machine.c revision 1.41
1/* $OpenBSD: machine.c,v 1.41 2004/06/11 05:29:28 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 %-8s %-6.6s %6s %5.2f%% %.11s" 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", "sleep", "stop", "zomb", "dead", "onproc" 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 143int ncpu; 144 145unsigned int maxslp; 146 147static int 148getstathz(void) 149{ 150 struct clockinfo cinf; 151 size_t size = sizeof(cinf); 152 int mib[2]; 153 154 mib[0] = CTL_KERN; 155 mib[1] = KERN_CLOCKRATE; 156 if (sysctl(mib, 2, &cinf, &size, NULL, 0) == -1) 157 return (-1); 158 return (cinf.stathz); 159} 160 161int 162machine_init(struct statics *statics) 163{ 164 size_t size = sizeof(ncpu); 165 int mib[2], pagesize; 166 167 mib[0] = CTL_HW; 168 mib[1] = HW_NCPU; 169 if (sysctl(mib, 2, &ncpu, &size, NULL, 0) == -1) 170 return (-1); 171 172 stathz = getstathz(); 173 if (stathz == -1) 174 return (-1); 175 176 pbase = NULL; 177 pref = NULL; 178 onproc = -1; 179 nproc = 0; 180 181 /* 182 * get the page size with "getpagesize" and calculate pageshift from 183 * it 184 */ 185 pagesize = getpagesize(); 186 pageshift = 0; 187 while (pagesize > 1) { 188 pageshift++; 189 pagesize >>= 1; 190 } 191 192 /* we only need the amount of log(2)1024 for our conversion */ 193 pageshift -= LOG1024; 194 195 /* fill in the statics information */ 196 statics->procstate_names = procstatenames; 197 statics->cpustate_names = cpustatenames; 198 statics->memory_names = memorynames; 199 statics->order_names = ordernames; 200 return (0); 201} 202 203char * 204format_header(char *uname_field) 205{ 206 char *ptr; 207 208 ptr = header + UNAME_START; 209 while (*uname_field != '\0') 210 *ptr++ = *uname_field++; 211 return (header); 212} 213 214void 215get_system_info(struct system_info *si) 216{ 217 static int sysload_mib[] = {CTL_VM, VM_LOADAVG}; 218 static int vmtotal_mib[] = {CTL_VM, VM_METER}; 219 static int cp_time_mib[] = {CTL_KERN, KERN_CPTIME}; 220 struct loadavg sysload; 221 struct vmtotal vmtotal; 222 double *infoloadp; 223 size_t size; 224 int i; 225 226 size = sizeof(cp_time); 227 if (sysctl(cp_time_mib, 2, &cp_time, &size, NULL, 0) < 0) 228 warn("sysctl kern.cp_time failed"); 229 230 size = sizeof(sysload); 231 if (sysctl(sysload_mib, 2, &sysload, &size, NULL, 0) < 0) 232 warn("sysctl failed"); 233 infoloadp = si->load_avg; 234 for (i = 0; i < 3; i++) 235 *infoloadp++ = ((double) sysload.ldavg[i]) / sysload.fscale; 236 237 /* convert cp_time counts to percentages */ 238 (void) percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff); 239 240 /* get total -- systemwide main memory usage structure */ 241 size = sizeof(vmtotal); 242 if (sysctl(vmtotal_mib, 2, &vmtotal, &size, NULL, 0) < 0) { 243 warn("sysctl failed"); 244 bzero(&vmtotal, sizeof(vmtotal)); 245 } 246 /* convert memory stats to Kbytes */ 247 memory_stats[0] = -1; 248 memory_stats[1] = pagetok(vmtotal.t_arm); 249 memory_stats[2] = pagetok(vmtotal.t_rm); 250 memory_stats[3] = -1; 251 memory_stats[4] = pagetok(vmtotal.t_free); 252 memory_stats[5] = -1; 253 254 if (!swapmode(&memory_stats[6], &memory_stats[7])) { 255 memory_stats[6] = 0; 256 memory_stats[7] = 0; 257 } 258 259 /* set arrays and strings */ 260 si->cpustates = cpu_states; 261 si->memory = memory_stats; 262 si->last_pid = -1; 263} 264 265static struct handle handle; 266 267static struct kinfo_proc2 * 268getprocs(int op, int arg, int *cnt) 269{ 270 size_t size; 271 int mib[6] = {CTL_KERN, KERN_PROC2, 0, 0, sizeof(struct kinfo_proc2), 0}; 272 static int maxslp_mib[] = {CTL_VM, VM_MAXSLP}; 273 static struct kinfo_proc2 *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 retry: 285 free(procbase); 286 st = sysctl(mib, 6, NULL, &size, NULL, 0); 287 if (st == -1) { 288 /* _kvm_syserr(kd, kd->program, "kvm_getproc2"); */ 289 return (0); 290 } 291 size = 5 * size / 4; /* extra slop */ 292 if ((procbase = malloc(size)) == NULL) 293 return (0); 294 mib[5] = (int)(size / sizeof(struct kinfo_proc2)); 295 st = sysctl(mib, 6, procbase, &size, NULL, 0); 296 if (st == -1) { 297 if (errno == ENOMEM) 298 goto retry; 299 /* _kvm_syserr(kd, kd->program, "kvm_getproc2"); */ 300 return (0); 301 } 302 *cnt = (int)(size / sizeof(struct kinfo_proc2)); 303 return (procbase); 304} 305 306caddr_t 307get_process_info(struct system_info *si, struct process_select *sel, 308 int (*compare) (const void *, const void *)) 309{ 310 int show_idle, show_system, show_uid; 311 int total_procs, active_procs, i; 312 struct kinfo_proc2 **prefp, *pp; 313 314 if ((pbase = getprocs(KERN_PROC_KTHREAD, 0, &nproc)) == NULL) { 315 /* warnx("%s", kvm_geterr(kd)); */ 316 quit(23); 317 } 318 if (nproc > onproc) 319 pref = (struct kinfo_proc2 **)realloc(pref, 320 sizeof(struct kinfo_proc2 *) * (onproc = nproc)); 321 if (pref == NULL) { 322 warnx("Out of memory."); 323 quit(23); 324 } 325 /* get a pointer to the states summary array */ 326 si->procstates = process_states; 327 328 /* set up flags which define what we are going to select */ 329 show_idle = sel->idle; 330 show_system = sel->system; 331 show_uid = sel->uid != (uid_t)-1; 332 333 /* count up process states and get pointers to interesting procs */ 334 total_procs = 0; 335 active_procs = 0; 336 memset((char *) process_states, 0, sizeof(process_states)); 337 prefp = pref; 338 for (pp = pbase, i = 0; i < nproc; pp++, i++) { 339 /* 340 * Place pointers to each valid proc structure in pref[]. 341 * Process slots that are actually in use have a non-zero 342 * status field. Processes with SSYS set are system 343 * processes---these get ignored unless show_sysprocs is set. 344 */ 345 if (pp->p_stat != 0 && 346 (show_system || (pp->p_flag & P_SYSTEM) == 0)) { 347 total_procs++; 348 process_states[(unsigned char) pp->p_stat]++; 349 if (pp->p_stat != SZOMB && 350 (show_idle || pp->p_pctcpu != 0 || 351 pp->p_stat == SRUN) && 352 (!show_uid || pp->p_ruid == sel->uid)) { 353 *prefp++ = pp; 354 active_procs++; 355 } 356 } 357 } 358 359 /* if requested, sort the "interesting" processes */ 360 if (compare != NULL) 361 qsort((char *) pref, active_procs, 362 sizeof(struct kinfo_proc2 *), compare); 363 /* remember active and total counts */ 364 si->p_total = total_procs; 365 si->p_active = pref_len = active_procs; 366 367 /* pass back a handle */ 368 handle.next_proc = pref; 369 handle.remaining = active_procs; 370 return ((caddr_t) & handle); 371} 372 373char fmt[MAX_COLS]; /* static area where result is built */ 374 375char * 376state_abbr(struct kinfo_proc2 *pp) 377{ 378 static char buf[10]; 379 380 if (ncpu > 1) 381 snprintf(buf, sizeof buf, "%s/%d", 382 state_abbrev[(unsigned char)pp->p_stat], pp->p_cpuid); 383 else 384 snprintf(buf, sizeof buf, "%s", 385 state_abbrev[(unsigned char)pp->p_stat]); 386 return buf; 387} 388 389char * 390format_next_process(caddr_t handle, char *(*get_userid)(uid_t)) 391{ 392 char *p_wait, waddr[sizeof(void *) * 2 + 3]; /* Hexify void pointer */ 393 struct kinfo_proc2 *pp; 394 struct handle *hp; 395 int cputime; 396 double pct; 397 398 /* find and remember the next proc structure */ 399 hp = (struct handle *) handle; 400 pp = *(hp->next_proc++); 401 hp->remaining--; 402 403 if ((pp->p_flag & P_INMEM) == 0) { 404 /* 405 * Print swapped processes as <pname> 406 */ 407 char buf[sizeof(pp->p_comm)]; 408 409 (void) strlcpy(buf, pp->p_comm, sizeof(buf)); 410 (void) snprintf(pp->p_comm, sizeof(pp->p_comm), "<%s>", buf); 411 } 412 cputime = (pp->p_uticks + pp->p_sticks + pp->p_iticks) / stathz; 413 414 /* calculate the base for cpu percentages */ 415 pct = pctdouble(pp->p_pctcpu); 416 417 if (pp->p_wchan) { 418 if (pp->p_wmesg) 419 p_wait = pp->p_wmesg; 420 else { 421 snprintf(waddr, sizeof(waddr), "%llx", 422 pp->p_wchan & ~KERNBASE); 423 p_wait = waddr; 424 } 425 } else 426 p_wait = "-"; 427 428 /* format this entry */ 429 snprintf(fmt, sizeof fmt, Proc_format, 430 pp->p_pid, (*get_userid)(pp->p_ruid), 431 pp->p_priority - PZERO, pp->p_nice - NZERO, 432 format_k(pagetok(PROCSIZE(pp))), 433 format_k(pagetok(pp->p_vm_rssize)), 434 (pp->p_stat == SSLEEP && pp->p_slptime > maxslp) ? 435 "idle" : state_abbr(pp), 436 p_wait, format_time(cputime), 100.0 * pct, 437 printable(pp->p_comm)); 438 439 /* return the result */ 440 return (fmt); 441} 442 443/* comparison routine for qsort */ 444static unsigned char sorted_state[] = 445{ 446 0, /* not used */ 447 4, /* start */ 448 5, /* run */ 449 2, /* sleep */ 450 3, /* stop */ 451 1 /* zombie */ 452}; 453 454/* 455 * proc_compares - comparison functions for "qsort" 456 */ 457 458/* 459 * First, the possible comparison keys. These are defined in such a way 460 * that they can be merely listed in the source code to define the actual 461 * desired ordering. 462 */ 463 464#define ORDERKEY_PCTCPU \ 465 if (lresult = (pctcpu)p2->p_pctcpu - (pctcpu)p1->p_pctcpu, \ 466 (result = lresult > 0 ? 1 : lresult < 0 ? -1 : 0) == 0) 467#define ORDERKEY_CPUTIME \ 468 if ((result = p2->p_rtime_sec - p1->p_rtime_sec) == 0) \ 469 if ((result = p2->p_rtime_usec - p1->p_rtime_usec) == 0) 470#define ORDERKEY_STATE \ 471 if ((result = sorted_state[(unsigned char)p2->p_stat] - \ 472 sorted_state[(unsigned char)p1->p_stat]) == 0) 473#define ORDERKEY_PRIO \ 474 if ((result = p2->p_priority - p1->p_priority) == 0) 475#define ORDERKEY_RSSIZE \ 476 if ((result = p2->p_vm_rssize - p1->p_vm_rssize) == 0) 477#define ORDERKEY_MEM \ 478 if ((result = PROCSIZE(p2) - PROCSIZE(p1)) == 0) 479 480/* compare_cpu - the comparison function for sorting by cpu percentage */ 481static int 482compare_cpu(const void *v1, const void *v2) 483{ 484 struct proc **pp1 = (struct proc **) v1; 485 struct proc **pp2 = (struct proc **) v2; 486 struct kinfo_proc2 *p1, *p2; 487 pctcpu lresult; 488 int result; 489 490 /* remove one level of indirection */ 491 p1 = *(struct kinfo_proc2 **) pp1; 492 p2 = *(struct kinfo_proc2 **) pp2; 493 494 ORDERKEY_PCTCPU 495 ORDERKEY_CPUTIME 496 ORDERKEY_STATE 497 ORDERKEY_PRIO 498 ORDERKEY_RSSIZE 499 ORDERKEY_MEM 500 ; 501 return (result); 502} 503 504/* compare_size - the comparison function for sorting by total memory usage */ 505static int 506compare_size(const void *v1, const void *v2) 507{ 508 struct proc **pp1 = (struct proc **) v1; 509 struct proc **pp2 = (struct proc **) v2; 510 struct kinfo_proc2 *p1, *p2; 511 pctcpu lresult; 512 int result; 513 514 /* remove one level of indirection */ 515 p1 = *(struct kinfo_proc2 **) pp1; 516 p2 = *(struct kinfo_proc2 **) pp2; 517 518 ORDERKEY_MEM 519 ORDERKEY_RSSIZE 520 ORDERKEY_PCTCPU 521 ORDERKEY_CPUTIME 522 ORDERKEY_STATE 523 ORDERKEY_PRIO 524 ; 525 return (result); 526} 527 528/* compare_res - the comparison function for sorting by resident set size */ 529static int 530compare_res(const void *v1, const void *v2) 531{ 532 struct proc **pp1 = (struct proc **) v1; 533 struct proc **pp2 = (struct proc **) v2; 534 struct kinfo_proc2 *p1, *p2; 535 pctcpu lresult; 536 int result; 537 538 /* remove one level of indirection */ 539 p1 = *(struct kinfo_proc2 **) pp1; 540 p2 = *(struct kinfo_proc2 **) pp2; 541 542 ORDERKEY_RSSIZE 543 ORDERKEY_MEM 544 ORDERKEY_PCTCPU 545 ORDERKEY_CPUTIME 546 ORDERKEY_STATE 547 ORDERKEY_PRIO 548 ; 549 return (result); 550} 551 552/* compare_time - the comparison function for sorting by CPU time */ 553static int 554compare_time(const void *v1, const void *v2) 555{ 556 struct proc **pp1 = (struct proc **) v1; 557 struct proc **pp2 = (struct proc **) v2; 558 struct kinfo_proc2 *p1, *p2; 559 pctcpu lresult; 560 int result; 561 562 /* remove one level of indirection */ 563 p1 = *(struct kinfo_proc2 **) pp1; 564 p2 = *(struct kinfo_proc2 **) pp2; 565 566 ORDERKEY_CPUTIME 567 ORDERKEY_PCTCPU 568 ORDERKEY_STATE 569 ORDERKEY_PRIO 570 ORDERKEY_MEM 571 ORDERKEY_RSSIZE 572 ; 573 return (result); 574} 575 576/* compare_prio - the comparison function for sorting by CPU time */ 577static int 578compare_prio(const void *v1, const void *v2) 579{ 580 struct proc **pp1 = (struct proc **) v1; 581 struct proc **pp2 = (struct proc **) v2; 582 struct kinfo_proc2 *p1, *p2; 583 pctcpu lresult; 584 int result; 585 586 /* remove one level of indirection */ 587 p1 = *(struct kinfo_proc2 **) pp1; 588 p2 = *(struct kinfo_proc2 **) pp2; 589 590 ORDERKEY_PRIO 591 ORDERKEY_PCTCPU 592 ORDERKEY_CPUTIME 593 ORDERKEY_STATE 594 ORDERKEY_RSSIZE 595 ORDERKEY_MEM 596 ; 597 return (result); 598} 599 600int (*proc_compares[])(const void *, const void *) = { 601 compare_cpu, 602 compare_size, 603 compare_res, 604 compare_time, 605 compare_prio, 606 NULL 607}; 608 609/* 610 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if 611 * the process does not exist. 612 * It is EXTREMLY IMPORTANT that this function work correctly. 613 * If top runs setuid root (as in SVR4), then this function 614 * is the only thing that stands in the way of a serious 615 * security problem. It validates requests for the "kill" 616 * and "renice" commands. 617 */ 618uid_t 619proc_owner(pid_t pid) 620{ 621 struct kinfo_proc2 **prefp, *pp; 622 int cnt; 623 624 prefp = pref; 625 cnt = pref_len; 626 while (--cnt >= 0) { 627 pp = *prefp++; 628 if (pp->p_pid == pid) 629 return ((uid_t)pp->p_ruid); 630 } 631 return (uid_t)(-1); 632} 633 634/* 635 * swapmode is rewritten by Tobias Weingartner <weingart@openbsd.org> 636 * to be based on the new swapctl(2) system call. 637 */ 638static int 639swapmode(int *used, int *total) 640{ 641 struct swapent *swdev; 642 int nswap, rnswap, i; 643 644 nswap = swapctl(SWAP_NSWAP, 0, 0); 645 if (nswap == 0) 646 return 0; 647 648 swdev = malloc(nswap * sizeof(*swdev)); 649 if (swdev == NULL) 650 return 0; 651 652 rnswap = swapctl(SWAP_STATS, swdev, nswap); 653 if (rnswap == -1) 654 return 0; 655 656 /* if rnswap != nswap, then what? */ 657 658 /* Total things up */ 659 *total = *used = 0; 660 for (i = 0; i < nswap; i++) { 661 if (swdev[i].se_flags & SWF_ENABLE) { 662 *used += (swdev[i].se_inuse / (1024 / DEV_BSIZE)); 663 *total += (swdev[i].se_nblks / (1024 / DEV_BSIZE)); 664 } 665 } 666 free(swdev); 667 return 1; 668} 669