vmstat.c revision 1.233
1/* $NetBSD: vmstat.c,v 1.233 2020/01/06 11:24:30 ad Exp $ */
2
3/*-
4 * Copyright (c) 1998, 2000, 2001, 2007, 2019 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 1.233 2020/01/06 11:24:30 ad Exp $");
74#endif
75#endif /* not lint */
76
77#define	__POOL_EXPOSE
78#define __NAMECACHE_PRIVATE
79
80#include <sys/param.h>
81#include <sys/types.h>
82#include <sys/mount.h>
83#include <sys/uio.h>
84
85#include <sys/buf.h>
86#include <sys/evcnt.h>
87#include <sys/ioctl.h>
88#include <sys/malloc.h>
89#include <sys/mallocvar.h>
90#include <sys/namei.h>
91#include <sys/pool.h>
92#include <sys/proc.h>
93#include <sys/sched.h>
94#include <sys/socket.h>
95#include <sys/sysctl.h>
96#include <sys/time.h>
97#include <sys/queue.h>
98#include <sys/kernhist.h>
99#include <sys/vnode.h>
100#include <sys/vnode_impl.h>
101
102#include <uvm/uvm_extern.h>
103#include <uvm/uvm_stat.h>
104
105#include <net/if.h>
106#include <netinet/in.h>
107#include <netinet/in_var.h>
108
109#include <ufs/ufs/inode.h>
110
111#include <nfs/rpcv2.h>
112#include <nfs/nfsproto.h>
113#include <nfs/nfsnode.h>
114
115#include <ctype.h>
116#include <err.h>
117#include <errno.h>
118#include <fcntl.h>
119#include <kvm.h>
120#include <limits.h>
121#include <nlist.h>
122#undef n_hash
123#include <paths.h>
124#include <signal.h>
125#include <stdio.h>
126#include <stddef.h>
127#include <stdlib.h>
128#include <string.h>
129#include <time.h>
130#include <unistd.h>
131#include <util.h>
132
133#include "drvstats.h"
134
135/*
136 * All this mess will go away once everything is converted.
137 */
138#ifdef __HAVE_CPU_DATA_FIRST
139
140# include <sys/cpu_data.h>
141struct cpu_info {
142	struct cpu_data ci_data;
143};
144#else
145# include <sys/cpu.h>
146#endif
147
148/*
149 * General namelist
150 */
151struct nlist namelist[] =
152{
153#define	X_HZ		0
154	{ .n_name = "_hz" },
155#define	X_STATHZ	1
156	{ .n_name = "_stathz" },
157#define	X_NCHSTATS	2
158	{ .n_name = "_nchstats" },
159#define	X_ALLEVENTS	3
160	{ .n_name = "_allevents" },
161#define	X_POOLHEAD	4
162	{ .n_name = "_pool_head" },
163#define	X_UVMEXP	5
164	{ .n_name = "_uvmexp" },
165#define X_CPU_INFOS	6
166	{ .n_name = "_cpu_infos" },
167#define	X_NL_SIZE	7
168	{ .n_name = NULL },
169};
170
171/*
172 * Namelist for time data.
173 */
174struct nlist timenl[] =
175{
176#define	X_TIMEBASEBIN	0
177	{ .n_name = "_timebasebin" },
178#define	X_TIME_SECOND	1
179	{ .n_name = "_time_second" },
180#define X_TIME		2
181	{ .n_name = "_time" },
182#define	X_TIMENL_SIZE	3
183	{ .n_name = NULL },
184};
185
186/*
187 * Namelist for pre-evcnt interrupt counters.
188 */
189struct nlist intrnl[] =
190{
191#define	X_INTRNAMES	0
192	{ .n_name = "_intrnames" },
193#define	X_EINTRNAMES	1
194	{ .n_name = "_eintrnames" },
195#define	X_INTRCNT	2
196	{ .n_name = "_intrcnt" },
197#define	X_EINTRCNT	3
198	{ .n_name = "_eintrcnt" },
199#define	X_INTRNL_SIZE	4
200	{ .n_name = NULL },
201};
202
203
204/*
205 * Namelist for hash statistics
206 */
207struct nlist hashnl[] =
208{
209#define	X_NFSNODE	0
210	{ .n_name = "_nfsnodehash" },
211#define	X_NFSNODETBL	1
212	{ .n_name = "_nfsnodehashtbl" },
213#define	X_IHASH		2
214	{ .n_name = "_ihash" },
215#define	X_IHASHTBL	3
216	{ .n_name = "_ihashtbl" },
217#define	X_BUFHASH	4
218	{ .n_name = "_bufhash" },
219#define	X_BUFHASHTBL	5
220	{ .n_name = "_bufhashtbl" },
221#define	X_UIHASH	6
222	{ .n_name = "_uihash" },
223#define	X_UIHASHTBL	7
224	{ .n_name = "_uihashtbl" },
225#define	X_IFADDRHASH	8
226	{ .n_name = "_in_ifaddrhash" },
227#define	X_IFADDRHASHTBL	9
228	{ .n_name = "_in_ifaddrhashtbl" },
229#define	X_NCHASH	10
230	{ .n_name = "_nchash" },
231#define	X_NCHASHTBL	11
232	{ .n_name = "_nchashtbl" },
233#define	X_NCVCACHEHASH	12
234	{ .n_name = "_vcache_hashmask" },
235#define	X_NCVCACHETBL	13
236	{ .n_name = "_vcache_hashtab" },
237#define X_HASHNL_SIZE	14	/* must be last */
238	{ .n_name = NULL },
239};
240
241/*
242 * Namelist for kernel histories
243 */
244struct nlist histnl[] =
245{
246	{ .n_name = "_kern_histories" },
247#define	X_KERN_HISTORIES		0
248	{ .n_name = NULL },
249};
250
251
252#define KILO	1024
253
254struct cpu_counter {
255	uint64_t nintr;
256	uint64_t nsyscall;
257	uint64_t nswtch;
258	uint64_t nfault;
259	uint64_t ntrap;
260	uint64_t nsoft;
261} cpucounter, ocpucounter;
262
263struct	uvmexp_sysctl uvmexp, ouvmexp;
264int	ndrives;
265
266int	winlines = 20;
267
268kvm_t *kd;
269
270
271#define	FORKSTAT	0x001
272#define	INTRSTAT	0x002
273#define	MEMSTAT		0x004
274#define	SUMSTAT		0x008
275#define	EVCNTSTAT	0x010
276#define	VMSTAT		0x020
277#define	HISTLIST	0x040
278#define	HISTDUMP	0x080
279#define	HASHSTAT	0x100
280#define	HASHLIST	0x200
281#define	VMTOTAL		0x400
282#define	POOLCACHESTAT	0x800
283
284/*
285 * Print single word.  `ovflow' is number of characters didn't fit
286 * on the last word.  `fmt' is a format string to print this word.
287 * It must contain asterisk for field width.  `width' is a width
288 * occupied by this word.  `fixed' is a number of constant chars in
289 * `fmt'.  `val' is a value to be printed using format string `fmt'.
290 */
291#define	PRWORD(ovflw, fmt, width, fixed, val) do {	\
292	(ovflw) += printf((fmt),			\
293	    (width) - (fixed) - (ovflw) > 0 ?		\
294	    (width) - (fixed) - (ovflw) : 0,		\
295	    (val)) - (width);				\
296	if ((ovflw) < 0)				\
297		(ovflw) = 0;				\
298} while (/* CONSTCOND */0)
299
300void	cpustats(int *);
301void	cpucounters(struct cpu_counter *);
302void	deref_kptr(const void *, void *, size_t, const char *);
303void	drvstats(int *);
304void	doevcnt(int verbose, int type);
305void	dohashstat(int, int, const char *);
306void	dointr(int verbose);
307void	dopool(int, int);
308void	dopoolcache(int);
309void	dosum(void);
310void	dovmstat(struct timespec *, int);
311void	print_total_hdr(void);
312void	dovmtotal(struct timespec *, int);
313void	kread(struct nlist *, int, void *, size_t);
314int	kreadc(struct nlist *, int, void *, size_t);
315void	needhdr(int);
316void	getnlist(int);
317long	getuptime(void);
318void	printhdr(void);
319long	pct(u_long, u_long);
320__dead static void	usage(void);
321void	doforkst(void);
322
323void	hist_traverse(int, const char *);
324void	hist_dodump(struct kern_history *);
325void	hist_traverse_sysctl(int, const char *);
326void	hist_dodump_sysctl(int[], unsigned int);
327
328char	**choosedrives(char **);
329
330/* Namelist and memory file names. */
331char	*nlistf, *memf;
332
333/* allow old usage [vmstat 1] */
334#define	BACKWARD_COMPATIBILITY
335
336static const int clockrate_mib[] = { CTL_KERN, KERN_CLOCKRATE };
337static const int vmmeter_mib[] = { CTL_VM, VM_METER };
338static const int uvmexp2_mib[] = { CTL_VM, VM_UVMEXP2 };
339static const int boottime_mib[] = { CTL_KERN, KERN_BOOTTIME };
340static char kvm_errbuf[_POSIX2_LINE_MAX];
341
342int
343main(int argc, char *argv[])
344{
345	int c, todo, verbose, wide;
346	struct timespec interval;
347	int reps;
348	gid_t egid = getegid();
349	const char *histname, *hashname;
350
351	histname = hashname = NULL;
352	(void)setegid(getgid());
353	memf = nlistf = NULL;
354	reps = todo = verbose = wide = 0;
355	interval.tv_sec = 0;
356	interval.tv_nsec = 0;
357	while ((c = getopt(argc, argv, "Cc:efh:HilLM:mN:stu:UvWw:")) != -1) {
358		switch (c) {
359		case 'c':
360			reps = atoi(optarg);
361			break;
362		case 'C':
363			todo |= POOLCACHESTAT;
364			break;
365		case 'e':
366			todo |= EVCNTSTAT;
367			break;
368		case 'f':
369			todo |= FORKSTAT;
370			break;
371		case 'h':
372			hashname = optarg;
373			/* FALLTHROUGH */
374		case 'H':
375			todo |= HASHSTAT;
376			break;
377		case 'i':
378			todo |= INTRSTAT;
379			break;
380		case 'l':
381			todo |= HISTLIST;
382			break;
383		case 'L':
384			todo |= HASHLIST;
385			break;
386		case 'M':
387			memf = optarg;
388			break;
389		case 'm':
390			todo |= MEMSTAT;
391			break;
392		case 'N':
393			nlistf = optarg;
394			break;
395		case 's':
396			todo |= SUMSTAT;
397			break;
398		case 't':
399			todo |= VMTOTAL;
400			break;
401		case 'u':
402			histname = optarg;
403			/* FALLTHROUGH */
404		case 'U':
405			todo |= HISTDUMP;
406			break;
407		case 'v':
408			verbose++;
409			break;
410		case 'W':
411			wide++;
412			break;
413		case 'w':
414			interval.tv_sec = atol(optarg);
415			break;
416		case '?':
417		default:
418			usage();
419		}
420	}
421	argc -= optind;
422	argv += optind;
423
424	if (todo == 0)
425		todo = VMSTAT;
426
427	/*
428	 * Discard setgid privileges.  If not the running kernel, we toss
429	 * them away totally so that bad guys can't print interesting stuff
430	 * from kernel memory, otherwise switch back to kmem for the
431	 * duration of the kvm_openfiles() call.
432	 */
433	if (nlistf != NULL || memf != NULL)
434		(void)setgid(getgid());
435	else
436		(void)setegid(egid);
437
438	kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, kvm_errbuf);
439	if (kd == NULL) {
440		if (nlistf != NULL || memf != NULL) {
441			errx(1, "kvm_openfiles: %s", kvm_errbuf);
442		}
443	}
444
445	if (nlistf == NULL && memf == NULL)
446		(void)setgid(getgid());
447
448
449	if (todo & VMSTAT) {
450		struct winsize winsize;
451
452		(void)drvinit(0);/* Initialize disk stats, no disks selected. */
453
454		(void)setgid(getgid()); /* don't need privs anymore */
455
456		argv = choosedrives(argv);	/* Select disks. */
457		winsize.ws_row = 0;
458		(void)ioctl(STDOUT_FILENO, TIOCGWINSZ, &winsize);
459		if (winsize.ws_row > 0)
460			winlines = winsize.ws_row;
461
462	}
463
464#ifdef	BACKWARD_COMPATIBILITY
465	if (*argv) {
466		interval.tv_sec = atol(*argv);
467		if (*++argv)
468			reps = atoi(*argv);
469	}
470#endif
471
472	if (interval.tv_sec) {
473		if (!reps)
474			reps = -1;
475	} else if (reps)
476		interval.tv_sec = 1;
477
478
479	getnlist(todo);
480	/*
481	 * Statistics dumping is incompatible with the default
482	 * VMSTAT/dovmstat() output. So perform the interval/reps handling
483	 * for it here.
484	 */
485	if ((todo & (VMSTAT|VMTOTAL)) == 0) {
486		for (;;) {
487			if (todo & (HISTLIST|HISTDUMP)) {
488				if ((todo & (HISTLIST|HISTDUMP)) ==
489				    (HISTLIST|HISTDUMP))
490					errx(1, "you may list or dump,"
491					    " but not both!");
492				if (memf != NULL)
493					hist_traverse(todo, histname);
494				else
495					hist_traverse_sysctl(todo, histname);
496				(void)putchar('\n');
497			}
498			if (todo & FORKSTAT) {
499				doforkst();
500				(void)putchar('\n');
501			}
502			if (todo & MEMSTAT) {
503				dopool(verbose, wide);
504				(void)putchar('\n');
505			}
506			if (todo & POOLCACHESTAT) {
507				dopoolcache(verbose);
508				(void)putchar('\n');
509			}
510			if (todo & SUMSTAT) {
511				dosum();
512				(void)putchar('\n');
513			}
514			if (todo & INTRSTAT) {
515				dointr(verbose);
516				(void)putchar('\n');
517			}
518			if (todo & EVCNTSTAT) {
519				doevcnt(verbose, EVCNT_TYPE_ANY);
520				(void)putchar('\n');
521			}
522			if (todo & (HASHLIST|HASHSTAT)) {
523				if ((todo & (HASHLIST|HASHSTAT)) ==
524				    (HASHLIST|HASHSTAT))
525					errx(1, "you may list or display,"
526					    " but not both!");
527				dohashstat(verbose, todo, hashname);
528				(void)putchar('\n');
529			}
530
531			fflush(stdout);
532			if (reps >= 0 && --reps <=0)
533				break;
534			(void)nanosleep(&interval, NULL);
535		}
536	} else {
537		if ((todo & (VMSTAT|VMTOTAL)) == (VMSTAT|VMTOTAL)) {
538			errx(1, "you may not both do vmstat and vmtotal");
539		}
540		if (todo & VMSTAT)
541			dovmstat(&interval, reps);
542		if (todo & VMTOTAL)
543			dovmtotal(&interval, reps);
544	}
545	return 0;
546}
547
548void
549getnlist(int todo)
550{
551	static int namelist_done = 0;
552	static int done = 0;
553	int c;
554	size_t i;
555
556	if (kd == NULL)
557		errx(1, "kvm_openfiles: %s", kvm_errbuf);
558
559	if (!namelist_done) {
560		namelist_done = 1;
561		if ((c = kvm_nlist(kd, namelist)) != 0) {
562			int doexit = 0;
563			if (c == -1)
564				errx(1, "kvm_nlist: %s %s",
565				    "namelist", kvm_geterr(kd));
566			for (i = 0; i < __arraycount(namelist)-1; i++)
567				if (namelist[i].n_type == 0) {
568					if (doexit++ == 0)
569						(void)fprintf(stderr,
570						    "%s: undefined symbols:",
571						    getprogname());
572					(void)fprintf(stderr, " %s",
573					    namelist[i].n_name);
574				}
575			if (doexit) {
576				(void)fputc('\n', stderr);
577				exit(1);
578			}
579		}
580	}
581	if ((todo & (VMSTAT|INTRSTAT)) && !(done & (VMSTAT))) {
582		done |= VMSTAT;
583		if ((c = kvm_nlist(kd, timenl)) == -1 || c == X_TIMENL_SIZE)
584			errx(1, "kvm_nlist: %s %s", "timenl", kvm_geterr(kd));
585	}
586	if ((todo & (SUMSTAT|INTRSTAT)) && !(done & (SUMSTAT|INTRSTAT))) {
587		done |= SUMSTAT|INTRSTAT;
588		(void) kvm_nlist(kd, intrnl);
589	}
590	if ((todo & (HASHLIST|HASHSTAT)) && !(done & (HASHLIST|HASHSTAT))) {
591		done |= HASHLIST|HASHSTAT;
592		if ((c = kvm_nlist(kd, hashnl)) == -1 || c == X_HASHNL_SIZE)
593			errx(1, "kvm_nlist: %s %s", "hashnl", kvm_geterr(kd));
594	}
595	if ((todo & (HISTLIST|HISTDUMP)) && !(done & (HISTLIST|HISTDUMP))) {
596		done |= HISTLIST|HISTDUMP;
597		if (kvm_nlist(kd, histnl) == -1)
598			errx(1, "kvm_nlist: %s %s", "histnl", kvm_geterr(kd));
599	}
600}
601
602char **
603choosedrives(char **argv)
604{
605	size_t i;
606
607	/*
608	 * Choose drives to be displayed.  Priority goes to (in order) drives
609	 * supplied as arguments, default drives.  If everything isn't filled
610	 * in and there are drives not taken care of, display the first few
611	 * that fit.
612	 */
613#define	BACKWARD_COMPATIBILITY
614	for (ndrives = 0; *argv; ++argv) {
615#ifdef	BACKWARD_COMPATIBILITY
616		if (isdigit((unsigned char)**argv))
617			break;
618#endif
619		for (i = 0; i < ndrive; i++) {
620			if (strcmp(dr_name[i], *argv))
621				continue;
622			drv_select[i] = 1;
623			++ndrives;
624			break;
625		}
626	}
627	for (i = 0; i < ndrive && ndrives < 2; i++) {
628		if (drv_select[i])
629			continue;
630		drv_select[i] = 1;
631		++ndrives;
632	}
633
634	return (argv);
635}
636
637long
638getuptime(void)
639{
640	static struct timespec boottime;
641	struct timespec now;
642	time_t uptime, nowsec;
643
644	if (memf == NULL) {
645		if (boottime.tv_sec == 0) {
646			size_t buflen = sizeof(boottime);
647			if (sysctl(boottime_mib, __arraycount(boottime_mib),
648			    &boottime, &buflen, NULL, 0) == -1)
649				warn("Can't get boottime");
650		}
651		clock_gettime(CLOCK_REALTIME, &now);
652	} else {
653		if (boottime.tv_sec == 0) {
654			struct bintime bt;
655
656			kread(timenl, X_TIMEBASEBIN, &bt, sizeof(bt));
657			bintime2timespec(&bt, &boottime);
658		}
659		if (kreadc(timenl, X_TIME_SECOND, &nowsec, sizeof(nowsec))) {
660			/*
661			 * XXX this assignment dance can be removed once
662			 * timeval tv_sec is SUS mandated time_t
663			 */
664			now.tv_sec = nowsec;
665			now.tv_nsec = 0;
666		} else {
667			kread(timenl, X_TIME, &now, sizeof(now));
668		}
669	}
670	uptime = now.tv_sec - boottime.tv_sec;
671	if (uptime <= 0 || uptime > 60*60*24*365*10)
672		errx(1, "time makes no sense; namelist must be wrong.");
673	return (uptime);
674}
675
676int	hz, hdrcnt;
677
678void
679print_total_hdr(void)
680{
681
682	(void)printf("procs         memory\n");
683	(void)printf("ru dw pw sl");
684	(void)printf("   total-v  active-v  active-r");
685	(void)printf(" vm-sh avm-sh rm-sh arm-sh free\n");
686	hdrcnt = winlines - 2;
687}
688
689void
690dovmtotal(struct timespec *interval, int reps)
691{
692	struct vmtotal total;
693	size_t size;
694
695	(void)signal(SIGCONT, needhdr);
696
697	for (hdrcnt = 1;;) {
698		if (!--hdrcnt)
699			print_total_hdr();
700		if (memf != NULL) {
701			warnx("Unable to get vmtotals from crash dump.");
702			(void)memset(&total, 0, sizeof(total));
703		} else {
704			size = sizeof(total);
705			if (sysctl(vmmeter_mib, __arraycount(vmmeter_mib),
706			    &total, &size, NULL, 0) == -1) {
707				warn("Can't get vmtotals");
708				(void)memset(&total, 0, sizeof(total));
709			}
710		}
711		(void)printf("%2d ", total.t_rq);
712		(void)printf("%2d ", total.t_dw);
713		(void)printf("%2d ", total.t_pw);
714		(void)printf("%2d ", total.t_sl);
715
716		(void)printf("%9d ", total.t_vm);
717		(void)printf("%9d ", total.t_avm);
718		(void)printf("%9d ", total.t_arm);
719		(void)printf("%5d ", total.t_vmshr);
720		(void)printf("%6d ", total.t_avmshr);
721		(void)printf("%5d ", total.t_rmshr);
722		(void)printf("%6d ", total.t_armshr);
723		(void)printf("%5d",  total.t_free);
724
725		(void)putchar('\n');
726
727		(void)fflush(stdout);
728		if (reps >= 0 && --reps <= 0)
729			break;
730
731		(void)nanosleep(interval, NULL);
732	}
733}
734
735void
736dovmstat(struct timespec *interval, int reps)
737{
738	struct vmtotal total;
739	time_t uptime, halfuptime;
740	size_t size;
741	int pagesize = getpagesize();
742	int ovflw;
743
744	uptime = getuptime();
745	halfuptime = uptime / 2;
746	(void)signal(SIGCONT, needhdr);
747
748	if (memf != NULL) {
749		if (namelist[X_STATHZ].n_type != 0 && namelist[X_STATHZ].n_value != 0)
750			kread(namelist, X_STATHZ, &hz, sizeof(hz));
751		if (!hz)
752			kread(namelist, X_HZ, &hz, sizeof(hz));
753	} else {
754		struct clockinfo clockinfo;
755		size = sizeof(clockinfo);
756		if (sysctl(clockrate_mib, 2, &clockinfo, &size, NULL, 0) == -1)
757			err(1, "sysctl kern.clockrate failed");
758		hz = clockinfo.stathz;
759		if (!hz)
760			hz = clockinfo.hz;
761	}
762
763	for (hdrcnt = 1;;) {
764		if (!--hdrcnt)
765			printhdr();
766		/* Read new disk statistics */
767		cpureadstats();
768		drvreadstats();
769		tkreadstats();
770		if (memf != NULL) {
771			struct uvmexp uvmexp_kernel;
772			/*
773			 * XXX Can't do this if we're reading a crash
774			 * XXX dump because they're lazily-calculated.
775			 */
776			warnx("Unable to get vmtotals from crash dump.");
777			(void)memset(&total, 0, sizeof(total));
778			kread(namelist, X_UVMEXP, &uvmexp_kernel, sizeof(uvmexp_kernel));
779#define COPY(field) uvmexp.field = uvmexp_kernel.field
780			COPY(pdreact);
781			COPY(pageins);
782			COPY(pgswapout);
783			COPY(pdfreed);
784			COPY(pdscans);
785#undef COPY
786		} else {
787			size = sizeof(total);
788			if (sysctl(vmmeter_mib, __arraycount(vmmeter_mib),
789			    &total, &size, NULL, 0) == -1) {
790				warn("Can't get vmtotals");
791				(void)memset(&total, 0, sizeof(total));
792			}
793			size = sizeof(uvmexp);
794			if (sysctl(uvmexp2_mib, __arraycount(uvmexp2_mib), &uvmexp,
795			    &size, NULL, 0) == -1)
796				warn("sysctl vm.uvmexp2 failed");
797		}
798		cpucounters(&cpucounter);
799		ovflw = 0;
800		PRWORD(ovflw, " %*d", 2, 1, total.t_rq - 1);
801		PRWORD(ovflw, " %*d", 2, 1, total.t_dw + total.t_pw);
802#define	pgtok(a) (long)((a) * ((uint32_t)pagesize >> 10))
803#define	rate(x)	(u_long)(((x) + halfuptime) / uptime)	/* round */
804		PRWORD(ovflw, " %*ld", 9, 1, pgtok(total.t_avm));
805		PRWORD(ovflw, " %*ld", 7, 1, pgtok(total.t_free));
806		PRWORD(ovflw, " %*ld", 5, 1,
807		    rate(cpucounter.nfault - ocpucounter.nfault));
808		PRWORD(ovflw, " %*ld", 4, 1,
809		    rate(uvmexp.pdreact - ouvmexp.pdreact));
810		PRWORD(ovflw, " %*ld", 4, 1,
811		    rate(uvmexp.pageins - ouvmexp.pageins));
812		PRWORD(ovflw, " %*ld", 5, 1,
813		    rate(uvmexp.pgswapout - ouvmexp.pgswapout));
814		PRWORD(ovflw, " %*ld", 5, 1,
815		    rate(uvmexp.pdfreed - ouvmexp.pdfreed));
816		PRWORD(ovflw, " %*ld", 6, 2,
817		    rate(uvmexp.pdscans - ouvmexp.pdscans));
818		drvstats(&ovflw);
819		PRWORD(ovflw, " %*ld", 5, 1,
820		    rate(cpucounter.nintr - ocpucounter.nintr));
821		PRWORD(ovflw, " %*ld", 5, 1,
822		    rate(cpucounter.nsyscall - ocpucounter.nsyscall));
823		PRWORD(ovflw, " %*ld", 4, 1,
824		    rate(cpucounter.nswtch - ocpucounter.nswtch));
825		cpustats(&ovflw);
826		(void)putchar('\n');
827		(void)fflush(stdout);
828		if (reps >= 0 && --reps <= 0)
829			break;
830		ouvmexp = uvmexp;
831		ocpucounter = cpucounter;
832		uptime = interval->tv_sec;
833		/*
834		 * We round upward to avoid losing low-frequency events
835		 * (i.e., >= 1 per interval but < 1 per second).
836		 */
837		halfuptime = uptime == 1 ? 0 : (uptime + 1) / 2;
838		(void)nanosleep(interval, NULL);
839	}
840}
841
842void
843printhdr(void)
844{
845	size_t i;
846
847	(void)printf(" procs    memory      page%*s", 23, "");
848	if (ndrives > 0)
849		(void)printf("%s %*sfaults      cpu\n",
850		    ((ndrives > 1) ? "disks" : "disk"),
851		    ((ndrives > 1) ? ndrives * 3 - 4 : 0), "");
852	else
853		(void)printf("%*s  faults   cpu\n",
854		    ndrives * 3, "");
855
856	(void)printf(" r b      avm    fre  flt  re  pi   po   fr   sr ");
857	for (i = 0; i < ndrive; i++)
858		if (drv_select[i])
859			(void)printf("%c%c ", dr_name[i][0],
860			    dr_name[i][strlen(dr_name[i]) - 1]);
861	(void)printf("  in   sy  cs us sy id\n");
862	hdrcnt = winlines - 2;
863}
864
865/*
866 * Force a header to be prepended to the next output.
867 */
868void
869/*ARGSUSED*/
870needhdr(int dummy)
871{
872
873	hdrcnt = 1;
874}
875
876long
877pct(u_long top, u_long bot)
878{
879	long ans;
880
881	if (bot == 0)
882		return (0);
883	ans = (long)((quad_t)top * 100 / bot);
884	return (ans);
885}
886
887#define	PCT(top, bot) (int)pct((u_long)(top), (u_long)(bot))
888
889void
890dosum(void)
891{
892	struct nchstats nch_stats;
893	uint64_t nchtotal;
894	size_t ssize;
895	int active_kernel;
896	struct cpu_counter cc;
897
898	/*
899	 * The "active" and "inactive" variables
900	 * are now estimated by the kernel and sadly
901	 * can not easily be dug out of a crash dump.
902	 */
903	ssize = sizeof(uvmexp);
904	memset(&uvmexp, 0, ssize);
905	active_kernel = (memf == NULL);
906	if (active_kernel) {
907		/* only on active kernel */
908		if (sysctl(uvmexp2_mib, __arraycount(uvmexp2_mib), &uvmexp,
909		    &ssize, NULL, 0) == -1)
910			warn("sysctl vm.uvmexp2 failed");
911	} else {
912		struct uvmexp uvmexp_kernel;
913		struct pool pool, *pp = &pool;
914		struct pool_allocator pa;
915		TAILQ_HEAD(,pool) pool_head;
916		void *addr;
917		uint64_t bytes;
918
919		kread(namelist, X_UVMEXP, &uvmexp_kernel, sizeof(uvmexp_kernel));
920#define COPY(field) uvmexp.field = uvmexp_kernel.field
921		COPY(pagesize);
922		COPY(ncolors);
923		COPY(npages);
924		COPY(free);
925		COPY(paging);
926		COPY(wired);
927		COPY(zeropages);
928		COPY(reserve_pagedaemon);
929		COPY(reserve_kernel);
930		COPY(anonpages);
931		COPY(filepages);
932		COPY(execpages);
933		COPY(freemin);
934		COPY(freetarg);
935		COPY(wiredmax);
936		COPY(nswapdev);
937		COPY(swpages);
938		COPY(swpginuse);
939		COPY(nswget);
940		COPY(pageins);
941		COPY(pdpageouts);
942		COPY(pgswapin);
943		COPY(pgswapout);
944		COPY(forks);
945		COPY(forks_ppwait);
946		COPY(forks_sharevm);
947		COPY(pga_zerohit);
948		COPY(pga_zeromiss);
949		COPY(zeroaborts);
950		COPY(colorhit);
951		COPY(colormiss);
952		COPY(cpuhit);
953		COPY(cpumiss);
954		COPY(fltnoram);
955		COPY(fltnoanon);
956		COPY(fltpgwait);
957		COPY(fltpgrele);
958		COPY(fltrelck);
959		COPY(fltrelckok);
960		COPY(fltanget);
961		COPY(fltanretry);
962		COPY(fltamcopy);
963		COPY(fltamcopy);
964		COPY(fltnomap);
965		COPY(fltlget);
966		COPY(fltget);
967		COPY(flt_anon);
968		COPY(flt_acow);
969		COPY(flt_obj);
970		COPY(flt_prcopy);
971		COPY(flt_przero);
972		COPY(pdwoke);
973		COPY(pdrevs);
974		COPY(pdfreed);
975		COPY(pdscans);
976		COPY(pdanscan);
977		COPY(pdobscan);
978		COPY(pdreact);
979		COPY(pdbusy);
980		COPY(pdpending);
981		COPY(pddeact);
982		COPY(bootpages);
983#undef COPY
984		kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
985		addr = TAILQ_FIRST(&pool_head);
986		uvmexp.poolpages = 0;
987		for (; addr != NULL; addr = TAILQ_NEXT(pp, pr_poollist)) {
988			deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
989			deref_kptr(pp->pr_alloc, &pa, sizeof(pa),
990			    "pool allocator trashed");
991			bytes = pp->pr_npages * pa.pa_pagesz;
992			if ((pp->pr_roflags & PR_RECURSIVE) != 0)
993				bytes -= (pp->pr_nout * pp->pr_size);
994			uvmexp.poolpages += bytes / uvmexp.pagesize;
995		}
996	}
997
998
999	(void)printf("%9" PRIu64 " bytes per page\n", uvmexp.pagesize);
1000
1001	(void)printf("%9" PRIu64 " page color%s\n",
1002	    uvmexp.ncolors, uvmexp.ncolors == 1 ? "" : "s");
1003
1004	(void)printf("%9" PRIu64 " pages managed\n", uvmexp.npages);
1005	(void)printf("%9" PRIu64 " pages free\n", uvmexp.free);
1006	if (active_kernel) {
1007		(void)printf("%9" PRIu64 " pages active\n", uvmexp.active);
1008		(void)printf("%9" PRIu64 " pages inactive\n", uvmexp.inactive);
1009	}
1010	(void)printf("%9" PRIu64 " pages paging\n", uvmexp.paging);
1011	(void)printf("%9" PRIu64 " pages wired\n", uvmexp.wired);
1012	(void)printf("%9" PRIu64 " zero pages\n", uvmexp.zeropages);
1013	(void)printf("%9" PRIu64 " reserve pagedaemon pages\n",
1014	    uvmexp.reserve_pagedaemon);
1015	(void)printf("%9" PRIu64 " reserve kernel pages\n", uvmexp.reserve_kernel);
1016	(void)printf("%9" PRIu64 " boot kernel pages\n", uvmexp.bootpages);
1017	(void)printf("%9" PRIu64 " kernel pool pages\n", uvmexp.poolpages);
1018	(void)printf("%9" PRIu64 " anonymous pages\n", uvmexp.anonpages);
1019	(void)printf("%9" PRIu64 " cached file pages\n", uvmexp.filepages);
1020	(void)printf("%9" PRIu64 " cached executable pages\n", uvmexp.execpages);
1021
1022	(void)printf("%9" PRIu64 " minimum free pages\n", uvmexp.freemin);
1023	(void)printf("%9" PRIu64 " target free pages\n", uvmexp.freetarg);
1024	(void)printf("%9" PRIu64 " maximum wired pages\n", uvmexp.wiredmax);
1025
1026	(void)printf("%9" PRIu64 " swap devices\n", uvmexp.nswapdev);
1027	(void)printf("%9" PRIu64 " swap pages\n", uvmexp.swpages);
1028	(void)printf("%9" PRIu64 " swap pages in use\n", uvmexp.swpginuse);
1029	(void)printf("%9" PRIu64 " swap allocations\n", uvmexp.nswget);
1030
1031	cpucounters(&cc);
1032
1033	(void)printf("%9" PRIu64 " total faults taken\n", cc.nfault);
1034	(void)printf("%9" PRIu64 " traps\n", cc.ntrap);
1035	(void)printf("%9" PRIu64 " device interrupts\n", cc.nintr);
1036	(void)printf("%9" PRIu64 " CPU context switches\n", cc.nswtch);
1037	(void)printf("%9" PRIu64 " software interrupts\n", cc.nsoft);
1038	(void)printf("%9" PRIu64 " system calls\n", cc.nsyscall);
1039	(void)printf("%9" PRIu64 " pagein requests\n", uvmexp.pageins);
1040	(void)printf("%9" PRIu64 " pageout requests\n", uvmexp.pdpageouts);
1041	(void)printf("%9" PRIu64 " pages swapped in\n", uvmexp.pgswapin);
1042	(void)printf("%9" PRIu64 " pages swapped out\n", uvmexp.pgswapout);
1043	(void)printf("%9" PRIu64 " forks total\n", uvmexp.forks);
1044	(void)printf("%9" PRIu64 " forks blocked parent\n", uvmexp.forks_ppwait);
1045	(void)printf("%9" PRIu64 " forks shared address space with parent\n",
1046	    uvmexp.forks_sharevm);
1047	(void)printf("%9" PRIu64 " pagealloc zero wanted and avail\n",
1048	    uvmexp.pga_zerohit);
1049	(void)printf("%9" PRIu64 " pagealloc zero wanted and not avail\n",
1050	    uvmexp.pga_zeromiss);
1051	(void)printf("%9" PRIu64 " aborts of idle page zeroing\n",
1052	    uvmexp.zeroaborts);
1053	(void)printf("%9" PRIu64 " pagealloc desired color avail\n",
1054	    uvmexp.colorhit);
1055	(void)printf("%9" PRIu64 " pagealloc desired color not avail\n",
1056	    uvmexp.colormiss);
1057	(void)printf("%9" PRIu64 " pagealloc local cpu avail\n",
1058	    uvmexp.cpuhit);
1059	(void)printf("%9" PRIu64 " pagealloc local cpu not avail\n",
1060	    uvmexp.cpumiss);
1061
1062	(void)printf("%9" PRIu64 " faults with no memory\n", uvmexp.fltnoram);
1063	(void)printf("%9" PRIu64 " faults with no anons\n", uvmexp.fltnoanon);
1064	(void)printf("%9" PRIu64 " faults had to wait on pages\n", uvmexp.fltpgwait);
1065	(void)printf("%9" PRIu64 " faults found released page\n", uvmexp.fltpgrele);
1066	(void)printf("%9" PRIu64 " faults relock (%" PRIu64 " ok)\n", uvmexp.fltrelck,
1067	    uvmexp.fltrelckok);
1068	(void)printf("%9" PRIu64 " anon page faults\n", uvmexp.fltanget);
1069	(void)printf("%9" PRIu64 " anon retry faults\n", uvmexp.fltanretry);
1070	(void)printf("%9" PRIu64 " amap copy faults\n", uvmexp.fltamcopy);
1071	(void)printf("%9" PRIu64 " neighbour anon page faults\n", uvmexp.fltnamap);
1072	(void)printf("%9" PRIu64 " neighbour object page faults\n", uvmexp.fltnomap);
1073	(void)printf("%9" PRIu64 " locked pager get faults\n", uvmexp.fltlget);
1074	(void)printf("%9" PRIu64 " unlocked pager get faults\n", uvmexp.fltget);
1075	(void)printf("%9" PRIu64 " anon faults\n", uvmexp.flt_anon);
1076	(void)printf("%9" PRIu64 " anon copy on write faults\n", uvmexp.flt_acow);
1077	(void)printf("%9" PRIu64 " object faults\n", uvmexp.flt_obj);
1078	(void)printf("%9" PRIu64 " promote copy faults\n", uvmexp.flt_prcopy);
1079	(void)printf("%9" PRIu64 " promote zero fill faults\n", uvmexp.flt_przero);
1080
1081	(void)printf("%9" PRIu64 " times daemon wokeup\n",uvmexp.pdwoke);
1082	(void)printf("%9" PRIu64 " revolutions of the clock hand\n", uvmexp.pdrevs);
1083	(void)printf("%9" PRIu64 " pages freed by daemon\n", uvmexp.pdfreed);
1084	(void)printf("%9" PRIu64 " pages scanned by daemon\n", uvmexp.pdscans);
1085	(void)printf("%9" PRIu64 " anonymous pages scanned by daemon\n",
1086	    uvmexp.pdanscan);
1087	(void)printf("%9" PRIu64 " object pages scanned by daemon\n", uvmexp.pdobscan);
1088	(void)printf("%9" PRIu64 " pages reactivated\n", uvmexp.pdreact);
1089	(void)printf("%9" PRIu64 " pages found busy by daemon\n", uvmexp.pdbusy);
1090	(void)printf("%9" PRIu64 " total pending pageouts\n", uvmexp.pdpending);
1091	(void)printf("%9" PRIu64 " pages deactivated\n", uvmexp.pddeact);
1092
1093	if (active_kernel) {
1094		ssize = sizeof(nch_stats);
1095		if (sysctlbyname("vfs.namecache_stats", &nch_stats, &ssize,
1096		    NULL, 0)) {
1097			warn("vfs.namecache_stats failed");
1098			memset(&nch_stats, 0, sizeof(nch_stats));
1099		}
1100	} else {
1101		kread(namelist, X_NCHSTATS, &nch_stats, sizeof(nch_stats));
1102	}
1103
1104	nchtotal = nch_stats.ncs_goodhits + nch_stats.ncs_neghits +
1105	    nch_stats.ncs_badhits + nch_stats.ncs_falsehits +
1106	    nch_stats.ncs_miss + nch_stats.ncs_long;
1107	(void)printf("%9" PRIu64 " total name lookups\n", nchtotal);
1108	(void)printf("%9" PRIu64 " good hits\n", nch_stats.ncs_goodhits);
1109	(void)printf("%9" PRIu64 " negative hits\n", nch_stats.ncs_neghits);
1110	(void)printf("%9" PRIu64 " bad hits\n", nch_stats.ncs_badhits);
1111	(void)printf("%9" PRIu64 " false hits\n", nch_stats.ncs_falsehits);
1112	(void)printf("%9" PRIu64 " miss\n", nch_stats.ncs_miss);
1113	(void)printf("%9" PRIu64 " too long\n", nch_stats.ncs_long);
1114	(void)printf("%9" PRIu64 " pass2 hits\n", nch_stats.ncs_pass2);
1115	(void)printf("%9" PRIu64 " 2passes\n", nch_stats.ncs_2passes);
1116	(void)printf("%9" PRIu64 " reverse hits\n", nch_stats.ncs_revhits);
1117	(void)printf(
1118	    "%9s cache hits (%d%% pos + %d%% neg) system %d%% per-process\n",
1119	    "", PCT(nch_stats.ncs_goodhits, nchtotal),
1120	    PCT(nch_stats.ncs_neghits, nchtotal),
1121	    PCT(nch_stats.ncs_pass2, nchtotal));
1122	(void)printf("%9s deletions %d%%, falsehits %d%%, toolong %d%%\n", "",
1123	    PCT(nch_stats.ncs_badhits, nchtotal),
1124	    PCT(nch_stats.ncs_falsehits, nchtotal),
1125	    PCT(nch_stats.ncs_long, nchtotal));
1126}
1127
1128void
1129doforkst(void)
1130{
1131	if (memf != NULL) {
1132		struct uvmexp uvmexp_kernel;
1133		kread(namelist, X_UVMEXP, &uvmexp_kernel, sizeof(uvmexp_kernel));
1134#define COPY(field) uvmexp.field = uvmexp_kernel.field
1135		COPY(forks);
1136		COPY(forks_ppwait);
1137		COPY(forks_sharevm);
1138#undef COPY
1139	} else {
1140		size_t size = sizeof(uvmexp);
1141		if (sysctl(uvmexp2_mib, __arraycount(uvmexp2_mib), &uvmexp,
1142		    &size, NULL, 0) == -1)
1143			warn("sysctl vm.uvmexp2 failed");
1144	}
1145
1146	(void)printf("%" PRIu64 " forks total\n", uvmexp.forks);
1147	(void)printf("%" PRIu64 " forks blocked parent\n", uvmexp.forks_ppwait);
1148	(void)printf("%" PRIu64 " forks shared address space with parent\n",
1149	    uvmexp.forks_sharevm);
1150}
1151
1152void
1153drvstats(int *ovflwp)
1154{
1155	size_t dn;
1156	double dtime;
1157	int ovflw = *ovflwp;
1158
1159	/* Calculate disk stat deltas. */
1160	cpuswap();
1161	drvswap();
1162	tkswap();
1163
1164	for (dn = 0; dn < ndrive; ++dn) {
1165		/* elapsed time for disk stats */
1166		dtime = cur.cp_etime;
1167		if (cur.timestamp[dn].tv_sec || cur.timestamp[dn].tv_usec) {
1168			dtime = (double)cur.timestamp[dn].tv_sec +
1169				((double)cur.timestamp[dn].tv_usec / (double)1000000);
1170		}
1171
1172		if (!drv_select[dn])
1173	 		continue;
1174		PRWORD(ovflw, " %*.0f", 3, 1,
1175		    (cur.rxfer[dn] + cur.wxfer[dn]) / dtime);
1176	}
1177	*ovflwp = ovflw;
1178}
1179
1180void
1181cpucounters(struct cpu_counter *cc)
1182{
1183	static struct cpu_info **cpu_infos;
1184	static int initialised;
1185	struct cpu_info **slot;
1186
1187	if (memf == NULL) {
1188		cc->nintr = uvmexp.intrs;
1189		cc->nsyscall = uvmexp.syscalls;
1190		cc->nswtch = uvmexp.swtch;
1191		cc->nfault = uvmexp.faults;
1192		cc->ntrap = uvmexp.traps;
1193		cc->nsoft = uvmexp.softs;
1194		return;
1195	}
1196
1197	if (!initialised) {
1198		kread(namelist, X_CPU_INFOS, &cpu_infos, sizeof(cpu_infos));
1199		initialised = 1;
1200	}
1201
1202	slot = cpu_infos;
1203
1204	memset(cc, 0, sizeof(*cc));
1205
1206	for (;;) {
1207		struct cpu_info tci, *ci = NULL;
1208
1209		deref_kptr(slot++, &ci, sizeof(ci), "CPU array trashed");
1210		if (!ci) {
1211			break;
1212		}
1213
1214		if ((size_t)kvm_read(kd, (u_long)ci, &tci, sizeof(tci))
1215		    != sizeof(tci)) {
1216			warnx("Can't read cpu info from %p (%s)",
1217			    ci, kvm_geterr(kd));
1218			memset(cc, 0, sizeof(*cc));
1219			return;
1220		}
1221		cc->nintr += tci.ci_data.cpu_nintr;
1222		cc->nsyscall += tci.ci_data.cpu_nsyscall;
1223		cc->nswtch = tci.ci_data.cpu_nswtch;
1224		cc->nfault = tci.ci_data.cpu_nfault;
1225		cc->ntrap = tci.ci_data.cpu_ntrap;
1226		cc->nsoft = tci.ci_data.cpu_nsoft;
1227	}
1228}
1229
1230void
1231cpustats(int *ovflwp)
1232{
1233	int state;
1234	double pcnt, total;
1235	double stat_us, stat_sy, stat_id;
1236	int ovflw = *ovflwp;
1237
1238	total = 0;
1239	for (state = 0; state < CPUSTATES; ++state)
1240		total += cur.cp_time[state];
1241	if (total)
1242		pcnt = 100 / total;
1243	else
1244		pcnt = 0;
1245	stat_us = (cur.cp_time[CP_USER] + cur.cp_time[CP_NICE]) * pcnt;
1246	stat_sy = (cur.cp_time[CP_SYS] + cur.cp_time[CP_INTR]) * pcnt;
1247	stat_id = cur.cp_time[CP_IDLE] * pcnt;
1248	PRWORD(ovflw, " %*.0f", ((stat_sy >= 100) ? 2 : 3), 1, stat_us);
1249	PRWORD(ovflw, " %*.0f", ((stat_us >= 100 || stat_id >= 100) ? 2 : 3), 1,
1250	    stat_sy);
1251	PRWORD(ovflw, " %*.0f", 3, 1, stat_id);
1252	*ovflwp = ovflw;
1253}
1254
1255void
1256dointr(int verbose)
1257{
1258	unsigned long *intrcnt, *ointrcnt;
1259	unsigned long long inttotal, uptime;
1260	int nintr, inamlen;
1261	char *intrname, *ointrname;
1262
1263	inttotal = 0;
1264	uptime = getuptime();
1265	nintr = intrnl[X_EINTRCNT].n_value - intrnl[X_INTRCNT].n_value;
1266	inamlen = intrnl[X_EINTRNAMES].n_value - intrnl[X_INTRNAMES].n_value;
1267	if (nintr != 0 && inamlen != 0) {
1268		(void)printf("%-34s %16s %8s\n", "interrupt", "total", "rate");
1269
1270		ointrcnt = intrcnt = malloc((size_t)nintr);
1271		ointrname = intrname = malloc((size_t)inamlen);
1272		if (intrcnt == NULL || intrname == NULL)
1273			errx(1, "%s", "");
1274		kread(intrnl, X_INTRCNT, intrcnt, (size_t)nintr);
1275		kread(intrnl, X_INTRNAMES, intrname, (size_t)inamlen);
1276		nintr /= sizeof(long);
1277		while (--nintr >= 0) {
1278			if (*intrcnt || verbose)
1279				(void)printf("%-34s %16llu %8llu\n", intrname,
1280					     (unsigned long long)*intrcnt,
1281					     (unsigned long long)
1282					     (*intrcnt / uptime));
1283			intrname += strlen(intrname) + 1;
1284			inttotal += *intrcnt++;
1285		}
1286		free(ointrcnt);
1287		free(ointrname);
1288	}
1289
1290	doevcnt(verbose, EVCNT_TYPE_INTR);
1291}
1292
1293void
1294doevcnt(int verbose, int type)
1295{
1296	static const char * const evtypes [] = { "misc", "intr", "trap" };
1297	uint64_t counttotal, uptime;
1298	struct evcntlist allevents;
1299	struct evcnt evcnt, *evptr;
1300	size_t evlen_max, total_max, rate_max;
1301	char evgroup[EVCNT_STRING_MAX], evname[EVCNT_STRING_MAX];
1302
1303	counttotal = 0;
1304	uptime = getuptime();
1305
1306	if (memf == NULL) do {
1307		const int mib[4] = { CTL_KERN, KERN_EVCNT, type,
1308		    verbose ? KERN_EVCNT_COUNT_ANY : KERN_EVCNT_COUNT_NONZERO };
1309		size_t buflen0, buflen = 0;
1310		void *buf0, *buf = NULL;
1311		const struct evcnt_sysctl *evs, *last_evs;
1312		for (;;) {
1313			size_t newlen;
1314			int error;
1315			if (buflen)
1316				buf = malloc(buflen);
1317			error = sysctl(mib, __arraycount(mib),
1318			    buf, &newlen, NULL, 0);
1319			if (error) {
1320				err(1, "kern.evcnt");
1321				if (buf)
1322					free(buf);
1323				return;
1324			}
1325			if (newlen <= buflen) {
1326				buflen = newlen;
1327				break;
1328			}
1329			if (buf)
1330				free(buf);
1331			buflen = newlen;
1332		}
1333		buflen0 = buflen;
1334		evs = buf0 = buf;
1335		last_evs = (void *)((char *)buf + buflen);
1336		buflen /= sizeof(uint64_t);
1337		/* calc columns */
1338		evlen_max = 0;
1339		total_max = sizeof("total") - 1;
1340		rate_max = sizeof("rate") - 1;
1341		while (evs < last_evs
1342		    && buflen >= sizeof(*evs)/sizeof(uint64_t)
1343		    && buflen >= evs->ev_len) {
1344			char cbuf[64];
1345			size_t len;
1346			len = strlen(evs->ev_strings + evs->ev_grouplen + 1);
1347			len += evs->ev_grouplen + 1;
1348			if (evlen_max < len)
1349				evlen_max= len;
1350			len = snprintf(cbuf, sizeof(cbuf), "%"PRIu64,
1351			    evs->ev_count);
1352			if (total_max < len)
1353				total_max = len;
1354			len = snprintf(cbuf, sizeof(cbuf), "%"PRIu64,
1355			    evs->ev_count / uptime);
1356			if (rate_max < len)
1357				rate_max = len;
1358			buflen -= evs->ev_len;
1359			evs = (const void *)
1360			    ((const uint64_t *)evs + evs->ev_len);
1361		}
1362
1363		(void)printf(type == EVCNT_TYPE_ANY ?
1364		    "%-*s  %*s %*s %s\n" :
1365		    "%-*s  %*s %*s\n",
1366		    (int)evlen_max, "interrupt",
1367		    (int)total_max, "total",
1368		    (int)rate_max, "rate",
1369		    "type");
1370
1371		buflen = buflen0;
1372		evs = buf0;
1373		last_evs = (void *)((char *)buf + buflen);
1374		buflen /= sizeof(uint64_t);
1375		while (evs < last_evs
1376		    && buflen >= sizeof(*evs)/sizeof(uint64_t)
1377		    && buflen >= evs->ev_len) {
1378			(void)printf(type == EVCNT_TYPE_ANY ?
1379			    "%s %s%*s  %*"PRIu64" %*"PRIu64" %s\n" :
1380			    "%s %s%*s  %*"PRIu64" %*"PRIu64"\n",
1381			    evs->ev_strings,
1382			    evs->ev_strings + evs->ev_grouplen + 1,
1383			    (int)evlen_max - (evs->ev_grouplen + 1
1384			    + evs->ev_namelen), "",
1385			    (int)total_max, evs->ev_count,
1386			    (int)rate_max, evs->ev_count / uptime,
1387			    (evs->ev_type < __arraycount(evtypes) ?
1388			    evtypes[evs->ev_type] : "?"));
1389			buflen -= evs->ev_len;
1390			counttotal += evs->ev_count;
1391			evs = (const void *)
1392			    ((const uint64_t *)evs + evs->ev_len);
1393		}
1394		free(buf);
1395		if (type != EVCNT_TYPE_ANY)
1396			(void)printf("%-*s  %*"PRIu64" %*"PRIu64"\n",
1397			    (int)evlen_max, "Total",
1398			    (int)total_max, counttotal,
1399			    (int)rate_max, counttotal / uptime);
1400		return;
1401	} while (/*CONSTCOND*/ 0);
1402
1403	if (type == EVCNT_TYPE_ANY)
1404		(void)printf("%-34s %16s %8s %s\n", "event", "total", "rate",
1405		    "type");
1406
1407	kread(namelist, X_ALLEVENTS, &allevents, sizeof allevents);
1408	evptr = TAILQ_FIRST(&allevents);
1409	while (evptr) {
1410		deref_kptr(evptr, &evcnt, sizeof(evcnt), "event chain trashed");
1411
1412		evptr = TAILQ_NEXT(&evcnt, ev_list);
1413		if (evcnt.ev_count == 0 && !verbose)
1414			continue;
1415		if (type != EVCNT_TYPE_ANY && evcnt.ev_type != type)
1416			continue;
1417
1418		deref_kptr(evcnt.ev_group, evgroup,
1419		    (size_t)evcnt.ev_grouplen + 1, "event chain trashed");
1420		deref_kptr(evcnt.ev_name, evname,
1421		    (size_t)evcnt.ev_namelen + 1, "event chain trashed");
1422
1423		(void)printf(type == EVCNT_TYPE_ANY ?
1424		    "%s %s%*s %16"PRIu64" %8"PRIu64" %s\n" :
1425		    "%s %s%*s %16"PRIu64" %8"PRIu64"\n",
1426		    evgroup, evname,
1427		    34 - (evcnt.ev_grouplen + 1 + evcnt.ev_namelen), "",
1428		    evcnt.ev_count,
1429		    (evcnt.ev_count / uptime),
1430		    (evcnt.ev_type < __arraycount(evtypes) ?
1431			evtypes[evcnt.ev_type] : "?"));
1432
1433		counttotal += evcnt.ev_count;
1434	}
1435	if (type != EVCNT_TYPE_ANY)
1436		(void)printf("%-34s %16"PRIu64" %8"PRIu64"\n",
1437		    "Total", counttotal, counttotal / uptime);
1438}
1439
1440static void
1441dopool_sysctl(int verbose, int wide)
1442{
1443	uint64_t total, inuse, this_total, this_inuse;
1444	struct {
1445		uint64_t pt_nget;
1446		uint64_t pt_nfail;
1447		uint64_t pt_nput;
1448		uint64_t pt_nout;
1449		uint64_t pt_nitems;
1450		uint64_t pt_npagealloc;
1451		uint64_t pt_npagefree;
1452		uint64_t pt_npages;
1453	} pool_totals;
1454	size_t i, len;
1455	int name_len, ovflw;
1456	struct pool_sysctl *pp, *data;
1457	char maxp[32];
1458
1459	data = asysctlbyname("kern.pool", &len);
1460	if (data == NULL)
1461		err(1, "failed to read kern.pool");
1462
1463	memset(&pool_totals, 0, sizeof pool_totals);
1464	total = inuse = 0;
1465	len /= sizeof(*data);
1466
1467	(void)printf("Memory resource pool statistics\n");
1468	(void)printf(
1469	    "%-*s%*s%*s%*s%*s%s%s%*s%*s%*s%s%*s%6s%*s%5s%s%s\n",
1470	    wide ? 16 : 11, "Name",
1471	    wide ? 7 : 5, "Size",
1472	    wide ? 12 : 9, "Requests",
1473	    wide ? 8 : 5, "Fail",
1474	    wide ? 12 : 9, "Releases",
1475	    wide ? "    InUse" : "",
1476	    wide ? "    Avail" : "",
1477	    wide ? 11 : 6, "Pgreq",
1478	    wide ? 11 : 6, "Pgrel",
1479	    wide ? 8 : 6, "Npage",
1480	    wide ? " PageSz" : "",
1481	    wide ? 7 : 6, "Hiwat",
1482	    "Minpg",
1483	    wide ? 7 : 6, "Maxpg",
1484	    "Idle",
1485	    wide ? "  Flags" : "",
1486	    wide ? "   Util" : "");
1487
1488	name_len = MIN((int)sizeof(pp->pr_wchan), wide ? 16 : 11);
1489	for (i = 0; i < len; ++i) {
1490		pp = &data[i];
1491		if (pp->pr_nget == 0 && !verbose)
1492			continue;
1493		if (pp->pr_maxpages == UINT_MAX)
1494			(void)snprintf(maxp, sizeof(maxp), "inf");
1495		else
1496			(void)snprintf(maxp, sizeof(maxp), "%" PRIu64,
1497			    pp->pr_maxpages);
1498		ovflw = 0;
1499		PRWORD(ovflw, "%-*s", name_len, 0, pp->pr_wchan);
1500		PRWORD(ovflw, " %*" PRIu64, wide ? 7 : 5, 1, pp->pr_size);
1501		PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pp->pr_nget);
1502		pool_totals.pt_nget += pp->pr_nget;
1503		PRWORD(ovflw, " %*" PRIu64, wide ? 8 : 5, 1, pp->pr_nfail);
1504		pool_totals.pt_nfail += pp->pr_nfail;
1505		PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pp->pr_nput);
1506		pool_totals.pt_nput += pp->pr_nput;
1507		if (wide) {
1508			PRWORD(ovflw, " %*" PRIu64, 9, 1, pp->pr_nout);
1509			pool_totals.pt_nout += pp->pr_nout;
1510			PRWORD(ovflw, " %*" PRIu64, 9, 1, pp->pr_nitems);
1511			pool_totals.pt_nitems += pp->pr_nitems;
1512		}
1513		PRWORD(ovflw, " %*" PRIu64, wide ? 11 : 6, 1, pp->pr_npagealloc);
1514		pool_totals.pt_npagealloc += pp->pr_npagealloc;
1515		PRWORD(ovflw, " %*" PRIu64, wide ? 11 : 6, 1, pp->pr_npagefree);
1516		pool_totals.pt_npagefree += pp->pr_npagefree;
1517		PRWORD(ovflw, " %*" PRIu64, wide ? 8 : 6, 1, pp->pr_npages);
1518		pool_totals.pt_npages += pp->pr_npages;
1519		if (wide)
1520			PRWORD(ovflw, " %*" PRIu64, 7, 1, pp->pr_pagesize);
1521		PRWORD(ovflw, " %*" PRIu64, wide ? 7 : 6, 1, pp->pr_hiwat);
1522		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_minpages);
1523		PRWORD(ovflw, " %*s", wide ? 7 : 6, 1, maxp);
1524		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_nidle);
1525		if (wide)
1526			PRWORD(ovflw, " 0x%0*" PRIx64, 5, 1,
1527			    pp->pr_flags);
1528
1529		this_inuse = pp->pr_nout * pp->pr_size;
1530		this_total = pp->pr_npages * pp->pr_pagesize;
1531		if (pp->pr_flags & PR_RECURSIVE) {
1532			/*
1533			 * Don't count in-use memory, since it's part
1534			 * of another pool and will be accounted for
1535			 * there.
1536			 */
1537			total += (this_total - this_inuse);
1538		} else {
1539			inuse += this_inuse;
1540			total += this_total;
1541		}
1542		if (wide) {
1543			if (this_total == 0)
1544				(void)printf("   ---");
1545			else
1546				(void)printf(" %5.1f%%",
1547				    (100.0 * this_inuse) / this_total);
1548		}
1549		(void)printf("\n");
1550	}
1551	ovflw = 0;
1552	PRWORD(ovflw, "%-*s", name_len, 0, "Totals");
1553	PRWORD(ovflw, " %*s", wide ? 7 : 5, 1, "");
1554	PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pool_totals.pt_nget);
1555	PRWORD(ovflw, " %*" PRIu64, wide ? 8 : 5, 1, pool_totals.pt_nfail);
1556	PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pool_totals.pt_nput);
1557	if (wide) {
1558		PRWORD(ovflw, " %*" PRIu64, 9, 1, pool_totals.pt_nout);
1559		PRWORD(ovflw, " %*" PRIu64, 9, 1, pool_totals.pt_nitems);
1560	}
1561	PRWORD(ovflw, " %*" PRIu64, wide ? 11 : 6, 1, pool_totals.pt_npagealloc);
1562	PRWORD(ovflw, " %*" PRIu64, wide ? 11 : 6, 1, pool_totals.pt_npagefree);
1563	PRWORD(ovflw, " %*" PRIu64, wide ? 8 : 6, 1, pool_totals.pt_npages);
1564	(void)printf("\n");
1565
1566	inuse /= KILO;
1567	total /= KILO;
1568	(void)printf(
1569	    "\nIn use %" PRIu64 "K, "
1570	    "total allocated %" PRIu64 "K; utilization %.1f%%\n",
1571	    inuse, total, (100.0 * inuse) / total);
1572
1573	free(data);
1574}
1575
1576void
1577dopool(int verbose, int wide)
1578{
1579	int first, ovflw;
1580	void *addr;
1581	long total, inuse, this_total, this_inuse;
1582	struct {
1583		uint64_t pt_nget;
1584		uint64_t pt_nfail;
1585		uint64_t pt_nput;
1586		uint64_t pt_nout;
1587		uint64_t pt_nitems;
1588		uint64_t pt_npagealloc;
1589		uint64_t pt_npagefree;
1590		uint64_t pt_npages;
1591	} pool_totals;
1592	TAILQ_HEAD(,pool) pool_head;
1593	struct pool pool, *pp = &pool;
1594	struct pool_allocator pa;
1595	char maxp[32], name[32];
1596
1597	if (memf == NULL)
1598		return dopool_sysctl(verbose, wide);
1599
1600	memset(&pool_totals, 0, sizeof pool_totals);
1601	kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
1602	addr = TAILQ_FIRST(&pool_head);
1603
1604	total = inuse = 0;
1605
1606	for (first = 1; addr != NULL; addr = TAILQ_NEXT(pp, pr_poollist) ) {
1607		deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
1608		deref_kptr(pp->pr_alloc, &pa, sizeof(pa),
1609		    "pool allocator trashed");
1610		deref_kptr(pp->pr_wchan, name, sizeof(name),
1611		    "pool wait channel trashed");
1612		name[sizeof(name)-1] = '\0';
1613
1614		if (first) {
1615			(void)printf("Memory resource pool statistics\n");
1616			(void)printf(
1617			    "%-*s%*s%*s%*s%*s%s%s%*s%*s%*s%s%*s%6s%*s%5s%s%s\n",
1618			    wide ? 16 : 11, "Name",
1619			    wide ? 7 : 5, "Size",
1620			    wide ? 12 : 9, "Requests",
1621			    wide ? 8 : 5, "Fail",
1622			    wide ? 12 : 9, "Releases",
1623			    wide ? "    InUse" : "",
1624			    wide ? "    Avail" : "",
1625			    wide ? 11 : 6, "Pgreq",
1626			    wide ? 11 : 6, "Pgrel",
1627			    wide ? 8 : 6, "Npage",
1628			    wide ? " PageSz" : "",
1629			    wide ? 7 : 6, "Hiwat",
1630			    "Minpg",
1631			    wide ? 7 : 6, "Maxpg",
1632			    "Idle",
1633			    wide ? "  Flags" : "",
1634			    wide ? "   Util" : "");
1635			first = 0;
1636		}
1637		if (pp->pr_nget == 0 && !verbose)
1638			continue;
1639		if (pp->pr_maxpages == UINT_MAX)
1640			(void)snprintf(maxp, sizeof(maxp), "inf");
1641		else
1642			(void)snprintf(maxp, sizeof(maxp), "%u",
1643			    pp->pr_maxpages);
1644		ovflw = 0;
1645		PRWORD(ovflw, "%-*s", wide ? 16 : 11, 0, name);
1646		PRWORD(ovflw, " %*u", wide ? 7 : 5, 1, pp->pr_size);
1647		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nget);
1648		pool_totals.pt_nget += pp->pr_nget;
1649		PRWORD(ovflw, " %*lu", wide ? 8 : 5, 1, pp->pr_nfail);
1650		pool_totals.pt_nfail += pp->pr_nfail;
1651		PRWORD(ovflw, " %*lu", wide ? 12 : 9, 1, pp->pr_nput);
1652		pool_totals.pt_nput += pp->pr_nput;
1653		if (wide) {
1654			PRWORD(ovflw, " %*u", 9, 1, pp->pr_nout);
1655			pool_totals.pt_nout += pp->pr_nout;
1656			PRWORD(ovflw, " %*u", 9, 1, pp->pr_nitems);
1657			pool_totals.pt_nitems += pp->pr_nitems;
1658		}
1659		PRWORD(ovflw, " %*lu", wide ? 11 : 6, 1, pp->pr_npagealloc);
1660		pool_totals.pt_npagealloc += pp->pr_npagealloc;
1661		PRWORD(ovflw, " %*lu", wide ? 11 : 6, 1, pp->pr_npagefree);
1662		pool_totals.pt_npagefree += pp->pr_npagefree;
1663		PRWORD(ovflw, " %*u", wide ? 8 : 6, 1, pp->pr_npages);
1664		pool_totals.pt_npages += pp->pr_npages;
1665		if (wide)
1666			PRWORD(ovflw, " %*u", 7, 1, pa.pa_pagesz);
1667		PRWORD(ovflw, " %*u", wide ? 7 : 6, 1, pp->pr_hiwat);
1668		PRWORD(ovflw, " %*u", 6, 1, pp->pr_minpages);
1669		PRWORD(ovflw, " %*s", wide ? 7 : 6, 1, maxp);
1670		PRWORD(ovflw, " %*lu", 5, 1, pp->pr_nidle);
1671		if (wide)
1672			PRWORD(ovflw, " 0x%0*x", 5, 1,
1673			    pp->pr_flags | pp->pr_roflags);
1674
1675		this_inuse = pp->pr_nout * pp->pr_size;
1676		this_total = pp->pr_npages * pa.pa_pagesz;
1677		if (pp->pr_roflags & PR_RECURSIVE) {
1678			/*
1679			 * Don't count in-use memory, since it's part
1680			 * of another pool and will be accounted for
1681			 * there.
1682			 */
1683			total += (this_total - this_inuse);
1684		} else {
1685			inuse += this_inuse;
1686			total += this_total;
1687		}
1688		if (wide) {
1689			if (this_total == 0)
1690				(void)printf("   ---");
1691			else
1692				(void)printf(" %5.1f%%",
1693				    (100.0 * this_inuse) / this_total);
1694		}
1695		(void)printf("\n");
1696	}
1697	ovflw = 0;
1698	PRWORD(ovflw, "%-*s", wide ? 16 : 11, 0, "Totals");
1699	PRWORD(ovflw, " %*s", wide ? 7 : 5, 1, "");
1700	PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pool_totals.pt_nget);
1701	PRWORD(ovflw, " %*" PRIu64, wide ? 8 : 5, 1, pool_totals.pt_nfail);
1702	PRWORD(ovflw, " %*" PRIu64, wide ? 12 : 9, 1, pool_totals.pt_nput);
1703 	if (wide) {
1704		PRWORD(ovflw, " %*" PRIu64, 9, 1, pool_totals.pt_nout);
1705		PRWORD(ovflw, " %*" PRIu64, 9, 1, pool_totals.pt_nitems);
1706 	}
1707	PRWORD(ovflw, " %*" PRIu64, wide ? 11 : 6, 1, pool_totals.pt_npagealloc);
1708	PRWORD(ovflw, " %*" PRIu64, wide ? 11 : 6, 1, pool_totals.pt_npagefree);
1709	PRWORD(ovflw, " %*" PRIu64, wide ? 8 : 6, 1, pool_totals.pt_npages);
1710	(void)printf("\n");
1711
1712	inuse /= KILO;
1713	total /= KILO;
1714	(void)printf(
1715	    "\nIn use %ldK, total allocated %ldK; utilization %.1f%%\n",
1716	    inuse, total, (100.0 * inuse) / total);
1717}
1718
1719static void
1720dopoolcache_sysctl(int verbose)
1721{
1722	struct pool_sysctl *data, *pp;
1723	size_t i, len;
1724	bool first = true;
1725	int ovflw;
1726	uint64_t tot;
1727	double p;
1728
1729	data = asysctlbyname("kern.pool", &len);
1730	if (data == NULL)
1731		err(1, "failed to read kern.pool");
1732	len /= sizeof(*data);
1733
1734	for (i = 0; i < len; ++i) {
1735		pp = &data[i];
1736		if (pp->pr_cache_meta_size == 0)
1737			continue;
1738
1739		if (pp->pr_cache_nmiss_global == 0 && !verbose)
1740			continue;
1741
1742		if (first) {
1743			(void)printf("Pool cache statistics.\n");
1744			(void)printf("%-*s%*s%*s%*s%*s%*s%*s%*s%*s%*s\n",
1745			    12, "Name",
1746			    6, "Spin",
1747			    6, "GrpSz",
1748			    5, "Full",
1749			    5, "Emty",
1750			    10, "PoolLayer",
1751			    11, "CacheLayer",
1752			    6, "Hit%",
1753			    12, "CpuLayer",
1754			    6, "Hit%"
1755			);
1756			first = false;
1757		}
1758
1759		ovflw = 0;
1760		PRWORD(ovflw, "%-*s", MIN((int)sizeof(pp->pr_wchan), 13), 1,
1761		    pp->pr_wchan);
1762		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_cache_ncontended);
1763		PRWORD(ovflw, " %*" PRIu64, 6, 1, pp->pr_cache_meta_size);
1764		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_cache_nfull);
1765		PRWORD(ovflw, " %*" PRIu64, 5, 1, pp->pr_cache_nempty);
1766		PRWORD(ovflw, " %*" PRIu64, 10, 1, pp->pr_cache_nmiss_global);
1767
1768		tot = pp->pr_cache_nhit_global + pp->pr_cache_nmiss_global;
1769		p = pp->pr_cache_nhit_global * 100.0 / tot;
1770		PRWORD(ovflw, " %*" PRIu64, 11, 1, tot);
1771		PRWORD(ovflw, " %*.1f", 6, 1, p);
1772
1773		tot = pp->pr_cache_nhit_pcpu + pp->pr_cache_nmiss_pcpu;
1774		p = pp->pr_cache_nhit_pcpu * 100.0 / tot;
1775		PRWORD(ovflw, " %*" PRIu64, 12, 1, tot);
1776		PRWORD(ovflw, " %*.1f", 6, 1, p);
1777		printf("\n");
1778	}
1779}
1780
1781void
1782dopoolcache(int verbose)
1783{
1784	struct pool_cache pool_cache, *pc = &pool_cache;
1785	pool_cache_cpu_t cache_cpu, *cc = &cache_cpu;
1786	TAILQ_HEAD(,pool) pool_head;
1787	struct pool pool, *pp = &pool;
1788	char name[32];
1789	uint64_t cpuhit, cpumiss, tot;
1790	void *addr;
1791	int first, ovflw;
1792	size_t i;
1793	double p;
1794
1795	if (memf == NULL)
1796		return dopoolcache_sysctl(verbose);
1797
1798	kread(namelist, X_POOLHEAD, &pool_head, sizeof(pool_head));
1799	addr = TAILQ_FIRST(&pool_head);
1800
1801	for (first = 1; addr != NULL; addr = TAILQ_NEXT(pp, pr_poollist) ) {
1802		deref_kptr(addr, pp, sizeof(*pp), "pool chain trashed");
1803		if (pp->pr_cache == NULL)
1804			continue;
1805		deref_kptr(pp->pr_wchan, name, sizeof(name),
1806		    "pool wait channel trashed");
1807		deref_kptr(pp->pr_cache, pc, sizeof(*pc), "pool cache trashed");
1808		if (pc->pc_misses == 0 && !verbose)
1809			continue;
1810		name[sizeof(name)-1] = '\0';
1811
1812		cpuhit = 0;
1813		cpumiss = 0;
1814		for (i = 0; i < __arraycount(pc->pc_cpus); i++) {
1815		    	if ((addr = pc->pc_cpus[i]) == NULL)
1816		    		continue;
1817			deref_kptr(addr, cc, sizeof(*cc),
1818			    "pool cache cpu trashed");
1819			cpuhit += cc->cc_hits;
1820			cpumiss += cc->cc_misses;
1821		}
1822
1823		if (first) {
1824			(void)printf("Pool cache statistics.\n");
1825			(void)printf("%-*s%*s%*s%*s%*s%*s%*s%*s%*s%*s\n",
1826			    12, "Name",
1827			    6, "Spin",
1828			    6, "GrpSz",
1829			    5, "Full",
1830			    5, "Emty",
1831			    10, "PoolLayer",
1832			    11, "CacheLayer",
1833			    6, "Hit%",
1834			    12, "CpuLayer",
1835			    6, "Hit%"
1836			);
1837			first = 0;
1838		}
1839
1840		ovflw = 0;
1841		PRWORD(ovflw, "%-*s", 13, 1, name);
1842		PRWORD(ovflw, " %*llu", 6, 1, (long long)pc->pc_contended);
1843		PRWORD(ovflw, " %*u", 6, 1, pc->pc_pcgsize);
1844		PRWORD(ovflw, " %*u", 5, 1, pc->pc_nfull);
1845		PRWORD(ovflw, " %*u", 5, 1, pc->pc_nempty);
1846		PRWORD(ovflw, " %*llu", 10, 1, (long long)pc->pc_misses);
1847
1848		tot = pc->pc_hits + pc->pc_misses;
1849		p = pc->pc_hits * 100.0 / (tot);
1850		PRWORD(ovflw, " %*llu", 11, 1, (long long)tot);
1851		PRWORD(ovflw, " %*.1f", 6, 1, p);
1852
1853		tot = cpuhit + cpumiss;
1854		p = cpuhit * 100.0 / (tot);
1855		PRWORD(ovflw, " %*llu", 12, 1, (long long)tot);
1856		PRWORD(ovflw, " %*.1f", 6, 1, p);
1857		printf("\n");
1858	}
1859}
1860
1861enum hashtype {			/* from <sys/systm.h> */
1862	HASH_LIST,
1863	HASH_SLIST,
1864	HASH_TAILQ,
1865	HASH_PSLIST
1866};
1867
1868struct uidinfo {		/* XXX: no kernel header file */
1869	LIST_ENTRY(uidinfo) ui_hash;
1870	uid_t	ui_uid;
1871	long	ui_proccnt;
1872};
1873
1874struct kernel_hash {
1875	const char *	description;	/* description */
1876	int		hashsize;	/* nlist index for hash size */
1877	int		hashtbl;	/* nlist index for hash table */
1878	enum hashtype	type;		/* type of hash table */
1879	size_t		offset;		/* offset of {LIST,TAILQ}_NEXT */
1880} khashes[] =
1881{
1882	{
1883		"buffer hash",
1884		X_BUFHASH, X_BUFHASHTBL,
1885		HASH_LIST, offsetof(struct buf, b_hash)
1886	}, {
1887		"ipv4 address -> interface hash",
1888		X_IFADDRHASH, X_IFADDRHASHTBL,
1889		HASH_LIST, offsetof(struct in_ifaddr, ia_hash),
1890	}, {
1891		"name cache hash",
1892		X_NCHASH, X_NCHASHTBL,
1893		HASH_LIST, offsetof(struct namecache, nc_hash),
1894	}, {
1895		"user info (uid -> used processes) hash",
1896		X_UIHASH, X_UIHASHTBL,
1897		HASH_LIST, offsetof(struct uidinfo, ui_hash),
1898	}, {
1899		"vnode cache hash",
1900		X_NCVCACHEHASH, X_NCVCACHETBL,
1901		HASH_SLIST, offsetof(struct vnode_impl, vi_hash),
1902	}, {
1903		NULL, -1, -1, 0, 0,
1904	}
1905};
1906
1907void
1908dohashstat(int verbose, int todo, const char *hashname)
1909{
1910	LIST_HEAD(, generic)	*hashtbl_list;
1911	SLIST_HEAD(, generic)	*hashtbl_slist;
1912	TAILQ_HEAD(, generic)	*hashtbl_tailq;
1913	struct kernel_hash	*curhash;
1914	void	*hashaddr, *hashbuf, *nhashbuf, *nextaddr;
1915	size_t	elemsize, hashbufsize, thissize;
1916	u_long	hashsize, i;
1917	int	used, items, chain, maxchain;
1918
1919	hashbuf = NULL;
1920	hashbufsize = 0;
1921
1922	if (todo & HASHLIST) {
1923		(void)printf("Supported hashes:\n");
1924		for (curhash = khashes; curhash->description; curhash++) {
1925			if (hashnl[curhash->hashsize].n_value == 0 ||
1926			    hashnl[curhash->hashtbl].n_value == 0)
1927				continue;
1928			(void)printf("\t%-16s%s\n",
1929			    hashnl[curhash->hashsize].n_name + 1,
1930			    curhash->description);
1931		}
1932		return;
1933	}
1934
1935	if (hashname != NULL) {
1936		for (curhash = khashes; curhash->description; curhash++) {
1937			if (strcmp(hashnl[curhash->hashsize].n_name + 1,
1938			    hashname) == 0 &&
1939			    hashnl[curhash->hashsize].n_value != 0 &&
1940			    hashnl[curhash->hashtbl].n_value != 0)
1941				break;
1942		}
1943		if (curhash->description == NULL) {
1944			warnx("%s: no such hash", hashname);
1945			return;
1946		}
1947	}
1948
1949	(void)printf(
1950	    "%-16s %8s %8s %8s %8s %8s %8s\n"
1951	    "%-16s %8s %8s %8s %8s %8s %8s\n",
1952	    "", "total", "used", "util", "num", "average", "maximum",
1953	    "hash table", "buckets", "buckets", "%", "items", "chain",
1954	    "chain");
1955
1956	for (curhash = khashes; curhash->description; curhash++) {
1957		if (hashnl[curhash->hashsize].n_value == 0 ||
1958		    hashnl[curhash->hashtbl].n_value == 0)
1959			continue;
1960		if (hashname != NULL &&
1961		    strcmp(hashnl[curhash->hashsize].n_name + 1, hashname))
1962			continue;
1963		elemsize = curhash->type == HASH_LIST ?
1964		    sizeof(*hashtbl_list) : sizeof(*hashtbl_tailq);
1965		deref_kptr((void *)hashnl[curhash->hashsize].n_value,
1966		    &hashsize, sizeof(hashsize),
1967		    hashnl[curhash->hashsize].n_name);
1968		hashsize++;
1969		deref_kptr((void *)hashnl[curhash->hashtbl].n_value,
1970		    &hashaddr, sizeof(hashaddr),
1971		    hashnl[curhash->hashtbl].n_name);
1972		if (verbose)
1973			(void)printf(
1974			    "%s %lu, %s %p, offset %ld, elemsize %llu\n",
1975			    hashnl[curhash->hashsize].n_name + 1, hashsize,
1976			    hashnl[curhash->hashtbl].n_name + 1, hashaddr,
1977			    (long)curhash->offset,
1978			    (unsigned long long)elemsize);
1979		thissize = hashsize * elemsize;
1980		if (hashbuf == NULL || thissize > hashbufsize) {
1981			if ((nhashbuf = realloc(hashbuf, thissize)) == NULL)
1982				errx(1, "malloc hashbuf %llu",
1983				    (unsigned long long)hashbufsize);
1984			hashbuf = nhashbuf;
1985			hashbufsize = thissize;
1986		}
1987		deref_kptr(hashaddr, hashbuf, thissize,
1988		    hashnl[curhash->hashtbl].n_name);
1989		used = 0;
1990		items = maxchain = 0;
1991		if (curhash->type == HASH_LIST) {
1992			hashtbl_list = hashbuf;
1993			hashtbl_slist = NULL;
1994			hashtbl_tailq = NULL;
1995		} else if (curhash->type == HASH_SLIST) {
1996			hashtbl_list = NULL;
1997			hashtbl_slist = hashbuf;
1998			hashtbl_tailq = NULL;
1999		} else {
2000			hashtbl_list = NULL;
2001			hashtbl_slist = NULL;
2002			hashtbl_tailq = hashbuf;
2003		}
2004		for (i = 0; i < hashsize; i++) {
2005			if (curhash->type == HASH_LIST)
2006				nextaddr = LIST_FIRST(&hashtbl_list[i]);
2007			else if (curhash->type == HASH_SLIST)
2008				nextaddr = SLIST_FIRST(&hashtbl_slist[i]);
2009			else
2010				nextaddr = TAILQ_FIRST(&hashtbl_tailq[i]);
2011			if (nextaddr == NULL)
2012				continue;
2013			if (verbose)
2014				(void)printf("%5lu: %p\n", i, nextaddr);
2015			used++;
2016			chain = 0;
2017			do {
2018				chain++;
2019				deref_kptr((char *)nextaddr + curhash->offset,
2020				    &nextaddr, sizeof(void *),
2021				    "hash chain corrupted");
2022				if (verbose > 1)
2023					(void)printf("got nextaddr as %p\n",
2024					    nextaddr);
2025			} while (nextaddr != NULL);
2026			items += chain;
2027			if (verbose && chain > 1)
2028				(void)printf("\tchain = %d\n", chain);
2029			if (chain > maxchain)
2030				maxchain = chain;
2031		}
2032		(void)printf("%-16s %8ld %8d %8.2f %8d %8.2f %8d\n",
2033		    hashnl[curhash->hashsize].n_name + 1,
2034		    hashsize, used, used * 100.0 / hashsize,
2035		    items, used ? (double)items / used : 0.0, maxchain);
2036	}
2037}
2038
2039/*
2040 * kreadc like kread but returns 1 if successful, 0 otherwise
2041 */
2042int
2043kreadc(struct nlist *nl, int nlx, void *addr, size_t size)
2044{
2045	const char *sym;
2046
2047	sym = nl[nlx].n_name;
2048	if (*sym == '_')
2049		++sym;
2050	if (nl[nlx].n_type == 0 || nl[nlx].n_value == 0)
2051		return 0;
2052	deref_kptr((void *)nl[nlx].n_value, addr, size, sym);
2053	return 1;
2054}
2055
2056/*
2057 * kread reads something from the kernel, given its nlist index in namelist[].
2058 */
2059void
2060kread(struct nlist *nl, int nlx, void *addr, size_t size)
2061{
2062	const char *sym;
2063
2064	sym = nl[nlx].n_name;
2065	if (*sym == '_')
2066		++sym;
2067	if (nl[nlx].n_type == 0 || nl[nlx].n_value == 0)
2068		errx(1, "symbol %s not defined", sym);
2069	deref_kptr((void *)nl[nlx].n_value, addr, size, sym);
2070}
2071
2072/*
2073 * Dereference the kernel pointer `kptr' and fill in the local copy
2074 * pointed to by `ptr'.  The storage space must be pre-allocated,
2075 * and the size of the copy passed in `len'.
2076 */
2077void
2078deref_kptr(const void *kptr, void *ptr, size_t len, const char *msg)
2079{
2080
2081	if (*msg == '_')
2082		msg++;
2083	if ((size_t)kvm_read(kd, (u_long)kptr, (char *)ptr, len) != len)
2084		errx(1, "kptr %lx: %s: %s", (u_long)kptr, msg, kvm_geterr(kd));
2085}
2086
2087/*
2088 * Traverse the kernel history buffers, performing the requested action.
2089 *
2090 * Note, we assume that if we're not listing, we're dumping.
2091 */
2092void
2093hist_traverse(int todo, const char *histname)
2094{
2095	struct kern_history_head histhead;
2096	struct kern_history hist, *histkva;
2097	char *name = NULL;
2098	size_t namelen = 0;
2099
2100	if (histnl[0].n_value == 0) {
2101		warnx("kernel history is not compiled into the kernel.");
2102		return;
2103	}
2104
2105	deref_kptr((void *)histnl[X_KERN_HISTORIES].n_value, &histhead,
2106	    sizeof(histhead), histnl[X_KERN_HISTORIES].n_name);
2107
2108	if (histhead.lh_first == NULL) {
2109		warnx("No active kernel history logs.");
2110		return;
2111	}
2112
2113	if (todo & HISTLIST)
2114		(void)printf("Active kernel histories:");
2115
2116	for (histkva = LIST_FIRST(&histhead); histkva != NULL;
2117	    histkva = LIST_NEXT(&hist, list)) {
2118		deref_kptr(histkva, &hist, sizeof(hist), "histkva");
2119		if (name == NULL || hist.namelen > namelen) {
2120			if (name != NULL)
2121				free(name);
2122			namelen = hist.namelen;
2123			if ((name = malloc(namelen + 1)) == NULL)
2124				err(1, "malloc history name");
2125		}
2126
2127		deref_kptr(hist.name, name, namelen, "history name");
2128		name[namelen] = '\0';
2129		if (todo & HISTLIST)
2130			(void)printf(" %s", name);
2131		else {
2132			/*
2133			 * If we're dumping all histories, do it, else
2134			 * check to see if this is the one we want.
2135			 */
2136			if (histname == NULL || strcmp(histname, name) == 0) {
2137				if (histname == NULL)
2138					(void)printf(
2139					    "\nkernel history `%s':\n", name);
2140				hist_dodump(&hist);
2141			}
2142		}
2143	}
2144
2145	if (todo & HISTLIST)
2146		(void)putchar('\n');
2147
2148	if (name != NULL)
2149		free(name);
2150}
2151
2152/*
2153 * Actually dump the history buffer at the specified KVA.
2154 */
2155void
2156hist_dodump(struct kern_history *histp)
2157{
2158	struct kern_history_ent *histents, *e;
2159	struct timeval tv;
2160	size_t histsize;
2161	char *fmt = NULL, *fn = NULL;
2162	size_t fmtlen = 0, fnlen = 0;
2163	unsigned i;
2164
2165	histsize = sizeof(struct kern_history_ent) * histp->n;
2166
2167	if ((histents = malloc(histsize)) == NULL)
2168		err(1, "malloc history entries");
2169
2170	(void)memset(histents, 0, histsize);
2171
2172	(void)printf("%"PRIu32" entries, next is %"PRIu32"\n",
2173	    histp->n, histp->f);
2174
2175	deref_kptr(histp->e, histents, histsize, "history entries");
2176	i = histp->f;
2177	do {
2178		e = &histents[i];
2179		if (e->fmt != NULL) {
2180			if (fmt == NULL || e->fmtlen > fmtlen) {
2181				free(fmt);
2182				fmtlen = e->fmtlen;
2183				if ((fmt = malloc(fmtlen + 1)) == NULL)
2184					err(1, "malloc printf format");
2185			}
2186			if (fn == NULL || e->fnlen > fnlen) {
2187				free(fn);
2188				fnlen = e->fnlen;
2189				if ((fn = malloc(fnlen + 1)) == NULL)
2190					err(1, "malloc function name");
2191			}
2192
2193			deref_kptr(e->fmt, fmt, fmtlen, "printf format");
2194			fmt[fmtlen] = '\0';
2195			for (unsigned z = 0; z < fmtlen - 1; z++) {
2196				if (fmt[z] == '%' && fmt[z+1] == 's')
2197					fmt[z+1] = 'p';
2198			}
2199
2200			deref_kptr(e->fn, fn, fnlen, "function name");
2201			fn[fnlen] = '\0';
2202
2203			bintime2timeval(&e->bt, &tv);
2204			(void)printf("%06ld.%06ld ", (long int)tv.tv_sec,
2205			    (long int)tv.tv_usec);
2206			(void)printf("%s#%" PRId32 "@%" PRId32 "d: ",
2207			    fn, e->call, e->cpunum);
2208			(void)printf(fmt, e->v[0], e->v[1], e->v[2], e->v[3]);
2209			(void)putchar('\n');
2210		}
2211		i = (i + 1) % histp->n;
2212	} while (i != histp->f);
2213
2214	free(histents);
2215	free(fmt);
2216	free(fn);
2217}
2218
2219void
2220hist_traverse_sysctl(int todo, const char *histname)
2221{
2222	int error;
2223	int mib[4];
2224	unsigned int i;
2225	size_t len, miblen;
2226	struct sysctlnode query, histnode[32];
2227
2228	/* retrieve names of available histories */
2229	miblen = __arraycount(mib);
2230	error = sysctlnametomib("kern.hist", mib, &miblen);
2231	if (error != 0) {
2232		if (errno == ENOENT) {
2233 			warnx("kernel history is not compiled into the kernel.");
2234			return;
2235		} else
2236			err(EXIT_FAILURE, "nametomib failed");
2237	}
2238
2239	/* get the list of nodenames below kern.hist */
2240	mib[2] = CTL_QUERY;
2241	memset(&query, 0, sizeof(query));
2242	query.sysctl_flags = SYSCTL_VERSION;
2243	len = sizeof(histnode);
2244	error = sysctl(mib, 3, &histnode[0], &len, &query, sizeof(query));
2245	if (error != 0) {
2246		err(1, "query failed");
2247		return;
2248	}
2249	if (len == 0) {
2250 		warnx("No active kernel history logs.");
2251 		return;
2252 	}
2253
2254	len = len / sizeof(histnode[0]);	/* get # of entries returned */
2255
2256 	if (todo & HISTLIST)
2257 		(void)printf("Active kernel histories:");
2258
2259	for (i = 0; i < len; i++) {
2260 		if (todo & HISTLIST)
2261			(void)printf(" %s", histnode[i].sysctl_name);
2262 		else {
2263 			/*
2264 			 * If we're dumping all histories, do it, else
2265 			 * check to see if this is the one we want.
2266 			 */
2267			if (histname == NULL ||
2268			    strcmp(histname, histnode[i].sysctl_name) == 0) {
2269 				if (histname == NULL)
2270 					(void)printf(
2271					    "\nkernel history `%s':\n",
2272					    histnode[i].sysctl_name);
2273				mib[2] = histnode[i].sysctl_num;
2274				mib[3] = CTL_EOL;
2275				hist_dodump_sysctl(mib, 4);
2276 			}
2277 		}
2278 	}
2279
2280 	if (todo & HISTLIST)
2281 		(void)putchar('\n');
2282	else if (mib[2] == CTL_QUERY)
2283		warnx("history %s not found", histname);
2284 }
2285
2286 /*
2287  * Actually dump the history buffer at the specified KVA.
2288  */
2289void
2290hist_dodump_sysctl(int mib[], unsigned int miblen)
2291{
2292	struct sysctl_history *hist;
2293	struct timeval tv;
2294	struct sysctl_history_event *e;
2295 	size_t histsize;
2296	char *strp;
2297 	unsigned i;
2298	char *fmt = NULL, *fn = NULL;
2299
2300	hist = NULL;
2301	histsize = 0;
2302 	do {
2303		errno = 0;
2304		if (sysctl(mib, miblen, hist, &histsize, NULL, 0) == 0)
2305			break;
2306		if (errno != ENOMEM)
2307			break;
2308		if ((hist = realloc(hist, histsize)) == NULL)
2309			errx(1, "realloc history buffer");
2310	} while (errno == ENOMEM);
2311	if (errno != 0)
2312		err(1, "sysctl failed");
2313
2314	strp = (char *)(&hist->sh_events[hist->sh_numentries]);
2315
2316	(void)printf("%"PRIu32" entries, next is %"PRIu32"\n",
2317	    hist->sh_numentries,
2318	    hist->sh_nextfree);
2319
2320	i = hist->sh_nextfree;
2321
2322	do {
2323		e = &hist->sh_events[i];
2324		if (e->she_fmtoffset != 0) {
2325			fmt = &strp[e->she_fmtoffset];
2326			size_t fmtlen = strlen(fmt);
2327			for (unsigned z = 0; z < fmtlen - 1; z++) {
2328				if (fmt[z] == '%' && fmt[z+1] == 's')
2329					fmt[z+1] = 'p';
2330			}
2331			fn = &strp[e->she_funcoffset];
2332			bintime2timeval(&e->she_bintime, &tv);
2333			(void)printf("%06ld.%06ld %s#%"PRIu32"@%"PRIu32": ",
2334			    (long int)tv.tv_sec, (long int)tv.tv_usec,
2335			    fn, e->she_callnumber, e->she_cpunum);
2336			(void)printf(fmt, e->she_values[0], e->she_values[1],
2337			     e->she_values[2], e->she_values[3]);
2338 			(void)putchar('\n');
2339 		}
2340		i = (i + 1) % hist->sh_numentries;
2341	} while (i != hist->sh_nextfree);
2342
2343	free(hist);
2344 }
2345
2346static void
2347usage(void)
2348{
2349
2350	(void)fprintf(stderr,
2351	    "usage: %s [-CefHiLlmstUvW] [-c count] [-h hashname] [-M core] [-N system]\n"
2352	    "\t\t[-u histname] [-w wait] [disks]\n", getprogname());
2353	exit(1);
2354}
2355