machine.c revision 1.34
1/* $OpenBSD: machine.c,v 1.34 2003/06/17 00:51:29 jfb Exp $ */ 2 3/*- 4 * Copyright (c) 1994 Thorsten Lockert <tholo@sigmasoft.com> 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. The name of the author may not be used to endorse or promote products 16 * derived from this software without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, 19 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY 20 * AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL 21 * THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 22 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 23 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 24 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 25 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 26 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 27 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 28 * 29 * AUTHOR: Thorsten Lockert <tholo@sigmasoft.com> 30 * Adapted from BSD4.4 by Christos Zoulas <christos@ee.cornell.edu> 31 * Patch for process wait display by Jarl F. Greipsland <jarle@idt.unit.no> 32 * Patch for -DORDER by Kenneth Stailey <kstailey@disclosure.com> 33 * Patch for new swapctl(2) by Tobias Weingartner <weingart@openbsd.org> 34 */ 35 36#include <sys/types.h> 37#include <sys/signal.h> 38#include <sys/param.h> 39#include <stdio.h> 40#include <stdlib.h> 41#include <string.h> 42#include <limits.h> 43#include <err.h> 44#include <math.h> 45#include <unistd.h> 46#include <sys/errno.h> 47#include <sys/sysctl.h> 48#include <sys/dir.h> 49#include <sys/dkstat.h> 50#include <sys/file.h> 51#include <sys/time.h> 52#include <sys/resource.h> 53#include <sys/swap.h> 54#include <err.h> 55 56#include "top.h" 57#include "display.h" 58#include "machine.h" 59#include "utils.h" 60#include "loadavg.h" 61 62static int swapmode(int *, int *); 63 64/* get_process_info passes back a handle. This is what it looks like: */ 65 66struct handle { 67 struct kinfo_proc **next_proc; /* points to next valid proc pointer */ 68 int remaining; /* number of pointers remaining */ 69}; 70 71#define PP(pp, field) ((pp)->kp_proc . field) 72#define EP(pp, field) ((pp)->kp_eproc . field) 73#define VP(pp, field) ((pp)->kp_eproc.e_vm . field) 74 75/* what we consider to be process size: */ 76#define PROCSIZE(pp) (VP((pp), vm_tsize) + VP((pp), vm_dsize) + VP((pp), vm_ssize)) 77 78/* 79 * These definitions control the format of the per-process area 80 */ 81static char header[] = 82 " PID X PRI NICE SIZE RES STATE WAIT TIME CPU COMMAND"; 83 84/* 0123456 -- field to fill in starts at header+6 */ 85#define UNAME_START 6 86 87#define Proc_format \ 88 "%5d %-8.8s %3d %4d %5s %5s %-5s %-6.6s %6s %5.2f%% %.14s" 89 90/* process state names for the "STATE" column of the display */ 91/* 92 * the extra nulls in the string "run" are for adding a slash and the 93 * processor number when needed 94 */ 95 96char *state_abbrev[] = { 97 "", "start", "run\0\0\0", "sleep", "stop", "zomb", 98}; 99 100static int stathz; 101 102/* these are for calculating cpu state percentages */ 103static long cp_time[CPUSTATES]; 104static long cp_old[CPUSTATES]; 105static long cp_diff[CPUSTATES]; 106 107/* these are for detailing the process states */ 108int process_states[7]; 109char *procstatenames[] = { 110 "", " starting, ", " running, ", " idle, ", 111 " stopped, ", " zombie, ", 112 NULL 113}; 114 115/* these are for detailing the cpu states */ 116int cpu_states[CPUSTATES]; 117char *cpustatenames[] = { 118 "user", "nice", "system", "interrupt", "idle", NULL 119}; 120 121/* these are for detailing the memory statistics */ 122int memory_stats[8]; 123char *memorynames[] = { 124 "Real: ", "K/", "K act/tot ", "Free: ", "K ", 125 "Swap: ", "K/", "K used/tot", 126 NULL 127}; 128 129/* these are names given to allowed sorting orders -- first is default */ 130char *ordernames[] = { 131 "cpu", "size", "res", "time", "pri", NULL 132}; 133 134/* these are for keeping track of the proc array */ 135static int nproc; 136static int onproc = -1; 137static int pref_len; 138static struct kinfo_proc *pbase; 139static struct kinfo_proc **pref; 140 141/* these are for getting the memory statistics */ 142static int pageshift; /* log base 2 of the pagesize */ 143 144/* define pagetok in terms of pageshift */ 145#define pagetok(size) ((size) << pageshift) 146 147unsigned int maxslp; 148 149int 150getstathz(void) 151{ 152 struct clockinfo cinf; 153 size_t size = sizeof(cinf); 154 int mib[2]; 155 156 mib[0] = CTL_KERN; 157 mib[1] = KERN_CLOCKRATE; 158 if (sysctl(mib, 2, &cinf, &size, NULL, 0) == -1) 159 return (-1); 160 return (cinf.stathz); 161} 162 163int 164machine_init(struct statics *statics) 165{ 166 int pagesize; 167 168 stathz = getstathz(); 169 if (stathz == -1) 170 return (-1); 171 172 pbase = NULL; 173 pref = NULL; 174 onproc = -1; 175 nproc = 0; 176 177 /* 178 * get the page size with "getpagesize" and calculate pageshift from 179 * it 180 */ 181 pagesize = getpagesize(); 182 pageshift = 0; 183 while (pagesize > 1) { 184 pageshift++; 185 pagesize >>= 1; 186 } 187 188 /* we only need the amount of log(2)1024 for our conversion */ 189 pageshift -= LOG1024; 190 191 /* fill in the statics information */ 192 statics->procstate_names = procstatenames; 193 statics->cpustate_names = cpustatenames; 194 statics->memory_names = memorynames; 195 statics->order_names = ordernames; 196 return (0); 197} 198 199char * 200format_header(char *uname_field) 201{ 202 char *ptr; 203 204 ptr = header + UNAME_START; 205 while (*uname_field != '\0') 206 *ptr++ = *uname_field++; 207 return (header); 208} 209 210void 211get_system_info(struct system_info *si) 212{ 213 static int sysload_mib[] = {CTL_VM, VM_LOADAVG}; 214 static int vmtotal_mib[] = {CTL_VM, VM_METER}; 215 static int cp_time_mib[] = {CTL_KERN, KERN_CPTIME}; 216 struct loadavg sysload; 217 struct vmtotal vmtotal; 218 double *infoloadp; 219 int total, i; 220 size_t size; 221 222 size = sizeof(cp_time); 223 if (sysctl(cp_time_mib, 2, &cp_time, &size, NULL, 0) < 0) { 224 warn("sysctl kern.cp_time failed"); 225 total = 0; 226 } 227 size = sizeof(sysload); 228 if (sysctl(sysload_mib, 2, &sysload, &size, NULL, 0) < 0) { 229 warn("sysctl failed"); 230 total = 0; 231 } 232 infoloadp = si->load_avg; 233 for (i = 0; i < 3; i++) 234 *infoloadp++ = ((double) sysload.ldavg[i]) / sysload.fscale; 235 236 /* convert cp_time counts to percentages */ 237 total = percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff); 238 239 /* get total -- systemwide main memory usage structure */ 240 size = sizeof(vmtotal); 241 if (sysctl(vmtotal_mib, 2, &vmtotal, &size, NULL, 0) < 0) { 242 warn("sysctl failed"); 243 bzero(&vmtotal, sizeof(vmtotal)); 244 } 245 /* convert memory stats to Kbytes */ 246 memory_stats[0] = -1; 247 memory_stats[1] = pagetok(vmtotal.t_arm); 248 memory_stats[2] = pagetok(vmtotal.t_rm); 249 memory_stats[3] = -1; 250 memory_stats[4] = pagetok(vmtotal.t_free); 251 memory_stats[5] = -1; 252 253 if (!swapmode(&memory_stats[6], &memory_stats[7])) { 254 memory_stats[6] = 0; 255 memory_stats[7] = 0; 256 } 257 258 /* set arrays and strings */ 259 si->cpustates = cpu_states; 260 si->memory = memory_stats; 261 si->last_pid = -1; 262} 263 264static struct handle handle; 265 266struct kinfo_proc * 267getprocs(int op, int arg, int *cnt) 268{ 269 size_t size = sizeof(int); 270 int mib[4] = {CTL_KERN, KERN_PROC, 0, 0}; 271 int smib[2] = {CTL_KERN, KERN_NPROCS}; 272 static int maxslp_mib[] = {CTL_VM, VM_MAXSLP}; 273 static struct kinfo_proc *procbase; 274 int st; 275 276 mib[2] = op; 277 mib[3] = arg; 278 279 size = sizeof(maxslp); 280 if (sysctl(maxslp_mib, 2, &maxslp, &size, NULL, 0) < 0) { 281 warn("sysctl vm.maxslp failed"); 282 return (0); 283 } 284 st = sysctl(smib, 2, cnt, &size, NULL, 0); 285 if (st == -1) { 286 /* _kvm_syserr(kd, kd->program, "kvm_getprocs"); */ 287 return (0); 288 } 289 if (procbase) 290 free(procbase); 291 size = (6 * (*cnt) * sizeof(struct kinfo_proc)) / 5; 292 procbase = (struct kinfo_proc *) malloc(size); 293 if (procbase == NULL) 294 return (0); 295 st = sysctl(mib, 4, procbase, &size, NULL, 0); 296 if (st == -1) { 297 /* _kvm_syserr(kd, kd->program, "kvm_getprocs"); */ 298 return (0); 299 } 300 if (size % sizeof(struct kinfo_proc) != 0) { 301 /* 302 * _kvm_err(kd, kd->program, "proc size mismatch (%d total, 303 * %d chunks)", size, sizeof(struct kinfo_proc)); 304 */ 305 return (0); 306 } 307 return (procbase); 308} 309 310caddr_t 311get_process_info(struct system_info *si, struct process_select *sel, 312 int (*compare) (const void *, const void *)) 313{ 314 int show_idle, show_system, show_uid; 315 int total_procs, active_procs, i; 316 struct kinfo_proc **prefp, *pp; 317 318 if ((pbase = getprocs(KERN_PROC_KTHREAD, 0, &nproc)) == NULL) { 319 /* warnx("%s", kvm_geterr(kd)); */ 320 quit(23); 321 } 322 if (nproc > onproc) 323 pref = (struct kinfo_proc **)realloc(pref, 324 sizeof(struct kinfo_proc *) * (onproc = nproc)); 325 if (pref == NULL) { 326 warnx("Out of memory."); 327 quit(23); 328 } 329 /* get a pointer to the states summary array */ 330 si->procstates = process_states; 331 332 /* set up flags which define what we are going to select */ 333 show_idle = sel->idle; 334 show_system = sel->system; 335 show_uid = sel->uid != (uid_t)-1; 336 337 /* count up process states and get pointers to interesting procs */ 338 total_procs = 0; 339 active_procs = 0; 340 memset((char *) process_states, 0, sizeof(process_states)); 341 prefp = pref; 342 for (pp = pbase, i = 0; i < nproc; pp++, i++) { 343 /* 344 * Place pointers to each valid proc structure in pref[]. 345 * Process slots that are actually in use have a non-zero 346 * status field. Processes with SSYS set are system 347 * processes---these get ignored unless show_sysprocs is set. 348 */ 349 if (PP(pp, p_stat) != 0 && 350 (show_system || ((PP(pp, p_flag) & P_SYSTEM) == 0))) { 351 total_procs++; 352 process_states[(unsigned char) PP(pp, p_stat)]++; 353 if ((PP(pp, p_stat) != SZOMB) && 354 (show_idle || (PP(pp, p_pctcpu) != 0) || 355 (PP(pp, p_stat) == SRUN)) && 356 (!show_uid || EP(pp, e_pcred.p_ruid) == sel->uid)) { 357 *prefp++ = pp; 358 active_procs++; 359 } 360 } 361 } 362 363 /* if requested, sort the "interesting" processes */ 364 if (compare != NULL) 365 qsort((char *) pref, active_procs, 366 sizeof(struct kinfo_proc *), compare); 367 /* remember active and total counts */ 368 si->p_total = total_procs; 369 si->p_active = pref_len = active_procs; 370 371 /* pass back a handle */ 372 handle.next_proc = pref; 373 handle.remaining = active_procs; 374 return ((caddr_t) & handle); 375} 376 377char fmt[MAX_COLS]; /* static area where result is built */ 378 379char * 380format_next_process(caddr_t handle, char *(*get_userid)()) 381{ 382 char *p_wait, waddr[sizeof(void *) * 2 + 3]; /* Hexify void pointer */ 383 struct kinfo_proc *pp; 384 struct handle *hp; 385 int cputime; 386 double pct; 387 388 /* find and remember the next proc structure */ 389 hp = (struct handle *) handle; 390 pp = *(hp->next_proc++); 391 hp->remaining--; 392 393 /* get the process's user struct and set cputime */ 394 if ((PP(pp, p_flag) & P_INMEM) == 0) { 395 /* 396 * Print swapped processes as <pname> 397 */ 398 char *comm = PP(pp, p_comm); 399 char buf[sizeof(PP(pp, p_comm))]; 400 401 (void) strlcpy(buf, comm, sizeof buf); 402 comm[0] = '<'; 403 (void) strlcpy(&comm[1], buf, sizeof buf - 1); 404 (void) strlcat(comm, ">", sizeof buf); 405 } 406 cputime = (PP(pp, p_uticks) + PP(pp, p_sticks) + PP(pp, p_iticks)) / stathz; 407 408 /* calculate the base for cpu percentages */ 409 pct = pctdouble(PP(pp, p_pctcpu)); 410 411 if (PP(pp, p_wchan)) { 412 if (PP(pp, p_wmesg)) 413 p_wait = EP(pp, e_wmesg); 414 else { 415 snprintf(waddr, sizeof(waddr), "%lx", 416 (unsigned long) (PP(pp, p_wchan)) & ~KERNBASE); 417 p_wait = waddr; 418 } 419 } else 420 p_wait = "-"; 421 422 /* format this entry */ 423 snprintf(fmt, sizeof fmt, Proc_format, 424 PP(pp, p_pid), (*get_userid) (EP(pp, e_pcred.p_ruid)), 425 PP(pp, p_priority) - PZERO, PP(pp, p_nice) - NZERO, 426 format_k(pagetok(PROCSIZE(pp))), 427 format_k(pagetok(VP(pp, vm_rssize))), 428 (PP(pp, p_stat) == SSLEEP && PP(pp, p_slptime) > maxslp) ? 429 "idle" : state_abbrev[(unsigned char) PP(pp, p_stat)], 430 p_wait, format_time(cputime), 100.0 * pct, 431 printable(PP(pp, p_comm))); 432 433 /* return the result */ 434 return (fmt); 435} 436 437/* comparison routine for qsort */ 438static unsigned char sorted_state[] = 439{ 440 0, /* not used */ 441 4, /* start */ 442 5, /* run */ 443 2, /* sleep */ 444 3, /* stop */ 445 1 /* zombie */ 446}; 447 448/* 449 * proc_compares - comparison functions for "qsort" 450 */ 451 452/* 453 * First, the possible comparison keys. These are defined in such a way 454 * that they can be merely listed in the source code to define the actual 455 * desired ordering. 456 */ 457 458#define ORDERKEY_PCTCPU \ 459 if (lresult = (pctcpu)PP(p2, p_pctcpu) - (pctcpu)PP(p1, p_pctcpu), \ 460 (result = lresult > 0 ? 1 : lresult < 0 ? -1 : 0) == 0) 461#define ORDERKEY_CPUTIME \ 462 if ((result = PP(p2, p_rtime.tv_sec) - PP(p1, p_rtime.tv_sec)) == 0) \ 463 if ((result = PP(p2, p_rtime.tv_usec) - \ 464 PP(p1, p_rtime.tv_usec)) == 0) 465#define ORDERKEY_STATE \ 466 if ((result = sorted_state[(unsigned char) PP(p2, p_stat)] - \ 467 sorted_state[(unsigned char) PP(p1, p_stat)]) == 0) 468#define ORDERKEY_PRIO \ 469 if ((result = PP(p2, p_priority) - PP(p1, p_priority)) == 0) 470#define ORDERKEY_RSSIZE \ 471 if ((result = VP(p2, vm_rssize) - VP(p1, vm_rssize)) == 0) 472#define ORDERKEY_MEM \ 473 if ((result = PROCSIZE(p2) - PROCSIZE(p1)) == 0) 474 475/* compare_cpu - the comparison function for sorting by cpu percentage */ 476int 477compare_cpu(const void *v1, const void *v2) 478{ 479 struct proc **pp1 = (struct proc **) v1; 480 struct proc **pp2 = (struct proc **) v2; 481 struct kinfo_proc *p1, *p2; 482 pctcpu lresult; 483 int result; 484 485 /* remove one level of indirection */ 486 p1 = *(struct kinfo_proc **) pp1; 487 p2 = *(struct kinfo_proc **) pp2; 488 489 ORDERKEY_PCTCPU 490 ORDERKEY_CPUTIME 491 ORDERKEY_STATE 492 ORDERKEY_PRIO 493 ORDERKEY_RSSIZE 494 ORDERKEY_MEM 495 ; 496 return (result); 497} 498 499/* compare_size - the comparison function for sorting by total memory usage */ 500int 501compare_size(const void *v1, const void *v2) 502{ 503 struct proc **pp1 = (struct proc **) v1; 504 struct proc **pp2 = (struct proc **) v2; 505 struct kinfo_proc *p1, *p2; 506 pctcpu lresult; 507 int result; 508 509 /* remove one level of indirection */ 510 p1 = *(struct kinfo_proc **) pp1; 511 p2 = *(struct kinfo_proc **) pp2; 512 513 ORDERKEY_MEM 514 ORDERKEY_RSSIZE 515 ORDERKEY_PCTCPU 516 ORDERKEY_CPUTIME 517 ORDERKEY_STATE 518 ORDERKEY_PRIO 519 ; 520 return (result); 521} 522 523/* compare_res - the comparison function for sorting by resident set size */ 524int 525compare_res(const void *v1, const void *v2) 526{ 527 struct proc **pp1 = (struct proc **) v1; 528 struct proc **pp2 = (struct proc **) v2; 529 struct kinfo_proc *p1, *p2; 530 pctcpu lresult; 531 int result; 532 533 /* remove one level of indirection */ 534 p1 = *(struct kinfo_proc **) pp1; 535 p2 = *(struct kinfo_proc **) pp2; 536 537 ORDERKEY_RSSIZE 538 ORDERKEY_MEM 539 ORDERKEY_PCTCPU 540 ORDERKEY_CPUTIME 541 ORDERKEY_STATE 542 ORDERKEY_PRIO 543 ; 544 return (result); 545} 546 547/* compare_time - the comparison function for sorting by CPU time */ 548int 549compare_time(const void *v1, const void *v2) 550{ 551 struct proc **pp1 = (struct proc **) v1; 552 struct proc **pp2 = (struct proc **) v2; 553 struct kinfo_proc *p1, *p2; 554 pctcpu lresult; 555 int result; 556 557 /* remove one level of indirection */ 558 p1 = *(struct kinfo_proc **) pp1; 559 p2 = *(struct kinfo_proc **) pp2; 560 561 ORDERKEY_CPUTIME 562 ORDERKEY_PCTCPU 563 ORDERKEY_STATE 564 ORDERKEY_PRIO 565 ORDERKEY_MEM 566 ORDERKEY_RSSIZE 567 ; 568 return (result); 569} 570 571/* compare_prio - the comparison function for sorting by CPU time */ 572int 573compare_prio(const void *v1, const void *v2) 574{ 575 struct proc **pp1 = (struct proc **) v1; 576 struct proc **pp2 = (struct proc **) v2; 577 struct kinfo_proc *p1, *p2; 578 pctcpu lresult; 579 int result; 580 581 /* remove one level of indirection */ 582 p1 = *(struct kinfo_proc **) pp1; 583 p2 = *(struct kinfo_proc **) pp2; 584 585 ORDERKEY_PRIO 586 ORDERKEY_PCTCPU 587 ORDERKEY_CPUTIME 588 ORDERKEY_STATE 589 ORDERKEY_RSSIZE 590 ORDERKEY_MEM 591 ; 592 return (result); 593} 594 595int (*proc_compares[])(const void *, const void *) = { 596 compare_cpu, 597 compare_size, 598 compare_res, 599 compare_time, 600 compare_prio, 601 NULL 602}; 603 604/* 605 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if 606 * the process does not exist. 607 * It is EXTREMLY IMPORTANT that this function work correctly. 608 * If top runs setuid root (as in SVR4), then this function 609 * is the only thing that stands in the way of a serious 610 * security problem. It validates requests for the "kill" 611 * and "renice" commands. 612 */ 613uid_t 614proc_owner(pid_t pid) 615{ 616 struct kinfo_proc **prefp, *pp; 617 int cnt; 618 619 prefp = pref; 620 cnt = pref_len; 621 while (--cnt >= 0) { 622 pp = *prefp++; 623 if (PP(pp, p_pid) == pid) 624 return ((uid_t) EP(pp, e_pcred.p_ruid)); 625 } 626 return (uid_t)(-1); 627} 628 629/* 630 * swapmode is rewritten by Tobias Weingartner <weingart@openbsd.org> 631 * to be based on the new swapctl(2) system call. 632 */ 633static int 634swapmode(int *used, int *total) 635{ 636 struct swapent *swdev; 637 int nswap, rnswap, i; 638 639 nswap = swapctl(SWAP_NSWAP, 0, 0); 640 if (nswap == 0) 641 return 0; 642 643 swdev = malloc(nswap * sizeof(*swdev)); 644 if (swdev == NULL) 645 return 0; 646 647 rnswap = swapctl(SWAP_STATS, swdev, nswap); 648 if (rnswap == -1) 649 return 0; 650 651 /* if rnswap != nswap, then what? */ 652 653 /* Total things up */ 654 *total = *used = 0; 655 for (i = 0; i < nswap; i++) { 656 if (swdev[i].se_flags & SWF_ENABLE) { 657 *used += (swdev[i].se_inuse / (1024 / DEV_BSIZE)); 658 *total += (swdev[i].se_nblks / (1024 / DEV_BSIZE)); 659 } 660 } 661 free(swdev); 662 return 1; 663} 664