machine.c revision 1.35
1/* $OpenBSD: machine.c,v 1.35 2003/06/18 08:36:31 deraadt 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 size_t size; 220 int i; 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"); 225 226 size = sizeof(sysload); 227 if (sysctl(sysload_mib, 2, &sysload, &size, NULL, 0) < 0) 228 warn("sysctl failed"); 229 infoloadp = si->load_avg; 230 for (i = 0; i < 3; i++) 231 *infoloadp++ = ((double) sysload.ldavg[i]) / sysload.fscale; 232 233 /* convert cp_time counts to percentages */ 234 (void) percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff); 235 236 /* get total -- systemwide main memory usage structure */ 237 size = sizeof(vmtotal); 238 if (sysctl(vmtotal_mib, 2, &vmtotal, &size, NULL, 0) < 0) { 239 warn("sysctl failed"); 240 bzero(&vmtotal, sizeof(vmtotal)); 241 } 242 /* convert memory stats to Kbytes */ 243 memory_stats[0] = -1; 244 memory_stats[1] = pagetok(vmtotal.t_arm); 245 memory_stats[2] = pagetok(vmtotal.t_rm); 246 memory_stats[3] = -1; 247 memory_stats[4] = pagetok(vmtotal.t_free); 248 memory_stats[5] = -1; 249 250 if (!swapmode(&memory_stats[6], &memory_stats[7])) { 251 memory_stats[6] = 0; 252 memory_stats[7] = 0; 253 } 254 255 /* set arrays and strings */ 256 si->cpustates = cpu_states; 257 si->memory = memory_stats; 258 si->last_pid = -1; 259} 260 261static struct handle handle; 262 263struct kinfo_proc * 264getprocs(int op, int arg, int *cnt) 265{ 266 size_t size = sizeof(int); 267 int mib[4] = {CTL_KERN, KERN_PROC, 0, 0}; 268 int smib[2] = {CTL_KERN, KERN_NPROCS}; 269 static int maxslp_mib[] = {CTL_VM, VM_MAXSLP}; 270 static struct kinfo_proc *procbase; 271 int st; 272 273 mib[2] = op; 274 mib[3] = arg; 275 276 size = sizeof(maxslp); 277 if (sysctl(maxslp_mib, 2, &maxslp, &size, NULL, 0) < 0) { 278 warn("sysctl vm.maxslp failed"); 279 return (0); 280 } 281 st = sysctl(smib, 2, cnt, &size, NULL, 0); 282 if (st == -1) { 283 /* _kvm_syserr(kd, kd->program, "kvm_getprocs"); */ 284 return (0); 285 } 286 if (procbase) 287 free(procbase); 288 size = (6 * (*cnt) * sizeof(struct kinfo_proc)) / 5; 289 procbase = (struct kinfo_proc *) malloc(size); 290 if (procbase == NULL) 291 return (0); 292 st = sysctl(mib, 4, procbase, &size, NULL, 0); 293 if (st == -1) { 294 /* _kvm_syserr(kd, kd->program, "kvm_getprocs"); */ 295 return (0); 296 } 297 if (size % sizeof(struct kinfo_proc) != 0) { 298 /* 299 * _kvm_err(kd, kd->program, "proc size mismatch (%d total, 300 * %d chunks)", size, sizeof(struct kinfo_proc)); 301 */ 302 return (0); 303 } 304 return (procbase); 305} 306 307caddr_t 308get_process_info(struct system_info *si, struct process_select *sel, 309 int (*compare) (const void *, const void *)) 310{ 311 int show_idle, show_system, show_uid; 312 int total_procs, active_procs, i; 313 struct kinfo_proc **prefp, *pp; 314 315 if ((pbase = getprocs(KERN_PROC_KTHREAD, 0, &nproc)) == NULL) { 316 /* warnx("%s", kvm_geterr(kd)); */ 317 quit(23); 318 } 319 if (nproc > onproc) 320 pref = (struct kinfo_proc **)realloc(pref, 321 sizeof(struct kinfo_proc *) * (onproc = nproc)); 322 if (pref == NULL) { 323 warnx("Out of memory."); 324 quit(23); 325 } 326 /* get a pointer to the states summary array */ 327 si->procstates = process_states; 328 329 /* set up flags which define what we are going to select */ 330 show_idle = sel->idle; 331 show_system = sel->system; 332 show_uid = sel->uid != (uid_t)-1; 333 334 /* count up process states and get pointers to interesting procs */ 335 total_procs = 0; 336 active_procs = 0; 337 memset((char *) process_states, 0, sizeof(process_states)); 338 prefp = pref; 339 for (pp = pbase, i = 0; i < nproc; pp++, i++) { 340 /* 341 * Place pointers to each valid proc structure in pref[]. 342 * Process slots that are actually in use have a non-zero 343 * status field. Processes with SSYS set are system 344 * processes---these get ignored unless show_sysprocs is set. 345 */ 346 if (PP(pp, p_stat) != 0 && 347 (show_system || ((PP(pp, p_flag) & P_SYSTEM) == 0))) { 348 total_procs++; 349 process_states[(unsigned char) PP(pp, p_stat)]++; 350 if ((PP(pp, p_stat) != SZOMB) && 351 (show_idle || (PP(pp, p_pctcpu) != 0) || 352 (PP(pp, p_stat) == SRUN)) && 353 (!show_uid || EP(pp, e_pcred.p_ruid) == sel->uid)) { 354 *prefp++ = pp; 355 active_procs++; 356 } 357 } 358 } 359 360 /* if requested, sort the "interesting" processes */ 361 if (compare != NULL) 362 qsort((char *) pref, active_procs, 363 sizeof(struct kinfo_proc *), compare); 364 /* remember active and total counts */ 365 si->p_total = total_procs; 366 si->p_active = pref_len = active_procs; 367 368 /* pass back a handle */ 369 handle.next_proc = pref; 370 handle.remaining = active_procs; 371 return ((caddr_t) & handle); 372} 373 374char fmt[MAX_COLS]; /* static area where result is built */ 375 376char * 377format_next_process(caddr_t handle, char *(*get_userid)()) 378{ 379 char *p_wait, waddr[sizeof(void *) * 2 + 3]; /* Hexify void pointer */ 380 struct kinfo_proc *pp; 381 struct handle *hp; 382 int cputime; 383 double pct; 384 385 /* find and remember the next proc structure */ 386 hp = (struct handle *) handle; 387 pp = *(hp->next_proc++); 388 hp->remaining--; 389 390 /* get the process's user struct and set cputime */ 391 if ((PP(pp, p_flag) & P_INMEM) == 0) { 392 /* 393 * Print swapped processes as <pname> 394 */ 395 char *comm = PP(pp, p_comm); 396 char buf[sizeof(PP(pp, p_comm))]; 397 398 (void) strlcpy(buf, comm, sizeof buf); 399 comm[0] = '<'; 400 (void) strlcpy(&comm[1], buf, sizeof buf - 1); 401 (void) strlcat(comm, ">", sizeof buf); 402 } 403 cputime = (PP(pp, p_uticks) + PP(pp, p_sticks) + PP(pp, p_iticks)) / stathz; 404 405 /* calculate the base for cpu percentages */ 406 pct = pctdouble(PP(pp, p_pctcpu)); 407 408 if (PP(pp, p_wchan)) { 409 if (PP(pp, p_wmesg)) 410 p_wait = EP(pp, e_wmesg); 411 else { 412 snprintf(waddr, sizeof(waddr), "%lx", 413 (unsigned long) (PP(pp, p_wchan)) & ~KERNBASE); 414 p_wait = waddr; 415 } 416 } else 417 p_wait = "-"; 418 419 /* format this entry */ 420 snprintf(fmt, sizeof fmt, Proc_format, 421 PP(pp, p_pid), (*get_userid) (EP(pp, e_pcred.p_ruid)), 422 PP(pp, p_priority) - PZERO, PP(pp, p_nice) - NZERO, 423 format_k(pagetok(PROCSIZE(pp))), 424 format_k(pagetok(VP(pp, vm_rssize))), 425 (PP(pp, p_stat) == SSLEEP && PP(pp, p_slptime) > maxslp) ? 426 "idle" : state_abbrev[(unsigned char) PP(pp, p_stat)], 427 p_wait, format_time(cputime), 100.0 * pct, 428 printable(PP(pp, p_comm))); 429 430 /* return the result */ 431 return (fmt); 432} 433 434/* comparison routine for qsort */ 435static unsigned char sorted_state[] = 436{ 437 0, /* not used */ 438 4, /* start */ 439 5, /* run */ 440 2, /* sleep */ 441 3, /* stop */ 442 1 /* zombie */ 443}; 444 445/* 446 * proc_compares - comparison functions for "qsort" 447 */ 448 449/* 450 * First, the possible comparison keys. These are defined in such a way 451 * that they can be merely listed in the source code to define the actual 452 * desired ordering. 453 */ 454 455#define ORDERKEY_PCTCPU \ 456 if (lresult = (pctcpu)PP(p2, p_pctcpu) - (pctcpu)PP(p1, p_pctcpu), \ 457 (result = lresult > 0 ? 1 : lresult < 0 ? -1 : 0) == 0) 458#define ORDERKEY_CPUTIME \ 459 if ((result = PP(p2, p_rtime.tv_sec) - PP(p1, p_rtime.tv_sec)) == 0) \ 460 if ((result = PP(p2, p_rtime.tv_usec) - \ 461 PP(p1, p_rtime.tv_usec)) == 0) 462#define ORDERKEY_STATE \ 463 if ((result = sorted_state[(unsigned char) PP(p2, p_stat)] - \ 464 sorted_state[(unsigned char) PP(p1, p_stat)]) == 0) 465#define ORDERKEY_PRIO \ 466 if ((result = PP(p2, p_priority) - PP(p1, p_priority)) == 0) 467#define ORDERKEY_RSSIZE \ 468 if ((result = VP(p2, vm_rssize) - VP(p1, vm_rssize)) == 0) 469#define ORDERKEY_MEM \ 470 if ((result = PROCSIZE(p2) - PROCSIZE(p1)) == 0) 471 472/* compare_cpu - the comparison function for sorting by cpu percentage */ 473int 474compare_cpu(const void *v1, const void *v2) 475{ 476 struct proc **pp1 = (struct proc **) v1; 477 struct proc **pp2 = (struct proc **) v2; 478 struct kinfo_proc *p1, *p2; 479 pctcpu lresult; 480 int result; 481 482 /* remove one level of indirection */ 483 p1 = *(struct kinfo_proc **) pp1; 484 p2 = *(struct kinfo_proc **) pp2; 485 486 ORDERKEY_PCTCPU 487 ORDERKEY_CPUTIME 488 ORDERKEY_STATE 489 ORDERKEY_PRIO 490 ORDERKEY_RSSIZE 491 ORDERKEY_MEM 492 ; 493 return (result); 494} 495 496/* compare_size - the comparison function for sorting by total memory usage */ 497int 498compare_size(const void *v1, const void *v2) 499{ 500 struct proc **pp1 = (struct proc **) v1; 501 struct proc **pp2 = (struct proc **) v2; 502 struct kinfo_proc *p1, *p2; 503 pctcpu lresult; 504 int result; 505 506 /* remove one level of indirection */ 507 p1 = *(struct kinfo_proc **) pp1; 508 p2 = *(struct kinfo_proc **) pp2; 509 510 ORDERKEY_MEM 511 ORDERKEY_RSSIZE 512 ORDERKEY_PCTCPU 513 ORDERKEY_CPUTIME 514 ORDERKEY_STATE 515 ORDERKEY_PRIO 516 ; 517 return (result); 518} 519 520/* compare_res - the comparison function for sorting by resident set size */ 521int 522compare_res(const void *v1, const void *v2) 523{ 524 struct proc **pp1 = (struct proc **) v1; 525 struct proc **pp2 = (struct proc **) v2; 526 struct kinfo_proc *p1, *p2; 527 pctcpu lresult; 528 int result; 529 530 /* remove one level of indirection */ 531 p1 = *(struct kinfo_proc **) pp1; 532 p2 = *(struct kinfo_proc **) pp2; 533 534 ORDERKEY_RSSIZE 535 ORDERKEY_MEM 536 ORDERKEY_PCTCPU 537 ORDERKEY_CPUTIME 538 ORDERKEY_STATE 539 ORDERKEY_PRIO 540 ; 541 return (result); 542} 543 544/* compare_time - the comparison function for sorting by CPU time */ 545int 546compare_time(const void *v1, const void *v2) 547{ 548 struct proc **pp1 = (struct proc **) v1; 549 struct proc **pp2 = (struct proc **) v2; 550 struct kinfo_proc *p1, *p2; 551 pctcpu lresult; 552 int result; 553 554 /* remove one level of indirection */ 555 p1 = *(struct kinfo_proc **) pp1; 556 p2 = *(struct kinfo_proc **) pp2; 557 558 ORDERKEY_CPUTIME 559 ORDERKEY_PCTCPU 560 ORDERKEY_STATE 561 ORDERKEY_PRIO 562 ORDERKEY_MEM 563 ORDERKEY_RSSIZE 564 ; 565 return (result); 566} 567 568/* compare_prio - the comparison function for sorting by CPU time */ 569int 570compare_prio(const void *v1, const void *v2) 571{ 572 struct proc **pp1 = (struct proc **) v1; 573 struct proc **pp2 = (struct proc **) v2; 574 struct kinfo_proc *p1, *p2; 575 pctcpu lresult; 576 int result; 577 578 /* remove one level of indirection */ 579 p1 = *(struct kinfo_proc **) pp1; 580 p2 = *(struct kinfo_proc **) pp2; 581 582 ORDERKEY_PRIO 583 ORDERKEY_PCTCPU 584 ORDERKEY_CPUTIME 585 ORDERKEY_STATE 586 ORDERKEY_RSSIZE 587 ORDERKEY_MEM 588 ; 589 return (result); 590} 591 592int (*proc_compares[])(const void *, const void *) = { 593 compare_cpu, 594 compare_size, 595 compare_res, 596 compare_time, 597 compare_prio, 598 NULL 599}; 600 601/* 602 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if 603 * the process does not exist. 604 * It is EXTREMLY IMPORTANT that this function work correctly. 605 * If top runs setuid root (as in SVR4), then this function 606 * is the only thing that stands in the way of a serious 607 * security problem. It validates requests for the "kill" 608 * and "renice" commands. 609 */ 610uid_t 611proc_owner(pid_t pid) 612{ 613 struct kinfo_proc **prefp, *pp; 614 int cnt; 615 616 prefp = pref; 617 cnt = pref_len; 618 while (--cnt >= 0) { 619 pp = *prefp++; 620 if (PP(pp, p_pid) == pid) 621 return ((uid_t) EP(pp, e_pcred.p_ruid)); 622 } 623 return (uid_t)(-1); 624} 625 626/* 627 * swapmode is rewritten by Tobias Weingartner <weingart@openbsd.org> 628 * to be based on the new swapctl(2) system call. 629 */ 630static int 631swapmode(int *used, int *total) 632{ 633 struct swapent *swdev; 634 int nswap, rnswap, i; 635 636 nswap = swapctl(SWAP_NSWAP, 0, 0); 637 if (nswap == 0) 638 return 0; 639 640 swdev = malloc(nswap * sizeof(*swdev)); 641 if (swdev == NULL) 642 return 0; 643 644 rnswap = swapctl(SWAP_STATS, swdev, nswap); 645 if (rnswap == -1) 646 return 0; 647 648 /* if rnswap != nswap, then what? */ 649 650 /* Total things up */ 651 *total = *used = 0; 652 for (i = 0; i < nswap; i++) { 653 if (swdev[i].se_flags & SWF_ENABLE) { 654 *used += (swdev[i].se_inuse / (1024 / DEV_BSIZE)); 655 *total += (swdev[i].se_nblks / (1024 / DEV_BSIZE)); 656 } 657 } 658 free(swdev); 659 return 1; 660} 661