vmstat.c revision 1.215
1/* $NetBSD: vmstat.c,v 1.215 2017/01/05 03:42:27 pgoyette Exp $ */
2
3/*-
4 * Copyright (c) 1998, 2000, 2001, 2007 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 *
21 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
23 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
24 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 */
33
34/*
35 * Copyright (c) 1980, 1986, 1991, 1993
36 *	The Regents of the University of California.  All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 *    notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 *    notice, this list of conditions and the following disclaimer in the
45 *    documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 *    may be used to endorse or promote products derived from this software
48 *    without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 */
62
63#include <sys/cdefs.h>
64#ifndef lint
65__COPYRIGHT("@(#) Copyright (c) 1980, 1986, 1991, 1993\
66 The Regents of the University of California.  All rights reserved.");
67#endif /* not lint */
68
69#ifndef lint
70#if 0
71static char sccsid[] = "@(#)vmstat.c	8.2 (Berkeley) 3/1/95";
72#else
73__RCSID("$NetBSD: vmstat.c,v 1.215 2017/01/05 03:42:27 pgoyette Exp $");
74#endif
75#endif /* not lint */
76
77#define	__POOL_EXPOSE
78
79#include <sys/param.h>
80#include <sys/types.h>
81#include <sys/mount.h>
82#include <sys/uio.h>
83
84#include <sys/buf.h>
85#include <sys/evcnt.h>
86#include <sys/ioctl.h>
87#include <sys/malloc.h>
88#include <sys/mallocvar.h>
89#include <sys/namei.h>
90#include <sys/pool.h>
91#include <sys/proc.h>
92#include <sys/sched.h>
93#include <sys/socket.h>
94#include <sys/sysctl.h>
95#include <sys/time.h>
96#include <sys/queue.h>
97#include <sys/kernhist.h>
98
99#include <uvm/uvm_extern.h>
100#include <uvm/uvm_stat.h>
101
102#include <net/if.h>
103#include <netinet/in.h>
104#include <netinet/in_var.h>
105
106#include <ufs/ufs/inode.h>
107
108#include <nfs/rpcv2.h>
109#include <nfs/nfsproto.h>
110#include <nfs/nfsnode.h>
111
112#include <ctype.h>
113#include <err.h>
114#include <errno.h>
115#include <fcntl.h>
116#include <kvm.h>
117#include <limits.h>
118#include <nlist.h>
119#undef n_hash
120#include <paths.h>
121#include <signal.h>
122#include <stdio.h>
123#include <stddef.h>
124#include <stdlib.h>
125#include <string.h>
126#include <time.h>
127#include <unistd.h>
128#include <util.h>
129
130#include "drvstats.h"
131
132/*
133 * All this mess will go away once everything is converted.
134 */
135#ifdef __HAVE_CPU_DATA_FIRST
136
137# include <sys/cpu_data.h>
138struct cpu_info {
139	struct cpu_data ci_data;
140};
141#else
142# include <sys/cpu.h>
143#endif
144
145/*
146 * General namelist
147 */
148struct nlist namelist[] =
149{
150#define	X_BOOTTIME	0
151	{ .n_name = "_boottime" },
152#define	X_HZ		1
153	{ .n_name = "_hz" },
154#define	X_STATHZ	2
155	{ .n_name = "_stathz" },
156#define	X_NCHSTATS	3
157	{ .n_name = "_nchstats" },
158#define	X_ALLEVENTS	4
159	{ .n_name = "_allevents" },
160#define	X_POOLHEAD	5
161	{ .n_name = "_pool_head" },
162#define	X_UVMEXP	6
163	{ .n_name = "_uvmexp" },
164#define	X_TIME_SECOND	7
165	{ .n_name = "_time_second" },
166#define X_TIME		8
167	{ .n_name = "_time" },
168#define X_CPU_INFOS	9
169	{ .n_name = "_cpu_infos" },
170#define	X_NL_SIZE	10
171	{ .n_name = NULL },
172};
173
174/*
175 * Namelist for pre-evcnt interrupt counters.
176 */
177struct nlist intrnl[] =
178{
179#define	X_INTRNAMES	0
180	{ .n_name = "_intrnames" },
181#define	X_EINTRNAMES	1
182	{ .n_name = "_eintrnames" },
183#define	X_INTRCNT	2
184	{ .n_name = "_intrcnt" },
185#define	X_EINTRCNT	3
186	{ .n_name = "_eintrcnt" },
187#define	X_INTRNL_SIZE	4
188	{ .n_name = NULL },
189};
190
191
192/*
193 * Namelist for hash statistics
194 */
195struct nlist hashnl[] =
196{
197#define	X_NFSNODE	0
198	{ .n_name = "_nfsnodehash" },
199#define	X_NFSNODETBL	1
200	{ .n_name = "_nfsnodehashtbl" },
201#define	X_IHASH		2
202	{ .n_name = "_ihash" },
203#define	X_IHASHTBL	3
204	{ .n_name = "_ihashtbl" },
205#define	X_BUFHASH	4
206	{ .n_name = "_bufhash" },
207#define	X_BUFHASHTBL	5
208	{ .n_name = "_bufhashtbl" },
209#define	X_UIHASH	6
210	{ .n_name = "_uihash" },
211#define	X_UIHASHTBL	7
212	{ .n_name = "_uihashtbl" },
213#define	X_IFADDRHASH	8
214	{ .n_name = "_in_ifaddrhash" },
215#define	X_IFADDRHASHTBL	9
216	{ .n_name = "_in_ifaddrhashtbl" },
217#define	X_NCHASH	10
218	{ .n_name = "_nchash" },
219#define	X_NCHASHTBL	11
220	{ .n_name = "_nchashtbl" },
221#define	X_NCVHASH	12
222	{ .n_name = "_ncvhash" },
223#define	X_NCVHASHTBL	13
224	{ .n_name = "_ncvhashtbl" },
225#define X_HASHNL_SIZE	14	/* must be last */
226	{ .n_name = NULL },
227};
228
229/*
230 * Namelist for kernel histories
231 */
232struct nlist histnl[] =
233{
234	{ .n_name = "_kern_histories" },
235#define	X_KERN_HISTORIES		0
236	{ .n_name = NULL },
237};
238
239
240#define KILO	1024
241
242struct cpu_counter {
243	uint64_t nintr;
244	uint64_t nsyscall;
245	uint64_t nswtch;
246	uint64_t nfault;
247	uint64_t ntrap;
248	uint64_t nsoft;
249} cpucounter, ocpucounter;
250
251struct	uvmexp_sysctl uvmexp, ouvmexp;
252int	ndrives;
253
254int	winlines = 20;
255
256kvm_t *kd;
257
258
259#define	FORKSTAT	0x001
260#define	INTRSTAT	0x002
261#define	MEMSTAT		0x004
262#define	SUMSTAT		0x008
263#define	EVCNTSTAT	0x010
264#define	VMSTAT		0x020
265#define	HISTLIST	0x040
266#define	HISTDUMP	0x080
267#define	HASHSTAT	0x100
268#define	HASHLIST	0x200
269#define	VMTOTAL		0x400
270#define	POOLCACHESTAT	0x800
271
272/*
273 * Print single word.  `ovflow' is number of characters didn't fit
274 * on the last word.  `fmt' is a format string to print this word.
275 * It must contain asterisk for field width.  `width' is a width
276 * occupied by this word.  `fixed' is a number of constant chars in
277 * `fmt'.  `val' is a value to be printed using format string `fmt'.
278 */
279#define	PRWORD(ovflw, fmt, width, fixed, val) do {	\
280	(ovflw) += printf((fmt),			\
281	    (width) - (fixed) - (ovflw) > 0 ?		\
282	    (width) - (fixed) - (ovflw) : 0,		\
283	    (val)) - (width);				\
284	if ((ovflw) < 0)				\
285		(ovflw) = 0;				\
286} while (/* CONSTCOND */0)
287
288void	cpustats(int *);
289void	cpucounters(struct cpu_counter *);
290void	deref_kptr(const void *, void *, size_t, const char *);
291void	drvstats(int *);
292void	doevcnt(int verbose, int type);
293void	dohashstat(int, int, const char *);
294void	dointr(int verbose);
295void	dopool(int, int);
296void	dopoolcache(int);
297void	dosum(void);
298void	dovmstat(struct timespec *, int);
299void	print_total_hdr(void);
300void	dovmtotal(struct timespec *, int);
301void	kread(struct nlist *, int, void *, size_t);
302int	kreadc(struct nlist *, int, void *, size_t);
303void	needhdr(int);
304void	getnlist(int);
305long	getuptime(void);
306void	printhdr(void);
307long	pct(u_long, u_long);
308__dead static void	usage(void);
309void	doforkst(void);
310
311void	hist_traverse(int, const char *);
312void	hist_dodump(struct kern_history *);
313void	hist_traverse_sysctl(int, const char *);
314void	hist_dodump_sysctl(int[], unsigned int);
315
316char	**choosedrives(char **);
317
318/* Namelist and memory file names. */
319char	*nlistf, *memf;
320
321/* allow old usage [vmstat 1] */
322#define	BACKWARD_COMPATIBILITY
323
324static const int clockrate_mib[] = { CTL_KERN, KERN_CLOCKRATE };
325static const int vmmeter_mib[] = { CTL_VM, VM_METER };
326static const int uvmexp2_mib[] = { CTL_VM, VM_UVMEXP2 };
327static const int boottime_mib[] = { CTL_KERN, KERN_BOOTTIME };
328static char kvm_errbuf[_POSIX2_LINE_MAX];
329
330int
331main(int argc, char *argv[])
332{
333	int c, todo, verbose, wide;
334	struct timespec interval;
335	int reps;
336	gid_t egid = getegid();
337	const char *histname, *hashname;
338
339	histname = hashname = NULL;
340	(void)setegid(getgid());
341	memf = nlistf = NULL;
342	reps = todo = verbose = wide = 0;
343	interval.tv_sec = 0;
344	interval.tv_nsec = 0;
345	while ((c = getopt(argc, argv, "Cc:efh:HilLM:mN:stu:UvWw:")) != -1) {
346		switch (c) {
347		case 'c':
348			reps = atoi(optarg);
349			break;
350		case 'C':
351			todo |= POOLCACHESTAT;
352			break;
353		case 'e':
354			todo |= EVCNTSTAT;
355			break;
356		case 'f':
357			todo |= FORKSTAT;
358			break;
359		case 'h':
360			hashname = optarg;
361			/* FALLTHROUGH */
362		case 'H':
363			todo |= HASHSTAT;
364			break;
365		case 'i':
366			todo |= INTRSTAT;
367			break;
368		case 'l':
369			todo |= HISTLIST;
370			break;
371		case 'L':
372			todo |= HASHLIST;
373			break;
374		case 'M':
375			memf = optarg;
376			break;
377		case 'm':
378			todo |= MEMSTAT;
379			break;
380		case 'N':
381			nlistf = optarg;
382			break;
383		case 's':
384			todo |= SUMSTAT;
385			break;
386		case 't':
387			todo |= VMTOTAL;
388			break;
389		case 'u':
390			histname = optarg;
391			/* FALLTHROUGH */
392		case 'U':
393			todo |= HISTDUMP;
394			break;
395		case 'v':
396			verbose++;
397			break;
398		case 'W':
399			wide++;
400			break;
401		case 'w':
402			interval.tv_sec = atol(optarg);
403			break;
404		case '?':
405		default:
406			usage();
407		}
408	}
409	argc -= optind;
410	argv += optind;
411
412	if (todo == 0)
413		todo = VMSTAT;
414
415	/*
416	 * Discard setgid privileges.  If not the running kernel, we toss
417	 * them away totally so that bad guys can't print interesting stuff
418	 * from kernel memory, otherwise switch back to kmem for the
419	 * duration of the kvm_openfiles() call.
420	 */
421	if (nlistf != NULL || memf != NULL)
422		(void)setgid(getgid());
423	else
424		(void)setegid(egid);
425
426	kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, kvm_errbuf);
427	if (kd == NULL) {
428		if (nlistf != NULL || memf != NULL) {
429			errx(1, "kvm_openfiles: %s", kvm_errbuf);
430		}
431	}
432
433	if (nlistf == NULL && memf == NULL)
434		(void)setgid(getgid());
435
436
437	if (todo & VMSTAT) {
438		struct winsize winsize;
439
440		(void)drvinit(0);/* Initialize disk stats, no disks selected. */
441
442		(void)setgid(getgid()); /* don't need privs anymore */
443
444		argv = choosedrives(argv);	/* Select disks. */
445		winsize.ws_row = 0;
446		(void)ioctl(STDOUT_FILENO, TIOCGWINSZ, &winsize);
447		if (winsize.ws_row > 0)
448			winlines = winsize.ws_row;
449
450	}
451
452#ifdef	BACKWARD_COMPATIBILITY
453	if (*argv) {
454		interval.tv_sec = atol(*argv);
455		if (*++argv)
456			reps = atoi(*argv);
457	}
458#endif
459
460	if (interval.tv_sec) {
461		if (!reps)
462			reps = -1;
463	} else if (reps)
464		interval.tv_sec = 1;
465
466
467	getnlist(todo);
468	/*
469	 * Statistics dumping is incompatible with the default
470	 * VMSTAT/dovmstat() output. So perform the interval/reps handling
471	 * for it here.
472	 */
473	if ((todo & (VMSTAT|VMTOTAL)) == 0) {
474		for (;;) {
475			if (todo & (HISTLIST|HISTDUMP)) {
476				if ((todo & (HISTLIST|HISTDUMP)) ==
477				    (HISTLIST|HISTDUMP))
478					errx(1, "you may list or dump,"
479					    " but not both!");
480				if (memf != NULL)
481					hist_traverse(todo, histname);
482				else
483					hist_traverse_sysctl(todo, histname);
484				(void)putchar('\n');
485			}
486			if (todo & FORKSTAT) {
487				doforkst();
488				(void)putchar('\n');
489			}
490			if (todo & MEMSTAT) {
491				dopool(verbose, wide);
492				(void)putchar('\n');
493			}
494			if (todo & POOLCACHESTAT) {
495				dopoolcache(verbose);
496				(void)putchar('\n');
497			}
498			if (todo & SUMSTAT) {
499				dosum();
500				(void)putchar('\n');
501			}
502			if (todo & INTRSTAT) {
503				dointr(verbose);
504				(void)putchar('\n');
505			}
506			if (todo & EVCNTSTAT) {
507				doevcnt(verbose, EVCNT_TYPE_ANY);
508				(void)putchar('\n');
509			}
510			if (todo & (HASHLIST|HASHSTAT)) {
511				if ((todo & (HASHLIST|HASHSTAT)) ==
512				    (HASHLIST|HASHSTAT))
513					errx(1, "you may list or display,"
514					    " but not both!");
515				dohashstat(verbose, todo, hashname);
516				(void)putchar('\n');
517			}
518
519			fflush(stdout);
520			if (reps >= 0 && --reps <=0)
521				break;
522			(void)nanosleep(&interval, NULL);
523		}
524	} else {
525		if ((todo & (VMSTAT|VMTOTAL)) == (VMSTAT|VMTOTAL)) {
526			errx(1, "you may not both do vmstat and vmtotal");
527		}
528		if (todo & VMSTAT)
529			dovmstat(&interval, reps);
530		if (todo & VMTOTAL)
531			dovmtotal(&interval, reps);
532	}
533	return 0;
534}
535
536void
537getnlist(int todo)
538{
539	static int namelist_done = 0;
540	static int done = 0;
541	int c;
542	size_t i;
543
544	if (kd == NULL)
545		errx(1, "kvm_openfiles: %s", kvm_errbuf);
546
547	if (!namelist_done) {
548		namelist_done = 1;
549		if ((c = kvm_nlist(kd, namelist)) != 0) {
550			int doexit = 0;
551			if (c == -1)
552				errx(1, "kvm_nlist: %s %s",
553				    "namelist", kvm_geterr(kd));
554			for (i = 0; i < __arraycount(namelist)-1; i++)
555				if (namelist[i].n_type == 0 &&
556				    i != X_TIME_SECOND &&
557				    i != X_TIME) {
558					if (doexit++ == 0)
559						(void)fprintf(stderr,
560						    "%s: undefined symbols:",
561						    getprogname());
562					(void)fprintf(stderr, " %s",
563					    namelist[i].n_name);
564				}
565			if (doexit) {
566				(void)fputc('\n', stderr);
567				exit(1);
568			}
569		}
570	}
571	if ((todo & (SUMSTAT|INTRSTAT)) && !(done & (SUMSTAT|INTRSTAT))) {
572		done |= SUMSTAT|INTRSTAT;
573		(void) kvm_nlist(kd, intrnl);
574	}
575	if ((todo & (HASHLIST|HASHSTAT)) && !(done & (HASHLIST|HASHSTAT))) {
576		done |= HASHLIST|HASHSTAT;
577		if ((c = kvm_nlist(kd, hashnl)) == -1 || c == X_HASHNL_SIZE)
578			errx(1, "kvm_nlist: %s %s", "hashnl", kvm_geterr(kd));
579	}
580	if ((todo & (HISTLIST|HISTDUMP)) && !(done & (HISTLIST|HISTDUMP))) {
581		done |= HISTLIST|HISTDUMP;
582		if (kvm_nlist(kd, histnl) == -1)
583			errx(1, "kvm_nlist: %s %s", "histnl", kvm_geterr(kd));
584	}
585}
586
587char **
588choosedrives(char **argv)
589{
590	size_t i;
591
592	/*
593	 * Choose drives to be displayed.  Priority goes to (in order) drives
594	 * supplied as arguments, default drives.  If everything isn't filled
595	 * in and there are drives not taken care of, display the first few
596	 * that fit.
597	 */
598#define	BACKWARD_COMPATIBILITY
599	for (ndrives = 0; *argv; ++argv) {
600#ifdef	BACKWARD_COMPATIBILITY
601		if (isdigit((unsigned char)**argv))
602			break;
603#endif
604		for (i = 0; i < ndrive; i++) {
605			if (strcmp(dr_name[i], *argv))
606				continue;
607			drv_select[i] = 1;
608			++ndrives;
609			break;
610		}
611	}
612	for (i = 0; i < ndrive && ndrives < 2; i++) {
613		if (drv_select[i])
614			continue;
615		drv_select[i] = 1;
616		++ndrives;
617	}
618
619	return (argv);
620}
621
622long
623getuptime(void)
624{
625	static struct timespec boottime;
626	struct timespec now;
627	time_t uptime, nowsec;
628
629	if (memf == NULL) {
630		if (boottime.tv_sec == 0) {
631			size_t buflen = sizeof(boottime);
632			if (sysctl(boottime_mib, __arraycount(boottime_mib),
633			    &boottime, &buflen, NULL, 0) == -1)
634				warn("Can't get boottime");
635		}
636		clock_gettime(CLOCK_REALTIME, &now);
637	} else {
638		if (boottime.tv_sec == 0)
639			kread(namelist, X_BOOTTIME, &boottime,
640			    sizeof(boottime));
641		if (kreadc(namelist, X_TIME_SECOND, &nowsec, sizeof(nowsec))) {
642			/*
643			 * XXX this assignment dance can be removed once
644			 * timeval tv_sec is SUS mandated time_t
645			 */
646			now.tv_sec = nowsec;
647			now.tv_nsec = 0;
648		} else {
649			kread(namelist, X_TIME, &now, sizeof(now));
650		}
651	}
652	uptime = now.tv_sec - boottime.tv_sec;
653	if (uptime <= 0 || uptime > 60*60*24*365*10)
654		errx(1, "time makes no sense; namelist must be wrong.");
655	return (uptime);
656}
657
658int	hz, hdrcnt;
659
660void
661print_total_hdr(void)
662{
663
664	(void)printf("procs         memory\n");
665	(void)printf("ru dw pw sl");
666	(void)printf("   total-v  active-v  active-r");
667	(void)printf(" vm-sh avm-sh rm-sh arm-sh free\n");
668	hdrcnt = winlines - 2;
669}
670
671void
672dovmtotal(struct timespec *interval, int reps)
673{
674	struct vmtotal total;
675	size_t size;
676
677	(void)signal(SIGCONT, needhdr);
678
679	for (hdrcnt = 1;;) {
680		if (!--hdrcnt)
681			print_total_hdr();
682		if (memf != NULL) {
683			warnx("Unable to get vmtotals from crash dump.");
684			(void)memset(&total, 0, sizeof(total));
685		} else {
686			size = sizeof(total);
687			if (sysctl(vmmeter_mib, __arraycount(vmmeter_mib),
688			    &total, &size, NULL, 0) == -1) {
689				warn("Can't get vmtotals");
690				(void)memset(&total, 0, sizeof(total));
691			}
692		}
693		(void)printf("%2d ", total.t_rq);
694		(void)printf("%2d ", total.t_dw);
695		(void)printf("%2d ", total.t_pw);
696		(void)printf("%2d ", total.t_sl);
697
698		(void)printf("%9d ", total.t_vm);
699		(void)printf("%9d ", total.t_avm);
700		(void)printf("%9d ", total.t_arm);
701		(void)printf("%5d ", total.t_vmshr);
702		(void)printf("%6d ", total.t_avmshr);
703		(void)printf("%5d ", total.t_rmshr);
704		(void)printf("%6d ", total.t_armshr);
705		(void)printf("%5d",  total.t_free);
706
707		(void)putchar('\n');
708
709		(void)fflush(stdout);
710		if (reps >= 0 && --reps <= 0)
711			break;
712
713		(void)nanosleep(interval, NULL);
714	}
715}
716
717void
718dovmstat(struct timespec *interval, int reps)
719{
720	struct vmtotal total;
721	time_t uptime, halfuptime;
722	size_t size;
723	int pagesize = getpagesize();
724	int ovflw;
725
726	uptime = getuptime();
727	halfuptime = uptime / 2;
728	(void)signal(SIGCONT, needhdr);
729
730	if (memf != NULL) {
731		if (namelist[X_STATHZ].n_type != 0 && namelist[X_STATHZ].n_value != 0)
732			kread(namelist, X_STATHZ, &hz, sizeof(hz));
733		if (!hz)
734			kread(namelist, X_HZ, &hz, sizeof(hz));
735	} else {
736		struct clockinfo clockinfo;
737		size = sizeof(clockinfo);
738		if (sysctl(clockrate_mib, 2, &clockinfo, &size, NULL, 0) == -1)
739			err(1, "sysctl kern.clockrate failed");
740		hz = clockinfo.stathz;
741		if (!hz)
742			hz = clockinfo.hz;
743	}
744
745	for (hdrcnt = 1;;) {
746		if (!--hdrcnt)
747			printhdr();
748		/* Read new disk statistics */
749		cpureadstats();
750		drvreadstats();
751		tkreadstats();
752		if (memf != NULL) {
753			struct uvmexp uvmexp_kernel;
754			/*
755			 * XXX Can't do this if we're reading a crash
756			 * XXX dump because they're lazily-calculated.
757			 */
758			warnx("Unable to get vmtotals from crash dump.");
759			(void)memset(&total, 0, sizeof(total));
760			kread(namelist, X_UVMEXP, &uvmexp_kernel, sizeof(uvmexp_kernel));
761#define COPY(field) uvmexp.field = uvmexp_kernel.field
762			COPY(pdreact);
763			COPY(pageins);
764			COPY(pgswapout);
765			COPY(pdfreed);
766			COPY(pdscans);
767#undef COPY
768		} else {
769			size = sizeof(total);
770			if (sysctl(vmmeter_mib, __arraycount(vmmeter_mib),
771			    &total, &size, NULL, 0) == -1) {
772				warn("Can't get vmtotals");
773				(void)memset(&total, 0, sizeof(total));
774			}
775			size = sizeof(uvmexp);
776			if (sysctl(uvmexp2_mib, __arraycount(uvmexp2_mib), &uvmexp,
777			    &size, NULL, 0) == -1)
778				warn("sysctl vm.uvmexp2 failed");
779		}
780		cpucounters(&cpucounter);
781		ovflw = 0;
782		PRWORD(ovflw, " %*d", 2, 1, total.t_rq - 1);
783		PRWORD(ovflw, " %*d", 2, 1, total.t_dw + total.t_pw);
784#define	pgtok(a) (long)((a) * ((uint32_t)pagesize >> 10))
785#define	rate(x)	(u_long)(((x) + halfuptime) / uptime)	/* round */
786		PRWORD(ovflw, " %*ld", 9, 1, pgtok(total.t_avm));
787		PRWORD(ovflw, " %*ld", 7, 1, pgtok(total.t_free));
788		PRWORD(ovflw, " %*ld", 5, 1,
789		    rate(cpucounter.nfault - ocpucounter.nfault));
790		PRWORD(ovflw, " %*ld", 4, 1,
791		    rate(uvmexp.pdreact - ouvmexp.pdreact));
792		PRWORD(ovflw, " %*ld", 4, 1,
793		    rate(uvmexp.pageins - ouvmexp.pageins));
794		PRWORD(ovflw, " %*ld", 5, 1,
795		    rate(uvmexp.pgswapout - ouvmexp.pgswapout));
796		PRWORD(ovflw, " %*ld", 5, 1,
797		    rate(uvmexp.pdfreed - ouvmexp.pdfreed));
798		PRWORD(ovflw, " %*ld", 6, 2,
799		    rate(uvmexp.pdscans - ouvmexp.pdscans));
800		drvstats(&ovflw);
801		PRWORD(ovflw, " %*ld", 5, 1,
802		    rate(cpucounter.nintr - ocpucounter.nintr));
803		PRWORD(ovflw, " %*ld", 5, 1,
804		    rate(cpucounter.nsyscall - ocpucounter.nsyscall));
805		PRWORD(ovflw, " %*ld", 4, 1,
806		    rate(cpucounter.nswtch - ocpucounter.nswtch));
807		cpustats(&ovflw);
808		(void)putchar('\n');
809		(void)fflush(stdout);
810		if (reps >= 0 && --reps <= 0)
811			break;
812		ouvmexp = uvmexp;
813		ocpucounter = cpucounter;
814		uptime = interval->tv_sec;
815		/*
816		 * We round upward to avoid losing low-frequency events
817		 * (i.e., >= 1 per interval but < 1 per second).
818		 */
819		halfuptime = uptime == 1 ? 0 : (uptime + 1) / 2;
820		(void)nanosleep(interval, NULL);
821	}
822}
823
824void
825printhdr(void)
826{
827	size_t i;
828
829	(void)printf(" procs    memory      page%*s", 23, "");
830	if (ndrives > 0)
831		(void)printf("%s %*sfaults      cpu\n",
832		    ((ndrives > 1) ? "disks" : "disk"),
833		    ((ndrives > 1) ? ndrives * 3 - 4 : 0), "");
834	else
835		(void)printf("%*s  faults   cpu\n",
836		    ndrives * 3, "");
837
838	(void)printf(" r b      avm    fre  flt  re  pi   po   fr   sr ");
839	for (i = 0; i < ndrive; i++)
840		if (drv_select[i])
841			(void)printf("%c%c ", dr_name[i][0],
842			    dr_name[i][strlen(dr_name[i]) - 1]);
843	(void)printf("  in   sy  cs us sy id\n");
844	hdrcnt = winlines - 2;
845}
846
847/*
848 * Force a header to be prepended to the next output.
849 */
850void
851/*ARGSUSED*/
852needhdr(int dummy)
853{
854
855	hdrcnt = 1;
856}
857
858long
859pct(u_long top, u_long bot)
860{
861	long ans;
862
863	if (bot == 0)
864		return (0);
865	ans = (long)((quad_t)top * 100 / bot);
866	return (ans);
867}
868
869#define	PCT(top, bot) (int)pct((u_long)(top), (u_long)(bot))
870
871void
872dosum(void)
873{
874	struct nchstats nch_stats;
875	uint64_t nchtotal;
876	size_t ssize;
877	int active_kernel;
878	struct cpu_counter cc;
879
880	/*
881	 * The "active" and "inactive" variables
882	 * are now estimated by the kernel and sadly
883	 * can not easily be dug out of a crash dump.
884	 */
885	ssize = sizeof(uvmexp);
886	memset(&uvmexp, 0, ssize);
887	active_kernel = (memf == NULL);
888	if (active_kernel) {
889		/* only on active kernel */
890		if (sysctl(uvmexp2_mib, __arraycount(uvmexp2_mib), &uvmexp,
891		    &ssize, NULL, 0) == -1)
892			warn("sysctl vm.uvmexp2 failed");
893	} else {
894		struct uvmexp uvmexp_kernel;
895		kread(namelist, X_UVMEXP, &uvmexp_kernel, sizeof(uvmexp_kernel));
896#define COPY(field) uvmexp.field = uvmexp_kernel.field
897		COPY(pagesize);
898		COPY(ncolors);
899		COPY(npages);
900		COPY(free);
901		COPY(paging);
902		COPY(wired);
903		COPY(zeropages);
904		COPY(reserve_pagedaemon);
905		COPY(reserve_kernel);
906		COPY(anonpages);
907		COPY(filepages);
908		COPY(execpages);
909		COPY(freemin);
910		COPY(freetarg);
911		COPY(wiredmax);
912		COPY(nswapdev);
913		COPY(swpages);
914		COPY(swpginuse);
915		COPY(nswget);
916		COPY(pageins);
917		COPY(pdpageouts);
918		COPY(pgswapin);
919		COPY(pgswapout);
920		COPY(forks);
921		COPY(forks_ppwait);
922		COPY(forks_sharevm);
923		COPY(pga_zerohit);
924		COPY(pga_zeromiss);
925		COPY(zeroaborts);
926		COPY(colorhit);
927		COPY(colormiss);
928		COPY(cpuhit);
929		COPY(cpumiss);
930		COPY(fltnoram);
931		COPY(fltnoanon);
932		COPY(fltpgwait);
933		COPY(fltpgrele);
934		COPY(fltrelck);
935		COPY(fltrelckok);
936		COPY(fltanget);
937		COPY(fltanretry);
938		COPY(fltamcopy);
939		COPY(fltamcopy);
940		COPY(fltnomap);
941		COPY(fltlget);
942		COPY(fltget);
943		COPY(flt_anon);
944		COPY(flt_acow);
945		COPY(flt_obj);
946		COPY(flt_prcopy);
947		COPY(flt_przero);
948		COPY(pdwoke);
949		COPY(pdrevs);
950		COPY(pdfreed);
951		COPY(pdscans);
952		COPY(pdanscan);
953		COPY(pdobscan);
954		COPY(pdreact);
955		COPY(pdbusy);
956		COPY(pdpending);
957		COPY(pddeact);
958#undef COPY
959	}
960
961
962	(void)printf("%9" PRIu64 " bytes per page\n", uvmexp.pagesize);
963
964	(void)printf("%9" PRIu64 " page color%s\n",
965	    uvmexp.ncolors, uvmexp.ncolors == 1 ? "" : "s");
966
967	(void)printf("%9" PRIu64 " pages managed\n", uvmexp.npages);
968	(void)printf("%9" PRIu64 " pages free\n", uvmexp.free);
969	if (active_kernel) {
970		(void)printf("%9" PRIu64 " pages active\n", uvmexp.active);
971		(void)printf("%9" PRIu64 " pages inactive\n", uvmexp.inactive);
972	}
973	(void)printf("%9" PRIu64 " pages paging\n", uvmexp.paging);
974	(void)printf("%9" PRIu64 " pages wired\n", uvmexp.wired);
975	(void)printf("%9" PRIu64 " zero pages\n", uvmexp.zeropages);
976	(void)printf("%9" PRIu64 " reserve pagedaemon pages\n",
977	    uvmexp.reserve_pagedaemon);
978	(void)printf("%9" PRIu64 " reserve kernel pages\n", uvmexp.reserve_kernel);
979	(void)printf("%9" PRIu64 " anonymous pages\n", uvmexp.anonpages);
980	(void)printf("%9" PRIu64 " cached file pages\n", uvmexp.filepages);
981	(void)printf("%9" PRIu64 " cached executable pages\n", uvmexp.execpages);
982
983	(void)printf("%9" PRIu64 " minimum free pages\n", uvmexp.freemin);
984	(void)printf("%9" PRIu64 " target free pages\n", uvmexp.freetarg);
985	(void)printf("%9" PRIu64 " maximum wired pages\n", uvmexp.wiredmax);
986
987	(void)printf("%9" PRIu64 " swap devices\n", uvmexp.nswapdev);
988	(void)printf("%9" PRIu64 " swap pages\n", uvmexp.swpages);
989	(void)printf("%9" PRIu64 " swap pages in use\n", uvmexp.swpginuse);
990	(void)printf("%9" PRIu64 " swap allocations\n", uvmexp.nswget);
991
992	cpucounters(&cc);
993
994	(void)printf("%9" PRIu64 " total faults taken\n", cc.nfault);
995	(void)printf("%9" PRIu64 " traps\n", cc.ntrap);
996	(void)printf("%9" PRIu64 " device interrupts\n", cc.nintr);
997	(void)printf("%9" PRIu64 " CPU context switches\n", cc.nswtch);
998	(void)printf("%9" PRIu64 " software interrupts\n", cc.nsoft);
999	(void)printf("%9" PRIu64 " system calls\n", cc.nsyscall);
1000	(void)printf("%9" PRIu64 " pagein requests\n", uvmexp.pageins);
1001	(void)printf("%9" PRIu64 " pageout requests\n", uvmexp.pdpageouts);
1002	(void)printf("%9" PRIu64 " pages swapped in\n", uvmexp.pgswapin);
1003	(void)printf("%9" PRIu64 " pages swapped out\n", uvmexp.pgswapout);
1004	(void)printf("%9" PRIu64 " forks total\n", uvmexp.forks);
1005	(void)printf("%9" PRIu64 " forks blocked parent\n", uvmexp.forks_ppwait);
1006	(void)printf("%9" PRIu64 " forks shared address space with parent\n",
1007	    uvmexp.forks_sharevm);
1008	(void)printf("%9" PRIu64 " pagealloc zero wanted and avail\n",
1009	    uvmexp.pga_zerohit);
1010	(void)printf("%9" PRIu64 " pagealloc zero wanted and not avail\n",
1011	    uvmexp.pga_zeromiss);
1012	(void)printf("%9" PRIu64 " aborts of idle page zeroing\n",
1013	    uvmexp.zeroaborts);
1014	(void)printf("%9" PRIu64 " pagealloc desired color avail\n",
1015	    uvmexp.colorhit);
1016	(void)printf("%9" PRIu64 " pagealloc desired color not avail\n",
1017	    uvmexp.colormiss);
1018	(void)printf("%9" PRIu64 " pagealloc local cpu avail\n",
1019	    uvmexp.cpuhit);
1020	(void)printf("%9" PRIu64 " pagealloc local cpu not avail\n",
1021	    uvmexp.cpumiss);
1022
1023	(void)printf("%9" PRIu64 " faults with no memory\n", uvmexp.fltnoram);
1024	(void)printf("%9" PRIu64 " faults with no anons\n", uvmexp.fltnoanon);
1025	(void)printf("%9" PRIu64 " faults had to wait on pages\n", uvmexp.fltpgwait);
1026	(void)printf("%9" PRIu64 " faults found released page\n", uvmexp.fltpgrele);
1027	(void)printf("%9" PRIu64 " faults relock (%" PRIu64 " ok)\n", uvmexp.fltrelck,
1028	    uvmexp.fltrelckok);
1029	(void)printf("%9" PRIu64 " anon page faults\n", uvmexp.fltanget);
1030	(void)printf("%9" PRIu64 " anon retry faults\n", uvmexp.fltanretry);
1031	(void)printf("%9" PRIu64 " amap copy faults\n", uvmexp.fltamcopy);
1032	(void)printf("%9" PRIu64 " neighbour anon page faults\n", uvmexp.fltnamap);
1033	(void)printf("%9" PRIu64 " neighbour object page faults\n", uvmexp.fltnomap);
1034	(void)printf("%9" PRIu64 " locked pager get faults\n", uvmexp.fltlget);
1035	(void)printf("%9" PRIu64 " unlocked pager get faults\n", uvmexp.fltget);
1036	(void)printf("%9" PRIu64 " anon faults\n", uvmexp.flt_anon);
1037	(void)printf("%9" PRIu64 " anon copy on write faults\n", uvmexp.flt_acow);
1038	(void)printf("%9" PRIu64 " object faults\n", uvmexp.flt_obj);
1039	(void)printf("%9" PRIu64 " promote copy faults\n", uvmexp.flt_prcopy);
1040	(void)printf("%9" PRIu64 " promote zero fill faults\n", uvmexp.flt_przero);
1041
1042	(void)printf("%9" PRIu64 " times daemon wokeup\n",uvmexp.pdwoke);
1043	(void)printf("%9" PRIu64 " revolutions of the clock hand\n", uvmexp.pdrevs);
1044	(void)printf("%9" PRIu64 " pages freed by daemon\n", uvmexp.pdfreed);
1045	(void)printf("%9" PRIu64 " pages scanned by daemon\n", uvmexp.pdscans);
1046	(void)printf("%9" PRIu64 " anonymous pages scanned by daemon\n",
1047	    uvmexp.pdanscan);
1048	(void)printf("%9" PRIu64 " object pages scanned by daemon\n", uvmexp.pdobscan);
1049	(void)printf("%9" PRIu64 " pages reactivated\n", uvmexp.pdreact);
1050	(void)printf("%9" PRIu64 " pages found busy by daemon\n", uvmexp.pdbusy);
1051	(void)printf("%9" PRIu64 " total pending pageouts\n", uvmexp.pdpending);
1052	(void)printf("%9" PRIu64 " pages deactivated\n", uvmexp.pddeact);
1053
1054	if (active_kernel) {
1055		ssize = sizeof(nch_stats);
1056		if (sysctlbyname("vfs.namecache_stats", &nch_stats, &ssize,
1057		    NULL, 0)) {
1058			warn("vfs.namecache_stats failed");
1059			memset(&nch_stats, 0, sizeof(nch_stats));
1060		}
1061	} else {
1062		kread(namelist, X_NCHSTATS, &nch_stats, sizeof(nch_stats));
1063	}
1064
1065	nchtotal = nch_stats.ncs_goodhits + nch_stats.ncs_neghits +
1066	    nch_stats.ncs_badhits + nch_stats.ncs_falsehits +
1067	    nch_stats.ncs_miss + nch_stats.ncs_long;
1068	(void)printf("%9" PRIu64 " total name lookups\n", nchtotal);
1069	(void)printf("%9" PRIu64 " good hits\n", nch_stats.ncs_goodhits);
1070	(void)printf("%9" PRIu64 " negative hits\n", nch_stats.ncs_neghits);
1071	(void)printf("%9" PRIu64 " bad hits\n", nch_stats.ncs_badhits);
1072	(void)printf("%9" PRIu64 " false hits\n", nch_stats.ncs_falsehits);
1073	(void)printf("%9" PRIu64 " miss\n", nch_stats.ncs_miss);
1074	(void)printf("%9" PRIu64 " too long\n", nch_stats.ncs_long);
1075	(void)printf("%9" PRIu64 " pass2 hits\n", nch_stats.ncs_pass2);
1076	(void)printf("%9" PRIu64 " 2passes\n", nch_stats.ncs_2passes);
1077	(void)printf(
1078	    "%9s cache hits (%d%% pos + %d%% neg) system %d%% per-process\n",
1079	    "", PCT(nch_stats.ncs_goodhits, nchtotal),
1080	    PCT(nch_stats.ncs_neghits, nchtotal),
1081	    PCT(nch_stats.ncs_pass2, nchtotal));
1082	(void)printf("%9s deletions %d%%, falsehits %d%%, toolong %d%%\n", "",
1083	    PCT(nch_stats.ncs_badhits, nchtotal),
1084	    PCT(nch_stats.ncs_falsehits, nchtotal),
1085	    PCT(nch_stats.ncs_long, nchtotal));
1086}
1087
1088void
1089doforkst(void)
1090{
1091	if (memf != NULL) {
1092		struct uvmexp uvmexp_kernel;
1093		kread(namelist, X_UVMEXP, &uvmexp_kernel, sizeof(uvmexp_kernel));
1094#define COPY(field) uvmexp.field = uvmexp_kernel.field
1095		COPY(forks);
1096		COPY(forks_ppwait);
1097		COPY(forks_sharevm);
1098#undef COPY
1099	} else {
1100		size_t size = sizeof(uvmexp);
1101		if (sysctl(uvmexp2_mib, __arraycount(uvmexp2_mib), &uvmexp,
1102		    &size, NULL, 0) == -1)
1103			warn("sysctl vm.uvmexp2 failed");
1104	}
1105
1106	(void)printf("%" PRIu64 " forks total\n", uvmexp.forks);
1107	(void)printf("%" PRIu64 " forks blocked parent\n", uvmexp.forks_ppwait);
1108	(void)printf("%" PRIu64 " forks shared address space with parent\n",
1109	    uvmexp.forks_sharevm);
1110}
1111
1112void
1113drvstats(int *ovflwp)
1114{
1115	size_t dn;
1116	double etime;
1117	int ovflw = *ovflwp;
1118
1119	/* Calculate disk stat deltas. */
1120	cpuswap();
1121	drvswap();
1122	tkswap();
1123	etime = cur.cp_etime;
1124
1125	for (dn = 0; dn < ndrive; ++dn) {
1126		if (!drv_select[dn])
1127	 		continue;
1128		PRWORD(ovflw, " %*.0f", 3, 1,
1129		    (cur.rxfer[dn] + cur.wxfer[dn]) / etime);
1130	}
1131	*ovflwp = ovflw;
1132}
1133
1134void
1135cpucounters(struct cpu_counter *cc)
1136{
1137	static struct cpu_info **cpu_infos;
1138	static int initialised;
1139	struct cpu_info **slot;
1140
1141	if (memf == NULL) {
1142		cc->nintr = uvmexp.intrs;
1143		cc->nsyscall = uvmexp.syscalls;
1144		cc->nswtch = uvmexp.swtch;
1145		cc->nfault = uvmexp.faults;
1146		cc->ntrap = uvmexp.traps;
1147		cc->nsoft = uvmexp.softs;
1148		return;
1149	}
1150
1151	if (!initialised) {
1152		kread(namelist, X_CPU_INFOS, &cpu_infos, sizeof(cpu_infos));
1153		initialised = 1;
1154	}
1155
1156	slot = cpu_infos;
1157
1158	memset(cc, 0, sizeof(*cc));
1159
1160	for (;;) {
1161		struct cpu_info tci, *ci = NULL;
1162
1163		deref_kptr(slot++, &ci, sizeof(ci), "CPU array trashed");
1164		if (!ci) {
1165			break;
1166		}
1167
1168		if ((size_t)kvm_read(kd, (u_long)ci, &tci, sizeof(tci))
1169		    != sizeof(tci)) {
1170			warnx("Can't read cpu info from %p (%s)",
1171			    ci, kvm_geterr(kd));
1172			memset(cc, 0, sizeof(*cc));
1173			return;
1174		}
1175		cc->nintr += tci.ci_data.cpu_nintr;
1176		cc->nsyscall += tci.ci_data.cpu_nsyscall;
1177		cc->nswtch = tci.ci_data.cpu_nswtch;
1178		cc->nfault = tci.ci_data.cpu_nfault;
1179		cc->ntrap = tci.ci_data.cpu_ntrap;
1180		cc->nsoft = tci.ci_data.cpu_nsoft;
1181	}
1182}
1183
1184void
1185cpustats(int *ovflwp)
1186{
1187	int state;
1188	double pcnt, total;
1189	double stat_us, stat_sy, stat_id;
1190	int ovflw = *ovflwp;
1191
1192	total = 0;
1193	for (state = 0; state < CPUSTATES; ++state)
1194		total += cur.cp_time[state];
1195	if (total)
1196		pcnt = 100 / total;
1197	else
1198		pcnt = 0;
1199	stat_us = (cur.cp_time[CP_USER] + cur.cp_time[CP_NICE]) * pcnt;
1200	stat_sy = (cur.cp_time[CP_SYS] + cur.cp_time[CP_INTR]) * pcnt;
1201	stat_id = cur.cp_time[CP_IDLE] * pcnt;
1202	PRWORD(ovflw, " %*.0f", ((stat_sy >= 100) ? 2 : 3), 1, stat_us);
1203	PRWORD(ovflw, " %*.0f", ((stat_us >= 100 || stat_id >= 100) ? 2 : 3), 1,
1204	    stat_sy);
1205	PRWORD(ovflw, " %*.0f", 3, 1, stat_id);
1206	*ovflwp = ovflw;
1207}
1208
1209void
1210dointr(int verbose)
1211{
1212	unsigned long *intrcnt, *ointrcnt;
1213	unsigned long long inttotal, uptime;
1214	int nintr, inamlen;
1215	char *intrname, *ointrname;
1216
1217	inttotal = 0;
1218	uptime = getuptime();
1219	(void)printf("%-34s %16s %8s\n", "interrupt", "total", "rate");
1220	nintr = intrnl[X_EINTRCNT].n_value - intrnl[X_INTRCNT].n_value;
1221	inamlen = intrnl[X_EINTRNAMES].n_value - intrnl[X_INTRNAMES].n_value;
1222	if (nintr != 0 && inamlen != 0) {
1223		ointrcnt = intrcnt = malloc((size_t)nintr);
1224		ointrname = intrname = malloc((size_t)inamlen);
1225		if (intrcnt == NULL || intrname == NULL)
1226			errx(1, "%s", "");
1227		kread(intrnl, X_INTRCNT, intrcnt, (size_t)nintr);
1228		kread(intrnl, X_INTRNAMES, intrname, (size_t)inamlen);
1229		nintr /= sizeof(long);
1230		while (--nintr >= 0) {
1231			if (*intrcnt || verbose)
1232				(void)printf("%-34s %16llu %8llu\n", intrname,
1233					     (unsigned long long)*intrcnt,
1234					     (unsigned long long)
1235					     (*intrcnt / uptime));
1236			intrname += strlen(intrname) + 1;
1237			inttotal += *intrcnt++;
1238		}
1239		free(ointrcnt);
1240		free(ointrname);
1241	}
1242
1243	doevcnt(verbose, EVCNT_TYPE_INTR);
1244}
1245
1246void
1247doevcnt(int verbose, int type)
1248{
1249	static const char * const evtypes [] = { "misc", "intr", "trap" };
1250	uint64_t counttotal, uptime;
1251	struct evcntlist allevents;
1252	struct evcnt evcnt, *evptr;
1253	char evgroup[EVCNT_STRING_MAX], evname[EVCNT_STRING_MAX];
1254
1255	counttotal = 0;
1256	uptime = getuptime();
1257	if (type == EVCNT_TYPE_ANY)
1258		(void)printf("%-34s %16s %8s %s\n", "event", "total", "rate",
1259		    "type");
1260
1261	if (memf == NULL) do {
1262		const int mib[4] = { CTL_KERN, KERN_EVCNT, type,
1263		    verbose ? KERN_EVCNT_COUNT_ANY : KERN_EVCNT_COUNT_NONZERO };
1264		size_t buflen = 0;
1265		void *buf = NULL;
1266		const struct evcnt_sysctl *evs, *last_evs;
1267		for (;;) {
1268			size_t newlen;
1269			int error;
1270			if (buflen)
1271				buf = malloc(buflen);
1272			error = sysctl(mib, __arraycount(mib),
1273			    buf, &newlen, NULL, 0);
1274			if (error) {
1275				err(1, "kern.evcnt");
1276				if (buf)
1277					free(buf);
1278				return;
1279			}
1280			if (newlen <= buflen) {
1281				buflen = newlen;
1282				break;
1283			}
1284			if (buf)
1285				free(buf);
1286			buflen = newlen;
1287		}
1288		evs = buf;
1289		last_evs = (void *)((char *)buf + buflen);
1290		buflen /= sizeof(uint64_t);
1291		while (evs < last_evs
1292		    && buflen >= sizeof(*evs)/sizeof(uint64_t)
1293		    && buflen >= evs->ev_len) {
1294			(void)printf(type == EVCNT_TYPE_ANY ?
1295			    "%s %s%*s %16"PRIu64" %8"PRIu64" %s\n" :
1296			    "%s %s%*s %16"PRIu64" %8"PRIu64"\n",
1297			    evs->ev_strings,
1298			    evs->ev_strings + evs->ev_grouplen + 1,
1299			    34 - (evs->ev_grouplen + 1 + evs->ev_namelen), "",
1300			    evs->ev_count,
1301			    evs->ev_count / uptime,
1302			    (evs->ev_type < __arraycount(evtypes) ?
1303				evtypes[evs->ev_type] : "?"));
1304			buflen -= evs->ev_len;
1305			counttotal += evs->ev_count;
1306			evs = (const void *)((const uint64_t *)evs + evs->ev_len);
1307		}
1308		free(buf);
1309		if (type != EVCNT_TYPE_ANY)
1310			(void)printf("%-34s %16"PRIu64" %8"PRIu64"\n",
1311			    "Total", counttotal, counttotal / uptime);
1312		return;
1313	} while (/*CONSTCOND*/ 0);
1314
1315	kread(namelist, X_ALLEVENTS, &allevents, sizeof allevents);
1316	evptr = TAILQ_FIRST(&allevents);
1317	while (evptr) {
1318		deref_kptr(evptr, &evcnt, sizeof(evcnt), "event chain trashed");
1319
1320		evptr = TAILQ_NEXT(&evcnt, ev_list);
1321		if (evcnt.ev_count == 0 && !verbose)
1322			continue;
1323		if (type != EVCNT_TYPE_ANY && evcnt.ev_type != type)
1324			continue;
1325
1326		deref_kptr(evcnt.ev_group, evgroup,
1327		    (size_t)evcnt.ev_grouplen + 1, "event chain trashed");
1328		deref_kptr(evcnt.ev_name, evname,
1329		    (size_t)evcnt.ev_namelen + 1, "event chain trashed");
1330
1331		(void)printf(type == EVCNT_TYPE_ANY ?
1332		    "%s %s%*s %16"PRIu64" %8"PRIu64" %s\n" :
1333		    "%s %s%*s %16"PRIu64" %8"PRIu64"\n",
1334		    evgroup, evname,
1335		    34 - (evcnt.ev_grouplen + 1 + evcnt.ev_namelen), "",
1336		    evcnt.ev_count,
1337		    (evcnt.ev_count / uptime),
1338		    (evcnt.ev_type < __arraycount(evtypes) ?
1339			evtypes[evcnt.ev_type] : "?"));
1340
1341		counttotal += evcnt.ev_count;
1342	}
1343	if (type != EVCNT_TYPE_ANY)
1344		(void)printf("%-34s %16"PRIu64" %8"PRIu64"\n",
1345		    "Total", counttotal, counttotal / uptime);
1346}
1347
1348static void
1349dopool_sysctl(int verbose, int wide)
1350{
1351	uint64_t total, inuse, this_total, this_inuse;
1352	struct {
1353		uint64_t pt_nget;
1354		uint64_t pt_nfail;
1355		uint64_t pt_nput;
1356		uint64_t pt_nout;
1357		uint64_t pt_nitems;
1358		uint64_t pt_npagealloc;
1359		uint64_t pt_npagefree;
1360		uint64_t pt_npages;
1361	} pool_totals;
1362	size_t i, len;
1363	int name_len, ovflw;
1364	struct pool_sysctl *pp, *data;
1365	char in_use[8], avail[8], maxp[32];
1366
1367	data = asysctlbyname("kern.pool", &len);
1368	if (data == NULL)
1369		err(1, "failed to reead kern.pool");
1370
1371	memset(&pool_totals, 0, sizeof pool_totals);
1372	total = inuse = 0;
1373	len /= sizeof(*data);
1374
1375	(void)printf("Memory resource pool statistics\n");
1376	(void)printf(
1377	    "%-*s%*s%*s%5s%*s%s%s%*s%*s%6s%s%6s%6s%6s%5s%s%s\n",
1378	    wide ? 16 : 11, "Name",
1379	    wide ? 6 : 5, "Size",
1380	    wide ? 12 : 9, "Requests",
1381	    "Fail",
1382	    wide ? 12 : 9, "Releases",
1383	    wide ? "  InUse" : "",
1384	    wide ? " Avail" : "",
1385	    wide ? 7 : 6, "Pgreq",
1386	    wide ? 7 : 6, "Pgrel",
1387	    "Npage",
1388	    wide ? " PageSz" : "",
1389	    "Hiwat",
1390	    "Minpg",
1391	    "Maxpg",
1392	    "Idle",
1393	    wide ? " Flags" : "",
1394	    wide ? "   Util" : "");
1395
1396	name_len = MIN((int)sizeof(pp->pr_wchan), wide ? 16 : 11);
1397	for (i = 0; i < len; ++i) {
1398		pp = &data[i];
1399		if (pp->pr_nget == 0 && !verbose)
1400			continue;
1401		if (pp->pr_maxpages == UINT_MAX)
1402			(void)snprintf(maxp, sizeof(maxp), "inf");
1403		else
1404			(void)snprintf(maxp, sizeof(maxp), "%" PRIu64,
1405			    pp->pr_maxpages);
1406		ovflw = 0;
1407		PRWORD(ovflw, "%-*s", name_len, 0, pp->pr_wchan);
1408		PRWORD(ovflw, " %*" PRIu64, wide ? 6 : 5, 1, pp->pr_size);
1409		PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pp->pr_nget);
1410		pool_totals.pt_nget += pp->pr_nget;
1411		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_nfail);
1412		pool_totals.pt_nfail += pp->pr_nfail;
1413		PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pp->pr_nput);
1414		pool_totals.pt_nput += pp->pr_nput;
1415		if (wide) {
1416			PRWORD(ovflw, " %*" PRIu64, 7, 1, pp->pr_nout);
1417			pool_totals.pt_nout += pp->pr_nout;
1418		}
1419		if (wide) {
1420			PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_nitems);
1421			pool_totals.pt_nitems += pp->pr_nitems;
1422		}
1423		PRWORD(ovflw, " %*" PRIu64, wide ? 7 : 6, 1, pp->pr_npagealloc);
1424		pool_totals.pt_npagealloc += pp->pr_npagealloc;
1425		PRWORD(ovflw, " %*" PRIu64, wide ? 7 : 6, 1, pp->pr_npagefree);
1426		pool_totals.pt_npagefree += pp->pr_npagefree;
1427		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_npages);
1428		pool_totals.pt_npages += pp->pr_npages;
1429		if (wide)
1430			PRWORD(ovflw, " %*" PRIu64, 7, 1, pp->pr_pagesize);
1431		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_hiwat);
1432		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_minpages);
1433		PRWORD(ovflw, " %*s", 6, 1, maxp);
1434		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_nidle);
1435		if (wide)
1436			PRWORD(ovflw, " 0x%0*" PRIx64, 4, 1,
1437			    pp->pr_flags);
1438
1439		this_inuse = pp->pr_nout * pp->pr_size;
1440		this_total = pp->pr_npages * pp->pr_pagesize;
1441		if (pp->pr_flags & PR_RECURSIVE) {
1442			/*
1443			 * Don't count in-use memory, since it's part
1444			 * of another pool and will be accounted for
1445			 * there.
1446			 */
1447			total += (this_total - this_inuse);
1448		} else {
1449			inuse += this_inuse;
1450			total += this_total;
1451		}
1452		if (wide) {
1453			if (this_total == 0)
1454				(void)printf("   ---");
1455			else
1456				(void)printf(" %5.1f%%",
1457				    (100.0 * this_inuse) / this_total);
1458		}
1459		(void)printf("\n");
1460	}
1461	if (wide) {
1462		snprintf(in_use, sizeof in_use, "%7"PRId64, pool_totals.pt_nout);
1463		snprintf(avail, sizeof avail, "%6"PRId64, pool_totals.pt_nitems);
1464	} else {
1465		in_use[0] = '\0';
1466		avail[0] = '\0';
1467	}
1468	(void)printf(
1469	    "%-*s%*s%*"PRId64"%5"PRId64"%*"PRId64"%s%s%*"PRId64"%*"PRId64"%6"PRId64"\n",
1470	    wide ? 16 : 11, "Totals",
1471	    wide ? 6 : 5, "",
1472	    wide ? 12 : 9, pool_totals.pt_nget,
1473	    pool_totals.pt_nfail,
1474	    wide ? 12 : 9, pool_totals.pt_nput,
1475	    in_use,
1476	    avail,
1477	    wide ? 7 : 6, pool_totals.pt_npagealloc,
1478	    wide ? 7 : 6, pool_totals.pt_npagefree,
1479	    pool_totals.pt_npages);
1480
1481	inuse /= KILO;
1482	total /= KILO;
1483	(void)printf(
1484	    "\nIn use %" PRIu64 "K, "
1485	    "total allocated %" PRIu64 "K; utilization %.1f%%\n",
1486	    inuse, total, (100.0 * inuse) / total);
1487
1488	free(data);
1489}
1490
1491void
1492dopool(int verbose, int wide)
1493{
1494	int first, ovflw;
1495	void *addr;
1496	long total, inuse, this_total, this_inuse;
1497	struct {
1498		uint64_t pt_nget;
1499		uint64_t pt_nfail;
1500		uint64_t pt_nput;
1501		uint64_t pt_nout;
1502		uint64_t pt_nitems;
1503		uint64_t pt_npagealloc;
1504		uint64_t pt_npagefree;
1505		uint64_t pt_npages;
1506	} pool_totals;
1507	char in_use[8];
1508	char avail[8];
1509	TAILQ_HEAD(,pool) pool_head;
1510	struct pool pool, *pp = &pool;
1511	struct pool_allocator pa;
1512	char name[32], maxp[32];
1513
1514	if (memf == NULL)
1515		return dopool_sysctl(verbose, wide);
1516
1517	memset(&pool_totals, 0, sizeof pool_totals);
1518	kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
1519	addr = TAILQ_FIRST(&pool_head);
1520
1521	total = inuse = 0;
1522
1523	for (first = 1; addr != NULL; addr = TAILQ_NEXT(pp, pr_poollist) ) {
1524		deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
1525		deref_kptr(pp->pr_alloc, &pa, sizeof(pa),
1526		    "pool allocator trashed");
1527		deref_kptr(pp->pr_wchan, name, sizeof(name),
1528		    "pool wait channel trashed");
1529		name[sizeof(name)-1] = '\0';
1530
1531		if (first) {
1532			(void)printf("Memory resource pool statistics\n");
1533			(void)printf(
1534			    "%-*s%*s%*s%5s%*s%s%s%*s%*s%6s%s%6s%6s%6s%5s%s%s\n",
1535			    wide ? 16 : 11, "Name",
1536			    wide ? 6 : 5, "Size",
1537			    wide ? 12 : 9, "Requests",
1538			    "Fail",
1539			    wide ? 12 : 9, "Releases",
1540			    wide ? "  InUse" : "",
1541			    wide ? " Avail" : "",
1542			    wide ? 7 : 6, "Pgreq",
1543			    wide ? 7 : 6, "Pgrel",
1544			    "Npage",
1545			    wide ? " PageSz" : "",
1546			    "Hiwat",
1547			    "Minpg",
1548			    "Maxpg",
1549			    "Idle",
1550			    wide ? " Flags" : "",
1551			    wide ? "   Util" : "");
1552			first = 0;
1553		}
1554		if (pp->pr_nget == 0 && !verbose)
1555			continue;
1556		if (pp->pr_maxpages == UINT_MAX)
1557			(void)snprintf(maxp, sizeof(maxp), "inf");
1558		else
1559			(void)snprintf(maxp, sizeof(maxp), "%u",
1560			    pp->pr_maxpages);
1561		ovflw = 0;
1562		PRWORD(ovflw, "%-*s", wide ? 16 : 11, 0, name);
1563		PRWORD(ovflw, " %*u", wide ? 6 : 5, 1, pp->pr_size);
1564		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nget);
1565		pool_totals.pt_nget += pp->pr_nget;
1566		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nfail);
1567		pool_totals.pt_nfail += pp->pr_nfail;
1568		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nput);
1569		pool_totals.pt_nput += pp->pr_nput;
1570		if (wide) {
1571			PRWORD(ovflw, " %*u", 7, 1, pp->pr_nout);
1572			pool_totals.pt_nout += pp->pr_nout;
1573		}
1574		if (wide) {
1575			PRWORD(ovflw, " %*u", 6, 1, pp->pr_nitems);
1576			pool_totals.pt_nitems += pp->pr_nitems;
1577		}
1578		PRWORD(ovflw, " %*lu", wide ? 7 : 6, 1, pp->pr_npagealloc);
1579		pool_totals.pt_npagealloc += pp->pr_npagealloc;
1580		PRWORD(ovflw, " %*lu", wide ? 7 : 6, 1, pp->pr_npagefree);
1581		pool_totals.pt_npagefree += pp->pr_npagefree;
1582		PRWORD(ovflw, " %*u", 6, 1, pp->pr_npages);
1583		pool_totals.pt_npages += pp->pr_npages;
1584		if (wide)
1585			PRWORD(ovflw, " %*u", 7, 1, pa.pa_pagesz);
1586		PRWORD(ovflw, " %*u", 6, 1, pp->pr_hiwat);
1587		PRWORD(ovflw, " %*u", 6, 1, pp->pr_minpages);
1588		PRWORD(ovflw, " %*s", 6, 1, maxp);
1589		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nidle);
1590		if (wide)
1591			PRWORD(ovflw, " 0x%0*x", 4, 1,
1592			    pp->pr_flags | pp->pr_roflags);
1593
1594		this_inuse = pp->pr_nout * pp->pr_size;
1595		this_total = pp->pr_npages * pa.pa_pagesz;
1596		if (pp->pr_roflags & PR_RECURSIVE) {
1597			/*
1598			 * Don't count in-use memory, since it's part
1599			 * of another pool and will be accounted for
1600			 * there.
1601			 */
1602			total += (this_total - this_inuse);
1603		} else {
1604			inuse += this_inuse;
1605			total += this_total;
1606		}
1607		if (wide) {
1608			if (this_total == 0)
1609				(void)printf("   ---");
1610			else
1611				(void)printf(" %5.1f%%",
1612				    (100.0 * this_inuse) / this_total);
1613		}
1614		(void)printf("\n");
1615	}
1616	if (wide) {
1617		snprintf(in_use, sizeof in_use, "%7"PRId64, pool_totals.pt_nout);
1618		snprintf(avail, sizeof avail, "%6"PRId64, pool_totals.pt_nitems);
1619	} else {
1620		in_use[0] = '\0';
1621		avail[0] = '\0';
1622	}
1623	(void)printf(
1624	    "%-*s%*s%*"PRId64"%5"PRId64"%*"PRId64"%s%s%*"PRId64"%*"PRId64"%6"PRId64"\n",
1625	    wide ? 16 : 11, "Totals",
1626	    wide ? 6 : 5, "",
1627	    wide ? 12 : 9, pool_totals.pt_nget,
1628	    pool_totals.pt_nfail,
1629	    wide ? 12 : 9, pool_totals.pt_nput,
1630	    in_use,
1631	    avail,
1632	    wide ? 7 : 6, pool_totals.pt_npagealloc,
1633	    wide ? 7 : 6, pool_totals.pt_npagefree,
1634	    pool_totals.pt_npages);
1635
1636	inuse /= KILO;
1637	total /= KILO;
1638	(void)printf(
1639	    "\nIn use %ldK, total allocated %ldK; utilization %.1f%%\n",
1640	    inuse, total, (100.0 * inuse) / total);
1641}
1642
1643static void
1644dopoolcache_sysctl(int verbose)
1645{
1646	struct pool_sysctl *data, *pp;
1647	size_t i, len;
1648	bool first = true;
1649	int ovflw;
1650	uint64_t tot;
1651	double p;
1652
1653	data = asysctlbyname("kern.pool", &len);
1654	if (data == NULL)
1655		err(1, "failed to reead kern.pool");
1656	len /= sizeof(*data);
1657
1658	for (i = 0; i < len; ++i) {
1659		pp = &data[i];
1660		if (pp->pr_cache_meta_size == 0)
1661			continue;
1662
1663		if (pp->pr_cache_nmiss_global == 0 && !verbose)
1664			continue;
1665
1666		if (first) {
1667			(void)printf("Pool cache statistics.\n");
1668			(void)printf("%-*s%*s%*s%*s%*s%*s%*s%*s%*s%*s\n",
1669			    12, "Name",
1670			    6, "Spin",
1671			    6, "GrpSz",
1672			    5, "Full",
1673			    5, "Emty",
1674			    10, "PoolLayer",
1675			    11, "CacheLayer",
1676			    6, "Hit%",
1677			    12, "CpuLayer",
1678			    6, "Hit%"
1679			);
1680			first = false;
1681		}
1682
1683		ovflw = 0;
1684		PRWORD(ovflw, "%-*s", MIN((int)sizeof(pp->pr_wchan), 13), 1,
1685		    pp->pr_wchan);
1686		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_cache_ncontended);
1687		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_cache_meta_size);
1688		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_cache_nfull);
1689		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_cache_nempty);
1690		PRWORD(ovflw, " %*" PRIu64, 10, 1, pp->pr_cache_nmiss_global);
1691
1692		tot = pp->pr_cache_nhit_global + pp->pr_cache_nmiss_global;
1693		p = pp->pr_cache_nhit_global * 100.0 / tot;
1694		PRWORD(ovflw, " %*" PRIu64, 11, 1, tot);
1695		PRWORD(ovflw, " %*.1f", 6, 1, p);
1696
1697		tot = pp->pr_cache_nhit_pcpu + pp->pr_cache_nmiss_pcpu;
1698		p = pp->pr_cache_nhit_pcpu * 100.0 / tot;
1699		PRWORD(ovflw, " %*" PRIu64, 12, 1, tot);
1700		PRWORD(ovflw, " %*.1f", 6, 1, p);
1701		printf("\n");
1702	}
1703}
1704
1705void
1706dopoolcache(int verbose)
1707{
1708	struct pool_cache pool_cache, *pc = &pool_cache;
1709	pool_cache_cpu_t cache_cpu, *cc = &cache_cpu;
1710	TAILQ_HEAD(,pool) pool_head;
1711	struct pool pool, *pp = &pool;
1712	char name[32];
1713	uint64_t cpuhit, cpumiss, tot;
1714	void *addr;
1715	int first, ovflw;
1716	size_t i;
1717	double p;
1718
1719	if (memf == NULL)
1720		return dopoolcache_sysctl(verbose);
1721
1722	kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
1723	addr = TAILQ_FIRST(&pool_head);
1724
1725	for (first = 1; addr != NULL; addr = TAILQ_NEXT(pp, pr_poollist) ) {
1726		deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
1727		if (pp->pr_cache == NULL)
1728			continue;
1729		deref_kptr(pp->pr_wchan, name, sizeof(name),
1730		    "pool wait channel trashed");
1731		deref_kptr(pp->pr_cache, pc, sizeof(*pc), "pool cache trashed");
1732		if (pc->pc_misses == 0 && !verbose)
1733			continue;
1734		name[sizeof(name)-1] = '\0';
1735
1736		cpuhit = 0;
1737		cpumiss = 0;
1738		for (i = 0; i < __arraycount(pc->pc_cpus); i++) {
1739		    	if ((addr = pc->pc_cpus[i]) == NULL)
1740		    		continue;
1741			deref_kptr(addr, cc, sizeof(*cc),
1742			    "pool cache cpu trashed");
1743			cpuhit += cc->cc_hits;
1744			cpumiss += cc->cc_misses;
1745		}
1746
1747		if (first) {
1748			(void)printf("Pool cache statistics.\n");
1749			(void)printf("%-*s%*s%*s%*s%*s%*s%*s%*s%*s%*s\n",
1750			    12, "Name",
1751			    6, "Spin",
1752			    6, "GrpSz",
1753			    5, "Full",
1754			    5, "Emty",
1755			    10, "PoolLayer",
1756			    11, "CacheLayer",
1757			    6, "Hit%",
1758			    12, "CpuLayer",
1759			    6, "Hit%"
1760			);
1761			first = 0;
1762		}
1763
1764		ovflw = 0;
1765		PRWORD(ovflw, "%-*s", 13, 1, name);
1766		PRWORD(ovflw, " %*llu", 6, 1, (long long)pc->pc_contended);
1767		PRWORD(ovflw, " %*u", 6, 1, pc->pc_pcgsize);
1768		PRWORD(ovflw, " %*u", 5, 1, pc->pc_nfull);
1769		PRWORD(ovflw, " %*u", 5, 1, pc->pc_nempty);
1770		PRWORD(ovflw, " %*llu", 10, 1, (long long)pc->pc_misses);
1771
1772		tot = pc->pc_hits + pc->pc_misses;
1773		p = pc->pc_hits * 100.0 / (tot);
1774		PRWORD(ovflw, " %*llu", 11, 1, (long long)tot);
1775		PRWORD(ovflw, " %*.1f", 6, 1, p);
1776
1777		tot = cpuhit + cpumiss;
1778		p = cpuhit * 100.0 / (tot);
1779		PRWORD(ovflw, " %*llu", 12, 1, (long long)tot);
1780		PRWORD(ovflw, " %*.1f", 6, 1, p);
1781		printf("\n");
1782	}
1783}
1784
1785enum hashtype {			/* from <sys/systm.h> */
1786	HASH_LIST,
1787	HASH_TAILQ
1788};
1789
1790struct uidinfo {		/* XXX: no kernel header file */
1791	LIST_ENTRY(uidinfo) ui_hash;
1792	uid_t	ui_uid;
1793	long	ui_proccnt;
1794};
1795
1796struct kernel_hash {
1797	const char *	description;	/* description */
1798	int		hashsize;	/* nlist index for hash size */
1799	int		hashtbl;	/* nlist index for hash table */
1800	enum hashtype	type;		/* type of hash table */
1801	size_t		offset;		/* offset of {LIST,TAILQ}_NEXT */
1802} khashes[] =
1803{
1804	{
1805		"buffer hash",
1806		X_BUFHASH, X_BUFHASHTBL,
1807		HASH_LIST, offsetof(struct buf, b_hash)
1808	}, {
1809		"ipv4 address -> interface hash",
1810		X_IFADDRHASH, X_IFADDRHASHTBL,
1811		HASH_LIST, offsetof(struct in_ifaddr, ia_hash),
1812	}, {
1813		"name cache hash",
1814		X_NCHASH, X_NCHASHTBL,
1815		HASH_LIST, offsetof(struct namecache, nc_hash),
1816	}, {
1817		"name cache directory hash",
1818		X_NCVHASH, X_NCVHASHTBL,
1819		HASH_LIST, offsetof(struct namecache, nc_vhash),
1820	}, {
1821		"user info (uid -> used processes) hash",
1822		X_UIHASH, X_UIHASHTBL,
1823		HASH_LIST, offsetof(struct uidinfo, ui_hash),
1824	}, {
1825		NULL, -1, -1, 0, 0,
1826	}
1827};
1828
1829void
1830dohashstat(int verbose, int todo, const char *hashname)
1831{
1832	LIST_HEAD(, generic)	*hashtbl_list;
1833	TAILQ_HEAD(, generic)	*hashtbl_tailq;
1834	struct kernel_hash	*curhash;
1835	void	*hashaddr, *hashbuf, *nhashbuf, *nextaddr;
1836	size_t	elemsize, hashbufsize, thissize;
1837	u_long	hashsize, i;
1838	int	used, items, chain, maxchain;
1839
1840	hashbuf = NULL;
1841	hashbufsize = 0;
1842
1843	if (todo & HASHLIST) {
1844		(void)printf("Supported hashes:\n");
1845		for (curhash = khashes; curhash->description; curhash++) {
1846			if (hashnl[curhash->hashsize].n_value == 0 ||
1847			    hashnl[curhash->hashtbl].n_value == 0)
1848				continue;
1849			(void)printf("\t%-16s%s\n",
1850			    hashnl[curhash->hashsize].n_name + 1,
1851			    curhash->description);
1852		}
1853		return;
1854	}
1855
1856	if (hashname != NULL) {
1857		for (curhash = khashes; curhash->description; curhash++) {
1858			if (strcmp(hashnl[curhash->hashsize].n_name + 1,
1859			    hashname) == 0 &&
1860			    hashnl[curhash->hashsize].n_value != 0 &&
1861			    hashnl[curhash->hashtbl].n_value != 0)
1862				break;
1863		}
1864		if (curhash->description == NULL) {
1865			warnx("%s: no such hash", hashname);
1866			return;
1867		}
1868	}
1869
1870	(void)printf(
1871	    "%-16s %8s %8s %8s %8s %8s %8s\n"
1872	    "%-16s %8s %8s %8s %8s %8s %8s\n",
1873	    "", "total", "used", "util", "num", "average", "maximum",
1874	    "hash table", "buckets", "buckets", "%", "items", "chain",
1875	    "chain");
1876
1877	for (curhash = khashes; curhash->description; curhash++) {
1878		if (hashnl[curhash->hashsize].n_value == 0 ||
1879		    hashnl[curhash->hashtbl].n_value == 0)
1880			continue;
1881		if (hashname != NULL &&
1882		    strcmp(hashnl[curhash->hashsize].n_name + 1, hashname))
1883			continue;
1884		elemsize = curhash->type == HASH_LIST ?
1885		    sizeof(*hashtbl_list) : sizeof(*hashtbl_tailq);
1886		deref_kptr((void *)hashnl[curhash->hashsize].n_value,
1887		    &hashsize, sizeof(hashsize),
1888		    hashnl[curhash->hashsize].n_name);
1889		hashsize++;
1890		deref_kptr((void *)hashnl[curhash->hashtbl].n_value,
1891		    &hashaddr, sizeof(hashaddr),
1892		    hashnl[curhash->hashtbl].n_name);
1893		if (verbose)
1894			(void)printf(
1895			    "%s %lu, %s %p, offset %ld, elemsize %llu\n",
1896			    hashnl[curhash->hashsize].n_name + 1, hashsize,
1897			    hashnl[curhash->hashtbl].n_name + 1, hashaddr,
1898			    (long)curhash->offset,
1899			    (unsigned long long)elemsize);
1900		thissize = hashsize * elemsize;
1901		if (hashbuf == NULL || thissize > hashbufsize) {
1902			if ((nhashbuf = realloc(hashbuf, thissize)) == NULL)
1903				errx(1, "malloc hashbuf %llu",
1904				    (unsigned long long)hashbufsize);
1905			hashbuf = nhashbuf;
1906			hashbufsize = thissize;
1907		}
1908		deref_kptr(hashaddr, hashbuf, thissize,
1909		    hashnl[curhash->hashtbl].n_name);
1910		used = 0;
1911		items = maxchain = 0;
1912		if (curhash->type == HASH_LIST) {
1913			hashtbl_list = hashbuf;
1914			hashtbl_tailq = NULL;
1915		} else {
1916			hashtbl_list = NULL;
1917			hashtbl_tailq = hashbuf;
1918		}
1919		for (i = 0; i < hashsize; i++) {
1920			if (curhash->type == HASH_LIST)
1921				nextaddr = LIST_FIRST(&hashtbl_list[i]);
1922			else
1923				nextaddr = TAILQ_FIRST(&hashtbl_tailq[i]);
1924			if (nextaddr == NULL)
1925				continue;
1926			if (verbose)
1927				(void)printf("%5lu: %p\n", i, nextaddr);
1928			used++;
1929			chain = 0;
1930			do {
1931				chain++;
1932				deref_kptr((char *)nextaddr + curhash->offset,
1933				    &nextaddr, sizeof(void *),
1934				    "hash chain corrupted");
1935				if (verbose > 1)
1936					(void)printf("got nextaddr as %p\n",
1937					    nextaddr);
1938			} while (nextaddr != NULL);
1939			items += chain;
1940			if (verbose && chain > 1)
1941				(void)printf("\tchain = %d\n", chain);
1942			if (chain > maxchain)
1943				maxchain = chain;
1944		}
1945		(void)printf("%-16s %8ld %8d %8.2f %8d %8.2f %8d\n",
1946		    hashnl[curhash->hashsize].n_name + 1,
1947		    hashsize, used, used * 100.0 / hashsize,
1948		    items, used ? (double)items / used : 0.0, maxchain);
1949	}
1950}
1951
1952/*
1953 * kreadc like kread but returns 1 if sucessful, 0 otherwise
1954 */
1955int
1956kreadc(struct nlist *nl, int nlx, void *addr, size_t size)
1957{
1958	const char *sym;
1959
1960	sym = nl[nlx].n_name;
1961	if (*sym == '_')
1962		++sym;
1963	if (nl[nlx].n_type == 0 || nl[nlx].n_value == 0)
1964		return 0;
1965	deref_kptr((void *)nl[nlx].n_value, addr, size, sym);
1966	return 1;
1967}
1968
1969/*
1970 * kread reads something from the kernel, given its nlist index in namelist[].
1971 */
1972void
1973kread(struct nlist *nl, int nlx, void *addr, size_t size)
1974{
1975	const char *sym;
1976
1977	sym = nl[nlx].n_name;
1978	if (*sym == '_')
1979		++sym;
1980	if (nl[nlx].n_type == 0 || nl[nlx].n_value == 0)
1981		errx(1, "symbol %s not defined", sym);
1982	deref_kptr((void *)nl[nlx].n_value, addr, size, sym);
1983}
1984
1985/*
1986 * Dereference the kernel pointer `kptr' and fill in the local copy
1987 * pointed to by `ptr'.  The storage space must be pre-allocated,
1988 * and the size of the copy passed in `len'.
1989 */
1990void
1991deref_kptr(const void *kptr, void *ptr, size_t len, const char *msg)
1992{
1993
1994	if (*msg == '_')
1995		msg++;
1996	if ((size_t)kvm_read(kd, (u_long)kptr, (char *)ptr, len) != len)
1997		errx(1, "kptr %lx: %s: %s", (u_long)kptr, msg, kvm_geterr(kd));
1998}
1999
2000/*
2001 * Traverse the kernel history buffers, performing the requested action.
2002 *
2003 * Note, we assume that if we're not listing, we're dumping.
2004 */
2005void
2006hist_traverse(int todo, const char *histname)
2007{
2008	struct kern_history_head histhead;
2009	struct kern_history hist, *histkva;
2010	char *name = NULL;
2011	size_t namelen = 0;
2012
2013	if (histnl[0].n_value == 0) {
2014		warnx("kernel history is not compiled into the kernel.");
2015		return;
2016	}
2017
2018	deref_kptr((void *)histnl[X_KERN_HISTORIES].n_value, &histhead,
2019	    sizeof(histhead), histnl[X_KERN_HISTORIES].n_name);
2020
2021	if (histhead.lh_first == NULL) {
2022		warnx("No active kernel history logs.");
2023		return;
2024	}
2025
2026	if (todo & HISTLIST)
2027		(void)printf("Active kernel histories:");
2028
2029	for (histkva = LIST_FIRST(&histhead); histkva != NULL;
2030	    histkva = LIST_NEXT(&hist, list)) {
2031		deref_kptr(histkva, &hist, sizeof(hist), "histkva");
2032		if (name == NULL || hist.namelen > namelen) {
2033			if (name != NULL)
2034				free(name);
2035			namelen = hist.namelen;
2036			if ((name = malloc(namelen + 1)) == NULL)
2037				err(1, "malloc history name");
2038		}
2039
2040		deref_kptr(hist.name, name, namelen, "history name");
2041		name[namelen] = '\0';
2042		if (todo & HISTLIST)
2043			(void)printf(" %s", name);
2044		else {
2045			/*
2046			 * If we're dumping all histories, do it, else
2047			 * check to see if this is the one we want.
2048			 */
2049			if (histname == NULL || strcmp(histname, name) == 0) {
2050				if (histname == NULL)
2051					(void)printf(
2052					    "\nkernel history `%s':\n", name);
2053				hist_dodump(&hist);
2054			}
2055		}
2056	}
2057
2058	if (todo & HISTLIST)
2059		(void)putchar('\n');
2060
2061	if (name != NULL)
2062		free(name);
2063}
2064
2065/*
2066 * Actually dump the history buffer at the specified KVA.
2067 */
2068void
2069hist_dodump(struct kern_history *histp)
2070{
2071	struct kern_history_ent *histents, *e;
2072	struct timeval tv;
2073	size_t histsize;
2074	char *fmt = NULL, *fn = NULL;
2075	size_t fmtlen = 0, fnlen = 0;
2076	unsigned i;
2077
2078	histsize = sizeof(struct kern_history_ent) * histp->n;
2079
2080	if ((histents = malloc(histsize)) == NULL)
2081		err(1, "malloc history entries");
2082
2083	(void)memset(histents, 0, histsize);
2084
2085	(void)printf("%"PRIu32" entries, next is %"PRIu32"\n",
2086	    histp->n, histp->f);
2087
2088	deref_kptr(histp->e, histents, histsize, "history entries");
2089	i = histp->f;
2090	do {
2091		e = &histents[i];
2092		if (e->fmt != NULL) {
2093			if (fmt == NULL || e->fmtlen > fmtlen) {
2094				if (fmt != NULL)
2095					free(fmt);
2096				fmtlen = e->fmtlen;
2097				if ((fmt = malloc(fmtlen + 1)) == NULL)
2098					err(1, "malloc printf format");
2099			}
2100			if (fn == NULL || e->fnlen > fnlen) {
2101				if (fn != NULL)
2102					free(fn);
2103				fnlen = e->fnlen;
2104				if ((fn = malloc(fnlen + 1)) == NULL)
2105					err(1, "malloc function name");
2106			}
2107
2108			deref_kptr(e->fmt, fmt, fmtlen, "printf format");
2109			fmt[fmtlen] = '\0';
2110
2111			deref_kptr(e->fn, fn, fnlen, "function name");
2112			fn[fnlen] = '\0';
2113
2114			bintime2timeval(&e->bt, &tv);
2115			(void)printf("%06ld.%06ld ", (long int)tv.tv_sec,
2116			    (long int)tv.tv_usec);
2117			(void)printf("%s#%ld@%d: ", fn, e->call, e->cpunum);
2118			(void)printf(fmt, e->v[0], e->v[1], e->v[2], e->v[3]);
2119			(void)putchar('\n');
2120		}
2121		i = (i + 1) % histp->n;
2122	} while (i != histp->f);
2123
2124	free(histents);
2125	if (fmt != NULL)
2126		free(fmt);
2127	if (fn != NULL)
2128		free(fn);
2129}
2130
2131void
2132hist_traverse_sysctl(int todo, const char *histname)
2133{
2134	int error;
2135	int mib[4];
2136	unsigned int i;
2137	size_t len, miblen;
2138	struct sysctlnode query, histnode[32];
2139
2140	/* retrieve names of available histories */
2141	miblen = __arraycount(mib);
2142	error = sysctlnametomib("kern.hist", mib, &miblen);
2143	if (error != 0) {
2144		if (errno == ENOENT) {
2145 			warnx("kernel history is not compiled into the kernel.");
2146			return;
2147		} else
2148			err(EXIT_FAILURE, "nametomib failed");
2149	}
2150
2151	/* get the list of nodenames below kern.hist */
2152	mib[2] = CTL_QUERY;
2153	memset(&query, 0, sizeof(query));
2154	query.sysctl_flags = SYSCTL_VERSION;
2155	len = sizeof(histnode);
2156	error = sysctl(mib, 3, &histnode[0], &len, &query, sizeof(query));
2157	if (error != 0) {
2158		err(1, "query failed");
2159		return;
2160	}
2161	if (len == 0) {
2162 		warnx("No active kernel history logs.");
2163 		return;
2164 	}
2165
2166	len = len / sizeof(histnode[0]);	/* get # of entries returned */
2167
2168 	if (todo & HISTLIST)
2169 		(void)printf("Active kernel histories:");
2170
2171	for (i = 0; i < len; i++) {
2172 		if (todo & HISTLIST)
2173			(void)printf(" %s", histnode[i].sysctl_name);
2174 		else {
2175 			/*
2176 			 * If we're dumping all histories, do it, else
2177 			 * check to see if this is the one we want.
2178 			 */
2179			if (histname == NULL ||
2180			    strcmp(histname, histnode[i].sysctl_name) == 0) {
2181 				if (histname == NULL)
2182 					(void)printf(
2183					    "\nkernel history `%s':\n",
2184					    histnode[i].sysctl_name);
2185				mib[2] = histnode[i].sysctl_num;
2186				mib[3] = CTL_EOL;
2187				hist_dodump_sysctl(mib, 4);
2188 			}
2189 		}
2190 	}
2191
2192 	if (todo & HISTLIST)
2193 		(void)putchar('\n');
2194 }
2195
2196 /*
2197  * Actually dump the history buffer at the specified KVA.
2198  */
2199 void
2200hist_dodump_sysctl(int mib[], unsigned int miblen)
2201 {
2202	struct sysctl_history *hist;
2203	struct timeval tv;
2204	struct sysctl_history_event *e;
2205 	size_t histsize;
2206	char *strp;
2207 	unsigned i;
2208	char *fmt = NULL, *fn = NULL;
2209
2210	hist = NULL;
2211	histsize = 0;
2212 	do {
2213		errno = 0;
2214		if (sysctl(mib, miblen, hist, &histsize, NULL, 0) == 0)
2215			break;
2216		if (errno != ENOMEM)
2217			break;
2218		if ((hist = realloc(hist, histsize)) == NULL)
2219			errx(1, "realloc history buffer");
2220	} while (errno == ENOMEM);
2221	if (errno != 0)
2222		err(1, "sysctl failed");
2223
2224	strp = (char *)(&hist->sh_events[hist->sh_listentry.shle_numentries]);
2225
2226	(void)printf("%"PRIu32" entries, next is %"PRIu32"\n",
2227	    hist->sh_listentry.shle_numentries,
2228	    hist->sh_listentry.shle_nextfree);
2229
2230	i = hist->sh_listentry.shle_nextfree;
2231
2232	do {
2233		e = &hist->sh_events[i];
2234		if (e->she_fmtoffset != 0) {
2235			fmt = &strp[e->she_fmtoffset];
2236			fn = &strp[e->she_funcoffset];
2237			bintime2timeval(&e->she_bintime, &tv);
2238			(void)printf("%06ld.%06ld %s#%"PRIu64"@%"PRIu32": ",
2239			    (long int)tv.tv_sec, (long int)tv.tv_usec,
2240			    fn, e->she_callnumber, e->she_cpunum);
2241			(void)printf(fmt, e->she_values[0], e->she_values[1],
2242			     e->she_values[2], e->she_values[3]);
2243 			(void)putchar('\n');
2244 		}
2245		i = (i + 1) % hist->sh_listentry.shle_numentries;
2246	} while (i != hist->sh_listentry.shle_nextfree);
2247
2248	free(hist);
2249 }
2250
2251static void
2252usage(void)
2253{
2254
2255	(void)fprintf(stderr,
2256	    "usage: %s [-CefHiLlmstUvW] [-c count] [-h hashname] [-M core] [-N system]\n"
2257	    "\t\t[-u histname] [-w wait] [disks]\n", getprogname());
2258	exit(1);
2259}
2260