machine.c revision 178116
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: head/usr.bin/top/machine.c 178116 2008-04-11 11:39:26Z ru $
24 */
25
26#include <sys/param.h>
27#include <sys/errno.h>
28#include <sys/file.h>
29#include <sys/proc.h>
30#include <sys/resource.h>
31#include <sys/rtprio.h>
32#include <sys/signal.h>
33#include <sys/sysctl.h>
34#include <sys/time.h>
35#include <sys/user.h>
36#include <sys/vmmeter.h>
37
38#include <kvm.h>
39#include <math.h>
40#include <nlist.h>
41#include <paths.h>
42#include <pwd.h>
43#include <stdio.h>
44#include <stdlib.h>
45#include <string.h>
46#include <strings.h>
47#include <unistd.h>
48#include <vis.h>
49
50#include "top.h"
51#include "machine.h"
52#include "screen.h"
53#include "utils.h"
54#include "layout.h"
55
56#define GETSYSCTL(name, var) getsysctl(name, &(var), sizeof(var))
57#define	SMPUNAMELEN	13
58#define	UPUNAMELEN	15
59
60extern struct process_select ps;
61extern char* printable(char *);
62static int smpmode;
63enum displaymodes displaymode;
64#ifdef TOP_USERNAME_LEN
65static int namelength = TOP_USERNAME_LEN;
66#else
67static int namelength = 8;
68#endif
69static int cmdlengthdelta;
70
71/* Prototypes for top internals */
72void quit(int);
73
74/* get_process_info passes back a handle.  This is what it looks like: */
75
76struct handle {
77	struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
78	int remaining;			/* number of pointers remaining */
79};
80
81/* declarations for load_avg */
82#include "loadavg.h"
83
84/* define what weighted cpu is.  */
85#define weighted_cpu(pct, pp) ((pp)->ki_swtime == 0 ? 0.0 : \
86			 ((pct) / (1.0 - exp((pp)->ki_swtime * logcpu))))
87
88/* what we consider to be process size: */
89#define PROCSIZE(pp) ((pp)->ki_size / 1024)
90
91#define RU(pp)	(&(pp)->ki_rusage)
92#define RUTOT(pp) \
93	(RU(pp)->ru_inblock + RU(pp)->ru_oublock + RU(pp)->ru_majflt)
94
95
96/* definitions for indices in the nlist array */
97
98/*
99 *  These definitions control the format of the per-process area
100 */
101
102static char io_header[] =
103    "  PID%s %-*.*s   VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
104
105#define io_Proc_format \
106    "%5d%s %-*.*s %6ld %6ld %6ld %6ld %6ld %6ld %6.2f%% %.*s"
107
108static char smp_header_thr[] =
109    "  PID%s %-*.*s  THR PRI NICE   SIZE    RES STATE  C   TIME %6s COMMAND";
110static char smp_header[] =
111    "  PID%s %-*.*s "   "PRI NICE   SIZE    RES STATE  C   TIME %6s COMMAND";
112
113#define smp_Proc_format \
114    "%5d%s %-*.*s %s%3d %4s%7s %6s %-6.6s %1x%7s %5.2f%% %.*s"
115
116static char up_header_thr[] =
117    "  PID%s %-*.*s  THR PRI NICE   SIZE    RES STATE    TIME %6s COMMAND";
118static char up_header[] =
119    "  PID%s %-*.*s "   "PRI NICE   SIZE    RES STATE    TIME %6s COMMAND";
120
121#define up_Proc_format \
122    "%5d%s %-*.*s %s%3d %4s%7s %6s %-6.6s%.0d%7s %5.2f%% %.*s"
123
124
125/* process state names for the "STATE" column of the display */
126/* the extra nulls in the string "run" are for adding a slash and
127   the processor number when needed */
128
129char *state_abbrev[] = {
130	"", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB", "WAIT", "LOCK"
131};
132
133
134static kvm_t *kd;
135
136/* values that we stash away in _init and use in later routines */
137
138static double logcpu;
139
140/* these are retrieved from the kernel in _init */
141
142static load_avg  ccpu;
143
144/* these are used in the get_ functions */
145
146static int lastpid;
147
148/* these are for calculating cpu state percentages */
149
150static long cp_time[CPUSTATES];
151static long cp_old[CPUSTATES];
152static long cp_diff[CPUSTATES];
153
154/* these are for detailing the process states */
155
156int process_states[8];
157char *procstatenames[] = {
158	"", " starting, ", " running, ", " sleeping, ", " stopped, ",
159	" zombie, ", " waiting, ", " lock, ",
160	NULL
161};
162
163/* these are for detailing the cpu states */
164
165int cpu_states[CPUSTATES];
166char *cpustatenames[] = {
167	"user", "nice", "system", "interrupt", "idle", NULL
168};
169
170/* these are for detailing the memory statistics */
171
172int memory_stats[7];
173char *memorynames[] = {
174	"K Active, ", "K Inact, ", "K Wired, ", "K Cache, ", "K Buf, ",
175	"K Free", NULL
176};
177
178int swap_stats[7];
179char *swapnames[] = {
180	"K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
181	NULL
182};
183
184
185/* these are for keeping track of the proc array */
186
187static int nproc;
188static int onproc = -1;
189static int pref_len;
190static struct kinfo_proc *pbase;
191static struct kinfo_proc **pref;
192static struct kinfo_proc *previous_procs;
193static struct kinfo_proc **previous_pref;
194static int previous_proc_count = 0;
195static int previous_proc_count_max = 0;
196
197/* total number of io operations */
198static long total_inblock;
199static long total_oublock;
200static long total_majflt;
201
202/* these are for getting the memory statistics */
203
204static int pageshift;		/* log base 2 of the pagesize */
205
206/* define pagetok in terms of pageshift */
207
208#define pagetok(size) ((size) << pageshift)
209
210/* useful externals */
211long percentages();
212
213#ifdef ORDER
214/*
215 * Sorting orders.  The first element is the default.
216 */
217char *ordernames[] = {
218	"cpu", "size", "res", "time", "pri", "threads",
219	"total", "read", "write", "fault", "vcsw", "ivcsw",
220	"jid", NULL
221};
222#endif
223
224/* Per-cpu time states */
225static int maxcpu;
226static int maxid;
227static int ncpus;
228static u_long cpumask;
229static long *times;
230static long *pcpu_cp_time;
231static long *pcpu_cp_old;
232static long *pcpu_cp_diff;
233static int *pcpu_cpu_states;
234
235static int compare_jid(const void *a, const void *b);
236static int compare_pid(const void *a, const void *b);
237static const char *format_nice(const struct kinfo_proc *pp);
238static void getsysctl(const char *name, void *ptr, size_t len);
239static int swapmode(int *retavail, int *retfree);
240
241int
242machine_init(struct statics *statics, char do_unames)
243{
244	int pagesize;
245	size_t modelen;
246	struct passwd *pw;
247
248	modelen = sizeof(smpmode);
249	if ((sysctlbyname("machdep.smp_active", &smpmode, &modelen,
250	    NULL, 0) != 0 &&
251	    sysctlbyname("kern.smp.active", &smpmode, &modelen,
252	    NULL, 0) != 0) ||
253	    modelen != sizeof(smpmode))
254		smpmode = 0;
255
256	if (do_unames) {
257	    while ((pw = getpwent()) != NULL) {
258		if (strlen(pw->pw_name) > namelength)
259			namelength = strlen(pw->pw_name);
260	    }
261	}
262	if (smpmode && namelength > SMPUNAMELEN)
263		namelength = SMPUNAMELEN;
264	else if (namelength > UPUNAMELEN)
265		namelength = UPUNAMELEN;
266
267	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
268	if (kd == NULL)
269		return (-1);
270
271	GETSYSCTL("kern.ccpu", ccpu);
272
273	/* this is used in calculating WCPU -- calculate it ahead of time */
274	logcpu = log(loaddouble(ccpu));
275
276	pbase = NULL;
277	pref = NULL;
278	nproc = 0;
279	onproc = -1;
280
281	/* get the page size and calculate pageshift from it */
282	pagesize = getpagesize();
283	pageshift = 0;
284	while (pagesize > 1) {
285		pageshift++;
286		pagesize >>= 1;
287	}
288
289	/* we only need the amount of log(2)1024 for our conversion */
290	pageshift -= LOG1024;
291
292	/* fill in the statics information */
293	statics->procstate_names = procstatenames;
294	statics->cpustate_names = cpustatenames;
295	statics->memory_names = memorynames;
296	statics->swap_names = swapnames;
297#ifdef ORDER
298	statics->order_names = ordernames;
299#endif
300
301	/* Adjust display based on ncpus */
302	if (pcpu_stats) {
303		int i, j, empty;
304		size_t size;
305
306		cpumask = 0;
307		ncpus = 0;
308		GETSYSCTL("kern.smp.maxcpus", maxcpu);
309		size = sizeof(long) * maxcpu * CPUSTATES;
310		times = malloc(size);
311		if (times == NULL)
312			err(1, "malloc %zd bytes", size);
313		if (sysctlbyname("kern.cp_times", times, &size, NULL, 0) == -1)
314			err(1, "sysctlbyname kern.cp_times");
315		pcpu_cp_time = calloc(1, size);
316		maxid = (size / CPUSTATES / sizeof(long)) - 1;
317		for (i = 0; i <= maxid; i++) {
318			empty = 1;
319			for (j = 0; empty && j < CPUSTATES; j++) {
320				if (times[i * CPUSTATES + j] != 0)
321					empty = 0;
322			}
323			if (!empty) {
324				cpumask |= (1ul << i);
325				ncpus++;
326			}
327		}
328
329		if (ncpus > 1) {
330			y_mem += ncpus - 1;	/* 3 */
331			y_swap += ncpus - 1;	/* 4 */
332			y_idlecursor += ncpus - 1; /* 5 */
333			y_message += ncpus - 1;	/* 5 */
334			y_header += ncpus - 1;	/* 6 */
335			y_procs += ncpus - 1;	/* 7 */
336			Header_lines += ncpus - 1; /* 7 */
337		}
338		size = sizeof(long) * ncpus * CPUSTATES;
339		pcpu_cp_old = calloc(1, size);
340		pcpu_cp_diff = calloc(1, size);
341		pcpu_cpu_states = calloc(1, size);
342		statics->ncpus = ncpus;
343	} else {
344		statics->ncpus = 1;
345	}
346
347	/* all done! */
348	return (0);
349}
350
351char *
352format_header(char *uname_field)
353{
354	static char Header[128];
355	const char *prehead;
356
357	switch (displaymode) {
358	case DISP_CPU:
359		/*
360		 * The logic of picking the right header format seems reverse
361		 * here because we only want to display a THR column when
362		 * "thread mode" is off (and threads are not listed as
363		 * separate lines).
364		 */
365		prehead = smpmode ?
366		    (ps.thread ? smp_header : smp_header_thr) :
367		    (ps.thread ? up_header : up_header_thr);
368		snprintf(Header, sizeof(Header), prehead,
369		    ps.jail ? " JID" : "",
370		    namelength, namelength, uname_field,
371		    ps.wcpu ? "WCPU" : "CPU");
372		break;
373	case DISP_IO:
374		prehead = io_header;
375		snprintf(Header, sizeof(Header), prehead,
376		    ps.jail ? " JID" : "",
377		    namelength, namelength, uname_field);
378		break;
379	}
380	cmdlengthdelta = strlen(Header) - 7;
381	return (Header);
382}
383
384static int swappgsin = -1;
385static int swappgsout = -1;
386extern struct timeval timeout;
387
388
389void
390get_system_info(struct system_info *si)
391{
392	long total;
393	struct loadavg sysload;
394	int mib[2];
395	struct timeval boottime;
396	size_t bt_size;
397	int i, j;
398	size_t size;
399
400	/* get the cp_time array */
401	if (pcpu_stats) {
402		size = (maxid + 1) * CPUSTATES * sizeof(long);
403		if (sysctlbyname("kern.cp_times", pcpu_cp_time, &size, NULL, 0) == -1)
404			err(1, "sysctlbyname kern.cp_times");
405	} else {
406		GETSYSCTL("kern.cp_time", cp_time);
407	}
408	GETSYSCTL("vm.loadavg", sysload);
409	GETSYSCTL("kern.lastpid", lastpid);
410
411	/* convert load averages to doubles */
412	for (i = 0; i < 3; i++)
413		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
414
415	if (pcpu_stats) {
416		for (i = j = 0; i <= maxid; i++) {
417			if ((cpumask & (1ul << i)) == 0)
418				continue;
419			/* convert cp_time counts to percentages */
420			percentages(CPUSTATES, &pcpu_cpu_states[j * CPUSTATES],
421			    &pcpu_cp_time[j * CPUSTATES],
422			    &pcpu_cp_old[j * CPUSTATES],
423			    &pcpu_cp_diff[j * CPUSTATES]);
424			j++;
425		}
426	} else {
427		/* convert cp_time counts to percentages */
428		percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
429	}
430
431	/* sum memory & swap statistics */
432	{
433		static unsigned int swap_delay = 0;
434		static int swapavail = 0;
435		static int swapfree = 0;
436		static int bufspace = 0;
437		static int nspgsin, nspgsout;
438
439		GETSYSCTL("vfs.bufspace", bufspace);
440		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
441		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
442		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[2]);
443		GETSYSCTL("vm.stats.vm.v_cache_count", memory_stats[3]);
444		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
445		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
446		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
447		/* convert memory stats to Kbytes */
448		memory_stats[0] = pagetok(memory_stats[0]);
449		memory_stats[1] = pagetok(memory_stats[1]);
450		memory_stats[2] = pagetok(memory_stats[2]);
451		memory_stats[3] = pagetok(memory_stats[3]);
452		memory_stats[4] = bufspace / 1024;
453		memory_stats[5] = pagetok(memory_stats[5]);
454		memory_stats[6] = -1;
455
456		/* first interval */
457		if (swappgsin < 0) {
458			swap_stats[4] = 0;
459			swap_stats[5] = 0;
460		}
461
462		/* compute differences between old and new swap statistic */
463		else {
464			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
465			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
466		}
467
468		swappgsin = nspgsin;
469		swappgsout = nspgsout;
470
471		/* call CPU heavy swapmode() only for changes */
472		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
473			swap_stats[3] = swapmode(&swapavail, &swapfree);
474			swap_stats[0] = swapavail;
475			swap_stats[1] = swapavail - swapfree;
476			swap_stats[2] = swapfree;
477		}
478		swap_delay = 1;
479		swap_stats[6] = -1;
480	}
481
482	/* set arrays and strings */
483	if (pcpu_stats) {
484		si->cpustates = pcpu_cpu_states;
485		si->ncpus = ncpus;
486	} else {
487		si->cpustates = cpu_states;
488		si->ncpus = 1;
489	}
490	si->memory = memory_stats;
491	si->swap = swap_stats;
492
493
494	if (lastpid > 0) {
495		si->last_pid = lastpid;
496	} else {
497		si->last_pid = -1;
498	}
499
500	/*
501	 * Print how long system has been up.
502	 * (Found by looking getting "boottime" from the kernel)
503	 */
504	mib[0] = CTL_KERN;
505	mib[1] = KERN_BOOTTIME;
506	bt_size = sizeof(boottime);
507	if (sysctl(mib, 2, &boottime, &bt_size, NULL, 0) != -1 &&
508	    boottime.tv_sec != 0) {
509		si->boottime = boottime;
510	} else {
511		si->boottime.tv_sec = -1;
512	}
513}
514
515#define NOPROC	((void *)-1)
516
517/*
518 * We need to compare data from the old process entry with the new
519 * process entry.
520 * To facilitate doing this quickly we stash a pointer in the kinfo_proc
521 * structure to cache the mapping.  We also use a negative cache pointer
522 * of NOPROC to avoid duplicate lookups.
523 * XXX: this could be done when the actual processes are fetched, we do
524 * it here out of laziness.
525 */
526const struct kinfo_proc *
527get_old_proc(struct kinfo_proc *pp)
528{
529	struct kinfo_proc **oldpp, *oldp;
530
531	/*
532	 * If this is the first fetch of the kinfo_procs then we don't have
533	 * any previous entries.
534	 */
535	if (previous_proc_count == 0)
536		return (NULL);
537	/* negative cache? */
538	if (pp->ki_udata == NOPROC)
539		return (NULL);
540	/* cached? */
541	if (pp->ki_udata != NULL)
542		return (pp->ki_udata);
543	/*
544	 * Not cached,
545	 * 1) look up based on pid.
546	 * 2) compare process start.
547	 * If we fail here, then setup a negative cache entry, otherwise
548	 * cache it.
549	 */
550	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
551	    sizeof(*previous_pref), compare_pid);
552	if (oldpp == NULL) {
553		pp->ki_udata = NOPROC;
554		return (NULL);
555	}
556	oldp = *oldpp;
557	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
558		pp->ki_udata = NOPROC;
559		return (NULL);
560	}
561	pp->ki_udata = oldp;
562	return (oldp);
563}
564
565/*
566 * Return the total amount of IO done in blocks in/out and faults.
567 * store the values individually in the pointers passed in.
568 */
569long
570get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp,
571    long *vcsw, long *ivcsw)
572{
573	const struct kinfo_proc *oldp;
574	static struct kinfo_proc dummy;
575	long ret;
576
577	oldp = get_old_proc(pp);
578	if (oldp == NULL) {
579		bzero(&dummy, sizeof(dummy));
580		oldp = &dummy;
581	}
582	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
583	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
584	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
585	*vcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
586	*ivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
587	ret =
588	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
589	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
590	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
591	return (ret);
592}
593
594/*
595 * Return the total number of block in/out and faults by a process.
596 */
597long
598get_io_total(struct kinfo_proc *pp)
599{
600	long dummy;
601
602	return (get_io_stats(pp, &dummy, &dummy, &dummy, &dummy, &dummy));
603}
604
605static struct handle handle;
606
607caddr_t
608get_process_info(struct system_info *si, struct process_select *sel,
609    int (*compare)(const void *, const void *))
610{
611	int i;
612	int total_procs;
613	long p_io;
614	long p_inblock, p_oublock, p_majflt, p_vcsw, p_ivcsw;
615	int active_procs;
616	struct kinfo_proc **prefp;
617	struct kinfo_proc *pp;
618	struct kinfo_proc *prev_pp = NULL;
619
620	/* these are copied out of sel for speed */
621	int show_idle;
622	int show_self;
623	int show_system;
624	int show_uid;
625	int show_command;
626
627	/*
628	 * Save the previous process info.
629	 */
630	if (previous_proc_count_max < nproc) {
631		free(previous_procs);
632		previous_procs = malloc(nproc * sizeof(*previous_procs));
633		free(previous_pref);
634		previous_pref = malloc(nproc * sizeof(*previous_pref));
635		if (previous_procs == NULL || previous_pref == NULL) {
636			(void) fprintf(stderr, "top: Out of memory.\n");
637			quit(23);
638		}
639		previous_proc_count_max = nproc;
640	}
641	if (nproc) {
642		for (i = 0; i < nproc; i++)
643			previous_pref[i] = &previous_procs[i];
644		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
645		qsort(previous_pref, nproc, sizeof(*previous_pref),
646		    compare_pid);
647	}
648	previous_proc_count = nproc;
649
650	pbase = kvm_getprocs(kd, KERN_PROC_ALL, 0, &nproc);
651	if (nproc > onproc)
652		pref = realloc(pref, sizeof(*pref) * (onproc = nproc));
653	if (pref == NULL || pbase == NULL) {
654		(void) fprintf(stderr, "top: Out of memory.\n");
655		quit(23);
656	}
657	/* get a pointer to the states summary array */
658	si->procstates = process_states;
659
660	/* set up flags which define what we are going to select */
661	show_idle = sel->idle;
662	show_self = sel->self == -1;
663	show_system = sel->system;
664	show_uid = sel->uid != -1;
665	show_command = sel->command != NULL;
666
667	/* count up process states and get pointers to interesting procs */
668	total_procs = 0;
669	active_procs = 0;
670	total_inblock = 0;
671	total_oublock = 0;
672	total_majflt = 0;
673	memset((char *)process_states, 0, sizeof(process_states));
674	prefp = pref;
675	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
676
677		if (pp->ki_stat == 0)
678			/* not in use */
679			continue;
680
681		if (!show_self && pp->ki_pid == sel->self)
682			/* skip self */
683			continue;
684
685		if (!show_system && (pp->ki_flag & P_SYSTEM))
686			/* skip system process */
687			continue;
688
689		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt,
690		    &p_vcsw, &p_ivcsw);
691		total_inblock += p_inblock;
692		total_oublock += p_oublock;
693		total_majflt += p_majflt;
694		total_procs++;
695		process_states[pp->ki_stat]++;
696
697		if (pp->ki_stat == SZOMB)
698			/* skip zombies */
699			continue;
700
701		if (displaymode == DISP_CPU && !show_idle &&
702		    (pp->ki_pctcpu == 0 ||
703		     pp->ki_stat == SSTOP || pp->ki_stat == SIDL))
704			/* skip idle or non-running processes */
705			continue;
706
707		if (displaymode == DISP_IO && !show_idle && p_io == 0)
708			/* skip processes that aren't doing I/O */
709			continue;
710
711		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
712			/* skip proc. that don't belong to the selected UID */
713			continue;
714
715		/*
716		 * When not showing threads, take the first thread
717		 * for output and add the fields that we can from
718		 * the rest of the process's threads rather than
719		 * using the system's mostly-broken KERN_PROC_PROC.
720		 */
721		if (sel->thread || prev_pp == NULL ||
722		    prev_pp->ki_pid != pp->ki_pid) {
723			*prefp++ = pp;
724			active_procs++;
725			prev_pp = pp;
726		} else {
727			prev_pp->ki_pctcpu += pp->ki_pctcpu;
728		}
729	}
730
731	/* if requested, sort the "interesting" processes */
732	if (compare != NULL)
733		qsort(pref, active_procs, sizeof(*pref), compare);
734
735	/* remember active and total counts */
736	si->p_total = total_procs;
737	si->p_active = pref_len = active_procs;
738
739	/* pass back a handle */
740	handle.next_proc = pref;
741	handle.remaining = active_procs;
742	return ((caddr_t)&handle);
743}
744
745static char fmt[128];	/* static area where result is built */
746
747char *
748format_next_process(caddr_t handle, char *(*get_userid)(int), int flags)
749{
750	struct kinfo_proc *pp;
751	const struct kinfo_proc *oldp;
752	long cputime;
753	double pct;
754	struct handle *hp;
755	char status[16];
756	int state;
757	struct rusage ru, *rup;
758	long p_tot, s_tot;
759	char *proc_fmt, thr_buf[6], jid_buf[6];
760	char *cmdbuf = NULL;
761	char **args;
762
763	/* find and remember the next proc structure */
764	hp = (struct handle *)handle;
765	pp = *(hp->next_proc++);
766	hp->remaining--;
767
768	/* get the process's command name */
769	if ((pp->ki_flag & P_INMEM) == 0) {
770		/*
771		 * Print swapped processes as <pname>
772		 */
773		size_t len;
774
775		len = strlen(pp->ki_comm);
776		if (len > sizeof(pp->ki_comm) - 3)
777			len = sizeof(pp->ki_comm) - 3;
778		memmove(pp->ki_comm + 1, pp->ki_comm, len);
779		pp->ki_comm[0] = '<';
780		pp->ki_comm[len + 1] = '>';
781		pp->ki_comm[len + 2] = '\0';
782	}
783
784	/*
785	 * Convert the process's runtime from microseconds to seconds.  This
786	 * time includes the interrupt time although that is not wanted here.
787	 * ps(1) is similarly sloppy.
788	 */
789	cputime = (pp->ki_runtime + 500000) / 1000000;
790
791	/* calculate the base for cpu percentages */
792	pct = pctdouble(pp->ki_pctcpu);
793
794	/* generate "STATE" field */
795	switch (state = pp->ki_stat) {
796	case SRUN:
797		if (smpmode && pp->ki_oncpu != 0xff)
798			sprintf(status, "CPU%d", pp->ki_oncpu);
799		else
800			strcpy(status, "RUN");
801		break;
802	case SLOCK:
803		if (pp->ki_kiflag & KI_LOCKBLOCK) {
804			sprintf(status, "*%.6s", pp->ki_lockname);
805			break;
806		}
807		/* fall through */
808	case SSLEEP:
809		if (pp->ki_wmesg != NULL) {
810			sprintf(status, "%.6s", pp->ki_wmesg);
811			break;
812		}
813		/* FALLTHROUGH */
814	default:
815
816		if (state >= 0 &&
817		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
818			sprintf(status, "%.6s", state_abbrev[state]);
819		else
820			sprintf(status, "?%5d", state);
821		break;
822	}
823
824	cmdbuf = (char *)malloc(cmdlengthdelta + 1);
825	if (cmdbuf == NULL) {
826		warn("malloc(%d)", cmdlengthdelta + 1);
827		return NULL;
828	}
829
830	if (!(flags & FMT_SHOWARGS)) {
831		if (ps.thread && pp->ki_flag & P_HADTHREADS &&
832		    pp->ki_ocomm[0]) {
833			snprintf(cmdbuf, cmdlengthdelta, "{%s}", pp->ki_ocomm);
834		} else {
835			snprintf(cmdbuf, cmdlengthdelta, "%s", pp->ki_comm);
836		}
837	} else {
838		if (pp->ki_flag & P_SYSTEM ||
839		    pp->ki_args == NULL ||
840		    (args = kvm_getargv(kd, pp, cmdlengthdelta)) == NULL ||
841		    !(*args)) {
842			if (ps.thread && pp->ki_flag & P_HADTHREADS &&
843		    	pp->ki_ocomm[0]) {
844				snprintf(cmdbuf, cmdlengthdelta,
845				    "{%s}", pp->ki_ocomm);
846			} else {
847				snprintf(cmdbuf, cmdlengthdelta,
848				    "[%s]", pp->ki_comm);
849			}
850		} else {
851			char *src, *dst, *argbuf;
852			char *cmd;
853			size_t argbuflen;
854			size_t len;
855
856			argbuflen = cmdlengthdelta * 4;
857			argbuf = (char *)malloc(argbuflen + 1);
858			if (argbuf == NULL) {
859				warn("malloc(%d)", argbuflen + 1);
860				free(cmdbuf);
861				return NULL;
862			}
863
864			dst = argbuf;
865
866			/* Extract cmd name from argv */
867			cmd = strrchr(*args, '/');
868			if (cmd == NULL)
869				cmd = *args;
870			else
871				cmd++;
872
873			for (; (src = *args++) != NULL; ) {
874				if (*src == '\0')
875					continue;
876				len = (argbuflen - (dst - argbuf) - 1) / 4;
877				strvisx(dst, src,
878				    strlen(src) < len ? strlen(src) : len,
879				    VIS_NL | VIS_CSTYLE);
880				while (*dst != '\0')
881					dst++;
882				if ((argbuflen - (dst - argbuf) - 1) / 4 > 0)
883					*dst++ = ' '; /* add delimiting space */
884			}
885			if (dst != argbuf && dst[-1] == ' ')
886				dst--;
887			*dst = '\0';
888
889			if (strcmp(cmd, pp->ki_comm) != 0 )
890				snprintf(cmdbuf, cmdlengthdelta,
891				    "%s (%s)",argbuf,  pp->ki_comm);
892			else
893				strlcpy(cmdbuf, argbuf, cmdlengthdelta);
894
895			free(argbuf);
896		}
897	}
898
899	if (ps.jail == 0)
900		jid_buf[0] = '\0';
901	else
902		snprintf(jid_buf, sizeof(jid_buf), " %*d",
903		    sizeof(jid_buf) - 3, pp->ki_jid);
904
905	if (displaymode == DISP_IO) {
906		oldp = get_old_proc(pp);
907		if (oldp != NULL) {
908			ru.ru_inblock = RU(pp)->ru_inblock -
909			    RU(oldp)->ru_inblock;
910			ru.ru_oublock = RU(pp)->ru_oublock -
911			    RU(oldp)->ru_oublock;
912			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
913			ru.ru_nvcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
914			ru.ru_nivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
915			rup = &ru;
916		} else {
917			rup = RU(pp);
918		}
919		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
920		s_tot = total_inblock + total_oublock + total_majflt;
921
922		sprintf(fmt, io_Proc_format,
923		    pp->ki_pid,
924		    jid_buf,
925		    namelength, namelength, (*get_userid)(pp->ki_ruid),
926		    rup->ru_nvcsw,
927		    rup->ru_nivcsw,
928		    rup->ru_inblock,
929		    rup->ru_oublock,
930		    rup->ru_majflt,
931		    p_tot,
932		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
933		    screen_width > cmdlengthdelta ?
934		    screen_width - cmdlengthdelta : 0,
935		    printable(cmdbuf));
936
937		free(cmdbuf);
938
939		return (fmt);
940	}
941
942	/* format this entry */
943	proc_fmt = smpmode ? smp_Proc_format : up_Proc_format;
944	if (ps.thread != 0)
945		thr_buf[0] = '\0';
946	else
947		snprintf(thr_buf, sizeof(thr_buf), "%*d ",
948		    sizeof(thr_buf) - 2, pp->ki_numthreads);
949
950	sprintf(fmt, proc_fmt,
951	    pp->ki_pid,
952	    jid_buf,
953	    namelength, namelength, (*get_userid)(pp->ki_ruid),
954	    thr_buf,
955	    pp->ki_pri.pri_level - PZERO,
956	    format_nice(pp),
957	    format_k2(PROCSIZE(pp)),
958	    format_k2(pagetok(pp->ki_rssize)),
959	    status,
960	    smpmode ? pp->ki_lastcpu : 0,
961	    format_time(cputime),
962	    ps.wcpu ? 100.0 * weighted_cpu(pct, pp) : 100.0 * pct,
963	    screen_width > cmdlengthdelta ? screen_width - cmdlengthdelta : 0,
964	    printable(cmdbuf));
965
966	free(cmdbuf);
967
968	/* return the result */
969	return (fmt);
970}
971
972static void
973getsysctl(const char *name, void *ptr, size_t len)
974{
975	size_t nlen = len;
976
977	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
978		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
979		    strerror(errno));
980		quit(23);
981	}
982	if (nlen != len) {
983		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n",
984		    name, (unsigned long)len, (unsigned long)nlen);
985		quit(23);
986	}
987}
988
989static const char *
990format_nice(const struct kinfo_proc *pp)
991{
992	const char *fifo, *kthread;
993	int rtpri;
994	static char nicebuf[4 + 1];
995
996	fifo = PRI_NEED_RR(pp->ki_pri.pri_class) ? "" : "F";
997	kthread = (pp->ki_flag & P_KTHREAD) ? "k" : "";
998	switch (PRI_BASE(pp->ki_pri.pri_class)) {
999	case PRI_ITHD:
1000		return ("-");
1001	case PRI_REALTIME:
1002		/*
1003		 * XXX: the kernel doesn't tell us the original rtprio and
1004		 * doesn't really know what it was, so to recover it we
1005		 * must be more chummy with the implementation than the
1006		 * implementation is with itself.  pri_user gives a
1007		 * constant "base" priority, but is only initialized
1008		 * properly for user threads.  pri_native gives what the
1009		 * kernel calls the "base" priority, but it isn't constant
1010		 * since it is changed by priority propagation.  pri_native
1011		 * also isn't properly initialized for all threads, but it
1012		 * is properly initialized for kernel realtime and idletime
1013		 * threads.  Thus we use pri_user for the base priority of
1014		 * user threads (it is always correct) and pri_native for
1015		 * the base priority of kernel realtime and idletime threads
1016		 * (there is nothing better, and it is usually correct).
1017		 *
1018		 * The field width and thus the buffer are too small for
1019		 * values like "kr31F", but such values shouldn't occur,
1020		 * and if they do then the tailing "F" is not displayed.
1021		 */
1022		rtpri = ((pp->ki_flag & P_KTHREAD) ? pp->ki_pri.pri_native :
1023		    pp->ki_pri.pri_user) - PRI_MIN_REALTIME;
1024		snprintf(nicebuf, sizeof(nicebuf), "%sr%d%s",
1025		    kthread, rtpri, fifo);
1026		break;
1027	case PRI_TIMESHARE:
1028		if (pp->ki_flag & P_KTHREAD)
1029			return ("-");
1030		snprintf(nicebuf, sizeof(nicebuf), "%d", pp->ki_nice - NZERO);
1031		break;
1032	case PRI_IDLE:
1033		/* XXX: as above. */
1034		rtpri = ((pp->ki_flag & P_KTHREAD) ? pp->ki_pri.pri_native :
1035		    pp->ki_pri.pri_user) - PRI_MIN_IDLE;
1036		snprintf(nicebuf, sizeof(nicebuf), "%si%d%s",
1037		    kthread, rtpri, fifo);
1038		break;
1039	default:
1040		return ("?");
1041	}
1042	return (nicebuf);
1043}
1044
1045/* comparison routines for qsort */
1046
1047static int
1048compare_pid(const void *p1, const void *p2)
1049{
1050	const struct kinfo_proc * const *pp1 = p1;
1051	const struct kinfo_proc * const *pp2 = p2;
1052
1053	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
1054		abort();
1055
1056	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
1057}
1058
1059/*
1060 *  proc_compare - comparison function for "qsort"
1061 *	Compares the resource consumption of two processes using five
1062 *	distinct keys.  The keys (in descending order of importance) are:
1063 *	percent cpu, cpu ticks, state, resident set size, total virtual
1064 *	memory usage.  The process states are ordered as follows (from least
1065 *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
1066 *	array declaration below maps a process state index into a number
1067 *	that reflects this ordering.
1068 */
1069
1070static int sorted_state[] = {
1071	0,	/* not used		*/
1072	3,	/* sleep		*/
1073	1,	/* ABANDONED (WAIT)	*/
1074	6,	/* run			*/
1075	5,	/* start		*/
1076	2,	/* zombie		*/
1077	4	/* stop			*/
1078};
1079
1080
1081#define ORDERKEY_PCTCPU(a, b) do { \
1082	long diff; \
1083	if (ps.wcpu) \
1084		diff = floor(1.0E6 * weighted_cpu(pctdouble((b)->ki_pctcpu), \
1085		    (b))) - \
1086		    floor(1.0E6 * weighted_cpu(pctdouble((a)->ki_pctcpu), \
1087		    (a))); \
1088	else \
1089		diff = (long)(b)->ki_pctcpu - (long)(a)->ki_pctcpu; \
1090	if (diff != 0) \
1091		return (diff > 0 ? 1 : -1); \
1092} while (0)
1093
1094#define ORDERKEY_CPTICKS(a, b) do { \
1095	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
1096	if (diff != 0) \
1097		return (diff > 0 ? 1 : -1); \
1098} while (0)
1099
1100#define ORDERKEY_STATE(a, b) do { \
1101	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
1102	if (diff != 0) \
1103		return (diff > 0 ? 1 : -1); \
1104} while (0)
1105
1106#define ORDERKEY_PRIO(a, b) do { \
1107	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
1108	if (diff != 0) \
1109		return (diff > 0 ? 1 : -1); \
1110} while (0)
1111
1112#define	ORDERKEY_THREADS(a, b) do { \
1113	int diff = (int)(b)->ki_numthreads - (int)(a)->ki_numthreads; \
1114	if (diff != 0) \
1115		return (diff > 0 ? 1 : -1); \
1116} while (0)
1117
1118#define ORDERKEY_RSSIZE(a, b) do { \
1119	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
1120	if (diff != 0) \
1121		return (diff > 0 ? 1 : -1); \
1122} while (0)
1123
1124#define ORDERKEY_MEM(a, b) do { \
1125	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
1126	if (diff != 0) \
1127		return (diff > 0 ? 1 : -1); \
1128} while (0)
1129
1130#define ORDERKEY_JID(a, b) do { \
1131	int diff = (int)(b)->ki_jid - (int)(a)->ki_jid; \
1132	if (diff != 0) \
1133		return (diff > 0 ? 1 : -1); \
1134} while (0)
1135
1136/* compare_cpu - the comparison function for sorting by cpu percentage */
1137
1138int
1139#ifdef ORDER
1140compare_cpu(void *arg1, void *arg2)
1141#else
1142proc_compare(void *arg1, void *arg2)
1143#endif
1144{
1145	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1146	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1147
1148	ORDERKEY_PCTCPU(p1, p2);
1149	ORDERKEY_CPTICKS(p1, p2);
1150	ORDERKEY_STATE(p1, p2);
1151	ORDERKEY_PRIO(p1, p2);
1152	ORDERKEY_RSSIZE(p1, p2);
1153	ORDERKEY_MEM(p1, p2);
1154
1155	return (0);
1156}
1157
1158#ifdef ORDER
1159/* "cpu" compare routines */
1160int compare_size(), compare_res(), compare_time(), compare_prio(),
1161    compare_threads();
1162
1163/*
1164 * "io" compare routines.  Context switches aren't i/o, but are displayed
1165 * on the "io" display.
1166 */
1167int compare_iototal(), compare_ioread(), compare_iowrite(), compare_iofault(),
1168    compare_vcsw(), compare_ivcsw();
1169
1170int (*compares[])() = {
1171	compare_cpu,
1172	compare_size,
1173	compare_res,
1174	compare_time,
1175	compare_prio,
1176	compare_threads,
1177	compare_iototal,
1178	compare_ioread,
1179	compare_iowrite,
1180	compare_iofault,
1181	compare_vcsw,
1182	compare_ivcsw,
1183	compare_jid,
1184	NULL
1185};
1186
1187/* compare_size - the comparison function for sorting by total memory usage */
1188
1189int
1190compare_size(void *arg1, void *arg2)
1191{
1192	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1193	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1194
1195	ORDERKEY_MEM(p1, p2);
1196	ORDERKEY_RSSIZE(p1, p2);
1197	ORDERKEY_PCTCPU(p1, p2);
1198	ORDERKEY_CPTICKS(p1, p2);
1199	ORDERKEY_STATE(p1, p2);
1200	ORDERKEY_PRIO(p1, p2);
1201
1202	return (0);
1203}
1204
1205/* compare_res - the comparison function for sorting by resident set size */
1206
1207int
1208compare_res(void *arg1, void *arg2)
1209{
1210	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1211	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1212
1213	ORDERKEY_RSSIZE(p1, p2);
1214	ORDERKEY_MEM(p1, p2);
1215	ORDERKEY_PCTCPU(p1, p2);
1216	ORDERKEY_CPTICKS(p1, p2);
1217	ORDERKEY_STATE(p1, p2);
1218	ORDERKEY_PRIO(p1, p2);
1219
1220	return (0);
1221}
1222
1223/* compare_time - the comparison function for sorting by total cpu time */
1224
1225int
1226compare_time(void *arg1, void *arg2)
1227{
1228	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1229	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1230
1231	ORDERKEY_CPTICKS(p1, p2);
1232	ORDERKEY_PCTCPU(p1, p2);
1233	ORDERKEY_STATE(p1, p2);
1234	ORDERKEY_PRIO(p1, p2);
1235	ORDERKEY_RSSIZE(p1, p2);
1236	ORDERKEY_MEM(p1, p2);
1237
1238	return (0);
1239}
1240
1241/* compare_prio - the comparison function for sorting by priority */
1242
1243int
1244compare_prio(void *arg1, void *arg2)
1245{
1246	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1247	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1248
1249	ORDERKEY_PRIO(p1, p2);
1250	ORDERKEY_CPTICKS(p1, p2);
1251	ORDERKEY_PCTCPU(p1, p2);
1252	ORDERKEY_STATE(p1, p2);
1253	ORDERKEY_RSSIZE(p1, p2);
1254	ORDERKEY_MEM(p1, p2);
1255
1256	return (0);
1257}
1258
1259/* compare_threads - the comparison function for sorting by threads */
1260int
1261compare_threads(void *arg1, void *arg2)
1262{
1263	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1264	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1265
1266	ORDERKEY_THREADS(p1, p2);
1267	ORDERKEY_PCTCPU(p1, p2);
1268	ORDERKEY_CPTICKS(p1, p2);
1269	ORDERKEY_STATE(p1, p2);
1270	ORDERKEY_PRIO(p1, p2);
1271	ORDERKEY_RSSIZE(p1, p2);
1272	ORDERKEY_MEM(p1, p2);
1273
1274	return (0);
1275}
1276
1277/* compare_jid - the comparison function for sorting by jid */
1278static int
1279compare_jid(const void *arg1, const void *arg2)
1280{
1281	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1282	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1283
1284	ORDERKEY_JID(p1, p2);
1285	ORDERKEY_PCTCPU(p1, p2);
1286	ORDERKEY_CPTICKS(p1, p2);
1287	ORDERKEY_STATE(p1, p2);
1288	ORDERKEY_PRIO(p1, p2);
1289	ORDERKEY_RSSIZE(p1, p2);
1290	ORDERKEY_MEM(p1, p2);
1291
1292	return (0);
1293}
1294#endif /* ORDER */
1295
1296/* assorted comparison functions for sorting by i/o */
1297
1298int
1299#ifdef ORDER
1300compare_iototal(void *arg1, void *arg2)
1301#else
1302io_compare(void *arg1, void *arg2)
1303#endif
1304{
1305	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1306	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1307
1308	return (get_io_total(p2) - get_io_total(p1));
1309}
1310
1311#ifdef ORDER
1312int
1313compare_ioread(void *arg1, void *arg2)
1314{
1315	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1316	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1317	long dummy, inp1, inp2;
1318
1319	(void) get_io_stats(p1, &inp1, &dummy, &dummy, &dummy, &dummy);
1320	(void) get_io_stats(p2, &inp2, &dummy, &dummy, &dummy, &dummy);
1321
1322	return (inp2 - inp1);
1323}
1324
1325int
1326compare_iowrite(void *arg1, void *arg2)
1327{
1328	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1329	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1330	long dummy, oup1, oup2;
1331
1332	(void) get_io_stats(p1, &dummy, &oup1, &dummy, &dummy, &dummy);
1333	(void) get_io_stats(p2, &dummy, &oup2, &dummy, &dummy, &dummy);
1334
1335	return (oup2 - oup1);
1336}
1337
1338int
1339compare_iofault(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, flp1, flp2;
1344
1345	(void) get_io_stats(p1, &dummy, &dummy, &flp1, &dummy, &dummy);
1346	(void) get_io_stats(p2, &dummy, &dummy, &flp2, &dummy, &dummy);
1347
1348	return (flp2 - flp1);
1349}
1350
1351int
1352compare_vcsw(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, flp1, flp2;
1357
1358	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &flp1, &dummy);
1359	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &flp2, &dummy);
1360
1361	return (flp2 - flp1);
1362}
1363
1364int
1365compare_ivcsw(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, &dummy, &dummy, &flp1);
1372	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &dummy, &flp2);
1373
1374	return (flp2 - flp1);
1375}
1376#endif /* ORDER */
1377
1378/*
1379 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1380 *		the process does not exist.
1381 *		It is EXTREMLY IMPORTANT that this function work correctly.
1382 *		If top runs setuid root (as in SVR4), then this function
1383 *		is the only thing that stands in the way of a serious
1384 *		security problem.  It validates requests for the "kill"
1385 *		and "renice" commands.
1386 */
1387
1388int
1389proc_owner(int pid)
1390{
1391	int cnt;
1392	struct kinfo_proc **prefp;
1393	struct kinfo_proc *pp;
1394
1395	prefp = pref;
1396	cnt = pref_len;
1397	while (--cnt >= 0) {
1398		pp = *prefp++;
1399		if (pp->ki_pid == (pid_t)pid)
1400			return ((int)pp->ki_ruid);
1401	}
1402	return (-1);
1403}
1404
1405static int
1406swapmode(int *retavail, int *retfree)
1407{
1408	int n;
1409	int pagesize = getpagesize();
1410	struct kvm_swap swapary[1];
1411
1412	*retavail = 0;
1413	*retfree = 0;
1414
1415#define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1416
1417	n = kvm_getswapinfo(kd, swapary, 1, 0);
1418	if (n < 0 || swapary[0].ksw_total == 0)
1419		return (0);
1420
1421	*retavail = CONVERT(swapary[0].ksw_total);
1422	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1423
1424	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1425	return (n);
1426}
1427