xref: /netbsd-src/usr.bin/vmstat/vmstat.c (revision e5548b402ae4c44fb816de42c7bba9581ce23ef5)
1 /* $NetBSD: vmstat.c,v 1.138 2005/10/22 15:32:48 nonaka Exp $ */
2 
3 /*-
4  * Copyright (c) 1998, 2000, 2001 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation by:
8  *	- Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  *	  NASA Ames Research Center.
10  *	- Simon Burge and Luke Mewburn of Wasabi Systems, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the NetBSD
23  *	Foundation, Inc. and its contributors.
24  * 4. Neither the name of The NetBSD Foundation nor the names of its
25  *    contributors may be used to endorse or promote products derived
26  *    from this software without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
29  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
30  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
31  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
32  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
33  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
34  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
35  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
36  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
37  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
38  * POSSIBILITY OF SUCH DAMAGE.
39  */
40 
41 /*
42  * Copyright (c) 1980, 1986, 1991, 1993
43  *	The Regents of the University of California.  All rights reserved.
44  *
45  * Redistribution and use in source and binary forms, with or without
46  * modification, are permitted provided that the following conditions
47  * are met:
48  * 1. Redistributions of source code must retain the above copyright
49  *    notice, this list of conditions and the following disclaimer.
50  * 2. Redistributions in binary form must reproduce the above copyright
51  *    notice, this list of conditions and the following disclaimer in the
52  *    documentation and/or other materials provided with the distribution.
53  * 3. Neither the name of the University nor the names of its contributors
54  *    may be used to endorse or promote products derived from this software
55  *    without specific prior written permission.
56  *
57  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67  * SUCH DAMAGE.
68  */
69 
70 #include <sys/cdefs.h>
71 #ifndef lint
72 __COPYRIGHT("@(#) Copyright (c) 1980, 1986, 1991, 1993\n\
73 	The Regents of the University of California.  All rights reserved.\n");
74 #endif /* not lint */
75 
76 #ifndef lint
77 #if 0
78 static char sccsid[] = "@(#)vmstat.c	8.2 (Berkeley) 3/1/95";
79 #else
80 __RCSID("$NetBSD: vmstat.c,v 1.138 2005/10/22 15:32:48 nonaka Exp $");
81 #endif
82 #endif /* not lint */
83 
84 #define	__POOL_EXPOSE
85 
86 #include <sys/param.h>
87 #include <sys/mount.h>
88 #include <sys/uio.h>
89 
90 #include <sys/buf.h>
91 #include <sys/device.h>
92 #include <sys/ioctl.h>
93 #include <sys/malloc.h>
94 #include <sys/mallocvar.h>
95 #include <sys/namei.h>
96 #include <sys/pool.h>
97 #include <sys/proc.h>
98 #include <sys/sched.h>
99 #include <sys/socket.h>
100 #include <sys/sysctl.h>
101 #include <sys/time.h>
102 #include <sys/user.h>
103 
104 #include <uvm/uvm_extern.h>
105 #include <uvm/uvm_stat.h>
106 
107 #include <net/if.h>
108 #include <netinet/in.h>
109 #include <netinet/in_var.h>
110 
111 #include <ufs/ufs/inode.h>
112 
113 #include <nfs/rpcv2.h>
114 #include <nfs/nfsproto.h>
115 #include <nfs/nfsnode.h>
116 
117 #include <ctype.h>
118 #include <err.h>
119 #include <errno.h>
120 #include <fcntl.h>
121 #include <kvm.h>
122 #include <limits.h>
123 #include <nlist.h>
124 #undef n_hash
125 #include <paths.h>
126 #include <signal.h>
127 #include <stdio.h>
128 #include <stddef.h>
129 #include <stdlib.h>
130 #include <string.h>
131 #include <time.h>
132 #include <unistd.h>
133 #include <util.h>
134 
135 #include "dkstats.h"
136 #include "tpstats.h"
137 
138 /*
139  * General namelist
140  */
141 struct nlist namelist[] =
142 {
143 #define	X_BOOTTIME	0
144 	{ "_boottime" },
145 #define	X_HZ		1
146 	{ "_hz" },
147 #define	X_STATHZ	2
148 	{ "_stathz" },
149 #define	X_NCHSTATS	3
150 	{ "_nchstats" },
151 #define	X_KMEMSTAT	4
152 	{ "_kmemstatistics" },
153 #define	X_KMEMBUCKETS	5
154 	{ "_kmembuckets" },
155 #define	X_ALLEVENTS	6
156 	{ "_allevents" },
157 #define	X_POOLHEAD	7
158 	{ "_pool_head" },
159 #define	X_UVMEXP	8
160 	{ "_uvmexp" },
161 #define	X_TIME		9
162 	{ "_time" },
163 #define	X_NL_SIZE	10
164 	{ NULL },
165 };
166 
167 /*
168  * Namelist for pre-evcnt interrupt counters.
169  */
170 struct nlist intrnl[] =
171 {
172 #define	X_INTRNAMES	0
173 	{ "_intrnames" },
174 #define	X_EINTRNAMES	1
175 	{ "_eintrnames" },
176 #define	X_INTRCNT	2
177 	{ "_intrcnt" },
178 #define	X_EINTRCNT	3
179 	{ "_eintrcnt" },
180 #define	X_INTRNL_SIZE	4
181 	{ NULL },
182 };
183 
184 
185 /*
186  * Namelist for hash statistics
187  */
188 struct nlist hashnl[] =
189 {
190 #define	X_NFSNODE	0
191 	{ "_nfsnodehash" },
192 #define	X_NFSNODETBL	1
193 	{ "_nfsnodehashtbl" },
194 #define	X_IHASH		2
195 	{ "_ihash" },
196 #define	X_IHASHTBL	3
197 	{ "_ihashtbl" },
198 #define	X_BUFHASH	4
199 	{ "_bufhash" },
200 #define	X_BUFHASHTBL	5
201 	{ "_bufhashtbl" },
202 #define	X_UIHASH	6
203 	{ "_uihash" },
204 #define	X_UIHASHTBL	7
205 	{ "_uihashtbl" },
206 #define	X_IFADDRHASH	8
207 	{ "_in_ifaddrhash" },
208 #define	X_IFADDRHASHTBL	9
209 	{ "_in_ifaddrhashtbl" },
210 #define	X_NCHASH	10
211 	{ "_nchash" },
212 #define	X_NCHASHTBL	11
213 	{ "_nchashtbl" },
214 #define	X_NCVHASH	12
215 	{ "_ncvhash" },
216 #define	X_NCVHASHTBL	13
217 	{ "_ncvhashtbl" },
218 #define X_HASHNL_SIZE	14	/* must be last */
219 	{ NULL },
220 };
221 
222 /*
223  * Namelist for UVM histories
224  */
225 struct nlist histnl[] =
226 {
227 	{ "_uvm_histories" },
228 #define	X_UVM_HISTORIES		0
229 	{ NULL },
230 };
231 
232 
233 
234 struct	uvmexp uvmexp, ouvmexp;
235 int	ndrives;
236 
237 int	winlines = 20;
238 
239 kvm_t *kd;
240 
241 #define	FORKSTAT	1<<0
242 #define	INTRSTAT	1<<1
243 #define	MEMSTAT		1<<2
244 #define	SUMSTAT		1<<3
245 #define	EVCNTSTAT	1<<4
246 #define	VMSTAT		1<<5
247 #define	HISTLIST	1<<6
248 #define	HISTDUMP	1<<7
249 #define	HASHSTAT	1<<8
250 #define	HASHLIST	1<<9
251 #define	VMTOTAL		1<<10
252 
253 void	cpustats(void);
254 void	deref_kptr(const void *, void *, size_t, const char *);
255 void	dkstats(void);
256 void	tpstats(void);
257 void	doevcnt(int verbose);
258 void	dohashstat(int, int, const char *);
259 void	dointr(int verbose);
260 void	domem(void);
261 void	dopool(int, int);
262 void	dopoolcache(struct pool *, int);
263 void	dosum(void);
264 void	dovmstat(struct timespec *, int);
265 void	print_total_hdr(void);
266 void	dovmtotal(struct timespec *, int);
267 void	kread(struct nlist *, int, void *, size_t);
268 void	needhdr(int);
269 long	getuptime(void);
270 void	printhdr(void);
271 long	pct(long, long);
272 void	usage(void);
273 void	doforkst(void);
274 
275 void	hist_traverse(int, const char *);
276 void	hist_dodump(struct uvm_history *);
277 
278 int	main(int, char **);
279 char	**choosedrives(char **);
280 
281 /* Namelist and memory file names. */
282 char	*nlistf, *memf;
283 
284 /* allow old usage [vmstat 1] */
285 #define	BACKWARD_COMPATIBILITY
286 
287 int
288 main(int argc, char *argv[])
289 {
290 	int c, todo, verbose, wide;
291 	struct timespec interval;
292 	int reps;
293 	char errbuf[_POSIX2_LINE_MAX];
294 	gid_t egid = getegid();
295 	const char *histname, *hashname;
296 
297 	histname = hashname = NULL;
298 	(void)setegid(getgid());
299 	memf = nlistf = NULL;
300 	reps = todo = verbose = wide = 0;
301 	interval.tv_sec = 0;
302 	interval.tv_nsec = 0;
303 	while ((c = getopt(argc, argv, "c:efh:HilLM:mN:stu:UvWw:")) != -1) {
304 		switch (c) {
305 		case 'c':
306 			reps = atoi(optarg);
307 			break;
308 		case 'e':
309 			todo |= EVCNTSTAT;
310 			break;
311 		case 'f':
312 			todo |= FORKSTAT;
313 			break;
314 		case 'h':
315 			hashname = optarg;
316 			/* FALLTHROUGH */
317 		case 'H':
318 			todo |= HASHSTAT;
319 			break;
320 		case 'i':
321 			todo |= INTRSTAT;
322 			break;
323 		case 'l':
324 			todo |= HISTLIST;
325 			break;
326 		case 'L':
327 			todo |= HASHLIST;
328 			break;
329 		case 'M':
330 			memf = optarg;
331 			break;
332 		case 'm':
333 			todo |= MEMSTAT;
334 			break;
335 		case 'N':
336 			nlistf = optarg;
337 			break;
338 		case 's':
339 			todo |= SUMSTAT;
340 			break;
341 		case 't':
342 			todo |= VMTOTAL;
343 			break;
344 		case 'u':
345 			histname = optarg;
346 			/* FALLTHROUGH */
347 		case 'U':
348 			todo |= HISTDUMP;
349 			break;
350 		case 'v':
351 			verbose++;
352 			break;
353 		case 'W':
354 			wide++;
355 			break;
356 		case 'w':
357 			interval.tv_sec = atol(optarg);
358 			break;
359 		case '?':
360 		default:
361 			usage();
362 		}
363 	}
364 	argc -= optind;
365 	argv += optind;
366 
367 	if (todo == 0)
368 		todo = VMSTAT;
369 
370 	/*
371 	 * Discard setgid privileges.  If not the running kernel, we toss
372 	 * them away totally so that bad guys can't print interesting stuff
373 	 * from kernel memory, otherwise switch back to kmem for the
374 	 * duration of the kvm_openfiles() call.
375 	 */
376 	if (nlistf != NULL || memf != NULL)
377 		(void)setgid(getgid());
378 	else
379 		(void)setegid(egid);
380 
381 	kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, errbuf);
382 	if (kd == NULL)
383 		errx(1, "kvm_openfiles: %s", errbuf);
384 
385 	if (nlistf == NULL && memf == NULL)
386 		(void)setgid(getgid());
387 
388 	if ((c = kvm_nlist(kd, namelist)) != 0) {
389 		if (c == -1)
390 			errx(1, "kvm_nlist: %s %s", "namelist", kvm_geterr(kd));
391 		(void)fprintf(stderr, "vmstat: undefined symbols:");
392 		for (c = 0; c < sizeof(namelist) / sizeof(namelist[0])-1; c++)
393 			if (namelist[c].n_type == 0)
394 				fprintf(stderr, " %s", namelist[c].n_name);
395 		(void)fputc('\n', stderr);
396 		exit(1);
397 	}
398 	if (todo & INTRSTAT)
399 		(void) kvm_nlist(kd, intrnl);
400 	if ((c = kvm_nlist(kd, hashnl)) == -1 || c == X_HASHNL_SIZE)
401 		errx(1, "kvm_nlist: %s %s", "hashnl", kvm_geterr(kd));
402 	if (kvm_nlist(kd, histnl) == -1)
403 		errx(1, "kvm_nlist: %s %s", "histnl", kvm_geterr(kd));
404 
405 	if (todo & VMSTAT) {
406 		struct winsize winsize;
407 
408 		dkinit(0);	/* Initialize disk stats, no disks selected. */
409 		tpinit(0);
410 
411 		(void)setgid(getgid()); /* don't need privs anymore */
412 
413 		argv = choosedrives(argv);	/* Select disks. */
414 		winsize.ws_row = 0;
415 		(void)ioctl(STDOUT_FILENO, TIOCGWINSZ, &winsize);
416 		if (winsize.ws_row > 0)
417 			winlines = winsize.ws_row;
418 
419 	}
420 
421 #ifdef	BACKWARD_COMPATIBILITY
422 	if (*argv) {
423 		interval.tv_sec = atol(*argv);
424 		if (*++argv)
425 			reps = atoi(*argv);
426 	}
427 #endif
428 
429 	if (interval.tv_sec) {
430 		if (!reps)
431 			reps = -1;
432 	} else if (reps)
433 		interval.tv_sec = 1;
434 
435 
436 	/*
437 	 * Statistics dumping is incompatible with the default
438 	 * VMSTAT/dovmstat() output. So perform the interval/reps handling
439 	 * for it here.
440 	 */
441 	if ((todo & (VMSTAT|VMTOTAL)) == 0) {
442 		for (;;) {
443 			if (todo & (HISTLIST|HISTDUMP)) {
444 				if ((todo & (HISTLIST|HISTDUMP)) ==
445 				    (HISTLIST|HISTDUMP))
446 					errx(1, "you may list or dump,"
447 					    " but not both!");
448 				hist_traverse(todo, histname);
449 				putchar('\n');
450 			}
451 			if (todo & FORKSTAT) {
452 				doforkst();
453 				putchar('\n');
454 			}
455 			if (todo & MEMSTAT) {
456 				domem();
457 				dopool(verbose, wide);
458 				putchar('\n');
459 			}
460 			if (todo & SUMSTAT) {
461 				dosum();
462 				putchar('\n');
463 			}
464 			if (todo & INTRSTAT) {
465 				dointr(verbose);
466 				putchar('\n');
467 			}
468 			if (todo & EVCNTSTAT) {
469 				doevcnt(verbose);
470 				putchar('\n');
471 			}
472 			if (todo & (HASHLIST|HASHSTAT)) {
473 				if ((todo & (HASHLIST|HASHSTAT)) ==
474 				    (HASHLIST|HASHSTAT))
475 					errx(1, "you may list or display,"
476 					    " but not both!");
477 				dohashstat(verbose, todo, hashname);
478 				putchar('\n');
479 			}
480 
481 			if (reps >= 0 && --reps <=0)
482 				break;
483 			nanosleep(&interval, NULL);
484 		}
485 	} else {
486 		if ((todo & (VMSTAT|VMTOTAL)) == (VMSTAT|VMTOTAL)) {
487 			errx(1, "you may not both do vmstat and vmtotal");
488 		}
489 		if (todo & VMSTAT)
490 			dovmstat(&interval, reps);
491 		if (todo & VMTOTAL)
492 			dovmtotal(&interval, reps);
493 	}
494 	exit(0);
495 }
496 
497 char **
498 choosedrives(char **argv)
499 {
500 	int i;
501 
502 	/*
503 	 * Choose drives to be displayed.  Priority goes to (in order) drives
504 	 * supplied as arguments, default drives.  If everything isn't filled
505 	 * in and there are drives not taken care of, display the first few
506 	 * that fit.
507 	 */
508 #define	BACKWARD_COMPATIBILITY
509 	for (ndrives = 0; *argv; ++argv) {
510 #ifdef	BACKWARD_COMPATIBILITY
511 		if (isdigit((unsigned char)**argv))
512 			break;
513 #endif
514 		for (i = 0; i < dk_ndrive; i++) {
515 			if (strcmp(dr_name[i], *argv))
516 				continue;
517 			dk_select[i] = 1;
518 			++ndrives;
519 			break;
520 		}
521 		for (i = 0; i < tp_ndrive; i++) {
522 			if (strcmp(tp_name[i], *argv))
523 				continue;
524 			tp_select[i] = 1;
525 			++ndrives;
526 			break;
527 		}
528 	}
529 	for (i = 0; i < dk_ndrive && ndrives < 3; i++) {
530 		if (dk_select[i])
531 			continue;
532 		dk_select[i] = 1;
533 		++ndrives;
534 	}
535 
536 	for (i = 0; i < tp_ndrive && ndrives < 3; i++) {
537 		if (tp_select[i])
538 			continue;
539 		tp_select[i] = 1;
540 		++ndrives;
541 	}
542 
543 	return (argv);
544 }
545 
546 long
547 getuptime(void)
548 {
549 	static struct timeval boottime;
550 	struct timeval now, diff;
551 	time_t uptime;
552 
553 	if (boottime.tv_sec == 0)
554 		kread(namelist, X_BOOTTIME, &boottime, sizeof(boottime));
555 	kread(namelist, X_TIME, &now, sizeof(now));
556 	timersub(&now, &boottime, &diff);
557 	uptime = diff.tv_sec;
558 	if (uptime <= 0 || uptime > 60*60*24*365*10)
559 		errx(1, "time makes no sense; namelist must be wrong.");
560 	return (uptime);
561 }
562 
563 int	hz, hdrcnt;
564 
565 void
566 print_total_hdr()
567 {
568 
569 	(void)printf("procs            memory\n");
570 	(void)printf("ru dw pw sl sw");
571 	(void)printf("   total-v  active-v  active-r");
572 	(void)printf(" vm-sh avm-sh rm-sh arm-sh free\n");
573 	hdrcnt = winlines - 2;
574 }
575 
576 void
577 dovmtotal(struct timespec *interval, int reps)
578 {
579 	struct vmtotal total;
580 	int mib[2];
581 	size_t size;
582 
583 	(void)signal(SIGCONT, needhdr);
584 
585 	for (hdrcnt = 1;;) {
586 		if (!--hdrcnt)
587 			print_total_hdr();
588 		if (memf != NULL) {
589 			printf("Unable to get vmtotals from crash dump.\n");
590 			memset(&total, 0, sizeof(total));
591 		} else {
592 			size = sizeof(total);
593 			mib[0] = CTL_VM;
594 			mib[1] = VM_METER;
595 			if (sysctl(mib, 2, &total, &size, NULL, 0) < 0) {
596 				printf("Can't get vmtotals: %s\n",
597 				    strerror(errno));
598 				memset(&total, 0, sizeof(total));
599 			}
600 		}
601 		printf("%2d ", total.t_rq);
602 		printf("%2d ", total.t_dw);
603 		printf("%2d ", total.t_pw);
604 		printf("%2d ", total.t_sl);
605 		printf("%2d ", total.t_sw);
606 
607 		printf("%9d ", total.t_vm);
608 		printf("%9d ", total.t_avm);
609 		printf("%9d ", total.t_arm);
610 		printf("%5d ", total.t_vmshr);
611 		printf("%6d ", total.t_avmshr);
612 		printf("%5d ", total.t_rmshr);
613 		printf("%6d ", total.t_armshr);
614 		printf("%5d",  total.t_free);
615 
616 		putchar('\n');
617 
618 		(void)fflush(stdout);
619 		if (reps >= 0 && --reps <= 0)
620 			break;
621 
622 		nanosleep(interval, NULL);
623 	}
624 }
625 
626 void
627 dovmstat(struct timespec *interval, int reps)
628 {
629 	struct vmtotal total;
630 	time_t uptime, halfuptime;
631 	int mib[2];
632 	size_t size;
633 	int pagesize = getpagesize();
634 
635 	uptime = getuptime();
636 	halfuptime = uptime / 2;
637 	(void)signal(SIGCONT, needhdr);
638 
639 	if (namelist[X_STATHZ].n_type != 0 && namelist[X_STATHZ].n_value != 0)
640 		kread(namelist, X_STATHZ, &hz, sizeof(hz));
641 	if (!hz)
642 		kread(namelist, X_HZ, &hz, sizeof(hz));
643 
644 	for (hdrcnt = 1;;) {
645 		if (!--hdrcnt)
646 			printhdr();
647 		/* Read new disk statistics */
648 		dkreadstats();
649 		kread(namelist, X_UVMEXP, &uvmexp, sizeof(uvmexp));
650 		if (memf != NULL) {
651 			/*
652 			 * XXX Can't do this if we're reading a crash
653 			 * XXX dump because they're lazily-calculated.
654 			 */
655 			printf("Unable to get vmtotals from crash dump.\n");
656 			memset(&total, 0, sizeof(total));
657 		} else {
658 			size = sizeof(total);
659 			mib[0] = CTL_VM;
660 			mib[1] = VM_METER;
661 			if (sysctl(mib, 2, &total, &size, NULL, 0) < 0) {
662 				printf("Can't get vmtotals: %s\n",
663 				    strerror(errno));
664 				memset(&total, 0, sizeof(total));
665 			}
666 		}
667 		(void)printf("%2d %d %d",
668 		    total.t_rq - 1, total.t_dw + total.t_pw, total.t_sw);
669 #define	pgtok(a) (long)((a) * (pagesize >> 10))
670 #define	rate(x)	(u_long)(((x) + halfuptime) / uptime)	/* round */
671 		(void)printf(" %6ld %6ld ",
672 		    pgtok(total.t_avm), pgtok(total.t_free));
673 		(void)printf("%4lu ", rate(uvmexp.faults - ouvmexp.faults));
674 		(void)printf("%3lu ", rate(uvmexp.pdreact - ouvmexp.pdreact));
675 		(void)printf("%3lu ", rate(uvmexp.pageins - ouvmexp.pageins));
676 		(void)printf("%4lu ",
677 		    rate(uvmexp.pgswapout - ouvmexp.pgswapout));
678 		(void)printf("%4lu ", rate(uvmexp.pdfreed - ouvmexp.pdfreed));
679 		(void)printf("%4lu ", rate(uvmexp.pdscans - ouvmexp.pdscans));
680 		dkstats();
681 		tpstats();
682 		(void)printf("%4lu %4lu %3lu ",
683 		    rate(uvmexp.intrs - ouvmexp.intrs),
684 		    rate(uvmexp.syscalls - ouvmexp.syscalls),
685 		    rate(uvmexp.swtch - ouvmexp.swtch));
686 		cpustats();
687 		putchar('\n');
688 		(void)fflush(stdout);
689 		if (reps >= 0 && --reps <= 0)
690 			break;
691 		ouvmexp = uvmexp;
692 		uptime = interval->tv_sec;
693 		/*
694 		 * We round upward to avoid losing low-frequency events
695 		 * (i.e., >= 1 per interval but < 1 per second).
696 		 */
697 		halfuptime = uptime == 1 ? 0 : (uptime + 1) / 2;
698 		nanosleep(interval, NULL);
699 	}
700 }
701 
702 void
703 printhdr(void)
704 {
705 	int i;
706 
707 	(void)printf(" procs    memory      page%*s", 23, "");
708 	if (ndrives > 0)
709 		(void)printf("%s %*sfaults      cpu\n",
710 		    ((ndrives > 1) ? "disks" : "disk"),
711 		    ((ndrives > 1) ? ndrives * 3 - 4 : 0), "");
712 	else
713 		(void)printf("%*s  faults   cpu\n",
714 		    ndrives * 3, "");
715 
716 	(void)printf(" r b w    avm    fre  flt  re  pi   po   fr   sr ");
717 	for (i = 0; i < dk_ndrive; i++)
718 		if (dk_select[i])
719 			(void)printf("%c%c ", dr_name[i][0],
720 			    dr_name[i][strlen(dr_name[i]) - 1]);
721 	(void)printf("  in   sy  cs us sy id\n");
722 	hdrcnt = winlines - 2;
723 }
724 
725 /*
726  * Force a header to be prepended to the next output.
727  */
728 void
729 needhdr(int dummy)
730 {
731 
732 	hdrcnt = 1;
733 }
734 
735 long
736 pct(long top, long bot)
737 {
738 	long ans;
739 
740 	if (bot == 0)
741 		return (0);
742 	ans = (quad_t)top * 100 / bot;
743 	return (ans);
744 }
745 
746 #define	PCT(top, bot) (int)pct((long)(top), (long)(bot))
747 
748 void
749 dosum(void)
750 {
751 	struct nchstats nchstats;
752 	u_long nchtotal;
753 
754 	kread(namelist, X_UVMEXP, &uvmexp, sizeof(uvmexp));
755 
756 	(void)printf("%9u bytes per page\n", uvmexp.pagesize);
757 
758 	(void)printf("%9u page color%s\n",
759 	    uvmexp.ncolors, uvmexp.ncolors == 1 ? "" : "s");
760 
761 	(void)printf("%9u pages managed\n", uvmexp.npages);
762 	(void)printf("%9u pages free\n", uvmexp.free);
763 	(void)printf("%9u pages active\n", uvmexp.active);
764 	(void)printf("%9u pages inactive\n", uvmexp.inactive);
765 	(void)printf("%9u pages paging\n", uvmexp.paging);
766 	(void)printf("%9u pages wired\n", uvmexp.wired);
767 	(void)printf("%9u zero pages\n", uvmexp.zeropages);
768 	(void)printf("%9u reserve pagedaemon pages\n",
769 	    uvmexp.reserve_pagedaemon);
770 	(void)printf("%9u reserve kernel pages\n", uvmexp.reserve_kernel);
771 	(void)printf("%9u anonymous pages\n", uvmexp.anonpages);
772 	(void)printf("%9u cached file pages\n", uvmexp.filepages);
773 	(void)printf("%9u cached executable pages\n", uvmexp.execpages);
774 
775 	(void)printf("%9u minimum free pages\n", uvmexp.freemin);
776 	(void)printf("%9u target free pages\n", uvmexp.freetarg);
777 	(void)printf("%9u target inactive pages\n", uvmexp.inactarg);
778 	(void)printf("%9u maximum wired pages\n", uvmexp.wiredmax);
779 
780 	(void)printf("%9u swap devices\n", uvmexp.nswapdev);
781 	(void)printf("%9u swap pages\n", uvmexp.swpages);
782 	(void)printf("%9u swap pages in use\n", uvmexp.swpginuse);
783 	(void)printf("%9u swap allocations\n", uvmexp.nswget);
784 
785 	(void)printf("%9u total faults taken\n", uvmexp.faults);
786 	(void)printf("%9u traps\n", uvmexp.traps);
787 	(void)printf("%9u device interrupts\n", uvmexp.intrs);
788 	(void)printf("%9u CPU context switches\n", uvmexp.swtch);
789 	(void)printf("%9u software interrupts\n", uvmexp.softs);
790 	(void)printf("%9u system calls\n", uvmexp.syscalls);
791 	(void)printf("%9u pagein requests\n", uvmexp.pageins);
792 	(void)printf("%9u pageout requests\n", uvmexp.pdpageouts);
793 	(void)printf("%9u swap ins\n", uvmexp.swapins);
794 	(void)printf("%9u swap outs\n", uvmexp.swapouts);
795 	(void)printf("%9u pages swapped in\n", uvmexp.pgswapin);
796 	(void)printf("%9u pages swapped out\n", uvmexp.pgswapout);
797 	(void)printf("%9u forks total\n", uvmexp.forks);
798 	(void)printf("%9u forks blocked parent\n", uvmexp.forks_ppwait);
799 	(void)printf("%9u forks shared address space with parent\n",
800 	    uvmexp.forks_sharevm);
801 	(void)printf("%9u pagealloc zero wanted and avail\n",
802 	    uvmexp.pga_zerohit);
803 	(void)printf("%9u pagealloc zero wanted and not avail\n",
804 	    uvmexp.pga_zeromiss);
805 	(void)printf("%9u aborts of idle page zeroing\n",
806 	    uvmexp.zeroaborts);
807 	(void)printf("%9u pagealloc desired color avail\n",
808 	    uvmexp.colorhit);
809 	(void)printf("%9u pagealloc desired color not avail\n",
810 	    uvmexp.colormiss);
811 
812 	(void)printf("%9u faults with no memory\n", uvmexp.fltnoram);
813 	(void)printf("%9u faults with no anons\n", uvmexp.fltnoanon);
814 	(void)printf("%9u faults had to wait on pages\n", uvmexp.fltpgwait);
815 	(void)printf("%9u faults found released page\n", uvmexp.fltpgrele);
816 	(void)printf("%9u faults relock (%u ok)\n", uvmexp.fltrelck,
817 	    uvmexp.fltrelckok);
818 	(void)printf("%9u anon page faults\n", uvmexp.fltanget);
819 	(void)printf("%9u anon retry faults\n", uvmexp.fltanretry);
820 	(void)printf("%9u amap copy faults\n", uvmexp.fltamcopy);
821 	(void)printf("%9u neighbour anon page faults\n", uvmexp.fltnamap);
822 	(void)printf("%9u neighbour object page faults\n", uvmexp.fltnomap);
823 	(void)printf("%9u locked pager get faults\n", uvmexp.fltlget);
824 	(void)printf("%9u unlocked pager get faults\n", uvmexp.fltget);
825 	(void)printf("%9u anon faults\n", uvmexp.flt_anon);
826 	(void)printf("%9u anon copy on write faults\n", uvmexp.flt_acow);
827 	(void)printf("%9u object faults\n", uvmexp.flt_obj);
828 	(void)printf("%9u promote copy faults\n", uvmexp.flt_prcopy);
829 	(void)printf("%9u promote zero fill faults\n", uvmexp.flt_przero);
830 
831 	(void)printf("%9u times daemon wokeup\n",uvmexp.pdwoke);
832 	(void)printf("%9u revolutions of the clock hand\n", uvmexp.pdrevs);
833 	(void)printf("%9u times daemon attempted swapout\n", uvmexp.pdswout);
834 	(void)printf("%9u pages freed by daemon\n", uvmexp.pdfreed);
835 	(void)printf("%9u pages scanned by daemon\n", uvmexp.pdscans);
836 	(void)printf("%9u anonymous pages scanned by daemon\n",
837 	    uvmexp.pdanscan);
838 	(void)printf("%9u object pages scanned by daemon\n", uvmexp.pdobscan);
839 	(void)printf("%9u pages reactivated\n", uvmexp.pdreact);
840 	(void)printf("%9u anonymous pages reactivated\n", uvmexp.pdreanon);
841 	(void)printf("%9u cached file pages reactivated\n", uvmexp.pdrefile);
842 	(void)printf("%9u cached executable pages reactivated\n",
843 	    uvmexp.pdreexec);
844 	(void)printf("%9u pages found busy by daemon\n", uvmexp.pdbusy);
845 	(void)printf("%9u total pending pageouts\n", uvmexp.pdpending);
846 	(void)printf("%9u pages deactivated\n", uvmexp.pddeact);
847 
848 	kread(namelist, X_NCHSTATS, &nchstats, sizeof(nchstats));
849 	nchtotal = nchstats.ncs_goodhits + nchstats.ncs_neghits +
850 	    nchstats.ncs_badhits + nchstats.ncs_falsehits +
851 	    nchstats.ncs_miss + nchstats.ncs_long;
852 	(void)printf("%9lu total name lookups\n", nchtotal);
853 	(void)printf("%9lu good hits\n", nchstats.ncs_goodhits);
854 	(void)printf("%9lu negative hits\n", nchstats.ncs_neghits);
855 	(void)printf("%9lu bad hits\n", nchstats.ncs_badhits);
856 	(void)printf("%9lu false hits\n", nchstats.ncs_falsehits);
857 	(void)printf("%9lu miss\n", nchstats.ncs_miss);
858 	(void)printf("%9lu too long\n", nchstats.ncs_long);
859 	(void)printf("%9lu pass2 hits\n", nchstats.ncs_pass2);
860 	(void)printf("%9lu 2passes\n", nchstats.ncs_2passes);
861 	(void)printf(
862 	    "%9s cache hits (%d%% pos + %d%% neg) system %d%% per-process\n",
863 	    "", PCT(nchstats.ncs_goodhits, nchtotal),
864 	    PCT(nchstats.ncs_neghits, nchtotal),
865 	    PCT(nchstats.ncs_pass2, nchtotal));
866 	(void)printf("%9s deletions %d%%, falsehits %d%%, toolong %d%%\n", "",
867 	    PCT(nchstats.ncs_badhits, nchtotal),
868 	    PCT(nchstats.ncs_falsehits, nchtotal),
869 	    PCT(nchstats.ncs_long, nchtotal));
870 }
871 
872 void
873 doforkst(void)
874 {
875 
876 	kread(namelist, X_UVMEXP, &uvmexp, sizeof(uvmexp));
877 
878 	(void)printf("%u forks total\n", uvmexp.forks);
879 	(void)printf("%u forks blocked parent\n", uvmexp.forks_ppwait);
880 	(void)printf("%u forks shared address space with parent\n",
881 	    uvmexp.forks_sharevm);
882 }
883 
884 void
885 dkstats(void)
886 {
887 	int dn;
888 	double etime;
889 
890 	/* Calculate disk stat deltas. */
891 	dkswap();
892 	etime = cur.cp_etime;
893 
894 	for (dn = 0; dn < dk_ndrive; ++dn) {
895 		if (!dk_select[dn])
896 			continue;
897 		(void)printf("%2.0f ",
898 		    (cur.dk_rxfer[dn] + cur.dk_wxfer[dn]) / etime);
899 	}
900 }
901 
902 void
903 tpstats(void)
904 {
905 	int dn;
906 	double etime;
907 
908 	/* Calculate tape stat deltas. */
909 	tpswap();
910 	etime = cur.cp_etime;
911 
912 	for (dn = 0; dn < tp_ndrive; ++dn) {
913 		if (!tp_select[dn])
914 			continue;
915 		(void)printf("%2.0f ",
916 		    (cur_tape.rxfer[dn] + cur_tape.wxfer[dn]) / etime);
917 	}
918 }
919 
920 void
921 cpustats(void)
922 {
923 	int state;
924 	double pcnt, total;
925 	double stat_us, stat_sy, stat_id;
926 
927 	total = 0;
928 	for (state = 0; state < CPUSTATES; ++state)
929 		total += cur.cp_time[state];
930 	if (total)
931 		pcnt = 100 / total;
932 	else
933 		pcnt = 0;
934 	stat_us = (cur.cp_time[CP_USER] + cur.cp_time[CP_NICE]) * pcnt;
935 	stat_sy = (cur.cp_time[CP_SYS] + cur.cp_time[CP_INTR]) * pcnt;
936 	stat_id = cur.cp_time[CP_IDLE] * pcnt;
937 	(void)printf("%*.0f ", ((stat_sy >= 100) ? 1 : 2), stat_us);
938 	(void)printf("%*.0f ", ((stat_us >= 100 || stat_id >= 100) ? 1 : 2),
939 		     stat_sy);
940 	(void)printf("%2.0f", stat_id);
941 }
942 
943 void
944 dointr(int verbose)
945 {
946 	unsigned long *intrcnt, *ointrcnt;
947 	unsigned long long inttotal, uptime;
948 	int nintr, inamlen;
949 	char *intrname, *ointrname;
950 	struct evcntlist allevents;
951 	struct evcnt evcnt, *evptr;
952 	char evgroup[EVCNT_STRING_MAX], evname[EVCNT_STRING_MAX];
953 
954 	inttotal = 0;
955 	uptime = getuptime();
956 	(void)printf("%-34s %16s %8s\n", "interrupt", "total", "rate");
957 	nintr = intrnl[X_EINTRCNT].n_value - intrnl[X_INTRCNT].n_value;
958 	inamlen = intrnl[X_EINTRNAMES].n_value - intrnl[X_INTRNAMES].n_value;
959 	if (nintr != 0 && inamlen != 0) {
960 		ointrcnt = intrcnt = malloc((size_t)nintr);
961 		ointrname = intrname = malloc((size_t)inamlen);
962 		if (intrcnt == NULL || intrname == NULL)
963 			errx(1, "%s", "");
964 		kread(intrnl, X_INTRCNT, intrcnt, (size_t)nintr);
965 		kread(intrnl, X_INTRNAMES, intrname, (size_t)inamlen);
966 		nintr /= sizeof(long);
967 		while (--nintr >= 0) {
968 			if (*intrcnt || verbose)
969 				(void)printf("%-34s %16llu %8llu\n", intrname,
970 					     (unsigned long long)*intrcnt,
971 					     (unsigned long long)
972 					     (*intrcnt / uptime));
973 			intrname += strlen(intrname) + 1;
974 			inttotal += *intrcnt++;
975 		}
976 		free(ointrcnt);
977 		free(ointrname);
978 	}
979 
980 	kread(namelist, X_ALLEVENTS, &allevents, sizeof allevents);
981 	evptr = TAILQ_FIRST(&allevents);
982 	while (evptr) {
983 		deref_kptr(evptr, &evcnt, sizeof(evcnt), "event chain trashed");
984 		evptr = TAILQ_NEXT(&evcnt, ev_list);
985 		if (evcnt.ev_type != EVCNT_TYPE_INTR)
986 			continue;
987 
988 		if (evcnt.ev_count == 0 && !verbose)
989 			continue;
990 
991 		deref_kptr(evcnt.ev_group, evgroup, evcnt.ev_grouplen + 1,
992 		    "event chain trashed");
993 		deref_kptr(evcnt.ev_name, evname, evcnt.ev_namelen + 1,
994 		    "event chain trashed");
995 
996 		(void)printf("%s %s%*s %16llu %8llu\n", evgroup, evname,
997 		    34 - (evcnt.ev_grouplen + 1 + evcnt.ev_namelen), "",
998 		    (unsigned long long)evcnt.ev_count,
999 		    (unsigned long long)(evcnt.ev_count / uptime));
1000 
1001 		inttotal += evcnt.ev_count++;
1002 	}
1003 	(void)printf("%-34s %16llu %8llu\n", "Total", inttotal,
1004 	    (unsigned long long)(inttotal / uptime));
1005 }
1006 
1007 void
1008 doevcnt(int verbose)
1009 {
1010 	static const char * evtypes [] = { "misc", "intr", "trap" };
1011 	unsigned long long uptime;
1012 	struct evcntlist allevents;
1013 	struct evcnt evcnt, *evptr;
1014 	char evgroup[EVCNT_STRING_MAX], evname[EVCNT_STRING_MAX];
1015 
1016 	/* XXX should print type! */
1017 
1018 	uptime = getuptime();
1019 	(void)printf("%-34s %16s %8s %s\n", "event", "total", "rate", "type");
1020 	kread(namelist, X_ALLEVENTS, &allevents, sizeof allevents);
1021 	evptr = TAILQ_FIRST(&allevents);
1022 	while (evptr) {
1023 		deref_kptr(evptr, &evcnt, sizeof(evcnt), "event chain trashed");
1024 
1025 		evptr = TAILQ_NEXT(&evcnt, ev_list);
1026 		if (evcnt.ev_count == 0 && !verbose)
1027 			continue;
1028 
1029 		deref_kptr(evcnt.ev_group, evgroup, evcnt.ev_grouplen + 1,
1030 		    "event chain trashed");
1031 		deref_kptr(evcnt.ev_name, evname, evcnt.ev_namelen + 1,
1032 		    "event chain trashed");
1033 
1034 		(void)printf("%s %s%*s %16llu %8llu %s\n", evgroup, evname,
1035 		    34 - (evcnt.ev_grouplen + 1 + evcnt.ev_namelen), "",
1036 		    (unsigned long long)evcnt.ev_count,
1037 		    (unsigned long long)(evcnt.ev_count / uptime),
1038 		    (evcnt.ev_type < sizeof(evtypes)/sizeof(evtypes[0]) ?
1039 			evtypes[evcnt.ev_type] : "?"));
1040 	}
1041 }
1042 
1043 static char memname[64];
1044 
1045 void
1046 domem(void)
1047 {
1048 	struct kmembuckets *kp;
1049 	struct malloc_type ks, *ksp;
1050 	int i, j;
1051 	int len, size, first;
1052 	long totuse = 0, totfree = 0, totreq = 0;
1053 	struct kmembuckets buckets[MINBUCKET + 16];
1054 
1055 	kread(namelist, X_KMEMBUCKETS, buckets, sizeof(buckets));
1056 	for (first = 1, i = MINBUCKET, kp = &buckets[i]; i < MINBUCKET + 16;
1057 	    i++, kp++) {
1058 		if (kp->kb_calls == 0)
1059 			continue;
1060 		if (first) {
1061 			(void)printf("Memory statistics by bucket size\n");
1062 			(void)printf(
1063 		 "    Size   In Use   Free   Requests  HighWater  Couldfree\n");
1064 			first = 0;
1065 		}
1066 		size = 1 << i;
1067 		(void)printf("%8d %8ld %6ld %10ld %7ld %10ld\n", size,
1068 		    kp->kb_total - kp->kb_totalfree,
1069 		    kp->kb_totalfree, kp->kb_calls,
1070 		    kp->kb_highwat, kp->kb_couldfree);
1071 		totfree += size * kp->kb_totalfree;
1072 	}
1073 
1074 	/*
1075 	 * If kmem statistics are not being gathered by the kernel,
1076 	 * first will still be 1.
1077 	 */
1078 	if (first) {
1079 		warnx("Kmem statistics are not being gathered by the kernel.");
1080 		return;
1081 	}
1082 
1083 	(void)printf("\nMemory usage type by bucket size\n");
1084 	(void)printf("    Size  Type(s)\n");
1085 	kp = &buckets[MINBUCKET];
1086 	for (j =  1 << MINBUCKET; j < 1 << (MINBUCKET + 16); j <<= 1, kp++) {
1087 		if (kp->kb_calls == 0)
1088 			continue;
1089 		first = 1;
1090 		len = 8;
1091 		for (kread(namelist, X_KMEMSTAT, &ksp, sizeof(ksp));
1092 		     ksp != NULL; ksp = ks.ks_next) {
1093 			deref_kptr(ksp, &ks, sizeof(ks), "malloc type");
1094 			if (ks.ks_calls == 0)
1095 				continue;
1096 			if ((ks.ks_size & j) == 0)
1097 				continue;
1098 			deref_kptr(ks.ks_shortdesc, memname,
1099 			    sizeof(memname), "malloc type name");
1100 			len += 2 + strlen(memname);
1101 			if (first)
1102 				printf("%8d  %s", j, memname);
1103 			else
1104 				printf(",");
1105 			if (len >= 80) {
1106 				printf("\n\t ");
1107 				len = 10 + strlen(memname);
1108 			}
1109 			if (!first)
1110 				printf(" %s", memname);
1111 			first = 0;
1112 		}
1113 		putchar('\n');
1114 	}
1115 
1116 	(void)printf(
1117 	    "\nMemory statistics by type                           Type  Kern\n");
1118 	(void)printf(
1119 "         Type  InUse MemUse HighUse  Limit Requests Limit Limit Size(s)\n");
1120 	for (kread(namelist, X_KMEMSTAT, &ksp, sizeof(ksp));
1121 	     ksp != NULL; ksp = ks.ks_next) {
1122 		deref_kptr(ksp, &ks, sizeof(ks), "malloc type");
1123 		if (ks.ks_calls == 0)
1124 			continue;
1125 		deref_kptr(ks.ks_shortdesc, memname,
1126 		    sizeof(memname), "malloc type name");
1127 		(void)printf("%14s%6ld%6ldK%7ldK%6ldK%9ld%5u%6u",
1128 		    memname,
1129 		    ks.ks_inuse, (ks.ks_memuse + 1023) / 1024,
1130 		    (ks.ks_maxused + 1023) / 1024,
1131 		    (ks.ks_limit + 1023) / 1024, ks.ks_calls,
1132 		    ks.ks_limblocks, ks.ks_mapblocks);
1133 		first = 1;
1134 		for (j =  1 << MINBUCKET; j < 1 << (MINBUCKET + 16); j <<= 1) {
1135 			if ((ks.ks_size & j) == 0)
1136 				continue;
1137 			if (first)
1138 				printf("  %d", j);
1139 			else
1140 				printf(",%d", j);
1141 			first = 0;
1142 		}
1143 		printf("\n");
1144 		totuse += ks.ks_memuse;
1145 		totreq += ks.ks_calls;
1146 	}
1147 	(void)printf("\nMemory totals:  In Use    Free    Requests\n");
1148 	(void)printf("              %7ldK %6ldK    %8ld\n\n",
1149 	    (totuse + 1023) / 1024, (totfree + 1023) / 1024, totreq);
1150 }
1151 
1152 void
1153 dopool(int verbose, int wide)
1154 {
1155 	int first, ovflw;
1156 	void *addr;
1157 	long total, inuse, this_total, this_inuse;
1158 	LIST_HEAD(,pool) pool_head;
1159 	struct pool pool, *pp = &pool;
1160 	struct pool_allocator pa;
1161 	char name[32], maxp[32];
1162 
1163 	kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
1164 	addr = LIST_FIRST(&pool_head);
1165 
1166 	total = inuse = 0;
1167 
1168 	for (first = 1; addr != NULL; addr = LIST_NEXT(pp, pr_poollist) ) {
1169 		deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
1170 		deref_kptr(pp->pr_alloc, &pa, sizeof(pa),
1171 		    "pool allocator trashed");
1172 		deref_kptr(pp->pr_wchan, name, sizeof(name),
1173 		    "pool wait channel trashed");
1174 		name[sizeof(name)-1] = '\0';
1175 
1176 		if (first) {
1177 			(void)printf("Memory resource pool statistics\n");
1178 			(void)printf(
1179 			    "%-*s%*s%*s%5s%*s%s%s%*s%*s%6s%s%6s%6s%6s%5s%s%s\n",
1180 			    wide ? 16 : 11, "Name",
1181 			    wide ? 6 : 5, "Size",
1182 			    wide ? 12 : 9, "Requests",
1183 			    "Fail",
1184 			    wide ? 12 : 9, "Releases",
1185 			    wide ? " InUse" : "",
1186 			    wide ? " Avail" : "",
1187 			    wide ? 7 : 6, "Pgreq",
1188 			    wide ? 7 : 6, "Pgrel",
1189 			    "Npage",
1190 			    wide ? " PageSz" : "",
1191 			    "Hiwat",
1192 			    "Minpg",
1193 			    "Maxpg",
1194 			    "Idle",
1195 			    wide ? " Flags" : "",
1196 			    wide ? "   Util" : "");
1197 			first = 0;
1198 		}
1199 		if (pp->pr_nget == 0 && !verbose)
1200 			continue;
1201 		if (pp->pr_maxpages == UINT_MAX)
1202 			snprintf(maxp, sizeof(maxp), "inf");
1203 		else
1204 			snprintf(maxp, sizeof(maxp), "%u", pp->pr_maxpages);
1205 /*
1206  * Print single word.  `ovflow' is number of characters didn't fit
1207  * on the last word.  `fmt' is a format string to print this word.
1208  * It must contain asterisk for field width.  `width' is a width
1209  * occupied by this word.  `fixed' is a number of constant chars in
1210  * `fmt'.  `val' is a value to be printed using format string `fmt'.
1211  */
1212 #define	PRWORD(ovflw, fmt, width, fixed, val) do {	\
1213 	(ovflw) += printf((fmt),			\
1214 	    (width) - (fixed) - (ovflw) > 0 ?		\
1215 	    (width) - (fixed) - (ovflw) : 0,		\
1216 	    (val)) - (width);				\
1217 	if ((ovflw) < 0)				\
1218 		(ovflw) = 0;				\
1219 } while (/* CONSTCOND */0)
1220 		ovflw = 0;
1221 		PRWORD(ovflw, "%-*s", wide ? 16 : 11, 0, name);
1222 		PRWORD(ovflw, " %*u", wide ? 6 : 5, 1, pp->pr_size);
1223 		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nget);
1224 		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nfail);
1225 		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nput);
1226 		if (wide)
1227 			PRWORD(ovflw, " %*u", 6, 1, pp->pr_nout);
1228 		if (wide)
1229 			PRWORD(ovflw, " %*u", 6, 1, pp->pr_nitems);
1230 		PRWORD(ovflw, " %*lu", wide ? 7 : 6, 1, pp->pr_npagealloc);
1231 		PRWORD(ovflw, " %*lu", wide ? 7 : 6, 1, pp->pr_npagefree);
1232 		PRWORD(ovflw, " %*u", 6, 1, pp->pr_npages);
1233 		if (wide)
1234 			PRWORD(ovflw, " %*u", 7, 1, pa.pa_pagesz);
1235 		PRWORD(ovflw, " %*u", 6, 1, pp->pr_hiwat);
1236 		PRWORD(ovflw, " %*u", 6, 1, pp->pr_minpages);
1237 		PRWORD(ovflw, " %*s", 6, 1, maxp);
1238 		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nidle);
1239 		if (wide)
1240 			PRWORD(ovflw, " 0x%0*x", 4, 1,
1241 			    pp->pr_flags | pp->pr_roflags);
1242 
1243 		this_inuse = pp->pr_nout * pp->pr_size;
1244 		this_total = pp->pr_npages * pa.pa_pagesz;
1245 		if (pp->pr_roflags & PR_RECURSIVE) {
1246 			/*
1247 			 * Don't count in-use memory, since it's part
1248 			 * of another pool and will be accounted for
1249 			 * there.
1250 			 */
1251 			total += (this_total - this_inuse);
1252 		} else {
1253 			inuse += this_inuse;
1254 			total += this_total;
1255 		}
1256 		if (wide) {
1257 			if (this_total == 0)
1258 				printf("   ---");
1259 			else
1260 				printf(" %5.1f%%",
1261 				    (100.0 * this_inuse) / this_total);
1262 		}
1263 		printf("\n");
1264 		dopoolcache(pp, verbose);
1265 	}
1266 
1267 	inuse /= 1024;
1268 	total /= 1024;
1269 	printf("\nIn use %ldK, total allocated %ldK; utilization %.1f%%\n",
1270 	    inuse, total, (100.0 * inuse) / total);
1271 }
1272 
1273 void
1274 dopoolcache(struct pool *pp, int verbose)
1275 {
1276 	struct pool_cache pool_cache, *pc = &pool_cache;
1277 	struct pool_cache_group pool_cache_group, *pcg = &pool_cache_group;
1278 	void *addr, *pcg_addr;
1279 	int i;
1280 
1281 	if (verbose < 1)
1282 		return;
1283 
1284 #define PR_GROUPLIST							\
1285 	deref_kptr(pcg_addr, pcg, sizeof(*pcg),				\
1286 	    "pool cache group trashed");				\
1287 	printf("\t\tgroup %p: avail %d\n", pcg_addr,			\
1288 	    pcg->pcg_avail);						\
1289 	for (i = 0; i < PCG_NOBJECTS; i++) {				\
1290 		if (pcg->pcg_objects[i].pcgo_pa !=			\
1291 		    POOL_PADDR_INVALID) {				\
1292 			printf("\t\t\t%p, 0x%llx\n",			\
1293 			    pcg->pcg_objects[i].pcgo_va,		\
1294 			    (unsigned long long)			\
1295 			    pcg->pcg_objects[i].pcgo_pa);		\
1296 		} else {						\
1297 			printf("\t\t\t%p\n",				\
1298 			    pcg->pcg_objects[i].pcgo_va);		\
1299 		}							\
1300 	}
1301 
1302 	for (addr = LIST_FIRST(&pp->pr_cachelist); addr != NULL;
1303 	    addr = LIST_NEXT(pc, pc_poollist)) {
1304 		deref_kptr(addr, pc, sizeof(*pc), "pool cache trashed");
1305 		printf("\t    hits %lu misses %lu ngroups %lu nitems %lu\n",
1306 		    pc->pc_hits, pc->pc_misses, pc->pc_ngroups, pc->pc_nitems);
1307 		if (verbose < 2)
1308 			continue;
1309 		printf("\t    full groups:\n");
1310 		for (pcg_addr = LIST_FIRST(&pc->pc_fullgroups);
1311 		    pcg_addr != NULL; pcg_addr = LIST_NEXT(pcg, pcg_list)) {
1312 			PR_GROUPLIST;
1313 		}
1314 		printf("\t    partial groups:\n");
1315 		for (pcg_addr = LIST_FIRST(&pc->pc_partgroups);
1316 		    pcg_addr != NULL; pcg_addr = LIST_NEXT(pcg, pcg_list)) {
1317 			PR_GROUPLIST;
1318 		}
1319 		printf("\t    empty groups:\n");
1320 		for (pcg_addr = LIST_FIRST(&pc->pc_emptygroups);
1321 		    pcg_addr != NULL; pcg_addr = LIST_NEXT(pcg, pcg_list)) {
1322 			PR_GROUPLIST;
1323 		}
1324 	}
1325 
1326 #undef PR_GROUPLIST
1327 
1328 }
1329 
1330 enum hashtype {			/* from <sys/systm.h> */
1331 	HASH_LIST,
1332 	HASH_TAILQ
1333 };
1334 
1335 struct uidinfo {		/* XXX: no kernel header file */
1336 	LIST_ENTRY(uidinfo) ui_hash;
1337 	uid_t	ui_uid;
1338 	long	ui_proccnt;
1339 };
1340 
1341 struct kernel_hash {
1342 	const char *	description;	/* description */
1343 	int		hashsize;	/* nlist index for hash size */
1344 	int		hashtbl;	/* nlist index for hash table */
1345 	enum hashtype	type;		/* type of hash table */
1346 	size_t		offset;		/* offset of {LIST,TAILQ}_NEXT */
1347 } khashes[] =
1348 {
1349 	{
1350 		"buffer hash",
1351 		X_BUFHASH, X_BUFHASHTBL,
1352 		HASH_LIST, offsetof(struct buf, b_hash)
1353 	}, {
1354 		"inode cache (ihash)",
1355 		X_IHASH, X_IHASHTBL,
1356 		HASH_LIST, offsetof(struct inode, i_hash)
1357 	}, {
1358 		"ipv4 address -> interface hash",
1359 		X_IFADDRHASH, X_IFADDRHASHTBL,
1360 		HASH_LIST, offsetof(struct in_ifaddr, ia_hash),
1361 	}, {
1362 		"name cache hash",
1363 		X_NCHASH, X_NCHASHTBL,
1364 		HASH_LIST, offsetof(struct namecache, nc_hash),
1365 	}, {
1366 		"name cache directory hash",
1367 		X_NCVHASH, X_NCVHASHTBL,
1368 		HASH_LIST, offsetof(struct namecache, nc_vhash),
1369 	}, {
1370 		"nfs client node cache",
1371 		X_NFSNODE, X_NFSNODETBL,
1372 		HASH_LIST, offsetof(struct nfsnode, n_hash)
1373 	}, {
1374 		"user info (uid -> used processes) hash",
1375 		X_UIHASH, X_UIHASHTBL,
1376 		HASH_LIST, offsetof(struct uidinfo, ui_hash),
1377 	}, {
1378 		NULL, -1, -1, 0, 0,
1379 	}
1380 };
1381 
1382 void
1383 dohashstat(int verbose, int todo, const char *hashname)
1384 {
1385 	LIST_HEAD(, generic)	*hashtbl_list;
1386 	TAILQ_HEAD(, generic)	*hashtbl_tailq;
1387 	struct kernel_hash	*curhash;
1388 	void	*hashaddr, *hashbuf, *nhashbuf, *nextaddr;
1389 	size_t	elemsize, hashbufsize, thissize;
1390 	u_long	hashsize;
1391 	int	i, used, items, chain, maxchain;
1392 
1393 	hashbuf = NULL;
1394 	hashbufsize = 0;
1395 
1396 	if (todo & HASHLIST) {
1397 		printf("Supported hashes:\n");
1398 		for (curhash = khashes; curhash->description; curhash++) {
1399 			if (hashnl[curhash->hashsize].n_value == 0 ||
1400 			    hashnl[curhash->hashtbl].n_value == 0)
1401 				continue;
1402 			printf("\t%-16s%s\n",
1403 			    hashnl[curhash->hashsize].n_name + 1,
1404 			    curhash->description);
1405 		}
1406 		return;
1407 	}
1408 
1409 	if (hashname != NULL) {
1410 		for (curhash = khashes; curhash->description; curhash++) {
1411 			if (strcmp(hashnl[curhash->hashsize].n_name + 1,
1412 			    hashname) == 0 &&
1413 			    hashnl[curhash->hashsize].n_value != 0 &&
1414 			    hashnl[curhash->hashtbl].n_value != 0)
1415 				break;
1416 		}
1417 		if (curhash->description == NULL) {
1418 			warnx("%s: no such hash", hashname);
1419 			return;
1420 		}
1421 	}
1422 
1423 	printf(
1424 	    "%-16s %8s %8s %8s %8s %8s %8s\n"
1425 	    "%-16s %8s %8s %8s %8s %8s %8s\n",
1426 	    "", "total", "used", "util", "num", "average", "maximum",
1427 	    "hash table", "buckets", "buckets", "%", "items", "chain",
1428 	    "chain");
1429 
1430 	for (curhash = khashes; curhash->description; curhash++) {
1431 		if (hashnl[curhash->hashsize].n_value == 0 ||
1432 		    hashnl[curhash->hashtbl].n_value == 0)
1433 			continue;
1434 		if (hashname != NULL &&
1435 		    strcmp(hashnl[curhash->hashsize].n_name + 1, hashname))
1436 			continue;
1437 		elemsize = curhash->type == HASH_LIST ?
1438 		    sizeof(*hashtbl_list) : sizeof(*hashtbl_tailq);
1439 		deref_kptr((void *)hashnl[curhash->hashsize].n_value,
1440 		    &hashsize, sizeof(hashsize),
1441 		    hashnl[curhash->hashsize].n_name);
1442 		hashsize++;
1443 		deref_kptr((void *)hashnl[curhash->hashtbl].n_value,
1444 		    &hashaddr, sizeof(hashaddr),
1445 		    hashnl[curhash->hashtbl].n_name);
1446 		if (verbose)
1447 			printf("%s %lu, %s %p, offset %ld, elemsize %llu\n",
1448 			    hashnl[curhash->hashsize].n_name + 1, hashsize,
1449 			    hashnl[curhash->hashtbl].n_name + 1, hashaddr,
1450 			    (long)curhash->offset,
1451 			    (unsigned long long)elemsize);
1452 		thissize = hashsize * elemsize;
1453 		if (thissize > hashbufsize) {
1454 			if ((nhashbuf = realloc(hashbuf, thissize)) == NULL)
1455 				errx(1, "malloc hashbuf %llu",
1456 				    (unsigned long long)hashbufsize);
1457 			hashbuf = nhashbuf;
1458 			hashbufsize = thissize;
1459 		}
1460 		deref_kptr(hashaddr, hashbuf, thissize,
1461 		    hashnl[curhash->hashtbl].n_name);
1462 		used = 0;
1463 		items = maxchain = 0;
1464 		if (curhash->type == HASH_LIST) {
1465 			hashtbl_list = hashbuf;
1466 			hashtbl_tailq = NULL;
1467 		} else {
1468 			hashtbl_list = NULL;
1469 			hashtbl_tailq = hashbuf;
1470 		}
1471 		for (i = 0; i < hashsize; i++) {
1472 			if (curhash->type == HASH_LIST)
1473 				nextaddr = LIST_FIRST(&hashtbl_list[i]);
1474 			else
1475 				nextaddr = TAILQ_FIRST(&hashtbl_tailq[i]);
1476 			if (nextaddr == NULL)
1477 				continue;
1478 			if (verbose)
1479 				printf("%5d: %p\n", i, nextaddr);
1480 			used++;
1481 			chain = 0;
1482 			do {
1483 				chain++;
1484 				deref_kptr((char *)nextaddr + curhash->offset,
1485 				    &nextaddr, sizeof(void *),
1486 				    "hash chain corrupted");
1487 				if (verbose > 1)
1488 					printf("got nextaddr as %p\n",
1489 					    nextaddr);
1490 			} while (nextaddr != NULL);
1491 			items += chain;
1492 			if (verbose && chain > 1)
1493 				printf("\tchain = %d\n", chain);
1494 			if (chain > maxchain)
1495 				maxchain = chain;
1496 		}
1497 		printf("%-16s %8ld %8d %8.2f %8d %8.2f %8d\n",
1498 		    hashnl[curhash->hashsize].n_name + 1,
1499 		    hashsize, used, used * 100.0 / hashsize,
1500 		    items, used ? (double)items / used : 0.0, maxchain);
1501 	}
1502 }
1503 
1504 /*
1505  * kread reads something from the kernel, given its nlist index in namelist[].
1506  */
1507 void
1508 kread(struct nlist *nl, int nlx, void *addr, size_t size)
1509 {
1510 	const char *sym;
1511 
1512 	sym = nl[nlx].n_name;
1513 	if (*sym == '_')
1514 		++sym;
1515 	if (nl[nlx].n_type == 0 || nl[nlx].n_value == 0)
1516 		errx(1, "symbol %s not defined", sym);
1517 	deref_kptr((void *)nl[nlx].n_value, addr, size, sym);
1518 }
1519 
1520 /*
1521  * Dereference the kernel pointer `kptr' and fill in the local copy
1522  * pointed to by `ptr'.  The storage space must be pre-allocated,
1523  * and the size of the copy passed in `len'.
1524  */
1525 void
1526 deref_kptr(const void *kptr, void *ptr, size_t len, const char *msg)
1527 {
1528 
1529 	if (*msg == '_')
1530 		msg++;
1531 	if (kvm_read(kd, (u_long)kptr, (char *)ptr, len) != len)
1532 		errx(1, "kptr %lx: %s: %s", (u_long)kptr, msg, kvm_geterr(kd));
1533 }
1534 
1535 /*
1536  * Traverse the UVM history buffers, performing the requested action.
1537  *
1538  * Note, we assume that if we're not listing, we're dumping.
1539  */
1540 void
1541 hist_traverse(int todo, const char *histname)
1542 {
1543 	struct uvm_history_head histhead;
1544 	struct uvm_history hist, *histkva;
1545 	char *name = NULL;
1546 	size_t namelen = 0;
1547 
1548 	if (histnl[0].n_value == 0) {
1549 		warnx("UVM history is not compiled into the kernel.");
1550 		return;
1551 	}
1552 
1553 	deref_kptr((void *)histnl[X_UVM_HISTORIES].n_value, &histhead,
1554 	    sizeof(histhead), histnl[X_UVM_HISTORIES].n_name);
1555 
1556 	if (histhead.lh_first == NULL) {
1557 		warnx("No active UVM history logs.");
1558 		return;
1559 	}
1560 
1561 	if (todo & HISTLIST)
1562 		printf("Active UVM histories:");
1563 
1564 	for (histkva = LIST_FIRST(&histhead); histkva != NULL;
1565 	    histkva = LIST_NEXT(&hist, list)) {
1566 		deref_kptr(histkva, &hist, sizeof(hist), "histkva");
1567 		if (hist.namelen > namelen) {
1568 			if (name != NULL)
1569 				free(name);
1570 			namelen = hist.namelen;
1571 			if ((name = malloc(namelen + 1)) == NULL)
1572 				err(1, "malloc history name");
1573 		}
1574 
1575 		deref_kptr(hist.name, name, namelen, "history name");
1576 		name[namelen] = '\0';
1577 		if (todo & HISTLIST)
1578 			printf(" %s", name);
1579 		else {
1580 			/*
1581 			 * If we're dumping all histories, do it, else
1582 			 * check to see if this is the one we want.
1583 			 */
1584 			if (histname == NULL || strcmp(histname, name) == 0) {
1585 				if (histname == NULL)
1586 					printf("\nUVM history `%s':\n", name);
1587 				hist_dodump(&hist);
1588 			}
1589 		}
1590 	}
1591 
1592 	if (todo & HISTLIST)
1593 		putchar('\n');
1594 
1595 	if (name != NULL)
1596 		free(name);
1597 }
1598 
1599 /*
1600  * Actually dump the history buffer at the specified KVA.
1601  */
1602 void
1603 hist_dodump(struct uvm_history *histp)
1604 {
1605 	struct uvm_history_ent *histents, *e;
1606 	size_t histsize;
1607 	char *fmt = NULL, *fn = NULL;
1608 	size_t fmtlen = 0, fnlen = 0;
1609 	int i;
1610 
1611 	histsize = sizeof(struct uvm_history_ent) * histp->n;
1612 
1613 	if ((histents = malloc(histsize)) == NULL)
1614 		err(1, "malloc history entries");
1615 
1616 	memset(histents, 0, histsize);
1617 
1618 	deref_kptr(histp->e, histents, histsize, "history entries");
1619 	i = histp->f;
1620 	do {
1621 		e = &histents[i];
1622 		if (e->fmt != NULL) {
1623 			if (e->fmtlen > fmtlen) {
1624 				if (fmt != NULL)
1625 					free(fmt);
1626 				fmtlen = e->fmtlen;
1627 				if ((fmt = malloc(fmtlen + 1)) == NULL)
1628 					err(1, "malloc printf format");
1629 			}
1630 			if (e->fnlen > fnlen) {
1631 				if (fn != NULL)
1632 					free(fn);
1633 				fnlen = e->fnlen;
1634 				if ((fn = malloc(fnlen + 1)) == NULL)
1635 					err(1, "malloc function name");
1636 			}
1637 
1638 			deref_kptr(e->fmt, fmt, fmtlen, "printf format");
1639 			fmt[fmtlen] = '\0';
1640 
1641 			deref_kptr(e->fn, fn, fnlen, "function name");
1642 			fn[fnlen] = '\0';
1643 
1644 			printf("%06ld.%06ld ", (long int)e->tv.tv_sec,
1645 			    (long int)e->tv.tv_usec);
1646 			printf("%s#%ld: ", fn, e->call);
1647 			printf(fmt, e->v[0], e->v[1], e->v[2], e->v[3]);
1648 			putchar('\n');
1649 		}
1650 		i = (i + 1) % histp->n;
1651 	} while (i != histp->f);
1652 
1653 	free(histents);
1654 	if (fmt != NULL)
1655 		free(fmt);
1656 	if (fn != NULL)
1657 		free(fn);
1658 }
1659 
1660 void
1661 usage(void)
1662 {
1663 
1664 	(void)fprintf(stderr,
1665 	    "usage: %s [-efHiLlmstUvW] [-c count] [-h hashname] [-M core] [-N system]\n"
1666 	    "\t\t[-u histname] [-w wait] [disks]\n", getprogname());
1667 	exit(1);
1668 }
1669