1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1980, 1986, 1991, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 *    may be used to endorse or promote products derived from this software
17 *    without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32#ifndef lint
33static const char copyright[] =
34"@(#) Copyright (c) 1980, 1986, 1991, 1993\n\
35	The Regents of the University of California.  All rights reserved.\n";
36#endif /* not lint */
37
38#if 0
39#ifndef lint
40static char sccsid[] = "@(#)vmstat.c	8.1 (Berkeley) 6/6/93";
41#endif /* not lint */
42#endif
43
44#include <sys/cdefs.h>
45__FBSDID("$FreeBSD$");
46
47#include <sys/param.h>
48#include <sys/proc.h>
49#include <sys/uio.h>
50#include <sys/namei.h>
51#include <sys/malloc.h>
52#include <sys/signal.h>
53#include <sys/fcntl.h>
54#include <sys/ioctl.h>
55#include <sys/resource.h>
56#include <sys/sysctl.h>
57#include <sys/time.h>
58#include <sys/user.h>
59#define	_WANT_VMMETER
60#include <sys/vmmeter.h>
61#include <sys/pcpu.h>
62
63#include <vm/vm_param.h>
64
65#include <ctype.h>
66#include <devstat.h>
67#include <err.h>
68#include <errno.h>
69#include <inttypes.h>
70#include <kvm.h>
71#include <limits.h>
72#include <memstat.h>
73#include <nlist.h>
74#include <paths.h>
75#include <stdio.h>
76#include <stdlib.h>
77#include <string.h>
78#include <sysexits.h>
79#include <time.h>
80#include <unistd.h>
81#include <libutil.h>
82#include <libxo/xo.h>
83
84#define VMSTAT_XO_VERSION "1"
85
86static char da[] = "da";
87
88enum x_stats { X_SUM, X_HZ, X_STATHZ, X_NCHSTATS, X_INTRNAMES, X_SINTRNAMES,
89    X_INTRCNT, X_SINTRCNT, X_NINTRCNT };
90
91static struct nlist namelist[] = {
92	[X_SUM] = { .n_name = "_vm_cnt", },
93	[X_HZ] = { .n_name = "_hz", },
94	[X_STATHZ] = { .n_name = "_stathz", },
95	[X_NCHSTATS] = { .n_name = "_nchstats", },
96	[X_INTRNAMES] = { .n_name = "_intrnames", },
97	[X_SINTRNAMES] = { .n_name = "_sintrnames", },
98	[X_INTRCNT] = { .n_name = "_intrcnt", },
99	[X_SINTRCNT] = { .n_name = "_sintrcnt", },
100	[X_NINTRCNT] = { .n_name = "_nintrcnt", },
101	{ .n_name = NULL, },
102};
103
104static struct devstat_match *matches;
105static struct device_selection *dev_select;
106static struct statinfo cur, last;
107static devstat_select_mode select_mode;
108static size_t size_cp_times;
109static long *cur_cp_times, *last_cp_times;
110static long generation, select_generation;
111static int hz, hdrcnt, maxshowdevs;
112static int num_devices, num_devices_specified;
113static int num_matches, num_selected, num_selections;
114static char **specified_devices;
115
116static struct __vmmeter {
117	uint64_t v_swtch;
118	uint64_t v_trap;
119	uint64_t v_syscall;
120	uint64_t v_intr;
121	uint64_t v_soft;
122	uint64_t v_vm_faults;
123	uint64_t v_io_faults;
124	uint64_t v_cow_faults;
125	uint64_t v_cow_optim;
126	uint64_t v_zfod;
127	uint64_t v_ozfod;
128	uint64_t v_swapin;
129	uint64_t v_swapout;
130	uint64_t v_swappgsin;
131	uint64_t v_swappgsout;
132	uint64_t v_vnodein;
133	uint64_t v_vnodeout;
134	uint64_t v_vnodepgsin;
135	uint64_t v_vnodepgsout;
136	uint64_t v_intrans;
137	uint64_t v_reactivated;
138	uint64_t v_pdwakeups;
139	uint64_t v_pdpages;
140	uint64_t v_pdshortfalls;
141	uint64_t v_dfree;
142	uint64_t v_pfree;
143	uint64_t v_tfree;
144	uint64_t v_forks;
145	uint64_t v_vforks;
146	uint64_t v_rforks;
147	uint64_t v_kthreads;
148	uint64_t v_forkpages;
149	uint64_t v_vforkpages;
150	uint64_t v_rforkpages;
151	uint64_t v_kthreadpages;
152	u_int v_page_size;
153	u_int v_page_count;
154	u_int v_free_reserved;
155	u_int v_free_target;
156	u_int v_free_min;
157	u_int v_free_count;
158	u_int v_wire_count;
159	u_long v_user_wire_count;
160	u_int v_active_count;
161	u_int v_inactive_target;
162	u_int v_inactive_count;
163	u_int v_laundry_count;
164	u_int v_pageout_free_min;
165	u_int v_interrupt_free_min;
166	u_int v_free_severe;
167} sum, osum;
168
169#define	VMSTAT_DEFAULT_LINES	20	/* Default number of `winlines'. */
170static volatile sig_atomic_t wresized;		/* Tty resized when non-zero. */
171static int winlines = VMSTAT_DEFAULT_LINES; /* Current number of tty rows. */
172
173static int	aflag;
174static int	nflag;
175static int	Pflag;
176static int	hflag;
177
178static kvm_t	*kd;
179
180#define	FORKSTAT	0x01
181#define	INTRSTAT	0x02
182#define	MEMSTAT		0x04
183#define	SUMSTAT		0x08
184#define	TIMESTAT	0x10
185#define	VMSTAT		0x20
186#define	ZMEMSTAT	0x40
187#define	OBJSTAT		0x80
188
189static void	cpustats(void);
190static void	pcpustats(u_long, int);
191static void	devstats(void);
192static void	doforkst(void);
193static void	dointr(unsigned int, int);
194static void	doobjstat(void);
195static void	dosum(void);
196static void	dovmstat(unsigned int, int);
197static void	domemstat_malloc(void);
198static void	domemstat_zone(void);
199static void	kread(int, void *, size_t);
200static void	kreado(int, void *, size_t, size_t);
201static void	kreadptr(uintptr_t, void *, size_t);
202static void	needhdr(int);
203static void	needresize(int);
204static void	doresize(void);
205static void	printhdr(int, u_long);
206static void	usage(void);
207
208static long	pct(long, long);
209static long long	getuptime(void);
210
211static char	**getdrivedata(char **);
212
213int
214main(int argc, char *argv[])
215{
216	char *bp, *buf, *memf, *nlistf;
217	float f;
218	int bufsize, c, reps, todo;
219	size_t len;
220	unsigned int interval;
221	char errbuf[_POSIX2_LINE_MAX];
222
223	memf = nlistf = NULL;
224	interval = reps = todo = 0;
225	maxshowdevs = 2;
226	hflag = isatty(1);
227
228	argc = xo_parse_args(argc, argv);
229	if (argc < 0)
230		return (argc);
231
232	while ((c = getopt(argc, argv, "ac:fhHiM:mN:n:oPp:sw:z")) != -1) {
233		switch (c) {
234		case 'a':
235			aflag++;
236			break;
237		case 'c':
238			reps = atoi(optarg);
239			break;
240		case 'P':
241			Pflag++;
242			break;
243		case 'f':
244			todo |= FORKSTAT;
245			break;
246		case 'h':
247			hflag = 1;
248			break;
249		case 'H':
250			hflag = 0;
251			break;
252		case 'i':
253			todo |= INTRSTAT;
254			break;
255		case 'M':
256			memf = optarg;
257			break;
258		case 'm':
259			todo |= MEMSTAT;
260			break;
261		case 'N':
262			nlistf = optarg;
263			break;
264		case 'n':
265			nflag = 1;
266			maxshowdevs = atoi(optarg);
267			if (maxshowdevs < 0)
268				xo_errx(1, "number of devices %d is < 0",
269				    maxshowdevs);
270			break;
271		case 'o':
272			todo |= OBJSTAT;
273			break;
274		case 'p':
275			if (devstat_buildmatch(optarg, &matches, &num_matches)
276			    != 0)
277				xo_errx(1, "%s", devstat_errbuf);
278			break;
279		case 's':
280			todo |= SUMSTAT;
281			break;
282		case 'w':
283			/* Convert to milliseconds. */
284			f = atof(optarg);
285			interval = f * 1000;
286			break;
287		case 'z':
288			todo |= ZMEMSTAT;
289			break;
290		case '?':
291		default:
292			usage();
293		}
294	}
295	argc -= optind;
296	argv += optind;
297
298	xo_set_version(VMSTAT_XO_VERSION);
299	if (todo == 0)
300		todo = VMSTAT;
301
302	if (memf != NULL) {
303		kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, errbuf);
304		if (kd == NULL)
305			xo_errx(1, "kvm_openfiles: %s", errbuf);
306	}
307
308retry_nlist:
309	if (kd != NULL && (c = kvm_nlist(kd, namelist)) != 0) {
310		if (c > 0) {
311			bufsize = 0;
312			len = 0;
313
314			/*
315			 * 'cnt' was renamed to 'vm_cnt'.  If 'vm_cnt' is not
316			 * found try looking up older 'cnt' symbol.
317			 * */
318			if (namelist[X_SUM].n_type == 0 &&
319			    strcmp(namelist[X_SUM].n_name, "_vm_cnt") == 0) {
320				namelist[X_SUM].n_name = "_cnt";
321				goto retry_nlist;
322			}
323
324			/*
325			 * 'nintrcnt' doesn't exist in older kernels, but
326			 * that isn't fatal.
327			 */
328			if (namelist[X_NINTRCNT].n_type == 0 && c == 1)
329				goto nlist_ok;
330
331			for (c = 0; c < (int)(nitems(namelist)); c++)
332				if (namelist[c].n_type == 0)
333					bufsize += strlen(namelist[c].n_name)
334					    + 1;
335			bufsize += len + 1;
336			buf = bp = alloca(bufsize);
337
338			for (c = 0; c < (int)(nitems(namelist)); c++)
339				if (namelist[c].n_type == 0) {
340					xo_error(" %s",
341					    namelist[c].n_name);
342					len = strlen(namelist[c].n_name);
343					*bp++ = ' ';
344					memcpy(bp, namelist[c].n_name, len);
345					bp += len;
346				}
347			*bp = '\0';
348			xo_error("undefined symbols:\n", buf);
349		} else
350			xo_warnx("kvm_nlist: %s", kvm_geterr(kd));
351		xo_finish();
352		exit(1);
353	}
354nlist_ok:
355	if (kd && Pflag)
356		xo_errx(1, "Cannot use -P with crash dumps");
357
358	if (todo & VMSTAT) {
359		/*
360		 * Make sure that the userland devstat version matches the
361		 * kernel devstat version.  If not, exit and print a
362		 * message informing the user of his mistake.
363		 */
364		if (devstat_checkversion(NULL) < 0)
365			xo_errx(1, "%s", devstat_errbuf);
366
367
368		argv = getdrivedata(argv);
369	}
370
371	if (*argv) {
372		f = atof(*argv);
373		interval = f * 1000;
374		if (*++argv)
375			reps = atoi(*argv);
376	}
377
378	if (interval) {
379		if (!reps)
380			reps = -1;
381	} else if (reps)
382		interval = 1 * 1000;
383
384	if (todo & FORKSTAT)
385		doforkst();
386	if (todo & MEMSTAT)
387		domemstat_malloc();
388	if (todo & ZMEMSTAT)
389		domemstat_zone();
390	if (todo & SUMSTAT)
391		dosum();
392	if (todo & OBJSTAT)
393		doobjstat();
394	if (todo & INTRSTAT)
395		dointr(interval, reps);
396	if (todo & VMSTAT)
397		dovmstat(interval, reps);
398	xo_finish();
399	exit(0);
400}
401
402static int
403mysysctl(const char *name, void *oldp, size_t *oldlenp)
404{
405	int error;
406
407	error = sysctlbyname(name, oldp, oldlenp, NULL, 0);
408	if (error != 0 && errno != ENOMEM)
409		xo_err(1, "sysctl(%s)", name);
410	return (error);
411}
412
413static char **
414getdrivedata(char **argv)
415{
416
417	if ((num_devices = devstat_getnumdevs(NULL)) < 0)
418		xo_errx(1, "%s", devstat_errbuf);
419
420	cur.dinfo = (struct devinfo *)calloc(1, sizeof(struct devinfo));
421	last.dinfo = (struct devinfo *)calloc(1, sizeof(struct devinfo));
422
423	if (devstat_getdevs(NULL, &cur) == -1)
424		xo_errx(1, "%s", devstat_errbuf);
425
426	num_devices = cur.dinfo->numdevs;
427	generation = cur.dinfo->generation;
428
429	specified_devices = malloc(sizeof(char *));
430	for (num_devices_specified = 0; *argv; ++argv) {
431		if (isdigit(**argv))
432			break;
433		num_devices_specified++;
434		specified_devices = reallocf(specified_devices,
435		    sizeof(char *) * num_devices_specified);
436		if (specified_devices == NULL) {
437			xo_errx(1, "%s", "reallocf (specified_devices)");
438		}
439		specified_devices[num_devices_specified - 1] = *argv;
440	}
441	dev_select = NULL;
442
443	if (nflag == 0 && maxshowdevs < num_devices_specified)
444		maxshowdevs = num_devices_specified;
445
446	/*
447	 * People are generally only interested in disk statistics when
448	 * they're running vmstat.  So, that's what we're going to give
449	 * them if they don't specify anything by default.  We'll also give
450	 * them any other random devices in the system so that we get to
451	 * maxshowdevs devices, if that many devices exist.  If the user
452	 * specifies devices on the command line, either through a pattern
453	 * match or by naming them explicitly, we will give the user only
454	 * those devices.
455	 */
456	if ((num_devices_specified == 0) && (num_matches == 0)) {
457		if (devstat_buildmatch(da, &matches, &num_matches) != 0)
458			xo_errx(1, "%s", devstat_errbuf);
459		select_mode = DS_SELECT_ADD;
460	} else
461		select_mode = DS_SELECT_ONLY;
462
463	/*
464	 * At this point, selectdevs will almost surely indicate that the
465	 * device list has changed, so we don't look for return values of 0
466	 * or 1.  If we get back -1, though, there is an error.
467	 */
468	if (devstat_selectdevs(&dev_select, &num_selected, &num_selections,
469	    &select_generation, generation, cur.dinfo->devices,
470	    num_devices, matches, num_matches, specified_devices,
471	    num_devices_specified, select_mode,
472	    maxshowdevs, 0) == -1)
473		xo_errx(1, "%s", devstat_errbuf);
474
475	return(argv);
476}
477
478/* Return system uptime in nanoseconds */
479static long long
480getuptime(void)
481{
482	struct timespec sp;
483
484	(void)clock_gettime(CLOCK_UPTIME, &sp);
485	return((long long)sp.tv_sec * 1000000000LL + sp.tv_nsec);
486}
487
488static void
489fill_vmmeter(struct __vmmeter *vmmp)
490{
491	struct vmmeter vm_cnt;
492	size_t size;
493
494	if (kd != NULL) {
495		kread(X_SUM, &vm_cnt, sizeof(vm_cnt));
496#define	GET_COUNTER(name) \
497		vmmp->name = kvm_counter_u64_fetch(kd, (u_long)vm_cnt.name)
498		GET_COUNTER(v_swtch);
499		GET_COUNTER(v_trap);
500		GET_COUNTER(v_syscall);
501		GET_COUNTER(v_intr);
502		GET_COUNTER(v_soft);
503		GET_COUNTER(v_vm_faults);
504		GET_COUNTER(v_io_faults);
505		GET_COUNTER(v_cow_faults);
506		GET_COUNTER(v_cow_optim);
507		GET_COUNTER(v_zfod);
508		GET_COUNTER(v_ozfod);
509		GET_COUNTER(v_swapin);
510		GET_COUNTER(v_swapout);
511		GET_COUNTER(v_swappgsin);
512		GET_COUNTER(v_swappgsout);
513		GET_COUNTER(v_vnodein);
514		GET_COUNTER(v_vnodeout);
515		GET_COUNTER(v_vnodepgsin);
516		GET_COUNTER(v_vnodepgsout);
517		GET_COUNTER(v_intrans);
518		GET_COUNTER(v_tfree);
519		GET_COUNTER(v_forks);
520		GET_COUNTER(v_vforks);
521		GET_COUNTER(v_rforks);
522		GET_COUNTER(v_kthreads);
523		GET_COUNTER(v_forkpages);
524		GET_COUNTER(v_vforkpages);
525		GET_COUNTER(v_rforkpages);
526		GET_COUNTER(v_kthreadpages);
527#undef GET_COUNTER
528	} else {
529#define GET_VM_STATS(cat, name)	do {					\
530	size = sizeof(vmmp->name);					\
531	mysysctl("vm.stats." #cat "." #name, &vmmp->name, &size);	\
532} while (0)
533		/* sys */
534		GET_VM_STATS(sys, v_swtch);
535		GET_VM_STATS(sys, v_trap);
536		GET_VM_STATS(sys, v_syscall);
537		GET_VM_STATS(sys, v_intr);
538		GET_VM_STATS(sys, v_soft);
539
540		/* vm */
541		GET_VM_STATS(vm, v_vm_faults);
542		GET_VM_STATS(vm, v_io_faults);
543		GET_VM_STATS(vm, v_cow_faults);
544		GET_VM_STATS(vm, v_cow_optim);
545		GET_VM_STATS(vm, v_zfod);
546		GET_VM_STATS(vm, v_ozfod);
547		GET_VM_STATS(vm, v_swapin);
548		GET_VM_STATS(vm, v_swapout);
549		GET_VM_STATS(vm, v_swappgsin);
550		GET_VM_STATS(vm, v_swappgsout);
551		GET_VM_STATS(vm, v_vnodein);
552		GET_VM_STATS(vm, v_vnodeout);
553		GET_VM_STATS(vm, v_vnodepgsin);
554		GET_VM_STATS(vm, v_vnodepgsout);
555		GET_VM_STATS(vm, v_intrans);
556		GET_VM_STATS(vm, v_reactivated);
557		GET_VM_STATS(vm, v_pdwakeups);
558		GET_VM_STATS(vm, v_pdpages);
559		GET_VM_STATS(vm, v_pdshortfalls);
560		GET_VM_STATS(vm, v_dfree);
561		GET_VM_STATS(vm, v_pfree);
562		GET_VM_STATS(vm, v_tfree);
563		GET_VM_STATS(vm, v_page_size);
564		GET_VM_STATS(vm, v_page_count);
565		GET_VM_STATS(vm, v_free_reserved);
566		GET_VM_STATS(vm, v_free_target);
567		GET_VM_STATS(vm, v_free_min);
568		GET_VM_STATS(vm, v_free_count);
569		GET_VM_STATS(vm, v_wire_count);
570		GET_VM_STATS(vm, v_user_wire_count);
571		GET_VM_STATS(vm, v_active_count);
572		GET_VM_STATS(vm, v_inactive_target);
573		GET_VM_STATS(vm, v_inactive_count);
574		GET_VM_STATS(vm, v_laundry_count);
575		GET_VM_STATS(vm, v_pageout_free_min);
576		GET_VM_STATS(vm, v_interrupt_free_min);
577		/*GET_VM_STATS(vm, v_free_severe);*/
578		GET_VM_STATS(vm, v_forks);
579		GET_VM_STATS(vm, v_vforks);
580		GET_VM_STATS(vm, v_rforks);
581		GET_VM_STATS(vm, v_kthreads);
582		GET_VM_STATS(vm, v_forkpages);
583		GET_VM_STATS(vm, v_vforkpages);
584		GET_VM_STATS(vm, v_rforkpages);
585		GET_VM_STATS(vm, v_kthreadpages);
586#undef GET_VM_STATS
587	}
588}
589
590static void
591fill_vmtotal(struct vmtotal *vmtp)
592{
593	size_t size;
594
595	if (kd != NULL) {
596		/* XXX fill vmtp */
597		xo_errx(1, "not implemented");
598	} else {
599		size = sizeof(*vmtp);
600		mysysctl("vm.vmtotal", vmtp, &size);
601		if (size != sizeof(*vmtp))
602			xo_errx(1, "vm.total size mismatch");
603	}
604}
605
606/* Determine how many cpu columns, and what index they are in kern.cp_times */
607static int
608getcpuinfo(u_long *maskp, int *maxidp)
609{
610	long *times;
611	u_long mask;
612	size_t size;
613	int empty, i, j, maxcpu, maxid, ncpus;
614
615	if (kd != NULL)
616		xo_errx(1, "not implemented");
617	mask = 0;
618	ncpus = 0;
619	size = sizeof(maxcpu);
620	mysysctl("kern.smp.maxcpus", &maxcpu, &size);
621	if (size != sizeof(maxcpu))
622		xo_errx(1, "sysctl kern.smp.maxcpus");
623	size = sizeof(long) * maxcpu * CPUSTATES;
624	times = malloc(size);
625	if (times == NULL)
626		xo_err(1, "malloc %zd bytes", size);
627	mysysctl("kern.cp_times", times, &size);
628	maxid = (size / CPUSTATES / sizeof(long)) - 1;
629	for (i = 0; i <= maxid; i++) {
630		empty = 1;
631		for (j = 0; empty && j < CPUSTATES; j++) {
632			if (times[i * CPUSTATES + j] != 0)
633				empty = 0;
634		}
635		if (!empty) {
636			mask |= (1ul << i);
637			ncpus++;
638		}
639	}
640	if (maskp)
641		*maskp = mask;
642	if (maxidp)
643		*maxidp = maxid;
644	return (ncpus);
645}
646
647
648static void
649prthuman(const char *name, uint64_t val, int size, int flags)
650{
651	char buf[10];
652	char fmt[128];
653
654	snprintf(fmt, sizeof(fmt), "{:%s/%%*s}", name);
655
656	if (size < 5 || size > 9)
657		xo_errx(1, "doofus");
658	flags |= HN_NOSPACE | HN_DECIMAL;
659	humanize_number(buf, size, val, "", HN_AUTOSCALE, flags);
660	xo_attr("value", "%ju", (uintmax_t) val);
661	xo_emit(fmt, size, buf);
662}
663
664static void
665dovmstat(unsigned int interval, int reps)
666{
667	struct clockinfo clockrate;
668	struct vmtotal total;
669	struct devinfo *tmp_dinfo;
670	u_long cpumask;
671	size_t size;
672	time_t uptime, halfuptime;
673	int ncpus, maxid, rate_adj, retval;
674
675	uptime = getuptime() / 1000000000LL;
676	halfuptime = uptime / 2;
677	rate_adj = 1;
678	ncpus = 1;
679	maxid = 0;
680	cpumask = 0;
681
682	/*
683	 * If the user stops the program (control-Z) and then resumes it,
684	 * print out the header again.
685	 */
686	(void)signal(SIGCONT, needhdr);
687
688	/*
689	 * If our standard output is a tty, then install a SIGWINCH handler
690	 * and set wresized so that our first iteration through the main
691	 * vmstat loop will peek at the terminal's current rows to find out
692	 * how many lines can fit in a screenful of output.
693	 */
694	if (isatty(fileno(stdout)) != 0) {
695		wresized = 1;
696		(void)signal(SIGWINCH, needresize);
697	} else {
698		wresized = 0;
699		winlines = VMSTAT_DEFAULT_LINES;
700	}
701
702	if (kd != NULL) {
703		if (namelist[X_STATHZ].n_type != 0 &&
704		    namelist[X_STATHZ].n_value != 0)
705			kread(X_STATHZ, &hz, sizeof(hz));
706		if (!hz)
707			kread(X_HZ, &hz, sizeof(hz));
708	} else {
709		size = sizeof(clockrate);
710		mysysctl("kern.clockrate", &clockrate, &size);
711		if (size != sizeof(clockrate))
712			xo_errx(1, "clockrate size mismatch");
713		hz = clockrate.hz;
714	}
715
716	if (Pflag) {
717		ncpus = getcpuinfo(&cpumask, &maxid);
718		size_cp_times = sizeof(long) * (maxid + 1) * CPUSTATES;
719		cur_cp_times = calloc(1, size_cp_times);
720		last_cp_times = calloc(1, size_cp_times);
721	}
722	for (hdrcnt = 1;;) {
723		if (!--hdrcnt)
724			printhdr(maxid, cpumask);
725		if (kd != NULL) {
726			if (kvm_getcptime(kd, cur.cp_time) < 0)
727				xo_errx(1, "kvm_getcptime: %s", kvm_geterr(kd));
728		} else {
729			size = sizeof(cur.cp_time);
730			mysysctl("kern.cp_time", &cur.cp_time, &size);
731			if (size != sizeof(cur.cp_time))
732				xo_errx(1, "cp_time size mismatch");
733		}
734		if (Pflag) {
735			size = size_cp_times;
736			mysysctl("kern.cp_times", cur_cp_times, &size);
737			if (size != size_cp_times)
738				xo_errx(1, "cp_times mismatch");
739		}
740
741		tmp_dinfo = last.dinfo;
742		last.dinfo = cur.dinfo;
743		cur.dinfo = tmp_dinfo;
744		last.snap_time = cur.snap_time;
745
746		/*
747		 * Here what we want to do is refresh our device stats.
748		 * getdevs() returns 1 when the device list has changed.
749		 * If the device list has changed, we want to go through
750		 * the selection process again, in case a device that we
751		 * were previously displaying has gone away.
752		 */
753		switch (devstat_getdevs(NULL, &cur)) {
754		case -1:
755			xo_errx(1, "%s", devstat_errbuf);
756			break;
757		case 1:
758			num_devices = cur.dinfo->numdevs;
759			generation = cur.dinfo->generation;
760
761			retval = devstat_selectdevs(&dev_select, &num_selected,
762			    &num_selections, &select_generation,
763			    generation, cur.dinfo->devices,
764			    num_devices, matches, num_matches,
765			    specified_devices,
766			    num_devices_specified, select_mode,
767			    maxshowdevs, 0);
768			switch (retval) {
769			case -1:
770				xo_errx(1, "%s", devstat_errbuf);
771				break;
772			case 1:
773				printhdr(maxid, cpumask);
774				break;
775			default:
776				break;
777			}
778			break;
779		default:
780			break;
781		}
782
783		fill_vmmeter(&sum);
784		fill_vmtotal(&total);
785		xo_open_container("processes");
786		xo_emit("{:runnable/%2d} {:waiting/%2ld} "
787		    "{:swapped-out/%2ld}", total.t_rq - 1, total.t_dw +
788		    total.t_pw, total.t_sw);
789		xo_close_container("processes");
790		xo_open_container("memory");
791#define vmstat_pgtok(a) ((uintmax_t)(a) * (sum.v_page_size >> 10))
792#define	rate(x)	(unsigned long)(((x) * rate_adj + halfuptime) / uptime)
793		if (hflag) {
794			prthuman("available-memory",
795			    total.t_avm * (uint64_t)sum.v_page_size, 5, HN_B);
796			prthuman("free-memory",
797			    total.t_free * (uint64_t)sum.v_page_size, 5, HN_B);
798			prthuman("total-page-faults",
799			    rate(sum.v_vm_faults - osum.v_vm_faults), 5, 0);
800			xo_emit(" ");
801		} else {
802			xo_emit(" ");
803			xo_emit("{:available-memory/%7ju}",
804			    vmstat_pgtok(total.t_avm));
805			xo_emit(" ");
806			xo_emit("{:free-memory/%7ju}",
807			    vmstat_pgtok(total.t_free));
808			xo_emit(" ");
809			xo_emit("{:total-page-faults/%5lu} ",
810			    rate(sum.v_vm_faults - osum.v_vm_faults));
811		}
812		xo_close_container("memory");
813
814		xo_open_container("paging-rates");
815		xo_emit("{:page-reactivated/%3lu} ",
816		    rate(sum.v_reactivated - osum.v_reactivated));
817		xo_emit("{:paged-in/%3lu} ",
818		    rate(sum.v_swapin + sum.v_vnodein -
819		    (osum.v_swapin + osum.v_vnodein)));
820		xo_emit("{:paged-out/%3lu}",
821		    rate(sum.v_swapout + sum.v_vnodeout -
822		    (osum.v_swapout + osum.v_vnodeout)));
823		if (hflag) {
824			prthuman("freed",
825			    rate(sum.v_tfree - osum.v_tfree), 5, 0);
826			prthuman("scanned",
827			    rate(sum.v_pdpages - osum.v_pdpages), 5, 0);
828			xo_emit(" ");
829		} else {
830			xo_emit(" ");
831			xo_emit("{:freed/%5lu} ",
832			    rate(sum.v_tfree - osum.v_tfree));
833			xo_emit("{:scanned/%4lu} ",
834			    rate(sum.v_pdpages - osum.v_pdpages));
835		}
836		xo_close_container("paging-rates");
837
838		devstats();
839		xo_open_container("fault-rates");
840		xo_emit("{:interrupts/%4lu}", rate(sum.v_intr - osum.v_intr));
841		if (hflag) {
842			prthuman("system-calls",
843			    rate(sum.v_syscall - osum.v_syscall), 5, 0);
844			prthuman("context-switches",
845			    rate(sum.v_swtch - osum.v_swtch), 5, 0);
846		} else {
847			xo_emit(" ");
848			xo_emit("{:system-calls/%5lu} "
849			    "{:context-switches/%5lu}",
850			    rate(sum.v_syscall - osum.v_syscall),
851			    rate(sum.v_swtch - osum.v_swtch));
852		}
853		xo_close_container("fault-rates");
854		if (Pflag)
855			pcpustats(cpumask, maxid);
856		else
857			cpustats();
858		xo_emit("\n");
859		xo_flush();
860		if (reps >= 0 && --reps <= 0)
861			break;
862		osum = sum;
863		uptime = interval;
864		rate_adj = 1000;
865		/*
866		 * We round upward to avoid losing low-frequency events
867		 * (i.e., >= 1 per interval but < 1 per millisecond).
868		 */
869		if (interval != 1)
870			halfuptime = (uptime + 1) / 2;
871		else
872			halfuptime = 0;
873		(void)usleep(interval * 1000);
874	}
875}
876
877static void
878printhdr(int maxid, u_long cpumask)
879{
880	int i, num_shown;
881
882	num_shown = MIN(num_selected, maxshowdevs);
883	if (hflag)
884		xo_emit(" {T:procs}    {T:memory}    {T:/page%*s}", 19, "");
885	else
886		xo_emit("{T:procs}     {T:memory}       {T:/page%*s}", 19, "");
887	if (num_shown > 1)
888		xo_emit("   {T:/disks %*s}  ", num_shown * 4 - 7, "");
889	else if (num_shown == 1)
890		xo_emit("   {T:disks}");
891	xo_emit(" {T:faults}      ");
892	if (Pflag) {
893		for (i = 0; i <= maxid; i++) {
894			if (cpumask & (1ul << i))
895				xo_emit("  {T:/cpu%d}   ", i);
896		}
897		xo_emit("\n");
898	} else
899		xo_emit(" {T:cpu}\n");
900	if (hflag) {
901		xo_emit(" {T:r}  {T:b}  {T:w}  {T:avm}  {T:fre}  {T:flt}  {T:re}"
902		    "  {T:pi}  {T:po}   {T:fr}   {T:sr} ");
903	} else {
904		xo_emit("{T:r} {T:b} {T:w}     {T:avm}     {T:fre}  {T:flt}  "
905		    "{T:re}  {T:pi}  {T:po}    {T:fr}   {T:sr} ");
906	}
907	for (i = 0; i < num_devices; i++)
908		if ((dev_select[i].selected) &&
909		    (dev_select[i].selected <= maxshowdevs))
910			xo_emit("{T:/%c%c%d} ", dev_select[i].device_name[0],
911			    dev_select[i].device_name[1],
912			    dev_select[i].unit_number);
913	xo_emit("  {T:in}   {T:sy}   {T:cs}");
914	if (Pflag) {
915		for (i = 0; i <= maxid; i++) {
916			if (cpumask & (1ul << i))
917				xo_emit(" {T:us} {T:sy} {T:id}");
918		}
919		xo_emit("\n");
920	} else
921		xo_emit(" {T:us} {T:sy} {T:id}\n");
922	if (wresized != 0)
923		doresize();
924	hdrcnt = winlines;
925}
926
927/*
928 * Force a header to be prepended to the next output.
929 */
930static void
931needhdr(int dummy __unused)
932{
933
934	hdrcnt = 1;
935}
936
937/*
938 * When the terminal is resized, force an update of the maximum number of rows
939 * printed between each header repetition.  Then force a new header to be
940 * prepended to the next output.
941 */
942void
943needresize(int signo __unused)
944{
945
946	wresized = 1;
947	hdrcnt = 1;
948}
949
950/*
951 * Update the global `winlines' count of terminal rows.
952 */
953void
954doresize(void)
955{
956	struct winsize w;
957	int status;
958
959	for (;;) {
960		status = ioctl(fileno(stdout), TIOCGWINSZ, &w);
961		if (status == -1 && errno == EINTR)
962			continue;
963		else if (status == -1)
964			xo_err(1, "ioctl");
965		if (w.ws_row > 3)
966			winlines = w.ws_row - 3;
967		else
968			winlines = VMSTAT_DEFAULT_LINES;
969		break;
970	}
971
972	/*
973	 * Inhibit doresize() calls until we are rescheduled by SIGWINCH.
974	 */
975	wresized = 0;
976}
977
978static long
979pct(long top, long bot)
980{
981	long ans;
982
983	if (bot == 0)
984		return(0);
985	ans = (quad_t)top * 100 / bot;
986	return (ans);
987}
988
989#define	PCT(top, bot) pct((long)(top), (long)(bot))
990
991static void
992dosum(void)
993{
994	struct nchstats lnchstats;
995	size_t size;
996	long nchtotal;
997
998	fill_vmmeter(&sum);
999	xo_open_container("summary-statistics");
1000	xo_emit("{:context-switches/%9u} {N:cpu context switches}\n",
1001	    sum.v_swtch);
1002	xo_emit("{:interrupts/%9u} {N:device interrupts}\n",
1003	    sum.v_intr);
1004	xo_emit("{:software-interrupts/%9u} {N:software interrupts}\n",
1005	    sum.v_soft);
1006	xo_emit("{:traps/%9u} {N:traps}\n", sum.v_trap);
1007	xo_emit("{:system-calls/%9u} {N:system calls}\n",
1008	    sum.v_syscall);
1009	xo_emit("{:kernel-threads/%9u} {N:kernel threads created}\n",
1010	    sum.v_kthreads);
1011	xo_emit("{:forks/%9u} {N: fork() calls}\n", sum.v_forks);
1012	xo_emit("{:vforks/%9u} {N:vfork() calls}\n",
1013	    sum.v_vforks);
1014	xo_emit("{:rforks/%9u} {N:rfork() calls}\n",
1015	    sum.v_rforks);
1016	xo_emit("{:swap-ins/%9u} {N:swap pager pageins}\n",
1017	    sum.v_swapin);
1018	xo_emit("{:swap-in-pages/%9u} {N:swap pager pages paged in}\n",
1019	    sum.v_swappgsin);
1020	xo_emit("{:swap-outs/%9u} {N:swap pager pageouts}\n",
1021	    sum.v_swapout);
1022	xo_emit("{:swap-out-pages/%9u} {N:swap pager pages paged out}\n",
1023	    sum.v_swappgsout);
1024	xo_emit("{:vnode-page-ins/%9u} {N:vnode pager pageins}\n",
1025	    sum.v_vnodein);
1026	xo_emit("{:vnode-page-in-pages/%9u} {N:vnode pager pages paged in}\n",
1027	    sum.v_vnodepgsin);
1028	xo_emit("{:vnode-page-outs/%9u} {N:vnode pager pageouts}\n",
1029	    sum.v_vnodeout);
1030	xo_emit("{:vnode-page-out-pages/%9u} {N:vnode pager pages paged out}\n",
1031	    sum.v_vnodepgsout);
1032	xo_emit("{:page-daemon-wakeups/%9u} {N:page daemon wakeups}\n",
1033	    sum.v_pdwakeups);
1034	xo_emit("{:page-daemon-pages/%9u} {N:pages examined by the page "
1035	    "daemon}\n", sum.v_pdpages);
1036	xo_emit("{:page-reclamation-shortfalls/%9u} {N:clean page reclamation "
1037	    "shortfalls}\n", sum.v_pdshortfalls);
1038	xo_emit("{:reactivated/%9u} {N:pages reactivated by the page daemon}\n",
1039	    sum.v_reactivated);
1040	xo_emit("{:copy-on-write-faults/%9u} {N:copy-on-write faults}\n",
1041	    sum.v_cow_faults);
1042	xo_emit("{:copy-on-write-optimized-faults/%9u} {N:copy-on-write "
1043	    "optimized faults}\n", sum.v_cow_optim);
1044	xo_emit("{:zero-fill-pages/%9u} {N:zero fill pages zeroed}\n",
1045	    sum.v_zfod);
1046	xo_emit("{:zero-fill-prezeroed/%9u} {N:zero fill pages prezeroed}\n",
1047	    sum.v_ozfod);
1048	xo_emit("{:intransit-blocking/%9u} {N:intransit blocking page faults}\n",
1049	    sum.v_intrans);
1050	xo_emit("{:total-faults/%9u} {N:total VM faults taken}\n",
1051	    sum.v_vm_faults);
1052	xo_emit("{:faults-requiring-io/%9u} {N:page faults requiring I\\/O}\n",
1053	    sum.v_io_faults);
1054	xo_emit("{:faults-from-thread-creation/%9u} {N:pages affected by "
1055	    "kernel thread creation}\n", sum.v_kthreadpages);
1056	xo_emit("{:faults-from-fork/%9u} {N:pages affected by  fork}()\n",
1057	    sum.v_forkpages);
1058	xo_emit("{:faults-from-vfork/%9u} {N:pages affected by vfork}()\n",
1059	    sum.v_vforkpages);
1060	xo_emit("{:pages-rfork/%9u} {N:pages affected by rfork}()\n",
1061	    sum.v_rforkpages);
1062	xo_emit("{:pages-freed/%9u} {N:pages freed}\n",
1063	    sum.v_tfree);
1064	xo_emit("{:pages-freed-by-daemon/%9u} {N:pages freed by daemon}\n",
1065	    sum.v_dfree);
1066	xo_emit("{:pages-freed-on-exit/%9u} {N:pages freed by exiting processes}\n",
1067	    sum.v_pfree);
1068	xo_emit("{:active-pages/%9u} {N:pages active}\n",
1069	    sum.v_active_count);
1070	xo_emit("{:inactive-pages/%9u} {N:pages inactive}\n",
1071	    sum.v_inactive_count);
1072	xo_emit("{:laundry-pages/%9u} {N:pages in the laundry queue}\n",
1073	    sum.v_laundry_count);
1074	xo_emit("{:wired-pages/%9u} {N:pages wired down}\n",
1075	    sum.v_wire_count);
1076	xo_emit("{:virtual-user-wired-pages/%9lu} {N:virtual user pages wired "
1077	    "down}\n", sum.v_user_wire_count);
1078	xo_emit("{:free-pages/%9u} {N:pages free}\n",
1079	    sum.v_free_count);
1080	xo_emit("{:bytes-per-page/%9u} {N:bytes per page}\n", sum.v_page_size);
1081	if (kd != NULL) {
1082		kread(X_NCHSTATS, &lnchstats, sizeof(lnchstats));
1083	} else {
1084		size = sizeof(lnchstats);
1085		mysysctl("vfs.cache.nchstats", &lnchstats, &size);
1086		if (size != sizeof(lnchstats))
1087			xo_errx(1, "vfs.cache.nchstats size mismatch");
1088	}
1089	nchtotal = lnchstats.ncs_goodhits + lnchstats.ncs_neghits +
1090	    lnchstats.ncs_badhits + lnchstats.ncs_falsehits +
1091	    lnchstats.ncs_miss + lnchstats.ncs_long;
1092	xo_emit("{:total-name-lookups/%9ld} {N:total name lookups}\n",
1093	    nchtotal);
1094	xo_emit("{P:/%9s} {N:cache hits} "
1095	    "({:positive-cache-hits/%ld}% pos + "
1096	    "{:negative-cache-hits/%ld}% {N:neg}) "
1097	    "system {:cache-hit-percent/%ld}% per-directory\n",
1098	    "", PCT(lnchstats.ncs_goodhits, nchtotal),
1099	    PCT(lnchstats.ncs_neghits, nchtotal),
1100	    PCT(lnchstats.ncs_pass2, nchtotal));
1101	xo_emit("{P:/%9s} {L:deletions} {:deletions/%ld}%, "
1102	    "{L:falsehits} {:false-hits/%ld}%, "
1103	    "{L:toolong} {:too-long/%ld}%\n", "",
1104	    PCT(lnchstats.ncs_badhits, nchtotal),
1105	    PCT(lnchstats.ncs_falsehits, nchtotal),
1106	    PCT(lnchstats.ncs_long, nchtotal));
1107	xo_close_container("summary-statistics");
1108}
1109
1110static void
1111doforkst(void)
1112{
1113
1114	fill_vmmeter(&sum);
1115	xo_open_container("fork-statistics");
1116	xo_emit("{:fork/%u} {N:forks}, {:fork-pages/%u} {N:pages}, "
1117	    "{L:average} {:fork-average/%.2f}\n",
1118	    sum.v_forks, sum.v_forkpages,
1119	    sum.v_forks == 0 ? 0.0 :
1120	    (double)sum.v_forkpages / sum.v_forks);
1121	xo_emit("{:vfork/%u} {N:vforks}, {:vfork-pages/%u} {N:pages}, "
1122	    "{L:average} {:vfork-average/%.2f}\n",
1123	    sum.v_vforks, sum.v_vforkpages,
1124	    sum.v_vforks == 0 ? 0.0 :
1125	    (double)sum.v_vforkpages / sum.v_vforks);
1126	xo_emit("{:rfork/%u} {N:rforks}, {:rfork-pages/%u} {N:pages}, "
1127	    "{L:average} {:rfork-average/%.2f}\n",
1128	    sum.v_rforks, sum.v_rforkpages,
1129	    sum.v_rforks == 0 ? 0.0 :
1130	    (double)sum.v_rforkpages / sum.v_rforks);
1131	xo_close_container("fork-statistics");
1132}
1133
1134static void
1135devstats(void)
1136{
1137	long double busy_seconds, transfers_per_second;
1138	long tmp;
1139	int di, dn, state;
1140
1141	for (state = 0; state < CPUSTATES; ++state) {
1142		tmp = cur.cp_time[state];
1143		cur.cp_time[state] -= last.cp_time[state];
1144		last.cp_time[state] = tmp;
1145	}
1146
1147	busy_seconds = cur.snap_time - last.snap_time;
1148
1149	xo_open_list("device");
1150	for (dn = 0; dn < num_devices; dn++) {
1151		if (dev_select[dn].selected == 0 ||
1152		    dev_select[dn].selected > maxshowdevs)
1153			continue;
1154
1155		di = dev_select[dn].position;
1156
1157		if (devstat_compute_statistics(&cur.dinfo->devices[di],
1158		    &last.dinfo->devices[di], busy_seconds,
1159		    DSM_TRANSFERS_PER_SECOND, &transfers_per_second,
1160		    DSM_NONE) != 0)
1161			xo_errx(1, "%s", devstat_errbuf);
1162
1163		xo_open_instance("device");
1164		xo_emit("{ekq:name/%c%c%d}{:transfers/%3.0Lf} ",
1165		    dev_select[dn].device_name[0],
1166		    dev_select[dn].device_name[1],
1167		    dev_select[dn].unit_number,
1168		    transfers_per_second);
1169		xo_close_instance("device");
1170	}
1171	xo_close_list("device");
1172}
1173
1174static void
1175percent(const char *name, double pctv, int *over)
1176{
1177	int l;
1178	char buf[10];
1179	char fmt[128];
1180
1181	snprintf(fmt, sizeof(fmt), " {:%s/%%*s}", name);
1182	l = snprintf(buf, sizeof(buf), "%.0f", pctv);
1183	if (l == 1 && *over) {
1184		xo_emit(fmt, 1, buf);
1185		(*over)--;
1186	} else
1187		xo_emit(fmt, 2, buf);
1188	if (l > 2)
1189		(*over)++;
1190}
1191
1192static void
1193cpustats(void)
1194{
1195	double lpct, total;
1196	int state, over;
1197
1198	total = 0;
1199	for (state = 0; state < CPUSTATES; ++state)
1200		total += cur.cp_time[state];
1201	if (total > 0)
1202		lpct = 100.0 / total;
1203	else
1204		lpct = 0.0;
1205	over = 0;
1206	xo_open_container("cpu-statistics");
1207	percent("user", (cur.cp_time[CP_USER] + cur.cp_time[CP_NICE]) * lpct,
1208	    &over);
1209	percent("system", (cur.cp_time[CP_SYS] + cur.cp_time[CP_INTR]) * lpct,
1210	    &over);
1211	percent("idle", cur.cp_time[CP_IDLE] * lpct, &over);
1212	xo_close_container("cpu-statistics");
1213}
1214
1215static void
1216pcpustats(u_long cpumask, int maxid)
1217{
1218	double lpct, total;
1219	long tmp;
1220	int i, over, state;
1221
1222	/* devstats does this for cp_time */
1223	for (i = 0; i <= maxid; i++) {
1224		if ((cpumask & (1ul << i)) == 0)
1225			continue;
1226		for (state = 0; state < CPUSTATES; ++state) {
1227			tmp = cur_cp_times[i * CPUSTATES + state];
1228			cur_cp_times[i * CPUSTATES + state] -= last_cp_times[i *
1229			    CPUSTATES + state];
1230			last_cp_times[i * CPUSTATES + state] = tmp;
1231		}
1232	}
1233
1234	over = 0;
1235	xo_open_list("cpu");
1236	for (i = 0; i <= maxid; i++) {
1237		if ((cpumask & (1ul << i)) == 0)
1238			continue;
1239		xo_open_instance("cpu");
1240		xo_emit("{ke:name/%d}", i);
1241		total = 0;
1242		for (state = 0; state < CPUSTATES; ++state)
1243			total += cur_cp_times[i * CPUSTATES + state];
1244		if (total)
1245			lpct = 100.0 / total;
1246		else
1247			lpct = 0.0;
1248		percent("user", (cur_cp_times[i * CPUSTATES + CP_USER] +
1249		    cur_cp_times[i * CPUSTATES + CP_NICE]) * lpct, &over);
1250		percent("system", (cur_cp_times[i * CPUSTATES + CP_SYS] +
1251		    cur_cp_times[i * CPUSTATES + CP_INTR]) * lpct, &over);
1252		percent("idle", cur_cp_times[i * CPUSTATES + CP_IDLE] * lpct,
1253		    &over);
1254		xo_close_instance("cpu");
1255	}
1256	xo_close_list("cpu");
1257}
1258
1259static unsigned int
1260read_intrcnts(unsigned long **intrcnts)
1261{
1262	size_t intrcntlen;
1263	uintptr_t kaddr;
1264
1265	if (kd != NULL) {
1266		kread(X_SINTRCNT, &intrcntlen, sizeof(intrcntlen));
1267		if ((*intrcnts = malloc(intrcntlen)) == NULL)
1268			err(1, "malloc()");
1269		if (namelist[X_NINTRCNT].n_type == 0)
1270			kread(X_INTRCNT, *intrcnts, intrcntlen);
1271		else {
1272			kread(X_INTRCNT, &kaddr, sizeof(kaddr));
1273			kreadptr(kaddr, *intrcnts, intrcntlen);
1274		}
1275	} else {
1276		for (*intrcnts = NULL, intrcntlen = 1024; ; intrcntlen *= 2) {
1277			*intrcnts = reallocf(*intrcnts, intrcntlen);
1278			if (*intrcnts == NULL)
1279				err(1, "reallocf()");
1280			if (mysysctl("hw.intrcnt", *intrcnts, &intrcntlen) == 0)
1281				break;
1282		}
1283	}
1284
1285	return (intrcntlen / sizeof(unsigned long));
1286}
1287
1288static void
1289print_intrcnts(unsigned long *intrcnts, unsigned long *old_intrcnts,
1290    char *intrnames, unsigned int nintr, size_t istrnamlen, long long period_ms)
1291{
1292	unsigned long *intrcnt, *old_intrcnt;
1293	char *intrname;
1294	uint64_t inttotal, old_inttotal, total_count, total_rate;
1295	unsigned long count, rate;
1296	unsigned int i;
1297
1298	inttotal = 0;
1299	old_inttotal = 0;
1300	intrname = intrnames;
1301	xo_open_list("interrupt");
1302	for (i = 0, intrcnt=intrcnts, old_intrcnt=old_intrcnts; i < nintr; i++) {
1303		if (intrname[0] != '\0' && (*intrcnt != 0 || aflag)) {
1304			count = *intrcnt - *old_intrcnt;
1305			rate = ((uint64_t)count * 1000 + period_ms / 2) / period_ms;
1306			xo_open_instance("interrupt");
1307			xo_emit("{d:name/%-*s}{ket:name/%s} "
1308			    "{:total/%20lu} {:rate/%10lu}\n",
1309			    (int)istrnamlen, intrname, intrname, count, rate);
1310			xo_close_instance("interrupt");
1311		}
1312		intrname += strlen(intrname) + 1;
1313		inttotal += *intrcnt++;
1314		old_inttotal += *old_intrcnt++;
1315	}
1316	total_count = inttotal - old_inttotal;
1317	total_rate = (total_count * 1000 + period_ms / 2) / period_ms;
1318	xo_close_list("interrupt");
1319	xo_emit("{L:/%-*s} {:total-interrupts/%20ju} "
1320	    "{:total-rate/%10ju}\n", (int)istrnamlen,
1321	    "Total", (uintmax_t)total_count, (uintmax_t)total_rate);
1322}
1323
1324static void
1325dointr(unsigned int interval, int reps)
1326{
1327	unsigned long *intrcnts, *old_intrcnts;
1328	char *intrname, *intrnames;
1329	long long period_ms, old_uptime, uptime;
1330	size_t clen, inamlen, istrnamlen;
1331	uintptr_t kaddr;
1332	unsigned int nintr;
1333
1334	old_intrcnts = NULL;
1335	uptime = getuptime();
1336
1337	/* Get the names of each interrupt source */
1338	if (kd != NULL) {
1339		kread(X_SINTRNAMES, &inamlen, sizeof(inamlen));
1340		if ((intrnames = malloc(inamlen)) == NULL)
1341			xo_err(1, "malloc()");
1342		if (namelist[X_NINTRCNT].n_type == 0)
1343			kread(X_INTRNAMES, intrnames, inamlen);
1344		else {
1345			kread(X_INTRNAMES, &kaddr, sizeof(kaddr));
1346			kreadptr(kaddr, intrnames, inamlen);
1347		}
1348	} else {
1349		for (intrnames = NULL, inamlen = 1024; ; inamlen *= 2) {
1350			if ((intrnames = reallocf(intrnames, inamlen)) == NULL)
1351				xo_err(1, "reallocf()");
1352			if (mysysctl("hw.intrnames", intrnames, &inamlen) == 0)
1353				break;
1354		}
1355	}
1356
1357	/* Determine the length of the longest interrupt name */
1358	intrname = intrnames;
1359	istrnamlen = strlen("interrupt");
1360	while(*intrname != '\0') {
1361		clen = strlen(intrname);
1362		if (clen > istrnamlen)
1363			istrnamlen = clen;
1364		intrname += strlen(intrname) + 1;
1365	}
1366	xo_emit("{T:/%-*s} {T:/%20s} {T:/%10s}\n",
1367	    (int)istrnamlen, "interrupt", "total", "rate");
1368
1369	/*
1370	 * Loop reps times printing differential interrupt counts.  If reps is
1371	 * zero, then run just once, printing total counts
1372	 */
1373	xo_open_container("interrupt-statistics");
1374
1375	period_ms = uptime / 1000000;
1376	while(1) {
1377		nintr = read_intrcnts(&intrcnts);
1378		/*
1379		 * Initialize old_intrcnts to 0 for the first pass, so
1380		 * print_intrcnts will print total interrupts since boot
1381		 */
1382		if (old_intrcnts == NULL) {
1383			old_intrcnts = calloc(nintr, sizeof(unsigned long));
1384			if (old_intrcnts == NULL)
1385				xo_err(1, "calloc()");
1386		}
1387
1388		print_intrcnts(intrcnts, old_intrcnts, intrnames, nintr,
1389		    istrnamlen, period_ms);
1390		xo_flush();
1391
1392		free(old_intrcnts);
1393		old_intrcnts = intrcnts;
1394		if (reps >= 0 && --reps <= 0)
1395			break;
1396		usleep(interval * 1000);
1397		old_uptime = uptime;
1398		uptime = getuptime();
1399		period_ms = (uptime - old_uptime) / 1000000;
1400	}
1401
1402	xo_close_container("interrupt-statistics");
1403}
1404
1405static void
1406domemstat_malloc(void)
1407{
1408	struct memory_type_list *mtlp;
1409	struct memory_type *mtp;
1410	size_t i, zones;
1411	int error, first;
1412
1413	mtlp = memstat_mtl_alloc();
1414	if (mtlp == NULL) {
1415		xo_warn("memstat_mtl_alloc");
1416		return;
1417	}
1418	if (kd == NULL) {
1419		if (memstat_sysctl_malloc(mtlp, 0) < 0) {
1420			xo_warnx("memstat_sysctl_malloc: %s",
1421			    memstat_strerror(memstat_mtl_geterror(mtlp)));
1422			return;
1423		}
1424	} else {
1425		if (memstat_kvm_malloc(mtlp, kd) < 0) {
1426			error = memstat_mtl_geterror(mtlp);
1427			if (error == MEMSTAT_ERROR_KVM)
1428				xo_warnx("memstat_kvm_malloc: %s",
1429				    kvm_geterr(kd));
1430			else
1431				xo_warnx("memstat_kvm_malloc: %s",
1432				    memstat_strerror(error));
1433		}
1434	}
1435	xo_open_container("malloc-statistics");
1436	xo_emit("{T:/%13s} {T:/%5s} {T:/%6s} {T:/%8s}  {T:Size(s)}\n",
1437	    "Type", "InUse", "MemUse", "Requests");
1438	xo_open_list("memory");
1439	zones = memstat_malloc_zone_get_count();
1440	for (mtp = memstat_mtl_first(mtlp); mtp != NULL;
1441	    mtp = memstat_mtl_next(mtp)) {
1442		if (memstat_get_numallocs(mtp) == 0 &&
1443		    memstat_get_count(mtp) == 0)
1444			continue;
1445		xo_open_instance("memory");
1446		xo_emit("{k:type/%13s/%s} {:in-use/%5ju} "
1447		    "{:memory-use/%5ju}{U:K} {:requests/%8ju}  ",
1448		    memstat_get_name(mtp), (uintmax_t)memstat_get_count(mtp),
1449		    ((uintmax_t)memstat_get_bytes(mtp) + 1023) / 1024,
1450		    (uintmax_t)memstat_get_numallocs(mtp));
1451		first = 1;
1452		xo_open_list("size");
1453		for (i = 0; i < zones; i++) {
1454			if (memstat_malloc_zone_used(mtp, i)) {
1455				if (!first)
1456					xo_emit(",");
1457				xo_emit("{l:size/%d}", memstat_malloc_zone_get_size(i));
1458				first = 0;
1459			}
1460		}
1461		xo_close_list("size");
1462		xo_close_instance("memory");
1463		xo_emit("\n");
1464	}
1465	xo_close_list("memory");
1466	xo_close_container("malloc-statistics");
1467	memstat_mtl_free(mtlp);
1468}
1469
1470static void
1471domemstat_zone(void)
1472{
1473	struct memory_type_list *mtlp;
1474	struct memory_type *mtp;
1475	int error;
1476	char name[MEMTYPE_MAXNAME + 1];
1477
1478	mtlp = memstat_mtl_alloc();
1479	if (mtlp == NULL) {
1480		xo_warn("memstat_mtl_alloc");
1481		return;
1482	}
1483	if (kd == NULL) {
1484		if (memstat_sysctl_uma(mtlp, 0) < 0) {
1485			xo_warnx("memstat_sysctl_uma: %s",
1486			    memstat_strerror(memstat_mtl_geterror(mtlp)));
1487			return;
1488		}
1489	} else {
1490		if (memstat_kvm_uma(mtlp, kd) < 0) {
1491			error = memstat_mtl_geterror(mtlp);
1492			if (error == MEMSTAT_ERROR_KVM)
1493				xo_warnx("memstat_kvm_uma: %s",
1494				    kvm_geterr(kd));
1495			else
1496				xo_warnx("memstat_kvm_uma: %s",
1497				    memstat_strerror(error));
1498		}
1499	}
1500	xo_open_container("memory-zone-statistics");
1501	xo_emit("{T:/%-20s} {T:/%6s} {T:/%6s} {T:/%8s} {T:/%8s} {T:/%8s} {T:/%8s}"
1502	    "{T:/%4s} {T:/%4s}\n", "ITEM", "SIZE",
1503	    "LIMIT", "USED", "FREE", "REQ", "FAIL", "SLEEP", "XDOMAIN");
1504	xo_open_list("zone");
1505	for (mtp = memstat_mtl_first(mtlp); mtp != NULL;
1506	    mtp = memstat_mtl_next(mtp)) {
1507		strlcpy(name, memstat_get_name(mtp), MEMTYPE_MAXNAME);
1508		strcat(name, ":");
1509		xo_open_instance("zone");
1510		xo_emit("{d:name/%-20s}{ke:name/%s} {:size/%6ju}, "
1511		    "{:limit/%6ju},{:used/%8ju},"
1512		    "{:free/%8ju},{:requests/%8ju},"
1513		    "{:fail/%4ju},{:sleep/%4ju},{:xdomain/%4ju}\n", name,
1514		    memstat_get_name(mtp),
1515		    (uintmax_t)memstat_get_size(mtp),
1516		    (uintmax_t)memstat_get_countlimit(mtp),
1517		    (uintmax_t)memstat_get_count(mtp),
1518		    (uintmax_t)memstat_get_free(mtp),
1519		    (uintmax_t)memstat_get_numallocs(mtp),
1520		    (uintmax_t)memstat_get_failures(mtp),
1521		    (uintmax_t)memstat_get_sleeps(mtp),
1522		    (uintmax_t)memstat_get_xdomain(mtp));
1523		xo_close_instance("zone");
1524	}
1525	memstat_mtl_free(mtlp);
1526	xo_close_list("zone");
1527	xo_close_container("memory-zone-statistics");
1528}
1529
1530static void
1531display_object(struct kinfo_vmobject *kvo)
1532{
1533	const char *str;
1534
1535	xo_open_instance("object");
1536	xo_emit("{:resident/%5ju} ", (uintmax_t)kvo->kvo_resident);
1537	xo_emit("{:active/%5ju} ", (uintmax_t)kvo->kvo_active);
1538	xo_emit("{:inactive/%5ju} ", (uintmax_t)kvo->kvo_inactive);
1539	xo_emit("{:refcount/%3d} ", kvo->kvo_ref_count);
1540	xo_emit("{:shadowcount/%3d} ", kvo->kvo_shadow_count);
1541
1542#define	MEMATTR_STR(type, val)					\
1543	if (kvo->kvo_memattr == (type)) {			\
1544		str = (val);					\
1545	} else
1546#ifdef VM_MEMATTR_UNCACHEABLE
1547	MEMATTR_STR(VM_MEMATTR_UNCACHEABLE, "UC")
1548#endif
1549#ifdef VM_MEMATTR_WRITE_COMBINING
1550	MEMATTR_STR(VM_MEMATTR_WRITE_COMBINING, "WC")
1551#endif
1552#ifdef VM_MEMATTR_WRITE_THROUGH
1553	MEMATTR_STR(VM_MEMATTR_WRITE_THROUGH, "WT")
1554#endif
1555#ifdef VM_MEMATTR_WRITE_PROTECTED
1556	MEMATTR_STR(VM_MEMATTR_WRITE_PROTECTED, "WP")
1557#endif
1558#ifdef VM_MEMATTR_WRITE_BACK
1559	MEMATTR_STR(VM_MEMATTR_WRITE_BACK, "WB")
1560#endif
1561#ifdef VM_MEMATTR_WEAK_UNCACHEABLE
1562	MEMATTR_STR(VM_MEMATTR_WEAK_UNCACHEABLE, "UC-")
1563#endif
1564#ifdef VM_MEMATTR_WB_WA
1565	MEMATTR_STR(VM_MEMATTR_WB_WA, "WB")
1566#endif
1567#ifdef VM_MEMATTR_NOCACHE
1568	MEMATTR_STR(VM_MEMATTR_NOCACHE, "NC")
1569#endif
1570#ifdef VM_MEMATTR_DEVICE
1571	MEMATTR_STR(VM_MEMATTR_DEVICE, "DEV")
1572#endif
1573#ifdef VM_MEMATTR_CACHEABLE
1574	MEMATTR_STR(VM_MEMATTR_CACHEABLE, "C")
1575#endif
1576#ifdef VM_MEMATTR_PREFETCHABLE
1577	MEMATTR_STR(VM_MEMATTR_PREFETCHABLE, "PRE")
1578#endif
1579	{
1580		str = "??";
1581	}
1582#undef MEMATTR_STR
1583	xo_emit("{:attribute/%-3s} ", str);
1584	switch (kvo->kvo_type) {
1585	case KVME_TYPE_NONE:
1586		str = "--";
1587		break;
1588	case KVME_TYPE_DEFAULT:
1589		str = "df";
1590		break;
1591	case KVME_TYPE_VNODE:
1592		str = "vn";
1593		break;
1594	case KVME_TYPE_SWAP:
1595		str = "sw";
1596		break;
1597	case KVME_TYPE_DEVICE:
1598		str = "dv";
1599		break;
1600	case KVME_TYPE_PHYS:
1601		str = "ph";
1602		break;
1603	case KVME_TYPE_DEAD:
1604		str = "dd";
1605		break;
1606	case KVME_TYPE_SG:
1607		str = "sg";
1608		break;
1609	case KVME_TYPE_MGTDEVICE:
1610		str = "md";
1611		break;
1612	case KVME_TYPE_UNKNOWN:
1613	default:
1614		str = "??";
1615		break;
1616	}
1617	xo_emit("{:type/%-2s} ", str);
1618	xo_emit("{:path/%-s}\n", kvo->kvo_path);
1619	xo_close_instance("object");
1620}
1621
1622static void
1623doobjstat(void)
1624{
1625	struct kinfo_vmobject *kvo;
1626	int cnt, i;
1627
1628	kvo = kinfo_getvmobject(&cnt);
1629	if (kvo == NULL) {
1630		xo_warn("Failed to fetch VM object list");
1631		return;
1632	}
1633	xo_emit("{T:RES/%5s} {T:ACT/%5s} {T:INACT/%5s} {T:REF/%3s} {T:SHD/%3s} "
1634	    "{T:CM/%3s} {T:TP/%2s} {T:PATH/%s}\n");
1635	xo_open_list("object");
1636	for (i = 0; i < cnt; i++)
1637		display_object(&kvo[i]);
1638	free(kvo);
1639	xo_close_list("object");
1640}
1641
1642/*
1643 * kread reads something from the kernel, given its nlist index.
1644 */
1645static void
1646kreado(int nlx, void *addr, size_t size, size_t offset)
1647{
1648	const char *sym;
1649
1650	if (namelist[nlx].n_type == 0 || namelist[nlx].n_value == 0) {
1651		sym = namelist[nlx].n_name;
1652		if (*sym == '_')
1653			++sym;
1654		xo_errx(1, "symbol %s not defined", sym);
1655	}
1656	if ((size_t)kvm_read(kd, namelist[nlx].n_value + offset, addr,
1657	    size) != size) {
1658		sym = namelist[nlx].n_name;
1659		if (*sym == '_')
1660			++sym;
1661		xo_errx(1, "%s: %s", sym, kvm_geterr(kd));
1662	}
1663}
1664
1665static void
1666kread(int nlx, void *addr, size_t size)
1667{
1668
1669	kreado(nlx, addr, size, 0);
1670}
1671
1672static void
1673kreadptr(uintptr_t addr, void *buf, size_t size)
1674{
1675
1676	if ((size_t)kvm_read(kd, addr, buf, size) != size)
1677		xo_errx(1, "%s", kvm_geterr(kd));
1678}
1679
1680static void __dead2
1681usage(void)
1682{
1683	xo_error("%s%s",
1684	    "usage: vmstat [-afHhimoPsz] [-M core [-N system]] [-c count] [-n devs]\n",
1685	    "              [-p type,if,pass] [-w wait] [disks] [wait [count]]\n");
1686	xo_finish();
1687	exit(1);
1688}
1689