machine.c revision 319982
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/11/usr.bin/top/machine.c 319982 2017-06-15 17:43:40Z allanjude $
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
72#define TOP_SWAP_LEN 6
73static int jidlength;
74static int swaplength;
75static int cmdlengthdelta;
76
77/* Prototypes for top internals */
78void quit(int);
79
80/* get_process_info passes back a handle.  This is what it looks like: */
81
82struct handle {
83	struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
84	int remaining;			/* number of pointers remaining */
85};
86
87/* declarations for load_avg */
88#include "loadavg.h"
89
90/* define what weighted cpu is.  */
91#define weighted_cpu(pct, pp) ((pp)->ki_swtime == 0 ? 0.0 : \
92			 ((pct) / (1.0 - exp((pp)->ki_swtime * logcpu))))
93
94/* what we consider to be process size: */
95#define PROCSIZE(pp) ((pp)->ki_size / 1024)
96
97#define RU(pp)	(&(pp)->ki_rusage)
98#define RUTOT(pp) \
99	(RU(pp)->ru_inblock + RU(pp)->ru_oublock + RU(pp)->ru_majflt)
100
101#define	PCTCPU(pp) (pcpu[pp - pbase])
102
103/* definitions for indices in the nlist array */
104
105/*
106 *  These definitions control the format of the per-process area
107 */
108
109static char io_header[] =
110    "  PID%*s %-*.*s   VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
111
112#define io_Proc_format \
113    "%5d%*s %-*.*s %6ld %6ld %6ld %6ld %6ld %6ld %6.2f%% %.*s"
114
115static char smp_header_thr[] =
116    "  PID%*s %-*.*s  THR PRI NICE   SIZE    RES%*s STATE   C   TIME %7s COMMAND";
117static char smp_header[] =
118    "  PID%*s %-*.*s "   "PRI NICE   SIZE    RES%*s STATE   C   TIME %7s COMMAND";
119
120#define smp_Proc_format \
121    "%5d%*s %-*.*s %s%3d %4s%7s %6s%*.*s %-6.6s %2d%7s %6.2f%% %.*s"
122
123static char up_header_thr[] =
124    "  PID%*s %-*.*s  THR PRI NICE   SIZE    RES%*s STATE    TIME %7s COMMAND";
125static char up_header[] =
126    "  PID%*s %-*.*s "   "PRI NICE   SIZE    RES%*s STATE    TIME %7s COMMAND";
127
128#define up_Proc_format \
129    "%5d%*s %-*.*s %s%3d %4s%7s %6s%*.*s %-6.6s%.0d%7s %6.2f%% %.*s"
130
131
132/* process state names for the "STATE" column of the display */
133/* the extra nulls in the string "run" are for adding a slash and
134   the processor number when needed */
135
136char *state_abbrev[] = {
137	"", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB", "WAIT", "LOCK"
138};
139
140
141static kvm_t *kd;
142
143/* values that we stash away in _init and use in later routines */
144
145static double logcpu;
146
147/* these are retrieved from the kernel in _init */
148
149static load_avg  ccpu;
150
151/* these are used in the get_ functions */
152
153static int lastpid;
154
155/* these are for calculating cpu state percentages */
156
157static long cp_time[CPUSTATES];
158static long cp_old[CPUSTATES];
159static long cp_diff[CPUSTATES];
160
161/* these are for detailing the process states */
162
163int process_states[8];
164char *procstatenames[] = {
165	"", " starting, ", " running, ", " sleeping, ", " stopped, ",
166	" zombie, ", " waiting, ", " lock, ",
167	NULL
168};
169
170/* these are for detailing the cpu states */
171
172int cpu_states[CPUSTATES];
173char *cpustatenames[] = {
174	"user", "nice", "system", "interrupt", "idle", NULL
175};
176
177/* these are for detailing the memory statistics */
178
179int memory_stats[7];
180char *memorynames[] = {
181	"K Active, ", "K Inact, ", "K Laundry, ", "K Wired, ", "K Buf, ",
182	"K Free", NULL
183};
184
185int arc_stats[7];
186char *arcnames[] = {
187	"K Total, ", "K MFU, ", "K MRU, ", "K Anon, ", "K Header, ", "K Other",
188	NULL
189};
190
191int carc_stats[4];
192char *carcnames[] = {
193	"K Compressed, ", "K Uncompressed, ", ":1 Ratio, ",
194	NULL
195};
196
197int swap_stats[7];
198char *swapnames[] = {
199	"K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
200	NULL
201};
202
203
204/* these are for keeping track of the proc array */
205
206static int nproc;
207static int onproc = -1;
208static int pref_len;
209static struct kinfo_proc *pbase;
210static struct kinfo_proc **pref;
211static struct kinfo_proc *previous_procs;
212static struct kinfo_proc **previous_pref;
213static int previous_proc_count = 0;
214static int previous_proc_count_max = 0;
215static int previous_thread;
216
217/* data used for recalculating pctcpu */
218static double *pcpu;
219static struct timespec proc_uptime;
220static struct timeval proc_wall_time;
221static struct timeval previous_wall_time;
222static uint64_t previous_interval = 0;
223
224/* total number of io operations */
225static long total_inblock;
226static long total_oublock;
227static long total_majflt;
228
229/* these are for getting the memory statistics */
230
231static int arc_enabled;
232static int carc_enabled;
233static int pageshift;		/* log base 2 of the pagesize */
234
235/* define pagetok in terms of pageshift */
236
237#define pagetok(size) ((size) << pageshift)
238
239/* swap usage */
240#define ki_swap(kip) \
241    ((kip)->ki_swrss > (kip)->ki_rssize ? (kip)->ki_swrss - (kip)->ki_rssize : 0)
242
243/* useful externals */
244long percentages();
245
246#ifdef ORDER
247/*
248 * Sorting orders.  The first element is the default.
249 */
250char *ordernames[] = {
251	"cpu", "size", "res", "time", "pri", "threads",
252	"total", "read", "write", "fault", "vcsw", "ivcsw",
253	"jid", "swap", "pid", NULL
254};
255#endif
256
257/* Per-cpu time states */
258static int maxcpu;
259static int maxid;
260static int ncpus;
261static u_long cpumask;
262static long *times;
263static long *pcpu_cp_time;
264static long *pcpu_cp_old;
265static long *pcpu_cp_diff;
266static int *pcpu_cpu_states;
267
268static int compare_swap(const void *a, const void *b);
269static int compare_jid(const void *a, const void *b);
270static int compare_pid(const void *a, const void *b);
271static int compare_tid(const void *a, const void *b);
272static const char *format_nice(const struct kinfo_proc *pp);
273static void getsysctl(const char *name, void *ptr, size_t len);
274static int swapmode(int *retavail, int *retfree);
275static void update_layout(void);
276
277void
278toggle_pcpustats(void)
279{
280
281	if (ncpus == 1)
282		return;
283	update_layout();
284}
285
286/* Adjust display based on ncpus and the ARC state. */
287static void
288update_layout(void)
289{
290
291	y_mem = 3;
292	y_arc = 4;
293	y_carc = 5;
294	y_swap = 4 + arc_enabled + carc_enabled;
295	y_idlecursor = 5 + arc_enabled + carc_enabled;
296	y_message = 5 + arc_enabled + carc_enabled;
297	y_header = 6 + arc_enabled + carc_enabled;
298	y_procs = 7 + arc_enabled + carc_enabled;
299	Header_lines = 7 + arc_enabled + carc_enabled;
300
301	if (pcpu_stats) {
302		y_mem += ncpus - 1;
303		y_arc += ncpus - 1;
304		y_carc += ncpus - 1;
305		y_swap += ncpus - 1;
306		y_idlecursor += ncpus - 1;
307		y_message += ncpus - 1;
308		y_header += ncpus - 1;
309		y_procs += ncpus - 1;
310		Header_lines += ncpus - 1;
311	}
312}
313
314int
315machine_init(struct statics *statics, char do_unames)
316{
317	int i, j, empty, pagesize;
318	uint64_t arc_size;
319	boolean_t carc_en;
320	size_t size;
321	struct passwd *pw;
322
323	size = sizeof(smpmode);
324	if ((sysctlbyname("machdep.smp_active", &smpmode, &size,
325	    NULL, 0) != 0 &&
326	    sysctlbyname("kern.smp.active", &smpmode, &size,
327	    NULL, 0) != 0) ||
328	    size != sizeof(smpmode))
329		smpmode = 0;
330
331	size = sizeof(carc_en);
332	if (sysctlbyname("vfs.zfs.compressed_arc_enabled", &carc_en, &size,
333	    NULL, 0) == 0 && carc_en == 1)
334		carc_enabled = 1;
335	size = sizeof(arc_size);
336	if (sysctlbyname("kstat.zfs.misc.arcstats.size", &arc_size, &size,
337	    NULL, 0) == 0 && arc_size != 0)
338		arc_enabled = 1;
339
340	if (do_unames) {
341	    while ((pw = getpwent()) != NULL) {
342		if (strlen(pw->pw_name) > namelength)
343			namelength = strlen(pw->pw_name);
344	    }
345	}
346	if (smpmode && namelength > SMPUNAMELEN)
347		namelength = SMPUNAMELEN;
348	else if (namelength > UPUNAMELEN)
349		namelength = UPUNAMELEN;
350
351	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
352	if (kd == NULL)
353		return (-1);
354
355	GETSYSCTL("kern.ccpu", ccpu);
356
357	/* this is used in calculating WCPU -- calculate it ahead of time */
358	logcpu = log(loaddouble(ccpu));
359
360	pbase = NULL;
361	pref = NULL;
362	pcpu = NULL;
363	nproc = 0;
364	onproc = -1;
365
366	/* get the page size and calculate pageshift from it */
367	pagesize = getpagesize();
368	pageshift = 0;
369	while (pagesize > 1) {
370		pageshift++;
371		pagesize >>= 1;
372	}
373
374	/* we only need the amount of log(2)1024 for our conversion */
375	pageshift -= LOG1024;
376
377	/* fill in the statics information */
378	statics->procstate_names = procstatenames;
379	statics->cpustate_names = cpustatenames;
380	statics->memory_names = memorynames;
381	if (arc_enabled)
382		statics->arc_names = arcnames;
383	else
384		statics->arc_names = NULL;
385	if (carc_enabled)
386		statics->carc_names = carcnames;
387	else
388		statics->carc_names = NULL;
389	statics->swap_names = swapnames;
390#ifdef ORDER
391	statics->order_names = ordernames;
392#endif
393
394	/* Allocate state for per-CPU stats. */
395	cpumask = 0;
396	ncpus = 0;
397	GETSYSCTL("kern.smp.maxcpus", maxcpu);
398	size = sizeof(long) * maxcpu * CPUSTATES;
399	times = malloc(size);
400	if (times == NULL)
401		err(1, "malloc %zu bytes", size);
402	if (sysctlbyname("kern.cp_times", times, &size, NULL, 0) == -1)
403		err(1, "sysctlbyname kern.cp_times");
404	pcpu_cp_time = calloc(1, size);
405	maxid = (size / CPUSTATES / sizeof(long)) - 1;
406	for (i = 0; i <= maxid; i++) {
407		empty = 1;
408		for (j = 0; empty && j < CPUSTATES; j++) {
409			if (times[i * CPUSTATES + j] != 0)
410				empty = 0;
411		}
412		if (!empty) {
413			cpumask |= (1ul << i);
414			ncpus++;
415		}
416	}
417	size = sizeof(long) * ncpus * CPUSTATES;
418	pcpu_cp_old = calloc(1, size);
419	pcpu_cp_diff = calloc(1, size);
420	pcpu_cpu_states = calloc(1, size);
421	statics->ncpus = ncpus;
422
423	update_layout();
424
425	/* all done! */
426	return (0);
427}
428
429char *
430format_header(char *uname_field)
431{
432	static char Header[128];
433	const char *prehead;
434
435	if (ps.jail)
436		jidlength = TOP_JID_LEN + 1;	/* +1 for extra left space. */
437	else
438		jidlength = 0;
439
440	if (ps.swap)
441		swaplength = TOP_SWAP_LEN + 1;  /* +1 for extra left space */
442	else
443		swaplength = 0;
444
445	switch (displaymode) {
446	case DISP_CPU:
447		/*
448		 * The logic of picking the right header format seems reverse
449		 * here because we only want to display a THR column when
450		 * "thread mode" is off (and threads are not listed as
451		 * separate lines).
452		 */
453		prehead = smpmode ?
454		    (ps.thread ? smp_header : smp_header_thr) :
455		    (ps.thread ? up_header : up_header_thr);
456		snprintf(Header, sizeof(Header), prehead,
457		    jidlength, ps.jail ? " JID" : "",
458		    namelength, namelength, uname_field,
459		    swaplength, ps.swap ? " SWAP" : "",
460		    ps.wcpu ? "WCPU" : "CPU");
461		break;
462	case DISP_IO:
463		prehead = io_header;
464		snprintf(Header, sizeof(Header), prehead,
465		    jidlength, ps.jail ? " JID" : "",
466		    namelength, namelength, uname_field);
467		break;
468	}
469	cmdlengthdelta = strlen(Header) - 7;
470	return (Header);
471}
472
473static int swappgsin = -1;
474static int swappgsout = -1;
475extern struct timeval timeout;
476
477
478void
479get_system_info(struct system_info *si)
480{
481	long total;
482	struct loadavg sysload;
483	int mib[2];
484	struct timeval boottime;
485	uint64_t arc_stat, arc_stat2;
486	int i, j;
487	size_t size;
488
489	/* get the CPU stats */
490	size = (maxid + 1) * CPUSTATES * sizeof(long);
491	if (sysctlbyname("kern.cp_times", pcpu_cp_time, &size, NULL, 0) == -1)
492		err(1, "sysctlbyname kern.cp_times");
493	GETSYSCTL("kern.cp_time", cp_time);
494	GETSYSCTL("vm.loadavg", sysload);
495	GETSYSCTL("kern.lastpid", lastpid);
496
497	/* convert load averages to doubles */
498	for (i = 0; i < 3; i++)
499		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
500
501	/* convert cp_time counts to percentages */
502	for (i = j = 0; i <= maxid; i++) {
503		if ((cpumask & (1ul << i)) == 0)
504			continue;
505		percentages(CPUSTATES, &pcpu_cpu_states[j * CPUSTATES],
506		    &pcpu_cp_time[j * CPUSTATES],
507		    &pcpu_cp_old[j * CPUSTATES],
508		    &pcpu_cp_diff[j * CPUSTATES]);
509		j++;
510	}
511	percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
512
513	/* sum memory & swap statistics */
514	{
515		static unsigned int swap_delay = 0;
516		static int swapavail = 0;
517		static int swapfree = 0;
518		static long bufspace = 0;
519		static int nspgsin, nspgsout;
520
521		GETSYSCTL("vfs.bufspace", bufspace);
522		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
523		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
524		GETSYSCTL("vm.stats.vm.v_laundry_count", memory_stats[2]);
525		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[3]);
526		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
527		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
528		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
529		/* convert memory stats to Kbytes */
530		memory_stats[0] = pagetok(memory_stats[0]);
531		memory_stats[1] = pagetok(memory_stats[1]);
532		memory_stats[2] = pagetok(memory_stats[2]);
533		memory_stats[3] = pagetok(memory_stats[3]);
534		memory_stats[4] = bufspace / 1024;
535		memory_stats[5] = pagetok(memory_stats[5]);
536		memory_stats[6] = -1;
537
538		/* first interval */
539		if (swappgsin < 0) {
540			swap_stats[4] = 0;
541			swap_stats[5] = 0;
542		}
543
544		/* compute differences between old and new swap statistic */
545		else {
546			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
547			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
548		}
549
550		swappgsin = nspgsin;
551		swappgsout = nspgsout;
552
553		/* call CPU heavy swapmode() only for changes */
554		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
555			swap_stats[3] = swapmode(&swapavail, &swapfree);
556			swap_stats[0] = swapavail;
557			swap_stats[1] = swapavail - swapfree;
558			swap_stats[2] = swapfree;
559		}
560		swap_delay = 1;
561		swap_stats[6] = -1;
562	}
563
564	if (arc_enabled) {
565		GETSYSCTL("kstat.zfs.misc.arcstats.size", arc_stat);
566		arc_stats[0] = arc_stat >> 10;
567		GETSYSCTL("vfs.zfs.mfu_size", arc_stat);
568		arc_stats[1] = arc_stat >> 10;
569		GETSYSCTL("vfs.zfs.mru_size", arc_stat);
570		arc_stats[2] = arc_stat >> 10;
571		GETSYSCTL("vfs.zfs.anon_size", arc_stat);
572		arc_stats[3] = arc_stat >> 10;
573		GETSYSCTL("kstat.zfs.misc.arcstats.hdr_size", arc_stat);
574		GETSYSCTL("kstat.zfs.misc.arcstats.l2_hdr_size", arc_stat2);
575		arc_stats[4] = arc_stat + arc_stat2 >> 10;
576		GETSYSCTL("kstat.zfs.misc.arcstats.other_size", arc_stat);
577		arc_stats[5] = arc_stat >> 10;
578		si->arc = arc_stats;
579	}
580	if (carc_enabled) {
581		GETSYSCTL("kstat.zfs.misc.arcstats.compressed_size", arc_stat);
582		carc_stats[0] = arc_stat >> 10;
583		carc_stats[2] = arc_stat >> 10; /* For ratio */
584		GETSYSCTL("kstat.zfs.misc.arcstats.uncompressed_size", arc_stat);
585		carc_stats[1] = arc_stat >> 10;
586		si->carc = carc_stats;
587	}
588
589	/* set arrays and strings */
590	if (pcpu_stats) {
591		si->cpustates = pcpu_cpu_states;
592		si->ncpus = ncpus;
593	} else {
594		si->cpustates = cpu_states;
595		si->ncpus = 1;
596	}
597	si->memory = memory_stats;
598	si->swap = swap_stats;
599
600
601	if (lastpid > 0) {
602		si->last_pid = lastpid;
603	} else {
604		si->last_pid = -1;
605	}
606
607	/*
608	 * Print how long system has been up.
609	 * (Found by looking getting "boottime" from the kernel)
610	 */
611	mib[0] = CTL_KERN;
612	mib[1] = KERN_BOOTTIME;
613	size = sizeof(boottime);
614	if (sysctl(mib, nitems(mib), &boottime, &size, NULL, 0) != -1 &&
615	    boottime.tv_sec != 0) {
616		si->boottime = boottime;
617	} else {
618		si->boottime.tv_sec = -1;
619	}
620}
621
622#define NOPROC	((void *)-1)
623
624/*
625 * We need to compare data from the old process entry with the new
626 * process entry.
627 * To facilitate doing this quickly we stash a pointer in the kinfo_proc
628 * structure to cache the mapping.  We also use a negative cache pointer
629 * of NOPROC to avoid duplicate lookups.
630 * XXX: this could be done when the actual processes are fetched, we do
631 * it here out of laziness.
632 */
633const struct kinfo_proc *
634get_old_proc(struct kinfo_proc *pp)
635{
636	struct kinfo_proc **oldpp, *oldp;
637
638	/*
639	 * If this is the first fetch of the kinfo_procs then we don't have
640	 * any previous entries.
641	 */
642	if (previous_proc_count == 0)
643		return (NULL);
644	/* negative cache? */
645	if (pp->ki_udata == NOPROC)
646		return (NULL);
647	/* cached? */
648	if (pp->ki_udata != NULL)
649		return (pp->ki_udata);
650	/*
651	 * Not cached,
652	 * 1) look up based on pid.
653	 * 2) compare process start.
654	 * If we fail here, then setup a negative cache entry, otherwise
655	 * cache it.
656	 */
657	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
658	    sizeof(*previous_pref), ps.thread ? compare_tid : compare_pid);
659	if (oldpp == NULL) {
660		pp->ki_udata = NOPROC;
661		return (NULL);
662	}
663	oldp = *oldpp;
664	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
665		pp->ki_udata = NOPROC;
666		return (NULL);
667	}
668	pp->ki_udata = oldp;
669	return (oldp);
670}
671
672/*
673 * Return the total amount of IO done in blocks in/out and faults.
674 * store the values individually in the pointers passed in.
675 */
676long
677get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp,
678    long *vcsw, long *ivcsw)
679{
680	const struct kinfo_proc *oldp;
681	static struct kinfo_proc dummy;
682	long ret;
683
684	oldp = get_old_proc(pp);
685	if (oldp == NULL) {
686		bzero(&dummy, sizeof(dummy));
687		oldp = &dummy;
688	}
689	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
690	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
691	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
692	*vcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
693	*ivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
694	ret =
695	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
696	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
697	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
698	return (ret);
699}
700
701/*
702 * If there was a previous update, use the delta in ki_runtime over
703 * the previous interval to calculate pctcpu.  Otherwise, fall back
704 * to using the kernel's ki_pctcpu.
705 */
706static double
707proc_calc_pctcpu(struct kinfo_proc *pp)
708{
709	const struct kinfo_proc *oldp;
710
711	if (previous_interval != 0) {
712		oldp = get_old_proc(pp);
713		if (oldp != NULL)
714			return ((double)(pp->ki_runtime - oldp->ki_runtime)
715			    / previous_interval);
716
717		/*
718		 * If this process/thread was created during the previous
719		 * interval, charge it's total runtime to the previous
720		 * interval.
721		 */
722		else if (pp->ki_start.tv_sec > previous_wall_time.tv_sec ||
723		    (pp->ki_start.tv_sec == previous_wall_time.tv_sec &&
724		    pp->ki_start.tv_usec >= previous_wall_time.tv_usec))
725			return ((double)pp->ki_runtime / previous_interval);
726	}
727	return (pctdouble(pp->ki_pctcpu));
728}
729
730/*
731 * Return true if this process has used any CPU time since the
732 * previous update.
733 */
734static int
735proc_used_cpu(struct kinfo_proc *pp)
736{
737	const struct kinfo_proc *oldp;
738
739	oldp = get_old_proc(pp);
740	if (oldp == NULL)
741		return (PCTCPU(pp) != 0);
742	return (pp->ki_runtime != oldp->ki_runtime ||
743	    RU(pp)->ru_nvcsw != RU(oldp)->ru_nvcsw ||
744	    RU(pp)->ru_nivcsw != RU(oldp)->ru_nivcsw);
745}
746
747/*
748 * Return the total number of block in/out and faults by a process.
749 */
750long
751get_io_total(struct kinfo_proc *pp)
752{
753	long dummy;
754
755	return (get_io_stats(pp, &dummy, &dummy, &dummy, &dummy, &dummy));
756}
757
758static struct handle handle;
759
760caddr_t
761get_process_info(struct system_info *si, struct process_select *sel,
762    int (*compare)(const void *, const void *))
763{
764	int i;
765	int total_procs;
766	long p_io;
767	long p_inblock, p_oublock, p_majflt, p_vcsw, p_ivcsw;
768	long nsec;
769	int active_procs;
770	struct kinfo_proc **prefp;
771	struct kinfo_proc *pp;
772	struct timespec previous_proc_uptime;
773
774	/* these are copied out of sel for speed */
775	int show_idle;
776	int show_jid;
777	int show_self;
778	int show_system;
779	int show_uid;
780	int show_command;
781	int show_kidle;
782
783	/*
784	 * If thread state was toggled, don't cache the previous processes.
785	 */
786	if (previous_thread != sel->thread)
787		nproc = 0;
788	previous_thread = sel->thread;
789
790	/*
791	 * Save the previous process info.
792	 */
793	if (previous_proc_count_max < nproc) {
794		free(previous_procs);
795		previous_procs = malloc(nproc * sizeof(*previous_procs));
796		free(previous_pref);
797		previous_pref = malloc(nproc * sizeof(*previous_pref));
798		if (previous_procs == NULL || previous_pref == NULL) {
799			(void) fprintf(stderr, "top: Out of memory.\n");
800			quit(23);
801		}
802		previous_proc_count_max = nproc;
803	}
804	if (nproc) {
805		for (i = 0; i < nproc; i++)
806			previous_pref[i] = &previous_procs[i];
807		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
808		qsort(previous_pref, nproc, sizeof(*previous_pref),
809		    ps.thread ? compare_tid : compare_pid);
810	}
811	previous_proc_count = nproc;
812	previous_proc_uptime = proc_uptime;
813	previous_wall_time = proc_wall_time;
814	previous_interval = 0;
815
816	pbase = kvm_getprocs(kd, sel->thread ? KERN_PROC_ALL : KERN_PROC_PROC,
817	    0, &nproc);
818	(void)gettimeofday(&proc_wall_time, NULL);
819	if (clock_gettime(CLOCK_UPTIME, &proc_uptime) != 0)
820		memset(&proc_uptime, 0, sizeof(proc_uptime));
821	else if (previous_proc_uptime.tv_sec != 0 &&
822	    previous_proc_uptime.tv_nsec != 0) {
823		previous_interval = (proc_uptime.tv_sec -
824		    previous_proc_uptime.tv_sec) * 1000000;
825		nsec = proc_uptime.tv_nsec - previous_proc_uptime.tv_nsec;
826		if (nsec < 0) {
827			previous_interval -= 1000000;
828			nsec += 1000000000;
829		}
830		previous_interval += nsec / 1000;
831	}
832	if (nproc > onproc) {
833		pref = realloc(pref, sizeof(*pref) * nproc);
834		pcpu = realloc(pcpu, sizeof(*pcpu) * nproc);
835		onproc = nproc;
836	}
837	if (pref == NULL || pbase == NULL || pcpu == NULL) {
838		(void) fprintf(stderr, "top: Out of memory.\n");
839		quit(23);
840	}
841	/* get a pointer to the states summary array */
842	si->procstates = process_states;
843
844	/* set up flags which define what we are going to select */
845	show_idle = sel->idle;
846	show_jid = sel->jid != -1;
847	show_self = sel->self == -1;
848	show_system = sel->system;
849	show_uid = sel->uid != -1;
850	show_command = sel->command != NULL;
851	show_kidle = sel->kidle;
852
853	/* count up process states and get pointers to interesting procs */
854	total_procs = 0;
855	active_procs = 0;
856	total_inblock = 0;
857	total_oublock = 0;
858	total_majflt = 0;
859	memset((char *)process_states, 0, sizeof(process_states));
860	prefp = pref;
861	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
862
863		if (pp->ki_stat == 0)
864			/* not in use */
865			continue;
866
867		if (!show_self && pp->ki_pid == sel->self)
868			/* skip self */
869			continue;
870
871		if (!show_system && (pp->ki_flag & P_SYSTEM))
872			/* skip system process */
873			continue;
874
875		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt,
876		    &p_vcsw, &p_ivcsw);
877		total_inblock += p_inblock;
878		total_oublock += p_oublock;
879		total_majflt += p_majflt;
880		total_procs++;
881		process_states[pp->ki_stat]++;
882
883		if (pp->ki_stat == SZOMB)
884			/* skip zombies */
885			continue;
886
887		if (!show_kidle && pp->ki_tdflags & TDF_IDLETD)
888			/* skip kernel idle process */
889			continue;
890
891		PCTCPU(pp) = proc_calc_pctcpu(pp);
892		if (sel->thread && PCTCPU(pp) > 1.0)
893			PCTCPU(pp) = 1.0;
894		if (displaymode == DISP_CPU && !show_idle &&
895		    (!proc_used_cpu(pp) ||
896		     pp->ki_stat == SSTOP || pp->ki_stat == SIDL))
897			/* skip idle or non-running processes */
898			continue;
899
900		if (displaymode == DISP_IO && !show_idle && p_io == 0)
901			/* skip processes that aren't doing I/O */
902			continue;
903
904		if (show_jid && pp->ki_jid != sel->jid)
905			/* skip proc. that don't belong to the selected JID */
906			continue;
907
908		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
909			/* skip proc. that don't belong to the selected UID */
910			continue;
911
912		*prefp++ = pp;
913		active_procs++;
914	}
915
916	/* if requested, sort the "interesting" processes */
917	if (compare != NULL)
918		qsort(pref, active_procs, sizeof(*pref), compare);
919
920	/* remember active and total counts */
921	si->p_total = total_procs;
922	si->p_active = pref_len = active_procs;
923
924	/* pass back a handle */
925	handle.next_proc = pref;
926	handle.remaining = active_procs;
927	return ((caddr_t)&handle);
928}
929
930static char fmt[512];	/* static area where result is built */
931
932char *
933format_next_process(caddr_t handle, char *(*get_userid)(int), int flags)
934{
935	struct kinfo_proc *pp;
936	const struct kinfo_proc *oldp;
937	long cputime;
938	double pct;
939	struct handle *hp;
940	char status[16];
941	int cpu, state;
942	struct rusage ru, *rup;
943	long p_tot, s_tot;
944	char *proc_fmt, thr_buf[6];
945	char jid_buf[TOP_JID_LEN + 1], swap_buf[TOP_SWAP_LEN + 1];
946	char *cmdbuf = NULL;
947	char **args;
948	const int cmdlen = 128;
949
950	/* find and remember the next proc structure */
951	hp = (struct handle *)handle;
952	pp = *(hp->next_proc++);
953	hp->remaining--;
954
955	/* get the process's command name */
956	if ((pp->ki_flag & P_INMEM) == 0) {
957		/*
958		 * Print swapped processes as <pname>
959		 */
960		size_t len;
961
962		len = strlen(pp->ki_comm);
963		if (len > sizeof(pp->ki_comm) - 3)
964			len = sizeof(pp->ki_comm) - 3;
965		memmove(pp->ki_comm + 1, pp->ki_comm, len);
966		pp->ki_comm[0] = '<';
967		pp->ki_comm[len + 1] = '>';
968		pp->ki_comm[len + 2] = '\0';
969	}
970
971	/*
972	 * Convert the process's runtime from microseconds to seconds.  This
973	 * time includes the interrupt time although that is not wanted here.
974	 * ps(1) is similarly sloppy.
975	 */
976	cputime = (pp->ki_runtime + 500000) / 1000000;
977
978	/* calculate the base for cpu percentages */
979	pct = PCTCPU(pp);
980
981	/* generate "STATE" field */
982	switch (state = pp->ki_stat) {
983	case SRUN:
984		if (smpmode && pp->ki_oncpu != NOCPU)
985			sprintf(status, "CPU%d", pp->ki_oncpu);
986		else
987			strcpy(status, "RUN");
988		break;
989	case SLOCK:
990		if (pp->ki_kiflag & KI_LOCKBLOCK) {
991			sprintf(status, "*%.6s", pp->ki_lockname);
992			break;
993		}
994		/* fall through */
995	case SSLEEP:
996		if (pp->ki_wmesg != NULL) {
997			sprintf(status, "%.6s", pp->ki_wmesg);
998			break;
999		}
1000		/* FALLTHROUGH */
1001	default:
1002
1003		if (state >= 0 &&
1004		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
1005			sprintf(status, "%.6s", state_abbrev[state]);
1006		else
1007			sprintf(status, "?%5d", state);
1008		break;
1009	}
1010
1011	cmdbuf = (char *)malloc(cmdlen + 1);
1012	if (cmdbuf == NULL) {
1013		warn("malloc(%d)", cmdlen + 1);
1014		return NULL;
1015	}
1016
1017	if (!(flags & FMT_SHOWARGS)) {
1018		if (ps.thread && pp->ki_flag & P_HADTHREADS &&
1019		    pp->ki_tdname[0]) {
1020			snprintf(cmdbuf, cmdlen, "%s{%s%s}", pp->ki_comm,
1021			    pp->ki_tdname, pp->ki_moretdname);
1022		} else {
1023			snprintf(cmdbuf, cmdlen, "%s", pp->ki_comm);
1024		}
1025	} else {
1026		if (pp->ki_flag & P_SYSTEM ||
1027		    pp->ki_args == NULL ||
1028		    (args = kvm_getargv(kd, pp, cmdlen)) == NULL ||
1029		    !(*args)) {
1030			if (ps.thread && pp->ki_flag & P_HADTHREADS &&
1031		    	    pp->ki_tdname[0]) {
1032				snprintf(cmdbuf, cmdlen,
1033				    "[%s{%s%s}]", pp->ki_comm, pp->ki_tdname,
1034				    pp->ki_moretdname);
1035			} else {
1036				snprintf(cmdbuf, cmdlen,
1037				    "[%s]", pp->ki_comm);
1038			}
1039		} else {
1040			char *src, *dst, *argbuf;
1041			char *cmd;
1042			size_t argbuflen;
1043			size_t len;
1044
1045			argbuflen = cmdlen * 4;
1046			argbuf = (char *)malloc(argbuflen + 1);
1047			if (argbuf == NULL) {
1048				warn("malloc(%zu)", argbuflen + 1);
1049				free(cmdbuf);
1050				return NULL;
1051			}
1052
1053			dst = argbuf;
1054
1055			/* Extract cmd name from argv */
1056			cmd = strrchr(*args, '/');
1057			if (cmd == NULL)
1058				cmd = *args;
1059			else
1060				cmd++;
1061
1062			for (; (src = *args++) != NULL; ) {
1063				if (*src == '\0')
1064					continue;
1065				len = (argbuflen - (dst - argbuf) - 1) / 4;
1066				strvisx(dst, src,
1067				    MIN(strlen(src), len),
1068				    VIS_NL | VIS_CSTYLE);
1069				while (*dst != '\0')
1070					dst++;
1071				if ((argbuflen - (dst - argbuf) - 1) / 4 > 0)
1072					*dst++ = ' '; /* add delimiting space */
1073			}
1074			if (dst != argbuf && dst[-1] == ' ')
1075				dst--;
1076			*dst = '\0';
1077
1078			if (strcmp(cmd, pp->ki_comm) != 0) {
1079				if (ps.thread && pp->ki_flag & P_HADTHREADS &&
1080				    pp->ki_tdname[0])
1081					snprintf(cmdbuf, cmdlen,
1082					    "%s (%s){%s%s}", argbuf,
1083					    pp->ki_comm, pp->ki_tdname,
1084					    pp->ki_moretdname);
1085				else
1086					snprintf(cmdbuf, cmdlen,
1087					    "%s (%s)", argbuf, pp->ki_comm);
1088			} else {
1089				if (ps.thread && pp->ki_flag & P_HADTHREADS &&
1090				    pp->ki_tdname[0])
1091					snprintf(cmdbuf, cmdlen,
1092					    "%s{%s%s}", argbuf, pp->ki_tdname,
1093					    pp->ki_moretdname);
1094				else
1095					strlcpy(cmdbuf, argbuf, cmdlen);
1096			}
1097			free(argbuf);
1098		}
1099	}
1100
1101	if (ps.jail == 0)
1102		jid_buf[0] = '\0';
1103	else
1104		snprintf(jid_buf, sizeof(jid_buf), "%*d",
1105		    jidlength - 1, pp->ki_jid);
1106
1107	if (ps.swap == 0)
1108		swap_buf[0] = '\0';
1109	else
1110		snprintf(swap_buf, sizeof(swap_buf), "%*s",
1111		    swaplength - 1,
1112		    format_k2(pagetok(ki_swap(pp)))); /* XXX */
1113
1114	if (displaymode == DISP_IO) {
1115		oldp = get_old_proc(pp);
1116		if (oldp != NULL) {
1117			ru.ru_inblock = RU(pp)->ru_inblock -
1118			    RU(oldp)->ru_inblock;
1119			ru.ru_oublock = RU(pp)->ru_oublock -
1120			    RU(oldp)->ru_oublock;
1121			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
1122			ru.ru_nvcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
1123			ru.ru_nivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
1124			rup = &ru;
1125		} else {
1126			rup = RU(pp);
1127		}
1128		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
1129		s_tot = total_inblock + total_oublock + total_majflt;
1130
1131		snprintf(fmt, sizeof(fmt), io_Proc_format,
1132		    pp->ki_pid,
1133		    jidlength, jid_buf,
1134		    namelength, namelength, (*get_userid)(pp->ki_ruid),
1135		    rup->ru_nvcsw,
1136		    rup->ru_nivcsw,
1137		    rup->ru_inblock,
1138		    rup->ru_oublock,
1139		    rup->ru_majflt,
1140		    p_tot,
1141		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
1142		    screen_width > cmdlengthdelta ?
1143		    screen_width - cmdlengthdelta : 0,
1144		    printable(cmdbuf));
1145
1146		free(cmdbuf);
1147
1148		return (fmt);
1149	}
1150
1151	/* format this entry */
1152	if (smpmode) {
1153		if (state == SRUN && pp->ki_oncpu != NOCPU)
1154			cpu = pp->ki_oncpu;
1155		else
1156			cpu = pp->ki_lastcpu;
1157	} else
1158		cpu = 0;
1159	proc_fmt = smpmode ? smp_Proc_format : up_Proc_format;
1160	if (ps.thread != 0)
1161		thr_buf[0] = '\0';
1162	else
1163		snprintf(thr_buf, sizeof(thr_buf), "%*d ",
1164		    (int)(sizeof(thr_buf) - 2), pp->ki_numthreads);
1165
1166	snprintf(fmt, sizeof(fmt), proc_fmt,
1167	    pp->ki_pid,
1168	    jidlength, jid_buf,
1169	    namelength, namelength, (*get_userid)(pp->ki_ruid),
1170	    thr_buf,
1171	    pp->ki_pri.pri_level - PZERO,
1172	    format_nice(pp),
1173	    format_k2(PROCSIZE(pp)),
1174	    format_k2(pagetok(pp->ki_rssize)),
1175	    swaplength, swaplength, swap_buf,
1176	    status,
1177	    cpu,
1178	    format_time(cputime),
1179	    ps.wcpu ? 100.0 * weighted_cpu(pct, pp) : 100.0 * pct,
1180	    screen_width > cmdlengthdelta ? screen_width - cmdlengthdelta : 0,
1181	    printable(cmdbuf));
1182
1183	free(cmdbuf);
1184
1185	/* return the result */
1186	return (fmt);
1187}
1188
1189static void
1190getsysctl(const char *name, void *ptr, size_t len)
1191{
1192	size_t nlen = len;
1193
1194	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
1195		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
1196		    strerror(errno));
1197		quit(23);
1198	}
1199	if (nlen != len) {
1200		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n",
1201		    name, (unsigned long)len, (unsigned long)nlen);
1202		quit(23);
1203	}
1204}
1205
1206static const char *
1207format_nice(const struct kinfo_proc *pp)
1208{
1209	const char *fifo, *kproc;
1210	int rtpri;
1211	static char nicebuf[4 + 1];
1212
1213	fifo = PRI_NEED_RR(pp->ki_pri.pri_class) ? "" : "F";
1214	kproc = (pp->ki_flag & P_KPROC) ? "k" : "";
1215	switch (PRI_BASE(pp->ki_pri.pri_class)) {
1216	case PRI_ITHD:
1217		return ("-");
1218	case PRI_REALTIME:
1219		/*
1220		 * XXX: the kernel doesn't tell us the original rtprio and
1221		 * doesn't really know what it was, so to recover it we
1222		 * must be more chummy with the implementation than the
1223		 * implementation is with itself.  pri_user gives a
1224		 * constant "base" priority, but is only initialized
1225		 * properly for user threads.  pri_native gives what the
1226		 * kernel calls the "base" priority, but it isn't constant
1227		 * since it is changed by priority propagation.  pri_native
1228		 * also isn't properly initialized for all threads, but it
1229		 * is properly initialized for kernel realtime and idletime
1230		 * threads.  Thus we use pri_user for the base priority of
1231		 * user threads (it is always correct) and pri_native for
1232		 * the base priority of kernel realtime and idletime threads
1233		 * (there is nothing better, and it is usually correct).
1234		 *
1235		 * The field width and thus the buffer are too small for
1236		 * values like "kr31F", but such values shouldn't occur,
1237		 * and if they do then the tailing "F" is not displayed.
1238		 */
1239		rtpri = ((pp->ki_flag & P_KPROC) ? pp->ki_pri.pri_native :
1240		    pp->ki_pri.pri_user) - PRI_MIN_REALTIME;
1241		snprintf(nicebuf, sizeof(nicebuf), "%sr%d%s",
1242		    kproc, rtpri, fifo);
1243		break;
1244	case PRI_TIMESHARE:
1245		if (pp->ki_flag & P_KPROC)
1246			return ("-");
1247		snprintf(nicebuf, sizeof(nicebuf), "%d", pp->ki_nice - NZERO);
1248		break;
1249	case PRI_IDLE:
1250		/* XXX: as above. */
1251		rtpri = ((pp->ki_flag & P_KPROC) ? pp->ki_pri.pri_native :
1252		    pp->ki_pri.pri_user) - PRI_MIN_IDLE;
1253		snprintf(nicebuf, sizeof(nicebuf), "%si%d%s",
1254		    kproc, rtpri, fifo);
1255		break;
1256	default:
1257		return ("?");
1258	}
1259	return (nicebuf);
1260}
1261
1262/* comparison routines for qsort */
1263
1264static int
1265compare_pid(const void *p1, const void *p2)
1266{
1267	const struct kinfo_proc * const *pp1 = p1;
1268	const struct kinfo_proc * const *pp2 = p2;
1269
1270	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
1271		abort();
1272
1273	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
1274}
1275
1276static int
1277compare_tid(const void *p1, const void *p2)
1278{
1279	const struct kinfo_proc * const *pp1 = p1;
1280	const struct kinfo_proc * const *pp2 = p2;
1281
1282	if ((*pp2)->ki_tid < 0 || (*pp1)->ki_tid < 0)
1283		abort();
1284
1285	return ((*pp1)->ki_tid - (*pp2)->ki_tid);
1286}
1287
1288/*
1289 *  proc_compare - comparison function for "qsort"
1290 *	Compares the resource consumption of two processes using five
1291 *	distinct keys.  The keys (in descending order of importance) are:
1292 *	percent cpu, cpu ticks, state, resident set size, total virtual
1293 *	memory usage.  The process states are ordered as follows (from least
1294 *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
1295 *	array declaration below maps a process state index into a number
1296 *	that reflects this ordering.
1297 */
1298
1299static int sorted_state[] = {
1300	0,	/* not used		*/
1301	3,	/* sleep		*/
1302	1,	/* ABANDONED (WAIT)	*/
1303	6,	/* run			*/
1304	5,	/* start		*/
1305	2,	/* zombie		*/
1306	4	/* stop			*/
1307};
1308
1309
1310#define ORDERKEY_PCTCPU(a, b) do { \
1311	double diff; \
1312	if (ps.wcpu) \
1313		diff = weighted_cpu(PCTCPU((b)), (b)) - \
1314		    weighted_cpu(PCTCPU((a)), (a)); \
1315	else \
1316		diff = PCTCPU((b)) - PCTCPU((a)); \
1317	if (diff != 0) \
1318		return (diff > 0 ? 1 : -1); \
1319} while (0)
1320
1321#define ORDERKEY_CPTICKS(a, b) do { \
1322	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
1323	if (diff != 0) \
1324		return (diff > 0 ? 1 : -1); \
1325} while (0)
1326
1327#define ORDERKEY_STATE(a, b) do { \
1328	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
1329	if (diff != 0) \
1330		return (diff > 0 ? 1 : -1); \
1331} while (0)
1332
1333#define ORDERKEY_PRIO(a, b) do { \
1334	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
1335	if (diff != 0) \
1336		return (diff > 0 ? 1 : -1); \
1337} while (0)
1338
1339#define	ORDERKEY_THREADS(a, b) do { \
1340	int diff = (int)(b)->ki_numthreads - (int)(a)->ki_numthreads; \
1341	if (diff != 0) \
1342		return (diff > 0 ? 1 : -1); \
1343} while (0)
1344
1345#define ORDERKEY_RSSIZE(a, b) do { \
1346	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
1347	if (diff != 0) \
1348		return (diff > 0 ? 1 : -1); \
1349} while (0)
1350
1351#define ORDERKEY_MEM(a, b) do { \
1352	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
1353	if (diff != 0) \
1354		return (diff > 0 ? 1 : -1); \
1355} while (0)
1356
1357#define ORDERKEY_JID(a, b) do { \
1358	int diff = (int)(b)->ki_jid - (int)(a)->ki_jid; \
1359	if (diff != 0) \
1360		return (diff > 0 ? 1 : -1); \
1361} while (0)
1362
1363#define ORDERKEY_SWAP(a, b) do { \
1364	int diff = (int)ki_swap(b) - (int)ki_swap(a); \
1365	if (diff != 0) \
1366		return (diff > 0 ? 1 : -1); \
1367} while (0)
1368
1369/* compare_cpu - the comparison function for sorting by cpu percentage */
1370
1371int
1372#ifdef ORDER
1373compare_cpu(void *arg1, void *arg2)
1374#else
1375proc_compare(void *arg1, void *arg2)
1376#endif
1377{
1378	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1379	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1380
1381	ORDERKEY_PCTCPU(p1, p2);
1382	ORDERKEY_CPTICKS(p1, p2);
1383	ORDERKEY_STATE(p1, p2);
1384	ORDERKEY_PRIO(p1, p2);
1385	ORDERKEY_RSSIZE(p1, p2);
1386	ORDERKEY_MEM(p1, p2);
1387
1388	return (0);
1389}
1390
1391#ifdef ORDER
1392/* "cpu" compare routines */
1393int compare_size(), compare_res(), compare_time(), compare_prio(),
1394    compare_threads();
1395
1396/*
1397 * "io" compare routines.  Context switches aren't i/o, but are displayed
1398 * on the "io" display.
1399 */
1400int compare_iototal(), compare_ioread(), compare_iowrite(), compare_iofault(),
1401    compare_vcsw(), compare_ivcsw();
1402
1403int (*compares[])() = {
1404	compare_cpu,
1405	compare_size,
1406	compare_res,
1407	compare_time,
1408	compare_prio,
1409	compare_threads,
1410	compare_iototal,
1411	compare_ioread,
1412	compare_iowrite,
1413	compare_iofault,
1414	compare_vcsw,
1415	compare_ivcsw,
1416	compare_jid,
1417	compare_swap,
1418	NULL
1419};
1420
1421/* compare_size - the comparison function for sorting by total memory usage */
1422
1423int
1424compare_size(void *arg1, void *arg2)
1425{
1426	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1427	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1428
1429	ORDERKEY_MEM(p1, p2);
1430	ORDERKEY_RSSIZE(p1, p2);
1431	ORDERKEY_PCTCPU(p1, p2);
1432	ORDERKEY_CPTICKS(p1, p2);
1433	ORDERKEY_STATE(p1, p2);
1434	ORDERKEY_PRIO(p1, p2);
1435
1436	return (0);
1437}
1438
1439/* compare_res - the comparison function for sorting by resident set size */
1440
1441int
1442compare_res(void *arg1, void *arg2)
1443{
1444	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1445	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1446
1447	ORDERKEY_RSSIZE(p1, p2);
1448	ORDERKEY_MEM(p1, p2);
1449	ORDERKEY_PCTCPU(p1, p2);
1450	ORDERKEY_CPTICKS(p1, p2);
1451	ORDERKEY_STATE(p1, p2);
1452	ORDERKEY_PRIO(p1, p2);
1453
1454	return (0);
1455}
1456
1457/* compare_time - the comparison function for sorting by total cpu time */
1458
1459int
1460compare_time(void *arg1, void *arg2)
1461{
1462	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1463	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1464
1465	ORDERKEY_CPTICKS(p1, p2);
1466	ORDERKEY_PCTCPU(p1, p2);
1467	ORDERKEY_STATE(p1, p2);
1468	ORDERKEY_PRIO(p1, p2);
1469	ORDERKEY_RSSIZE(p1, p2);
1470	ORDERKEY_MEM(p1, p2);
1471
1472	return (0);
1473}
1474
1475/* compare_prio - the comparison function for sorting by priority */
1476
1477int
1478compare_prio(void *arg1, void *arg2)
1479{
1480	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1481	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1482
1483	ORDERKEY_PRIO(p1, p2);
1484	ORDERKEY_CPTICKS(p1, p2);
1485	ORDERKEY_PCTCPU(p1, p2);
1486	ORDERKEY_STATE(p1, p2);
1487	ORDERKEY_RSSIZE(p1, p2);
1488	ORDERKEY_MEM(p1, p2);
1489
1490	return (0);
1491}
1492
1493/* compare_threads - the comparison function for sorting by threads */
1494int
1495compare_threads(void *arg1, void *arg2)
1496{
1497	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1498	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1499
1500	ORDERKEY_THREADS(p1, p2);
1501	ORDERKEY_PCTCPU(p1, p2);
1502	ORDERKEY_CPTICKS(p1, p2);
1503	ORDERKEY_STATE(p1, p2);
1504	ORDERKEY_PRIO(p1, p2);
1505	ORDERKEY_RSSIZE(p1, p2);
1506	ORDERKEY_MEM(p1, p2);
1507
1508	return (0);
1509}
1510
1511/* compare_jid - the comparison function for sorting by jid */
1512static int
1513compare_jid(const void *arg1, const void *arg2)
1514{
1515	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1516	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1517
1518	ORDERKEY_JID(p1, p2);
1519	ORDERKEY_PCTCPU(p1, p2);
1520	ORDERKEY_CPTICKS(p1, p2);
1521	ORDERKEY_STATE(p1, p2);
1522	ORDERKEY_PRIO(p1, p2);
1523	ORDERKEY_RSSIZE(p1, p2);
1524	ORDERKEY_MEM(p1, p2);
1525
1526	return (0);
1527}
1528
1529/* compare_swap - the comparison function for sorting by swap */
1530static int
1531compare_swap(const void *arg1, const void *arg2)
1532{
1533	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1534	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1535
1536	ORDERKEY_SWAP(p1, p2);
1537	ORDERKEY_PCTCPU(p1, p2);
1538	ORDERKEY_CPTICKS(p1, p2);
1539	ORDERKEY_STATE(p1, p2);
1540	ORDERKEY_PRIO(p1, p2);
1541	ORDERKEY_RSSIZE(p1, p2);
1542	ORDERKEY_MEM(p1, p2);
1543
1544	return (0);
1545}
1546#endif /* ORDER */
1547
1548/* assorted comparison functions for sorting by i/o */
1549
1550int
1551#ifdef ORDER
1552compare_iototal(void *arg1, void *arg2)
1553#else
1554io_compare(void *arg1, void *arg2)
1555#endif
1556{
1557	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1558	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1559
1560	return (get_io_total(p2) - get_io_total(p1));
1561}
1562
1563#ifdef ORDER
1564int
1565compare_ioread(void *arg1, void *arg2)
1566{
1567	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1568	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1569	long dummy, inp1, inp2;
1570
1571	(void) get_io_stats(p1, &inp1, &dummy, &dummy, &dummy, &dummy);
1572	(void) get_io_stats(p2, &inp2, &dummy, &dummy, &dummy, &dummy);
1573
1574	return (inp2 - inp1);
1575}
1576
1577int
1578compare_iowrite(void *arg1, void *arg2)
1579{
1580	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1581	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1582	long dummy, oup1, oup2;
1583
1584	(void) get_io_stats(p1, &dummy, &oup1, &dummy, &dummy, &dummy);
1585	(void) get_io_stats(p2, &dummy, &oup2, &dummy, &dummy, &dummy);
1586
1587	return (oup2 - oup1);
1588}
1589
1590int
1591compare_iofault(void *arg1, void *arg2)
1592{
1593	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1594	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1595	long dummy, flp1, flp2;
1596
1597	(void) get_io_stats(p1, &dummy, &dummy, &flp1, &dummy, &dummy);
1598	(void) get_io_stats(p2, &dummy, &dummy, &flp2, &dummy, &dummy);
1599
1600	return (flp2 - flp1);
1601}
1602
1603int
1604compare_vcsw(void *arg1, void *arg2)
1605{
1606	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1607	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1608	long dummy, flp1, flp2;
1609
1610	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &flp1, &dummy);
1611	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &flp2, &dummy);
1612
1613	return (flp2 - flp1);
1614}
1615
1616int
1617compare_ivcsw(void *arg1, void *arg2)
1618{
1619	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1620	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1621	long dummy, flp1, flp2;
1622
1623	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &dummy, &flp1);
1624	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &dummy, &flp2);
1625
1626	return (flp2 - flp1);
1627}
1628#endif /* ORDER */
1629
1630/*
1631 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1632 *		the process does not exist.
1633 *		It is EXTREMELY IMPORTANT that this function work correctly.
1634 *		If top runs setuid root (as in SVR4), then this function
1635 *		is the only thing that stands in the way of a serious
1636 *		security problem.  It validates requests for the "kill"
1637 *		and "renice" commands.
1638 */
1639
1640int
1641proc_owner(int pid)
1642{
1643	int cnt;
1644	struct kinfo_proc **prefp;
1645	struct kinfo_proc *pp;
1646
1647	prefp = pref;
1648	cnt = pref_len;
1649	while (--cnt >= 0) {
1650		pp = *prefp++;
1651		if (pp->ki_pid == (pid_t)pid)
1652			return ((int)pp->ki_ruid);
1653	}
1654	return (-1);
1655}
1656
1657static int
1658swapmode(int *retavail, int *retfree)
1659{
1660	int n;
1661	int pagesize = getpagesize();
1662	struct kvm_swap swapary[1];
1663
1664	*retavail = 0;
1665	*retfree = 0;
1666
1667#define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1668
1669	n = kvm_getswapinfo(kd, swapary, 1, 0);
1670	if (n < 0 || swapary[0].ksw_total == 0)
1671		return (0);
1672
1673	*retavail = CONVERT(swapary[0].ksw_total);
1674	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1675
1676	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1677	return (n);
1678}
1679