machine.c revision 132024 
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 132024 2004-07-12 08:22:32Z des $
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   READ  WRITE  FAULT  TOTAL PERCENT COMMAND";
98
99#define io_Proc_format \
100	"%5d %-*.*s %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 *proc_ordernames[] = {
213	"cpu", "size", "res", "time", "pri", NULL
214};
215char *io_ordernames[] = {
216	"total", "read", "write", "fault", NULL
217};
218#endif
219
220int
221machine_init(struct statics *statics)
222{
223	int pagesize;
224	size_t modelen;
225	struct passwd *pw;
226
227	modelen = sizeof(smpmode);
228	if ((sysctlbyname("machdep.smp_active", &smpmode, &modelen, NULL, 0) < 0 &&
229		sysctlbyname("kern.smp.active", &smpmode, &modelen, NULL, 0) < 0) ||
230	    modelen != sizeof(smpmode))
231		smpmode = 0;
232
233	while ((pw = getpwent()) != NULL) {
234		if (strlen(pw->pw_name) > namelength)
235			namelength = strlen(pw->pw_name);
236	}
237	if (namelength < 8)
238		namelength = 8;
239	if (smpmode && namelength > 13)
240		namelength = 13;
241	else if (namelength > 15)
242		namelength = 15;
243
244	kd = kvm_open(NULL, _PATH_DEVNULL, NULL, O_RDONLY, "kvm_open");
245	if (kd == NULL)
246		return (-1);
247
248	GETSYSCTL("kern.ccpu", ccpu);
249
250	/* this is used in calculating WCPU -- calculate it ahead of time */
251	logcpu = log(loaddouble(ccpu));
252
253	pbase = NULL;
254	pref = NULL;
255	nproc = 0;
256	onproc = -1;
257	/* get the page size with "getpagesize" and calculate pageshift from it */
258	pagesize = getpagesize();
259	pageshift = 0;
260	while (pagesize > 1) {
261		pageshift++;
262		pagesize >>= 1;
263	}
264
265	/* we only need the amount of log(2)1024 for our conversion */
266	pageshift -= LOG1024;
267
268	/* fill in the statics information */
269	statics->procstate_names = procstatenames;
270	statics->cpustate_names = cpustatenames;
271	statics->memory_names = memorynames;
272	statics->swap_names = swapnames;
273#ifdef ORDER
274	switch (displaymode) {
275	case DISP_IO:
276		statics->order_names = io_ordernames;
277		break;
278	case DISP_CPU:
279	default:
280		statics->order_names = proc_ordernames;
281		break;
282	}
283#endif
284
285	/* all done! */
286	return (0);
287}
288
289char *
290format_header(char *uname_field)
291{
292	static char Header[128];
293	const char *prehead;
294
295	switch (displaymode) {
296	case DISP_CPU:
297		prehead = smpmode ? smp_header : up_header;
298		break;
299	case DISP_IO:
300		prehead = io_header;
301		break;
302	}
303
304	snprintf(Header, sizeof(Header), prehead,
305	    namelength, namelength, uname_field);
306
307	cmdlengthdelta = strlen(Header) - 7;
308
309	return (Header);
310}
311
312static int swappgsin = -1;
313static int swappgsout = -1;
314extern struct timeval timeout;
315
316void
317get_system_info(struct system_info *si)
318{
319	long total;
320	struct loadavg sysload;
321	int mib[2];
322	struct timeval boottime;
323	size_t bt_size;
324	int i;
325
326	/* get the cp_time array */
327	GETSYSCTL("kern.cp_time", cp_time);
328	GETSYSCTL("vm.loadavg", sysload);
329	GETSYSCTL("kern.lastpid", lastpid);
330
331	/* convert load averages to doubles */
332	for (i = 0; i < 3; i++)
333		si->load_avg[i] = (double)sysload.ldavg[i] / sysload.fscale;
334
335	/* convert cp_time counts to percentages */
336	total = percentages(CPUSTATES, cpu_states, cp_time, cp_old, cp_diff);
337
338	/* sum memory & swap statistics */
339	{
340		static unsigned int swap_delay = 0;
341		static int swapavail = 0;
342		static int swapfree = 0;
343		static int bufspace = 0;
344		static int nspgsin, nspgsout;
345
346		GETSYSCTL("vfs.bufspace", bufspace);
347		GETSYSCTL("vm.stats.vm.v_active_count", memory_stats[0]);
348		GETSYSCTL("vm.stats.vm.v_inactive_count", memory_stats[1]);
349		GETSYSCTL("vm.stats.vm.v_wire_count", memory_stats[2]);
350		GETSYSCTL("vm.stats.vm.v_cache_count", memory_stats[3]);
351		GETSYSCTL("vm.stats.vm.v_free_count", memory_stats[5]);
352		GETSYSCTL("vm.stats.vm.v_swappgsin", nspgsin);
353		GETSYSCTL("vm.stats.vm.v_swappgsout", nspgsout);
354		/* convert memory stats to Kbytes */
355		memory_stats[0] = pagetok(memory_stats[0]);
356		memory_stats[1] = pagetok(memory_stats[1]);
357		memory_stats[2] = pagetok(memory_stats[2]);
358		memory_stats[3] = pagetok(memory_stats[3]);
359		memory_stats[4] = bufspace / 1024;
360		memory_stats[5] = pagetok(memory_stats[5]);
361		memory_stats[6] = -1;
362
363		/* first interval */
364		if (swappgsin < 0) {
365			swap_stats[4] = 0;
366			swap_stats[5] = 0;
367		}
368
369		/* compute differences between old and new swap statistic */
370		else {
371			swap_stats[4] = pagetok(((nspgsin - swappgsin)));
372			swap_stats[5] = pagetok(((nspgsout - swappgsout)));
373		}
374
375		swappgsin = nspgsin;
376		swappgsout = nspgsout;
377
378		/* call CPU heavy swapmode() only for changes */
379		if (swap_stats[4] > 0 || swap_stats[5] > 0 || swap_delay == 0) {
380			swap_stats[3] = swapmode(&swapavail, &swapfree);
381			swap_stats[0] = swapavail;
382			swap_stats[1] = swapavail - swapfree;
383			swap_stats[2] = swapfree;
384		}
385		swap_delay = 1;
386		swap_stats[6] = -1;
387	}
388
389	/* set arrays and strings */
390	si->cpustates = cpu_states;
391	si->memory = memory_stats;
392	si->swap = swap_stats;
393
394
395	if (lastpid > 0) {
396		si->last_pid = lastpid;
397	} else {
398		si->last_pid = -1;
399	}
400
401	/*
402	 * Print how long system has been up.
403	 * (Found by looking getting "boottime" from the kernel)
404	 */
405	mib[0] = CTL_KERN;
406	mib[1] = KERN_BOOTTIME;
407	bt_size = sizeof(boottime);
408	if (sysctl(mib, 2, &boottime, &bt_size, NULL, 0) != -1 &&
409	    boottime.tv_sec != 0) {
410		si->boottime = boottime;
411	} else {
412		si->boottime.tv_sec = -1;
413	}
414}
415
416#define NOPROC	((void *)-1)
417
418const struct kinfo_proc *
419get_old_proc(struct kinfo_proc *pp)
420{
421	struct kinfo_proc **oldpp, *oldp;
422
423	if (previous_proc_count == 0)
424		return (NULL);
425	if (pp->ki_udata == NOPROC)
426		return (NULL);
427	if (pp->ki_udata != NULL)
428		return (pp->ki_udata);
429	oldpp = bsearch(&pp, previous_pref, previous_proc_count,
430	    sizeof(*previous_pref), compare_pid);
431	if (oldpp == NULL) {
432		pp->ki_udata = NOPROC;
433		return (NULL);
434	}
435	oldp = *oldpp;
436	if (bcmp(&oldp->ki_start, &pp->ki_start, sizeof(pp->ki_start)) != 0) {
437		pp->ki_udata = NOPROC;
438		return (NULL);
439	}
440	pp->ki_udata = oldp;
441	return (oldp);
442}
443
444long
445get_io_stats(struct kinfo_proc *pp, long *inp, long *oup, long *flp)
446{
447	const struct kinfo_proc *oldp;
448	static struct kinfo_proc dummy;
449	long ret;
450
451	oldp = get_old_proc(pp);
452	if (oldp == NULL) {
453		bzero(&dummy, sizeof(dummy));
454		oldp = &dummy;
455	}
456
457	*inp = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
458	*oup = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
459	*flp = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
460	ret =
461	    (RU(pp)->ru_inblock - RU(oldp)->ru_inblock) +
462	    (RU(pp)->ru_oublock - RU(oldp)->ru_oublock) +
463	    (RU(pp)->ru_majflt - RU(oldp)->ru_majflt);
464	return (ret);
465}
466
467long
468get_io_total(struct kinfo_proc *pp)
469{
470	long dummy;
471
472	return (get_io_stats(pp, &dummy, &dummy, &dummy));
473}
474
475static struct handle handle;
476
477caddr_t
478get_process_info(struct system_info *si, struct process_select *sel,
479    int (*compare)(const void *, const void *))
480{
481	int i;
482	int total_procs;
483	long p_io;
484	long p_inblock, p_oublock, p_majflt;
485	int active_procs;
486	struct kinfo_proc **prefp;
487	struct kinfo_proc *pp;
488	struct kinfo_proc *prev_pp = NULL;
489
490	/* these are copied out of sel for speed */
491	int show_idle;
492	int show_self;
493	int show_system;
494	int show_uid;
495	int show_command;
496
497	/*
498	 * Save the previous process info.
499	 */
500	if (previous_proc_count_max < nproc) {
501		free(previous_procs);
502		previous_procs = malloc(nproc * sizeof(*previous_procs));
503		free(previous_pref);
504		previous_pref = malloc(nproc * sizeof(*previous_pref));
505		if (previous_procs == NULL || previous_pref == NULL) {
506			(void) fprintf(stderr, "top: Out of memory.\n");
507			quit(23);
508		}
509		previous_proc_count_max = nproc;
510	}
511	if (nproc) {
512		for (i = 0; i < nproc; i++)
513			previous_pref[i] = &previous_procs[i];
514		bcopy(pbase, previous_procs, nproc * sizeof(*previous_procs));
515		qsort(previous_pref, nproc, sizeof(*previous_pref), compare_pid);
516	}
517	previous_proc_count = nproc;
518
519	pbase = kvm_getprocs(kd, KERN_PROC_ALL, 0, &nproc);
520	if (nproc > onproc)
521		pref = realloc(pref, sizeof(*pref) * (onproc = nproc));
522	if (pref == NULL || pbase == NULL) {
523		(void) fprintf(stderr, "top: Out of memory.\n");
524		quit(23);
525	}
526	/* get a pointer to the states summary array */
527	si->procstates = process_states;
528
529	/* set up flags which define what we are going to select */
530	show_idle = sel->idle;
531	show_self = sel->self == -1;
532	show_system = sel->system;
533	show_uid = sel->uid != -1;
534	show_command = sel->command != NULL;
535
536	/* count up process states and get pointers to interesting procs */
537	total_procs = 0;
538	active_procs = 0;
539	total_inblock = 0;
540	total_oublock = 0;
541	total_majflt = 0;
542	memset((char *)process_states, 0, sizeof(process_states));
543	prefp = pref;
544	for (pp = pbase, i = 0; i < nproc; pp++, i++) {
545
546		if (pp->ki_stat == 0)
547			/* not in use */
548			continue;
549
550		if (!show_self && pp->ki_pid == sel->self)
551			/* skip self */
552			continue;
553
554		if (!show_system && (pp->ki_flag & P_SYSTEM))
555			/* skip system process */
556			continue;
557
558		p_io = get_io_stats(pp, &p_inblock, &p_oublock, &p_majflt);
559		total_inblock += p_inblock;
560		total_oublock += p_oublock;
561		total_majflt += p_majflt;
562		total_procs++;
563		process_states[pp->ki_stat]++;
564
565		if (pp->ki_stat == SZOMB)
566			/* skip zombies */
567			continue;
568
569		if (displaymode == DISP_CPU && !show_idle &&
570		    (pp->ki_pctcpu == 0 || pp->ki_stat != SRUN))
571			/* skip idle or non-running processes */
572			continue;
573
574		if (displaymode == DISP_IO && !show_idle && p_io == 0)
575			/* skip processes that aren't doing I/O */
576			continue;
577
578		if (show_uid && pp->ki_ruid != (uid_t)sel->uid)
579			/* skip processes which don't belong to the selected UID */
580			continue;
581
582		/*
583		 * When not showing threads, take the first thread
584		 * for output and add the fields that we can from
585		 * the rest of the process's threads rather than
586		 * using the system's mostly-broken KERN_PROC_PROC.
587		 */
588		if (sel->thread || prev_pp == NULL ||
589		    prev_pp->ki_pid != pp->ki_pid) {
590			*prefp++ = pp;
591			active_procs++;
592			prev_pp = pp;
593		} else {
594			prev_pp->ki_pctcpu += pp->ki_pctcpu;
595		}
596	}
597
598	/* if requested, sort the "interesting" processes */
599	if (compare != NULL)
600		qsort(pref, active_procs, sizeof(*pref), compare);
601
602	/* remember active and total counts */
603	si->p_total = total_procs;
604	si->p_active = pref_len = active_procs;
605
606	/* pass back a handle */
607	handle.next_proc = pref;
608	handle.remaining = active_procs;
609	return ((caddr_t)&handle);
610}
611
612static char fmt[128];	/* static area where result is built */
613
614char *
615format_next_process(caddr_t handle, char *(*get_userid)(int))
616{
617	struct kinfo_proc *pp;
618	const struct kinfo_proc *oldp;
619	long cputime;
620	double pct;
621	struct handle *hp;
622	char status[16];
623	int state;
624	struct rusage ru, *rup;
625	long p_tot, s_tot;
626
627	/* find and remember the next proc structure */
628	hp = (struct handle *)handle;
629	pp = *(hp->next_proc++);
630	hp->remaining--;
631
632	/* get the process's command name */
633	if ((pp->ki_sflag & PS_INMEM) == 0) {
634		/*
635		 * Print swapped processes as <pname>
636		 */
637		size_t len = strlen(pp->ki_comm);
638		if (len > sizeof(pp->ki_comm) - 3)
639			len = sizeof(pp->ki_comm) - 3;
640		memmove(pp->ki_comm + 1, pp->ki_comm, len);
641		pp->ki_comm[0] = '<';
642		pp->ki_comm[len + 1] = '>';
643		pp->ki_comm[len + 2] = '\0';
644	}
645
646	/*
647	 * Convert the process's runtime from microseconds to seconds.  This
648	 * time includes the interrupt time although that is not wanted here.
649	 * ps(1) is similarly sloppy.
650	 */
651	cputime = (pp->ki_runtime + 500000) / 1000000;
652
653	/* calculate the base for cpu percentages */
654	pct = pctdouble(pp->ki_pctcpu);
655
656	/* generate "STATE" field */
657	switch (state = pp->ki_stat) {
658	case SRUN:
659		if (smpmode && pp->ki_oncpu != 0xff)
660			sprintf(status, "CPU%d", pp->ki_oncpu);
661		else
662			strcpy(status, "RUN");
663		break;
664	case SLOCK:
665		if (pp->ki_kiflag & KI_LOCKBLOCK) {
666			sprintf(status, "*%.6s", pp->ki_lockname);
667			break;
668		}
669		/* fall through */
670	case SSLEEP:
671		if (pp->ki_wmesg != NULL) {
672			sprintf(status, "%.6s", pp->ki_wmesg);
673			break;
674		}
675		/* FALLTHROUGH */
676	default:
677
678		if (state >= 0 &&
679		    state < sizeof(state_abbrev) / sizeof(*state_abbrev))
680			sprintf(status, "%.6s", state_abbrev[state]);
681		else
682			sprintf(status, "?%5d", state);
683		break;
684	}
685
686	if (displaymode == DISP_IO) {
687		oldp = get_old_proc(pp);
688		if (oldp != NULL) {
689			ru.ru_inblock = RU(pp)->ru_inblock - RU(oldp)->ru_inblock;
690			ru.ru_oublock = RU(pp)->ru_oublock - RU(oldp)->ru_oublock;
691			ru.ru_majflt = RU(pp)->ru_majflt - RU(oldp)->ru_majflt;
692			rup = &ru;
693		} else {
694			rup = RU(pp);
695		}
696		p_tot = rup->ru_inblock + rup->ru_oublock + rup->ru_majflt;
697		s_tot = total_inblock + total_oublock + total_majflt;
698
699		sprintf(fmt, io_Proc_format,
700		    pp->ki_pid,
701		    namelength, namelength,
702		    (*get_userid)(pp->ki_ruid),
703		    rup->ru_inblock,
704		    rup->ru_oublock,
705		    rup->ru_majflt,
706		    p_tot,
707		    s_tot == 0 ? 0.0 : (p_tot * 100.0 / s_tot),
708		    screen_width > cmdlengthdelta ?
709		    screen_width - cmdlengthdelta : 0,
710		    printable(pp->ki_comm));
711		return (fmt);
712	}
713	/* format this entry */
714	sprintf(fmt,
715	    smpmode ? smp_Proc_format : up_Proc_format,
716	    pp->ki_pid,
717	    namelength, namelength,
718	    (*get_userid)(pp->ki_ruid),
719	    pp->ki_pri.pri_level - PZERO,
720
721	    /*
722	     * normal time      -> nice value -20 - +20
723	     * real time 0 - 31 -> nice value -52 - -21
724	     * idle time 0 - 31 -> nice value +21 - +52
725	     */
726	    (pp->ki_pri.pri_class ==  PRI_TIMESHARE ?
727		pp->ki_nice - NZERO :
728		(PRI_IS_REALTIME(pp->ki_pri.pri_class) ?
729		    (PRIO_MIN - 1 - (PRI_MAX_REALTIME - pp->ki_pri.pri_level)) :
730		    (PRIO_MAX + 1 + pp->ki_pri.pri_level - PRI_MIN_IDLE))),
731	    format_k2(PROCSIZE(pp)),
732	    format_k2(pagetok(pp->ki_rssize)),
733	    status,
734	    smpmode ? pp->ki_lastcpu : 0,
735	    format_time(cputime),
736	    100.0 * weighted_cpu(pct, pp),
737	    100.0 * pct,
738	    screen_width > cmdlengthdelta ?
739	    screen_width - cmdlengthdelta :
740	    0,
741	    printable(pp->ki_comm));
742
743	/* return the result */
744	return (fmt);
745}
746
747static void
748getsysctl(char *name, void *ptr, size_t len)
749{
750	size_t nlen = len;
751
752	if (sysctlbyname(name, ptr, &nlen, NULL, 0) == -1) {
753		fprintf(stderr, "top: sysctl(%s...) failed: %s\n", name,
754		    strerror(errno));
755		quit(23);
756	}
757	if (nlen != len) {
758		fprintf(stderr, "top: sysctl(%s...) expected %lu, got %lu\n", name,
759		    (unsigned long)len, (unsigned long)nlen);
760		quit(23);
761	}
762}
763
764/* comparison routines for qsort */
765
766int
767compare_pid(const void *p1, const void *p2)
768{
769	const struct kinfo_proc * const *pp1 = p1;
770	const struct kinfo_proc * const *pp2 = p2;
771
772	if ((*pp2)->ki_pid < 0 || (*pp1)->ki_pid < 0)
773		abort();
774
775	return ((*pp1)->ki_pid - (*pp2)->ki_pid);
776}
777
778/*
779 *  proc_compare - comparison function for "qsort"
780 *	Compares the resource consumption of two processes using five
781 *	distinct keys.  The keys (in descending order of importance) are:
782 *	percent cpu, cpu ticks, state, resident set size, total virtual
783 *	memory usage.  The process states are ordered as follows (from least
784 *	to most important):  WAIT, zombie, sleep, stop, start, run.  The
785 *	array declaration below maps a process state index into a number
786 *	that reflects this ordering.
787 */
788
789static int sorted_state[] =
790{
791	0,	/* not used		*/
792	3,	/* sleep		*/
793	1,	/* ABANDONED (WAIT)	*/
794	6,	/* run			*/
795	5,	/* start		*/
796	2,	/* zombie		*/
797	4	/* stop			*/
798};
799
800
801#define ORDERKEY_PCTCPU(a, b) do { \
802	long diff = (long)(b)->ki_pctcpu - (long)(a)->ki_pctcpu; \
803	if (diff != 0) \
804		return (diff > 0 ? 1 : -1); \
805} while (0)
806
807#define ORDERKEY_CPTICKS(a, b) do { \
808	int64_t diff = (int64_t)(b)->ki_runtime - (int64_t)(a)->ki_runtime; \
809	if (diff != 0) \
810		return (diff > 0 ? 1 : -1); \
811} while (0)
812
813#define ORDERKEY_STATE(a, b) do { \
814	int diff = sorted_state[(b)->ki_stat] - sorted_state[(a)->ki_stat]; \
815	if (diff != 0) \
816		return (diff > 0 ? 1 : -1); \
817} while (0)
818
819#define ORDERKEY_PRIO(a, b) do { \
820	int diff = (int)(b)->ki_pri.pri_level - (int)(a)->ki_pri.pri_level; \
821	if (diff != 0) \
822		return (diff > 0 ? 1 : -1); \
823} while (0)
824
825#define ORDERKEY_RSSIZE(a, b) do { \
826	long diff = (long)(b)->ki_rssize - (long)(a)->ki_rssize; \
827	if (diff != 0) \
828		return (diff > 0 ? 1 : -1); \
829} while (0)
830
831#define ORDERKEY_MEM(a, b) do { \
832	long diff = (long)PROCSIZE((b)) - (long)PROCSIZE((a)); \
833	if (diff != 0) \
834		return (diff > 0 ? 1 : -1); \
835} while (0)
836
837/* compare_cpu - the comparison function for sorting by cpu percentage */
838
839int
840#ifdef ORDER
841compare_cpu(void *arg1, void *arg2)
842#else
843proc_compare(void *arg1, void *arg2)
844#endif
845{
846	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
847	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
848
849	ORDERKEY_PCTCPU(p1, p2);
850	ORDERKEY_CPTICKS(p1, p2);
851	ORDERKEY_STATE(p1, p2);
852	ORDERKEY_PRIO(p1, p2);
853	ORDERKEY_RSSIZE(p1, p2);
854	ORDERKEY_MEM(p1, p2);
855
856	return (0);
857}
858
859#ifdef ORDER
860/* compare routines */
861int compare_size(), compare_res(), compare_time(), compare_prio();
862
863int (*proc_compares[])() = {
864	compare_cpu,
865	compare_size,
866	compare_res,
867	compare_time,
868	compare_prio,
869	NULL
870};
871
872/* compare_size - the comparison function for sorting by total memory usage */
873
874int
875compare_size(void *arg1, void *arg2)
876{
877	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
878	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
879
880	ORDERKEY_MEM(p1, p2);
881	ORDERKEY_RSSIZE(p1, p2);
882	ORDERKEY_PCTCPU(p1, p2);
883	ORDERKEY_CPTICKS(p1, p2);
884	ORDERKEY_STATE(p1, p2);
885	ORDERKEY_PRIO(p1, p2);
886
887	return (0);
888}
889
890/* compare_res - the comparison function for sorting by resident set size */
891
892int
893compare_res(void *arg1, void *arg2)
894{
895	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
896	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
897
898	ORDERKEY_RSSIZE(p1, p2);
899	ORDERKEY_MEM(p1, p2);
900	ORDERKEY_PCTCPU(p1, p2);
901	ORDERKEY_CPTICKS(p1, p2);
902	ORDERKEY_STATE(p1, p2);
903	ORDERKEY_PRIO(p1, p2);
904
905	return (0);
906}
907
908/* compare_time - the comparison function for sorting by total cpu time */
909
910int
911compare_time(void *arg1, void *arg2)
912{
913	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
914	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
915
916	ORDERKEY_CPTICKS(p1, p2);
917	ORDERKEY_PCTCPU(p1, p2);
918	ORDERKEY_STATE(p1, p2);
919	ORDERKEY_PRIO(p1, p2);
920	ORDERKEY_RSSIZE(p1, p2);
921	ORDERKEY_MEM(p1, p2);
922
923	return (0);
924}
925
926/* compare_prio - the comparison function for sorting by priority */
927
928int
929compare_prio(void *arg1, void *arg2)
930{
931	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
932	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
933
934	ORDERKEY_PRIO(p1, p2);
935	ORDERKEY_CPTICKS(p1, p2);
936	ORDERKEY_PCTCPU(p1, p2);
937	ORDERKEY_STATE(p1, p2);
938	ORDERKEY_RSSIZE(p1, p2);
939	ORDERKEY_MEM(p1, p2);
940
941	return (0);
942}
943#endif
944
945/* compare_io - the comparison function for sorting by total io */
946
947int
948#ifdef ORDER
949compare_iototal(void *arg1, void *arg2)
950#else
951io_compare(void *arg1, void *arg2)
952#endif
953{
954	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
955	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
956
957	return (get_io_total(p2) - get_io_total(p1));
958}
959
960#ifdef ORDER
961/* io compare routines */
962int compare_ioread(), compare_iowrite(), compare_iofault();
963
964int (*io_compares[])() = {
965	compare_iototal,
966	compare_ioread,
967	compare_iowrite,
968	compare_iofault,
969	NULL
970};
971
972int
973compare_ioread(void *arg1, void *arg2)
974{
975	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
976	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
977	long dummy, inp1, inp2;
978
979	(void) get_io_stats(p1, &inp1, &dummy, &dummy);
980	(void) get_io_stats(p2, &inp2, &dummy, &dummy);
981
982	return (inp2 - inp1);
983}
984
985int
986compare_iowrite(void *arg1, void *arg2)
987{
988	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
989	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
990	long dummy, oup1, oup2;
991
992	(void) get_io_stats(p1, &dummy, &oup1, &dummy);
993	(void) get_io_stats(p2, &dummy, &oup2, &dummy);
994
995	return (oup2 - oup1);
996}
997
998int
999compare_iofault(void *arg1, void *arg2)
1000{
1001	struct kinfo_proc *p1 = *(struct kinfo_proc **)arg1;
1002	struct kinfo_proc *p2 = *(struct kinfo_proc **)arg2;
1003	long dummy, flp1, flp2;
1004
1005	(void) get_io_stats(p1, &dummy, &dummy, &flp1);
1006	(void) get_io_stats(p2, &dummy, &dummy, &flp2);
1007
1008	return (flp2 - flp1);
1009}
1010
1011#endif /* ORDER */
1012
1013/*
1014 * proc_owner(pid) - returns the uid that owns process "pid", or -1 if
1015 *		the process does not exist.
1016 *		It is EXTREMLY IMPORTANT that this function work correctly.
1017 *		If top runs setuid root (as in SVR4), then this function
1018 *		is the only thing that stands in the way of a serious
1019 *		security problem.  It validates requests for the "kill"
1020 *		and "renice" commands.
1021 */
1022
1023int
1024proc_owner(int pid)
1025{
1026	int cnt;
1027	struct kinfo_proc **prefp;
1028	struct kinfo_proc *pp;
1029
1030	prefp = pref;
1031	cnt = pref_len;
1032	while (--cnt >= 0) {
1033		pp = *prefp++;
1034		if (pp->ki_pid == (pid_t)pid)
1035			return ((int)pp->ki_ruid);
1036	}
1037	return (-1);
1038}
1039
1040int
1041swapmode(int *retavail, int *retfree)
1042{
1043	int n;
1044	int pagesize = getpagesize();
1045	struct kvm_swap swapary[1];
1046
1047	*retavail = 0;
1048	*retfree = 0;
1049
1050#define CONVERT(v)	((quad_t)(v) * pagesize / 1024)
1051
1052	n = kvm_getswapinfo(kd, swapary, 1, 0);
1053	if (n < 0 || swapary[0].ksw_total == 0)
1054		return (0);
1055
1056	*retavail = CONVERT(swapary[0].ksw_total);
1057	*retfree = CONVERT(swapary[0].ksw_total - swapary[0].ksw_used);
1058
1059	n = (int)(swapary[0].ksw_used * 100.0 / swapary[0].ksw_total);
1060	return (n);
1061}
1062