xref: /netbsd-src/usr.bin/vmstat/vmstat.c (revision fad4c9f71477ae11cea2ee75ec82151ac770a534)
1 /* $NetBSD: vmstat.c,v 1.145 2006/06/07 20:58:23 kardel 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 1.145 2006/06/07 20:58:23 kardel 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 "drvstats.h"
136 
137 /*
138  * General namelist
139  */
140 struct nlist namelist[] =
141 {
142 #define	X_BOOTTIME	0
143 	{ "_boottime" },
144 #define	X_HZ		1
145 	{ "_hz" },
146 #define	X_STATHZ	2
147 	{ "_stathz" },
148 #define	X_NCHSTATS	3
149 	{ "_nchstats" },
150 #define	X_KMEMSTAT	4
151 	{ "_kmemstatistics" },
152 #define	X_KMEMBUCKETS	5
153 	{ "_kmembuckets" },
154 #define	X_ALLEVENTS	6
155 	{ "_allevents" },
156 #define	X_POOLHEAD	7
157 	{ "_pool_head" },
158 #define	X_UVMEXP	8
159 	{ "_uvmexp" },
160 #define	X_TIME_SECOND	9
161 #ifdef __HAVE_TIMECOUNTER
162 	{ "_time_second" },
163 #else
164 	{ "_time" },	/* XXX uses same array slot as "X_TIME_SECOND" */
165 #endif
166 #define	X_NL_SIZE	10
167 	{ NULL },
168 };
169 
170 /*
171  * Namelist for pre-evcnt interrupt counters.
172  */
173 struct nlist intrnl[] =
174 {
175 #define	X_INTRNAMES	0
176 	{ "_intrnames" },
177 #define	X_EINTRNAMES	1
178 	{ "_eintrnames" },
179 #define	X_INTRCNT	2
180 	{ "_intrcnt" },
181 #define	X_EINTRCNT	3
182 	{ "_eintrcnt" },
183 #define	X_INTRNL_SIZE	4
184 	{ NULL },
185 };
186 
187 
188 /*
189  * Namelist for hash statistics
190  */
191 struct nlist hashnl[] =
192 {
193 #define	X_NFSNODE	0
194 	{ "_nfsnodehash" },
195 #define	X_NFSNODETBL	1
196 	{ "_nfsnodehashtbl" },
197 #define	X_IHASH		2
198 	{ "_ihash" },
199 #define	X_IHASHTBL	3
200 	{ "_ihashtbl" },
201 #define	X_BUFHASH	4
202 	{ "_bufhash" },
203 #define	X_BUFHASHTBL	5
204 	{ "_bufhashtbl" },
205 #define	X_UIHASH	6
206 	{ "_uihash" },
207 #define	X_UIHASHTBL	7
208 	{ "_uihashtbl" },
209 #define	X_IFADDRHASH	8
210 	{ "_in_ifaddrhash" },
211 #define	X_IFADDRHASHTBL	9
212 	{ "_in_ifaddrhashtbl" },
213 #define	X_NCHASH	10
214 	{ "_nchash" },
215 #define	X_NCHASHTBL	11
216 	{ "_nchashtbl" },
217 #define	X_NCVHASH	12
218 	{ "_ncvhash" },
219 #define	X_NCVHASHTBL	13
220 	{ "_ncvhashtbl" },
221 #define X_HASHNL_SIZE	14	/* must be last */
222 	{ NULL },
223 };
224 
225 /*
226  * Namelist for UVM histories
227  */
228 struct nlist histnl[] =
229 {
230 	{ "_uvm_histories" },
231 #define	X_UVM_HISTORIES		0
232 	{ NULL },
233 };
234 
235 
236 
237 struct	uvmexp uvmexp, ouvmexp;
238 int	ndrives;
239 
240 int	winlines = 20;
241 
242 kvm_t *kd;
243 
244 #define	FORKSTAT	1<<0
245 #define	INTRSTAT	1<<1
246 #define	MEMSTAT		1<<2
247 #define	SUMSTAT		1<<3
248 #define	EVCNTSTAT	1<<4
249 #define	VMSTAT		1<<5
250 #define	HISTLIST	1<<6
251 #define	HISTDUMP	1<<7
252 #define	HASHSTAT	1<<8
253 #define	HASHLIST	1<<9
254 #define	VMTOTAL		1<<10
255 
256 void	cpustats(void);
257 void	deref_kptr(const void *, void *, size_t, const char *);
258 void	drvstats(void);
259 void	doevcnt(int verbose);
260 void	dohashstat(int, int, const char *);
261 void	dointr(int verbose);
262 void	domem(void);
263 void	dopool(int, int);
264 void	dopoolcache(struct pool *, int);
265 void	dosum(void);
266 void	dovmstat(struct timespec *, int);
267 void	print_total_hdr(void);
268 void	dovmtotal(struct timespec *, int);
269 void	kread(struct nlist *, int, void *, size_t);
270 void	needhdr(int);
271 long	getuptime(void);
272 void	printhdr(void);
273 long	pct(long, long);
274 void	usage(void);
275 void	doforkst(void);
276 
277 void	hist_traverse(int, const char *);
278 void	hist_dodump(struct uvm_history *);
279 
280 int	main(int, char **);
281 char	**choosedrives(char **);
282 
283 /* Namelist and memory file names. */
284 char	*nlistf, *memf;
285 
286 /* allow old usage [vmstat 1] */
287 #define	BACKWARD_COMPATIBILITY
288 
289 int
290 main(int argc, char *argv[])
291 {
292 	int c, todo, verbose, wide;
293 	struct timespec interval;
294 	int reps;
295 	char errbuf[_POSIX2_LINE_MAX];
296 	gid_t egid = getegid();
297 	const char *histname, *hashname;
298 
299 	histname = hashname = NULL;
300 	(void)setegid(getgid());
301 	memf = nlistf = NULL;
302 	reps = todo = verbose = wide = 0;
303 	interval.tv_sec = 0;
304 	interval.tv_nsec = 0;
305 	while ((c = getopt(argc, argv, "c:efh:HilLM:mN:stu:UvWw:")) != -1) {
306 		switch (c) {
307 		case 'c':
308 			reps = atoi(optarg);
309 			break;
310 		case 'e':
311 			todo |= EVCNTSTAT;
312 			break;
313 		case 'f':
314 			todo |= FORKSTAT;
315 			break;
316 		case 'h':
317 			hashname = optarg;
318 			/* FALLTHROUGH */
319 		case 'H':
320 			todo |= HASHSTAT;
321 			break;
322 		case 'i':
323 			todo |= INTRSTAT;
324 			break;
325 		case 'l':
326 			todo |= HISTLIST;
327 			break;
328 		case 'L':
329 			todo |= HASHLIST;
330 			break;
331 		case 'M':
332 			memf = optarg;
333 			break;
334 		case 'm':
335 			todo |= MEMSTAT;
336 			break;
337 		case 'N':
338 			nlistf = optarg;
339 			break;
340 		case 's':
341 			todo |= SUMSTAT;
342 			break;
343 		case 't':
344 			todo |= VMTOTAL;
345 			break;
346 		case 'u':
347 			histname = optarg;
348 			/* FALLTHROUGH */
349 		case 'U':
350 			todo |= HISTDUMP;
351 			break;
352 		case 'v':
353 			verbose++;
354 			break;
355 		case 'W':
356 			wide++;
357 			break;
358 		case 'w':
359 			interval.tv_sec = atol(optarg);
360 			break;
361 		case '?':
362 		default:
363 			usage();
364 		}
365 	}
366 	argc -= optind;
367 	argv += optind;
368 
369 	if (todo == 0)
370 		todo = VMSTAT;
371 
372 	/*
373 	 * Discard setgid privileges.  If not the running kernel, we toss
374 	 * them away totally so that bad guys can't print interesting stuff
375 	 * from kernel memory, otherwise switch back to kmem for the
376 	 * duration of the kvm_openfiles() call.
377 	 */
378 	if (nlistf != NULL || memf != NULL)
379 		(void)setgid(getgid());
380 	else
381 		(void)setegid(egid);
382 
383 	kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, errbuf);
384 	if (kd == NULL)
385 		errx(1, "kvm_openfiles: %s", errbuf);
386 
387 	if (nlistf == NULL && memf == NULL)
388 		(void)setgid(getgid());
389 
390 	if ((c = kvm_nlist(kd, namelist)) != 0) {
391 		if (c == -1)
392 			errx(1, "kvm_nlist: %s %s", "namelist", kvm_geterr(kd));
393 		(void)fprintf(stderr, "vmstat: undefined symbols:");
394 		for (c = 0; c < sizeof(namelist) / sizeof(namelist[0])-1; c++)
395 			if (namelist[c].n_type == 0)
396 				fprintf(stderr, " %s", namelist[c].n_name);
397 		(void)fputc('\n', stderr);
398 		exit(1);
399 	}
400 	if (todo & INTRSTAT)
401 		(void) kvm_nlist(kd, intrnl);
402 	if ((c = kvm_nlist(kd, hashnl)) == -1 || c == X_HASHNL_SIZE)
403 		errx(1, "kvm_nlist: %s %s", "hashnl", kvm_geterr(kd));
404 	if (kvm_nlist(kd, histnl) == -1)
405 		errx(1, "kvm_nlist: %s %s", "histnl", kvm_geterr(kd));
406 
407 	if (todo & VMSTAT) {
408 		struct winsize winsize;
409 
410 		drvinit(0);	/* Initialize disk stats, no disks selected. */
411 
412 		(void)setgid(getgid()); /* don't need privs anymore */
413 
414 		argv = choosedrives(argv);	/* Select disks. */
415 		winsize.ws_row = 0;
416 		(void)ioctl(STDOUT_FILENO, TIOCGWINSZ, &winsize);
417 		if (winsize.ws_row > 0)
418 			winlines = winsize.ws_row;
419 
420 	}
421 
422 #ifdef	BACKWARD_COMPATIBILITY
423 	if (*argv) {
424 		interval.tv_sec = atol(*argv);
425 		if (*++argv)
426 			reps = atoi(*argv);
427 	}
428 #endif
429 
430 	if (interval.tv_sec) {
431 		if (!reps)
432 			reps = -1;
433 	} else if (reps)
434 		interval.tv_sec = 1;
435 
436 
437 	/*
438 	 * Statistics dumping is incompatible with the default
439 	 * VMSTAT/dovmstat() output. So perform the interval/reps handling
440 	 * for it here.
441 	 */
442 	if ((todo & (VMSTAT|VMTOTAL)) == 0) {
443 		for (;;) {
444 			if (todo & (HISTLIST|HISTDUMP)) {
445 				if ((todo & (HISTLIST|HISTDUMP)) ==
446 				    (HISTLIST|HISTDUMP))
447 					errx(1, "you may list or dump,"
448 					    " but not both!");
449 				hist_traverse(todo, histname);
450 				putchar('\n');
451 			}
452 			if (todo & FORKSTAT) {
453 				doforkst();
454 				putchar('\n');
455 			}
456 			if (todo & MEMSTAT) {
457 				domem();
458 				dopool(verbose, wide);
459 				putchar('\n');
460 			}
461 			if (todo & SUMSTAT) {
462 				dosum();
463 				putchar('\n');
464 			}
465 			if (todo & INTRSTAT) {
466 				dointr(verbose);
467 				putchar('\n');
468 			}
469 			if (todo & EVCNTSTAT) {
470 				doevcnt(verbose);
471 				putchar('\n');
472 			}
473 			if (todo & (HASHLIST|HASHSTAT)) {
474 				if ((todo & (HASHLIST|HASHSTAT)) ==
475 				    (HASHLIST|HASHSTAT))
476 					errx(1, "you may list or display,"
477 					    " but not both!");
478 				dohashstat(verbose, todo, hashname);
479 				putchar('\n');
480 			}
481 
482 			if (reps >= 0 && --reps <=0)
483 				break;
484 			nanosleep(&interval, NULL);
485 		}
486 	} else {
487 		if ((todo & (VMSTAT|VMTOTAL)) == (VMSTAT|VMTOTAL)) {
488 			errx(1, "you may not both do vmstat and vmtotal");
489 		}
490 		if (todo & VMSTAT)
491 			dovmstat(&interval, reps);
492 		if (todo & VMTOTAL)
493 			dovmtotal(&interval, reps);
494 	}
495 	exit(0);
496 }
497 
498 char **
499 choosedrives(char **argv)
500 {
501 	int i;
502 
503 	/*
504 	 * Choose drives to be displayed.  Priority goes to (in order) drives
505 	 * supplied as arguments, default drives.  If everything isn't filled
506 	 * in and there are drives not taken care of, display the first few
507 	 * that fit.
508 	 */
509 #define	BACKWARD_COMPATIBILITY
510 	for (ndrives = 0; *argv; ++argv) {
511 #ifdef	BACKWARD_COMPATIBILITY
512 		if (isdigit((unsigned char)**argv))
513 			break;
514 #endif
515 		for (i = 0; i < ndrive; i++) {
516 			if (strcmp(dr_name[i], *argv))
517 				continue;
518 			drv_select[i] = 1;
519 			++ndrives;
520 			break;
521 		}
522 	}
523 	for (i = 0; i < ndrive && ndrives < 3; i++) {
524 		if (drv_select[i])
525 			continue;
526 		drv_select[i] = 1;
527 		++ndrives;
528 	}
529 
530 	return (argv);
531 }
532 
533 long
534 getuptime(void)
535 {
536 	static struct timeval boottime;
537 	struct timeval now;
538 	time_t uptime;
539 
540 	if (boottime.tv_sec == 0)
541 		kread(namelist, X_BOOTTIME, &boottime, sizeof(boottime));
542 #ifdef __HAVE_TIMECOUNTER
543 	kread(namelist, X_TIME_SECOND, &now.tv_sec, sizeof(now));
544 	now.tv_usec = 0;
545 #else
546 	kread(namelist, X_TIME_SECOND, &now, sizeof(now));
547 #endif
548 	uptime = now.tv_sec - boottime.tv_sec;
549 	if (uptime <= 0 || uptime > 60*60*24*365*10)
550 		errx(1, "time makes no sense; namelist must be wrong.");
551 	return (uptime);
552 }
553 
554 int	hz, hdrcnt;
555 
556 void
557 print_total_hdr()
558 {
559 
560 	(void)printf("procs            memory\n");
561 	(void)printf("ru dw pw sl sw");
562 	(void)printf("   total-v  active-v  active-r");
563 	(void)printf(" vm-sh avm-sh rm-sh arm-sh free\n");
564 	hdrcnt = winlines - 2;
565 }
566 
567 void
568 dovmtotal(struct timespec *interval, int reps)
569 {
570 	struct vmtotal total;
571 	int mib[2];
572 	size_t size;
573 
574 	(void)signal(SIGCONT, needhdr);
575 
576 	for (hdrcnt = 1;;) {
577 		if (!--hdrcnt)
578 			print_total_hdr();
579 		if (memf != NULL) {
580 			printf("Unable to get vmtotals from crash dump.\n");
581 			memset(&total, 0, sizeof(total));
582 		} else {
583 			size = sizeof(total);
584 			mib[0] = CTL_VM;
585 			mib[1] = VM_METER;
586 			if (sysctl(mib, 2, &total, &size, NULL, 0) < 0) {
587 				printf("Can't get vmtotals: %s\n",
588 				    strerror(errno));
589 				memset(&total, 0, sizeof(total));
590 			}
591 		}
592 		printf("%2d ", total.t_rq);
593 		printf("%2d ", total.t_dw);
594 		printf("%2d ", total.t_pw);
595 		printf("%2d ", total.t_sl);
596 		printf("%2d ", total.t_sw);
597 
598 		printf("%9d ", total.t_vm);
599 		printf("%9d ", total.t_avm);
600 		printf("%9d ", total.t_arm);
601 		printf("%5d ", total.t_vmshr);
602 		printf("%6d ", total.t_avmshr);
603 		printf("%5d ", total.t_rmshr);
604 		printf("%6d ", total.t_armshr);
605 		printf("%5d",  total.t_free);
606 
607 		putchar('\n');
608 
609 		(void)fflush(stdout);
610 		if (reps >= 0 && --reps <= 0)
611 			break;
612 
613 		nanosleep(interval, NULL);
614 	}
615 }
616 
617 void
618 dovmstat(struct timespec *interval, int reps)
619 {
620 	struct vmtotal total;
621 	time_t uptime, halfuptime;
622 	int mib[2];
623 	size_t size;
624 	int pagesize = getpagesize();
625 
626 	uptime = getuptime();
627 	halfuptime = uptime / 2;
628 	(void)signal(SIGCONT, needhdr);
629 
630 	if (namelist[X_STATHZ].n_type != 0 && namelist[X_STATHZ].n_value != 0)
631 		kread(namelist, X_STATHZ, &hz, sizeof(hz));
632 	if (!hz)
633 		kread(namelist, X_HZ, &hz, sizeof(hz));
634 
635 	for (hdrcnt = 1;;) {
636 		if (!--hdrcnt)
637 			printhdr();
638 		/* Read new disk statistics */
639 		cpureadstats();
640 		drvreadstats();
641 		tkreadstats();
642 		kread(namelist, X_UVMEXP, &uvmexp, sizeof(uvmexp));
643 		if (memf != NULL) {
644 			/*
645 			 * XXX Can't do this if we're reading a crash
646 			 * XXX dump because they're lazily-calculated.
647 			 */
648 			printf("Unable to get vmtotals from crash dump.\n");
649 			memset(&total, 0, sizeof(total));
650 		} else {
651 			size = sizeof(total);
652 			mib[0] = CTL_VM;
653 			mib[1] = VM_METER;
654 			if (sysctl(mib, 2, &total, &size, NULL, 0) < 0) {
655 				printf("Can't get vmtotals: %s\n",
656 				    strerror(errno));
657 				memset(&total, 0, sizeof(total));
658 			}
659 		}
660 		(void)printf("%2d %d %d",
661 		    total.t_rq - 1, total.t_dw + total.t_pw, total.t_sw);
662 #define	pgtok(a) (long)((a) * (pagesize >> 10))
663 #define	rate(x)	(u_long)(((x) + halfuptime) / uptime)	/* round */
664 		(void)printf(" %6ld %6ld ",
665 		    pgtok(total.t_avm), pgtok(total.t_free));
666 		(void)printf("%4lu ", rate(uvmexp.faults - ouvmexp.faults));
667 		(void)printf("%3lu ", rate(uvmexp.pdreact - ouvmexp.pdreact));
668 		(void)printf("%3lu ", rate(uvmexp.pageins - ouvmexp.pageins));
669 		(void)printf("%4lu ",
670 		    rate(uvmexp.pgswapout - ouvmexp.pgswapout));
671 		(void)printf("%4lu ", rate(uvmexp.pdfreed - ouvmexp.pdfreed));
672 		(void)printf("%4lu ", rate(uvmexp.pdscans - ouvmexp.pdscans));
673 		drvstats();
674 		(void)printf("%4lu %4lu %3lu ",
675 		    rate(uvmexp.intrs - ouvmexp.intrs),
676 		    rate(uvmexp.syscalls - ouvmexp.syscalls),
677 		    rate(uvmexp.swtch - ouvmexp.swtch));
678 		cpustats();
679 		putchar('\n');
680 		(void)fflush(stdout);
681 		if (reps >= 0 && --reps <= 0)
682 			break;
683 		ouvmexp = uvmexp;
684 		uptime = interval->tv_sec;
685 		/*
686 		 * We round upward to avoid losing low-frequency events
687 		 * (i.e., >= 1 per interval but < 1 per second).
688 		 */
689 		halfuptime = uptime == 1 ? 0 : (uptime + 1) / 2;
690 		nanosleep(interval, NULL);
691 	}
692 }
693 
694 void
695 printhdr(void)
696 {
697 	int i;
698 
699 	(void)printf(" procs    memory      page%*s", 23, "");
700 	if (ndrives > 0)
701 		(void)printf("%s %*sfaults      cpu\n",
702 		    ((ndrives > 1) ? "disks" : "disk"),
703 		    ((ndrives > 1) ? ndrives * 3 - 4 : 0), "");
704 	else
705 		(void)printf("%*s  faults   cpu\n",
706 		    ndrives * 3, "");
707 
708 	(void)printf(" r b w    avm    fre  flt  re  pi   po   fr   sr ");
709 	for (i = 0; i < ndrive; i++)
710 		if (drv_select[i])
711 			(void)printf("%c%c ", dr_name[i][0],
712 			    dr_name[i][strlen(dr_name[i]) - 1]);
713 	(void)printf("  in   sy  cs us sy id\n");
714 	hdrcnt = winlines - 2;
715 }
716 
717 /*
718  * Force a header to be prepended to the next output.
719  */
720 void
721 needhdr(int dummy)
722 {
723 
724 	hdrcnt = 1;
725 }
726 
727 long
728 pct(long top, long bot)
729 {
730 	long ans;
731 
732 	if (bot == 0)
733 		return (0);
734 	ans = (quad_t)top * 100 / bot;
735 	return (ans);
736 }
737 
738 #define	PCT(top, bot) (int)pct((long)(top), (long)(bot))
739 
740 void
741 dosum(void)
742 {
743 	struct nchstats nchstats;
744 	u_long nchtotal;
745 
746 	kread(namelist, X_UVMEXP, &uvmexp, sizeof(uvmexp));
747 
748 	(void)printf("%9u bytes per page\n", uvmexp.pagesize);
749 
750 	(void)printf("%9u page color%s\n",
751 	    uvmexp.ncolors, uvmexp.ncolors == 1 ? "" : "s");
752 
753 	(void)printf("%9u pages managed\n", uvmexp.npages);
754 	(void)printf("%9u pages free\n", uvmexp.free);
755 	(void)printf("%9u pages active\n", uvmexp.active);
756 	(void)printf("%9u pages inactive\n", uvmexp.inactive);
757 	(void)printf("%9u pages paging\n", uvmexp.paging);
758 	(void)printf("%9u pages wired\n", uvmexp.wired);
759 	(void)printf("%9u zero pages\n", uvmexp.zeropages);
760 	(void)printf("%9u reserve pagedaemon pages\n",
761 	    uvmexp.reserve_pagedaemon);
762 	(void)printf("%9u reserve kernel pages\n", uvmexp.reserve_kernel);
763 	(void)printf("%9u anonymous pages\n", uvmexp.anonpages);
764 	(void)printf("%9u cached file pages\n", uvmexp.filepages);
765 	(void)printf("%9u cached executable pages\n", uvmexp.execpages);
766 
767 	(void)printf("%9u minimum free pages\n", uvmexp.freemin);
768 	(void)printf("%9u target free pages\n", uvmexp.freetarg);
769 	(void)printf("%9u target inactive pages\n", uvmexp.inactarg);
770 	(void)printf("%9u maximum wired pages\n", uvmexp.wiredmax);
771 
772 	(void)printf("%9u swap devices\n", uvmexp.nswapdev);
773 	(void)printf("%9u swap pages\n", uvmexp.swpages);
774 	(void)printf("%9u swap pages in use\n", uvmexp.swpginuse);
775 	(void)printf("%9u swap allocations\n", uvmexp.nswget);
776 
777 	(void)printf("%9u total faults taken\n", uvmexp.faults);
778 	(void)printf("%9u traps\n", uvmexp.traps);
779 	(void)printf("%9u device interrupts\n", uvmexp.intrs);
780 	(void)printf("%9u CPU context switches\n", uvmexp.swtch);
781 	(void)printf("%9u software interrupts\n", uvmexp.softs);
782 	(void)printf("%9u system calls\n", uvmexp.syscalls);
783 	(void)printf("%9u pagein requests\n", uvmexp.pageins);
784 	(void)printf("%9u pageout requests\n", uvmexp.pdpageouts);
785 	(void)printf("%9u swap ins\n", uvmexp.swapins);
786 	(void)printf("%9u swap outs\n", uvmexp.swapouts);
787 	(void)printf("%9u pages swapped in\n", uvmexp.pgswapin);
788 	(void)printf("%9u pages swapped out\n", uvmexp.pgswapout);
789 	(void)printf("%9u forks total\n", uvmexp.forks);
790 	(void)printf("%9u forks blocked parent\n", uvmexp.forks_ppwait);
791 	(void)printf("%9u forks shared address space with parent\n",
792 	    uvmexp.forks_sharevm);
793 	(void)printf("%9u pagealloc zero wanted and avail\n",
794 	    uvmexp.pga_zerohit);
795 	(void)printf("%9u pagealloc zero wanted and not avail\n",
796 	    uvmexp.pga_zeromiss);
797 	(void)printf("%9u aborts of idle page zeroing\n",
798 	    uvmexp.zeroaborts);
799 	(void)printf("%9u pagealloc desired color avail\n",
800 	    uvmexp.colorhit);
801 	(void)printf("%9u pagealloc desired color not avail\n",
802 	    uvmexp.colormiss);
803 
804 	(void)printf("%9u faults with no memory\n", uvmexp.fltnoram);
805 	(void)printf("%9u faults with no anons\n", uvmexp.fltnoanon);
806 	(void)printf("%9u faults had to wait on pages\n", uvmexp.fltpgwait);
807 	(void)printf("%9u faults found released page\n", uvmexp.fltpgrele);
808 	(void)printf("%9u faults relock (%u ok)\n", uvmexp.fltrelck,
809 	    uvmexp.fltrelckok);
810 	(void)printf("%9u anon page faults\n", uvmexp.fltanget);
811 	(void)printf("%9u anon retry faults\n", uvmexp.fltanretry);
812 	(void)printf("%9u amap copy faults\n", uvmexp.fltamcopy);
813 	(void)printf("%9u neighbour anon page faults\n", uvmexp.fltnamap);
814 	(void)printf("%9u neighbour object page faults\n", uvmexp.fltnomap);
815 	(void)printf("%9u locked pager get faults\n", uvmexp.fltlget);
816 	(void)printf("%9u unlocked pager get faults\n", uvmexp.fltget);
817 	(void)printf("%9u anon faults\n", uvmexp.flt_anon);
818 	(void)printf("%9u anon copy on write faults\n", uvmexp.flt_acow);
819 	(void)printf("%9u object faults\n", uvmexp.flt_obj);
820 	(void)printf("%9u promote copy faults\n", uvmexp.flt_prcopy);
821 	(void)printf("%9u promote zero fill faults\n", uvmexp.flt_przero);
822 
823 	(void)printf("%9u times daemon wokeup\n",uvmexp.pdwoke);
824 	(void)printf("%9u revolutions of the clock hand\n", uvmexp.pdrevs);
825 	(void)printf("%9u times daemon attempted swapout\n", uvmexp.pdswout);
826 	(void)printf("%9u pages freed by daemon\n", uvmexp.pdfreed);
827 	(void)printf("%9u pages scanned by daemon\n", uvmexp.pdscans);
828 	(void)printf("%9u anonymous pages scanned by daemon\n",
829 	    uvmexp.pdanscan);
830 	(void)printf("%9u object pages scanned by daemon\n", uvmexp.pdobscan);
831 	(void)printf("%9u pages reactivated\n", uvmexp.pdreact);
832 	(void)printf("%9u anonymous pages reactivated\n", uvmexp.pdreanon);
833 	(void)printf("%9u cached file pages reactivated\n", uvmexp.pdrefile);
834 	(void)printf("%9u cached executable pages reactivated\n",
835 	    uvmexp.pdreexec);
836 	(void)printf("%9u pages found busy by daemon\n", uvmexp.pdbusy);
837 	(void)printf("%9u total pending pageouts\n", uvmexp.pdpending);
838 	(void)printf("%9u pages deactivated\n", uvmexp.pddeact);
839 
840 	kread(namelist, X_NCHSTATS, &nchstats, sizeof(nchstats));
841 	nchtotal = nchstats.ncs_goodhits + nchstats.ncs_neghits +
842 	    nchstats.ncs_badhits + nchstats.ncs_falsehits +
843 	    nchstats.ncs_miss + nchstats.ncs_long;
844 	(void)printf("%9lu total name lookups\n", nchtotal);
845 	(void)printf("%9lu good hits\n", nchstats.ncs_goodhits);
846 	(void)printf("%9lu negative hits\n", nchstats.ncs_neghits);
847 	(void)printf("%9lu bad hits\n", nchstats.ncs_badhits);
848 	(void)printf("%9lu false hits\n", nchstats.ncs_falsehits);
849 	(void)printf("%9lu miss\n", nchstats.ncs_miss);
850 	(void)printf("%9lu too long\n", nchstats.ncs_long);
851 	(void)printf("%9lu pass2 hits\n", nchstats.ncs_pass2);
852 	(void)printf("%9lu 2passes\n", nchstats.ncs_2passes);
853 	(void)printf(
854 	    "%9s cache hits (%d%% pos + %d%% neg) system %d%% per-process\n",
855 	    "", PCT(nchstats.ncs_goodhits, nchtotal),
856 	    PCT(nchstats.ncs_neghits, nchtotal),
857 	    PCT(nchstats.ncs_pass2, nchtotal));
858 	(void)printf("%9s deletions %d%%, falsehits %d%%, toolong %d%%\n", "",
859 	    PCT(nchstats.ncs_badhits, nchtotal),
860 	    PCT(nchstats.ncs_falsehits, nchtotal),
861 	    PCT(nchstats.ncs_long, nchtotal));
862 }
863 
864 void
865 doforkst(void)
866 {
867 
868 	kread(namelist, X_UVMEXP, &uvmexp, sizeof(uvmexp));
869 
870 	(void)printf("%u forks total\n", uvmexp.forks);
871 	(void)printf("%u forks blocked parent\n", uvmexp.forks_ppwait);
872 	(void)printf("%u forks shared address space with parent\n",
873 	    uvmexp.forks_sharevm);
874 }
875 
876 void
877 drvstats(void)
878 {
879 	int dn;
880 	double etime;
881 
882 	/* Calculate disk stat deltas. */
883 	cpuswap();
884 	drvswap();
885 	tkswap();
886 	etime = cur.cp_etime;
887 
888 	for (dn = 0; dn < ndrive; ++dn) {
889 		if (!drv_select[dn])
890 			continue;
891 		(void)printf("%2.0f ", (cur.rxfer[dn] + cur.wxfer[dn]) / etime);
892 	}
893 }
894 
895 void
896 cpustats(void)
897 {
898 	int state;
899 	double pcnt, total;
900 	double stat_us, stat_sy, stat_id;
901 
902 	total = 0;
903 	for (state = 0; state < CPUSTATES; ++state)
904 		total += cur.cp_time[state];
905 	if (total)
906 		pcnt = 100 / total;
907 	else
908 		pcnt = 0;
909 	stat_us = (cur.cp_time[CP_USER] + cur.cp_time[CP_NICE]) * pcnt;
910 	stat_sy = (cur.cp_time[CP_SYS] + cur.cp_time[CP_INTR]) * pcnt;
911 	stat_id = cur.cp_time[CP_IDLE] * pcnt;
912 	(void)printf("%*.0f ", ((stat_sy >= 100) ? 1 : 2), stat_us);
913 	(void)printf("%*.0f ", ((stat_us >= 100 || stat_id >= 100) ? 1 : 2),
914 		     stat_sy);
915 	(void)printf("%2.0f", stat_id);
916 }
917 
918 void
919 dointr(int verbose)
920 {
921 	unsigned long *intrcnt, *ointrcnt;
922 	unsigned long long inttotal, uptime;
923 	int nintr, inamlen;
924 	char *intrname, *ointrname;
925 	struct evcntlist allevents;
926 	struct evcnt evcnt, *evptr;
927 	char evgroup[EVCNT_STRING_MAX], evname[EVCNT_STRING_MAX];
928 
929 	inttotal = 0;
930 	uptime = getuptime();
931 	(void)printf("%-34s %16s %8s\n", "interrupt", "total", "rate");
932 	nintr = intrnl[X_EINTRCNT].n_value - intrnl[X_INTRCNT].n_value;
933 	inamlen = intrnl[X_EINTRNAMES].n_value - intrnl[X_INTRNAMES].n_value;
934 	if (nintr != 0 && inamlen != 0) {
935 		ointrcnt = intrcnt = malloc((size_t)nintr);
936 		ointrname = intrname = malloc((size_t)inamlen);
937 		if (intrcnt == NULL || intrname == NULL)
938 			errx(1, "%s", "");
939 		kread(intrnl, X_INTRCNT, intrcnt, (size_t)nintr);
940 		kread(intrnl, X_INTRNAMES, intrname, (size_t)inamlen);
941 		nintr /= sizeof(long);
942 		while (--nintr >= 0) {
943 			if (*intrcnt || verbose)
944 				(void)printf("%-34s %16llu %8llu\n", intrname,
945 					     (unsigned long long)*intrcnt,
946 					     (unsigned long long)
947 					     (*intrcnt / uptime));
948 			intrname += strlen(intrname) + 1;
949 			inttotal += *intrcnt++;
950 		}
951 		free(ointrcnt);
952 		free(ointrname);
953 	}
954 
955 	kread(namelist, X_ALLEVENTS, &allevents, sizeof allevents);
956 	evptr = TAILQ_FIRST(&allevents);
957 	while (evptr) {
958 		deref_kptr(evptr, &evcnt, sizeof(evcnt), "event chain trashed");
959 		evptr = TAILQ_NEXT(&evcnt, ev_list);
960 		if (evcnt.ev_type != EVCNT_TYPE_INTR)
961 			continue;
962 
963 		if (evcnt.ev_count == 0 && !verbose)
964 			continue;
965 
966 		deref_kptr(evcnt.ev_group, evgroup, evcnt.ev_grouplen + 1,
967 		    "event chain trashed");
968 		deref_kptr(evcnt.ev_name, evname, evcnt.ev_namelen + 1,
969 		    "event chain trashed");
970 
971 		(void)printf("%s %s%*s %16llu %8llu\n", evgroup, evname,
972 		    34 - (evcnt.ev_grouplen + 1 + evcnt.ev_namelen), "",
973 		    (unsigned long long)evcnt.ev_count,
974 		    (unsigned long long)(evcnt.ev_count / uptime));
975 
976 		inttotal += evcnt.ev_count++;
977 	}
978 	(void)printf("%-34s %16llu %8llu\n", "Total", inttotal,
979 	    (unsigned long long)(inttotal / uptime));
980 }
981 
982 void
983 doevcnt(int verbose)
984 {
985 	static const char * evtypes [] = { "misc", "intr", "trap" };
986 	unsigned long long uptime;
987 	struct evcntlist allevents;
988 	struct evcnt evcnt, *evptr;
989 	char evgroup[EVCNT_STRING_MAX], evname[EVCNT_STRING_MAX];
990 
991 	/* XXX should print type! */
992 
993 	uptime = getuptime();
994 	(void)printf("%-34s %16s %8s %s\n", "event", "total", "rate", "type");
995 	kread(namelist, X_ALLEVENTS, &allevents, sizeof allevents);
996 	evptr = TAILQ_FIRST(&allevents);
997 	while (evptr) {
998 		deref_kptr(evptr, &evcnt, sizeof(evcnt), "event chain trashed");
999 
1000 		evptr = TAILQ_NEXT(&evcnt, ev_list);
1001 		if (evcnt.ev_count == 0 && !verbose)
1002 			continue;
1003 
1004 		deref_kptr(evcnt.ev_group, evgroup, evcnt.ev_grouplen + 1,
1005 		    "event chain trashed");
1006 		deref_kptr(evcnt.ev_name, evname, evcnt.ev_namelen + 1,
1007 		    "event chain trashed");
1008 
1009 		(void)printf("%s %s%*s %16llu %8llu %s\n", evgroup, evname,
1010 		    34 - (evcnt.ev_grouplen + 1 + evcnt.ev_namelen), "",
1011 		    (unsigned long long)evcnt.ev_count,
1012 		    (unsigned long long)(evcnt.ev_count / uptime),
1013 		    (evcnt.ev_type < sizeof(evtypes)/sizeof(evtypes[0]) ?
1014 			evtypes[evcnt.ev_type] : "?"));
1015 	}
1016 }
1017 
1018 static char memname[64];
1019 
1020 void
1021 domem(void)
1022 {
1023 	struct kmembuckets *kp;
1024 	struct malloc_type ks, *ksp;
1025 	int i, j;
1026 	int len, size, first;
1027 	long totuse = 0, totfree = 0, totreq = 0;
1028 	struct kmembuckets buckets[MINBUCKET + 16];
1029 
1030 	kread(namelist, X_KMEMBUCKETS, buckets, sizeof(buckets));
1031 	for (first = 1, i = MINBUCKET, kp = &buckets[i]; i < MINBUCKET + 16;
1032 	    i++, kp++) {
1033 		if (kp->kb_calls == 0)
1034 			continue;
1035 		if (first) {
1036 			(void)printf("Memory statistics by bucket size\n");
1037 			(void)printf(
1038 		 "    Size   In Use   Free   Requests  HighWater  Couldfree\n");
1039 			first = 0;
1040 		}
1041 		size = 1 << i;
1042 		(void)printf("%8d %8ld %6ld %10ld %7ld %10ld\n", size,
1043 		    kp->kb_total - kp->kb_totalfree,
1044 		    kp->kb_totalfree, kp->kb_calls,
1045 		    kp->kb_highwat, kp->kb_couldfree);
1046 		totfree += size * kp->kb_totalfree;
1047 	}
1048 
1049 	/*
1050 	 * If kmem statistics are not being gathered by the kernel,
1051 	 * first will still be 1.
1052 	 */
1053 	if (first) {
1054 		warnx("Kmem statistics are not being gathered by the kernel.");
1055 		return;
1056 	}
1057 
1058 	(void)printf("\nMemory usage type by bucket size\n");
1059 	(void)printf("    Size  Type(s)\n");
1060 	kp = &buckets[MINBUCKET];
1061 	for (j =  1 << MINBUCKET; j < 1 << (MINBUCKET + 16); j <<= 1, kp++) {
1062 		if (kp->kb_calls == 0)
1063 			continue;
1064 		first = 1;
1065 		len = 8;
1066 		for (kread(namelist, X_KMEMSTAT, &ksp, sizeof(ksp));
1067 		     ksp != NULL; ksp = ks.ks_next) {
1068 			deref_kptr(ksp, &ks, sizeof(ks), "malloc type");
1069 			if (ks.ks_calls == 0)
1070 				continue;
1071 			if ((ks.ks_size & j) == 0)
1072 				continue;
1073 			deref_kptr(ks.ks_shortdesc, memname,
1074 			    sizeof(memname), "malloc type name");
1075 			len += 2 + strlen(memname);
1076 			if (first)
1077 				printf("%8d  %s", j, memname);
1078 			else
1079 				printf(",");
1080 			if (len >= 80) {
1081 				printf("\n\t ");
1082 				len = 10 + strlen(memname);
1083 			}
1084 			if (!first)
1085 				printf(" %s", memname);
1086 			first = 0;
1087 		}
1088 		putchar('\n');
1089 	}
1090 
1091 	(void)printf(
1092 	    "\nMemory statistics by type                           Type  Kern\n");
1093 	(void)printf(
1094 "         Type  InUse MemUse HighUse  Limit Requests Limit Limit Size(s)\n");
1095 	for (kread(namelist, X_KMEMSTAT, &ksp, sizeof(ksp));
1096 	     ksp != NULL; ksp = ks.ks_next) {
1097 		deref_kptr(ksp, &ks, sizeof(ks), "malloc type");
1098 		if (ks.ks_calls == 0)
1099 			continue;
1100 		deref_kptr(ks.ks_shortdesc, memname,
1101 		    sizeof(memname), "malloc type name");
1102 		(void)printf("%14s%6ld%6ldK%7ldK%6ldK%9ld%5u%6u",
1103 		    memname,
1104 		    ks.ks_inuse, (ks.ks_memuse + 1023) / 1024,
1105 		    (ks.ks_maxused + 1023) / 1024,
1106 		    (ks.ks_limit + 1023) / 1024, ks.ks_calls,
1107 		    ks.ks_limblocks, ks.ks_mapblocks);
1108 		first = 1;
1109 		for (j =  1 << MINBUCKET; j < 1 << (MINBUCKET + 16); j <<= 1) {
1110 			if ((ks.ks_size & j) == 0)
1111 				continue;
1112 			if (first)
1113 				printf("  %d", j);
1114 			else
1115 				printf(",%d", j);
1116 			first = 0;
1117 		}
1118 		printf("\n");
1119 		totuse += ks.ks_memuse;
1120 		totreq += ks.ks_calls;
1121 	}
1122 	(void)printf("\nMemory totals:  In Use    Free    Requests\n");
1123 	(void)printf("              %7ldK %6ldK    %8ld\n\n",
1124 	    (totuse + 1023) / 1024, (totfree + 1023) / 1024, totreq);
1125 }
1126 
1127 void
1128 dopool(int verbose, int wide)
1129 {
1130 	int first, ovflw;
1131 	void *addr;
1132 	long total, inuse, this_total, this_inuse;
1133 	LIST_HEAD(,pool) pool_head;
1134 	struct pool pool, *pp = &pool;
1135 	struct pool_allocator pa;
1136 	char name[32], maxp[32];
1137 
1138 	kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
1139 	addr = LIST_FIRST(&pool_head);
1140 
1141 	total = inuse = 0;
1142 
1143 	for (first = 1; addr != NULL; addr = LIST_NEXT(pp, pr_poollist) ) {
1144 		deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
1145 		deref_kptr(pp->pr_alloc, &pa, sizeof(pa),
1146 		    "pool allocator trashed");
1147 		deref_kptr(pp->pr_wchan, name, sizeof(name),
1148 		    "pool wait channel trashed");
1149 		name[sizeof(name)-1] = '\0';
1150 
1151 		if (first) {
1152 			(void)printf("Memory resource pool statistics\n");
1153 			(void)printf(
1154 			    "%-*s%*s%*s%5s%*s%s%s%*s%*s%6s%s%6s%6s%6s%5s%s%s\n",
1155 			    wide ? 16 : 11, "Name",
1156 			    wide ? 6 : 5, "Size",
1157 			    wide ? 12 : 9, "Requests",
1158 			    "Fail",
1159 			    wide ? 12 : 9, "Releases",
1160 			    wide ? " InUse" : "",
1161 			    wide ? " Avail" : "",
1162 			    wide ? 7 : 6, "Pgreq",
1163 			    wide ? 7 : 6, "Pgrel",
1164 			    "Npage",
1165 			    wide ? " PageSz" : "",
1166 			    "Hiwat",
1167 			    "Minpg",
1168 			    "Maxpg",
1169 			    "Idle",
1170 			    wide ? " Flags" : "",
1171 			    wide ? "   Util" : "");
1172 			first = 0;
1173 		}
1174 		if (pp->pr_nget == 0 && !verbose)
1175 			continue;
1176 		if (pp->pr_maxpages == UINT_MAX)
1177 			snprintf(maxp, sizeof(maxp), "inf");
1178 		else
1179 			snprintf(maxp, sizeof(maxp), "%u", pp->pr_maxpages);
1180 /*
1181  * Print single word.  `ovflow' is number of characters didn't fit
1182  * on the last word.  `fmt' is a format string to print this word.
1183  * It must contain asterisk for field width.  `width' is a width
1184  * occupied by this word.  `fixed' is a number of constant chars in
1185  * `fmt'.  `val' is a value to be printed using format string `fmt'.
1186  */
1187 #define	PRWORD(ovflw, fmt, width, fixed, val) do {	\
1188 	(ovflw) += printf((fmt),			\
1189 	    (width) - (fixed) - (ovflw) > 0 ?		\
1190 	    (width) - (fixed) - (ovflw) : 0,		\
1191 	    (val)) - (width);				\
1192 	if ((ovflw) < 0)				\
1193 		(ovflw) = 0;				\
1194 } while (/* CONSTCOND */0)
1195 		ovflw = 0;
1196 		PRWORD(ovflw, "%-*s", wide ? 16 : 11, 0, name);
1197 		PRWORD(ovflw, " %*u", wide ? 6 : 5, 1, pp->pr_size);
1198 		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nget);
1199 		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nfail);
1200 		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nput);
1201 		if (wide)
1202 			PRWORD(ovflw, " %*u", 6, 1, pp->pr_nout);
1203 		if (wide)
1204 			PRWORD(ovflw, " %*u", 6, 1, pp->pr_nitems);
1205 		PRWORD(ovflw, " %*lu", wide ? 7 : 6, 1, pp->pr_npagealloc);
1206 		PRWORD(ovflw, " %*lu", wide ? 7 : 6, 1, pp->pr_npagefree);
1207 		PRWORD(ovflw, " %*u", 6, 1, pp->pr_npages);
1208 		if (wide)
1209 			PRWORD(ovflw, " %*u", 7, 1, pa.pa_pagesz);
1210 		PRWORD(ovflw, " %*u", 6, 1, pp->pr_hiwat);
1211 		PRWORD(ovflw, " %*u", 6, 1, pp->pr_minpages);
1212 		PRWORD(ovflw, " %*s", 6, 1, maxp);
1213 		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nidle);
1214 		if (wide)
1215 			PRWORD(ovflw, " 0x%0*x", 4, 1,
1216 			    pp->pr_flags | pp->pr_roflags);
1217 
1218 		this_inuse = pp->pr_nout * pp->pr_size;
1219 		this_total = pp->pr_npages * pa.pa_pagesz;
1220 		if (pp->pr_roflags & PR_RECURSIVE) {
1221 			/*
1222 			 * Don't count in-use memory, since it's part
1223 			 * of another pool and will be accounted for
1224 			 * there.
1225 			 */
1226 			total += (this_total - this_inuse);
1227 		} else {
1228 			inuse += this_inuse;
1229 			total += this_total;
1230 		}
1231 		if (wide) {
1232 			if (this_total == 0)
1233 				printf("   ---");
1234 			else
1235 				printf(" %5.1f%%",
1236 				    (100.0 * this_inuse) / this_total);
1237 		}
1238 		printf("\n");
1239 		dopoolcache(pp, verbose);
1240 	}
1241 
1242 	inuse /= 1024;
1243 	total /= 1024;
1244 	printf("\nIn use %ldK, total allocated %ldK; utilization %.1f%%\n",
1245 	    inuse, total, (100.0 * inuse) / total);
1246 }
1247 
1248 void
1249 dopoolcache(struct pool *pp, int verbose)
1250 {
1251 	struct pool_cache pool_cache, *pc = &pool_cache;
1252 	struct pool_cache_group pool_cache_group, *pcg = &pool_cache_group;
1253 	void *addr, *pcg_addr;
1254 	int i;
1255 
1256 	if (verbose < 1)
1257 		return;
1258 
1259 #define PR_GROUPLIST							\
1260 	deref_kptr(pcg_addr, pcg, sizeof(*pcg),				\
1261 	    "pool cache group trashed");				\
1262 	printf("\t\tgroup %p: avail %d\n", pcg_addr,			\
1263 	    pcg->pcg_avail);						\
1264 	for (i = 0; i < PCG_NOBJECTS; i++) {				\
1265 		if (pcg->pcg_objects[i].pcgo_pa !=			\
1266 		    POOL_PADDR_INVALID) {				\
1267 			printf("\t\t\t%p, 0x%llx\n",			\
1268 			    pcg->pcg_objects[i].pcgo_va,		\
1269 			    (unsigned long long)			\
1270 			    pcg->pcg_objects[i].pcgo_pa);		\
1271 		} else {						\
1272 			printf("\t\t\t%p\n",				\
1273 			    pcg->pcg_objects[i].pcgo_va);		\
1274 		}							\
1275 	}
1276 
1277 	for (addr = LIST_FIRST(&pp->pr_cachelist); addr != NULL;
1278 	    addr = LIST_NEXT(pc, pc_poollist)) {
1279 		deref_kptr(addr, pc, sizeof(*pc), "pool cache trashed");
1280 		printf("\t    hits %lu misses %lu ngroups %lu nitems %lu\n",
1281 		    pc->pc_hits, pc->pc_misses, pc->pc_ngroups, pc->pc_nitems);
1282 		if (verbose < 2)
1283 			continue;
1284 		printf("\t    full groups:\n");
1285 		for (pcg_addr = LIST_FIRST(&pc->pc_fullgroups);
1286 		    pcg_addr != NULL; pcg_addr = LIST_NEXT(pcg, pcg_list)) {
1287 			PR_GROUPLIST;
1288 		}
1289 		printf("\t    partial groups:\n");
1290 		for (pcg_addr = LIST_FIRST(&pc->pc_partgroups);
1291 		    pcg_addr != NULL; pcg_addr = LIST_NEXT(pcg, pcg_list)) {
1292 			PR_GROUPLIST;
1293 		}
1294 		printf("\t    empty groups:\n");
1295 		for (pcg_addr = LIST_FIRST(&pc->pc_emptygroups);
1296 		    pcg_addr != NULL; pcg_addr = LIST_NEXT(pcg, pcg_list)) {
1297 			PR_GROUPLIST;
1298 		}
1299 	}
1300 
1301 #undef PR_GROUPLIST
1302 
1303 }
1304 
1305 enum hashtype {			/* from <sys/systm.h> */
1306 	HASH_LIST,
1307 	HASH_TAILQ
1308 };
1309 
1310 struct uidinfo {		/* XXX: no kernel header file */
1311 	LIST_ENTRY(uidinfo) ui_hash;
1312 	uid_t	ui_uid;
1313 	long	ui_proccnt;
1314 };
1315 
1316 struct kernel_hash {
1317 	const char *	description;	/* description */
1318 	int		hashsize;	/* nlist index for hash size */
1319 	int		hashtbl;	/* nlist index for hash table */
1320 	enum hashtype	type;		/* type of hash table */
1321 	size_t		offset;		/* offset of {LIST,TAILQ}_NEXT */
1322 } khashes[] =
1323 {
1324 	{
1325 		"buffer hash",
1326 		X_BUFHASH, X_BUFHASHTBL,
1327 		HASH_LIST, offsetof(struct buf, b_hash)
1328 	}, {
1329 		"inode cache (ihash)",
1330 		X_IHASH, X_IHASHTBL,
1331 		HASH_LIST, offsetof(struct inode, i_hash)
1332 	}, {
1333 		"ipv4 address -> interface hash",
1334 		X_IFADDRHASH, X_IFADDRHASHTBL,
1335 		HASH_LIST, offsetof(struct in_ifaddr, ia_hash),
1336 	}, {
1337 		"name cache hash",
1338 		X_NCHASH, X_NCHASHTBL,
1339 		HASH_LIST, offsetof(struct namecache, nc_hash),
1340 	}, {
1341 		"name cache directory hash",
1342 		X_NCVHASH, X_NCVHASHTBL,
1343 		HASH_LIST, offsetof(struct namecache, nc_vhash),
1344 	}, {
1345 		"nfs client node cache",
1346 		X_NFSNODE, X_NFSNODETBL,
1347 		HASH_LIST, offsetof(struct nfsnode, n_hash)
1348 	}, {
1349 		"user info (uid -> used processes) hash",
1350 		X_UIHASH, X_UIHASHTBL,
1351 		HASH_LIST, offsetof(struct uidinfo, ui_hash),
1352 	}, {
1353 		NULL, -1, -1, 0, 0,
1354 	}
1355 };
1356 
1357 void
1358 dohashstat(int verbose, int todo, const char *hashname)
1359 {
1360 	LIST_HEAD(, generic)	*hashtbl_list;
1361 	TAILQ_HEAD(, generic)	*hashtbl_tailq;
1362 	struct kernel_hash	*curhash;
1363 	void	*hashaddr, *hashbuf, *nhashbuf, *nextaddr;
1364 	size_t	elemsize, hashbufsize, thissize;
1365 	u_long	hashsize;
1366 	int	i, used, items, chain, maxchain;
1367 
1368 	hashbuf = NULL;
1369 	hashbufsize = 0;
1370 
1371 	if (todo & HASHLIST) {
1372 		printf("Supported hashes:\n");
1373 		for (curhash = khashes; curhash->description; curhash++) {
1374 			if (hashnl[curhash->hashsize].n_value == 0 ||
1375 			    hashnl[curhash->hashtbl].n_value == 0)
1376 				continue;
1377 			printf("\t%-16s%s\n",
1378 			    hashnl[curhash->hashsize].n_name + 1,
1379 			    curhash->description);
1380 		}
1381 		return;
1382 	}
1383 
1384 	if (hashname != NULL) {
1385 		for (curhash = khashes; curhash->description; curhash++) {
1386 			if (strcmp(hashnl[curhash->hashsize].n_name + 1,
1387 			    hashname) == 0 &&
1388 			    hashnl[curhash->hashsize].n_value != 0 &&
1389 			    hashnl[curhash->hashtbl].n_value != 0)
1390 				break;
1391 		}
1392 		if (curhash->description == NULL) {
1393 			warnx("%s: no such hash", hashname);
1394 			return;
1395 		}
1396 	}
1397 
1398 	printf(
1399 	    "%-16s %8s %8s %8s %8s %8s %8s\n"
1400 	    "%-16s %8s %8s %8s %8s %8s %8s\n",
1401 	    "", "total", "used", "util", "num", "average", "maximum",
1402 	    "hash table", "buckets", "buckets", "%", "items", "chain",
1403 	    "chain");
1404 
1405 	for (curhash = khashes; curhash->description; curhash++) {
1406 		if (hashnl[curhash->hashsize].n_value == 0 ||
1407 		    hashnl[curhash->hashtbl].n_value == 0)
1408 			continue;
1409 		if (hashname != NULL &&
1410 		    strcmp(hashnl[curhash->hashsize].n_name + 1, hashname))
1411 			continue;
1412 		elemsize = curhash->type == HASH_LIST ?
1413 		    sizeof(*hashtbl_list) : sizeof(*hashtbl_tailq);
1414 		deref_kptr((void *)hashnl[curhash->hashsize].n_value,
1415 		    &hashsize, sizeof(hashsize),
1416 		    hashnl[curhash->hashsize].n_name);
1417 		hashsize++;
1418 		deref_kptr((void *)hashnl[curhash->hashtbl].n_value,
1419 		    &hashaddr, sizeof(hashaddr),
1420 		    hashnl[curhash->hashtbl].n_name);
1421 		if (verbose)
1422 			printf("%s %lu, %s %p, offset %ld, elemsize %llu\n",
1423 			    hashnl[curhash->hashsize].n_name + 1, hashsize,
1424 			    hashnl[curhash->hashtbl].n_name + 1, hashaddr,
1425 			    (long)curhash->offset,
1426 			    (unsigned long long)elemsize);
1427 		thissize = hashsize * elemsize;
1428 		if (hashbuf == NULL || thissize > hashbufsize) {
1429 			if ((nhashbuf = realloc(hashbuf, thissize)) == NULL)
1430 				errx(1, "malloc hashbuf %llu",
1431 				    (unsigned long long)hashbufsize);
1432 			hashbuf = nhashbuf;
1433 			hashbufsize = thissize;
1434 		}
1435 		deref_kptr(hashaddr, hashbuf, thissize,
1436 		    hashnl[curhash->hashtbl].n_name);
1437 		used = 0;
1438 		items = maxchain = 0;
1439 		if (curhash->type == HASH_LIST) {
1440 			hashtbl_list = hashbuf;
1441 			hashtbl_tailq = NULL;
1442 		} else {
1443 			hashtbl_list = NULL;
1444 			hashtbl_tailq = hashbuf;
1445 		}
1446 		for (i = 0; i < hashsize; i++) {
1447 			if (curhash->type == HASH_LIST)
1448 				nextaddr = LIST_FIRST(&hashtbl_list[i]);
1449 			else
1450 				nextaddr = TAILQ_FIRST(&hashtbl_tailq[i]);
1451 			if (nextaddr == NULL)
1452 				continue;
1453 			if (verbose)
1454 				printf("%5d: %p\n", i, nextaddr);
1455 			used++;
1456 			chain = 0;
1457 			do {
1458 				chain++;
1459 				deref_kptr((char *)nextaddr + curhash->offset,
1460 				    &nextaddr, sizeof(void *),
1461 				    "hash chain corrupted");
1462 				if (verbose > 1)
1463 					printf("got nextaddr as %p\n",
1464 					    nextaddr);
1465 			} while (nextaddr != NULL);
1466 			items += chain;
1467 			if (verbose && chain > 1)
1468 				printf("\tchain = %d\n", chain);
1469 			if (chain > maxchain)
1470 				maxchain = chain;
1471 		}
1472 		printf("%-16s %8ld %8d %8.2f %8d %8.2f %8d\n",
1473 		    hashnl[curhash->hashsize].n_name + 1,
1474 		    hashsize, used, used * 100.0 / hashsize,
1475 		    items, used ? (double)items / used : 0.0, maxchain);
1476 	}
1477 }
1478 
1479 /*
1480  * kread reads something from the kernel, given its nlist index in namelist[].
1481  */
1482 void
1483 kread(struct nlist *nl, int nlx, void *addr, size_t size)
1484 {
1485 	const char *sym;
1486 
1487 	sym = nl[nlx].n_name;
1488 	if (*sym == '_')
1489 		++sym;
1490 	if (nl[nlx].n_type == 0 || nl[nlx].n_value == 0)
1491 		errx(1, "symbol %s not defined", sym);
1492 	deref_kptr((void *)nl[nlx].n_value, addr, size, sym);
1493 }
1494 
1495 /*
1496  * Dereference the kernel pointer `kptr' and fill in the local copy
1497  * pointed to by `ptr'.  The storage space must be pre-allocated,
1498  * and the size of the copy passed in `len'.
1499  */
1500 void
1501 deref_kptr(const void *kptr, void *ptr, size_t len, const char *msg)
1502 {
1503 
1504 	if (*msg == '_')
1505 		msg++;
1506 	if (kvm_read(kd, (u_long)kptr, (char *)ptr, len) != len)
1507 		errx(1, "kptr %lx: %s: %s", (u_long)kptr, msg, kvm_geterr(kd));
1508 }
1509 
1510 /*
1511  * Traverse the UVM history buffers, performing the requested action.
1512  *
1513  * Note, we assume that if we're not listing, we're dumping.
1514  */
1515 void
1516 hist_traverse(int todo, const char *histname)
1517 {
1518 	struct uvm_history_head histhead;
1519 	struct uvm_history hist, *histkva;
1520 	char *name = NULL;
1521 	size_t namelen = 0;
1522 
1523 	if (histnl[0].n_value == 0) {
1524 		warnx("UVM history is not compiled into the kernel.");
1525 		return;
1526 	}
1527 
1528 	deref_kptr((void *)histnl[X_UVM_HISTORIES].n_value, &histhead,
1529 	    sizeof(histhead), histnl[X_UVM_HISTORIES].n_name);
1530 
1531 	if (histhead.lh_first == NULL) {
1532 		warnx("No active UVM history logs.");
1533 		return;
1534 	}
1535 
1536 	if (todo & HISTLIST)
1537 		printf("Active UVM histories:");
1538 
1539 	for (histkva = LIST_FIRST(&histhead); histkva != NULL;
1540 	    histkva = LIST_NEXT(&hist, list)) {
1541 		deref_kptr(histkva, &hist, sizeof(hist), "histkva");
1542 		if (name == NULL || hist.namelen > namelen) {
1543 			if (name != NULL)
1544 				free(name);
1545 			namelen = hist.namelen;
1546 			if ((name = malloc(namelen + 1)) == NULL)
1547 				err(1, "malloc history name");
1548 		}
1549 
1550 		deref_kptr(hist.name, name, namelen, "history name");
1551 		name[namelen] = '\0';
1552 		if (todo & HISTLIST)
1553 			printf(" %s", name);
1554 		else {
1555 			/*
1556 			 * If we're dumping all histories, do it, else
1557 			 * check to see if this is the one we want.
1558 			 */
1559 			if (histname == NULL || strcmp(histname, name) == 0) {
1560 				if (histname == NULL)
1561 					printf("\nUVM history `%s':\n", name);
1562 				hist_dodump(&hist);
1563 			}
1564 		}
1565 	}
1566 
1567 	if (todo & HISTLIST)
1568 		putchar('\n');
1569 
1570 	if (name != NULL)
1571 		free(name);
1572 }
1573 
1574 /*
1575  * Actually dump the history buffer at the specified KVA.
1576  */
1577 void
1578 hist_dodump(struct uvm_history *histp)
1579 {
1580 	struct uvm_history_ent *histents, *e;
1581 	size_t histsize;
1582 	char *fmt = NULL, *fn = NULL;
1583 	size_t fmtlen = 0, fnlen = 0;
1584 	int i;
1585 
1586 	histsize = sizeof(struct uvm_history_ent) * histp->n;
1587 
1588 	if ((histents = malloc(histsize)) == NULL)
1589 		err(1, "malloc history entries");
1590 
1591 	memset(histents, 0, histsize);
1592 
1593 	deref_kptr(histp->e, histents, histsize, "history entries");
1594 	i = histp->f;
1595 	do {
1596 		e = &histents[i];
1597 		if (e->fmt != NULL) {
1598 			if (fmt == NULL || e->fmtlen > fmtlen) {
1599 				if (fmt != NULL)
1600 					free(fmt);
1601 				fmtlen = e->fmtlen;
1602 				if ((fmt = malloc(fmtlen + 1)) == NULL)
1603 					err(1, "malloc printf format");
1604 			}
1605 			if (fn == NULL || e->fnlen > fnlen) {
1606 				if (fn != NULL)
1607 					free(fn);
1608 				fnlen = e->fnlen;
1609 				if ((fn = malloc(fnlen + 1)) == NULL)
1610 					err(1, "malloc function name");
1611 			}
1612 
1613 			deref_kptr(e->fmt, fmt, fmtlen, "printf format");
1614 			fmt[fmtlen] = '\0';
1615 
1616 			deref_kptr(e->fn, fn, fnlen, "function name");
1617 			fn[fnlen] = '\0';
1618 
1619 			printf("%06ld.%06ld ", (long int)e->tv.tv_sec,
1620 			    (long int)e->tv.tv_usec);
1621 			printf("%s#%ld: ", fn, e->call);
1622 			printf(fmt, e->v[0], e->v[1], e->v[2], e->v[3]);
1623 			putchar('\n');
1624 		}
1625 		i = (i + 1) % histp->n;
1626 	} while (i != histp->f);
1627 
1628 	free(histents);
1629 	if (fmt != NULL)
1630 		free(fmt);
1631 	if (fn != NULL)
1632 		free(fn);
1633 }
1634 
1635 void
1636 usage(void)
1637 {
1638 
1639 	(void)fprintf(stderr,
1640 	    "usage: %s [-efHiLlmstUvW] [-c count] [-h hashname] [-M core] [-N system]\n"
1641 	    "\t\t[-u histname] [-w wait] [disks]\n", getprogname());
1642 	exit(1);
1643 }
1644