xref: /freebsd/usr.bin/top/machine.c (revision b3a1f9373a31b644f8a65de1ba35929af3f6a9fe)
1 /*
2  * top - a top users display for Unix
3  *
4  * SYNOPSIS:  For FreeBSD-2.x and later
5  *
6  * DESCRIPTION:
7  * Originally written for BSD4.4 system by Christos Zoulas.
8  * Ported to FreeBSD 2.x by Steven Wallace && Wolfram Schneider
9  * Order support hacked in from top-3.5beta6/machine/m_aix41.c
10  *   by Monte Mitzelfelt (for latest top see http://www.groupsys.com/topinfo/)
11  *
12  * This is the machine-dependent module for FreeBSD 2.2
13  * Works for:
14  *	FreeBSD 2.2.x, 3.x, 4.x, and probably FreeBSD 2.1.x
15  *
16  * LIBS: -lkvm
17  *
18  * AUTHOR:  Christos Zoulas <christos@ee.cornell.edu>
19  *          Steven Wallace  <swallace@freebsd.org>
20  *          Wolfram Schneider <wosch@FreeBSD.org>
21  *          Thomas Moestl <tmoestl@gmx.net>
22  *
23  * $FreeBSD$
24  */
25 
26 #include <sys/param.h>
27 #include <sys/errno.h>
28 #include <sys/file.h>
29 #include <sys/proc.h>
30 #include <sys/resource.h>
31 #include <sys/rtprio.h>
32 #include <sys/signal.h>
33 #include <sys/sysctl.h>
34 #include <sys/time.h>
35 #include <sys/user.h>
36 #include <sys/vmmeter.h>
37 
38 #include <kvm.h>
39 #include <math.h>
40 #include <nlist.h>
41 #include <paths.h>
42 #include <pwd.h>
43 #include <stdio.h>
44 #include <stdlib.h>
45 #include <string.h>
46 #include <strings.h>
47 #include <unistd.h>
48 
49 #include "top.h"
50 #include "machine.h"
51 #include "screen.h"
52 #include "utils.h"
53 
54 #define GETSYSCTL(name, var) getsysctl(name, &(var), sizeof(var))
55 #define	SMPUNAMELEN	13
56 #define	UPUNAMELEN	15
57 
58 extern struct process_select ps;
59 extern char* printable(char *);
60 static int smpmode;
61 enum displaymodes displaymode;
62 static int namelength = 8;
63 static int cmdlengthdelta;
64 
65 /* Prototypes for top internals */
66 void quit(int);
67 
68 /* get_process_info passes back a handle.  This is what it looks like: */
69 
70 struct handle {
71 	struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
72 	int remaining;			/* number of pointers remaining */
73 };
74 
75 /* declarations for load_avg */
76 #include "loadavg.h"
77 
78 /* define what weighted cpu is.  */
79 #define weighted_cpu(pct, pp) ((pp)->ki_swtime == 0 ? 0.0 : \
80 			 ((pct) / (1.0 - exp((pp)->ki_swtime * logcpu))))
81 
82 /* what we consider to be process size: */
83 #define PROCSIZE(pp) ((pp)->ki_size / 1024)
84 
85 #define RU(pp)	(&(pp)->ki_rusage)
86 #define RUTOT(pp) \
87 	(RU(pp)->ru_inblock + RU(pp)->ru_oublock + RU(pp)->ru_majflt)
88 
89 
90 /* definitions for indices in the nlist array */
91 
92 /*
93  *  These definitions control the format of the per-process area
94  */
95 
96 static char io_header[] =
97     "  PID %-*.*s   VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
98 
99 #define io_Proc_format \
100     "%5d %-*.*s %6ld %6ld %6ld %6ld %6ld %6ld %6.2f%% %.*s"
101 
102 static char smp_header_thr[] =
103     "  PID %-*.*s  THR PRI NICE   SIZE    RES STATE  C   TIME %6s COMMAND";
104 static char smp_header[] =
105     "  PID %-*.*s "   "PRI NICE   SIZE    RES STATE  C   TIME %6s COMMAND";
106 
107 #define smp_Proc_format \
108     "%5d %-*.*s %s%3d %4s%7s %6s %-6.6s %1x%7s %5.2f%% %.*s"
109 
110 static char up_header_thr[] =
111     "  PID %-*.*s  THR PRI NICE   SIZE    RES STATE    TIME %6s COMMAND";
112 static char up_header[] =
113     "  PID %-*.*s "   "PRI NICE   SIZE    RES STATE    TIME %6s COMMAND";
114 
115 #define up_Proc_format \
116     "%5d %-*.*s %s%3d %4s%7s %6s %-6.6s%.0d%7s %5.2f%% %.*s"
117 
118 
119 /* process state names for the "STATE" column of the display */
120 /* the extra nulls in the string "run" are for adding a slash and
121    the processor number when needed */
122 
123 char *state_abbrev[] = {
124 	"", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB", "WAIT", "LOCK"
125 };
126 
127 
128 static kvm_t *kd;
129 
130 /* values that we stash away in _init and use in later routines */
131 
132 static double logcpu;
133 
134 /* these are retrieved from the kernel in _init */
135 
136 static load_avg  ccpu;
137 
138 /* these are used in the get_ functions */
139 
140 static int lastpid;
141 
142 /* these are for calculating cpu state percentages */
143 
144 static long cp_time[CPUSTATES];
145 static long cp_old[CPUSTATES];
146 static long cp_diff[CPUSTATES];
147 
148 /* these are for detailing the process states */
149 
150 int process_states[8];
151 char *procstatenames[] = {
152 	"", " starting, ", " running, ", " sleeping, ", " stopped, ",
153 	" zombie, ", " waiting, ", " lock, ",
154 	NULL
155 };
156 
157 /* these are for detailing the cpu states */
158 
159 int cpu_states[CPUSTATES];
160 char *cpustatenames[] = {
161 	"user", "nice", "system", "interrupt", "idle", NULL
162 };
163 
164 /* these are for detailing the memory statistics */
165 
166 int memory_stats[7];
167 char *memorynames[] = {
168 	"K Active, ", "K Inact, ", "K Wired, ", "K Cache, ", "K Buf, ",
169 	"K Free", NULL
170 };
171 
172 int swap_stats[7];
173 char *swapnames[] = {
174 	"K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
175 	NULL
176 };
177 
178 
179 /* these are for keeping track of the proc array */
180 
181 static int nproc;
182 static int onproc = -1;
183 static int pref_len;
184 static struct kinfo_proc *pbase;
185 static struct kinfo_proc **pref;
186 static struct kinfo_proc *previous_procs;
187 static struct kinfo_proc **previous_pref;
188 static int previous_proc_count = 0;
189 static int previous_proc_count_max = 0;
190 
191 /* total number of io operations */
192 static long total_inblock;
193 static long total_oublock;
194 static long total_majflt;
195 
196 /* these are for getting the memory statistics */
197 
198 static int pageshift;		/* log base 2 of the pagesize */
199 
200 /* define pagetok in terms of pageshift */
201 
202 #define pagetok(size) ((size) << pageshift)
203 
204 /* useful externals */
205 long percentages();
206 
207 #ifdef ORDER
208 /*
209  * Sorting orders.  The first element is the default.
210  */
211 char *ordernames[] = {
212 	"cpu", "size", "res", "time", "pri", "threads",
213 	"total", "read", "write", "fault", "vcsw", "ivcsw", NULL
214 };
215 #endif
216 
217 static int compare_pid(const void *a, const void *b);
218 static const char *format_nice(const struct kinfo_proc *pp);
219 static void getsysctl(const char *name, void *ptr, size_t len);
220 static int swapmode(int *retavail, int *retfree);
221 
222 int
223 machine_init(struct statics *statics)
224 {
225 	int pagesize;
226 	size_t modelen;
227 	struct passwd *pw;
228 
229 	modelen = sizeof(smpmode);
230 	if ((sysctlbyname("machdep.smp_active", &smpmode, &modelen,
231 	    NULL, 0) != 0 &&
232 	    sysctlbyname("kern.smp.active", &smpmode, &modelen,
233 	    NULL, 0) != 0) ||
234 	    modelen != sizeof(smpmode))
235 		smpmode = 0;
236 
237 	while ((pw = getpwent()) != NULL) {
238 		if (strlen(pw->pw_name) > namelength)
239 			namelength = strlen(pw->pw_name);
240 	}
241 	if (smpmode && namelength > SMPUNAMELEN)
242 		namelength = SMPUNAMELEN;
243 	else if (namelength > UPUNAMELEN)
244 		namelength = UPUNAMELEN;
245 
246 	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
247 	if (kd == NULL)
248 		return (-1);
249 
250 	GETSYSCTL("kern.ccpu", ccpu);
251 
252 	/* this is used in calculating WCPU -- calculate it ahead of time */
253 	logcpu = log(loaddouble(ccpu));
254 
255 	pbase = NULL;
256 	pref = NULL;
257 	nproc = 0;
258 	onproc = -1;
259 
260 	/* get the page size and calculate pageshift from it */
261 	pagesize = getpagesize();
262 	pageshift = 0;
263 	while (pagesize > 1) {
264 		pageshift++;
265 		pagesize >>= 1;
266 	}
267 
268 	/* we only need the amount of log(2)1024 for our conversion */
269 	pageshift -= LOG1024;
270 
271 	/* fill in the statics information */
272 	statics->procstate_names = procstatenames;
273 	statics->cpustate_names = cpustatenames;
274 	statics->memory_names = memorynames;
275 	statics->swap_names = swapnames;
276 #ifdef ORDER
277 	statics->order_names = ordernames;
278 #endif
279 
280 	/* all done! */
281 	return (0);
282 }
283 
284 char *
285 format_header(char *uname_field)
286 {
287 	static char Header[128];
288 	const char *prehead;
289 
290 	switch (displaymode) {
291 	case DISP_CPU:
292 		/*
293 		 * The logic of picking the right header format seems reverse
294 		 * here because we only want to display a THR column when
295 		 * "thread mode" is off (and threads are not listed as
296 		 * separate lines).
297 		 */
298 		prehead = smpmode ?
299 		    (ps.thread ? smp_header : smp_header_thr) :
300 		    (ps.thread ? up_header : up_header_thr);
301 		snprintf(Header, sizeof(Header), prehead,
302 		    namelength, namelength, uname_field,
303 		    ps.wcpu ? "WCPU" : "CPU");
304 		break;
305 	case DISP_IO:
306 		prehead = io_header;
307 		snprintf(Header, sizeof(Header), prehead,
308 		    namelength, namelength, uname_field);
309 		break;
310 	}
311 	cmdlengthdelta = strlen(Header) - 7;
312 	return (Header);
313 }
314 
315 static int swappgsin = -1;
316 static int swappgsout = -1;
317 extern struct timeval timeout;
318 
319 void
320 get_system_info(struct system_info *si)
321 {
322 	long total;
323 	struct loadavg sysload;
324 	int mib[2];
325 	struct timeval boottime;
326 	size_t bt_size;
327 	int i;
328 
329 	/* get the cp_time array */
330 	GETSYSCTL("kern.cp_time", cp_time);
331 	GETSYSCTL("vm.loadavg", sysload);
332 	GETSYSCTL("kern.lastpid", lastpid);
333 
334 	/* convert load averages to doubles */
335 	for (i = 0; i < 3; i++)
336 		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
337 
338 	/* convert cp_time counts to percentages */
339 	total = percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
340 
341 	/* sum memory & swap statistics */
342 	{
343 		static unsigned int swap_delay = 0;
344 		static int swapavail = 0;
345 		static int swapfree = 0;
346 		static int bufspace = 0;
347 		static int nspgsin, nspgsout;
348 
349 		GETSYSCTL("vfs.bufspace", bufspace);
350 		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
351 		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
352 		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[2]);
353 		GETSYSCTL("vm.stats.vm.v_cache_count", memory_stats[3]);
354 		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
355 		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
356 		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
357 		/* convert memory stats to Kbytes */
358 		memory_stats[0] = pagetok(memory_stats[0]);
359 		memory_stats[1] = pagetok(memory_stats[1]);
360 		memory_stats[2] = pagetok(memory_stats[2]);
361 		memory_stats[3] = pagetok(memory_stats[3]);
362 		memory_stats[4] = bufspace / 1024;
363 		memory_stats[5] = pagetok(memory_stats[5]);
364 		memory_stats[6] = -1;
365 
366 		/* first interval */
367 		if (swappgsin < 0) {
368 			swap_stats[4] = 0;
369 			swap_stats[5] = 0;
370 		}
371 
372 		/* compute differences between old and new swap statistic */
373 		else {
374 			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
375 			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
376 		}
377 
378 		swappgsin = nspgsin;
379 		swappgsout = nspgsout;
380 
381 		/* call CPU heavy swapmode() only for changes */
382 		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
383 			swap_stats[3] = swapmode(&swapavail, &swapfree);
384 			swap_stats[0] = swapavail;
385 			swap_stats[1] = swapavail - swapfree;
386 			swap_stats[2] = swapfree;
387 		}
388 		swap_delay = 1;
389 		swap_stats[6] = -1;
390 	}
391 
392 	/* set arrays and strings */
393 	si->cpustates = cpu_states;
394 	si->memory = memory_stats;
395 	si->swap = swap_stats;
396 
397 
398 	if (lastpid > 0) {
399 		si->last_pid = lastpid;
400 	} else {
401 		si->last_pid = -1;
402 	}
403 
404 	/*
405 	 * Print how long system has been up.
406 	 * (Found by looking getting "boottime" from the kernel)
407 	 */
408 	mib[0] = CTL_KERN;
409 	mib[1] = KERN_BOOTTIME;
410 	bt_size = sizeof(boottime);
411 	if (sysctl(mib, 2, &boottime, &bt_size, NULL, 0) != -1 &&
412 	    boottime.tv_sec != 0) {
413 		si->boottime = boottime;
414 	} else {
415 		si->boottime.tv_sec = -1;
416 	}
417 }
418 
419 #define NOPROC	((void *)-1)
420 
421 /*
422  * We need to compare data from the old process entry with the new
423  * process entry.
424  * To facilitate doing this quickly we stash a pointer in the kinfo_proc
425  * structure to cache the mapping.  We also use a negative cache pointer
426  * of NOPROC to avoid duplicate lookups.
427  * XXX: this could be done when the actual processes are fetched, we do
428  * it here out of laziness.
429  */
430 const struct kinfo_proc *
431 get_old_proc(struct kinfo_proc *pp)
432 {
433 	struct kinfo_proc **oldpp, *oldp;
434 
435 	/*
436 	 * If this is the first fetch of the kinfo_procs then we don't have
437 	 * any previous entries.
438 	 */
439 	if (previous_proc_count == 0)
440 		return (NULL);
441 	/* negative cache? */
442 	if (pp->ki_udata == NOPROC)
443 		return (NULL);
444 	/* cached? */
445 	if (pp->ki_udata != NULL)
446 		return (pp->ki_udata);
447 	/*
448 	 * Not cached,
449 	 * 1) look up based on pid.
450 	 * 2) compare process start.
451 	 * If we fail here, then setup a negative cache entry, otherwise
452 	 * cache it.
453 	 */
454 	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
455 	    sizeof(*previous_pref), compare_pid);
456 	if (oldpp == NULL) {
457 		pp->ki_udata = NOPROC;
458 		return (NULL);
459 	}
460 	oldp = *oldpp;
461 	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
462 		pp->ki_udata = NOPROC;
463 		return (NULL);
464 	}
465 	pp->ki_udata = oldp;
466 	return (oldp);
467 }
468 
469 /*
470  * Return the total amount of IO done in blocks in/out and faults.
471  * store the values individually in the pointers passed in.
472  */
473 long
474 get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp,
475     long *vcsw, long *ivcsw)
476 {
477 	const struct kinfo_proc *oldp;
478 	static struct kinfo_proc dummy;
479 	long ret;
480 
481 	oldp = get_old_proc(pp);
482 	if (oldp == NULL) {
483 		bzero(&dummy, sizeof(dummy));
484 		oldp = &dummy;
485 	}
486 	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
487 	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
488 	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
489 	*vcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
490 	*ivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
491 	ret =
492 	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
493 	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
494 	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
495 	return (ret);
496 }
497 
498 /*
499  * Return the total number of block in/out and faults by a process.
500  */
501 long
502 get_io_total(struct kinfo_proc *pp)
503 {
504 	long dummy;
505 
506 	return (get_io_stats(pp, &dummy, &dummy, &dummy, &dummy, &dummy));
507 }
508 
509 static struct handle handle;
510 
511 caddr_t
512 get_process_info(struct system_info *si, struct process_select *sel,
513     int (*compare)(const void *, const void *))
514 {
515 	int i;
516 	int total_procs;
517 	long p_io;
518 	long p_inblock, p_oublock, p_majflt, p_vcsw, p_ivcsw;
519 	int active_procs;
520 	struct kinfo_proc **prefp;
521 	struct kinfo_proc *pp;
522 	struct kinfo_proc *prev_pp = NULL;
523 
524 	/* these are copied out of sel for speed */
525 	int show_idle;
526 	int show_self;
527 	int show_system;
528 	int show_uid;
529 	int show_command;
530 
531 	/*
532 	 * Save the previous process info.
533 	 */
534 	if (previous_proc_count_max < nproc) {
535 		free(previous_procs);
536 		previous_procs = malloc(nproc * sizeof(*previous_procs));
537 		free(previous_pref);
538 		previous_pref = malloc(nproc * sizeof(*previous_pref));
539 		if (previous_procs == NULL || previous_pref == NULL) {
540 			(void) fprintf(stderr, "top: Out of memory.\n");
541 			quit(23);
542 		}
543 		previous_proc_count_max = nproc;
544 	}
545 	if (nproc) {
546 		for (i = 0; i < nproc; i++)
547 			previous_pref[i] = &previous_procs[i];
548 		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
549 		qsort(previous_pref, nproc, sizeof(*previous_pref),
550 		    compare_pid);
551 	}
552 	previous_proc_count = nproc;
553 
554 	pbase = kvm_getprocs(kd, KERN_PROC_ALL, 0, &nproc);
555 	if (nproc > onproc)
556 		pref = realloc(pref, sizeof(*pref) * (onproc = nproc));
557 	if (pref == NULL || pbase == NULL) {
558 		(void) fprintf(stderr, "top: Out of memory.\n");
559 		quit(23);
560 	}
561 	/* get a pointer to the states summary array */
562 	si->procstates = process_states;
563 
564 	/* set up flags which define what we are going to select */
565 	show_idle = sel->idle;
566 	show_self = sel->self == -1;
567 	show_system = sel->system;
568 	show_uid = sel->uid != -1;
569 	show_command = sel->command != NULL;
570 
571 	/* count up process states and get pointers to interesting procs */
572 	total_procs = 0;
573 	active_procs = 0;
574 	total_inblock = 0;
575 	total_oublock = 0;
576 	total_majflt = 0;
577 	memset((char *)process_states, 0, sizeof(process_states));
578 	prefp = pref;
579 	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
580 
581 		if (pp->ki_stat == 0)
582 			/* not in use */
583 			continue;
584 
585 		if (!show_self && pp->ki_pid == sel->self)
586 			/* skip self */
587 			continue;
588 
589 		if (!show_system && (pp->ki_flag & P_SYSTEM))
590 			/* skip system process */
591 			continue;
592 
593 		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt,
594 		    &p_vcsw, &p_ivcsw);
595 		total_inblock += p_inblock;
596 		total_oublock += p_oublock;
597 		total_majflt += p_majflt;
598 		total_procs++;
599 		process_states[pp->ki_stat]++;
600 
601 		if (pp->ki_stat == SZOMB)
602 			/* skip zombies */
603 			continue;
604 
605 		if (displaymode == DISP_CPU && !show_idle &&
606 		    (pp->ki_pctcpu == 0 || pp->ki_stat != SRUN))
607 			/* skip idle or non-running processes */
608 			continue;
609 
610 		if (displaymode == DISP_IO && !show_idle && p_io == 0)
611 			/* skip processes that aren't doing I/O */
612 			continue;
613 
614 		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
615 			/* skip proc. that don't belong to the selected UID */
616 			continue;
617 
618 		/*
619 		 * When not showing threads, take the first thread
620 		 * for output and add the fields that we can from
621 		 * the rest of the process's threads rather than
622 		 * using the system's mostly-broken KERN_PROC_PROC.
623 		 */
624 		if (sel->thread || prev_pp == NULL ||
625 		    prev_pp->ki_pid != pp->ki_pid) {
626 			*prefp++ = pp;
627 			active_procs++;
628 			prev_pp = pp;
629 		} else {
630 			prev_pp->ki_pctcpu += pp->ki_pctcpu;
631 		}
632 	}
633 
634 	/* if requested, sort the "interesting" processes */
635 	if (compare != NULL)
636 		qsort(pref, active_procs, sizeof(*pref), compare);
637 
638 	/* remember active and total counts */
639 	si->p_total = total_procs;
640 	si->p_active = pref_len = active_procs;
641 
642 	/* pass back a handle */
643 	handle.next_proc = pref;
644 	handle.remaining = active_procs;
645 	return ((caddr_t)&handle);
646 }
647 
648 static char fmt[128];	/* static area where result is built */
649 
650 char *
651 format_next_process(caddr_t handle, char *(*get_userid)(int))
652 {
653 	struct kinfo_proc *pp;
654 	const struct kinfo_proc *oldp;
655 	long cputime;
656 	double pct;
657 	struct handle *hp;
658 	char status[16];
659 	int state;
660 	struct rusage ru, *rup;
661 	long p_tot, s_tot;
662 	char *proc_fmt, thr_buf[6];
663 
664 	/* find and remember the next proc structure */
665 	hp = (struct handle *)handle;
666 	pp = *(hp->next_proc++);
667 	hp->remaining--;
668 
669 	/* get the process's command name */
670 	if ((pp->ki_sflag & PS_INMEM) == 0) {
671 		/*
672 		 * Print swapped processes as <pname>
673 		 */
674 		size_t len;
675 
676 		len = strlen(pp->ki_comm);
677 		if (len > sizeof(pp->ki_comm) - 3)
678 			len = sizeof(pp->ki_comm) - 3;
679 		memmove(pp->ki_comm + 1, pp->ki_comm, len);
680 		pp->ki_comm[0] = '<';
681 		pp->ki_comm[len + 1] = '>';
682 		pp->ki_comm[len + 2] = '\0';
683 	}
684 
685 	/*
686 	 * Convert the process's runtime from microseconds to seconds.  This
687 	 * time includes the interrupt time although that is not wanted here.
688 	 * ps(1) is similarly sloppy.
689 	 */
690 	cputime = (pp->ki_runtime + 500000) / 1000000;
691 
692 	/* calculate the base for cpu percentages */
693 	pct = pctdouble(pp->ki_pctcpu);
694 
695 	/* generate "STATE" field */
696 	switch (state = pp->ki_stat) {
697 	case SRUN:
698 		if (smpmode && pp->ki_oncpu != 0xff)
699 			sprintf(status, "CPU%d", pp->ki_oncpu);
700 		else
701 			strcpy(status, "RUN");
702 		break;
703 	case SLOCK:
704 		if (pp->ki_kiflag & KI_LOCKBLOCK) {
705 			sprintf(status, "*%.6s", pp->ki_lockname);
706 			break;
707 		}
708 		/* fall through */
709 	case SSLEEP:
710 		if (pp->ki_wmesg != NULL) {
711 			sprintf(status, "%.6s", pp->ki_wmesg);
712 			break;
713 		}
714 		/* FALLTHROUGH */
715 	default:
716 
717 		if (state >= 0 &&
718 		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
719 			sprintf(status, "%.6s", state_abbrev[state]);
720 		else
721 			sprintf(status, "?%5d", state);
722 		break;
723 	}
724 
725 	if (displaymode == DISP_IO) {
726 		oldp = get_old_proc(pp);
727 		if (oldp != NULL) {
728 			ru.ru_inblock = RU(pp)->ru_inblock -
729 			    RU(oldp)->ru_inblock;
730 			ru.ru_oublock = RU(pp)->ru_oublock -
731 			    RU(oldp)->ru_oublock;
732 			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
733 			ru.ru_nvcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
734 			ru.ru_nivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
735 			rup = &ru;
736 		} else {
737 			rup = RU(pp);
738 		}
739 		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
740 		s_tot = total_inblock + total_oublock + total_majflt;
741 
742 		sprintf(fmt, io_Proc_format,
743 		    pp->ki_pid,
744 		    namelength, namelength, (*get_userid)(pp->ki_ruid),
745 		    rup->ru_nvcsw,
746 		    rup->ru_nivcsw,
747 		    rup->ru_inblock,
748 		    rup->ru_oublock,
749 		    rup->ru_majflt,
750 		    p_tot,
751 		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
752 		    screen_width > cmdlengthdelta ?
753 		    screen_width - cmdlengthdelta : 0,
754 		    printable(pp->ki_comm));
755 		return (fmt);
756 	}
757 
758 	/* format this entry */
759 	proc_fmt = smpmode ? smp_Proc_format : up_Proc_format;
760 	if (ps.thread != 0)
761 		thr_buf[0] = '\0';
762 	else
763 		snprintf(thr_buf, sizeof(thr_buf), "%*d ",
764 		    sizeof(thr_buf) - 2, pp->ki_numthreads);
765 
766 	sprintf(fmt, proc_fmt,
767 	    pp->ki_pid,
768 	    namelength, namelength, (*get_userid)(pp->ki_ruid),
769 	    thr_buf,
770 	    pp->ki_pri.pri_level - PZERO,
771 	    format_nice(pp),
772 	    format_k2(PROCSIZE(pp)),
773 	    format_k2(pagetok(pp->ki_rssize)),
774 	    status,
775 	    smpmode ? pp->ki_lastcpu : 0,
776 	    format_time(cputime),
777 	    ps.wcpu ? 100.0 * weighted_cpu(pct, pp) : 100.0 * pct,
778 	    screen_width > cmdlengthdelta ? screen_width - cmdlengthdelta : 0,
779 	    printable(pp->ki_comm));
780 
781 	/* return the result */
782 	return (fmt);
783 }
784 
785 static void
786 getsysctl(const char *name, void *ptr, size_t len)
787 {
788 	size_t nlen = len;
789 
790 	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
791 		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
792 		    strerror(errno));
793 		quit(23);
794 	}
795 	if (nlen != len) {
796 		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n",
797 		    name, (unsigned long)len, (unsigned long)nlen);
798 		quit(23);
799 	}
800 }
801 
802 static
803 const char *format_nice(const struct kinfo_proc *pp)
804 {
805 	static char nicebuf[5];
806 
807 	snprintf(nicebuf, sizeof(nicebuf), "%4d",
808 	    /*
809 	     * normal time      -> nice value -20 - +20
810 	     * real time 0 - 31 -> nice value -52 - -21
811 	     * idle time 0 - 31 -> nice value +21 - +52
812 	     */
813 	    (pp->ki_pri.pri_class ==  PRI_TIMESHARE ?
814 		pp->ki_nice - NZERO :
815 		(PRI_IS_REALTIME(pp->ki_pri.pri_class) ?
816 		    (PRIO_MIN - 1 - (PRI_MAX_REALTIME - pp->ki_pri.pri_level)) :
817 		    (PRIO_MAX + 1 + pp->ki_pri.pri_level - PRI_MIN_IDLE))));
818 	return (nicebuf);
819 }
820 
821 /* comparison routines for qsort */
822 
823 static int
824 compare_pid(const void *p1, const void *p2)
825 {
826 	const struct kinfo_proc * const *pp1 = p1;
827 	const struct kinfo_proc * const *pp2 = p2;
828 
829 	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
830 		abort();
831 
832 	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
833 }
834 
835 /*
836  *  proc_compare - comparison function for "qsort"
837  *	Compares the resource consumption of two processes using five
838  *	distinct keys.  The keys (in descending order of importance) are:
839  *	percent cpu, cpu ticks, state, resident set size, total virtual
840  *	memory usage.  The process states are ordered as follows (from least
841  *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
842  *	array declaration below maps a process state index into a number
843  *	that reflects this ordering.
844  */
845 
846 static int sorted_state[] = {
847 	0,	/* not used		*/
848 	3,	/* sleep		*/
849 	1,	/* ABANDONED (WAIT)	*/
850 	6,	/* run			*/
851 	5,	/* start		*/
852 	2,	/* zombie		*/
853 	4	/* stop			*/
854 };
855 
856 
857 #define ORDERKEY_PCTCPU(a, b) do { \
858 	long diff; \
859 	if (ps.wcpu) \
860 		diff = floor(1.0E6 * weighted_cpu(pctdouble((b)->ki_pctcpu), \
861 		    (b))) - \
862 		    floor(1.0E6 * weighted_cpu(pctdouble((a)->ki_pctcpu), \
863 		    (a))); \
864 	else \
865 		diff = (long)(b)->ki_pctcpu - (long)(a)->ki_pctcpu; \
866 	if (diff != 0) \
867 		return (diff > 0 ? 1 : -1); \
868 } while (0)
869 
870 #define ORDERKEY_CPTICKS(a, b) do { \
871 	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
872 	if (diff != 0) \
873 		return (diff > 0 ? 1 : -1); \
874 } while (0)
875 
876 #define ORDERKEY_STATE(a, b) do { \
877 	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
878 	if (diff != 0) \
879 		return (diff > 0 ? 1 : -1); \
880 } while (0)
881 
882 #define ORDERKEY_PRIO(a, b) do { \
883 	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
884 	if (diff != 0) \
885 		return (diff > 0 ? 1 : -1); \
886 } while (0)
887 
888 #define	ORDERKEY_THREADS(a, b) do { \
889 	int diff = (int)(b)->ki_numthreads - (int)(a)->ki_numthreads; \
890 	if (diff != 0) \
891 		return (diff > 0 ? 1 : -1); \
892 } while (0)
893 
894 #define ORDERKEY_RSSIZE(a, b) do { \
895 	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
896 	if (diff != 0) \
897 		return (diff > 0 ? 1 : -1); \
898 } while (0)
899 
900 #define ORDERKEY_MEM(a, b) do { \
901 	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
902 	if (diff != 0) \
903 		return (diff > 0 ? 1 : -1); \
904 } while (0)
905 
906 /* compare_cpu - the comparison function for sorting by cpu percentage */
907 
908 int
909 #ifdef ORDER
910 compare_cpu(void *arg1, void *arg2)
911 #else
912 proc_compare(void *arg1, void *arg2)
913 #endif
914 {
915 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
916 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
917 
918 	ORDERKEY_PCTCPU(p1, p2);
919 	ORDERKEY_CPTICKS(p1, p2);
920 	ORDERKEY_STATE(p1, p2);
921 	ORDERKEY_PRIO(p1, p2);
922 	ORDERKEY_RSSIZE(p1, p2);
923 	ORDERKEY_MEM(p1, p2);
924 
925 	return (0);
926 }
927 
928 #ifdef ORDER
929 /* "cpu" compare routines */
930 int compare_size(), compare_res(), compare_time(), compare_prio(),
931     compare_threads();
932 
933 /*
934  * "io" compare routines.  Context switches aren't i/o, but are displayed
935  * on the "io" display.
936  */
937 int compare_iototal(), compare_ioread(), compare_iowrite(), compare_iofault(),
938     compare_vcsw(), compare_ivcsw();
939 
940 int (*compares[])() = {
941 	compare_cpu,
942 	compare_size,
943 	compare_res,
944 	compare_time,
945 	compare_prio,
946 	compare_threads,
947 	compare_iototal,
948 	compare_ioread,
949 	compare_iowrite,
950 	compare_iofault,
951 	compare_vcsw,
952 	compare_ivcsw,
953 	NULL
954 };
955 
956 /* compare_size - the comparison function for sorting by total memory usage */
957 
958 int
959 compare_size(void *arg1, void *arg2)
960 {
961 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
962 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
963 
964 	ORDERKEY_MEM(p1, p2);
965 	ORDERKEY_RSSIZE(p1, p2);
966 	ORDERKEY_PCTCPU(p1, p2);
967 	ORDERKEY_CPTICKS(p1, p2);
968 	ORDERKEY_STATE(p1, p2);
969 	ORDERKEY_PRIO(p1, p2);
970 
971 	return (0);
972 }
973 
974 /* compare_res - the comparison function for sorting by resident set size */
975 
976 int
977 compare_res(void *arg1, void *arg2)
978 {
979 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
980 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
981 
982 	ORDERKEY_RSSIZE(p1, p2);
983 	ORDERKEY_MEM(p1, p2);
984 	ORDERKEY_PCTCPU(p1, p2);
985 	ORDERKEY_CPTICKS(p1, p2);
986 	ORDERKEY_STATE(p1, p2);
987 	ORDERKEY_PRIO(p1, p2);
988 
989 	return (0);
990 }
991 
992 /* compare_time - the comparison function for sorting by total cpu time */
993 
994 int
995 compare_time(void *arg1, void *arg2)
996 {
997 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
998 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
999 
1000 	ORDERKEY_CPTICKS(p1, p2);
1001 	ORDERKEY_PCTCPU(p1, p2);
1002 	ORDERKEY_STATE(p1, p2);
1003 	ORDERKEY_PRIO(p1, p2);
1004 	ORDERKEY_RSSIZE(p1, p2);
1005 	ORDERKEY_MEM(p1, p2);
1006 
1007 	return (0);
1008 }
1009 
1010 /* compare_prio - the comparison function for sorting by priority */
1011 
1012 int
1013 compare_prio(void *arg1, void *arg2)
1014 {
1015 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1016 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1017 
1018 	ORDERKEY_PRIO(p1, p2);
1019 	ORDERKEY_CPTICKS(p1, p2);
1020 	ORDERKEY_PCTCPU(p1, p2);
1021 	ORDERKEY_STATE(p1, p2);
1022 	ORDERKEY_RSSIZE(p1, p2);
1023 	ORDERKEY_MEM(p1, p2);
1024 
1025 	return (0);
1026 }
1027 
1028 /* compare_threads - the comparison function for sorting by threads */
1029 int
1030 compare_threads(void *arg1, void *arg2)
1031 {
1032 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1033 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1034 
1035 	ORDERKEY_THREADS(p1, p2);
1036 	ORDERKEY_PCTCPU(p1, p2);
1037 	ORDERKEY_CPTICKS(p1, p2);
1038 	ORDERKEY_STATE(p1, p2);
1039 	ORDERKEY_PRIO(p1, p2);
1040 	ORDERKEY_RSSIZE(p1, p2);
1041 	ORDERKEY_MEM(p1, p2);
1042 
1043 	return (0);
1044 }
1045 #endif /* ORDER */
1046 
1047 /* assorted comparison functions for sorting by i/o */
1048 
1049 int
1050 #ifdef ORDER
1051 compare_iototal(void *arg1, void *arg2)
1052 #else
1053 io_compare(void *arg1, void *arg2)
1054 #endif
1055 {
1056 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1057 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1058 
1059 	return (get_io_total(p2) - get_io_total(p1));
1060 }
1061 
1062 #ifdef ORDER
1063 int
1064 compare_ioread(void *arg1, void *arg2)
1065 {
1066 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1067 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1068 	long dummy, inp1, inp2;
1069 
1070 	(void) get_io_stats(p1, &inp1, &dummy, &dummy, &dummy, &dummy);
1071 	(void) get_io_stats(p2, &inp2, &dummy, &dummy, &dummy, &dummy);
1072 
1073 	return (inp2 - inp1);
1074 }
1075 
1076 int
1077 compare_iowrite(void *arg1, void *arg2)
1078 {
1079 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1080 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1081 	long dummy, oup1, oup2;
1082 
1083 	(void) get_io_stats(p1, &dummy, &oup1, &dummy, &dummy, &dummy);
1084 	(void) get_io_stats(p2, &dummy, &oup2, &dummy, &dummy, &dummy);
1085 
1086 	return (oup2 - oup1);
1087 }
1088 
1089 int
1090 compare_iofault(void *arg1, void *arg2)
1091 {
1092 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1093 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1094 	long dummy, flp1, flp2;
1095 
1096 	(void) get_io_stats(p1, &dummy, &dummy, &flp1, &dummy, &dummy);
1097 	(void) get_io_stats(p2, &dummy, &dummy, &flp2, &dummy, &dummy);
1098 
1099 	return (flp2 - flp1);
1100 }
1101 
1102 int
1103 compare_vcsw(void *arg1, void *arg2)
1104 {
1105 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1106 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1107 	long dummy, flp1, flp2;
1108 
1109 	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &flp1, &dummy);
1110 	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &flp2, &dummy);
1111 
1112 	return (flp2 - flp1);
1113 }
1114 
1115 int
1116 compare_ivcsw(void *arg1, void *arg2)
1117 {
1118 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1119 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1120 	long dummy, flp1, flp2;
1121 
1122 	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &dummy, &flp1);
1123 	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &dummy, &flp2);
1124 
1125 	return (flp2 - flp1);
1126 }
1127 #endif /* ORDER */
1128 
1129 /*
1130  * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1131  *		the process does not exist.
1132  *		It is EXTREMLY IMPORTANT that this function work correctly.
1133  *		If top runs setuid root (as in SVR4), then this function
1134  *		is the only thing that stands in the way of a serious
1135  *		security problem.  It validates requests for the "kill"
1136  *		and "renice" commands.
1137  */
1138 
1139 int
1140 proc_owner(int pid)
1141 {
1142 	int cnt;
1143 	struct kinfo_proc **prefp;
1144 	struct kinfo_proc *pp;
1145 
1146 	prefp = pref;
1147 	cnt = pref_len;
1148 	while (--cnt >= 0) {
1149 		pp = *prefp++;
1150 		if (pp->ki_pid == (pid_t)pid)
1151 			return ((int)pp->ki_ruid);
1152 	}
1153 	return (-1);
1154 }
1155 
1156 static int
1157 swapmode(int *retavail, int *retfree)
1158 {
1159 	int n;
1160 	int pagesize = getpagesize();
1161 	struct kvm_swap swapary[1];
1162 
1163 	*retavail = 0;
1164 	*retfree = 0;
1165 
1166 #define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1167 
1168 	n = kvm_getswapinfo(kd, swapary, 1, 0);
1169 	if (n < 0 || swapary[0].ksw_total == 0)
1170 		return (0);
1171 
1172 	*retavail = CONVERT(swapary[0].ksw_total);
1173 	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1174 
1175 	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1176 	return (n);
1177 }
1178