 357 struct thread *td; 358 struct kse *ke; 359 struct proc *p; 360 long rss; 361 362 td = curthread; 363 p = td->td_proc; 364 365 mtx_lock_spin_flags(&sched_lock, MTX_QUIET); 366 ke = td->td_kse; 367 if (CLKF_USERMODE(frame)) { 368 /* 369 * Charge the time as appropriate. 370 */ 371 if (p->p_flag & P_THREADED) 372 thread_statclock(1); 373 p->p_uticks++; 374 if (ke->ke_ksegrp->kg_nice > NZERO) 375 cp_time[CP_NICE]++; 376 else 377 cp_time[CP_USER]++; 378 } else { 379 /* 380 * Came from kernel mode, so we were: 381 * - handling an interrupt, 382 * - doing syscall or trap work on behalf of the current 383 * user process, or 384 * - spinning in the idle loop. 385 * Whichever it is, charge the time as appropriate. 386 * Note that we charge interrupts to the current process, 387 * regardless of whether they are ``for'' that process, 388 * so that we know how much of its real time was spent 389 * in ``non-process'' (i.e., interrupt) work. 390 */ 391 if ((td->td_ithd != NULL) || td->td_intr_nesting_level >= 2) { 392 p->p_iticks++; 393 cp_time[CP_INTR]++; 394 } else { 395 if (p->p_flag & P_THREADED) 396 thread_statclock(0); 397 td->td_sticks++; 398 p->p_sticks++; 399 if (p != PCPU_GET(idlethread)->td_proc) 400 cp_time[CP_SYS]++; 401 else 402 cp_time[CP_IDLE]++; 403 } 404 } 405 406 sched_clock(ke); 407 408 /* Update resource usage integrals and maximums. */ 409 if ((pstats = p->p_stats) != NULL && 410 (ru = &pstats->p_ru) != NULL && 411 (vm = p->p_vmspace) != NULL) { 412 ru->ru_ixrss += pgtok(vm->vm_tsize); 413 ru->ru_idrss += pgtok(vm->vm_dsize); 414 ru->ru_isrss += pgtok(vm->vm_ssize); 415 rss = pgtok(vmspace_resident_count(vm)); 416 if (ru->ru_maxrss < rss) 417 ru->ru_maxrss = rss; 418 } 419 mtx_unlock_spin_flags(&sched_lock, MTX_QUIET); 420 } 421 422 void 423 profclock(frame) 424 register struct clockframe *frame; 425 { 426 struct thread *td; 427 #ifdef GPROF 428 struct gmonparam *g; 429 int i; 430 #endif 431 432 td = curthread; 433 if (CLKF_USERMODE(frame)) { 434 /* 435 * Came from user mode; CPU was in user state. 436 * If this process is being profiled, record the tick. 437 * if there is no related user location yet, don't 438 * bother trying to count it. 439 */ 440 td = curthread; 441 if (td->td_proc->p_flag & P_PROFIL) 442 addupc_intr(td, CLKF_PC(frame), 1); 443 } 444 #ifdef GPROF 445 else { 446 /* 447 * Kernel statistics are just like addupc_intr, only easier. 448 */ 449 g = &_gmonparam; 450 if (g->state == GMON_PROF_ON) { 451 i = CLKF_PC(frame) - g->lowpc; 452 if (i < g->textsize) { 453 i /= HISTFRACTION * sizeof(*g->kcount); 454 g->kcount[i]++; 455 } 456 } 457 } 458 #endif 459 } 460 461 /* 462 * Return information about system clocks. 463 */ 464 static int 465 sysctl_kern_clockrate(SYSCTL_HANDLER_ARGS) 466 { 467 struct clockinfo clkinfo; 468 /* 469 * Construct clockinfo structure. 470 */ 471 bzero(&clkinfo, sizeof(clkinfo)); 472 clkinfo.hz = hz; 473 clkinfo.tick = tick; 474 clkinfo.profhz = profhz; 475 clkinfo.stathz = stathz ? stathz : hz; 476 return (sysctl_handle_opaque(oidp, &clkinfo, sizeof clkinfo, req)); 477 } 478 479 SYSCTL_PROC(_kern, KERN_CLOCKRATE, clockrate, CTLTYPE_STRUCT|CTLFLAG_RD, 480 0, 0, sysctl_kern_clockrate, "S,clockinfo", 481 "Rate and period of various kernel clocks"); 482