machine.c revision 133817
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 133817 2004-08-16 07:51:22Z alfred $
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 <unistd.h>
46
47#include "top.h"
48#include "machine.h"
49#include "screen.h"
50#include "utils.h"
51
52static void getsysctl(char *, void *, size_t);
53
54#define GETSYSCTL(name, var) getsysctl(name, &(var), sizeof(var))
55
56extern char* printable(char *);
57int swapmode(int *retavail, int *retfree);
58static int smpmode;
59enum displaymodes displaymode;
60static int namelength;
61static int cmdlengthdelta;
62
63/* Prototypes for top internals */
64void quit(int);
65int compare_pid(const void *a, const void *b);
66
67/* get_process_info passes back a handle.  This is what it looks like: */
68
69struct handle
70{
71	struct kinfo_proc **next_proc;	/* points to next valid proc pointer */
72	int remaining;			/* number of pointers remaining */
73};
74
75/* declarations for load_avg */
76#include "loadavg.h"
77
78/* define what weighted cpu is.  */
79#define weighted_cpu(pct, pp) ((pp)->ki_swtime == 0 ? 0.0 : \
80			 ((pct) / (1.0 - exp((pp)->ki_swtime * logcpu))))
81
82/* what we consider to be process size: */
83#define PROCSIZE(pp) ((pp)->ki_size / 1024)
84
85#define RU(pp)	(&(pp)->ki_rusage)
86#define RUTOT(pp) \
87	(RU(pp)->ru_inblock + RU(pp)->ru_oublock + RU(pp)->ru_majflt)
88
89
90/* definitions for indices in the nlist array */
91
92/*
93 *  These definitions control the format of the per-process area
94 */
95
96static char io_header[] =
97	"  PID %-*.*s   VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
98
99#define io_Proc_format \
100	"%5d %-*.*s %6ld %6ld %6ld %6ld %6ld %6ld %6.2f%% %.*s"
101
102static char smp_header[] =
103	"  PID %-*.*s PRI NICE   SIZE    RES STATE  C   TIME   WCPU    CPU COMMAND";
104
105#define smp_Proc_format \
106	"%5d %-*.*s %3d %4d%7s %6s %-6.6s %1x%7s %5.2f%% %5.2f%% %.*s"
107
108static char up_header[] =
109	"  PID %-*.*s PRI NICE   SIZE    RES STATE    TIME   WCPU    CPU COMMAND";
110
111#define up_Proc_format \
112	"%5d %-*.*s %3d %4d%7s %6s %-6.6s%.0d%7s %5.2f%% %5.2f%% %.*s"
113
114
115
116/* process state names for the "STATE" column of the display */
117/* the extra nulls in the string "run" are for adding a slash and
118   the processor number when needed */
119
120char *state_abbrev[] =
121{
122	"", "START", "RUN\0\0\0", "SLEEP", "STOP", "ZOMB", "WAIT", "LOCK"
123};
124
125
126static kvm_t *kd;
127
128/* values that we stash away in _init and use in later routines */
129
130static double logcpu;
131
132/* these are retrieved from the kernel in _init */
133
134static load_avg  ccpu;
135
136/* these are used in the get_ functions */
137
138static int lastpid;
139
140/* these are for calculating cpu state percentages */
141
142static long cp_time[CPUSTATES];
143static long cp_old[CPUSTATES];
144static long cp_diff[CPUSTATES];
145
146/* these are for detailing the process states */
147
148int process_states[8];
149char *procstatenames[] = {
150	"", " starting, ", " running, ", " sleeping, ", " stopped, ",
151	" zombie, ", " waiting, ", " lock, ",
152	NULL
153};
154
155/* these are for detailing the cpu states */
156
157int cpu_states[CPUSTATES];
158char *cpustatenames[] = {
159	"user", "nice", "system", "interrupt", "idle", NULL
160};
161
162/* these are for detailing the memory statistics */
163
164int memory_stats[7];
165char *memorynames[] = {
166	/* 0             1            2            3            4          5 */
167	"K Active, ", "K Inact, ", "K Wired, ", "K Cache, ", "K Buf, ", "K Free",
168	NULL
169};
170
171int swap_stats[7];
172char *swapnames[] = {
173	/* 0            1           2           3            4         5 */
174	"K Total, ", "K Used, ", "K Free, ", "% Inuse, ", "K In, ", "K Out",
175	NULL
176};
177
178
179/* these are for keeping track of the proc array */
180
181static int nproc;
182static int onproc = -1;
183static int pref_len;
184static struct kinfo_proc *pbase;
185static struct kinfo_proc **pref;
186static struct kinfo_proc *previous_procs;
187static struct kinfo_proc **previous_pref;
188static int previous_proc_count = 0;
189static int previous_proc_count_max = 0;
190
191/* total number of io operations */
192static long total_inblock;
193static long total_oublock;
194static long total_majflt;
195
196/* these are for getting the memory statistics */
197
198static int pageshift;		/* log base 2 of the pagesize */
199
200/* define pagetok in terms of pageshift */
201
202#define pagetok(size) ((size) << pageshift)
203
204/* useful externals */
205long percentages();
206
207#ifdef ORDER
208/*
209 * Sorting orders.  One vector per display mode.
210 * The first element is the default for each mode.
211 */
212char *ordernames[] = {
213	"cpu", "size", "res", "time", "pri",
214	"total", "read", "write", "fault", "vcsw", "ivcsw", NULL
215};
216#endif
217
218int
219machine_init(struct statics *statics)
220{
221	int pagesize;
222	size_t modelen;
223	struct passwd *pw;
224
225	modelen = sizeof(smpmode);
226	if ((sysctlbyname("machdep.smp_active", &smpmode, &modelen, NULL, 0) < 0 &&
227		sysctlbyname("kern.smp.active", &smpmode, &modelen, NULL, 0) < 0) ||
228	    modelen != sizeof(smpmode))
229		smpmode = 0;
230
231	while ((pw = getpwent()) != NULL) {
232		if (strlen(pw->pw_name) > namelength)
233			namelength = strlen(pw->pw_name);
234	}
235	if (namelength < 8)
236		namelength = 8;
237	if (smpmode && namelength > 13)
238		namelength = 13;
239	else if (namelength > 15)
240		namelength = 15;
241
242	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
243	if (kd == NULL)
244		return (-1);
245
246	GETSYSCTL("kern.ccpu", ccpu);
247
248	/* this is used in calculating WCPU -- calculate it ahead of time */
249	logcpu = log(loaddouble(ccpu));
250
251	pbase = NULL;
252	pref = NULL;
253	nproc = 0;
254	onproc = -1;
255	/* get the page size with "getpagesize" and calculate pageshift from it */
256	pagesize = getpagesize();
257	pageshift = 0;
258	while (pagesize > 1) {
259		pageshift++;
260		pagesize >>= 1;
261	}
262
263	/* we only need the amount of log(2)1024 for our conversion */
264	pageshift -= LOG1024;
265
266	/* fill in the statics information */
267	statics->procstate_names = procstatenames;
268	statics->cpustate_names = cpustatenames;
269	statics->memory_names = memorynames;
270	statics->swap_names = swapnames;
271#ifdef ORDER
272	statics->order_names = ordernames;
273#endif
274
275	/* all done! */
276	return (0);
277}
278
279char *
280format_header(char *uname_field)
281{
282	static char Header[128];
283	const char *prehead;
284
285	switch (displaymode) {
286	case DISP_CPU:
287		prehead = smpmode ? smp_header : up_header;
288		break;
289	case DISP_IO:
290		prehead = io_header;
291		break;
292	}
293
294	snprintf(Header, sizeof(Header), prehead,
295	    namelength, namelength, uname_field);
296
297	cmdlengthdelta = strlen(Header) - 7;
298
299	return (Header);
300}
301
302static int swappgsin = -1;
303static int swappgsout = -1;
304extern struct timeval timeout;
305
306void
307get_system_info(struct system_info *si)
308{
309	long total;
310	struct loadavg sysload;
311	int mib[2];
312	struct timeval boottime;
313	size_t bt_size;
314	int i;
315
316	/* get the cp_time array */
317	GETSYSCTL("kern.cp_time", cp_time);
318	GETSYSCTL("vm.loadavg", sysload);
319	GETSYSCTL("kern.lastpid", lastpid);
320
321	/* convert load averages to doubles */
322	for (i = 0; i < 3; i++)
323		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
324
325	/* convert cp_time counts to percentages */
326	total = percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
327
328	/* sum memory & swap statistics */
329	{
330		static unsigned int swap_delay = 0;
331		static int swapavail = 0;
332		static int swapfree = 0;
333		static int bufspace = 0;
334		static int nspgsin, nspgsout;
335
336		GETSYSCTL("vfs.bufspace", bufspace);
337		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
338		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
339		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[2]);
340		GETSYSCTL("vm.stats.vm.v_cache_count", memory_stats[3]);
341		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
342		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
343		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
344		/* convert memory stats to Kbytes */
345		memory_stats[0] = pagetok(memory_stats[0]);
346		memory_stats[1] = pagetok(memory_stats[1]);
347		memory_stats[2] = pagetok(memory_stats[2]);
348		memory_stats[3] = pagetok(memory_stats[3]);
349		memory_stats[4] = bufspace / 1024;
350		memory_stats[5] = pagetok(memory_stats[5]);
351		memory_stats[6] = -1;
352
353		/* first interval */
354		if (swappgsin < 0) {
355			swap_stats[4] = 0;
356			swap_stats[5] = 0;
357		}
358
359		/* compute differences between old and new swap statistic */
360		else {
361			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
362			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
363		}
364
365		swappgsin = nspgsin;
366		swappgsout = nspgsout;
367
368		/* call CPU heavy swapmode() only for changes */
369		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
370			swap_stats[3] = swapmode(&swapavail, &swapfree);
371			swap_stats[0] = swapavail;
372			swap_stats[1] = swapavail - swapfree;
373			swap_stats[2] = swapfree;
374		}
375		swap_delay = 1;
376		swap_stats[6] = -1;
377	}
378
379	/* set arrays and strings */
380	si->cpustates = cpu_states;
381	si->memory = memory_stats;
382	si->swap = swap_stats;
383
384
385	if (lastpid > 0) {
386		si->last_pid = lastpid;
387	} else {
388		si->last_pid = -1;
389	}
390
391	/*
392	 * Print how long system has been up.
393	 * (Found by looking getting "boottime" from the kernel)
394	 */
395	mib[0] = CTL_KERN;
396	mib[1] = KERN_BOOTTIME;
397	bt_size = sizeof(boottime);
398	if (sysctl(mib, 2, &boottime, &bt_size, NULL, 0) != -1 &&
399	    boottime.tv_sec != 0) {
400		si->boottime = boottime;
401	} else {
402		si->boottime.tv_sec = -1;
403	}
404}
405
406#define NOPROC	((void *)-1)
407
408/*
409 * We need to compare data from the old process entry with the new
410 * process entry.
411 * To facilitate doing this quickly we stash a pointer in the kinfo_proc
412 * structure to cache the mapping.  We also use a negative cache pointer
413 * of NOPROC to avoid duplicate lookups.
414 * XXX: this could be done when the actual processes are fetched, we do
415 * it here out of laziness.
416 */
417const struct kinfo_proc *
418get_old_proc(struct kinfo_proc *pp)
419{
420	struct kinfo_proc **oldpp, *oldp;
421
422	/*
423	 * If this is the first fetch of the kinfo_procs then we don't have
424	 * any previous entries.
425	 */
426	if (previous_proc_count == 0)
427		return (NULL);
428	/* negative cache? */
429	if (pp->ki_udata == NOPROC)
430		return (NULL);
431	/* cached? */
432	if (pp->ki_udata != NULL)
433		return (pp->ki_udata);
434	/*
435	 * Not cached,
436	 * 1) look up based on pid.
437	 * 2) compare process start.
438	 * If we fail here, then setup a negative cache entry, otherwise
439	 * cache it.
440	 */
441	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
442	    sizeof(*previous_pref), compare_pid);
443	if (oldpp == NULL) {
444		pp->ki_udata = NOPROC;
445		return (NULL);
446	}
447	oldp = *oldpp;
448	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
449		pp->ki_udata = NOPROC;
450		return (NULL);
451	}
452	pp->ki_udata = oldp;
453	return (oldp);
454}
455
456/*
457 * Return the total amount of IO done in blocks in/out and faults.
458 * store the values individually in the pointers passed in.
459 */
460long
461get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp, long *vcsw, long *ivcsw)
462{
463	const struct kinfo_proc *oldp;
464	static struct kinfo_proc dummy;
465	long ret;
466
467	oldp = get_old_proc(pp);
468	if (oldp == NULL) {
469		bzero(&dummy, sizeof(dummy));
470		oldp = &dummy;
471	}
472
473	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
474	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
475	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
476	*vcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
477	*ivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
478	ret =
479	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
480	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
481	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
482	return (ret);
483}
484
485/*
486 * Return the total number of block in/out and faults by a process.
487 */
488long
489get_io_total(struct kinfo_proc *pp)
490{
491	long dummy;
492
493	return (get_io_stats(pp, &dummy, &dummy, &dummy, &dummy, &dummy));
494}
495
496static struct handle handle;
497
498caddr_t
499get_process_info(struct system_info *si, struct process_select *sel,
500    int (*compare)(const void *, const void *))
501{
502	int i;
503	int total_procs;
504	long p_io;
505	long p_inblock, p_oublock, p_majflt, p_vcsw, p_ivcsw;
506	int active_procs;
507	struct kinfo_proc **prefp;
508	struct kinfo_proc *pp;
509	struct kinfo_proc *prev_pp = NULL;
510
511	/* these are copied out of sel for speed */
512	int show_idle;
513	int show_self;
514	int show_system;
515	int show_uid;
516	int show_command;
517
518	/*
519	 * Save the previous process info.
520	 */
521	if (previous_proc_count_max < nproc) {
522		free(previous_procs);
523		previous_procs = malloc(nproc * sizeof(*previous_procs));
524		free(previous_pref);
525		previous_pref = malloc(nproc * sizeof(*previous_pref));
526		if (previous_procs == NULL || previous_pref == NULL) {
527			(void) fprintf(stderr, "top: Out of memory.\n");
528			quit(23);
529		}
530		previous_proc_count_max = nproc;
531	}
532	if (nproc) {
533		for (i = 0; i < nproc; i++)
534			previous_pref[i] = &previous_procs[i];
535		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
536		qsort(previous_pref, nproc, sizeof(*previous_pref), compare_pid);
537	}
538	previous_proc_count = nproc;
539
540	pbase = kvm_getprocs(kd, KERN_PROC_ALL, 0, &nproc);
541	if (nproc > onproc)
542		pref = realloc(pref, sizeof(*pref) * (onproc = nproc));
543	if (pref == NULL || pbase == NULL) {
544		(void) fprintf(stderr, "top: Out of memory.\n");
545		quit(23);
546	}
547	/* get a pointer to the states summary array */
548	si->procstates = process_states;
549
550	/* set up flags which define what we are going to select */
551	show_idle = sel->idle;
552	show_self = sel->self == -1;
553	show_system = sel->system;
554	show_uid = sel->uid != -1;
555	show_command = sel->command != NULL;
556
557	/* count up process states and get pointers to interesting procs */
558	total_procs = 0;
559	active_procs = 0;
560	total_inblock = 0;
561	total_oublock = 0;
562	total_majflt = 0;
563	memset((char *)process_states, 0, sizeof(process_states));
564	prefp = pref;
565	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
566
567		if (pp->ki_stat == 0)
568			/* not in use */
569			continue;
570
571		if (!show_self && pp->ki_pid == sel->self)
572			/* skip self */
573			continue;
574
575		if (!show_system && (pp->ki_flag & P_SYSTEM))
576			/* skip system process */
577			continue;
578
579		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt, &p_vcsw, &p_ivcsw);
580		total_inblock += p_inblock;
581		total_oublock += p_oublock;
582		total_majflt += p_majflt;
583		total_procs++;
584		process_states[pp->ki_stat]++;
585
586		if (pp->ki_stat == SZOMB)
587			/* skip zombies */
588			continue;
589
590		if (displaymode == DISP_CPU && !show_idle &&
591		    (pp->ki_pctcpu == 0 || pp->ki_stat != SRUN))
592			/* skip idle or non-running processes */
593			continue;
594
595		if (displaymode == DISP_IO && !show_idle && p_io == 0)
596			/* skip processes that aren't doing I/O */
597			continue;
598
599		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
600			/* skip processes which don't belong to the selected UID */
601			continue;
602
603		/*
604		 * When not showing threads, take the first thread
605		 * for output and add the fields that we can from
606		 * the rest of the process's threads rather than
607		 * using the system's mostly-broken KERN_PROC_PROC.
608		 */
609		if (sel->thread || prev_pp == NULL ||
610		    prev_pp->ki_pid != pp->ki_pid) {
611			*prefp++ = pp;
612			active_procs++;
613			prev_pp = pp;
614		} else {
615			prev_pp->ki_pctcpu += pp->ki_pctcpu;
616		}
617	}
618
619	/* if requested, sort the "interesting" processes */
620	if (compare != NULL)
621		qsort(pref, active_procs, sizeof(*pref), compare);
622
623	/* remember active and total counts */
624	si->p_total = total_procs;
625	si->p_active = pref_len = active_procs;
626
627	/* pass back a handle */
628	handle.next_proc = pref;
629	handle.remaining = active_procs;
630	return ((caddr_t)&handle);
631}
632
633static char fmt[128];	/* static area where result is built */
634
635char *
636format_next_process(caddr_t handle, char *(*get_userid)(int))
637{
638	struct kinfo_proc *pp;
639	const struct kinfo_proc *oldp;
640	long cputime;
641	double pct;
642	struct handle *hp;
643	char status[16];
644	int state;
645	struct rusage ru, *rup;
646	long p_tot, s_tot;
647
648	/* find and remember the next proc structure */
649	hp = (struct handle *)handle;
650	pp = *(hp->next_proc++);
651	hp->remaining--;
652
653	/* get the process's command name */
654	if ((pp->ki_sflag & PS_INMEM) == 0) {
655		/*
656		 * Print swapped processes as <pname>
657		 */
658		size_t len = strlen(pp->ki_comm);
659		if (len > sizeof(pp->ki_comm) - 3)
660			len = sizeof(pp->ki_comm) - 3;
661		memmove(pp->ki_comm + 1, pp->ki_comm, len);
662		pp->ki_comm[0] = '<';
663		pp->ki_comm[len + 1] = '>';
664		pp->ki_comm[len + 2] = '\0';
665	}
666
667	/*
668	 * Convert the process's runtime from microseconds to seconds.  This
669	 * time includes the interrupt time although that is not wanted here.
670	 * ps(1) is similarly sloppy.
671	 */
672	cputime = (pp->ki_runtime + 500000) / 1000000;
673
674	/* calculate the base for cpu percentages */
675	pct = pctdouble(pp->ki_pctcpu);
676
677	/* generate "STATE" field */
678	switch (state = pp->ki_stat) {
679	case SRUN:
680		if (smpmode && pp->ki_oncpu != 0xff)
681			sprintf(status, "CPU%d", pp->ki_oncpu);
682		else
683			strcpy(status, "RUN");
684		break;
685	case SLOCK:
686		if (pp->ki_kiflag & KI_LOCKBLOCK) {
687			sprintf(status, "*%.6s", pp->ki_lockname);
688			break;
689		}
690		/* fall through */
691	case SSLEEP:
692		if (pp->ki_wmesg != NULL) {
693			sprintf(status, "%.6s", pp->ki_wmesg);
694			break;
695		}
696		/* FALLTHROUGH */
697	default:
698
699		if (state >= 0 &&
700		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
701			sprintf(status, "%.6s", state_abbrev[state]);
702		else
703			sprintf(status, "?%5d", state);
704		break;
705	}
706
707	if (displaymode == DISP_IO) {
708		oldp = get_old_proc(pp);
709		if (oldp != NULL) {
710			ru.ru_inblock = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
711			ru.ru_oublock = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
712			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
713			ru.ru_nvcsw = RU(pp)->ru_nvcsw - RU(oldp)->ru_nvcsw;
714			ru.ru_nivcsw = RU(pp)->ru_nivcsw - RU(oldp)->ru_nivcsw;
715			rup = &ru;
716		} else {
717			rup = RU(pp);
718		}
719		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
720		s_tot = total_inblock + total_oublock + total_majflt;
721
722		sprintf(fmt, io_Proc_format,
723		    pp->ki_pid,
724		    namelength, namelength,
725		    (*get_userid)(pp->ki_ruid),
726		    rup->ru_nvcsw,
727		    rup->ru_nivcsw,
728		    rup->ru_inblock,
729		    rup->ru_oublock,
730		    rup->ru_majflt,
731		    p_tot,
732		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
733		    screen_width > cmdlengthdelta ?
734		    screen_width - cmdlengthdelta : 0,
735		    printable(pp->ki_comm));
736		return (fmt);
737	}
738	/* format this entry */
739	sprintf(fmt,
740	    smpmode ? smp_Proc_format : up_Proc_format,
741	    pp->ki_pid,
742	    namelength, namelength,
743	    (*get_userid)(pp->ki_ruid),
744	    pp->ki_pri.pri_level - PZERO,
745
746	    /*
747	     * normal time      -> nice value -20 - +20
748	     * real time 0 - 31 -> nice value -52 - -21
749	     * idle time 0 - 31 -> nice value +21 - +52
750	     */
751	    (pp->ki_pri.pri_class ==  PRI_TIMESHARE ?
752		pp->ki_nice - NZERO :
753		(PRI_IS_REALTIME(pp->ki_pri.pri_class) ?
754		    (PRIO_MIN - 1 - (PRI_MAX_REALTIME - pp->ki_pri.pri_level)) :
755		    (PRIO_MAX + 1 + pp->ki_pri.pri_level - PRI_MIN_IDLE))),
756	    format_k2(PROCSIZE(pp)),
757	    format_k2(pagetok(pp->ki_rssize)),
758	    status,
759	    smpmode ? pp->ki_lastcpu : 0,
760	    format_time(cputime),
761	    100.0 * weighted_cpu(pct, pp),
762	    100.0 * pct,
763	    screen_width > cmdlengthdelta ?
764	    screen_width - cmdlengthdelta :
765	    0,
766	    printable(pp->ki_comm));
767
768	/* return the result */
769	return (fmt);
770}
771
772static void
773getsysctl(char *name, void *ptr, size_t len)
774{
775	size_t nlen = len;
776
777	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
778		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
779		    strerror(errno));
780		quit(23);
781	}
782	if (nlen != len) {
783		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n", name,
784		    (unsigned long)len, (unsigned long)nlen);
785		quit(23);
786	}
787}
788
789/* comparison routines for qsort */
790
791int
792compare_pid(const void *p1, const void *p2)
793{
794	const struct kinfo_proc * const *pp1 = p1;
795	const struct kinfo_proc * const *pp2 = p2;
796
797	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
798		abort();
799
800	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
801}
802
803/*
804 *  proc_compare - comparison function for "qsort"
805 *	Compares the resource consumption of two processes using five
806 *	distinct keys.  The keys (in descending order of importance) are:
807 *	percent cpu, cpu ticks, state, resident set size, total virtual
808 *	memory usage.  The process states are ordered as follows (from least
809 *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
810 *	array declaration below maps a process state index into a number
811 *	that reflects this ordering.
812 */
813
814static int sorted_state[] =
815{
816	0,	/* not used		*/
817	3,	/* sleep		*/
818	1,	/* ABANDONED (WAIT)	*/
819	6,	/* run			*/
820	5,	/* start		*/
821	2,	/* zombie		*/
822	4	/* stop			*/
823};
824
825
826#define ORDERKEY_PCTCPU(a, b) do { \
827	long diff = (long)(b)->ki_pctcpu - (long)(a)->ki_pctcpu; \
828	if (diff != 0) \
829		return (diff > 0 ? 1 : -1); \
830} while (0)
831
832#define ORDERKEY_CPTICKS(a, b) do { \
833	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
834	if (diff != 0) \
835		return (diff > 0 ? 1 : -1); \
836} while (0)
837
838#define ORDERKEY_STATE(a, b) do { \
839	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
840	if (diff != 0) \
841		return (diff > 0 ? 1 : -1); \
842} while (0)
843
844#define ORDERKEY_PRIO(a, b) do { \
845	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
846	if (diff != 0) \
847		return (diff > 0 ? 1 : -1); \
848} while (0)
849
850#define ORDERKEY_RSSIZE(a, b) do { \
851	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
852	if (diff != 0) \
853		return (diff > 0 ? 1 : -1); \
854} while (0)
855
856#define ORDERKEY_MEM(a, b) do { \
857	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
858	if (diff != 0) \
859		return (diff > 0 ? 1 : -1); \
860} while (0)
861
862/* compare_cpu - the comparison function for sorting by cpu percentage */
863
864int
865#ifdef ORDER
866compare_cpu(void *arg1, void *arg2)
867#else
868proc_compare(void *arg1, void *arg2)
869#endif
870{
871	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
872	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
873
874	ORDERKEY_PCTCPU(p1, p2);
875	ORDERKEY_CPTICKS(p1, p2);
876	ORDERKEY_STATE(p1, p2);
877	ORDERKEY_PRIO(p1, p2);
878	ORDERKEY_RSSIZE(p1, p2);
879	ORDERKEY_MEM(p1, p2);
880
881	return (0);
882}
883
884#ifdef ORDER
885/* compare routines */
886int compare_size(), compare_res(), compare_time(), compare_prio();
887/* io compare routines */
888int compare_iototal(), compare_ioread(), compare_iowrite(), compare_iofault(), compare_vcsw(), compare_ivcsw();
889
890int (*compares[])() = {
891	compare_cpu,
892	compare_size,
893	compare_res,
894	compare_time,
895	compare_prio,
896	compare_iototal,
897	compare_ioread,
898	compare_iowrite,
899	compare_iofault,
900	compare_vcsw,
901	compare_ivcsw,
902	NULL
903};
904
905/* compare_size - the comparison function for sorting by total memory usage */
906
907int
908compare_size(void *arg1, void *arg2)
909{
910	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
911	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
912
913	ORDERKEY_MEM(p1, p2);
914	ORDERKEY_RSSIZE(p1, p2);
915	ORDERKEY_PCTCPU(p1, p2);
916	ORDERKEY_CPTICKS(p1, p2);
917	ORDERKEY_STATE(p1, p2);
918	ORDERKEY_PRIO(p1, p2);
919
920	return (0);
921}
922
923/* compare_res - the comparison function for sorting by resident set size */
924
925int
926compare_res(void *arg1, void *arg2)
927{
928	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
929	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
930
931	ORDERKEY_RSSIZE(p1, p2);
932	ORDERKEY_MEM(p1, p2);
933	ORDERKEY_PCTCPU(p1, p2);
934	ORDERKEY_CPTICKS(p1, p2);
935	ORDERKEY_STATE(p1, p2);
936	ORDERKEY_PRIO(p1, p2);
937
938	return (0);
939}
940
941/* compare_time - the comparison function for sorting by total cpu time */
942
943int
944compare_time(void *arg1, void *arg2)
945{
946	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
947	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
948
949	ORDERKEY_CPTICKS(p1, p2);
950	ORDERKEY_PCTCPU(p1, p2);
951	ORDERKEY_STATE(p1, p2);
952	ORDERKEY_PRIO(p1, p2);
953	ORDERKEY_RSSIZE(p1, p2);
954	ORDERKEY_MEM(p1, p2);
955
956	return (0);
957}
958
959/* compare_prio - the comparison function for sorting by priority */
960
961int
962compare_prio(void *arg1, void *arg2)
963{
964	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
965	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
966
967	ORDERKEY_PRIO(p1, p2);
968	ORDERKEY_CPTICKS(p1, p2);
969	ORDERKEY_PCTCPU(p1, p2);
970	ORDERKEY_STATE(p1, p2);
971	ORDERKEY_RSSIZE(p1, p2);
972	ORDERKEY_MEM(p1, p2);
973
974	return (0);
975}
976#endif
977
978/* compare_io - the comparison function for sorting by total io */
979
980int
981#ifdef ORDER
982compare_iototal(void *arg1, void *arg2)
983#else
984io_compare(void *arg1, void *arg2)
985#endif
986{
987	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
988	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
989
990	return (get_io_total(p2) - get_io_total(p1));
991}
992
993#ifdef ORDER
994
995int
996compare_ioread(void *arg1, void *arg2)
997{
998	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
999	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1000	long dummy, inp1, inp2;
1001
1002	(void) get_io_stats(p1, &inp1, &dummy, &dummy, &dummy, &dummy);
1003	(void) get_io_stats(p2, &inp2, &dummy, &dummy, &dummy, &dummy);
1004
1005	return (inp2 - inp1);
1006}
1007
1008int
1009compare_iowrite(void *arg1, void *arg2)
1010{
1011	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1012	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1013	long dummy, oup1, oup2;
1014
1015	(void) get_io_stats(p1, &dummy, &oup1, &dummy, &dummy, &dummy);
1016	(void) get_io_stats(p2, &dummy, &oup2, &dummy, &dummy, &dummy);
1017
1018	return (oup2 - oup1);
1019}
1020
1021int
1022compare_iofault(void *arg1, void *arg2)
1023{
1024	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1025	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1026	long dummy, flp1, flp2;
1027
1028	(void) get_io_stats(p1, &dummy, &dummy, &flp1, &dummy, &dummy);
1029	(void) get_io_stats(p2, &dummy, &dummy, &flp2, &dummy, &dummy);
1030
1031	return (flp2 - flp1);
1032}
1033
1034int
1035compare_vcsw(void *arg1, void *arg2)
1036{
1037	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1038	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1039	long dummy, flp1, flp2;
1040
1041	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &flp1, &dummy);
1042	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &flp2, &dummy);
1043
1044	return (flp2 - flp1);
1045}
1046
1047int
1048compare_ivcsw(void *arg1, void *arg2)
1049{
1050	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1051	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1052	long dummy, flp1, flp2;
1053
1054	(void) get_io_stats(p1, &dummy, &dummy, &dummy, &dummy, &flp1);
1055	(void) get_io_stats(p2, &dummy, &dummy, &dummy, &dummy, &flp2);
1056
1057	return (flp2 - flp1);
1058}
1059
1060#endif /* ORDER */
1061
1062/*
1063 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1064 *		the process does not exist.
1065 *		It is EXTREMLY IMPORTANT that this function work correctly.
1066 *		If top runs setuid root (as in SVR4), then this function
1067 *		is the only thing that stands in the way of a serious
1068 *		security problem.  It validates requests for the "kill"
1069 *		and "renice" commands.
1070 */
1071
1072int
1073proc_owner(int pid)
1074{
1075	int cnt;
1076	struct kinfo_proc **prefp;
1077	struct kinfo_proc *pp;
1078
1079	prefp = pref;
1080	cnt = pref_len;
1081	while (--cnt >= 0) {
1082		pp = *prefp++;
1083		if (pp->ki_pid == (pid_t)pid)
1084			return ((int)pp->ki_ruid);
1085	}
1086	return (-1);
1087}
1088
1089int
1090swapmode(int *retavail, int *retfree)
1091{
1092	int n;
1093	int pagesize = getpagesize();
1094	struct kvm_swap swapary[1];
1095
1096	*retavail = 0;
1097	*retfree = 0;
1098
1099#define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1100
1101	n = kvm_getswapinfo(kd, swapary, 1, 0);
1102	if (n < 0 || swapary[0].ksw_total == 0)
1103		return (0);
1104
1105	*retavail = CONVERT(swapary[0].ksw_total);
1106	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1107
1108	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1109	return (n);
1110}
1111