xref: /freebsd/usr.bin/top/machine.c (revision 6472ac3d8a86336899b6cfb789a4cd9897e3fab5)
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 <err.h>
39 #include <kvm.h>
40 #include <math.h>
41 #include <nlist.h>
42 #include <paths.h>
43 #include <pwd.h>
44 #include <stdio.h>
45 #include <stdlib.h>
46 #include <string.h>
47 #include <strings.h>
48 #include <unistd.h>
49 #include <vis.h>
50 
51 #include "top.h"
52 #include "machine.h"
53 #include "screen.h"
54 #include "utils.h"
55 #include "layout.h"
56 
57 #define GETSYSCTL(name, var) getsysctl(name, &(var), sizeof(var))
58 #define	SMPUNAMELEN	13
59 #define	UPUNAMELEN	15
60 
61 extern struct process_select ps;
62 extern char* printable(char *);
63 static int smpmode;
64 enum displaymodes displaymode;
65 #ifdef TOP_USERNAME_LEN
66 static int namelength = TOP_USERNAME_LEN;
67 #else
68 static int namelength = 8;
69 #endif
70 static int cmdlengthdelta;
71 
72 /* Prototypes for top internals */
73 void quit(int);
74 
75 /* get_process_info passes back a handle.  This is what it looks like: */
76 
77 struct handle {
78 	struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
79 	int remaining;			/* number of pointers remaining */
80 };
81 
82 /* declarations for load_avg */
83 #include "loadavg.h"
84 
85 /* define what weighted cpu is.  */
86 #define weighted_cpu(pct, pp) ((pp)->ki_swtime == 0 ? 0.0 : \
87 			 ((pct) / (1.0 - exp((pp)->ki_swtime * logcpu))))
88 
89 /* what we consider to be process size: */
90 #define PROCSIZE(pp) ((pp)->ki_size / 1024)
91 
92 #define RU(pp)	(&(pp)->ki_rusage)
93 #define RUTOT(pp) \
94 	(RU(pp)->ru_inblock + RU(pp)->ru_oublock + RU(pp)->ru_majflt)
95 
96 
97 /* definitions for indices in the nlist array */
98 
99 /*
100  *  These definitions control the format of the per-process area
101  */
102 
103 static char io_header[] =
104     "  PID%s %-*.*s   VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
105 
106 #define io_Proc_format \
107     "%5d%s %-*.*s %6ld %6ld %6ld %6ld %6ld %6ld %6.2f%% %.*s"
108 
109 static char smp_header_thr[] =
110     "  PID%s %-*.*s  THR PRI NICE   SIZE    RES STATE   C   TIME %6s COMMAND";
111 static char smp_header[] =
112     "  PID%s %-*.*s "   "PRI NICE   SIZE    RES STATE   C   TIME %6s COMMAND";
113 
114 #define smp_Proc_format \
115     "%5d%s %-*.*s %s%3d %4s%7s %6s %-6.6s %2d%7s %5.2f%% %.*s"
116 
117 static char up_header_thr[] =
118     "  PID%s %-*.*s  THR PRI NICE   SIZE    RES STATE    TIME %6s COMMAND";
119 static char up_header[] =
120     "  PID%s %-*.*s "   "PRI NICE   SIZE    RES STATE    TIME %6s COMMAND";
121 
122 #define up_Proc_format \
123     "%5d%s %-*.*s %s%3d %4s%7s %6s %-6.6s%.0d%7s %5.2f%% %.*s"
124 
125 
126 /* process state names for the "STATE" column of the display */
127 /* the extra nulls in the string "run" are for adding a slash and
128    the processor number when needed */
129 
130 char *state_abbrev[] = {
131 	"", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB", "WAIT", "LOCK"
132 };
133 
134 
135 static kvm_t *kd;
136 
137 /* values that we stash away in _init and use in later routines */
138 
139 static double logcpu;
140 
141 /* these are retrieved from the kernel in _init */
142 
143 static load_avg  ccpu;
144 
145 /* these are used in the get_ functions */
146 
147 static int lastpid;
148 
149 /* these are for calculating cpu state percentages */
150 
151 static long cp_time[CPUSTATES];
152 static long cp_old[CPUSTATES];
153 static long cp_diff[CPUSTATES];
154 
155 /* these are for detailing the process states */
156 
157 int process_states[8];
158 char *procstatenames[] = {
159 	"", " starting, ", " running, ", " sleeping, ", " stopped, ",
160 	" zombie, ", " waiting, ", " lock, ",
161 	NULL
162 };
163 
164 /* these are for detailing the cpu states */
165 
166 int cpu_states[CPUSTATES];
167 char *cpustatenames[] = {
168 	"user", "nice", "system", "interrupt", "idle", NULL
169 };
170 
171 /* these are for detailing the memory statistics */
172 
173 int memory_stats[7];
174 char *memorynames[] = {
175 	"K Active, ", "K Inact, ", "K Wired, ", "K Cache, ", "K Buf, ",
176 	"K Free", NULL
177 };
178 
179 int swap_stats[7];
180 char *swapnames[] = {
181 	"K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
182 	NULL
183 };
184 
185 
186 /* these are for keeping track of the proc array */
187 
188 static int nproc;
189 static int onproc = -1;
190 static int pref_len;
191 static struct kinfo_proc *pbase;
192 static struct kinfo_proc **pref;
193 static struct kinfo_proc *previous_procs;
194 static struct kinfo_proc **previous_pref;
195 static int previous_proc_count = 0;
196 static int previous_proc_count_max = 0;
197 
198 /* total number of io operations */
199 static long total_inblock;
200 static long total_oublock;
201 static long total_majflt;
202 
203 /* these are for getting the memory statistics */
204 
205 static int pageshift;		/* log base 2 of the pagesize */
206 
207 /* define pagetok in terms of pageshift */
208 
209 #define pagetok(size) ((size) << pageshift)
210 
211 /* useful externals */
212 long percentages();
213 
214 #ifdef ORDER
215 /*
216  * Sorting orders.  The first element is the default.
217  */
218 char *ordernames[] = {
219 	"cpu", "size", "res", "time", "pri", "threads",
220 	"total", "read", "write", "fault", "vcsw", "ivcsw",
221 	"jid", NULL
222 };
223 #endif
224 
225 /* Per-cpu time states */
226 static int maxcpu;
227 static int maxid;
228 static int ncpus;
229 static u_long cpumask;
230 static long *times;
231 static long *pcpu_cp_time;
232 static long *pcpu_cp_old;
233 static long *pcpu_cp_diff;
234 static int *pcpu_cpu_states;
235 
236 static int compare_jid(const void *a, const void *b);
237 static int compare_pid(const void *a, const void *b);
238 static int compare_tid(const void *a, const void *b);
239 static const char *format_nice(const struct kinfo_proc *pp);
240 static void getsysctl(const char *name, void *ptr, size_t len);
241 static int swapmode(int *retavail, int *retfree);
242 
243 void
244 toggle_pcpustats(void)
245 {
246 
247 	if (ncpus == 1)
248 		return;
249 
250 	/* Adjust display based on ncpus */
251 	if (pcpu_stats) {
252 		y_mem += ncpus - 1;	/* 3 */
253 		y_swap += ncpus - 1;	/* 4 */
254 		y_idlecursor += ncpus - 1; /* 5 */
255 		y_message += ncpus - 1;	/* 5 */
256 		y_header += ncpus - 1;	/* 6 */
257 		y_procs += ncpus - 1;	/* 7 */
258 		Header_lines += ncpus - 1; /* 7 */
259 	} else {
260 		y_mem = 3;
261 		y_swap = 4;
262 		y_idlecursor = 5;
263 		y_message = 5;
264 		y_header = 6;
265 		y_procs = 7;
266 		Header_lines = 7;
267 	}
268 }
269 
270 int
271 machine_init(struct statics *statics, char do_unames)
272 {
273 	int i, j, empty, pagesize;
274 	size_t size;
275 	struct passwd *pw;
276 
277 	size = sizeof(smpmode);
278 	if ((sysctlbyname("machdep.smp_active", &smpmode, &size,
279 	    NULL, 0) != 0 &&
280 	    sysctlbyname("kern.smp.active", &smpmode, &size,
281 	    NULL, 0) != 0) ||
282 	    size != sizeof(smpmode))
283 		smpmode = 0;
284 
285 	if (do_unames) {
286 	    while ((pw = getpwent()) != NULL) {
287 		if (strlen(pw->pw_name) > namelength)
288 			namelength = strlen(pw->pw_name);
289 	    }
290 	}
291 	if (smpmode && namelength > SMPUNAMELEN)
292 		namelength = SMPUNAMELEN;
293 	else if (namelength > UPUNAMELEN)
294 		namelength = UPUNAMELEN;
295 
296 	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
297 	if (kd == NULL)
298 		return (-1);
299 
300 	GETSYSCTL("kern.ccpu", ccpu);
301 
302 	/* this is used in calculating WCPU -- calculate it ahead of time */
303 	logcpu = log(loaddouble(ccpu));
304 
305 	pbase = NULL;
306 	pref = NULL;
307 	nproc = 0;
308 	onproc = -1;
309 
310 	/* get the page size and calculate pageshift from it */
311 	pagesize = getpagesize();
312 	pageshift = 0;
313 	while (pagesize > 1) {
314 		pageshift++;
315 		pagesize >>= 1;
316 	}
317 
318 	/* we only need the amount of log(2)1024 for our conversion */
319 	pageshift -= LOG1024;
320 
321 	/* fill in the statics information */
322 	statics->procstate_names = procstatenames;
323 	statics->cpustate_names = cpustatenames;
324 	statics->memory_names = memorynames;
325 	statics->swap_names = swapnames;
326 #ifdef ORDER
327 	statics->order_names = ordernames;
328 #endif
329 
330 	/* Allocate state for per-CPU stats. */
331 	cpumask = 0;
332 	ncpus = 0;
333 	GETSYSCTL("kern.smp.maxcpus", maxcpu);
334 	size = sizeof(long) * maxcpu * CPUSTATES;
335 	times = malloc(size);
336 	if (times == NULL)
337 		err(1, "malloc %zd bytes", size);
338 	if (sysctlbyname("kern.cp_times", times, &size, NULL, 0) == -1)
339 		err(1, "sysctlbyname kern.cp_times");
340 	pcpu_cp_time = calloc(1, size);
341 	maxid = (size / CPUSTATES / sizeof(long)) - 1;
342 	for (i = 0; i <= maxid; i++) {
343 		empty = 1;
344 		for (j = 0; empty && j < CPUSTATES; j++) {
345 			if (times[i * CPUSTATES + j] != 0)
346 				empty = 0;
347 		}
348 		if (!empty) {
349 			cpumask |= (1ul << i);
350 			ncpus++;
351 		}
352 	}
353 	size = sizeof(long) * ncpus * CPUSTATES;
354 	pcpu_cp_old = calloc(1, size);
355 	pcpu_cp_diff = calloc(1, size);
356 	pcpu_cpu_states = calloc(1, size);
357 	statics->ncpus = ncpus;
358 
359 	if (pcpu_stats)
360 		toggle_pcpustats();
361 
362 	/* all done! */
363 	return (0);
364 }
365 
366 char *
367 format_header(char *uname_field)
368 {
369 	static char Header[128];
370 	const char *prehead;
371 
372 	switch (displaymode) {
373 	case DISP_CPU:
374 		/*
375 		 * The logic of picking the right header format seems reverse
376 		 * here because we only want to display a THR column when
377 		 * "thread mode" is off (and threads are not listed as
378 		 * separate lines).
379 		 */
380 		prehead = smpmode ?
381 		    (ps.thread ? smp_header : smp_header_thr) :
382 		    (ps.thread ? up_header : up_header_thr);
383 		snprintf(Header, sizeof(Header), prehead,
384 		    ps.jail ? " JID" : "",
385 		    namelength, namelength, uname_field,
386 		    ps.wcpu ? "WCPU" : "CPU");
387 		break;
388 	case DISP_IO:
389 		prehead = io_header;
390 		snprintf(Header, sizeof(Header), prehead,
391 		    ps.jail ? " JID" : "",
392 		    namelength, namelength, uname_field);
393 		break;
394 	}
395 	cmdlengthdelta = strlen(Header) - 7;
396 	return (Header);
397 }
398 
399 static int swappgsin = -1;
400 static int swappgsout = -1;
401 extern struct timeval timeout;
402 
403 
404 void
405 get_system_info(struct system_info *si)
406 {
407 	long total;
408 	struct loadavg sysload;
409 	int mib[2];
410 	struct timeval boottime;
411 	size_t bt_size;
412 	int i, j;
413 	size_t size;
414 
415 	/* get the CPU stats */
416 	size = (maxid + 1) * CPUSTATES * sizeof(long);
417 	if (sysctlbyname("kern.cp_times", pcpu_cp_time, &size, NULL, 0) == -1)
418 		err(1, "sysctlbyname kern.cp_times");
419 	GETSYSCTL("kern.cp_time", cp_time);
420 	GETSYSCTL("vm.loadavg", sysload);
421 	GETSYSCTL("kern.lastpid", lastpid);
422 
423 	/* convert load averages to doubles */
424 	for (i = 0; i < 3; i++)
425 		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
426 
427 	/* convert cp_time counts to percentages */
428 	for (i = j = 0; i <= maxid; i++) {
429 		if ((cpumask & (1ul << i)) == 0)
430 			continue;
431 		percentages(CPUSTATES, &pcpu_cpu_states[j * CPUSTATES],
432 		    &pcpu_cp_time[j * CPUSTATES],
433 		    &pcpu_cp_old[j * CPUSTATES],
434 		    &pcpu_cp_diff[j * CPUSTATES]);
435 		j++;
436 	}
437 	percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
438 
439 	/* sum memory & swap statistics */
440 	{
441 		static unsigned int swap_delay = 0;
442 		static int swapavail = 0;
443 		static int swapfree = 0;
444 		static long bufspace = 0;
445 		static int nspgsin, nspgsout;
446 
447 		GETSYSCTL("vfs.bufspace", bufspace);
448 		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
449 		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
450 		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[2]);
451 		GETSYSCTL("vm.stats.vm.v_cache_count", memory_stats[3]);
452 		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
453 		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
454 		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
455 		/* convert memory stats to Kbytes */
456 		memory_stats[0] = pagetok(memory_stats[0]);
457 		memory_stats[1] = pagetok(memory_stats[1]);
458 		memory_stats[2] = pagetok(memory_stats[2]);
459 		memory_stats[3] = pagetok(memory_stats[3]);
460 		memory_stats[4] = bufspace / 1024;
461 		memory_stats[5] = pagetok(memory_stats[5]);
462 		memory_stats[6] = -1;
463 
464 		/* first interval */
465 		if (swappgsin < 0) {
466 			swap_stats[4] = 0;
467 			swap_stats[5] = 0;
468 		}
469 
470 		/* compute differences between old and new swap statistic */
471 		else {
472 			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
473 			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
474 		}
475 
476 		swappgsin = nspgsin;
477 		swappgsout = nspgsout;
478 
479 		/* call CPU heavy swapmode() only for changes */
480 		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
481 			swap_stats[3] = swapmode(&swapavail, &swapfree);
482 			swap_stats[0] = swapavail;
483 			swap_stats[1] = swapavail - swapfree;
484 			swap_stats[2] = swapfree;
485 		}
486 		swap_delay = 1;
487 		swap_stats[6] = -1;
488 	}
489 
490 	/* set arrays and strings */
491 	if (pcpu_stats) {
492 		si->cpustates = pcpu_cpu_states;
493 		si->ncpus = ncpus;
494 	} else {
495 		si->cpustates = cpu_states;
496 		si->ncpus = 1;
497 	}
498 	si->memory = memory_stats;
499 	si->swap = swap_stats;
500 
501 
502 	if (lastpid > 0) {
503 		si->last_pid = lastpid;
504 	} else {
505 		si->last_pid = -1;
506 	}
507 
508 	/*
509 	 * Print how long system has been up.
510 	 * (Found by looking getting "boottime" from the kernel)
511 	 */
512 	mib[0] = CTL_KERN;
513 	mib[1] = KERN_BOOTTIME;
514 	bt_size = sizeof(boottime);
515 	if (sysctl(mib, 2, &boottime, &bt_size, NULL, 0) != -1 &&
516 	    boottime.tv_sec != 0) {
517 		si->boottime = boottime;
518 	} else {
519 		si->boottime.tv_sec = -1;
520 	}
521 }
522 
523 #define NOPROC	((void *)-1)
524 
525 /*
526  * We need to compare data from the old process entry with the new
527  * process entry.
528  * To facilitate doing this quickly we stash a pointer in the kinfo_proc
529  * structure to cache the mapping.  We also use a negative cache pointer
530  * of NOPROC to avoid duplicate lookups.
531  * XXX: this could be done when the actual processes are fetched, we do
532  * it here out of laziness.
533  */
534 const struct kinfo_proc *
535 get_old_proc(struct kinfo_proc *pp)
536 {
537 	struct kinfo_proc **oldpp, *oldp;
538 
539 	/*
540 	 * If this is the first fetch of the kinfo_procs then we don't have
541 	 * any previous entries.
542 	 */
543 	if (previous_proc_count == 0)
544 		return (NULL);
545 	/* negative cache? */
546 	if (pp->ki_udata == NOPROC)
547 		return (NULL);
548 	/* cached? */
549 	if (pp->ki_udata != NULL)
550 		return (pp->ki_udata);
551 	/*
552 	 * Not cached,
553 	 * 1) look up based on pid.
554 	 * 2) compare process start.
555 	 * If we fail here, then setup a negative cache entry, otherwise
556 	 * cache it.
557 	 */
558 	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
559 	    sizeof(*previous_pref), ps.thread ? compare_tid : compare_pid);
560 	if (oldpp == NULL) {
561 		pp->ki_udata = NOPROC;
562 		return (NULL);
563 	}
564 	oldp = *oldpp;
565 	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
566 		pp->ki_udata = NOPROC;
567 		return (NULL);
568 	}
569 	pp->ki_udata = oldp;
570 	return (oldp);
571 }
572 
573 /*
574  * Return the total amount of IO done in blocks in/out and faults.
575  * store the values individually in the pointers passed in.
576  */
577 long
578 get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp,
579     long *vcsw, long *ivcsw)
580 {
581 	const struct kinfo_proc *oldp;
582 	static struct kinfo_proc dummy;
583 	long ret;
584 
585 	oldp = get_old_proc(pp);
586 	if (oldp == NULL) {
587 		bzero(&dummy, sizeof(dummy));
588 		oldp = &dummy;
589 	}
590 	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
591 	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
592 	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
593 	*vcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
594 	*ivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
595 	ret =
596 	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
597 	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
598 	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
599 	return (ret);
600 }
601 
602 /*
603  * Return the total number of block in/out and faults by a process.
604  */
605 long
606 get_io_total(struct kinfo_proc *pp)
607 {
608 	long dummy;
609 
610 	return (get_io_stats(pp, &dummy, &dummy, &dummy, &dummy, &dummy));
611 }
612 
613 static struct handle handle;
614 
615 caddr_t
616 get_process_info(struct system_info *si, struct process_select *sel,
617     int (*compare)(const void *, const void *))
618 {
619 	int i;
620 	int total_procs;
621 	long p_io;
622 	long p_inblock, p_oublock, p_majflt, p_vcsw, p_ivcsw;
623 	int active_procs;
624 	struct kinfo_proc **prefp;
625 	struct kinfo_proc *pp;
626 
627 	/* these are copied out of sel for speed */
628 	int show_idle;
629 	int show_self;
630 	int show_system;
631 	int show_uid;
632 	int show_command;
633 	int show_kidle;
634 
635 	/*
636 	 * Save the previous process info.
637 	 */
638 	if (previous_proc_count_max < nproc) {
639 		free(previous_procs);
640 		previous_procs = malloc(nproc * sizeof(*previous_procs));
641 		free(previous_pref);
642 		previous_pref = malloc(nproc * sizeof(*previous_pref));
643 		if (previous_procs == NULL || previous_pref == NULL) {
644 			(void) fprintf(stderr, "top: Out of memory.\n");
645 			quit(23);
646 		}
647 		previous_proc_count_max = nproc;
648 	}
649 	if (nproc) {
650 		for (i = 0; i < nproc; i++)
651 			previous_pref[i] = &previous_procs[i];
652 		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
653 		qsort(previous_pref, nproc, sizeof(*previous_pref),
654 		    ps.thread ? compare_tid : compare_pid);
655 	}
656 	previous_proc_count = nproc;
657 
658 	pbase = kvm_getprocs(kd, sel->thread ? KERN_PROC_ALL : KERN_PROC_PROC,
659 	    0, &nproc);
660 	if (nproc > onproc)
661 		pref = realloc(pref, sizeof(*pref) * (onproc = nproc));
662 	if (pref == NULL || pbase == NULL) {
663 		(void) fprintf(stderr, "top: Out of memory.\n");
664 		quit(23);
665 	}
666 	/* get a pointer to the states summary array */
667 	si->procstates = process_states;
668 
669 	/* set up flags which define what we are going to select */
670 	show_idle = sel->idle;
671 	show_self = sel->self == -1;
672 	show_system = sel->system;
673 	show_uid = sel->uid != -1;
674 	show_command = sel->command != NULL;
675 	show_kidle = sel->kidle;
676 
677 	/* count up process states and get pointers to interesting procs */
678 	total_procs = 0;
679 	active_procs = 0;
680 	total_inblock = 0;
681 	total_oublock = 0;
682 	total_majflt = 0;
683 	memset((char *)process_states, 0, sizeof(process_states));
684 	prefp = pref;
685 	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
686 
687 		if (pp->ki_stat == 0)
688 			/* not in use */
689 			continue;
690 
691 		if (!show_self && pp->ki_pid == sel->self)
692 			/* skip self */
693 			continue;
694 
695 		if (!show_system && (pp->ki_flag & P_SYSTEM))
696 			/* skip system process */
697 			continue;
698 
699 		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt,
700 		    &p_vcsw, &p_ivcsw);
701 		total_inblock += p_inblock;
702 		total_oublock += p_oublock;
703 		total_majflt += p_majflt;
704 		total_procs++;
705 		process_states[pp->ki_stat]++;
706 
707 		if (pp->ki_stat == SZOMB)
708 			/* skip zombies */
709 			continue;
710 
711 		if (!show_kidle && pp->ki_tdflags & TDF_IDLETD)
712 			/* skip kernel idle process */
713 			continue;
714 
715 		if (displaymode == DISP_CPU && !show_idle &&
716 		    (pp->ki_pctcpu == 0 ||
717 		     pp->ki_stat == SSTOP || pp->ki_stat == SIDL))
718 			/* skip idle or non-running processes */
719 			continue;
720 
721 		if (displaymode == DISP_IO && !show_idle && p_io == 0)
722 			/* skip processes that aren't doing I/O */
723 			continue;
724 
725 		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
726 			/* skip proc. that don't belong to the selected UID */
727 			continue;
728 
729 		*prefp++ = pp;
730 		active_procs++;
731 	}
732 
733 	/* if requested, sort the "interesting" processes */
734 	if (compare != NULL)
735 		qsort(pref, active_procs, sizeof(*pref), compare);
736 
737 	/* remember active and total counts */
738 	si->p_total = total_procs;
739 	si->p_active = pref_len = active_procs;
740 
741 	/* pass back a handle */
742 	handle.next_proc = pref;
743 	handle.remaining = active_procs;
744 	return ((caddr_t)&handle);
745 }
746 
747 static char fmt[128];	/* static area where result is built */
748 
749 char *
750 format_next_process(caddr_t handle, char *(*get_userid)(int), int flags)
751 {
752 	struct kinfo_proc *pp;
753 	const struct kinfo_proc *oldp;
754 	long cputime;
755 	double pct;
756 	struct handle *hp;
757 	char status[16];
758 	int state;
759 	struct rusage ru, *rup;
760 	long p_tot, s_tot;
761 	char *proc_fmt, thr_buf[6], jid_buf[6];
762 	char *cmdbuf = NULL;
763 	char **args;
764 
765 	/* find and remember the next proc structure */
766 	hp = (struct handle *)handle;
767 	pp = *(hp->next_proc++);
768 	hp->remaining--;
769 
770 	/* get the process's command name */
771 	if ((pp->ki_flag & P_INMEM) == 0) {
772 		/*
773 		 * Print swapped processes as <pname>
774 		 */
775 		size_t len;
776 
777 		len = strlen(pp->ki_comm);
778 		if (len > sizeof(pp->ki_comm) - 3)
779 			len = sizeof(pp->ki_comm) - 3;
780 		memmove(pp->ki_comm + 1, pp->ki_comm, len);
781 		pp->ki_comm[0] = '<';
782 		pp->ki_comm[len + 1] = '>';
783 		pp->ki_comm[len + 2] = '\0';
784 	}
785 
786 	/*
787 	 * Convert the process's runtime from microseconds to seconds.  This
788 	 * time includes the interrupt time although that is not wanted here.
789 	 * ps(1) is similarly sloppy.
790 	 */
791 	cputime = (pp->ki_runtime + 500000) / 1000000;
792 
793 	/* calculate the base for cpu percentages */
794 	pct = pctdouble(pp->ki_pctcpu);
795 
796 	/* generate "STATE" field */
797 	switch (state = pp->ki_stat) {
798 	case SRUN:
799 		if (smpmode && pp->ki_oncpu != 0xff)
800 			sprintf(status, "CPU%d", pp->ki_oncpu);
801 		else
802 			strcpy(status, "RUN");
803 		break;
804 	case SLOCK:
805 		if (pp->ki_kiflag & KI_LOCKBLOCK) {
806 			sprintf(status, "*%.6s", pp->ki_lockname);
807 			break;
808 		}
809 		/* fall through */
810 	case SSLEEP:
811 		if (pp->ki_wmesg != NULL) {
812 			sprintf(status, "%.6s", pp->ki_wmesg);
813 			break;
814 		}
815 		/* FALLTHROUGH */
816 	default:
817 
818 		if (state >= 0 &&
819 		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
820 			sprintf(status, "%.6s", state_abbrev[state]);
821 		else
822 			sprintf(status, "?%5d", state);
823 		break;
824 	}
825 
826 	cmdbuf = (char *)malloc(cmdlengthdelta + 1);
827 	if (cmdbuf == NULL) {
828 		warn("malloc(%d)", cmdlengthdelta + 1);
829 		return NULL;
830 	}
831 
832 	if (!(flags & FMT_SHOWARGS)) {
833 		if (ps.thread && pp->ki_flag & P_HADTHREADS &&
834 		    pp->ki_tdname[0]) {
835 			snprintf(cmdbuf, cmdlengthdelta, "%s{%s}", pp->ki_comm,
836 			    pp->ki_tdname);
837 		} else {
838 			snprintf(cmdbuf, cmdlengthdelta, "%s", pp->ki_comm);
839 		}
840 	} else {
841 		if (pp->ki_flag & P_SYSTEM ||
842 		    pp->ki_args == NULL ||
843 		    (args = kvm_getargv(kd, pp, cmdlengthdelta)) == NULL ||
844 		    !(*args)) {
845 			if (ps.thread && pp->ki_flag & P_HADTHREADS &&
846 		    	    pp->ki_tdname[0]) {
847 				snprintf(cmdbuf, cmdlengthdelta,
848 				    "[%s{%s}]", pp->ki_comm, pp->ki_tdname);
849 			} else {
850 				snprintf(cmdbuf, cmdlengthdelta,
851 				    "[%s]", pp->ki_comm);
852 			}
853 		} else {
854 			char *src, *dst, *argbuf;
855 			char *cmd;
856 			size_t argbuflen;
857 			size_t len;
858 
859 			argbuflen = cmdlengthdelta * 4;
860 			argbuf = (char *)malloc(argbuflen + 1);
861 			if (argbuf == NULL) {
862 				warn("malloc(%d)", argbuflen + 1);
863 				free(cmdbuf);
864 				return NULL;
865 			}
866 
867 			dst = argbuf;
868 
869 			/* Extract cmd name from argv */
870 			cmd = strrchr(*args, '/');
871 			if (cmd == NULL)
872 				cmd = *args;
873 			else
874 				cmd++;
875 
876 			for (; (src = *args++) != NULL; ) {
877 				if (*src == '\0')
878 					continue;
879 				len = (argbuflen - (dst - argbuf) - 1) / 4;
880 				strvisx(dst, src,
881 				    strlen(src) < len ? strlen(src) : len,
882 				    VIS_NL | VIS_CSTYLE);
883 				while (*dst != '\0')
884 					dst++;
885 				if ((argbuflen - (dst - argbuf) - 1) / 4 > 0)
886 					*dst++ = ' '; /* add delimiting space */
887 			}
888 			if (dst != argbuf && dst[-1] == ' ')
889 				dst--;
890 			*dst = '\0';
891 
892 			if (strcmp(cmd, pp->ki_comm) != 0 ) {
893 				if (ps.thread && pp->ki_flag & P_HADTHREADS &&
894 				    pp->ki_tdname[0])
895 					snprintf(cmdbuf, cmdlengthdelta,
896 					    "%s (%s){%s}", argbuf, pp->ki_comm,
897 					    pp->ki_tdname);
898 				else
899 					snprintf(cmdbuf, cmdlengthdelta,
900 					    "%s (%s)", argbuf, pp->ki_comm);
901 			} else {
902 				if (ps.thread && pp->ki_flag & P_HADTHREADS &&
903 				    pp->ki_tdname[0])
904 					snprintf(cmdbuf, cmdlengthdelta,
905 					    "%s{%s}", argbuf, pp->ki_tdname);
906 				else
907 					strlcpy(cmdbuf, argbuf, cmdlengthdelta);
908 			}
909 			free(argbuf);
910 		}
911 	}
912 
913 	if (ps.jail == 0)
914 		jid_buf[0] = '\0';
915 	else
916 		snprintf(jid_buf, sizeof(jid_buf), " %*d",
917 		    sizeof(jid_buf) - 3, pp->ki_jid);
918 
919 	if (displaymode == DISP_IO) {
920 		oldp = get_old_proc(pp);
921 		if (oldp != NULL) {
922 			ru.ru_inblock = RU(pp)->ru_inblock -
923 			    RU(oldp)->ru_inblock;
924 			ru.ru_oublock = RU(pp)->ru_oublock -
925 			    RU(oldp)->ru_oublock;
926 			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
927 			ru.ru_nvcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
928 			ru.ru_nivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
929 			rup = &ru;
930 		} else {
931 			rup = RU(pp);
932 		}
933 		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
934 		s_tot = total_inblock + total_oublock + total_majflt;
935 
936 		sprintf(fmt, io_Proc_format,
937 		    pp->ki_pid,
938 		    jid_buf,
939 		    namelength, namelength, (*get_userid)(pp->ki_ruid),
940 		    rup->ru_nvcsw,
941 		    rup->ru_nivcsw,
942 		    rup->ru_inblock,
943 		    rup->ru_oublock,
944 		    rup->ru_majflt,
945 		    p_tot,
946 		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
947 		    screen_width > cmdlengthdelta ?
948 		    screen_width - cmdlengthdelta : 0,
949 		    printable(cmdbuf));
950 
951 		free(cmdbuf);
952 
953 		return (fmt);
954 	}
955 
956 	/* format this entry */
957 	proc_fmt = smpmode ? smp_Proc_format : up_Proc_format;
958 	if (ps.thread != 0)
959 		thr_buf[0] = '\0';
960 	else
961 		snprintf(thr_buf, sizeof(thr_buf), "%*d ",
962 		    sizeof(thr_buf) - 2, pp->ki_numthreads);
963 
964 	sprintf(fmt, proc_fmt,
965 	    pp->ki_pid,
966 	    jid_buf,
967 	    namelength, namelength, (*get_userid)(pp->ki_ruid),
968 	    thr_buf,
969 	    pp->ki_pri.pri_level - PZERO,
970 	    format_nice(pp),
971 	    format_k2(PROCSIZE(pp)),
972 	    format_k2(pagetok(pp->ki_rssize)),
973 	    status,
974 	    smpmode ? pp->ki_lastcpu : 0,
975 	    format_time(cputime),
976 	    ps.wcpu ? 100.0 * weighted_cpu(pct, pp) : 100.0 * pct,
977 	    screen_width > cmdlengthdelta ? screen_width - cmdlengthdelta : 0,
978 	    printable(cmdbuf));
979 
980 	free(cmdbuf);
981 
982 	/* return the result */
983 	return (fmt);
984 }
985 
986 static void
987 getsysctl(const char *name, void *ptr, size_t len)
988 {
989 	size_t nlen = len;
990 
991 	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
992 		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
993 		    strerror(errno));
994 		quit(23);
995 	}
996 	if (nlen != len) {
997 		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n",
998 		    name, (unsigned long)len, (unsigned long)nlen);
999 		quit(23);
1000 	}
1001 }
1002 
1003 static const char *
1004 format_nice(const struct kinfo_proc *pp)
1005 {
1006 	const char *fifo, *kthread;
1007 	int rtpri;
1008 	static char nicebuf[4 + 1];
1009 
1010 	fifo = PRI_NEED_RR(pp->ki_pri.pri_class) ? "" : "F";
1011 	kthread = (pp->ki_flag & P_KTHREAD) ? "k" : "";
1012 	switch (PRI_BASE(pp->ki_pri.pri_class)) {
1013 	case PRI_ITHD:
1014 		return ("-");
1015 	case PRI_REALTIME:
1016 		/*
1017 		 * XXX: the kernel doesn't tell us the original rtprio and
1018 		 * doesn't really know what it was, so to recover it we
1019 		 * must be more chummy with the implementation than the
1020 		 * implementation is with itself.  pri_user gives a
1021 		 * constant "base" priority, but is only initialized
1022 		 * properly for user threads.  pri_native gives what the
1023 		 * kernel calls the "base" priority, but it isn't constant
1024 		 * since it is changed by priority propagation.  pri_native
1025 		 * also isn't properly initialized for all threads, but it
1026 		 * is properly initialized for kernel realtime and idletime
1027 		 * threads.  Thus we use pri_user for the base priority of
1028 		 * user threads (it is always correct) and pri_native for
1029 		 * the base priority of kernel realtime and idletime threads
1030 		 * (there is nothing better, and it is usually correct).
1031 		 *
1032 		 * The field width and thus the buffer are too small for
1033 		 * values like "kr31F", but such values shouldn't occur,
1034 		 * and if they do then the tailing "F" is not displayed.
1035 		 */
1036 		rtpri = ((pp->ki_flag & P_KTHREAD) ? pp->ki_pri.pri_native :
1037 		    pp->ki_pri.pri_user) - PRI_MIN_REALTIME;
1038 		snprintf(nicebuf, sizeof(nicebuf), "%sr%d%s",
1039 		    kthread, rtpri, fifo);
1040 		break;
1041 	case PRI_TIMESHARE:
1042 		if (pp->ki_flag & P_KTHREAD)
1043 			return ("-");
1044 		snprintf(nicebuf, sizeof(nicebuf), "%d", pp->ki_nice - NZERO);
1045 		break;
1046 	case PRI_IDLE:
1047 		/* XXX: as above. */
1048 		rtpri = ((pp->ki_flag & P_KTHREAD) ? pp->ki_pri.pri_native :
1049 		    pp->ki_pri.pri_user) - PRI_MIN_IDLE;
1050 		snprintf(nicebuf, sizeof(nicebuf), "%si%d%s",
1051 		    kthread, rtpri, fifo);
1052 		break;
1053 	default:
1054 		return ("?");
1055 	}
1056 	return (nicebuf);
1057 }
1058 
1059 /* comparison routines for qsort */
1060 
1061 static int
1062 compare_pid(const void *p1, const void *p2)
1063 {
1064 	const struct kinfo_proc * const *pp1 = p1;
1065 	const struct kinfo_proc * const *pp2 = p2;
1066 
1067 	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
1068 		abort();
1069 
1070 	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
1071 }
1072 
1073 static int
1074 compare_tid(const void *p1, const void *p2)
1075 {
1076 	const struct kinfo_proc * const *pp1 = p1;
1077 	const struct kinfo_proc * const *pp2 = p2;
1078 
1079 	if ((*pp2)->ki_tid < 0 || (*pp1)->ki_tid < 0)
1080 		abort();
1081 
1082 	return ((*pp1)->ki_tid - (*pp2)->ki_tid);
1083 }
1084 
1085 /*
1086  *  proc_compare - comparison function for "qsort"
1087  *	Compares the resource consumption of two processes using five
1088  *	distinct keys.  The keys (in descending order of importance) are:
1089  *	percent cpu, cpu ticks, state, resident set size, total virtual
1090  *	memory usage.  The process states are ordered as follows (from least
1091  *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
1092  *	array declaration below maps a process state index into a number
1093  *	that reflects this ordering.
1094  */
1095 
1096 static int sorted_state[] = {
1097 	0,	/* not used		*/
1098 	3,	/* sleep		*/
1099 	1,	/* ABANDONED (WAIT)	*/
1100 	6,	/* run			*/
1101 	5,	/* start		*/
1102 	2,	/* zombie		*/
1103 	4	/* stop			*/
1104 };
1105 
1106 
1107 #define ORDERKEY_PCTCPU(a, b) do { \
1108 	long diff; \
1109 	if (ps.wcpu) \
1110 		diff = floor(1.0E6 * weighted_cpu(pctdouble((b)->ki_pctcpu), \
1111 		    (b))) - \
1112 		    floor(1.0E6 * weighted_cpu(pctdouble((a)->ki_pctcpu), \
1113 		    (a))); \
1114 	else \
1115 		diff = (long)(b)->ki_pctcpu - (long)(a)->ki_pctcpu; \
1116 	if (diff != 0) \
1117 		return (diff > 0 ? 1 : -1); \
1118 } while (0)
1119 
1120 #define ORDERKEY_CPTICKS(a, b) do { \
1121 	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
1122 	if (diff != 0) \
1123 		return (diff > 0 ? 1 : -1); \
1124 } while (0)
1125 
1126 #define ORDERKEY_STATE(a, b) do { \
1127 	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
1128 	if (diff != 0) \
1129 		return (diff > 0 ? 1 : -1); \
1130 } while (0)
1131 
1132 #define ORDERKEY_PRIO(a, b) do { \
1133 	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
1134 	if (diff != 0) \
1135 		return (diff > 0 ? 1 : -1); \
1136 } while (0)
1137 
1138 #define	ORDERKEY_THREADS(a, b) do { \
1139 	int diff = (int)(b)->ki_numthreads - (int)(a)->ki_numthreads; \
1140 	if (diff != 0) \
1141 		return (diff > 0 ? 1 : -1); \
1142 } while (0)
1143 
1144 #define ORDERKEY_RSSIZE(a, b) do { \
1145 	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
1146 	if (diff != 0) \
1147 		return (diff > 0 ? 1 : -1); \
1148 } while (0)
1149 
1150 #define ORDERKEY_MEM(a, b) do { \
1151 	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
1152 	if (diff != 0) \
1153 		return (diff > 0 ? 1 : -1); \
1154 } while (0)
1155 
1156 #define ORDERKEY_JID(a, b) do { \
1157 	int diff = (int)(b)->ki_jid - (int)(a)->ki_jid; \
1158 	if (diff != 0) \
1159 		return (diff > 0 ? 1 : -1); \
1160 } while (0)
1161 
1162 /* compare_cpu - the comparison function for sorting by cpu percentage */
1163 
1164 int
1165 #ifdef ORDER
1166 compare_cpu(void *arg1, void *arg2)
1167 #else
1168 proc_compare(void *arg1, void *arg2)
1169 #endif
1170 {
1171 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1172 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1173 
1174 	ORDERKEY_PCTCPU(p1, p2);
1175 	ORDERKEY_CPTICKS(p1, p2);
1176 	ORDERKEY_STATE(p1, p2);
1177 	ORDERKEY_PRIO(p1, p2);
1178 	ORDERKEY_RSSIZE(p1, p2);
1179 	ORDERKEY_MEM(p1, p2);
1180 
1181 	return (0);
1182 }
1183 
1184 #ifdef ORDER
1185 /* "cpu" compare routines */
1186 int compare_size(), compare_res(), compare_time(), compare_prio(),
1187     compare_threads();
1188 
1189 /*
1190  * "io" compare routines.  Context switches aren't i/o, but are displayed
1191  * on the "io" display.
1192  */
1193 int compare_iototal(), compare_ioread(), compare_iowrite(), compare_iofault(),
1194     compare_vcsw(), compare_ivcsw();
1195 
1196 int (*compares[])() = {
1197 	compare_cpu,
1198 	compare_size,
1199 	compare_res,
1200 	compare_time,
1201 	compare_prio,
1202 	compare_threads,
1203 	compare_iototal,
1204 	compare_ioread,
1205 	compare_iowrite,
1206 	compare_iofault,
1207 	compare_vcsw,
1208 	compare_ivcsw,
1209 	compare_jid,
1210 	NULL
1211 };
1212 
1213 /* compare_size - the comparison function for sorting by total memory usage */
1214 
1215 int
1216 compare_size(void *arg1, void *arg2)
1217 {
1218 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1219 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1220 
1221 	ORDERKEY_MEM(p1, p2);
1222 	ORDERKEY_RSSIZE(p1, p2);
1223 	ORDERKEY_PCTCPU(p1, p2);
1224 	ORDERKEY_CPTICKS(p1, p2);
1225 	ORDERKEY_STATE(p1, p2);
1226 	ORDERKEY_PRIO(p1, p2);
1227 
1228 	return (0);
1229 }
1230 
1231 /* compare_res - the comparison function for sorting by resident set size */
1232 
1233 int
1234 compare_res(void *arg1, void *arg2)
1235 {
1236 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1237 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1238 
1239 	ORDERKEY_RSSIZE(p1, p2);
1240 	ORDERKEY_MEM(p1, p2);
1241 	ORDERKEY_PCTCPU(p1, p2);
1242 	ORDERKEY_CPTICKS(p1, p2);
1243 	ORDERKEY_STATE(p1, p2);
1244 	ORDERKEY_PRIO(p1, p2);
1245 
1246 	return (0);
1247 }
1248 
1249 /* compare_time - the comparison function for sorting by total cpu time */
1250 
1251 int
1252 compare_time(void *arg1, void *arg2)
1253 {
1254 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1255 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1256 
1257 	ORDERKEY_CPTICKS(p1, p2);
1258 	ORDERKEY_PCTCPU(p1, p2);
1259 	ORDERKEY_STATE(p1, p2);
1260 	ORDERKEY_PRIO(p1, p2);
1261 	ORDERKEY_RSSIZE(p1, p2);
1262 	ORDERKEY_MEM(p1, p2);
1263 
1264 	return (0);
1265 }
1266 
1267 /* compare_prio - the comparison function for sorting by priority */
1268 
1269 int
1270 compare_prio(void *arg1, void *arg2)
1271 {
1272 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1273 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1274 
1275 	ORDERKEY_PRIO(p1, p2);
1276 	ORDERKEY_CPTICKS(p1, p2);
1277 	ORDERKEY_PCTCPU(p1, p2);
1278 	ORDERKEY_STATE(p1, p2);
1279 	ORDERKEY_RSSIZE(p1, p2);
1280 	ORDERKEY_MEM(p1, p2);
1281 
1282 	return (0);
1283 }
1284 
1285 /* compare_threads - the comparison function for sorting by threads */
1286 int
1287 compare_threads(void *arg1, void *arg2)
1288 {
1289 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1290 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1291 
1292 	ORDERKEY_THREADS(p1, p2);
1293 	ORDERKEY_PCTCPU(p1, p2);
1294 	ORDERKEY_CPTICKS(p1, p2);
1295 	ORDERKEY_STATE(p1, p2);
1296 	ORDERKEY_PRIO(p1, p2);
1297 	ORDERKEY_RSSIZE(p1, p2);
1298 	ORDERKEY_MEM(p1, p2);
1299 
1300 	return (0);
1301 }
1302 
1303 /* compare_jid - the comparison function for sorting by jid */
1304 static int
1305 compare_jid(const void *arg1, const void *arg2)
1306 {
1307 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1308 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1309 
1310 	ORDERKEY_JID(p1, p2);
1311 	ORDERKEY_PCTCPU(p1, p2);
1312 	ORDERKEY_CPTICKS(p1, p2);
1313 	ORDERKEY_STATE(p1, p2);
1314 	ORDERKEY_PRIO(p1, p2);
1315 	ORDERKEY_RSSIZE(p1, p2);
1316 	ORDERKEY_MEM(p1, p2);
1317 
1318 	return (0);
1319 }
1320 #endif /* ORDER */
1321 
1322 /* assorted comparison functions for sorting by i/o */
1323 
1324 int
1325 #ifdef ORDER
1326 compare_iototal(void *arg1, void *arg2)
1327 #else
1328 io_compare(void *arg1, void *arg2)
1329 #endif
1330 {
1331 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1332 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1333 
1334 	return (get_io_total(p2) - get_io_total(p1));
1335 }
1336 
1337 #ifdef ORDER
1338 int
1339 compare_ioread(void *arg1, void *arg2)
1340 {
1341 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1342 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1343 	long dummy, inp1, inp2;
1344 
1345 	(void) get_io_stats(p1, &inp1, &dummy, &dummy, &dummy, &dummy);
1346 	(void) get_io_stats(p2, &inp2, &dummy, &dummy, &dummy, &dummy);
1347 
1348 	return (inp2 - inp1);
1349 }
1350 
1351 int
1352 compare_iowrite(void *arg1, void *arg2)
1353 {
1354 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1355 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1356 	long dummy, oup1, oup2;
1357 
1358 	(void) get_io_stats(p1, &dummy, &oup1, &dummy, &dummy, &dummy);
1359 	(void) get_io_stats(p2, &dummy, &oup2, &dummy, &dummy, &dummy);
1360 
1361 	return (oup2 - oup1);
1362 }
1363 
1364 int
1365 compare_iofault(void *arg1, void *arg2)
1366 {
1367 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1368 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1369 	long dummy, flp1, flp2;
1370 
1371 	(void) get_io_stats(p1, &dummy, &dummy, &flp1, &dummy, &dummy);
1372 	(void) get_io_stats(p2, &dummy, &dummy, &flp2, &dummy, &dummy);
1373 
1374 	return (flp2 - flp1);
1375 }
1376 
1377 int
1378 compare_vcsw(void *arg1, void *arg2)
1379 {
1380 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1381 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1382 	long dummy, flp1, flp2;
1383 
1384 	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &flp1, &dummy);
1385 	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &flp2, &dummy);
1386 
1387 	return (flp2 - flp1);
1388 }
1389 
1390 int
1391 compare_ivcsw(void *arg1, void *arg2)
1392 {
1393 	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1394 	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1395 	long dummy, flp1, flp2;
1396 
1397 	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &dummy, &flp1);
1398 	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &dummy, &flp2);
1399 
1400 	return (flp2 - flp1);
1401 }
1402 #endif /* ORDER */
1403 
1404 /*
1405  * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1406  *		the process does not exist.
1407  *		It is EXTREMLY IMPORTANT that this function work correctly.
1408  *		If top runs setuid root (as in SVR4), then this function
1409  *		is the only thing that stands in the way of a serious
1410  *		security problem.  It validates requests for the "kill"
1411  *		and "renice" commands.
1412  */
1413 
1414 int
1415 proc_owner(int pid)
1416 {
1417 	int cnt;
1418 	struct kinfo_proc **prefp;
1419 	struct kinfo_proc *pp;
1420 
1421 	prefp = pref;
1422 	cnt = pref_len;
1423 	while (--cnt >= 0) {
1424 		pp = *prefp++;
1425 		if (pp->ki_pid == (pid_t)pid)
1426 			return ((int)pp->ki_ruid);
1427 	}
1428 	return (-1);
1429 }
1430 
1431 static int
1432 swapmode(int *retavail, int *retfree)
1433 {
1434 	int n;
1435 	int pagesize = getpagesize();
1436 	struct kvm_swap swapary[1];
1437 
1438 	*retavail = 0;
1439 	*retfree = 0;
1440 
1441 #define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1442 
1443 	n = kvm_getswapinfo(kd, swapary, 1, 0);
1444 	if (n < 0 || swapary[0].ksw_total == 0)
1445 		return (0);
1446 
1447 	*retavail = CONVERT(swapary[0].ksw_total);
1448 	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1449 
1450 	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1451 	return (n);
1452 }
1453