machine.c revision 310121
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: stable/10/usr.bin/top/machine.c 310121 2016-12-15 16:52:17Z vangyzen $
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
61extern struct process_select ps;
62extern char* printable(char *);
63static int smpmode;
64enum displaymodes displaymode;
65#ifdef TOP_USERNAME_LEN
66static int namelength = TOP_USERNAME_LEN;
67#else
68static int namelength = 8;
69#endif
70/* TOP_JID_LEN based on max of 999999 */
71#define TOP_JID_LEN 7
72static int jidlength;
73static int cmdlengthdelta;
74
75/* Prototypes for top internals */
76void quit(int);
77
78/* get_process_info passes back a handle.  This is what it looks like: */
79
80struct handle {
81	struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
82	int remaining;			/* number of pointers remaining */
83};
84
85/* declarations for load_avg */
86#include "loadavg.h"
87
88/* define what weighted cpu is.  */
89#define weighted_cpu(pct, pp) ((pp)->ki_swtime == 0 ? 0.0 : \
90			 ((pct) / (1.0 - exp((pp)->ki_swtime * logcpu))))
91
92/* what we consider to be process size: */
93#define PROCSIZE(pp) ((pp)->ki_size / 1024)
94
95#define RU(pp)	(&(pp)->ki_rusage)
96#define RUTOT(pp) \
97	(RU(pp)->ru_inblock + RU(pp)->ru_oublock + RU(pp)->ru_majflt)
98
99
100/* definitions for indices in the nlist array */
101
102/*
103 *  These definitions control the format of the per-process area
104 */
105
106static char io_header[] =
107    "  PID%*s %-*.*s   VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
108
109#define io_Proc_format \
110    "%5d%*s %-*.*s %6ld %6ld %6ld %6ld %6ld %6ld %6.2f%% %.*s"
111
112static char smp_header_thr[] =
113    "  PID%*s %-*.*s  THR PRI NICE   SIZE    RES STATE   C   TIME %7s COMMAND";
114static char smp_header[] =
115    "  PID%*s %-*.*s "   "PRI NICE   SIZE    RES STATE   C   TIME %7s COMMAND";
116
117#define smp_Proc_format \
118    "%5d%*s %-*.*s %s%3d %4s%7s %6s %-6.6s %2d%7s %6.2f%% %.*s"
119
120static char up_header_thr[] =
121    "  PID%*s %-*.*s  THR PRI NICE   SIZE    RES STATE    TIME %7s COMMAND";
122static char up_header[] =
123    "  PID%*s %-*.*s "   "PRI NICE   SIZE    RES STATE    TIME %7s COMMAND";
124
125#define up_Proc_format \
126    "%5d%*s %-*.*s %s%3d %4s%7s %6s %-6.6s%.0d%7s %6.2f%% %.*s"
127
128
129/* process state names for the "STATE" column of the display */
130/* the extra nulls in the string "run" are for adding a slash and
131   the processor number when needed */
132
133char *state_abbrev[] = {
134	"", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB", "WAIT", "LOCK"
135};
136
137
138static kvm_t *kd;
139
140/* values that we stash away in _init and use in later routines */
141
142static double logcpu;
143
144/* these are retrieved from the kernel in _init */
145
146static load_avg  ccpu;
147
148/* these are used in the get_ functions */
149
150static int lastpid;
151
152/* these are for calculating cpu state percentages */
153
154static long cp_time[CPUSTATES];
155static long cp_old[CPUSTATES];
156static long cp_diff[CPUSTATES];
157
158/* these are for detailing the process states */
159
160int process_states[8];
161char *procstatenames[] = {
162	"", " starting, ", " running, ", " sleeping, ", " stopped, ",
163	" zombie, ", " waiting, ", " lock, ",
164	NULL
165};
166
167/* these are for detailing the cpu states */
168
169int cpu_states[CPUSTATES];
170char *cpustatenames[] = {
171	"user", "nice", "system", "interrupt", "idle", NULL
172};
173
174/* these are for detailing the memory statistics */
175
176int memory_stats[7];
177char *memorynames[] = {
178	"K Active, ", "K Inact, ", "K Wired, ", "K Cache, ", "K Buf, ",
179	"K Free", NULL
180};
181
182int arc_stats[7];
183char *arcnames[] = {
184	"K Total, ", "K MFU, ", "K MRU, ", "K Anon, ", "K Header, ", "K Other",
185	NULL
186};
187
188int swap_stats[7];
189char *swapnames[] = {
190	"K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
191	NULL
192};
193
194
195/* these are for keeping track of the proc array */
196
197static int nproc;
198static int onproc = -1;
199static int pref_len;
200static struct kinfo_proc *pbase;
201static struct kinfo_proc **pref;
202static struct kinfo_proc *previous_procs;
203static struct kinfo_proc **previous_pref;
204static int previous_proc_count = 0;
205static int previous_proc_count_max = 0;
206static int arc_enabled;
207
208/* total number of io operations */
209static long total_inblock;
210static long total_oublock;
211static long total_majflt;
212
213/* these are for getting the memory statistics */
214
215static int pageshift;		/* log base 2 of the pagesize */
216
217/* define pagetok in terms of pageshift */
218
219#define pagetok(size) ((size) << pageshift)
220
221/* useful externals */
222long percentages();
223
224#ifdef ORDER
225/*
226 * Sorting orders.  The first element is the default.
227 */
228char *ordernames[] = {
229	"cpu", "size", "res", "time", "pri", "threads",
230	"total", "read", "write", "fault", "vcsw", "ivcsw",
231	"jid", "pid", NULL
232};
233#endif
234
235/* Per-cpu time states */
236static int maxcpu;
237static int maxid;
238static int ncpus;
239static u_long cpumask;
240static long *times;
241static long *pcpu_cp_time;
242static long *pcpu_cp_old;
243static long *pcpu_cp_diff;
244static int *pcpu_cpu_states;
245
246static int compare_jid(const void *a, const void *b);
247static int compare_pid(const void *a, const void *b);
248static int compare_tid(const void *a, const void *b);
249static const char *format_nice(const struct kinfo_proc *pp);
250static void getsysctl(const char *name, void *ptr, size_t len);
251static int swapmode(int *retavail, int *retfree);
252static void update_layout(void);
253
254void
255toggle_pcpustats(void)
256{
257
258	if (ncpus == 1)
259		return;
260	update_layout();
261}
262
263/* Adjust display based on ncpus and the ARC state. */
264static void
265update_layout(void)
266{
267
268	y_mem = 3;
269	y_arc = 4;
270	y_swap = 4 + arc_enabled;
271	y_idlecursor = 5 + arc_enabled;
272	y_message = 5 + arc_enabled;
273	y_header = 6 + arc_enabled;
274	y_procs = 7 + arc_enabled;
275	Header_lines = 7 + arc_enabled;
276
277	if (pcpu_stats) {
278		y_mem += ncpus - 1;
279		y_arc += ncpus - 1;
280		y_swap += ncpus - 1;
281		y_idlecursor += ncpus - 1;
282		y_message += ncpus - 1;
283		y_header += ncpus - 1;
284		y_procs += ncpus - 1;
285		Header_lines += ncpus - 1;
286	}
287}
288
289int
290machine_init(struct statics *statics, char do_unames)
291{
292	int i, j, empty, pagesize;
293	uint64_t arc_size;
294	size_t size;
295	struct passwd *pw;
296
297	size = sizeof(smpmode);
298	if ((sysctlbyname("machdep.smp_active", &smpmode, &size,
299	    NULL, 0) != 0 &&
300	    sysctlbyname("kern.smp.active", &smpmode, &size,
301	    NULL, 0) != 0) ||
302	    size != sizeof(smpmode))
303		smpmode = 0;
304
305	size = sizeof(arc_size);
306	if (sysctlbyname("kstat.zfs.misc.arcstats.size", &arc_size, &size,
307	    NULL, 0) == 0 && arc_size != 0)
308		arc_enabled = 1;
309
310	if (do_unames) {
311	    while ((pw = getpwent()) != NULL) {
312		if (strlen(pw->pw_name) > namelength)
313			namelength = strlen(pw->pw_name);
314	    }
315	}
316	if (smpmode && namelength > SMPUNAMELEN)
317		namelength = SMPUNAMELEN;
318	else if (namelength > UPUNAMELEN)
319		namelength = UPUNAMELEN;
320
321	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
322	if (kd == NULL)
323		return (-1);
324
325	GETSYSCTL("kern.ccpu", ccpu);
326
327	/* this is used in calculating WCPU -- calculate it ahead of time */
328	logcpu = log(loaddouble(ccpu));
329
330	pbase = NULL;
331	pref = NULL;
332	nproc = 0;
333	onproc = -1;
334
335	/* get the page size and calculate pageshift from it */
336	pagesize = getpagesize();
337	pageshift = 0;
338	while (pagesize > 1) {
339		pageshift++;
340		pagesize >>= 1;
341	}
342
343	/* we only need the amount of log(2)1024 for our conversion */
344	pageshift -= LOG1024;
345
346	/* fill in the statics information */
347	statics->procstate_names = procstatenames;
348	statics->cpustate_names = cpustatenames;
349	statics->memory_names = memorynames;
350	if (arc_enabled)
351		statics->arc_names = arcnames;
352	else
353		statics->arc_names = NULL;
354	statics->swap_names = swapnames;
355#ifdef ORDER
356	statics->order_names = ordernames;
357#endif
358
359	/* Allocate state for per-CPU stats. */
360	cpumask = 0;
361	ncpus = 0;
362	GETSYSCTL("kern.smp.maxcpus", maxcpu);
363	size = sizeof(long) * maxcpu * CPUSTATES;
364	times = malloc(size);
365	if (times == NULL)
366		err(1, "malloc %zd bytes", size);
367	if (sysctlbyname("kern.cp_times", times, &size, NULL, 0) == -1)
368		err(1, "sysctlbyname kern.cp_times");
369	pcpu_cp_time = calloc(1, size);
370	maxid = (size / CPUSTATES / sizeof(long)) - 1;
371	for (i = 0; i <= maxid; i++) {
372		empty = 1;
373		for (j = 0; empty && j < CPUSTATES; j++) {
374			if (times[i * CPUSTATES + j] != 0)
375				empty = 0;
376		}
377		if (!empty) {
378			cpumask |= (1ul << i);
379			ncpus++;
380		}
381	}
382	size = sizeof(long) * ncpus * CPUSTATES;
383	pcpu_cp_old = calloc(1, size);
384	pcpu_cp_diff = calloc(1, size);
385	pcpu_cpu_states = calloc(1, size);
386	statics->ncpus = ncpus;
387
388	update_layout();
389
390	/* all done! */
391	return (0);
392}
393
394char *
395format_header(char *uname_field)
396{
397	static char Header[128];
398	const char *prehead;
399
400	if (ps.jail)
401		jidlength = TOP_JID_LEN + 1;	/* +1 for extra left space. */
402	else
403		jidlength = 0;
404
405	switch (displaymode) {
406	case DISP_CPU:
407		/*
408		 * The logic of picking the right header format seems reverse
409		 * here because we only want to display a THR column when
410		 * "thread mode" is off (and threads are not listed as
411		 * separate lines).
412		 */
413		prehead = smpmode ?
414		    (ps.thread ? smp_header : smp_header_thr) :
415		    (ps.thread ? up_header : up_header_thr);
416		snprintf(Header, sizeof(Header), prehead,
417		    jidlength, ps.jail ? " JID" : "",
418		    namelength, namelength, uname_field,
419		    ps.wcpu ? "WCPU" : "CPU");
420		break;
421	case DISP_IO:
422		prehead = io_header;
423		snprintf(Header, sizeof(Header), prehead,
424		    jidlength, ps.jail ? " JID" : "",
425		    namelength, namelength, uname_field);
426		break;
427	}
428	cmdlengthdelta = strlen(Header) - 7;
429	return (Header);
430}
431
432static int swappgsin = -1;
433static int swappgsout = -1;
434extern struct timeval timeout;
435
436
437void
438get_system_info(struct system_info *si)
439{
440	long total;
441	struct loadavg sysload;
442	int mib[2];
443	struct timeval boottime;
444	uint64_t arc_stat, arc_stat2;
445	int i, j;
446	size_t size;
447
448	/* get the CPU stats */
449	size = (maxid + 1) * CPUSTATES * sizeof(long);
450	if (sysctlbyname("kern.cp_times", pcpu_cp_time, &size, NULL, 0) == -1)
451		err(1, "sysctlbyname kern.cp_times");
452	GETSYSCTL("kern.cp_time", cp_time);
453	GETSYSCTL("vm.loadavg", sysload);
454	GETSYSCTL("kern.lastpid", lastpid);
455
456	/* convert load averages to doubles */
457	for (i = 0; i < 3; i++)
458		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
459
460	/* convert cp_time counts to percentages */
461	for (i = j = 0; i <= maxid; i++) {
462		if ((cpumask & (1ul << i)) == 0)
463			continue;
464		percentages(CPUSTATES, &pcpu_cpu_states[j * CPUSTATES],
465		    &pcpu_cp_time[j * CPUSTATES],
466		    &pcpu_cp_old[j * CPUSTATES],
467		    &pcpu_cp_diff[j * CPUSTATES]);
468		j++;
469	}
470	percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
471
472	/* sum memory & swap statistics */
473	{
474		static unsigned int swap_delay = 0;
475		static int swapavail = 0;
476		static int swapfree = 0;
477		static long bufspace = 0;
478		static int nspgsin, nspgsout;
479
480		GETSYSCTL("vfs.bufspace", bufspace);
481		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
482		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
483		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[2]);
484		GETSYSCTL("vm.stats.vm.v_cache_count", memory_stats[3]);
485		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
486		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
487		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
488		/* convert memory stats to Kbytes */
489		memory_stats[0] = pagetok(memory_stats[0]);
490		memory_stats[1] = pagetok(memory_stats[1]);
491		memory_stats[2] = pagetok(memory_stats[2]);
492		memory_stats[3] = pagetok(memory_stats[3]);
493		memory_stats[4] = bufspace / 1024;
494		memory_stats[5] = pagetok(memory_stats[5]);
495		memory_stats[6] = -1;
496
497		/* first interval */
498		if (swappgsin < 0) {
499			swap_stats[4] = 0;
500			swap_stats[5] = 0;
501		}
502
503		/* compute differences between old and new swap statistic */
504		else {
505			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
506			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
507		}
508
509		swappgsin = nspgsin;
510		swappgsout = nspgsout;
511
512		/* call CPU heavy swapmode() only for changes */
513		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
514			swap_stats[3] = swapmode(&swapavail, &swapfree);
515			swap_stats[0] = swapavail;
516			swap_stats[1] = swapavail - swapfree;
517			swap_stats[2] = swapfree;
518		}
519		swap_delay = 1;
520		swap_stats[6] = -1;
521	}
522
523	if (arc_enabled) {
524		GETSYSCTL("kstat.zfs.misc.arcstats.size", arc_stat);
525		arc_stats[0] = arc_stat >> 10;
526		GETSYSCTL("vfs.zfs.mfu_size", arc_stat);
527		arc_stats[1] = arc_stat >> 10;
528		GETSYSCTL("vfs.zfs.mru_size", arc_stat);
529		arc_stats[2] = arc_stat >> 10;
530		GETSYSCTL("vfs.zfs.anon_size", arc_stat);
531		arc_stats[3] = arc_stat >> 10;
532		GETSYSCTL("kstat.zfs.misc.arcstats.hdr_size", arc_stat);
533		GETSYSCTL("kstat.zfs.misc.arcstats.l2_hdr_size", arc_stat2);
534		arc_stats[4] = arc_stat + arc_stat2 >> 10;
535		GETSYSCTL("kstat.zfs.misc.arcstats.other_size", arc_stat);
536		arc_stats[5] = arc_stat >> 10;
537		si->arc = arc_stats;
538	}
539
540	/* set arrays and strings */
541	if (pcpu_stats) {
542		si->cpustates = pcpu_cpu_states;
543		si->ncpus = ncpus;
544	} else {
545		si->cpustates = cpu_states;
546		si->ncpus = 1;
547	}
548	si->memory = memory_stats;
549	si->swap = swap_stats;
550
551
552	if (lastpid > 0) {
553		si->last_pid = lastpid;
554	} else {
555		si->last_pid = -1;
556	}
557
558	/*
559	 * Print how long system has been up.
560	 * (Found by looking getting "boottime" from the kernel)
561	 */
562	mib[0] = CTL_KERN;
563	mib[1] = KERN_BOOTTIME;
564	size = sizeof(boottime);
565	if (sysctl(mib, 2, &boottime, &size, NULL, 0) != -1 &&
566	    boottime.tv_sec != 0) {
567		si->boottime = boottime;
568	} else {
569		si->boottime.tv_sec = -1;
570	}
571}
572
573#define NOPROC	((void *)-1)
574
575/*
576 * We need to compare data from the old process entry with the new
577 * process entry.
578 * To facilitate doing this quickly we stash a pointer in the kinfo_proc
579 * structure to cache the mapping.  We also use a negative cache pointer
580 * of NOPROC to avoid duplicate lookups.
581 * XXX: this could be done when the actual processes are fetched, we do
582 * it here out of laziness.
583 */
584const struct kinfo_proc *
585get_old_proc(struct kinfo_proc *pp)
586{
587	struct kinfo_proc **oldpp, *oldp;
588
589	/*
590	 * If this is the first fetch of the kinfo_procs then we don't have
591	 * any previous entries.
592	 */
593	if (previous_proc_count == 0)
594		return (NULL);
595	/* negative cache? */
596	if (pp->ki_udata == NOPROC)
597		return (NULL);
598	/* cached? */
599	if (pp->ki_udata != NULL)
600		return (pp->ki_udata);
601	/*
602	 * Not cached,
603	 * 1) look up based on pid.
604	 * 2) compare process start.
605	 * If we fail here, then setup a negative cache entry, otherwise
606	 * cache it.
607	 */
608	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
609	    sizeof(*previous_pref), ps.thread ? compare_tid : compare_pid);
610	if (oldpp == NULL) {
611		pp->ki_udata = NOPROC;
612		return (NULL);
613	}
614	oldp = *oldpp;
615	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
616		pp->ki_udata = NOPROC;
617		return (NULL);
618	}
619	pp->ki_udata = oldp;
620	return (oldp);
621}
622
623/*
624 * Return the total amount of IO done in blocks in/out and faults.
625 * store the values individually in the pointers passed in.
626 */
627long
628get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp,
629    long *vcsw, long *ivcsw)
630{
631	const struct kinfo_proc *oldp;
632	static struct kinfo_proc dummy;
633	long ret;
634
635	oldp = get_old_proc(pp);
636	if (oldp == NULL) {
637		bzero(&dummy, sizeof(dummy));
638		oldp = &dummy;
639	}
640	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
641	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
642	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
643	*vcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
644	*ivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
645	ret =
646	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
647	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
648	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
649	return (ret);
650}
651
652/*
653 * Return the total number of block in/out and faults by a process.
654 */
655long
656get_io_total(struct kinfo_proc *pp)
657{
658	long dummy;
659
660	return (get_io_stats(pp, &dummy, &dummy, &dummy, &dummy, &dummy));
661}
662
663static struct handle handle;
664
665caddr_t
666get_process_info(struct system_info *si, struct process_select *sel,
667    int (*compare)(const void *, const void *))
668{
669	int i;
670	int total_procs;
671	long p_io;
672	long p_inblock, p_oublock, p_majflt, p_vcsw, p_ivcsw;
673	int active_procs;
674	struct kinfo_proc **prefp;
675	struct kinfo_proc *pp;
676
677	/* these are copied out of sel for speed */
678	int show_idle;
679	int show_jid;
680	int show_self;
681	int show_system;
682	int show_uid;
683	int show_command;
684	int show_kidle;
685
686	/*
687	 * Save the previous process info.
688	 */
689	if (previous_proc_count_max < nproc) {
690		free(previous_procs);
691		previous_procs = malloc(nproc * sizeof(*previous_procs));
692		free(previous_pref);
693		previous_pref = malloc(nproc * sizeof(*previous_pref));
694		if (previous_procs == NULL || previous_pref == NULL) {
695			(void) fprintf(stderr, "top: Out of memory.\n");
696			quit(23);
697		}
698		previous_proc_count_max = nproc;
699	}
700	if (nproc) {
701		for (i = 0; i < nproc; i++)
702			previous_pref[i] = &previous_procs[i];
703		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
704		qsort(previous_pref, nproc, sizeof(*previous_pref),
705		    ps.thread ? compare_tid : compare_pid);
706	}
707	previous_proc_count = nproc;
708
709	pbase = kvm_getprocs(kd, sel->thread ? KERN_PROC_ALL : KERN_PROC_PROC,
710	    0, &nproc);
711	if (nproc > onproc)
712		pref = realloc(pref, sizeof(*pref) * (onproc = nproc));
713	if (pref == NULL || pbase == NULL) {
714		(void) fprintf(stderr, "top: Out of memory.\n");
715		quit(23);
716	}
717	/* get a pointer to the states summary array */
718	si->procstates = process_states;
719
720	/* set up flags which define what we are going to select */
721	show_idle = sel->idle;
722	show_jid = sel->jid != -1;
723	show_self = sel->self == -1;
724	show_system = sel->system;
725	show_uid = sel->uid != -1;
726	show_command = sel->command != NULL;
727	show_kidle = sel->kidle;
728
729	/* count up process states and get pointers to interesting procs */
730	total_procs = 0;
731	active_procs = 0;
732	total_inblock = 0;
733	total_oublock = 0;
734	total_majflt = 0;
735	memset((char *)process_states, 0, sizeof(process_states));
736	prefp = pref;
737	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
738
739		if (pp->ki_stat == 0)
740			/* not in use */
741			continue;
742
743		if (!show_self && pp->ki_pid == sel->self)
744			/* skip self */
745			continue;
746
747		if (!show_system && (pp->ki_flag & P_SYSTEM))
748			/* skip system process */
749			continue;
750
751		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt,
752		    &p_vcsw, &p_ivcsw);
753		total_inblock += p_inblock;
754		total_oublock += p_oublock;
755		total_majflt += p_majflt;
756		total_procs++;
757		process_states[pp->ki_stat]++;
758
759		if (pp->ki_stat == SZOMB)
760			/* skip zombies */
761			continue;
762
763		if (!show_kidle && pp->ki_tdflags & TDF_IDLETD)
764			/* skip kernel idle process */
765			continue;
766
767		if (displaymode == DISP_CPU && !show_idle &&
768		    (pp->ki_pctcpu == 0 ||
769		     pp->ki_stat == SSTOP || pp->ki_stat == SIDL))
770			/* skip idle or non-running processes */
771			continue;
772
773		if (displaymode == DISP_IO && !show_idle && p_io == 0)
774			/* skip processes that aren't doing I/O */
775			continue;
776
777		if (show_jid && pp->ki_jid != sel->jid)
778			/* skip proc. that don't belong to the selected JID */
779			continue;
780
781		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
782			/* skip proc. that don't belong to the selected UID */
783			continue;
784
785		*prefp++ = pp;
786		active_procs++;
787	}
788
789	/* if requested, sort the "interesting" processes */
790	if (compare != NULL)
791		qsort(pref, active_procs, sizeof(*pref), compare);
792
793	/* remember active and total counts */
794	si->p_total = total_procs;
795	si->p_active = pref_len = active_procs;
796
797	/* pass back a handle */
798	handle.next_proc = pref;
799	handle.remaining = active_procs;
800	return ((caddr_t)&handle);
801}
802
803static char fmt[512];	/* static area where result is built */
804
805char *
806format_next_process(caddr_t handle, char *(*get_userid)(int), int flags)
807{
808	struct kinfo_proc *pp;
809	const struct kinfo_proc *oldp;
810	long cputime;
811	double pct;
812	struct handle *hp;
813	char status[16];
814	int cpu, state;
815	struct rusage ru, *rup;
816	long p_tot, s_tot;
817	char *proc_fmt, thr_buf[6], jid_buf[TOP_JID_LEN + 1];
818	char *cmdbuf = NULL;
819	char **args;
820	const int cmdlen = 128;
821
822	/* find and remember the next proc structure */
823	hp = (struct handle *)handle;
824	pp = *(hp->next_proc++);
825	hp->remaining--;
826
827	/* get the process's command name */
828	if ((pp->ki_flag & P_INMEM) == 0) {
829		/*
830		 * Print swapped processes as <pname>
831		 */
832		size_t len;
833
834		len = strlen(pp->ki_comm);
835		if (len > sizeof(pp->ki_comm) - 3)
836			len = sizeof(pp->ki_comm) - 3;
837		memmove(pp->ki_comm + 1, pp->ki_comm, len);
838		pp->ki_comm[0] = '<';
839		pp->ki_comm[len + 1] = '>';
840		pp->ki_comm[len + 2] = '\0';
841	}
842
843	/*
844	 * Convert the process's runtime from microseconds to seconds.  This
845	 * time includes the interrupt time although that is not wanted here.
846	 * ps(1) is similarly sloppy.
847	 */
848	cputime = (pp->ki_runtime + 500000) / 1000000;
849
850	/* calculate the base for cpu percentages */
851	pct = pctdouble(pp->ki_pctcpu);
852
853	/* generate "STATE" field */
854	switch (state = pp->ki_stat) {
855	case SRUN:
856		if (smpmode && pp->ki_oncpu != 0xff)
857			sprintf(status, "CPU%d", pp->ki_oncpu);
858		else
859			strcpy(status, "RUN");
860		break;
861	case SLOCK:
862		if (pp->ki_kiflag & KI_LOCKBLOCK) {
863			sprintf(status, "*%.6s", pp->ki_lockname);
864			break;
865		}
866		/* fall through */
867	case SSLEEP:
868		if (pp->ki_wmesg != NULL) {
869			sprintf(status, "%.6s", pp->ki_wmesg);
870			break;
871		}
872		/* FALLTHROUGH */
873	default:
874
875		if (state >= 0 &&
876		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
877			sprintf(status, "%.6s", state_abbrev[state]);
878		else
879			sprintf(status, "?%5d", state);
880		break;
881	}
882
883	cmdbuf = (char *)malloc(cmdlen + 1);
884	if (cmdbuf == NULL) {
885		warn("malloc(%d)", cmdlen + 1);
886		return NULL;
887	}
888
889	if (!(flags & FMT_SHOWARGS)) {
890		if (ps.thread && pp->ki_flag & P_HADTHREADS &&
891		    pp->ki_tdname[0]) {
892			snprintf(cmdbuf, cmdlen, "%s{%s%s}", pp->ki_comm,
893			    pp->ki_tdname, pp->ki_moretdname);
894		} else {
895			snprintf(cmdbuf, cmdlen, "%s", pp->ki_comm);
896		}
897	} else {
898		if (pp->ki_flag & P_SYSTEM ||
899		    pp->ki_args == NULL ||
900		    (args = kvm_getargv(kd, pp, cmdlen)) == NULL ||
901		    !(*args)) {
902			if (ps.thread && pp->ki_flag & P_HADTHREADS &&
903		    	    pp->ki_tdname[0]) {
904				snprintf(cmdbuf, cmdlen,
905				    "[%s{%s%s}]", pp->ki_comm, pp->ki_tdname,
906				    pp->ki_moretdname);
907			} else {
908				snprintf(cmdbuf, cmdlen,
909				    "[%s]", pp->ki_comm);
910			}
911		} else {
912			char *src, *dst, *argbuf;
913			char *cmd;
914			size_t argbuflen;
915			size_t len;
916
917			argbuflen = cmdlen * 4;
918			argbuf = (char *)malloc(argbuflen + 1);
919			if (argbuf == NULL) {
920				warn("malloc(%d)", argbuflen + 1);
921				free(cmdbuf);
922				return NULL;
923			}
924
925			dst = argbuf;
926
927			/* Extract cmd name from argv */
928			cmd = strrchr(*args, '/');
929			if (cmd == NULL)
930				cmd = *args;
931			else
932				cmd++;
933
934			for (; (src = *args++) != NULL; ) {
935				if (*src == '\0')
936					continue;
937				len = (argbuflen - (dst - argbuf) - 1) / 4;
938				strvisx(dst, src,
939				    strlen(src) < len ? strlen(src) : len,
940				    VIS_NL | VIS_CSTYLE);
941				while (*dst != '\0')
942					dst++;
943				if ((argbuflen - (dst - argbuf) - 1) / 4 > 0)
944					*dst++ = ' '; /* add delimiting space */
945			}
946			if (dst != argbuf && dst[-1] == ' ')
947				dst--;
948			*dst = '\0';
949
950			if (strcmp(cmd, pp->ki_comm) != 0) {
951				if (ps.thread && pp->ki_flag & P_HADTHREADS &&
952				    pp->ki_tdname[0])
953					snprintf(cmdbuf, cmdlen,
954					    "%s (%s){%s%s}", argbuf,
955					    pp->ki_comm, pp->ki_tdname,
956					    pp->ki_moretdname);
957				else
958					snprintf(cmdbuf, cmdlen,
959					    "%s (%s)", argbuf, pp->ki_comm);
960			} else {
961				if (ps.thread && pp->ki_flag & P_HADTHREADS &&
962				    pp->ki_tdname[0])
963					snprintf(cmdbuf, cmdlen,
964					    "%s{%s%s}", argbuf, pp->ki_tdname,
965					    pp->ki_moretdname);
966				else
967					strlcpy(cmdbuf, argbuf, cmdlen);
968			}
969			free(argbuf);
970		}
971	}
972
973	if (ps.jail == 0)
974		jid_buf[0] = '\0';
975	else
976		snprintf(jid_buf, sizeof(jid_buf), "%*d",
977		    jidlength - 1, pp->ki_jid);
978
979	if (displaymode == DISP_IO) {
980		oldp = get_old_proc(pp);
981		if (oldp != NULL) {
982			ru.ru_inblock = RU(pp)->ru_inblock -
983			    RU(oldp)->ru_inblock;
984			ru.ru_oublock = RU(pp)->ru_oublock -
985			    RU(oldp)->ru_oublock;
986			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
987			ru.ru_nvcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
988			ru.ru_nivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
989			rup = &ru;
990		} else {
991			rup = RU(pp);
992		}
993		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
994		s_tot = total_inblock + total_oublock + total_majflt;
995
996		snprintf(fmt, sizeof(fmt), io_Proc_format,
997		    pp->ki_pid,
998		    jidlength, jid_buf,
999		    namelength, namelength, (*get_userid)(pp->ki_ruid),
1000		    rup->ru_nvcsw,
1001		    rup->ru_nivcsw,
1002		    rup->ru_inblock,
1003		    rup->ru_oublock,
1004		    rup->ru_majflt,
1005		    p_tot,
1006		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
1007		    screen_width > cmdlengthdelta ?
1008		    screen_width - cmdlengthdelta : 0,
1009		    printable(cmdbuf));
1010
1011		free(cmdbuf);
1012
1013		return (fmt);
1014	}
1015
1016	/* format this entry */
1017	if (smpmode) {
1018		if (state == SRUN && pp->ki_oncpu != 0xff)
1019			cpu = pp->ki_oncpu;
1020		else
1021			cpu = pp->ki_lastcpu;
1022	} else
1023		cpu = 0;
1024	proc_fmt = smpmode ? smp_Proc_format : up_Proc_format;
1025	if (ps.thread != 0)
1026		thr_buf[0] = '\0';
1027	else
1028		snprintf(thr_buf, sizeof(thr_buf), "%*d ",
1029		    sizeof(thr_buf) - 2, pp->ki_numthreads);
1030
1031	snprintf(fmt, sizeof(fmt), proc_fmt,
1032	    pp->ki_pid,
1033	    jidlength, jid_buf,
1034	    namelength, namelength, (*get_userid)(pp->ki_ruid),
1035	    thr_buf,
1036	    pp->ki_pri.pri_level - PZERO,
1037	    format_nice(pp),
1038	    format_k2(PROCSIZE(pp)),
1039	    format_k2(pagetok(pp->ki_rssize)),
1040	    status,
1041	    cpu,
1042	    format_time(cputime),
1043	    ps.wcpu ? 100.0 * weighted_cpu(pct, pp) : 100.0 * pct,
1044	    screen_width > cmdlengthdelta ? screen_width - cmdlengthdelta : 0,
1045	    printable(cmdbuf));
1046
1047	free(cmdbuf);
1048
1049	/* return the result */
1050	return (fmt);
1051}
1052
1053static void
1054getsysctl(const char *name, void *ptr, size_t len)
1055{
1056	size_t nlen = len;
1057
1058	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
1059		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
1060		    strerror(errno));
1061		quit(23);
1062	}
1063	if (nlen != len) {
1064		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n",
1065		    name, (unsigned long)len, (unsigned long)nlen);
1066		quit(23);
1067	}
1068}
1069
1070static const char *
1071format_nice(const struct kinfo_proc *pp)
1072{
1073	const char *fifo, *kthread;
1074	int rtpri;
1075	static char nicebuf[4 + 1];
1076
1077	fifo = PRI_NEED_RR(pp->ki_pri.pri_class) ? "" : "F";
1078	kthread = (pp->ki_flag & P_KTHREAD) ? "k" : "";
1079	switch (PRI_BASE(pp->ki_pri.pri_class)) {
1080	case PRI_ITHD:
1081		return ("-");
1082	case PRI_REALTIME:
1083		/*
1084		 * XXX: the kernel doesn't tell us the original rtprio and
1085		 * doesn't really know what it was, so to recover it we
1086		 * must be more chummy with the implementation than the
1087		 * implementation is with itself.  pri_user gives a
1088		 * constant "base" priority, but is only initialized
1089		 * properly for user threads.  pri_native gives what the
1090		 * kernel calls the "base" priority, but it isn't constant
1091		 * since it is changed by priority propagation.  pri_native
1092		 * also isn't properly initialized for all threads, but it
1093		 * is properly initialized for kernel realtime and idletime
1094		 * threads.  Thus we use pri_user for the base priority of
1095		 * user threads (it is always correct) and pri_native for
1096		 * the base priority of kernel realtime and idletime threads
1097		 * (there is nothing better, and it is usually correct).
1098		 *
1099		 * The field width and thus the buffer are too small for
1100		 * values like "kr31F", but such values shouldn't occur,
1101		 * and if they do then the tailing "F" is not displayed.
1102		 */
1103		rtpri = ((pp->ki_flag & P_KTHREAD) ? pp->ki_pri.pri_native :
1104		    pp->ki_pri.pri_user) - PRI_MIN_REALTIME;
1105		snprintf(nicebuf, sizeof(nicebuf), "%sr%d%s",
1106		    kthread, rtpri, fifo);
1107		break;
1108	case PRI_TIMESHARE:
1109		if (pp->ki_flag & P_KTHREAD)
1110			return ("-");
1111		snprintf(nicebuf, sizeof(nicebuf), "%d", pp->ki_nice - NZERO);
1112		break;
1113	case PRI_IDLE:
1114		/* XXX: as above. */
1115		rtpri = ((pp->ki_flag & P_KTHREAD) ? pp->ki_pri.pri_native :
1116		    pp->ki_pri.pri_user) - PRI_MIN_IDLE;
1117		snprintf(nicebuf, sizeof(nicebuf), "%si%d%s",
1118		    kthread, rtpri, fifo);
1119		break;
1120	default:
1121		return ("?");
1122	}
1123	return (nicebuf);
1124}
1125
1126/* comparison routines for qsort */
1127
1128static int
1129compare_pid(const void *p1, const void *p2)
1130{
1131	const struct kinfo_proc * const *pp1 = p1;
1132	const struct kinfo_proc * const *pp2 = p2;
1133
1134	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
1135		abort();
1136
1137	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
1138}
1139
1140static int
1141compare_tid(const void *p1, const void *p2)
1142{
1143	const struct kinfo_proc * const *pp1 = p1;
1144	const struct kinfo_proc * const *pp2 = p2;
1145
1146	if ((*pp2)->ki_tid < 0 || (*pp1)->ki_tid < 0)
1147		abort();
1148
1149	return ((*pp1)->ki_tid - (*pp2)->ki_tid);
1150}
1151
1152/*
1153 *  proc_compare - comparison function for "qsort"
1154 *	Compares the resource consumption of two processes using five
1155 *	distinct keys.  The keys (in descending order of importance) are:
1156 *	percent cpu, cpu ticks, state, resident set size, total virtual
1157 *	memory usage.  The process states are ordered as follows (from least
1158 *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
1159 *	array declaration below maps a process state index into a number
1160 *	that reflects this ordering.
1161 */
1162
1163static int sorted_state[] = {
1164	0,	/* not used		*/
1165	3,	/* sleep		*/
1166	1,	/* ABANDONED (WAIT)	*/
1167	6,	/* run			*/
1168	5,	/* start		*/
1169	2,	/* zombie		*/
1170	4	/* stop			*/
1171};
1172
1173
1174#define ORDERKEY_PCTCPU(a, b) do { \
1175	long diff; \
1176	if (ps.wcpu) \
1177		diff = floor(1.0E6 * weighted_cpu(pctdouble((b)->ki_pctcpu), \
1178		    (b))) - \
1179		    floor(1.0E6 * weighted_cpu(pctdouble((a)->ki_pctcpu), \
1180		    (a))); \
1181	else \
1182		diff = (long)(b)->ki_pctcpu - (long)(a)->ki_pctcpu; \
1183	if (diff != 0) \
1184		return (diff > 0 ? 1 : -1); \
1185} while (0)
1186
1187#define ORDERKEY_CPTICKS(a, b) do { \
1188	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
1189	if (diff != 0) \
1190		return (diff > 0 ? 1 : -1); \
1191} while (0)
1192
1193#define ORDERKEY_STATE(a, b) do { \
1194	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
1195	if (diff != 0) \
1196		return (diff > 0 ? 1 : -1); \
1197} while (0)
1198
1199#define ORDERKEY_PRIO(a, b) do { \
1200	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
1201	if (diff != 0) \
1202		return (diff > 0 ? 1 : -1); \
1203} while (0)
1204
1205#define	ORDERKEY_THREADS(a, b) do { \
1206	int diff = (int)(b)->ki_numthreads - (int)(a)->ki_numthreads; \
1207	if (diff != 0) \
1208		return (diff > 0 ? 1 : -1); \
1209} while (0)
1210
1211#define ORDERKEY_RSSIZE(a, b) do { \
1212	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
1213	if (diff != 0) \
1214		return (diff > 0 ? 1 : -1); \
1215} while (0)
1216
1217#define ORDERKEY_MEM(a, b) do { \
1218	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
1219	if (diff != 0) \
1220		return (diff > 0 ? 1 : -1); \
1221} while (0)
1222
1223#define ORDERKEY_JID(a, b) do { \
1224	int diff = (int)(b)->ki_jid - (int)(a)->ki_jid; \
1225	if (diff != 0) \
1226		return (diff > 0 ? 1 : -1); \
1227} while (0)
1228
1229/* compare_cpu - the comparison function for sorting by cpu percentage */
1230
1231int
1232#ifdef ORDER
1233compare_cpu(void *arg1, void *arg2)
1234#else
1235proc_compare(void *arg1, void *arg2)
1236#endif
1237{
1238	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1239	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1240
1241	ORDERKEY_PCTCPU(p1, p2);
1242	ORDERKEY_CPTICKS(p1, p2);
1243	ORDERKEY_STATE(p1, p2);
1244	ORDERKEY_PRIO(p1, p2);
1245	ORDERKEY_RSSIZE(p1, p2);
1246	ORDERKEY_MEM(p1, p2);
1247
1248	return (0);
1249}
1250
1251#ifdef ORDER
1252/* "cpu" compare routines */
1253int compare_size(), compare_res(), compare_time(), compare_prio(),
1254    compare_threads();
1255
1256/*
1257 * "io" compare routines.  Context switches aren't i/o, but are displayed
1258 * on the "io" display.
1259 */
1260int compare_iototal(), compare_ioread(), compare_iowrite(), compare_iofault(),
1261    compare_vcsw(), compare_ivcsw();
1262
1263int (*compares[])() = {
1264	compare_cpu,
1265	compare_size,
1266	compare_res,
1267	compare_time,
1268	compare_prio,
1269	compare_threads,
1270	compare_iototal,
1271	compare_ioread,
1272	compare_iowrite,
1273	compare_iofault,
1274	compare_vcsw,
1275	compare_ivcsw,
1276	compare_jid,
1277	NULL
1278};
1279
1280/* compare_size - the comparison function for sorting by total memory usage */
1281
1282int
1283compare_size(void *arg1, void *arg2)
1284{
1285	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1286	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1287
1288	ORDERKEY_MEM(p1, p2);
1289	ORDERKEY_RSSIZE(p1, p2);
1290	ORDERKEY_PCTCPU(p1, p2);
1291	ORDERKEY_CPTICKS(p1, p2);
1292	ORDERKEY_STATE(p1, p2);
1293	ORDERKEY_PRIO(p1, p2);
1294
1295	return (0);
1296}
1297
1298/* compare_res - the comparison function for sorting by resident set size */
1299
1300int
1301compare_res(void *arg1, void *arg2)
1302{
1303	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1304	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1305
1306	ORDERKEY_RSSIZE(p1, p2);
1307	ORDERKEY_MEM(p1, p2);
1308	ORDERKEY_PCTCPU(p1, p2);
1309	ORDERKEY_CPTICKS(p1, p2);
1310	ORDERKEY_STATE(p1, p2);
1311	ORDERKEY_PRIO(p1, p2);
1312
1313	return (0);
1314}
1315
1316/* compare_time - the comparison function for sorting by total cpu time */
1317
1318int
1319compare_time(void *arg1, void *arg2)
1320{
1321	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1322	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1323
1324	ORDERKEY_CPTICKS(p1, p2);
1325	ORDERKEY_PCTCPU(p1, p2);
1326	ORDERKEY_STATE(p1, p2);
1327	ORDERKEY_PRIO(p1, p2);
1328	ORDERKEY_RSSIZE(p1, p2);
1329	ORDERKEY_MEM(p1, p2);
1330
1331	return (0);
1332}
1333
1334/* compare_prio - the comparison function for sorting by priority */
1335
1336int
1337compare_prio(void *arg1, void *arg2)
1338{
1339	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1340	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1341
1342	ORDERKEY_PRIO(p1, p2);
1343	ORDERKEY_CPTICKS(p1, p2);
1344	ORDERKEY_PCTCPU(p1, p2);
1345	ORDERKEY_STATE(p1, p2);
1346	ORDERKEY_RSSIZE(p1, p2);
1347	ORDERKEY_MEM(p1, p2);
1348
1349	return (0);
1350}
1351
1352/* compare_threads - the comparison function for sorting by threads */
1353int
1354compare_threads(void *arg1, void *arg2)
1355{
1356	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1357	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1358
1359	ORDERKEY_THREADS(p1, p2);
1360	ORDERKEY_PCTCPU(p1, p2);
1361	ORDERKEY_CPTICKS(p1, p2);
1362	ORDERKEY_STATE(p1, p2);
1363	ORDERKEY_PRIO(p1, p2);
1364	ORDERKEY_RSSIZE(p1, p2);
1365	ORDERKEY_MEM(p1, p2);
1366
1367	return (0);
1368}
1369
1370/* compare_jid - the comparison function for sorting by jid */
1371static int
1372compare_jid(const void *arg1, const void *arg2)
1373{
1374	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1375	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1376
1377	ORDERKEY_JID(p1, p2);
1378	ORDERKEY_PCTCPU(p1, p2);
1379	ORDERKEY_CPTICKS(p1, p2);
1380	ORDERKEY_STATE(p1, p2);
1381	ORDERKEY_PRIO(p1, p2);
1382	ORDERKEY_RSSIZE(p1, p2);
1383	ORDERKEY_MEM(p1, p2);
1384
1385	return (0);
1386}
1387#endif /* ORDER */
1388
1389/* assorted comparison functions for sorting by i/o */
1390
1391int
1392#ifdef ORDER
1393compare_iototal(void *arg1, void *arg2)
1394#else
1395io_compare(void *arg1, void *arg2)
1396#endif
1397{
1398	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1399	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1400
1401	return (get_io_total(p2) - get_io_total(p1));
1402}
1403
1404#ifdef ORDER
1405int
1406compare_ioread(void *arg1, void *arg2)
1407{
1408	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1409	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1410	long dummy, inp1, inp2;
1411
1412	(void) get_io_stats(p1, &inp1, &dummy, &dummy, &dummy, &dummy);
1413	(void) get_io_stats(p2, &inp2, &dummy, &dummy, &dummy, &dummy);
1414
1415	return (inp2 - inp1);
1416}
1417
1418int
1419compare_iowrite(void *arg1, void *arg2)
1420{
1421	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1422	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1423	long dummy, oup1, oup2;
1424
1425	(void) get_io_stats(p1, &dummy, &oup1, &dummy, &dummy, &dummy);
1426	(void) get_io_stats(p2, &dummy, &oup2, &dummy, &dummy, &dummy);
1427
1428	return (oup2 - oup1);
1429}
1430
1431int
1432compare_iofault(void *arg1, void *arg2)
1433{
1434	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1435	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1436	long dummy, flp1, flp2;
1437
1438	(void) get_io_stats(p1, &dummy, &dummy, &flp1, &dummy, &dummy);
1439	(void) get_io_stats(p2, &dummy, &dummy, &flp2, &dummy, &dummy);
1440
1441	return (flp2 - flp1);
1442}
1443
1444int
1445compare_vcsw(void *arg1, void *arg2)
1446{
1447	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1448	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1449	long dummy, flp1, flp2;
1450
1451	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &flp1, &dummy);
1452	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &flp2, &dummy);
1453
1454	return (flp2 - flp1);
1455}
1456
1457int
1458compare_ivcsw(void *arg1, void *arg2)
1459{
1460	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1461	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1462	long dummy, flp1, flp2;
1463
1464	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &dummy, &flp1);
1465	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &dummy, &flp2);
1466
1467	return (flp2 - flp1);
1468}
1469#endif /* ORDER */
1470
1471/*
1472 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1473 *		the process does not exist.
1474 *		It is EXTREMELY IMPORTANT that this function work correctly.
1475 *		If top runs setuid root (as in SVR4), then this function
1476 *		is the only thing that stands in the way of a serious
1477 *		security problem.  It validates requests for the "kill"
1478 *		and "renice" commands.
1479 */
1480
1481int
1482proc_owner(int pid)
1483{
1484	int cnt;
1485	struct kinfo_proc **prefp;
1486	struct kinfo_proc *pp;
1487
1488	prefp = pref;
1489	cnt = pref_len;
1490	while (--cnt >= 0) {
1491		pp = *prefp++;
1492		if (pp->ki_pid == (pid_t)pid)
1493			return ((int)pp->ki_ruid);
1494	}
1495	return (-1);
1496}
1497
1498static int
1499swapmode(int *retavail, int *retfree)
1500{
1501	int n;
1502	int pagesize = getpagesize();
1503	struct kvm_swap swapary[1];
1504
1505	*retavail = 0;
1506	*retfree = 0;
1507
1508#define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1509
1510	n = kvm_getswapinfo(kd, swapary, 1, 0);
1511	if (n < 0 || swapary[0].ksw_total == 0)
1512		return (0);
1513
1514	*retavail = CONVERT(swapary[0].ksw_total);
1515	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1516
1517	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1518	return (n);
1519}
1520