machine.c revision 1.38
1/* $OpenBSD: machine.c,v 1.38 2004/05/09 22:14:15 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_proc2 **next_proc; /* points to next valid proc pointer */ 68 int remaining; /* number of pointers remaining */ 69}; 70 71/* what we consider to be process size: */ 72#define PROCSIZE(pp) ((pp)->p_vm_tsize + (pp)->p_vm_dsize + (pp)->p_vm_ssize) 73 74/* 75 * These definitions control the format of the per-process area 76 */ 77static char header[] = 78 " PID X PRI NICE SIZE RES STATE WAIT TIME CPU COMMAND"; 79 80/* 0123456 -- field to fill in starts at header+6 */ 81#define UNAME_START 6 82 83#define Proc_format \ 84 "%5d %-8.8s %3d %4d %5s %5s %-5s %-6.6s %6s %5.2f%% %.14s" 85 86/* process state names for the "STATE" column of the display */ 87/* 88 * the extra nulls in the string "run" are for adding a slash and the 89 * processor number when needed 90 */ 91 92char *state_abbrev[] = { 93 "", "start", "run\0\0\0", "sleep", "stop", "zomb", 94}; 95 96static int stathz; 97 98/* these are for calculating cpu state percentages */ 99static long cp_time[CPUSTATES]; 100static long cp_old[CPUSTATES]; 101static long cp_diff[CPUSTATES]; 102 103/* these are for detailing the process states */ 104int process_states[7]; 105char *procstatenames[] = { 106 "", " starting, ", " running, ", " idle, ", 107 " stopped, ", " zombie, ", 108 NULL 109}; 110 111/* these are for detailing the cpu states */ 112int cpu_states[CPUSTATES]; 113char *cpustatenames[] = { 114 "user", "nice", "system", "interrupt", "idle", NULL 115}; 116 117/* these are for detailing the memory statistics */ 118int memory_stats[8]; 119char *memorynames[] = { 120 "Real: ", "K/", "K act/tot ", "Free: ", "K ", 121 "Swap: ", "K/", "K used/tot", 122 NULL 123}; 124 125/* these are names given to allowed sorting orders -- first is default */ 126char *ordernames[] = { 127 "cpu", "size", "res", "time", "pri", NULL 128}; 129 130/* these are for keeping track of the proc array */ 131static int nproc; 132static int onproc = -1; 133static int pref_len; 134static struct kinfo_proc2 *pbase; 135static struct kinfo_proc2 **pref; 136 137/* these are for getting the memory statistics */ 138static int pageshift; /* log base 2 of the pagesize */ 139 140/* define pagetok in terms of pageshift */ 141#define pagetok(size) ((size) << pageshift) 142 143unsigned int maxslp; 144 145static int 146getstathz(void) 147{ 148 struct clockinfo cinf; 149 size_t size = sizeof(cinf); 150 int mib[2]; 151 152 mib[0] = CTL_KERN; 153 mib[1] = KERN_CLOCKRATE; 154 if (sysctl(mib, 2, &cinf, &size, NULL, 0) == -1) 155 return (-1); 156 return (cinf.stathz); 157} 158 159int 160machine_init(struct statics *statics) 161{ 162 int pagesize; 163 164 stathz = getstathz(); 165 if (stathz == -1) 166 return (-1); 167 168 pbase = NULL; 169 pref = NULL; 170 onproc = -1; 171 nproc = 0; 172 173 /* 174 * get the page size with "getpagesize" and calculate pageshift from 175 * it 176 */ 177 pagesize = getpagesize(); 178 pageshift = 0; 179 while (pagesize > 1) { 180 pageshift++; 181 pagesize >>= 1; 182 } 183 184 /* we only need the amount of log(2)1024 for our conversion */ 185 pageshift -= LOG1024; 186 187 /* fill in the statics information */ 188 statics->procstate_names = procstatenames; 189 statics->cpustate_names = cpustatenames; 190 statics->memory_names = memorynames; 191 statics->order_names = ordernames; 192 return (0); 193} 194 195char * 196format_header(char *uname_field) 197{ 198 char *ptr; 199 200 ptr = header + UNAME_START; 201 while (*uname_field != '\0') 202 *ptr++ = *uname_field++; 203 return (header); 204} 205 206void 207get_system_info(struct system_info *si) 208{ 209 static int sysload_mib[] = {CTL_VM, VM_LOADAVG}; 210 static int vmtotal_mib[] = {CTL_VM, VM_METER}; 211 static int cp_time_mib[] = {CTL_KERN, KERN_CPTIME}; 212 struct loadavg sysload; 213 struct vmtotal vmtotal; 214 double *infoloadp; 215 size_t size; 216 int i; 217 218 size = sizeof(cp_time); 219 if (sysctl(cp_time_mib, 2, &cp_time, &size, NULL, 0) < 0) 220 warn("sysctl kern.cp_time failed"); 221 222 size = sizeof(sysload); 223 if (sysctl(sysload_mib, 2, &sysload, &size, NULL, 0) < 0) 224 warn("sysctl failed"); 225 infoloadp = si->load_avg; 226 for (i = 0; i < 3; i++) 227 *infoloadp++ = ((double) sysload.ldavg[i]) / sysload.fscale; 228 229 /* convert cp_time counts to percentages */ 230 (void) percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff); 231 232 /* get total -- systemwide main memory usage structure */ 233 size = sizeof(vmtotal); 234 if (sysctl(vmtotal_mib, 2, &vmtotal, &size, NULL, 0) < 0) { 235 warn("sysctl failed"); 236 bzero(&vmtotal, sizeof(vmtotal)); 237 } 238 /* convert memory stats to Kbytes */ 239 memory_stats[0] = -1; 240 memory_stats[1] = pagetok(vmtotal.t_arm); 241 memory_stats[2] = pagetok(vmtotal.t_rm); 242 memory_stats[3] = -1; 243 memory_stats[4] = pagetok(vmtotal.t_free); 244 memory_stats[5] = -1; 245 246 if (!swapmode(&memory_stats[6], &memory_stats[7])) { 247 memory_stats[6] = 0; 248 memory_stats[7] = 0; 249 } 250 251 /* set arrays and strings */ 252 si->cpustates = cpu_states; 253 si->memory = memory_stats; 254 si->last_pid = -1; 255} 256 257static struct handle handle; 258 259static struct kinfo_proc2 * 260getprocs(int op, int arg, int *cnt) 261{ 262 size_t size; 263 int mib[6] = {CTL_KERN, KERN_PROC2, 0, 0, sizeof(struct kinfo_proc2), 0}; 264 static int maxslp_mib[] = {CTL_VM, VM_MAXSLP}; 265 static struct kinfo_proc2 *procbase; 266 int st; 267 268 mib[2] = op; 269 mib[3] = arg; 270 271 size = sizeof(maxslp); 272 if (sysctl(maxslp_mib, 2, &maxslp, &size, NULL, 0) < 0) { 273 warn("sysctl vm.maxslp failed"); 274 return (0); 275 } 276 retry: 277 free(procbase); 278 st = sysctl(mib, 6, NULL, &size, NULL, 0); 279 if (st == -1) { 280 /* _kvm_syserr(kd, kd->program, "kvm_getproc2"); */ 281 return (0); 282 } 283 size = 5 * size / 4; /* extra slop */ 284 if ((procbase = malloc(size)) == NULL) 285 return (0); 286 mib[5] = (int)(size / sizeof(struct kinfo_proc2)); 287 st = sysctl(mib, 6, procbase, &size, NULL, 0); 288 if (st == -1) { 289 if (errno == ENOMEM) 290 goto retry; 291 /* _kvm_syserr(kd, kd->program, "kvm_getproc2"); */ 292 return (0); 293 } 294 *cnt = (int)(size / sizeof(struct kinfo_proc2)); 295 return (procbase); 296} 297 298caddr_t 299get_process_info(struct system_info *si, struct process_select *sel, 300 int (*compare) (const void *, const void *)) 301{ 302 int show_idle, show_system, show_uid; 303 int total_procs, active_procs, i; 304 struct kinfo_proc2 **prefp, *pp; 305 306 if ((pbase = getprocs(KERN_PROC_KTHREAD, 0, &nproc)) == NULL) { 307 /* warnx("%s", kvm_geterr(kd)); */ 308 quit(23); 309 } 310 if (nproc > onproc) 311 pref = (struct kinfo_proc2 **)realloc(pref, 312 sizeof(struct kinfo_proc2 *) * (onproc = nproc)); 313 if (pref == NULL) { 314 warnx("Out of memory."); 315 quit(23); 316 } 317 /* get a pointer to the states summary array */ 318 si->procstates = process_states; 319 320 /* set up flags which define what we are going to select */ 321 show_idle = sel->idle; 322 show_system = sel->system; 323 show_uid = sel->uid != (uid_t)-1; 324 325 /* count up process states and get pointers to interesting procs */ 326 total_procs = 0; 327 active_procs = 0; 328 memset((char *) process_states, 0, sizeof(process_states)); 329 prefp = pref; 330 for (pp = pbase, i = 0; i < nproc; pp++, i++) { 331 /* 332 * Place pointers to each valid proc structure in pref[]. 333 * Process slots that are actually in use have a non-zero 334 * status field. Processes with SSYS set are system 335 * processes---these get ignored unless show_sysprocs is set. 336 */ 337 if (pp->p_stat != 0 && 338 (show_system || (pp->p_flag & P_SYSTEM) == 0)) { 339 total_procs++; 340 process_states[(unsigned char) pp->p_stat]++; 341 if (pp->p_stat != SZOMB && 342 (show_idle || pp->p_pctcpu != 0 || 343 pp->p_stat == SRUN) && 344 (!show_uid || pp->p_ruid == sel->uid)) { 345 *prefp++ = pp; 346 active_procs++; 347 } 348 } 349 } 350 351 /* if requested, sort the "interesting" processes */ 352 if (compare != NULL) 353 qsort((char *) pref, active_procs, 354 sizeof(struct kinfo_proc2 *), compare); 355 /* remember active and total counts */ 356 si->p_total = total_procs; 357 si->p_active = pref_len = active_procs; 358 359 /* pass back a handle */ 360 handle.next_proc = pref; 361 handle.remaining = active_procs; 362 return ((caddr_t) & handle); 363} 364 365char fmt[MAX_COLS]; /* static area where result is built */ 366 367char * 368format_next_process(caddr_t handle, char *(*get_userid)(uid_t)) 369{ 370 char *p_wait, waddr[sizeof(void *) * 2 + 3]; /* Hexify void pointer */ 371 struct kinfo_proc2 *pp; 372 struct handle *hp; 373 int cputime; 374 double pct; 375 376 /* find and remember the next proc structure */ 377 hp = (struct handle *) handle; 378 pp = *(hp->next_proc++); 379 hp->remaining--; 380 381 if ((pp->p_flag & P_INMEM) == 0) { 382 /* 383 * Print swapped processes as <pname> 384 */ 385 char buf[sizeof(pp->p_comm)]; 386 387 (void) strlcpy(buf, pp->p_comm, sizeof(buf)); 388 (void) snprintf(pp->p_comm, sizeof(pp->p_comm), "<%s>", buf); 389 } 390 cputime = (pp->p_uticks + pp->p_sticks + pp->p_iticks) / stathz; 391 392 /* calculate the base for cpu percentages */ 393 pct = pctdouble(pp->p_pctcpu); 394 395 if (pp->p_wchan) { 396 if (pp->p_wmesg) 397 p_wait = pp->p_wmesg; 398 else { 399 snprintf(waddr, sizeof(waddr), "%llx", 400 pp->p_wchan & ~KERNBASE); 401 p_wait = waddr; 402 } 403 } else 404 p_wait = "-"; 405 406 /* format this entry */ 407 snprintf(fmt, sizeof fmt, Proc_format, 408 pp->p_pid, (*get_userid)(pp->p_ruid), 409 pp->p_priority - PZERO, pp->p_nice - NZERO, 410 format_k(pagetok(PROCSIZE(pp))), 411 format_k(pagetok(pp->p_vm_rssize)), 412 (pp->p_stat == SSLEEP && pp->p_slptime > maxslp) ? 413 "idle" : state_abbrev[(unsigned char)pp->p_stat], 414 p_wait, format_time(cputime), 100.0 * pct, 415 printable(pp->p_comm)); 416 417 /* return the result */ 418 return (fmt); 419} 420 421/* comparison routine for qsort */ 422static unsigned char sorted_state[] = 423{ 424 0, /* not used */ 425 4, /* start */ 426 5, /* run */ 427 2, /* sleep */ 428 3, /* stop */ 429 1 /* zombie */ 430}; 431 432/* 433 * proc_compares - comparison functions for "qsort" 434 */ 435 436/* 437 * First, the possible comparison keys. These are defined in such a way 438 * that they can be merely listed in the source code to define the actual 439 * desired ordering. 440 */ 441 442#define ORDERKEY_PCTCPU \ 443 if (lresult = (pctcpu)p2->p_pctcpu - (pctcpu)p1->p_pctcpu, \ 444 (result = lresult > 0 ? 1 : lresult < 0 ? -1 : 0) == 0) 445#define ORDERKEY_CPUTIME \ 446 if ((result = p2->p_rtime_sec - p1->p_rtime_sec) == 0) \ 447 if ((result = p2->p_rtime_usec - p1->p_rtime_usec) == 0) 448#define ORDERKEY_STATE \ 449 if ((result = sorted_state[(unsigned char)p2->p_stat] - \ 450 sorted_state[(unsigned char)p1->p_stat]) == 0) 451#define ORDERKEY_PRIO \ 452 if ((result = p2->p_priority - p1->p_priority) == 0) 453#define ORDERKEY_RSSIZE \ 454 if ((result = p2->p_vm_rssize - p1->p_vm_rssize) == 0) 455#define ORDERKEY_MEM \ 456 if ((result = PROCSIZE(p2) - PROCSIZE(p1)) == 0) 457 458/* compare_cpu - the comparison function for sorting by cpu percentage */ 459static int 460compare_cpu(const void *v1, const void *v2) 461{ 462 struct proc **pp1 = (struct proc **) v1; 463 struct proc **pp2 = (struct proc **) v2; 464 struct kinfo_proc2 *p1, *p2; 465 pctcpu lresult; 466 int result; 467 468 /* remove one level of indirection */ 469 p1 = *(struct kinfo_proc2 **) pp1; 470 p2 = *(struct kinfo_proc2 **) pp2; 471 472 ORDERKEY_PCTCPU 473 ORDERKEY_CPUTIME 474 ORDERKEY_STATE 475 ORDERKEY_PRIO 476 ORDERKEY_RSSIZE 477 ORDERKEY_MEM 478 ; 479 return (result); 480} 481 482/* compare_size - the comparison function for sorting by total memory usage */ 483static int 484compare_size(const void *v1, const void *v2) 485{ 486 struct proc **pp1 = (struct proc **) v1; 487 struct proc **pp2 = (struct proc **) v2; 488 struct kinfo_proc2 *p1, *p2; 489 pctcpu lresult; 490 int result; 491 492 /* remove one level of indirection */ 493 p1 = *(struct kinfo_proc2 **) pp1; 494 p2 = *(struct kinfo_proc2 **) pp2; 495 496 ORDERKEY_MEM 497 ORDERKEY_RSSIZE 498 ORDERKEY_PCTCPU 499 ORDERKEY_CPUTIME 500 ORDERKEY_STATE 501 ORDERKEY_PRIO 502 ; 503 return (result); 504} 505 506/* compare_res - the comparison function for sorting by resident set size */ 507static int 508compare_res(const void *v1, const void *v2) 509{ 510 struct proc **pp1 = (struct proc **) v1; 511 struct proc **pp2 = (struct proc **) v2; 512 struct kinfo_proc2 *p1, *p2; 513 pctcpu lresult; 514 int result; 515 516 /* remove one level of indirection */ 517 p1 = *(struct kinfo_proc2 **) pp1; 518 p2 = *(struct kinfo_proc2 **) pp2; 519 520 ORDERKEY_RSSIZE 521 ORDERKEY_MEM 522 ORDERKEY_PCTCPU 523 ORDERKEY_CPUTIME 524 ORDERKEY_STATE 525 ORDERKEY_PRIO 526 ; 527 return (result); 528} 529 530/* compare_time - the comparison function for sorting by CPU time */ 531static int 532compare_time(const void *v1, const void *v2) 533{ 534 struct proc **pp1 = (struct proc **) v1; 535 struct proc **pp2 = (struct proc **) v2; 536 struct kinfo_proc2 *p1, *p2; 537 pctcpu lresult; 538 int result; 539 540 /* remove one level of indirection */ 541 p1 = *(struct kinfo_proc2 **) pp1; 542 p2 = *(struct kinfo_proc2 **) pp2; 543 544 ORDERKEY_CPUTIME 545 ORDERKEY_PCTCPU 546 ORDERKEY_STATE 547 ORDERKEY_PRIO 548 ORDERKEY_MEM 549 ORDERKEY_RSSIZE 550 ; 551 return (result); 552} 553 554/* compare_prio - the comparison function for sorting by CPU time */ 555static int 556compare_prio(const void *v1, const void *v2) 557{ 558 struct proc **pp1 = (struct proc **) v1; 559 struct proc **pp2 = (struct proc **) v2; 560 struct kinfo_proc2 *p1, *p2; 561 pctcpu lresult; 562 int result; 563 564 /* remove one level of indirection */ 565 p1 = *(struct kinfo_proc2 **) pp1; 566 p2 = *(struct kinfo_proc2 **) pp2; 567 568 ORDERKEY_PRIO 569 ORDERKEY_PCTCPU 570 ORDERKEY_CPUTIME 571 ORDERKEY_STATE 572 ORDERKEY_RSSIZE 573 ORDERKEY_MEM 574 ; 575 return (result); 576} 577 578int (*proc_compares[])(const void *, const void *) = { 579 compare_cpu, 580 compare_size, 581 compare_res, 582 compare_time, 583 compare_prio, 584 NULL 585}; 586 587/* 588 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if 589 * the process does not exist. 590 * It is EXTREMLY IMPORTANT that this function work correctly. 591 * If top runs setuid root (as in SVR4), then this function 592 * is the only thing that stands in the way of a serious 593 * security problem. It validates requests for the "kill" 594 * and "renice" commands. 595 */ 596uid_t 597proc_owner(pid_t pid) 598{ 599 struct kinfo_proc2 **prefp, *pp; 600 int cnt; 601 602 prefp = pref; 603 cnt = pref_len; 604 while (--cnt >= 0) { 605 pp = *prefp++; 606 if (pp->p_pid == pid) 607 return ((uid_t)pp->p_ruid); 608 } 609 return (uid_t)(-1); 610} 611 612/* 613 * swapmode is rewritten by Tobias Weingartner <weingart@openbsd.org> 614 * to be based on the new swapctl(2) system call. 615 */ 616static int 617swapmode(int *used, int *total) 618{ 619 struct swapent *swdev; 620 int nswap, rnswap, i; 621 622 nswap = swapctl(SWAP_NSWAP, 0, 0); 623 if (nswap == 0) 624 return 0; 625 626 swdev = malloc(nswap * sizeof(*swdev)); 627 if (swdev == NULL) 628 return 0; 629 630 rnswap = swapctl(SWAP_STATS, swdev, nswap); 631 if (rnswap == -1) 632 return 0; 633 634 /* if rnswap != nswap, then what? */ 635 636 /* Total things up */ 637 *total = *used = 0; 638 for (i = 0; i < nswap; i++) { 639 if (swdev[i].se_flags & SWF_ENABLE) { 640 *used += (swdev[i].se_inuse / (1024 / DEV_BSIZE)); 641 *total += (swdev[i].se_nblks / (1024 / DEV_BSIZE)); 642 } 643 } 644 free(swdev); 645 return 1; 646} 647