xref: /freebsd/usr.bin/vmstat/vmstat.c (revision 525fe93dc7487a1e63a90f6a2b956abc601963c1)
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 #include <sys/param.h>
33 #include <sys/proc.h>
34 #include <sys/uio.h>
35 #include <sys/namei.h>
36 #include <sys/malloc.h>
37 #include <sys/signal.h>
38 #include <sys/fcntl.h>
39 #include <sys/ioctl.h>
40 #include <sys/resource.h>
41 #include <sys/sysctl.h>
42 #include <sys/time.h>
43 #include <sys/user.h>
44 #define	_WANT_VMMETER
45 #include <sys/vmmeter.h>
46 #include <sys/pcpu.h>
47 
48 #include <vm/vm_param.h>
49 
50 #include <ctype.h>
51 #include <devstat.h>
52 #include <err.h>
53 #include <errno.h>
54 #include <inttypes.h>
55 #include <kvm.h>
56 #include <limits.h>
57 #include <memstat.h>
58 #include <nlist.h>
59 #include <paths.h>
60 #include <stdio.h>
61 #include <stdlib.h>
62 #include <string.h>
63 #include <sysexits.h>
64 #include <time.h>
65 #include <unistd.h>
66 #include <libutil.h>
67 #include <libxo/xo.h>
68 
69 #define VMSTAT_XO_VERSION "1"
70 
71 static char da[] = "da";
72 
73 enum x_stats { X_SUM, X_HZ, X_STATHZ, X_NCHSTATS, X_INTRNAMES, X_SINTRNAMES,
74     X_INTRCNT, X_SINTRCNT, X_NINTRCNT };
75 
76 static struct nlist namelist[] = {
77 	[X_SUM] = { .n_name = "_vm_cnt", },
78 	[X_HZ] = { .n_name = "_hz", },
79 	[X_STATHZ] = { .n_name = "_stathz", },
80 	[X_NCHSTATS] = { .n_name = "_nchstats", },
81 	[X_INTRNAMES] = { .n_name = "_intrnames", },
82 	[X_SINTRNAMES] = { .n_name = "_sintrnames", },
83 	[X_INTRCNT] = { .n_name = "_intrcnt", },
84 	[X_SINTRCNT] = { .n_name = "_sintrcnt", },
85 	[X_NINTRCNT] = { .n_name = "_nintrcnt", },
86 	{ .n_name = NULL, },
87 };
88 
89 static struct devstat_match *matches;
90 static struct device_selection *dev_select;
91 static struct statinfo cur, last;
92 static devstat_select_mode select_mode;
93 static size_t size_cp_times;
94 static long *cur_cp_times, *last_cp_times;
95 static long generation, select_generation;
96 static int hz, hdrcnt, maxshowdevs;
97 static int num_devices, num_devices_specified;
98 static int num_matches, num_selected, num_selections;
99 static char **specified_devices;
100 
101 static struct __vmmeter {
102 	uint64_t v_swtch;
103 	uint64_t v_trap;
104 	uint64_t v_syscall;
105 	uint64_t v_intr;
106 	uint64_t v_soft;
107 	uint64_t v_vm_faults;
108 	uint64_t v_io_faults;
109 	uint64_t v_cow_faults;
110 	uint64_t v_cow_optim;
111 	uint64_t v_zfod;
112 	uint64_t v_ozfod;
113 	uint64_t v_swapin;
114 	uint64_t v_swapout;
115 	uint64_t v_swappgsin;
116 	uint64_t v_swappgsout;
117 	uint64_t v_vnodein;
118 	uint64_t v_vnodeout;
119 	uint64_t v_vnodepgsin;
120 	uint64_t v_vnodepgsout;
121 	uint64_t v_intrans;
122 	uint64_t v_reactivated;
123 	uint64_t v_pdwakeups;
124 	uint64_t v_pdpages;
125 	uint64_t v_pdshortfalls;
126 	uint64_t v_dfree;
127 	uint64_t v_pfree;
128 	uint64_t v_tfree;
129 	uint64_t v_forks;
130 	uint64_t v_vforks;
131 	uint64_t v_rforks;
132 	uint64_t v_kthreads;
133 	uint64_t v_forkpages;
134 	uint64_t v_vforkpages;
135 	uint64_t v_rforkpages;
136 	uint64_t v_kthreadpages;
137 	u_int v_page_size;
138 	u_int v_page_count;
139 	u_int v_free_reserved;
140 	u_int v_free_target;
141 	u_int v_free_min;
142 	u_int v_free_count;
143 	u_int v_wire_count;
144 	u_long v_user_wire_count;
145 	u_int v_active_count;
146 	u_int v_inactive_target;
147 	u_int v_inactive_count;
148 	u_int v_laundry_count;
149 	u_int v_pageout_free_min;
150 	u_int v_interrupt_free_min;
151 	u_int v_free_severe;
152 } sum, osum;
153 
154 #define	VMSTAT_DEFAULT_LINES	20	/* Default number of `winlines'. */
155 static volatile sig_atomic_t wresized;		/* Tty resized when non-zero. */
156 static int winlines = VMSTAT_DEFAULT_LINES; /* Current number of tty rows. */
157 
158 static int	aflag;
159 static int	nflag;
160 static int	Pflag;
161 static int	hflag;
162 
163 static kvm_t	*kd;
164 
165 #define	FORKSTAT	0x01
166 #define	INTRSTAT	0x02
167 #define	MEMSTAT		0x04
168 #define	SUMSTAT		0x08
169 #define	TIMESTAT	0x10
170 #define	VMSTAT		0x20
171 #define	ZMEMSTAT	0x40
172 #define	OBJSTAT		0x80
173 
174 static void	cpustats(void);
175 static void	pcpustats(u_long, int);
176 static void	devstats(void);
177 static void	doforkst(void);
178 static void	dointr(unsigned int, int);
179 static void	doobjstat(void);
180 static void	dosum(void);
181 static void	dovmstat(unsigned int, int);
182 static void	domemstat_malloc(void);
183 static void	domemstat_zone(void);
184 static void	kread(int, void *, size_t);
185 static void	kreado(int, void *, size_t, size_t);
186 static void	kreadptr(uintptr_t, void *, size_t);
187 static void	needhdr(int);
188 static void	needresize(int);
189 static void	doresize(void);
190 static void	printhdr(int, u_long);
191 static void	usage(void);
192 
193 static long	pct(long, long);
194 static long long	getuptime(void);
195 
196 static char	**getdrivedata(char **);
197 
198 int
199 main(int argc, char *argv[])
200 {
201 	char *bp, *buf, *memf, *nlistf;
202 	float f;
203 	int bufsize, c, reps, todo;
204 	size_t len;
205 	unsigned int interval;
206 	char errbuf[_POSIX2_LINE_MAX];
207 
208 	memf = nlistf = NULL;
209 	interval = reps = todo = 0;
210 	maxshowdevs = 2;
211 
212 	argc = xo_parse_args(argc, argv);
213 	if (argc < 0)
214 		return (argc);
215 
216 	hflag = (xo_get_style(NULL) == XO_STYLE_TEXT) && isatty(1);
217 
218 	while ((c = getopt(argc, argv, "ac:fhHiM:mN:n:oPp:sw:z")) != -1) {
219 		switch (c) {
220 		case 'a':
221 			aflag++;
222 			break;
223 		case 'c':
224 			reps = atoi(optarg);
225 			break;
226 		case 'P':
227 			Pflag++;
228 			break;
229 		case 'f':
230 			todo |= FORKSTAT;
231 			break;
232 		case 'h':
233 			hflag = 1;
234 			break;
235 		case 'H':
236 			hflag = 0;
237 			break;
238 		case 'i':
239 			todo |= INTRSTAT;
240 			break;
241 		case 'M':
242 			memf = optarg;
243 			break;
244 		case 'm':
245 			todo |= MEMSTAT;
246 			break;
247 		case 'N':
248 			nlistf = optarg;
249 			break;
250 		case 'n':
251 			nflag = 1;
252 			maxshowdevs = atoi(optarg);
253 			if (maxshowdevs < 0)
254 				xo_errx(1, "number of devices %d is < 0",
255 				    maxshowdevs);
256 			break;
257 		case 'o':
258 			todo |= OBJSTAT;
259 			break;
260 		case 'p':
261 			if (devstat_buildmatch(optarg, &matches, &num_matches)
262 			    != 0)
263 				xo_errx(1, "%s", devstat_errbuf);
264 			break;
265 		case 's':
266 			todo |= SUMSTAT;
267 			break;
268 		case 'w':
269 			/* Convert to milliseconds. */
270 			f = atof(optarg);
271 			interval = f * 1000;
272 			break;
273 		case 'z':
274 			todo |= ZMEMSTAT;
275 			break;
276 		case '?':
277 		default:
278 			usage();
279 		}
280 	}
281 	argc -= optind;
282 	argv += optind;
283 
284 	xo_set_version(VMSTAT_XO_VERSION);
285 	if (todo == 0)
286 		todo = VMSTAT;
287 
288 	if (memf != NULL) {
289 		kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, errbuf);
290 		if (kd == NULL)
291 			xo_errx(1, "kvm_openfiles: %s", errbuf);
292 	}
293 
294 retry_nlist:
295 	if (kd != NULL && (c = kvm_nlist(kd, namelist)) != 0) {
296 		if (c > 0) {
297 			bufsize = 0;
298 			len = 0;
299 
300 			/*
301 			 * 'cnt' was renamed to 'vm_cnt'.  If 'vm_cnt' is not
302 			 * found try looking up older 'cnt' symbol.
303 			 * */
304 			if (namelist[X_SUM].n_type == 0 &&
305 			    strcmp(namelist[X_SUM].n_name, "_vm_cnt") == 0) {
306 				namelist[X_SUM].n_name = "_cnt";
307 				goto retry_nlist;
308 			}
309 
310 			/*
311 			 * 'nintrcnt' doesn't exist in older kernels, but
312 			 * that isn't fatal.
313 			 */
314 			if (namelist[X_NINTRCNT].n_type == 0 && c == 1)
315 				goto nlist_ok;
316 
317 			for (c = 0; c < (int)(nitems(namelist)); c++)
318 				if (namelist[c].n_type == 0)
319 					bufsize += strlen(namelist[c].n_name)
320 					    + 1;
321 			bufsize += len + 1;
322 			buf = bp = alloca(bufsize);
323 
324 			for (c = 0; c < (int)(nitems(namelist)); c++)
325 				if (namelist[c].n_type == 0) {
326 					xo_error(" %s",
327 					    namelist[c].n_name);
328 					len = strlen(namelist[c].n_name);
329 					*bp++ = ' ';
330 					memcpy(bp, namelist[c].n_name, len);
331 					bp += len;
332 				}
333 			*bp = '\0';
334 			xo_error("undefined symbols:\n", buf);
335 		} else
336 			xo_warnx("kvm_nlist: %s", kvm_geterr(kd));
337 		xo_finish();
338 		exit(1);
339 	}
340 nlist_ok:
341 	if (kd && Pflag)
342 		xo_errx(1, "Cannot use -P with crash dumps");
343 
344 	if (todo & VMSTAT) {
345 		/*
346 		 * Make sure that the userland devstat version matches the
347 		 * kernel devstat version.  If not, exit and print a
348 		 * message informing the user of his mistake.
349 		 */
350 		if (devstat_checkversion(NULL) < 0)
351 			xo_errx(1, "%s", devstat_errbuf);
352 
353 
354 		argv = getdrivedata(argv);
355 	}
356 
357 	if (*argv) {
358 		f = atof(*argv);
359 		interval = f * 1000;
360 		if (*++argv)
361 			reps = atoi(*argv);
362 	}
363 
364 	if (interval) {
365 		if (!reps)
366 			reps = -1;
367 	} else if (reps)
368 		interval = 1 * 1000;
369 
370 	if (todo & FORKSTAT)
371 		doforkst();
372 	if (todo & MEMSTAT)
373 		domemstat_malloc();
374 	if (todo & ZMEMSTAT)
375 		domemstat_zone();
376 	if (todo & SUMSTAT)
377 		dosum();
378 	if (todo & OBJSTAT)
379 		doobjstat();
380 	if (todo & INTRSTAT)
381 		dointr(interval, reps);
382 	if (todo & VMSTAT)
383 		dovmstat(interval, reps);
384 	xo_finish();
385 	exit(0);
386 }
387 
388 static int
389 mysysctl(const char *name, void *oldp, size_t *oldlenp)
390 {
391 	int error;
392 
393 	error = sysctlbyname(name, oldp, oldlenp, NULL, 0);
394 	if (error != 0 && errno != ENOMEM)
395 		xo_err(1, "sysctl(%s)", name);
396 	return (error);
397 }
398 
399 static char **
400 getdrivedata(char **argv)
401 {
402 
403 	if ((num_devices = devstat_getnumdevs(NULL)) < 0)
404 		xo_errx(1, "%s", devstat_errbuf);
405 
406 	cur.dinfo = (struct devinfo *)calloc(1, sizeof(struct devinfo));
407 	last.dinfo = (struct devinfo *)calloc(1, sizeof(struct devinfo));
408 
409 	if (devstat_getdevs(NULL, &cur) == -1)
410 		xo_errx(1, "%s", devstat_errbuf);
411 
412 	num_devices = cur.dinfo->numdevs;
413 	generation = cur.dinfo->generation;
414 
415 	specified_devices = malloc(sizeof(char *));
416 	for (num_devices_specified = 0; *argv; ++argv) {
417 		if (isdigit(**argv))
418 			break;
419 		num_devices_specified++;
420 		specified_devices = reallocf(specified_devices,
421 		    sizeof(char *) * num_devices_specified);
422 		if (specified_devices == NULL) {
423 			xo_errx(1, "%s", "reallocf (specified_devices)");
424 		}
425 		specified_devices[num_devices_specified - 1] = *argv;
426 	}
427 	dev_select = NULL;
428 
429 	if (nflag == 0 && maxshowdevs < num_devices_specified)
430 		maxshowdevs = num_devices_specified;
431 
432 	/*
433 	 * People are generally only interested in disk statistics when
434 	 * they're running vmstat.  So, that's what we're going to give
435 	 * them if they don't specify anything by default.  We'll also give
436 	 * them any other random devices in the system so that we get to
437 	 * maxshowdevs devices, if that many devices exist.  If the user
438 	 * specifies devices on the command line, either through a pattern
439 	 * match or by naming them explicitly, we will give the user only
440 	 * those devices.
441 	 */
442 	if ((num_devices_specified == 0) && (num_matches == 0)) {
443 		if (devstat_buildmatch(da, &matches, &num_matches) != 0)
444 			xo_errx(1, "%s", devstat_errbuf);
445 		select_mode = DS_SELECT_ADD;
446 	} else
447 		select_mode = DS_SELECT_ONLY;
448 
449 	/*
450 	 * At this point, selectdevs will almost surely indicate that the
451 	 * device list has changed, so we don't look for return values of 0
452 	 * or 1.  If we get back -1, though, there is an error.
453 	 */
454 	if (devstat_selectdevs(&dev_select, &num_selected, &num_selections,
455 	    &select_generation, generation, cur.dinfo->devices,
456 	    num_devices, matches, num_matches, specified_devices,
457 	    num_devices_specified, select_mode,
458 	    maxshowdevs, 0) == -1)
459 		xo_errx(1, "%s", devstat_errbuf);
460 
461 	return(argv);
462 }
463 
464 /* Return system uptime in nanoseconds */
465 static long long
466 getuptime(void)
467 {
468 	struct timespec sp;
469 
470 	(void)clock_gettime(CLOCK_UPTIME, &sp);
471 	return((long long)sp.tv_sec * 1000000000LL + sp.tv_nsec);
472 }
473 
474 static void
475 fill_vmmeter(struct __vmmeter *vmmp)
476 {
477 	struct vmmeter vm_cnt;
478 	size_t size;
479 
480 	if (kd != NULL) {
481 		kread(X_SUM, &vm_cnt, sizeof(vm_cnt));
482 #define	GET_COUNTER(name) \
483 		vmmp->name = kvm_counter_u64_fetch(kd, (u_long)vm_cnt.name)
484 		GET_COUNTER(v_swtch);
485 		GET_COUNTER(v_trap);
486 		GET_COUNTER(v_syscall);
487 		GET_COUNTER(v_intr);
488 		GET_COUNTER(v_soft);
489 		GET_COUNTER(v_vm_faults);
490 		GET_COUNTER(v_io_faults);
491 		GET_COUNTER(v_cow_faults);
492 		GET_COUNTER(v_cow_optim);
493 		GET_COUNTER(v_zfod);
494 		GET_COUNTER(v_ozfod);
495 		GET_COUNTER(v_swapin);
496 		GET_COUNTER(v_swapout);
497 		GET_COUNTER(v_swappgsin);
498 		GET_COUNTER(v_swappgsout);
499 		GET_COUNTER(v_vnodein);
500 		GET_COUNTER(v_vnodeout);
501 		GET_COUNTER(v_vnodepgsin);
502 		GET_COUNTER(v_vnodepgsout);
503 		GET_COUNTER(v_intrans);
504 		GET_COUNTER(v_tfree);
505 		GET_COUNTER(v_forks);
506 		GET_COUNTER(v_vforks);
507 		GET_COUNTER(v_rforks);
508 		GET_COUNTER(v_kthreads);
509 		GET_COUNTER(v_forkpages);
510 		GET_COUNTER(v_vforkpages);
511 		GET_COUNTER(v_rforkpages);
512 		GET_COUNTER(v_kthreadpages);
513 #undef GET_COUNTER
514 	} else {
515 #define GET_VM_STATS(cat, name)	do {					\
516 	size = sizeof(vmmp->name);					\
517 	mysysctl("vm.stats." #cat "." #name, &vmmp->name, &size);	\
518 } while (0)
519 		/* sys */
520 		GET_VM_STATS(sys, v_swtch);
521 		GET_VM_STATS(sys, v_trap);
522 		GET_VM_STATS(sys, v_syscall);
523 		GET_VM_STATS(sys, v_intr);
524 		GET_VM_STATS(sys, v_soft);
525 
526 		/* vm */
527 		GET_VM_STATS(vm, v_vm_faults);
528 		GET_VM_STATS(vm, v_io_faults);
529 		GET_VM_STATS(vm, v_cow_faults);
530 		GET_VM_STATS(vm, v_cow_optim);
531 		GET_VM_STATS(vm, v_zfod);
532 		GET_VM_STATS(vm, v_ozfod);
533 		GET_VM_STATS(vm, v_swapin);
534 		GET_VM_STATS(vm, v_swapout);
535 		GET_VM_STATS(vm, v_swappgsin);
536 		GET_VM_STATS(vm, v_swappgsout);
537 		GET_VM_STATS(vm, v_vnodein);
538 		GET_VM_STATS(vm, v_vnodeout);
539 		GET_VM_STATS(vm, v_vnodepgsin);
540 		GET_VM_STATS(vm, v_vnodepgsout);
541 		GET_VM_STATS(vm, v_intrans);
542 		GET_VM_STATS(vm, v_reactivated);
543 		GET_VM_STATS(vm, v_pdwakeups);
544 		GET_VM_STATS(vm, v_pdpages);
545 		GET_VM_STATS(vm, v_pdshortfalls);
546 		GET_VM_STATS(vm, v_dfree);
547 		GET_VM_STATS(vm, v_pfree);
548 		GET_VM_STATS(vm, v_tfree);
549 		GET_VM_STATS(vm, v_page_size);
550 		GET_VM_STATS(vm, v_page_count);
551 		GET_VM_STATS(vm, v_free_reserved);
552 		GET_VM_STATS(vm, v_free_target);
553 		GET_VM_STATS(vm, v_free_min);
554 		GET_VM_STATS(vm, v_free_count);
555 		GET_VM_STATS(vm, v_wire_count);
556 		GET_VM_STATS(vm, v_user_wire_count);
557 		GET_VM_STATS(vm, v_active_count);
558 		GET_VM_STATS(vm, v_inactive_target);
559 		GET_VM_STATS(vm, v_inactive_count);
560 		GET_VM_STATS(vm, v_laundry_count);
561 		GET_VM_STATS(vm, v_pageout_free_min);
562 		GET_VM_STATS(vm, v_interrupt_free_min);
563 		/*GET_VM_STATS(vm, v_free_severe);*/
564 		GET_VM_STATS(vm, v_forks);
565 		GET_VM_STATS(vm, v_vforks);
566 		GET_VM_STATS(vm, v_rforks);
567 		GET_VM_STATS(vm, v_kthreads);
568 		GET_VM_STATS(vm, v_forkpages);
569 		GET_VM_STATS(vm, v_vforkpages);
570 		GET_VM_STATS(vm, v_rforkpages);
571 		GET_VM_STATS(vm, v_kthreadpages);
572 #undef GET_VM_STATS
573 	}
574 }
575 
576 static void
577 fill_vmtotal(struct vmtotal *vmtp)
578 {
579 	size_t size;
580 
581 	if (kd != NULL) {
582 		/* XXX fill vmtp */
583 		xo_errx(1, "not implemented");
584 	} else {
585 		size = sizeof(*vmtp);
586 		mysysctl("vm.vmtotal", vmtp, &size);
587 		if (size != sizeof(*vmtp))
588 			xo_errx(1, "vm.total size mismatch");
589 	}
590 }
591 
592 /* Determine how many cpu columns, and what index they are in kern.cp_times */
593 static void
594 getcpuinfo(u_long *maskp, int *maxidp)
595 {
596 	long *times;
597 	u_long mask;
598 	size_t size;
599 	int empty, i, j, maxcpu, maxid;
600 
601 	if (kd != NULL)
602 		xo_errx(1, "not implemented");
603 	mask = 0;
604 	size = sizeof(maxcpu);
605 	mysysctl("kern.smp.maxcpus", &maxcpu, &size);
606 	if (size != sizeof(maxcpu))
607 		xo_errx(1, "sysctl kern.smp.maxcpus");
608 	size = sizeof(long) * maxcpu * CPUSTATES;
609 	times = malloc(size);
610 	if (times == NULL)
611 		xo_err(1, "malloc %zd bytes", size);
612 	mysysctl("kern.cp_times", times, &size);
613 	maxid = (size / CPUSTATES / sizeof(long)) - 1;
614 	for (i = 0; i <= maxid; i++) {
615 		empty = 1;
616 		for (j = 0; empty && j < CPUSTATES; j++) {
617 			if (times[i * CPUSTATES + j] != 0)
618 				empty = 0;
619 		}
620 		if (!empty)
621 			mask |= (1ul << i);
622 	}
623 	if (maskp)
624 		*maskp = mask;
625 	if (maxidp)
626 		*maxidp = maxid;
627 }
628 
629 
630 static void
631 prthuman(const char *name, uint64_t val, int size, int flags)
632 {
633 	char buf[10];
634 	char fmt[128];
635 
636 	snprintf(fmt, sizeof(fmt), "{:%s/%%*s}", name);
637 
638 	if (size < 5 || size > 9)
639 		xo_errx(1, "doofus");
640 	flags |= HN_NOSPACE | HN_DECIMAL;
641 	humanize_number(buf, size, val, "", HN_AUTOSCALE, flags);
642 	xo_attr("value", "%ju", (uintmax_t) val);
643 	xo_emit(fmt, size, buf);
644 }
645 
646 static void
647 dovmstat(unsigned int interval, int reps)
648 {
649 	struct clockinfo clockrate;
650 	struct vmtotal total;
651 	struct devinfo *tmp_dinfo;
652 	u_long cpumask;
653 	size_t size;
654 	time_t uptime, halfuptime;
655 	int maxid, rate_adj, retval;
656 
657 	uptime = getuptime() / 1000000000LL;
658 	halfuptime = uptime / 2;
659 	rate_adj = 1;
660 	maxid = 0;
661 	cpumask = 0;
662 
663 	/*
664 	 * If the user stops the program (control-Z) and then resumes it,
665 	 * print out the header again.
666 	 */
667 	(void)signal(SIGCONT, needhdr);
668 
669 	/*
670 	 * If our standard output is a tty, then install a SIGWINCH handler
671 	 * and set wresized so that our first iteration through the main
672 	 * vmstat loop will peek at the terminal's current rows to find out
673 	 * how many lines can fit in a screenful of output.
674 	 */
675 	if (isatty(fileno(stdout)) != 0) {
676 		wresized = 1;
677 		(void)signal(SIGWINCH, needresize);
678 	} else {
679 		wresized = 0;
680 		winlines = VMSTAT_DEFAULT_LINES;
681 	}
682 
683 	if (kd != NULL) {
684 		if (namelist[X_STATHZ].n_type != 0 &&
685 		    namelist[X_STATHZ].n_value != 0)
686 			kread(X_STATHZ, &hz, sizeof(hz));
687 		if (!hz)
688 			kread(X_HZ, &hz, sizeof(hz));
689 	} else {
690 		size = sizeof(clockrate);
691 		mysysctl("kern.clockrate", &clockrate, &size);
692 		if (size != sizeof(clockrate))
693 			xo_errx(1, "clockrate size mismatch");
694 		hz = clockrate.hz;
695 	}
696 
697 	if (Pflag) {
698 		getcpuinfo(&cpumask, &maxid);
699 		size_cp_times = sizeof(long) * (maxid + 1) * CPUSTATES;
700 		cur_cp_times = calloc(1, size_cp_times);
701 		last_cp_times = calloc(1, size_cp_times);
702 	}
703 	for (hdrcnt = 1;;) {
704 		if (!--hdrcnt)
705 			printhdr(maxid, cpumask);
706 		if (kd != NULL) {
707 			if (kvm_getcptime(kd, cur.cp_time) < 0)
708 				xo_errx(1, "kvm_getcptime: %s", kvm_geterr(kd));
709 		} else {
710 			size = sizeof(cur.cp_time);
711 			mysysctl("kern.cp_time", &cur.cp_time, &size);
712 			if (size != sizeof(cur.cp_time))
713 				xo_errx(1, "cp_time size mismatch");
714 		}
715 		if (Pflag) {
716 			size = size_cp_times;
717 			mysysctl("kern.cp_times", cur_cp_times, &size);
718 			if (size != size_cp_times)
719 				xo_errx(1, "cp_times mismatch");
720 		}
721 
722 		tmp_dinfo = last.dinfo;
723 		last.dinfo = cur.dinfo;
724 		cur.dinfo = tmp_dinfo;
725 		last.snap_time = cur.snap_time;
726 
727 		/*
728 		 * Here what we want to do is refresh our device stats.
729 		 * getdevs() returns 1 when the device list has changed.
730 		 * If the device list has changed, we want to go through
731 		 * the selection process again, in case a device that we
732 		 * were previously displaying has gone away.
733 		 */
734 		switch (devstat_getdevs(NULL, &cur)) {
735 		case -1:
736 			xo_errx(1, "%s", devstat_errbuf);
737 			break;
738 		case 1:
739 			num_devices = cur.dinfo->numdevs;
740 			generation = cur.dinfo->generation;
741 
742 			retval = devstat_selectdevs(&dev_select, &num_selected,
743 			    &num_selections, &select_generation,
744 			    generation, cur.dinfo->devices,
745 			    num_devices, matches, num_matches,
746 			    specified_devices,
747 			    num_devices_specified, select_mode,
748 			    maxshowdevs, 0);
749 			switch (retval) {
750 			case -1:
751 				xo_errx(1, "%s", devstat_errbuf);
752 				break;
753 			case 1:
754 				printhdr(maxid, cpumask);
755 				break;
756 			default:
757 				break;
758 			}
759 			break;
760 		default:
761 			break;
762 		}
763 
764 		fill_vmmeter(&sum);
765 		fill_vmtotal(&total);
766 		xo_open_container("processes");
767 		xo_emit("{:runnable/%2d} {:waiting/%2ld} "
768 		    "{:swapped-out/%2ld}", total.t_rq - 1, total.t_dw +
769 		    total.t_pw, total.t_sw);
770 		xo_close_container("processes");
771 		xo_open_container("memory");
772 #define vmstat_pgtok(a) ((uintmax_t)(a) * (sum.v_page_size >> 10))
773 #define	rate(x)	(unsigned long)(((x) * rate_adj + halfuptime) / uptime)
774 		if (hflag) {
775 			prthuman("available-memory",
776 			    total.t_avm * (uint64_t)sum.v_page_size, 5, HN_B);
777 			prthuman("free-memory",
778 			    total.t_free * (uint64_t)sum.v_page_size, 5, HN_B);
779 			prthuman("total-page-faults",
780 			    rate(sum.v_vm_faults - osum.v_vm_faults), 5, 0);
781 			xo_emit(" ");
782 		} else {
783 			xo_emit(" ");
784 			xo_emit("{:available-memory/%7ju}",
785 			    vmstat_pgtok(total.t_avm));
786 			xo_emit(" ");
787 			xo_emit("{:free-memory/%7ju}",
788 			    vmstat_pgtok(total.t_free));
789 			xo_emit(" ");
790 			xo_emit("{:total-page-faults/%5lu} ",
791 			    rate(sum.v_vm_faults - osum.v_vm_faults));
792 		}
793 		xo_close_container("memory");
794 
795 		xo_open_container("paging-rates");
796 		xo_emit("{:page-reactivated/%3lu} ",
797 		    rate(sum.v_reactivated - osum.v_reactivated));
798 		xo_emit("{:paged-in/%3lu} ",
799 		    rate(sum.v_swapin + sum.v_vnodein -
800 		    (osum.v_swapin + osum.v_vnodein)));
801 		xo_emit("{:paged-out/%3lu}",
802 		    rate(sum.v_swapout + sum.v_vnodeout -
803 		    (osum.v_swapout + osum.v_vnodeout)));
804 		if (hflag) {
805 			prthuman("freed",
806 			    rate(sum.v_tfree - osum.v_tfree), 5, 0);
807 			prthuman("scanned",
808 			    rate(sum.v_pdpages - osum.v_pdpages), 5, 0);
809 		} else {
810 			xo_emit(" ");
811 			xo_emit("{:freed/%5lu} ",
812 			    rate(sum.v_tfree - osum.v_tfree));
813 			xo_emit("{:scanned/%4lu}",
814 			    rate(sum.v_pdpages - osum.v_pdpages));
815 		}
816 		xo_close_container("paging-rates");
817 
818 		devstats();
819 		xo_open_container("fault-rates");
820 		if (hflag) {
821 			prthuman("interrupts",
822 			    rate(sum.v_intr - osum.v_intr), 5, 0);
823 			prthuman("system-calls",
824 			    rate(sum.v_syscall - osum.v_syscall), 5, 0);
825 			prthuman("context-switches",
826 			    rate(sum.v_swtch - osum.v_swtch), 5, 0);
827 		} else {
828 			xo_emit(" ");
829 			xo_emit("{:interrupts/%4lu} "
830 			    "{:system-calls/%5lu} "
831 			    "{:context-switches/%5lu}",
832 			    rate(sum.v_intr - osum.v_intr),
833 			    rate(sum.v_syscall - osum.v_syscall),
834 			    rate(sum.v_swtch - osum.v_swtch));
835 		}
836 		xo_close_container("fault-rates");
837 		if (Pflag)
838 			pcpustats(cpumask, maxid);
839 		else
840 			cpustats();
841 		xo_emit("\n");
842 		xo_flush();
843 		if (reps >= 0 && --reps <= 0)
844 			break;
845 		osum = sum;
846 		uptime = interval;
847 		rate_adj = 1000;
848 		/*
849 		 * We round upward to avoid losing low-frequency events
850 		 * (i.e., >= 1 per interval but < 1 per millisecond).
851 		 */
852 		if (interval != 1)
853 			halfuptime = (uptime + 1) / 2;
854 		else
855 			halfuptime = 0;
856 		(void)usleep(interval * 1000);
857 	}
858 }
859 
860 static void
861 printhdr(int maxid, u_long cpumask)
862 {
863 	int i, num_shown;
864 
865 	num_shown = MIN(num_selected, maxshowdevs);
866 	if (hflag)
867 		xo_emit(" {T:procs}    {T:memory}    {T:/page%*s}", 19, "");
868 	else
869 		xo_emit("{T:procs}     {T:memory}       {T:/page%*s}", 19, "");
870 	if (num_shown > 1)
871 		xo_emit("   {T:/disks %*s}  ", num_shown * 5 - 7, "");
872 	else if (num_shown == 1)
873 		xo_emit("   {T:disks} ");
874 	xo_emit(" {T:faults}      ");
875 	if (Pflag) {
876 		for (i = 0; i <= maxid; i++) {
877 			if (cpumask & (1ul << i))
878 				xo_emit("  {T:/cpu%d}   ", i);
879 		}
880 		xo_emit("\n");
881 	} else
882 		xo_emit(" {T:cpu}\n");
883 	if (hflag) {
884 		xo_emit(" {T:r}  {T:b}  {T:w}  {T:avm}  {T:fre}  {T:flt}  {T:re}"
885 		    "  {T:pi}  {T:po}   {T:fr}   {T:sr} ");
886 	} else {
887 		xo_emit("{T:r} {T:b} {T:w}     {T:avm}     {T:fre}  {T:flt}  "
888 		    "{T:re}  {T:pi}  {T:po}    {T:fr}   {T:sr} ");
889 	}
890 	for (i = 0; i < num_devices; i++)
891 		if ((dev_select[i].selected) &&
892 		    (dev_select[i].selected <= maxshowdevs))
893 			xo_emit("{T:/%3.3s%d} ", dev_select[i].device_name,
894 			    dev_select[i].unit_number);
895 	xo_emit("  {T:in}   {T:sy}   {T:cs}");
896 	if (Pflag) {
897 		for (i = 0; i <= maxid; i++) {
898 			if (cpumask & (1ul << i))
899 				xo_emit(" {T:us} {T:sy} {T:id}");
900 		}
901 		xo_emit("\n");
902 	} else
903 		xo_emit(" {T:us} {T:sy} {T:id}\n");
904 	if (wresized != 0)
905 		doresize();
906 	hdrcnt = winlines;
907 }
908 
909 /*
910  * Force a header to be prepended to the next output.
911  */
912 static void
913 needhdr(int dummy __unused)
914 {
915 
916 	hdrcnt = 1;
917 }
918 
919 /*
920  * When the terminal is resized, force an update of the maximum number of rows
921  * printed between each header repetition.  Then force a new header to be
922  * prepended to the next output.
923  */
924 void
925 needresize(int signo __unused)
926 {
927 
928 	wresized = 1;
929 	hdrcnt = 1;
930 }
931 
932 /*
933  * Update the global `winlines' count of terminal rows.
934  */
935 void
936 doresize(void)
937 {
938 	struct winsize w;
939 	int status;
940 
941 	for (;;) {
942 		status = ioctl(fileno(stdout), TIOCGWINSZ, &w);
943 		if (status == -1 && errno == EINTR)
944 			continue;
945 		else if (status == -1)
946 			xo_err(1, "ioctl");
947 		if (w.ws_row > 3)
948 			winlines = w.ws_row - 3;
949 		else
950 			winlines = VMSTAT_DEFAULT_LINES;
951 		break;
952 	}
953 
954 	/*
955 	 * Inhibit doresize() calls until we are rescheduled by SIGWINCH.
956 	 */
957 	wresized = 0;
958 }
959 
960 static long
961 pct(long top, long bot)
962 {
963 	long ans;
964 
965 	if (bot == 0)
966 		return(0);
967 	ans = (quad_t)top * 100 / bot;
968 	return (ans);
969 }
970 
971 #define	PCT(top, bot) pct((long)(top), (long)(bot))
972 
973 static void
974 dosum(void)
975 {
976 	struct nchstats lnchstats;
977 	size_t size;
978 	long nchtotal;
979 
980 	fill_vmmeter(&sum);
981 	xo_open_container("summary-statistics");
982 	xo_emit("{:context-switches/%9u} {N:cpu context switches}\n",
983 	    sum.v_swtch);
984 	xo_emit("{:interrupts/%9u} {N:device interrupts}\n",
985 	    sum.v_intr);
986 	xo_emit("{:software-interrupts/%9u} {N:software interrupts}\n",
987 	    sum.v_soft);
988 	xo_emit("{:traps/%9u} {N:traps}\n", sum.v_trap);
989 	xo_emit("{:system-calls/%9u} {N:system calls}\n",
990 	    sum.v_syscall);
991 	xo_emit("{:kernel-threads/%9u} {N:kernel threads created}\n",
992 	    sum.v_kthreads);
993 	xo_emit("{:forks/%9u} {N: fork() calls}\n", sum.v_forks);
994 	xo_emit("{:vforks/%9u} {N:vfork() calls}\n",
995 	    sum.v_vforks);
996 	xo_emit("{:rforks/%9u} {N:rfork() calls}\n",
997 	    sum.v_rforks);
998 	xo_emit("{:swap-ins/%9u} {N:swap pager pageins}\n",
999 	    sum.v_swapin);
1000 	xo_emit("{:swap-in-pages/%9u} {N:swap pager pages paged in}\n",
1001 	    sum.v_swappgsin);
1002 	xo_emit("{:swap-outs/%9u} {N:swap pager pageouts}\n",
1003 	    sum.v_swapout);
1004 	xo_emit("{:swap-out-pages/%9u} {N:swap pager pages paged out}\n",
1005 	    sum.v_swappgsout);
1006 	xo_emit("{:vnode-page-ins/%9u} {N:vnode pager pageins}\n",
1007 	    sum.v_vnodein);
1008 	xo_emit("{:vnode-page-in-pages/%9u} {N:vnode pager pages paged in}\n",
1009 	    sum.v_vnodepgsin);
1010 	xo_emit("{:vnode-page-outs/%9u} {N:vnode pager pageouts}\n",
1011 	    sum.v_vnodeout);
1012 	xo_emit("{:vnode-page-out-pages/%9u} {N:vnode pager pages paged out}\n",
1013 	    sum.v_vnodepgsout);
1014 	xo_emit("{:page-daemon-wakeups/%9u} {N:page daemon wakeups}\n",
1015 	    sum.v_pdwakeups);
1016 	xo_emit("{:page-daemon-pages/%9u} {N:pages examined by the page "
1017 	    "daemon}\n", sum.v_pdpages);
1018 	xo_emit("{:page-reclamation-shortfalls/%9u} {N:clean page reclamation "
1019 	    "shortfalls}\n", sum.v_pdshortfalls);
1020 	xo_emit("{:reactivated/%9u} {N:pages reactivated by the page daemon}\n",
1021 	    sum.v_reactivated);
1022 	xo_emit("{:copy-on-write-faults/%9u} {N:copy-on-write faults}\n",
1023 	    sum.v_cow_faults);
1024 	xo_emit("{:copy-on-write-optimized-faults/%9u} {N:copy-on-write "
1025 	    "optimized faults}\n", sum.v_cow_optim);
1026 	xo_emit("{:zero-fill-pages/%9u} {N:zero fill pages zeroed}\n",
1027 	    sum.v_zfod);
1028 	xo_emit("{:zero-fill-prezeroed/%9u} {N:zero fill pages prezeroed}\n",
1029 	    sum.v_ozfod);
1030 	xo_emit("{:intransit-blocking/%9u} {N:intransit blocking page faults}\n",
1031 	    sum.v_intrans);
1032 	xo_emit("{:total-faults/%9u} {N:total VM faults taken}\n",
1033 	    sum.v_vm_faults);
1034 	xo_emit("{:faults-requiring-io/%9u} {N:page faults requiring I\\/O}\n",
1035 	    sum.v_io_faults);
1036 	xo_emit("{:faults-from-thread-creation/%9u} {N:pages affected by "
1037 	    "kernel thread creation}\n", sum.v_kthreadpages);
1038 	xo_emit("{:faults-from-fork/%9u} {N:pages affected by  fork}()\n",
1039 	    sum.v_forkpages);
1040 	xo_emit("{:faults-from-vfork/%9u} {N:pages affected by vfork}()\n",
1041 	    sum.v_vforkpages);
1042 	xo_emit("{:pages-rfork/%9u} {N:pages affected by rfork}()\n",
1043 	    sum.v_rforkpages);
1044 	xo_emit("{:pages-freed/%9u} {N:pages freed}\n",
1045 	    sum.v_tfree);
1046 	xo_emit("{:pages-freed-by-daemon/%9u} {N:pages freed by daemon}\n",
1047 	    sum.v_dfree);
1048 	xo_emit("{:pages-freed-on-exit/%9u} {N:pages freed by exiting processes}\n",
1049 	    sum.v_pfree);
1050 	xo_emit("{:active-pages/%9u} {N:pages active}\n",
1051 	    sum.v_active_count);
1052 	xo_emit("{:inactive-pages/%9u} {N:pages inactive}\n",
1053 	    sum.v_inactive_count);
1054 	xo_emit("{:laundry-pages/%9u} {N:pages in the laundry queue}\n",
1055 	    sum.v_laundry_count);
1056 	xo_emit("{:wired-pages/%9u} {N:pages wired down}\n",
1057 	    sum.v_wire_count);
1058 	xo_emit("{:virtual-user-wired-pages/%9lu} {N:virtual user pages wired "
1059 	    "down}\n", sum.v_user_wire_count);
1060 	xo_emit("{:free-pages/%9u} {N:pages free}\n",
1061 	    sum.v_free_count);
1062 	xo_emit("{:bytes-per-page/%9u} {N:bytes per page}\n", sum.v_page_size);
1063 	if (kd != NULL) {
1064 		kread(X_NCHSTATS, &lnchstats, sizeof(lnchstats));
1065 	} else {
1066 		size = sizeof(lnchstats);
1067 		mysysctl("vfs.cache.nchstats", &lnchstats, &size);
1068 		if (size != sizeof(lnchstats))
1069 			xo_errx(1, "vfs.cache.nchstats size mismatch");
1070 	}
1071 	nchtotal = lnchstats.ncs_goodhits + lnchstats.ncs_neghits +
1072 	    lnchstats.ncs_badhits + lnchstats.ncs_falsehits +
1073 	    lnchstats.ncs_miss + lnchstats.ncs_long;
1074 	xo_emit("{:total-name-lookups/%9ld} {N:total name lookups}\n",
1075 	    nchtotal);
1076 	xo_emit("{P:/%9s} {N:cache hits} "
1077 	    "({:positive-cache-hits/%ld}% pos + "
1078 	    "{:negative-cache-hits/%ld}% {N:neg}) "
1079 	    "system {:cache-hit-percent/%ld}% per-directory\n",
1080 	    "", PCT(lnchstats.ncs_goodhits, nchtotal),
1081 	    PCT(lnchstats.ncs_neghits, nchtotal),
1082 	    PCT(lnchstats.ncs_pass2, nchtotal));
1083 	xo_emit("{P:/%9s} {L:deletions} {:deletions/%ld}%, "
1084 	    "{L:falsehits} {:false-hits/%ld}%, "
1085 	    "{L:toolong} {:too-long/%ld}%\n", "",
1086 	    PCT(lnchstats.ncs_badhits, nchtotal),
1087 	    PCT(lnchstats.ncs_falsehits, nchtotal),
1088 	    PCT(lnchstats.ncs_long, nchtotal));
1089 	xo_close_container("summary-statistics");
1090 }
1091 
1092 static void
1093 doforkst(void)
1094 {
1095 
1096 	fill_vmmeter(&sum);
1097 	xo_open_container("fork-statistics");
1098 	xo_emit("{:fork/%u} {N:forks}, {:fork-pages/%u} {N:pages}, "
1099 	    "{L:average} {:fork-average/%.2f}\n",
1100 	    sum.v_forks, sum.v_forkpages,
1101 	    sum.v_forks == 0 ? 0.0 :
1102 	    (double)sum.v_forkpages / sum.v_forks);
1103 	xo_emit("{:vfork/%u} {N:vforks}, {:vfork-pages/%u} {N:pages}, "
1104 	    "{L:average} {:vfork-average/%.2f}\n",
1105 	    sum.v_vforks, sum.v_vforkpages,
1106 	    sum.v_vforks == 0 ? 0.0 :
1107 	    (double)sum.v_vforkpages / sum.v_vforks);
1108 	xo_emit("{:rfork/%u} {N:rforks}, {:rfork-pages/%u} {N:pages}, "
1109 	    "{L:average} {:rfork-average/%.2f}\n",
1110 	    sum.v_rforks, sum.v_rforkpages,
1111 	    sum.v_rforks == 0 ? 0.0 :
1112 	    (double)sum.v_rforkpages / sum.v_rforks);
1113 	xo_close_container("fork-statistics");
1114 }
1115 
1116 static void
1117 devstats(void)
1118 {
1119 	long double busy_seconds, transfers_per_second;
1120 	long tmp;
1121 	int di, dn, state;
1122 
1123 	for (state = 0; state < CPUSTATES; ++state) {
1124 		tmp = cur.cp_time[state];
1125 		cur.cp_time[state] -= last.cp_time[state];
1126 		last.cp_time[state] = tmp;
1127 	}
1128 
1129 	busy_seconds = cur.snap_time - last.snap_time;
1130 
1131 	xo_open_list("device");
1132 	for (dn = 0; dn < num_devices; dn++) {
1133 		if (dev_select[dn].selected == 0 ||
1134 		    dev_select[dn].selected > maxshowdevs)
1135 			continue;
1136 
1137 		di = dev_select[dn].position;
1138 
1139 		if (devstat_compute_statistics(&cur.dinfo->devices[di],
1140 		    &last.dinfo->devices[di], busy_seconds,
1141 		    DSM_TRANSFERS_PER_SECOND, &transfers_per_second,
1142 		    DSM_NONE) != 0)
1143 			xo_errx(1, "%s", devstat_errbuf);
1144 
1145 		xo_open_instance("device");
1146 		xo_emit("{ekq:name/%s%d}",
1147 		    dev_select[dn].device_name,
1148 		    dev_select[dn].unit_number);
1149 		if (hflag) {
1150 			prthuman("transfers", (uint64_t)transfers_per_second,
1151 			    5, HN_DIVISOR_1000);
1152 		} else {
1153 			xo_emit("{:transfers/%3.0Lf}", transfers_per_second);
1154 		}
1155 		xo_close_instance("device");
1156 	}
1157 	xo_close_list("device");
1158 }
1159 
1160 static void
1161 percent(const char *name, double pctv, int *over)
1162 {
1163 	int l;
1164 	char buf[10];
1165 	char fmt[128];
1166 
1167 	snprintf(fmt, sizeof(fmt), " {:%s/%%*s}", name);
1168 	l = snprintf(buf, sizeof(buf), "%.0f", pctv);
1169 	if (l == 1 && *over) {
1170 		xo_emit(fmt, 1, buf);
1171 		(*over)--;
1172 	} else
1173 		xo_emit(fmt, 2, buf);
1174 	if (l > 2)
1175 		(*over)++;
1176 }
1177 
1178 static void
1179 cpustats(void)
1180 {
1181 	double lpct, total;
1182 	int state, over;
1183 
1184 	total = 0;
1185 	for (state = 0; state < CPUSTATES; ++state)
1186 		total += cur.cp_time[state];
1187 	if (total > 0)
1188 		lpct = 100.0 / total;
1189 	else
1190 		lpct = 0.0;
1191 	over = 0;
1192 	xo_open_container("cpu-statistics");
1193 	percent("user", (cur.cp_time[CP_USER] + cur.cp_time[CP_NICE]) * lpct,
1194 	    &over);
1195 	percent("system", (cur.cp_time[CP_SYS] + cur.cp_time[CP_INTR]) * lpct,
1196 	    &over);
1197 	percent("idle", cur.cp_time[CP_IDLE] * lpct, &over);
1198 	xo_close_container("cpu-statistics");
1199 }
1200 
1201 static void
1202 pcpustats(u_long cpumask, int maxid)
1203 {
1204 	double lpct, total;
1205 	long tmp;
1206 	int i, over, state;
1207 
1208 	/* devstats does this for cp_time */
1209 	for (i = 0; i <= maxid; i++) {
1210 		if ((cpumask & (1ul << i)) == 0)
1211 			continue;
1212 		for (state = 0; state < CPUSTATES; ++state) {
1213 			tmp = cur_cp_times[i * CPUSTATES + state];
1214 			cur_cp_times[i * CPUSTATES + state] -= last_cp_times[i *
1215 			    CPUSTATES + state];
1216 			last_cp_times[i * CPUSTATES + state] = tmp;
1217 		}
1218 	}
1219 
1220 	over = 0;
1221 	xo_open_list("cpu");
1222 	for (i = 0; i <= maxid; i++) {
1223 		if ((cpumask & (1ul << i)) == 0)
1224 			continue;
1225 		xo_open_instance("cpu");
1226 		xo_emit("{ke:name/%d}", i);
1227 		total = 0;
1228 		for (state = 0; state < CPUSTATES; ++state)
1229 			total += cur_cp_times[i * CPUSTATES + state];
1230 		if (total)
1231 			lpct = 100.0 / total;
1232 		else
1233 			lpct = 0.0;
1234 		percent("user", (cur_cp_times[i * CPUSTATES + CP_USER] +
1235 		    cur_cp_times[i * CPUSTATES + CP_NICE]) * lpct, &over);
1236 		percent("system", (cur_cp_times[i * CPUSTATES + CP_SYS] +
1237 		    cur_cp_times[i * CPUSTATES + CP_INTR]) * lpct, &over);
1238 		percent("idle", cur_cp_times[i * CPUSTATES + CP_IDLE] * lpct,
1239 		    &over);
1240 		xo_close_instance("cpu");
1241 	}
1242 	xo_close_list("cpu");
1243 }
1244 
1245 static unsigned int
1246 read_intrcnts(unsigned long **intrcnts)
1247 {
1248 	size_t intrcntlen;
1249 	uintptr_t kaddr;
1250 
1251 	if (kd != NULL) {
1252 		kread(X_SINTRCNT, &intrcntlen, sizeof(intrcntlen));
1253 		if ((*intrcnts = malloc(intrcntlen)) == NULL)
1254 			err(1, "malloc()");
1255 		if (namelist[X_NINTRCNT].n_type == 0)
1256 			kread(X_INTRCNT, *intrcnts, intrcntlen);
1257 		else {
1258 			kread(X_INTRCNT, &kaddr, sizeof(kaddr));
1259 			kreadptr(kaddr, *intrcnts, intrcntlen);
1260 		}
1261 	} else {
1262 		for (*intrcnts = NULL, intrcntlen = 1024; ; intrcntlen *= 2) {
1263 			*intrcnts = reallocf(*intrcnts, intrcntlen);
1264 			if (*intrcnts == NULL)
1265 				err(1, "reallocf()");
1266 			if (mysysctl("hw.intrcnt", *intrcnts, &intrcntlen) == 0)
1267 				break;
1268 		}
1269 	}
1270 
1271 	return (intrcntlen / sizeof(unsigned long));
1272 }
1273 
1274 static void
1275 print_intrcnts(unsigned long *intrcnts, unsigned long *old_intrcnts,
1276     char *intrnames, unsigned int nintr, size_t istrnamlen, long long period_ms)
1277 {
1278 	uint64_t inttotal, old_inttotal, total_count, total_rate;
1279 	unsigned long count, rate;
1280 	unsigned int i;
1281 
1282 	inttotal = 0;
1283 	old_inttotal = 0;
1284 	xo_open_list("interrupt");
1285 	for (i = 0; i < nintr; i++) {
1286 		if (intrnames[0] != '\0' && (*intrcnts != 0 || aflag)) {
1287 			count = *intrcnts - *old_intrcnts;
1288 			rate = ((uint64_t)count * 1000 + period_ms / 2) / period_ms;
1289 			xo_open_instance("interrupt");
1290 			xo_emit("{d:name/%-*s}{ket:name/%s} "
1291 			    "{:total/%20lu} {:rate/%10lu}\n",
1292 			    (int)istrnamlen, intrnames, intrnames, count, rate);
1293 			xo_close_instance("interrupt");
1294 		}
1295 		intrnames += strlen(intrnames) + 1;
1296 		inttotal += *intrcnts++;
1297 		old_inttotal += *old_intrcnts++;
1298 	}
1299 	total_count = inttotal - old_inttotal;
1300 	total_rate = (total_count * 1000 + period_ms / 2) / period_ms;
1301 	xo_close_list("interrupt");
1302 	xo_emit("{L:/%-*s} {:total-interrupts/%20ju} "
1303 	    "{:total-rate/%10ju}\n", (int)istrnamlen,
1304 	    "Total", (uintmax_t)total_count, (uintmax_t)total_rate);
1305 }
1306 
1307 static void
1308 dointr(unsigned int interval, int reps)
1309 {
1310 	unsigned long *intrcnts, *old_intrcnts;
1311 	char *intrname, *intrnames;
1312 	long long period_ms, old_uptime, uptime;
1313 	size_t clen, inamlen, istrnamlen;
1314 	uintptr_t kaddr;
1315 	unsigned int nintr;
1316 
1317 	old_intrcnts = NULL;
1318 	uptime = getuptime();
1319 
1320 	/* Get the names of each interrupt source */
1321 	if (kd != NULL) {
1322 		kread(X_SINTRNAMES, &inamlen, sizeof(inamlen));
1323 		if ((intrnames = malloc(inamlen)) == NULL)
1324 			xo_err(1, "malloc()");
1325 		if (namelist[X_NINTRCNT].n_type == 0)
1326 			kread(X_INTRNAMES, intrnames, inamlen);
1327 		else {
1328 			kread(X_INTRNAMES, &kaddr, sizeof(kaddr));
1329 			kreadptr(kaddr, intrnames, inamlen);
1330 		}
1331 	} else {
1332 		for (intrnames = NULL, inamlen = 1024; ; inamlen *= 2) {
1333 			if ((intrnames = reallocf(intrnames, inamlen)) == NULL)
1334 				xo_err(1, "reallocf()");
1335 			if (mysysctl("hw.intrnames", intrnames, &inamlen) == 0)
1336 				break;
1337 		}
1338 	}
1339 
1340 	/* Determine the length of the longest interrupt name */
1341 	intrname = intrnames;
1342 	istrnamlen = strlen("interrupt");
1343 	while (intrname < intrnames + inamlen) {
1344 		clen = strlen(intrname);
1345 		if (clen > istrnamlen)
1346 			istrnamlen = clen;
1347 		intrname += strlen(intrname) + 1;
1348 	}
1349 	xo_emit("{T:/%-*s} {T:/%20s} {T:/%10s}\n",
1350 	    (int)istrnamlen, "interrupt", "total", "rate");
1351 
1352 	/*
1353 	 * Loop reps times printing differential interrupt counts.  If reps is
1354 	 * zero, then run just once, printing total counts
1355 	 */
1356 	xo_open_container("interrupt-statistics");
1357 
1358 	period_ms = uptime / 1000000;
1359 	while(1) {
1360 		nintr = read_intrcnts(&intrcnts);
1361 		/*
1362 		 * Initialize old_intrcnts to 0 for the first pass, so
1363 		 * print_intrcnts will print total interrupts since boot
1364 		 */
1365 		if (old_intrcnts == NULL) {
1366 			old_intrcnts = calloc(nintr, sizeof(unsigned long));
1367 			if (old_intrcnts == NULL)
1368 				xo_err(1, "calloc()");
1369 		}
1370 
1371 		print_intrcnts(intrcnts, old_intrcnts, intrnames, nintr,
1372 		    istrnamlen, period_ms);
1373 		xo_flush();
1374 
1375 		free(old_intrcnts);
1376 		old_intrcnts = intrcnts;
1377 		if (reps >= 0 && --reps <= 0)
1378 			break;
1379 		usleep(interval * 1000);
1380 		old_uptime = uptime;
1381 		uptime = getuptime();
1382 		period_ms = (uptime - old_uptime) / 1000000;
1383 	}
1384 
1385 	xo_close_container("interrupt-statistics");
1386 }
1387 
1388 static void
1389 domemstat_malloc(void)
1390 {
1391 	struct memory_type_list *mtlp;
1392 	struct memory_type *mtp;
1393 	size_t i, zones;
1394 	int error, first;
1395 
1396 	mtlp = memstat_mtl_alloc();
1397 	if (mtlp == NULL) {
1398 		xo_warn("memstat_mtl_alloc");
1399 		return;
1400 	}
1401 	if (kd == NULL) {
1402 		if (memstat_sysctl_malloc(mtlp, 0) < 0) {
1403 			xo_warnx("memstat_sysctl_malloc: %s",
1404 			    memstat_strerror(memstat_mtl_geterror(mtlp)));
1405 			return;
1406 		}
1407 	} else {
1408 		if (memstat_kvm_malloc(mtlp, kd) < 0) {
1409 			error = memstat_mtl_geterror(mtlp);
1410 			if (error == MEMSTAT_ERROR_KVM)
1411 				xo_warnx("memstat_kvm_malloc: %s",
1412 				    kvm_geterr(kd));
1413 			else
1414 				xo_warnx("memstat_kvm_malloc: %s",
1415 				    memstat_strerror(error));
1416 		}
1417 	}
1418 	xo_open_container("malloc-statistics");
1419 	xo_emit("{T:/%13s} {T:/%5s} {T:/%6s} {T:/%8s}  {T:Size(s)}\n",
1420 	    "Type", "InUse", "MemUse", "Requests");
1421 	xo_open_list("memory");
1422 	zones = memstat_malloc_zone_get_count();
1423 	for (mtp = memstat_mtl_first(mtlp); mtp != NULL;
1424 	    mtp = memstat_mtl_next(mtp)) {
1425 		if (memstat_get_numallocs(mtp) == 0 &&
1426 		    memstat_get_count(mtp) == 0)
1427 			continue;
1428 		xo_open_instance("memory");
1429 		xo_emit("{k:type/%13s/%s} {:in-use/%5ju} "
1430 		    "{:memory-use/%5ju}{U:K} {:requests/%8ju}  ",
1431 		    memstat_get_name(mtp), (uintmax_t)memstat_get_count(mtp),
1432 		    ((uintmax_t)memstat_get_bytes(mtp) + 1023) / 1024,
1433 		    (uintmax_t)memstat_get_numallocs(mtp));
1434 		first = 1;
1435 		xo_open_list("size");
1436 		for (i = 0; i < zones; i++) {
1437 			if (memstat_malloc_zone_used(mtp, i)) {
1438 				if (!first)
1439 					xo_emit(",");
1440 				xo_emit("{l:size/%d}", memstat_malloc_zone_get_size(i));
1441 				first = 0;
1442 			}
1443 		}
1444 		xo_close_list("size");
1445 		xo_close_instance("memory");
1446 		xo_emit("\n");
1447 	}
1448 	xo_close_list("memory");
1449 	xo_close_container("malloc-statistics");
1450 	memstat_mtl_free(mtlp);
1451 }
1452 
1453 static void
1454 domemstat_zone(void)
1455 {
1456 	struct memory_type_list *mtlp;
1457 	struct memory_type *mtp;
1458 	int error;
1459 	char name[MEMTYPE_MAXNAME + 1];
1460 
1461 	mtlp = memstat_mtl_alloc();
1462 	if (mtlp == NULL) {
1463 		xo_warn("memstat_mtl_alloc");
1464 		return;
1465 	}
1466 	if (kd == NULL) {
1467 		if (memstat_sysctl_uma(mtlp, 0) < 0) {
1468 			xo_warnx("memstat_sysctl_uma: %s",
1469 			    memstat_strerror(memstat_mtl_geterror(mtlp)));
1470 			return;
1471 		}
1472 	} else {
1473 		if (memstat_kvm_uma(mtlp, kd) < 0) {
1474 			error = memstat_mtl_geterror(mtlp);
1475 			if (error == MEMSTAT_ERROR_KVM)
1476 				xo_warnx("memstat_kvm_uma: %s",
1477 				    kvm_geterr(kd));
1478 			else
1479 				xo_warnx("memstat_kvm_uma: %s",
1480 				    memstat_strerror(error));
1481 		}
1482 	}
1483 	xo_open_container("memory-zone-statistics");
1484 	xo_emit("{T:/%-20s} {T:/%6s} {T:/%6s} {T:/%8s} {T:/%8s} {T:/%8s} {T:/%8s}"
1485 	    "{T:/%4s} {T:/%4s}\n", "ITEM", "SIZE",
1486 	    "LIMIT", "USED", "FREE", "REQ", "FAIL", "SLEEP", "XDOMAIN");
1487 	xo_open_list("zone");
1488 	for (mtp = memstat_mtl_first(mtlp); mtp != NULL;
1489 	    mtp = memstat_mtl_next(mtp)) {
1490 		strlcpy(name, memstat_get_name(mtp), MEMTYPE_MAXNAME);
1491 		strcat(name, ":");
1492 		xo_open_instance("zone");
1493 		xo_emit("{d:name/%-20s}{ke:name/%s} {:size/%6ju}, "
1494 		    "{:limit/%6ju},{:used/%8ju},"
1495 		    "{:free/%8ju},{:requests/%8ju},"
1496 		    "{:fail/%4ju},{:sleep/%4ju},{:xdomain/%4ju}\n", name,
1497 		    memstat_get_name(mtp),
1498 		    (uintmax_t)memstat_get_size(mtp),
1499 		    (uintmax_t)memstat_get_countlimit(mtp),
1500 		    (uintmax_t)memstat_get_count(mtp),
1501 		    (uintmax_t)memstat_get_free(mtp),
1502 		    (uintmax_t)memstat_get_numallocs(mtp),
1503 		    (uintmax_t)memstat_get_failures(mtp),
1504 		    (uintmax_t)memstat_get_sleeps(mtp),
1505 		    (uintmax_t)memstat_get_xdomain(mtp));
1506 		xo_close_instance("zone");
1507 	}
1508 	memstat_mtl_free(mtlp);
1509 	xo_close_list("zone");
1510 	xo_close_container("memory-zone-statistics");
1511 }
1512 
1513 static void
1514 display_object(struct kinfo_vmobject *kvo)
1515 {
1516 	const char *str;
1517 
1518 	xo_open_instance("object");
1519 	xo_emit("{:resident/%5ju} ", (uintmax_t)kvo->kvo_resident);
1520 	xo_emit("{:active/%5ju} ", (uintmax_t)kvo->kvo_active);
1521 	xo_emit("{:inactive/%5ju} ", (uintmax_t)kvo->kvo_inactive);
1522 	xo_emit("{:refcount/%3d} ", kvo->kvo_ref_count);
1523 	xo_emit("{:shadowcount/%3d} ", kvo->kvo_shadow_count);
1524 
1525 #define	MEMATTR_STR(type, val)					\
1526 	if (kvo->kvo_memattr == (type)) {			\
1527 		str = (val);					\
1528 	} else
1529 #ifdef VM_MEMATTR_UNCACHEABLE
1530 	MEMATTR_STR(VM_MEMATTR_UNCACHEABLE, "UC")
1531 #endif
1532 #ifdef VM_MEMATTR_WRITE_COMBINING
1533 	MEMATTR_STR(VM_MEMATTR_WRITE_COMBINING, "WC")
1534 #endif
1535 #ifdef VM_MEMATTR_WRITE_THROUGH
1536 	MEMATTR_STR(VM_MEMATTR_WRITE_THROUGH, "WT")
1537 #endif
1538 #ifdef VM_MEMATTR_WRITE_PROTECTED
1539 	MEMATTR_STR(VM_MEMATTR_WRITE_PROTECTED, "WP")
1540 #endif
1541 #ifdef VM_MEMATTR_WRITE_BACK
1542 	MEMATTR_STR(VM_MEMATTR_WRITE_BACK, "WB")
1543 #endif
1544 #ifdef VM_MEMATTR_WEAK_UNCACHEABLE
1545 	MEMATTR_STR(VM_MEMATTR_WEAK_UNCACHEABLE, "UC-")
1546 #endif
1547 #ifdef VM_MEMATTR_WB_WA
1548 	MEMATTR_STR(VM_MEMATTR_WB_WA, "WB")
1549 #endif
1550 #ifdef VM_MEMATTR_NOCACHE
1551 	MEMATTR_STR(VM_MEMATTR_NOCACHE, "NC")
1552 #endif
1553 #ifdef VM_MEMATTR_DEVICE
1554 	MEMATTR_STR(VM_MEMATTR_DEVICE, "DEV")
1555 #endif
1556 #ifdef VM_MEMATTR_DEVICE_NP
1557 	MEMATTR_STR(VM_MEMATTR_DEVICE, "NP")
1558 #endif
1559 #ifdef VM_MEMATTR_CACHEABLE
1560 	MEMATTR_STR(VM_MEMATTR_CACHEABLE, "C")
1561 #endif
1562 #ifdef VM_MEMATTR_PREFETCHABLE
1563 	MEMATTR_STR(VM_MEMATTR_PREFETCHABLE, "PRE")
1564 #endif
1565 	{
1566 		str = "??";
1567 	}
1568 #undef MEMATTR_STR
1569 	xo_emit("{:attribute/%-3s} ", str);
1570 	switch (kvo->kvo_type) {
1571 	case KVME_TYPE_NONE:
1572 		str = "--";
1573 		break;
1574 	case KVME_TYPE_DEFAULT:
1575 		str = "df";
1576 		break;
1577 	case KVME_TYPE_VNODE:
1578 		str = "vn";
1579 		break;
1580 	case KVME_TYPE_SWAP:
1581 		str = "sw";
1582 		break;
1583 	case KVME_TYPE_DEVICE:
1584 		str = "dv";
1585 		break;
1586 	case KVME_TYPE_PHYS:
1587 		str = "ph";
1588 		break;
1589 	case KVME_TYPE_DEAD:
1590 		str = "dd";
1591 		break;
1592 	case KVME_TYPE_SG:
1593 		str = "sg";
1594 		break;
1595 	case KVME_TYPE_MGTDEVICE:
1596 		str = "md";
1597 		break;
1598 	case KVME_TYPE_UNKNOWN:
1599 	default:
1600 		str = "??";
1601 		break;
1602 	}
1603 	xo_emit("{:type/%-2s} ", str);
1604 	xo_emit("{:path/%-s}\n", kvo->kvo_path);
1605 	xo_close_instance("object");
1606 }
1607 
1608 static void
1609 doobjstat(void)
1610 {
1611 	struct kinfo_vmobject *kvo;
1612 	int cnt, i;
1613 
1614 	kvo = kinfo_getvmobject(&cnt);
1615 	if (kvo == NULL) {
1616 		xo_warn("Failed to fetch VM object list");
1617 		return;
1618 	}
1619 	xo_emit("{T:RES/%5s} {T:ACT/%5s} {T:INACT/%5s} {T:REF/%3s} {T:SHD/%3s} "
1620 	    "{T:CM/%3s} {T:TP/%2s} {T:PATH/%s}\n");
1621 	xo_open_list("object");
1622 	for (i = 0; i < cnt; i++)
1623 		display_object(&kvo[i]);
1624 	free(kvo);
1625 	xo_close_list("object");
1626 }
1627 
1628 /*
1629  * kread reads something from the kernel, given its nlist index.
1630  */
1631 static void
1632 kreado(int nlx, void *addr, size_t size, size_t offset)
1633 {
1634 	const char *sym;
1635 
1636 	if (namelist[nlx].n_type == 0 || namelist[nlx].n_value == 0) {
1637 		sym = namelist[nlx].n_name;
1638 		if (*sym == '_')
1639 			++sym;
1640 		xo_errx(1, "symbol %s not defined", sym);
1641 	}
1642 	if ((size_t)kvm_read(kd, namelist[nlx].n_value + offset, addr,
1643 	    size) != size) {
1644 		sym = namelist[nlx].n_name;
1645 		if (*sym == '_')
1646 			++sym;
1647 		xo_errx(1, "%s: %s", sym, kvm_geterr(kd));
1648 	}
1649 }
1650 
1651 static void
1652 kread(int nlx, void *addr, size_t size)
1653 {
1654 
1655 	kreado(nlx, addr, size, 0);
1656 }
1657 
1658 static void
1659 kreadptr(uintptr_t addr, void *buf, size_t size)
1660 {
1661 
1662 	if ((size_t)kvm_read(kd, addr, buf, size) != size)
1663 		xo_errx(1, "%s", kvm_geterr(kd));
1664 }
1665 
1666 static void __dead2
1667 usage(void)
1668 {
1669 	xo_error("%s%s",
1670 	    "usage: vmstat [-afHhimoPsz] [-M core [-N system]] [-c count] [-n devs]\n",
1671 	    "              [-p type,if,pass] [-w wait] [disks] [wait [count]]\n");
1672 	xo_finish();
1673 	exit(1);
1674 }
1675