Move the world out of /zfs/ and seperate out module build tree

cmd/zdb/zdb_il.c

@@ -0,0 +1,363 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+/*
+ * Print intent log header and statistics.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/dmu.h>
+#include <sys/stat.h>
+#include <sys/resource.h>
+#include <sys/zil.h>
+#include <sys/zil_impl.h>
+
+extern uint8_t dump_opt[256];
+
+static void
+print_log_bp(const blkptr_t *bp, const char *prefix)
+{
+	char blkbuf[BP_SPRINTF_LEN];
+
+	sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
+	(void) printf("%s%s\n", prefix, blkbuf);
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_create(zilog_t *zilog, int txtype, lr_create_t *lr)
+{
+	time_t crtime = lr->lr_crtime[0];
+	char *name = (char *)(lr + 1);
+	char *link = name + strlen(name) + 1;
+
+	if (txtype == TX_SYMLINK)
+		(void) printf("\t\t\t%s -> %s\n", name, link);
+	else
+		(void) printf("\t\t\t%s\n", name);
+
+	(void) printf("\t\t\t%s", ctime(&crtime));
+	(void) printf("\t\t\tdoid %llu, foid %llu, mode %llo\n",
+	    (u_longlong_t)lr->lr_doid, (u_longlong_t)lr->lr_foid,
+	    (longlong_t)lr->lr_mode);
+	(void) printf("\t\t\tuid %llu, gid %llu, gen %llu, rdev 0x%llx\n",
+	    (u_longlong_t)lr->lr_uid, (u_longlong_t)lr->lr_gid,
+	    (u_longlong_t)lr->lr_gen, (u_longlong_t)lr->lr_rdev);
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_remove(zilog_t *zilog, int txtype, lr_remove_t *lr)
+{
+	(void) printf("\t\t\tdoid %llu, name %s\n",
+	    (u_longlong_t)lr->lr_doid, (char *)(lr + 1));
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_link(zilog_t *zilog, int txtype, lr_link_t *lr)
+{
+	(void) printf("\t\t\tdoid %llu, link_obj %llu, name %s\n",
+	    (u_longlong_t)lr->lr_doid, (u_longlong_t)lr->lr_link_obj,
+	    (char *)(lr + 1));
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_rename(zilog_t *zilog, int txtype, lr_rename_t *lr)
+{
+	char *snm = (char *)(lr + 1);
+	char *tnm = snm + strlen(snm) + 1;
+
+	(void) printf("\t\t\tsdoid %llu, tdoid %llu\n",
+	    (u_longlong_t)lr->lr_sdoid, (u_longlong_t)lr->lr_tdoid);
+	(void) printf("\t\t\tsrc %s tgt %s\n", snm, tnm);
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_write(zilog_t *zilog, int txtype, lr_write_t *lr)
+{
+	char *data, *dlimit;
+	blkptr_t *bp = &lr->lr_blkptr;
+	char buf[SPA_MAXBLOCKSIZE];
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
+	int error;
+
+	(void) printf("\t\t\tfoid %llu, offset 0x%llx,"
+	    " length 0x%llx, blkoff 0x%llx\n",
+	    (u_longlong_t)lr->lr_foid, (longlong_t)lr->lr_offset,
+	    (u_longlong_t)lr->lr_length, (u_longlong_t)lr->lr_blkoff);
+
+	if (verbose < 5)
+		return;
+
+	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
+		(void) printf("\t\t\thas blkptr, %s\n",
+		    bp->blk_birth >= spa_first_txg(zilog->zl_spa) ?
+		    "will claim" : "won't claim");
+		print_log_bp(bp, "\t\t\t");
+		if (bp->blk_birth == 0) {
+			bzero(buf, sizeof (buf));
+		} else {
+			zbookmark_t zb;
+
+			ASSERT3U(bp->blk_cksum.zc_word[ZIL_ZC_OBJSET], ==,
+			    dmu_objset_id(zilog->zl_os));
+
+			zb.zb_objset = bp->blk_cksum.zc_word[ZIL_ZC_OBJSET];
+			zb.zb_object = 0;
+			zb.zb_level = -1;
+			zb.zb_blkid = bp->blk_cksum.zc_word[ZIL_ZC_SEQ];
+
+			error = zio_wait(zio_read(NULL, zilog->zl_spa,
+			    bp, buf, BP_GET_LSIZE(bp), NULL, NULL,
+			    ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &zb));
+			if (error)
+				return;
+		}
+		data = buf + lr->lr_blkoff;
+	} else {
+		data = (char *)(lr + 1);
+	}
+
+	dlimit = data + MIN(lr->lr_length,
+	    (verbose < 6 ? 20 : SPA_MAXBLOCKSIZE));
+
+	(void) printf("\t\t\t");
+	while (data < dlimit) {
+		if (isprint(*data))
+			(void) printf("%c ", *data);
+		else
+			(void) printf("%2X", *data);
+		data++;
+	}
+	(void) printf("\n");
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_truncate(zilog_t *zilog, int txtype, lr_truncate_t *lr)
+{
+	(void) printf("\t\t\tfoid %llu, offset 0x%llx, length 0x%llx\n",
+	    (u_longlong_t)lr->lr_foid, (longlong_t)lr->lr_offset,
+	    (u_longlong_t)lr->lr_length);
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_setattr(zilog_t *zilog, int txtype, lr_setattr_t *lr)
+{
+	time_t atime = (time_t)lr->lr_atime[0];
+	time_t mtime = (time_t)lr->lr_mtime[0];
+
+	(void) printf("\t\t\tfoid %llu, mask 0x%llx\n",
+	    (u_longlong_t)lr->lr_foid, (u_longlong_t)lr->lr_mask);
+
+	if (lr->lr_mask & AT_MODE) {
+		(void) printf("\t\t\tAT_MODE  %llo\n",
+		    (longlong_t)lr->lr_mode);
+	}
+
+	if (lr->lr_mask & AT_UID) {
+		(void) printf("\t\t\tAT_UID   %llu\n",
+		    (u_longlong_t)lr->lr_uid);
+	}
+
+	if (lr->lr_mask & AT_GID) {
+		(void) printf("\t\t\tAT_GID   %llu\n",
+		    (u_longlong_t)lr->lr_gid);
+	}
+
+	if (lr->lr_mask & AT_SIZE) {
+		(void) printf("\t\t\tAT_SIZE  %llu\n",
+		    (u_longlong_t)lr->lr_size);
+	}
+
+	if (lr->lr_mask & AT_ATIME) {
+		(void) printf("\t\t\tAT_ATIME %llu.%09llu %s",
+		    (u_longlong_t)lr->lr_atime[0],
+		    (u_longlong_t)lr->lr_atime[1],
+		    ctime(&atime));
+	}
+
+	if (lr->lr_mask & AT_MTIME) {
+		(void) printf("\t\t\tAT_MTIME %llu.%09llu %s",
+		    (u_longlong_t)lr->lr_mtime[0],
+		    (u_longlong_t)lr->lr_mtime[1],
+		    ctime(&mtime));
+	}
+}
+
+/* ARGSUSED */
+static void
+zil_prt_rec_acl(zilog_t *zilog, int txtype, lr_acl_t *lr)
+{
+	(void) printf("\t\t\tfoid %llu, aclcnt %llu\n",
+	    (u_longlong_t)lr->lr_foid, (u_longlong_t)lr->lr_aclcnt);
+}
+
+typedef void (*zil_prt_rec_func_t)();
+typedef struct zil_rec_info {
+	zil_prt_rec_func_t	zri_print;
+	char			*zri_name;
+	uint64_t		zri_count;
+} zil_rec_info_t;
+
+static zil_rec_info_t zil_rec_info[TX_MAX_TYPE] = {
+	{	NULL,			"Total              " },
+	{	zil_prt_rec_create,	"TX_CREATE          " },
+	{	zil_prt_rec_create,	"TX_MKDIR           " },
+	{	zil_prt_rec_create,	"TX_MKXATTR         " },
+	{	zil_prt_rec_create,	"TX_SYMLINK         " },
+	{	zil_prt_rec_remove,	"TX_REMOVE          " },
+	{	zil_prt_rec_remove,	"TX_RMDIR           " },
+	{	zil_prt_rec_link,	"TX_LINK            " },
+	{	zil_prt_rec_rename,	"TX_RENAME          " },
+	{	zil_prt_rec_write,	"TX_WRITE           " },
+	{	zil_prt_rec_truncate,	"TX_TRUNCATE        " },
+	{	zil_prt_rec_setattr,	"TX_SETATTR         " },
+	{	zil_prt_rec_acl,	"TX_ACL_V0          " },
+	{	zil_prt_rec_acl,	"TX_ACL_ACL         " },
+	{	zil_prt_rec_create,	"TX_CREATE_ACL      " },
+	{	zil_prt_rec_create,	"TX_CREATE_ATTR     " },
+	{	zil_prt_rec_create,	"TX_CREATE_ACL_ATTR " },
+	{	zil_prt_rec_create,	"TX_MKDIR_ACL       " },
+	{	zil_prt_rec_create,	"TX_MKDIR_ATTR      " },
+	{	zil_prt_rec_create,	"TX_MKDIR_ACL_ATTR  " },
+};
+
+/* ARGSUSED */
+static void
+print_log_record(zilog_t *zilog, lr_t *lr, void *arg, uint64_t claim_txg)
+{
+	int txtype;
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
+
+	/* reduce size of txtype to strip off TX_CI bit */
+	txtype = lr->lrc_txtype;
+
+	ASSERT(txtype != 0 && (uint_t)txtype < TX_MAX_TYPE);
+	ASSERT(lr->lrc_txg);
+
+	(void) printf("\t\t%s%s len %6llu, txg %llu, seq %llu\n",
+	    (lr->lrc_txtype & TX_CI) ? "CI-" : "",
+	    zil_rec_info[txtype].zri_name,
+	    (u_longlong_t)lr->lrc_reclen,
+	    (u_longlong_t)lr->lrc_txg,
+	    (u_longlong_t)lr->lrc_seq);
+
+	if (txtype && verbose >= 3)
+		zil_rec_info[txtype].zri_print(zilog, txtype, lr);
+
+	zil_rec_info[txtype].zri_count++;
+	zil_rec_info[0].zri_count++;
+}
+
+/* ARGSUSED */
+static void
+print_log_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
+{
+	char blkbuf[BP_SPRINTF_LEN];
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
+	char *claim;
+
+	if (verbose <= 3)
+		return;
+
+	if (verbose >= 5) {
+		(void) strcpy(blkbuf, ", ");
+		sprintf_blkptr(blkbuf + strlen(blkbuf),
+		    BP_SPRINTF_LEN - strlen(blkbuf), bp);
+	} else {
+		blkbuf[0] = '\0';
+	}
+
+	if (claim_txg != 0)
+		claim = "already claimed";
+	else if (bp->blk_birth >= spa_first_txg(zilog->zl_spa))
+		claim = "will claim";
+	else
+		claim = "won't claim";
+
+	(void) printf("\tBlock seqno %llu, %s%s\n",
+	    (u_longlong_t)bp->blk_cksum.zc_word[ZIL_ZC_SEQ], claim, blkbuf);
+}
+
+static void
+print_log_stats(int verbose)
+{
+	int i, w, p10;
+
+	if (verbose > 3)
+		(void) printf("\n");
+
+	if (zil_rec_info[0].zri_count == 0)
+		return;
+
+	for (w = 1, p10 = 10; zil_rec_info[0].zri_count >= p10; p10 *= 10)
+		w++;
+
+	for (i = 0; i < TX_MAX_TYPE; i++)
+		if (zil_rec_info[i].zri_count || verbose >= 3)
+			(void) printf("\t\t%s %*llu\n",
+			    zil_rec_info[i].zri_name, w,
+			    (u_longlong_t)zil_rec_info[i].zri_count);
+	(void) printf("\n");
+}
+
+/* ARGSUSED */
+void
+dump_intent_log(zilog_t *zilog)
+{
+	const zil_header_t *zh = zilog->zl_header;
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
+	int i;
+
+	if (zh->zh_log.blk_birth == 0 || verbose < 2)
+		return;
+
+	(void) printf("\n    ZIL header: claim_txg %llu, seq %llu\n",
+	    (u_longlong_t)zh->zh_claim_txg, (u_longlong_t)zh->zh_replay_seq);
+
+	if (verbose >= 4)
+		print_log_bp(&zh->zh_log, "\n\tfirst block: ");
+
+	for (i = 0; i < TX_MAX_TYPE; i++)
+		zil_rec_info[i].zri_count = 0;
+
+	if (verbose >= 2) {
+		(void) printf("\n");
+		(void) zil_parse(zilog, print_log_block, print_log_record, NULL,
+		    zh->zh_claim_txg);
+		print_log_stats(verbose);
+	}
+}

cmd/zdump/zdump.c

@@ -0,0 +1,664 @@
+/*
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+/*
+ * zdump 7.24
+ * Taken from elsie.nci.nih.gov to replace the existing Solaris zdump,
+ * which was based on an earlier version of the elsie code.
+ *
+ * For zdump 7.24, the following changes were made to the elsie code:
+ *   locale/textdomain/messages to match existing Solaris style.
+ *   Solaris verbose mode is documented to display the current time first.
+ *   cstyle cleaned code.
+ *   removed old locale/textdomain code.
+ */
+
+static char	elsieid[] = "@(#)zdump.c	7.74";
+
+/*
+ * This code has been made independent of the rest of the time
+ * conversion package to increase confidence in the verification it provides.
+ * You can use this code to help in verifying other implementations.
+ */
+
+#include "stdio.h"	/* for stdout, stderr, perror */
+#include "string.h"	/* for strcpy */
+#include "sys/types.h"	/* for time_t */
+#include "time.h"	/* for struct tm */
+#include "stdlib.h"	/* for exit, malloc, atoi */
+#include "locale.h"	/* for setlocale, textdomain */
+#include "libintl.h"
+#include <ctype.h>
+#include "tzfile.h"	/* for defines */
+#include <limits.h>
+
+#ifndef ZDUMP_LO_YEAR
+#define	ZDUMP_LO_YEAR	(-500)
+#endif /* !defined ZDUMP_LO_YEAR */
+
+#ifndef ZDUMP_HI_YEAR
+#define	ZDUMP_HI_YEAR	2500
+#endif /* !defined ZDUMP_HI_YEAR */
+
+#ifndef MAX_STRING_LENGTH
+#define	MAX_STRING_LENGTH	1024
+#endif /* !defined MAX_STRING_LENGTH */
+
+#ifndef TRUE
+#define	TRUE		1
+#endif /* !defined TRUE */
+
+#ifndef FALSE
+#define	FALSE		0
+#endif /* !defined FALSE */
+
+#ifndef isleap_sum
+/*
+ * See tzfile.h for details on isleap_sum.
+ */
+#define	isleap_sum(a, b)	isleap((a) % 400 + (b) % 400)
+#endif /* !defined isleap_sum */
+
+#ifndef SECSPERDAY
+#define	SECSPERDAY	((long)SECSPERHOUR * HOURSPERDAY)
+#endif
+#define	SECSPERNYEAR	(SECSPERDAY * DAYSPERNYEAR)
+#define	SECSPERLYEAR	(SECSPERNYEAR + SECSPERDAY)
+
+#ifndef GNUC_or_lint
+#ifdef lint
+#define	GNUC_or_lint
+#else /* !defined lint */
+#ifdef __GNUC__
+#define	GNUC_or_lint
+#endif /* defined __GNUC__ */
+#endif /* !defined lint */
+#endif /* !defined GNUC_or_lint */
+
+#ifndef INITIALIZE
+#ifdef	GNUC_or_lint
+#define	INITIALIZE(x)	((x) = 0)
+#else /* !defined GNUC_or_lint */
+#define	INITIALIZE(x)
+#endif /* !defined GNUC_or_lint */
+#endif /* !defined INITIALIZE */
+
+static time_t	absolute_min_time;
+static time_t	absolute_max_time;
+static size_t	longest;
+static char	*progname;
+static int	warned;
+
+static char	*abbr(struct tm *);
+static void	abbrok(const char *, const char *);
+static long	delta(struct tm *, struct tm *);
+static void	dumptime(const struct tm *);
+static time_t	hunt(char *, time_t, time_t);
+static void	setabsolutes(void);
+static void	show(char *, time_t, int);
+static void	usage(void);
+static const char	*tformat(void);
+static time_t	yeartot(long y);
+
+#ifndef TYPECHECK
+#define	my_localtime	localtime
+#else /* !defined TYPECHECK */
+static struct tm *
+my_localtime(tp)
+time_t *tp;
+{
+	register struct tm *tmp;
+
+	tmp = localtime(tp);
+	if (tp != NULL && tmp != NULL) {
+		struct tm	tm;
+		register time_t	t;
+
+		tm = *tmp;
+		t = mktime(&tm);
+		if (t - *tp >= 1 || *tp - t >= 1) {
+			(void) fflush(stdout);
+			(void) fprintf(stderr, "\n%s: ", progname);
+			(void) fprintf(stderr, tformat(), *tp);
+			(void) fprintf(stderr, " ->");
+			(void) fprintf(stderr, " year=%d", tmp->tm_year);
+			(void) fprintf(stderr, " mon=%d", tmp->tm_mon);
+			(void) fprintf(stderr, " mday=%d", tmp->tm_mday);
+			(void) fprintf(stderr, " hour=%d", tmp->tm_hour);
+			(void) fprintf(stderr, " min=%d", tmp->tm_min);
+			(void) fprintf(stderr, " sec=%d", tmp->tm_sec);
+			(void) fprintf(stderr, " isdst=%d", tmp->tm_isdst);
+			(void) fprintf(stderr, " -> ");
+			(void) fprintf(stderr, tformat(), t);
+			(void) fprintf(stderr, "\n");
+		}
+	}
+	return (tmp);
+}
+#endif /* !defined TYPECHECK */
+
+static void
+abbrok(abbrp, zone)
+const char * const	abbrp;
+const char * const	zone;
+{
+	register const char *cp;
+	int error = 0;
+
+	if (warned)
+		return;
+	cp = abbrp;
+	while (isascii(*cp) && isalpha((unsigned char)*cp))
+		++cp;
+	(void) fflush(stdout);
+	if (cp - abbrp == 0) {
+		/*
+		 * TRANSLATION_NOTE
+		 * The first %s prints for the program name (zdump),
+		 * the second %s prints the timezone name, the third
+		 * %s prints the timezone abbreviation (tzname[0] or
+		 * tzname[1]).
+		 */
+		(void) fprintf(stderr, gettext("%s: warning: zone \"%s\" "
+		    "abbreviation \"%s\" lacks alphabetic at start\n"),
+		    progname, zone, abbrp);
+		error = 1;
+	} else if (cp - abbrp < 3) {
+		(void) fprintf(stderr, gettext("%s: warning: zone \"%s\" "
+		    "abbreviation \"%s\" has fewer than 3 alphabetics\n"),
+		    progname, zone, abbrp);
+		error = 1;
+	} else if (cp - abbrp > 6) {
+		(void) fprintf(stderr, gettext("%s: warning: zone \"%s\" "
+		    "abbreviation \"%s\" has more than 6 alphabetics\n"),
+		    progname, zone, abbrp);
+		error = 1;
+	}
+	if (error == 0 && (*cp == '+' || *cp == '-')) {
+		++cp;
+		if (isascii(*cp) && isdigit((unsigned char)*cp))
+			if (*cp++ == '1' && *cp >= '0' && *cp <= '4')
+				++cp;
+		if (*cp != '\0') {
+			(void) fprintf(stderr, gettext("%s: warning: "
+			    "zone \"%s\" abbreviation \"%s\" differs from "
+			    "POSIX standard\n"), progname, zone, abbrp);
+			error = 1;
+		}
+	}
+	if (error)
+		warned = TRUE;
+}
+
+int
+main(argc, argv)
+int	argc;
+char	*argv[];
+{
+	register int		i;
+	register int		c;
+	register int		vflag;
+	register char		*cutarg;
+	register long		cutloyear = ZDUMP_LO_YEAR;
+	register long		cuthiyear = ZDUMP_HI_YEAR;
+	register time_t		cutlotime;
+	register time_t		cuthitime;
+	time_t			now;
+	time_t			t;
+	time_t			newt;
+	struct tm		tm;
+	struct tm		newtm;
+	register struct tm	*tmp;
+	register struct tm	*newtmp;
+
+	INITIALIZE(cutlotime);
+	INITIALIZE(cuthitime);
+
+	(void) setlocale(LC_ALL, "");
+#if !defined(TEXT_DOMAIN)		/* Should be defined by cc -D */
+#define	TEXT_DOMAIN	"SYS_TEST"	/* Use this only if it weren't */
+#endif
+	(void) textdomain(TEXT_DOMAIN);
+
+	progname = argv[0];
+	for (i = 1; i < argc; ++i)
+		if (strcmp(argv[i], "--version") == 0) {
+			(void) printf("%s\n", elsieid);
+			exit(EXIT_SUCCESS);
+		}
+	vflag = 0;
+	cutarg = NULL;
+	while ((c = getopt(argc, argv, "c:v")) == 'c' || c == 'v')
+		if (c == 'v')
+			vflag = 1;
+		else	cutarg = optarg;
+	if (c != EOF ||
+		(optind == argc - 1 && strcmp(argv[optind], "=") == 0)) {
+			usage();
+			/* NOTREACHED */
+	}
+	if (vflag) {
+		if (cutarg != NULL) {
+			long	lo;
+			long	hi;
+			char	dummy;
+
+			if (sscanf(cutarg, "%ld%c", &hi, &dummy) == 1) {
+				cuthiyear = hi;
+			} else if (sscanf(cutarg, "%ld,%ld%c",
+				&lo, &hi, &dummy) == 2) {
+					cutloyear = lo;
+					cuthiyear = hi;
+			} else {
+(void) fprintf(stderr, gettext("%s: wild -c argument %s\n"),
+					progname, cutarg);
+				exit(EXIT_FAILURE);
+			}
+		}
+		setabsolutes();
+		cutlotime = yeartot(cutloyear);
+		cuthitime = yeartot(cuthiyear);
+	}
+	(void) time(&now);
+	longest = 0;
+	for (i = optind; i < argc; ++i)
+		if (strlen(argv[i]) > longest)
+			longest = strlen(argv[i]);
+
+	for (i = optind; i < argc; ++i) {
+		static char	buf[MAX_STRING_LENGTH];
+		static char	*tzp = NULL;
+
+		(void) unsetenv("TZ");
+		if (tzp != NULL)
+			free(tzp);
+		if ((tzp = malloc(3 + strlen(argv[i]) + 1)) == NULL) {
+			perror(progname);
+			exit(EXIT_FAILURE);
+		}
+		(void) strcpy(tzp, "TZ=");
+		(void) strcat(tzp, argv[i]);
+		if (putenv(tzp) != 0) {
+			perror(progname);
+			exit(EXIT_FAILURE);
+		}
+		if (!vflag) {
+			show(argv[i], now, FALSE);
+			continue;
+		}
+
+#if defined(sun)
+		/*
+		 * We show the current time first, probably because we froze
+		 * the behavior of zdump some time ago and then it got
+		 * changed.
+		 */
+		show(argv[i], now, TRUE);
+#endif
+		warned = FALSE;
+		t = absolute_min_time;
+		show(argv[i], t, TRUE);
+		t += SECSPERHOUR * HOURSPERDAY;
+		show(argv[i], t, TRUE);
+		if (t < cutlotime)
+			t = cutlotime;
+		tmp = my_localtime(&t);
+		if (tmp != NULL) {
+			tm = *tmp;
+			(void) strncpy(buf, abbr(&tm), sizeof (buf) - 1);
+		}
+		for (;;) {
+			if (t >= cuthitime)
+				break;
+			/* check if newt will overrun maximum time_t value */
+			if (t > LONG_MAX - (SECSPERHOUR * 12))
+				break;
+			newt = t + SECSPERHOUR * 12;
+			if (newt >= cuthitime)
+				break;
+			newtmp = localtime(&newt);
+			if (newtmp != NULL)
+				newtm = *newtmp;
+			if ((tmp == NULL || newtmp == NULL) ? (tmp != newtmp) :
+				(delta(&newtm, &tm) != (newt - t) ||
+				newtm.tm_isdst != tm.tm_isdst ||
+				strcmp(abbr(&newtm), buf) != 0)) {
+					newt = hunt(argv[i], t, newt);
+					newtmp = localtime(&newt);
+					if (newtmp != NULL) {
+						newtm = *newtmp;
+						(void) strncpy(buf,
+							abbr(&newtm),
+							sizeof (buf) - 1);
+					}
+			}
+			t = newt;
+			tm = newtm;
+			tmp = newtmp;
+		}
+		t = absolute_max_time;
+#if defined(sun)
+		show(argv[i], t, TRUE);
+		t -= SECSPERHOUR * HOURSPERDAY;
+		show(argv[i], t, TRUE);
+#else /* !defined(sun) */
+		t -= SECSPERHOUR * HOURSPERDAY;
+		show(argv[i], t, TRUE);
+		t += SECSPERHOUR * HOURSPERDAY;
+		show(argv[i], t, TRUE);
+#endif /* !defined(sun) */
+	}
+	if (fflush(stdout) || ferror(stdout)) {
+		(void) fprintf(stderr, "%s: ", progname);
+		(void) perror(gettext("Error writing standard output"));
+		exit(EXIT_FAILURE);
+	}
+	return (EXIT_SUCCESS);
+}
+
+static void
+setabsolutes()
+{
+#if defined(sun)
+	absolute_min_time = LONG_MIN;
+	absolute_max_time = LONG_MAX;
+#else
+	if (0.5 == (time_t)0.5) {
+		/*
+		 * time_t is floating.
+		 */
+		if (sizeof (time_t) == sizeof (float)) {
+			absolute_min_time = (time_t)-FLT_MAX;
+			absolute_max_time = (time_t)FLT_MAX;
+		} else if (sizeof (time_t) == sizeof (double)) {
+			absolute_min_time = (time_t)-DBL_MAX;
+			absolute_max_time = (time_t)DBL_MAX;
+		} else {
+			(void) fprintf(stderr, gettext("%s: use of -v on "
+			    "system with floating time_t other than float "
+			    "or double\n"), progname);
+			exit(EXIT_FAILURE);
+		}
+	} else
+	/*CONSTANTCONDITION*/
+	if (0 > (time_t)-1) {
+		/*
+		 * time_t is signed.
+		 */
+		register time_t	hibit;
+
+		for (hibit = 1; (hibit * 2) != 0; hibit *= 2)
+			continue;
+		absolute_min_time = hibit;
+		absolute_max_time = -(hibit + 1);
+	} else {
+		/*
+		 * time_t is unsigned.
+		 */
+		absolute_min_time = 0;
+		absolute_max_time = absolute_min_time - 1;
+	}
+#endif
+}
+
+static time_t
+yeartot(y)
+const long	y;
+{
+	register long	myy;
+	register long	seconds;
+	register time_t	t;
+
+	myy = EPOCH_YEAR;
+	t = 0;
+	while (myy != y) {
+		if (myy < y) {
+			seconds = isleap(myy) ? SECSPERLYEAR : SECSPERNYEAR;
+			++myy;
+			if (t > absolute_max_time - seconds) {
+				t = absolute_max_time;
+				break;
+			}
+			t += seconds;
+		} else {
+			--myy;
+			seconds = isleap(myy) ? SECSPERLYEAR : SECSPERNYEAR;
+			if (t < absolute_min_time + seconds) {
+				t = absolute_min_time;
+				break;
+			}
+			t -= seconds;
+		}
+	}
+	return (t);
+}
+
+static time_t
+hunt(name, lot, hit)
+char	*name;
+time_t	lot;
+time_t	hit;
+{
+	time_t			t;
+	long			diff;
+	struct tm		lotm;
+	register struct tm	*lotmp;
+	struct tm		tm;
+	register struct tm	*tmp;
+	char			loab[MAX_STRING_LENGTH];
+
+	lotmp = my_localtime(&lot);
+	if (lotmp != NULL) {
+		lotm = *lotmp;
+		(void) strncpy(loab, abbr(&lotm), sizeof (loab) - 1);
+	}
+	for (;;) {
+		diff = (long)(hit - lot);
+		if (diff < 2)
+			break;
+		t = lot;
+		t += diff / 2;
+		if (t <= lot)
+			++t;
+		else if (t >= hit)
+			--t;
+		tmp = my_localtime(&t);
+		if (tmp != NULL)
+			tm = *tmp;
+		if ((lotmp == NULL || tmp == NULL) ? (lotmp == tmp) :
+			(delta(&tm, &lotm) == (t - lot) &&
+			tm.tm_isdst == lotm.tm_isdst &&
+			strcmp(abbr(&tm), loab) == 0)) {
+				lot = t;
+				lotm = tm;
+				lotmp = tmp;
+		} else	hit = t;
+	}
+	show(name, lot, TRUE);
+	show(name, hit, TRUE);
+	return (hit);
+}
+
+/*
+ * Thanks to Paul Eggert for logic used in delta.
+ */
+
+static long
+delta(newp, oldp)
+struct tm	*newp;
+struct tm	*oldp;
+{
+	register long	result;
+	register int	tmy;
+
+	if (newp->tm_year < oldp->tm_year)
+		return (-delta(oldp, newp));
+	result = 0;
+	for (tmy = oldp->tm_year; tmy < newp->tm_year; ++tmy)
+		result += DAYSPERNYEAR + isleap_sum(tmy, TM_YEAR_BASE);
+	result += newp->tm_yday - oldp->tm_yday;
+	result *= HOURSPERDAY;
+	result += newp->tm_hour - oldp->tm_hour;
+	result *= MINSPERHOUR;
+	result += newp->tm_min - oldp->tm_min;
+	result *= SECSPERMIN;
+	result += newp->tm_sec - oldp->tm_sec;
+	return (result);
+}
+
+static void
+show(zone, t, v)
+char	*zone;
+time_t	t;
+int	v;
+{
+	register struct tm	*tmp;
+
+	(void) printf("%-*s  ", (int)longest, zone);
+	if (v) {
+		tmp = gmtime(&t);
+		if (tmp == NULL) {
+			(void) printf(tformat(), t);
+		} else {
+			dumptime(tmp);
+			(void) printf(" UTC");
+		}
+		(void) printf(" = ");
+	}
+	tmp = my_localtime(&t);
+	dumptime(tmp);
+	if (tmp != NULL) {
+		if (*abbr(tmp) != '\0')
+			(void) printf(" %s", abbr(tmp));
+		if (v) {
+			(void) printf(" isdst=%d", tmp->tm_isdst);
+#ifdef TM_GMTOFF
+			(void) printf(" gmtoff=%ld", tmp->TM_GMTOFF);
+#endif /* defined TM_GMTOFF */
+		}
+	}
+	(void) printf("\n");
+	if (tmp != NULL && *abbr(tmp) != '\0')
+		abbrok(abbr(tmp), zone);
+}
+
+static char *
+abbr(tmp)
+struct tm	*tmp;
+{
+	register char	*result;
+	static char	nada;
+
+	if (tmp->tm_isdst != 0 && tmp->tm_isdst != 1)
+		return (&nada);
+	result = tzname[tmp->tm_isdst];
+	return ((result == NULL) ? &nada : result);
+}
+
+/*
+ * The code below can fail on certain theoretical systems;
+ * it works on all known real-world systems as of 2004-12-30.
+ */
+
+static const char *
+tformat()
+{
+#if defined(sun)
+	/* time_t is signed long */
+	return ("%ld");
+#else
+	/*CONSTANTCONDITION*/
+	if (0.5 == (time_t)0.5) {	/* floating */
+		/*CONSTANTCONDITION*/
+		if (sizeof (time_t) > sizeof (double))
+			return ("%Lg");
+		return ("%g");
+	}
+	/*CONSTANTCONDITION*/
+	if (0 > (time_t)-1) {		/* signed */
+		/*CONSTANTCONDITION*/
+		if (sizeof (time_t) > sizeof (long))
+			return ("%lld");
+		/*CONSTANTCONDITION*/
+		if (sizeof (time_t) > sizeof (int))
+			return ("%ld");
+		return ("%d");
+	}
+	/*CONSTANTCONDITION*/
+	if (sizeof (time_t) > sizeof (unsigned long))
+		return ("%llu");
+	/*CONSTANTCONDITION*/
+	if (sizeof (time_t) > sizeof (unsigned int))
+		return ("%lu");
+	return ("%u");
+#endif
+}
+
+static void
+dumptime(timeptr)
+register const struct tm	*timeptr;
+{
+	static const char	wday_name[][3] = {
+		"Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"
+	};
+	static const char	mon_name[][3] = {
+		"Jan", "Feb", "Mar", "Apr", "May", "Jun",
+		"Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
+	};
+	register const char	*wn;
+	register const char	*mn;
+	register int		lead;
+	register int		trail;
+
+	if (timeptr == NULL) {
+		(void) printf("NULL");
+		return;
+	}
+	/*
+	 * The packaged versions of localtime and gmtime never put out-of-range
+	 * values in tm_wday or tm_mon, but since this code might be compiled
+	 * with other (perhaps experimental) versions, paranoia is in order.
+	 */
+	if (timeptr->tm_wday < 0 || timeptr->tm_wday >=
+		(int)(sizeof (wday_name) / sizeof (wday_name[0])))
+			wn = "???";
+	else		wn = wday_name[timeptr->tm_wday];
+	if (timeptr->tm_mon < 0 || timeptr->tm_mon >=
+		(int)(sizeof (mon_name) / sizeof (mon_name[0])))
+			mn = "???";
+	else		mn = mon_name[timeptr->tm_mon];
+	(void) printf("%.3s %.3s%3d %.2d:%.2d:%.2d ",
+		wn, mn,
+		timeptr->tm_mday, timeptr->tm_hour,
+		timeptr->tm_min, timeptr->tm_sec);
+#define	DIVISOR	10
+	trail = timeptr->tm_year % DIVISOR + TM_YEAR_BASE % DIVISOR;
+	lead = timeptr->tm_year / DIVISOR + TM_YEAR_BASE / DIVISOR +
+		trail / DIVISOR;
+	trail %= DIVISOR;
+	if (trail < 0 && lead > 0) {
+		trail += DIVISOR;
+		--lead;
+	} else if (lead < 0 && trail > 0) {
+		trail -= DIVISOR;
+		++lead;
+	}
+	if (lead == 0)
+		(void) printf("%d", trail);
+	else
+		(void) printf("%d%d", lead, ((trail < 0) ? -trail : trail));
+}
+
+static void
+usage()
+{
+	(void) fprintf(stderr, gettext(
+	    "%s: [ --version ] [ -v ] [ -c [loyear,]hiyear ] zonename ...\n"),
+		progname);
+	exit(EXIT_FAILURE);
+	/* NOTREACHED */
+}

cmd/zfs/zfs_iter.c

@@ -0,0 +1,420 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <libintl.h>
+#include <libuutil.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <strings.h>
+
+#include <libzfs.h>
+
+#include "zfs_util.h"
+#include "zfs_iter.h"
+
+/*
+ * This is a private interface used to gather up all the datasets specified on
+ * the command line so that we can iterate over them in order.
+ *
+ * First, we iterate over all filesystems, gathering them together into an
+ * AVL tree.  We report errors for any explicitly specified datasets
+ * that we couldn't open.
+ *
+ * When finished, we have an AVL tree of ZFS handles.  We go through and execute
+ * the provided callback for each one, passing whatever data the user supplied.
+ */
+
+typedef struct zfs_node {
+	zfs_handle_t	*zn_handle;
+	uu_avl_node_t	zn_avlnode;
+} zfs_node_t;
+
+typedef struct callback_data {
+	uu_avl_t	*cb_avl;
+	int		cb_flags;
+	zfs_type_t	cb_types;
+	zfs_sort_column_t *cb_sortcol;
+	zprop_list_t	**cb_proplist;
+} callback_data_t;
+
+uu_avl_pool_t *avl_pool;
+
+/*
+ * Include snaps if they were requested or if this a zfs list where types
+ * were not specified and the "listsnapshots" property is set on this pool.
+ */
+static int
+zfs_include_snapshots(zfs_handle_t *zhp, callback_data_t *cb)
+{
+	zpool_handle_t *zph;
+
+	if ((cb->cb_flags & ZFS_ITER_PROP_LISTSNAPS) == 0)
+		return (cb->cb_types & ZFS_TYPE_SNAPSHOT);
+
+	zph = zfs_get_pool_handle(zhp);
+	return (zpool_get_prop_int(zph, ZPOOL_PROP_LISTSNAPS, NULL));
+}
+
+/*
+ * Called for each dataset.  If the object is of an appropriate type,
+ * add it to the avl tree and recurse over any children as necessary.
+ */
+static int
+zfs_callback(zfs_handle_t *zhp, void *data)
+{
+	callback_data_t *cb = data;
+	int dontclose = 0;
+	int include_snaps = zfs_include_snapshots(zhp, cb);
+
+	if ((zfs_get_type(zhp) & cb->cb_types) ||
+	    ((zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT) && include_snaps)) {
+		uu_avl_index_t idx;
+		zfs_node_t *node = safe_malloc(sizeof (zfs_node_t));
+
+		node->zn_handle = zhp;
+		uu_avl_node_init(node, &node->zn_avlnode, avl_pool);
+		if (uu_avl_find(cb->cb_avl, node, cb->cb_sortcol,
+		    &idx) == NULL) {
+			if (cb->cb_proplist &&
+			    zfs_expand_proplist(zhp, cb->cb_proplist) != 0) {
+				free(node);
+				return (-1);
+			}
+			uu_avl_insert(cb->cb_avl, node, idx);
+			dontclose = 1;
+		} else {
+			free(node);
+		}
+	}
+
+	/*
+	 * Recurse if necessary.
+	 */
+	if (cb->cb_flags & ZFS_ITER_RECURSE) {
+		if (zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM)
+			(void) zfs_iter_filesystems(zhp, zfs_callback, data);
+		if ((zfs_get_type(zhp) != ZFS_TYPE_SNAPSHOT) && include_snaps)
+			(void) zfs_iter_snapshots(zhp, zfs_callback, data);
+	}
+
+	if (!dontclose)
+		zfs_close(zhp);
+
+	return (0);
+}
+
+int
+zfs_add_sort_column(zfs_sort_column_t **sc, const char *name,
+    boolean_t reverse)
+{
+	zfs_sort_column_t *col;
+	zfs_prop_t prop;
+
+	if ((prop = zfs_name_to_prop(name)) == ZPROP_INVAL &&
+	    !zfs_prop_user(name))
+		return (-1);
+
+	col = safe_malloc(sizeof (zfs_sort_column_t));
+
+	col->sc_prop = prop;
+	col->sc_reverse = reverse;
+	if (prop == ZPROP_INVAL) {
+		col->sc_user_prop = safe_malloc(strlen(name) + 1);
+		(void) strcpy(col->sc_user_prop, name);
+	}
+
+	if (*sc == NULL) {
+		col->sc_last = col;
+		*sc = col;
+	} else {
+		(*sc)->sc_last->sc_next = col;
+		(*sc)->sc_last = col;
+	}
+
+	return (0);
+}
+
+void
+zfs_free_sort_columns(zfs_sort_column_t *sc)
+{
+	zfs_sort_column_t *col;
+
+	while (sc != NULL) {
+		col = sc->sc_next;
+		free(sc->sc_user_prop);
+		free(sc);
+		sc = col;
+	}
+}
+
+/* ARGSUSED */
+static int
+zfs_compare(const void *larg, const void *rarg, void *unused)
+{
+	zfs_handle_t *l = ((zfs_node_t *)larg)->zn_handle;
+	zfs_handle_t *r = ((zfs_node_t *)rarg)->zn_handle;
+	const char *lname = zfs_get_name(l);
+	const char *rname = zfs_get_name(r);
+	char *lat, *rat;
+	uint64_t lcreate, rcreate;
+	int ret;
+
+	lat = (char *)strchr(lname, '@');
+	rat = (char *)strchr(rname, '@');
+
+	if (lat != NULL)
+		*lat = '\0';
+	if (rat != NULL)
+		*rat = '\0';
+
+	ret = strcmp(lname, rname);
+	if (ret == 0) {
+		/*
+		 * If we're comparing a dataset to one of its snapshots, we
+		 * always make the full dataset first.
+		 */
+		if (lat == NULL) {
+			ret = -1;
+		} else if (rat == NULL) {
+			ret = 1;
+		} else {
+			/*
+			 * If we have two snapshots from the same dataset, then
+			 * we want to sort them according to creation time.  We
+			 * use the hidden CREATETXG property to get an absolute
+			 * ordering of snapshots.
+			 */
+			lcreate = zfs_prop_get_int(l, ZFS_PROP_CREATETXG);
+			rcreate = zfs_prop_get_int(r, ZFS_PROP_CREATETXG);
+
+			if (lcreate < rcreate)
+				ret = -1;
+			else if (lcreate > rcreate)
+				ret = 1;
+		}
+	}
+
+	if (lat != NULL)
+		*lat = '@';
+	if (rat != NULL)
+		*rat = '@';
+
+	return (ret);
+}
+
+/*
+ * Sort datasets by specified columns.
+ *
+ * o  Numeric types sort in ascending order.
+ * o  String types sort in alphabetical order.
+ * o  Types inappropriate for a row sort that row to the literal
+ *    bottom, regardless of the specified ordering.
+ *
+ * If no sort columns are specified, or two datasets compare equally
+ * across all specified columns, they are sorted alphabetically by name
+ * with snapshots grouped under their parents.
+ */
+static int
+zfs_sort(const void *larg, const void *rarg, void *data)
+{
+	zfs_handle_t *l = ((zfs_node_t *)larg)->zn_handle;
+	zfs_handle_t *r = ((zfs_node_t *)rarg)->zn_handle;
+	zfs_sort_column_t *sc = (zfs_sort_column_t *)data;
+	zfs_sort_column_t *psc;
+
+	for (psc = sc; psc != NULL; psc = psc->sc_next) {
+		char lbuf[ZFS_MAXPROPLEN], rbuf[ZFS_MAXPROPLEN];
+		char *lstr, *rstr;
+		uint64_t lnum, rnum;
+		boolean_t lvalid, rvalid;
+		int ret = 0;
+
+		/*
+		 * We group the checks below the generic code.  If 'lstr' and
+		 * 'rstr' are non-NULL, then we do a string based comparison.
+		 * Otherwise, we compare 'lnum' and 'rnum'.
+		 */
+		lstr = rstr = NULL;
+		if (psc->sc_prop == ZPROP_INVAL) {
+			nvlist_t *luser, *ruser;
+			nvlist_t *lval, *rval;
+
+			luser = zfs_get_user_props(l);
+			ruser = zfs_get_user_props(r);
+
+			lvalid = (nvlist_lookup_nvlist(luser,
+			    psc->sc_user_prop, &lval) == 0);
+			rvalid = (nvlist_lookup_nvlist(ruser,
+			    psc->sc_user_prop, &rval) == 0);
+
+			if (lvalid)
+				verify(nvlist_lookup_string(lval,
+				    ZPROP_VALUE, &lstr) == 0);
+			if (rvalid)
+				verify(nvlist_lookup_string(rval,
+				    ZPROP_VALUE, &rstr) == 0);
+
+		} else if (zfs_prop_is_string(psc->sc_prop)) {
+			lvalid = (zfs_prop_get(l, psc->sc_prop, lbuf,
+			    sizeof (lbuf), NULL, NULL, 0, B_TRUE) == 0);
+			rvalid = (zfs_prop_get(r, psc->sc_prop, rbuf,
+			    sizeof (rbuf), NULL, NULL, 0, B_TRUE) == 0);
+
+			lstr = lbuf;
+			rstr = rbuf;
+		} else {
+			lvalid = zfs_prop_valid_for_type(psc->sc_prop,
+			    zfs_get_type(l));
+			rvalid = zfs_prop_valid_for_type(psc->sc_prop,
+			    zfs_get_type(r));
+
+			if (lvalid)
+				(void) zfs_prop_get_numeric(l, psc->sc_prop,
+				    &lnum, NULL, NULL, 0);
+			if (rvalid)
+				(void) zfs_prop_get_numeric(r, psc->sc_prop,
+				    &rnum, NULL, NULL, 0);
+		}
+
+		if (!lvalid && !rvalid)
+			continue;
+		else if (!lvalid)
+			return (1);
+		else if (!rvalid)
+			return (-1);
+
+		if (lstr)
+			ret = strcmp(lstr, rstr);
+		else if (lnum < rnum)
+			ret = -1;
+		else if (lnum > rnum)
+			ret = 1;
+
+		if (ret != 0) {
+			if (psc->sc_reverse == B_TRUE)
+				ret = (ret < 0) ? 1 : -1;
+			return (ret);
+		}
+	}
+
+	return (zfs_compare(larg, rarg, NULL));
+}
+
+int
+zfs_for_each(int argc, char **argv, int flags, zfs_type_t types,
+    zfs_sort_column_t *sortcol, zprop_list_t **proplist,
+    zfs_iter_f callback, void *data)
+{
+	callback_data_t cb;
+	int ret = 0;
+	zfs_node_t *node;
+	uu_avl_walk_t *walk;
+
+	avl_pool = uu_avl_pool_create("zfs_pool", sizeof (zfs_node_t),
+	    offsetof(zfs_node_t, zn_avlnode), zfs_sort, UU_DEFAULT);
+
+	if (avl_pool == NULL) {
+		(void) fprintf(stderr,
+		    gettext("internal error: out of memory\n"));
+		exit(1);
+	}
+
+	cb.cb_sortcol = sortcol;
+	cb.cb_flags = flags;
+	cb.cb_proplist = proplist;
+	cb.cb_types = types;
+	if ((cb.cb_avl = uu_avl_create(avl_pool, NULL, UU_DEFAULT)) == NULL) {
+		(void) fprintf(stderr,
+		    gettext("internal error: out of memory\n"));
+		exit(1);
+	}
+
+	if (argc == 0) {
+		/*
+		 * If given no arguments, iterate over all datasets.
+		 */
+		cb.cb_flags |= ZFS_ITER_RECURSE;
+		ret = zfs_iter_root(g_zfs, zfs_callback, &cb);
+	} else {
+		int i;
+		zfs_handle_t *zhp;
+		zfs_type_t argtype;
+
+		/*
+		 * If we're recursive, then we always allow filesystems as
+		 * arguments.  If we also are interested in snapshots, then we
+		 * can take volumes as well.
+		 */
+		argtype = types;
+		if (flags & ZFS_ITER_RECURSE) {
+			argtype |= ZFS_TYPE_FILESYSTEM;
+			if (types & ZFS_TYPE_SNAPSHOT)
+				argtype |= ZFS_TYPE_VOLUME;
+		}
+
+		for (i = 0; i < argc; i++) {
+			if (flags & ZFS_ITER_ARGS_CAN_BE_PATHS) {
+				zhp = zfs_path_to_zhandle(g_zfs, argv[i],
+				    argtype);
+			} else {
+				zhp = zfs_open(g_zfs, argv[i], argtype);
+			}
+			if (zhp != NULL)
+				ret |= zfs_callback(zhp, &cb);
+			else
+				ret = 1;
+		}
+	}
+
+	/*
+	 * At this point we've got our AVL tree full of zfs handles, so iterate
+	 * over each one and execute the real user callback.
+	 */
+	for (node = uu_avl_first(cb.cb_avl); node != NULL;
+	    node = uu_avl_next(cb.cb_avl, node))
+		ret |= callback(node->zn_handle, data);
+
+	/*
+	 * Finally, clean up the AVL tree.
+	 */
+	if ((walk = uu_avl_walk_start(cb.cb_avl, UU_WALK_ROBUST)) == NULL) {
+		(void) fprintf(stderr,
+		    gettext("internal error: out of memory"));
+		exit(1);
+	}
+
+	while ((node = uu_avl_walk_next(walk)) != NULL) {
+		uu_avl_remove(cb.cb_avl, node);
+		zfs_close(node->zn_handle);
+		free(node);
+	}
+
+	uu_avl_walk_end(walk);
+	uu_avl_destroy(cb.cb_avl);
+	uu_avl_pool_destroy(avl_pool);
+
+	return (ret);
+}

cmd/zfs/zfs_iter.h

@@ -0,0 +1,54 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef	ZFS_ITER_H
+#define	ZFS_ITER_H
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+typedef struct zfs_sort_column {
+	struct zfs_sort_column	*sc_next;
+	struct zfs_sort_column	*sc_last;
+	zfs_prop_t		sc_prop;
+	char			*sc_user_prop;
+	boolean_t		sc_reverse;
+} zfs_sort_column_t;
+
+#define	ZFS_ITER_RECURSE	   (1 << 0)
+#define	ZFS_ITER_ARGS_CAN_BE_PATHS (1 << 1)
+#define	ZFS_ITER_PROP_LISTSNAPS    (1 << 2)
+
+int zfs_for_each(int, char **, int options, zfs_type_t,
+    zfs_sort_column_t *, zprop_list_t **, zfs_iter_f, void *);
+int zfs_add_sort_column(zfs_sort_column_t **, const char *, boolean_t);
+void zfs_free_sort_columns(zfs_sort_column_t *);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* ZFS_ITER_H */

cmd/zfs/zfs_util.h

@@ -0,0 +1,44 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef	_ZFS_UTIL_H
+#define	_ZFS_UTIL_H
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <libzfs.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+void * safe_malloc(size_t size);
+libzfs_handle_t *g_zfs;
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _ZFS_UTIL_H */

cmd/zinject/translate.c

@@ -0,0 +1,474 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <libzfs.h>
+
+#undef verify	/* both libzfs.h and zfs_context.h want to define this */
+
+#include <sys/zfs_context.h>
+
+#include <errno.h>
+#include <fcntl.h>
+#include <stdarg.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <strings.h>
+#include <sys/file.h>
+#include <sys/mntent.h>
+#include <sys/mnttab.h>
+#include <sys/param.h>
+#include <sys/stat.h>
+
+#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
+#include <sys/dnode.h>
+#include <sys/vdev_impl.h>
+
+#include <sys/mkdev.h>
+
+#include "zinject.h"
+
+extern void kernel_init(int);
+extern void kernel_fini(void);
+
+static int debug;
+
+static void
+ziprintf(const char *fmt, ...)
+{
+	va_list ap;
+
+	if (!debug)
+		return;
+
+	va_start(ap, fmt);
+	(void) vprintf(fmt, ap);
+	va_end(ap);
+}
+
+/*
+ * Given a full path to a file, translate into a dataset name and a relative
+ * path within the dataset.  'dataset' must be at least MAXNAMELEN characters,
+ * and 'relpath' must be at least MAXPATHLEN characters.  We also pass a stat64
+ * buffer, which we need later to get the object ID.
+ */
+static int
+parse_pathname(const char *fullpath, char *dataset, char *relpath,
+    struct stat64 *statbuf)
+{
+	struct extmnttab mp;
+	FILE *fp;
+	int match;
+	const char *rel;
+
+	if (fullpath[0] != '/') {
+		(void) fprintf(stderr, "invalid object '%s': must be full "
+		    "path\n", fullpath);
+		usage();
+		return (-1);
+	}
+
+	if (strlen(fullpath) >= MAXPATHLEN) {
+		(void) fprintf(stderr, "invalid object; pathname too long\n");
+		return (-1);
+	}
+
+	if (stat64(fullpath, statbuf) != 0) {
+		(void) fprintf(stderr, "cannot open '%s': %s\n",
+		    fullpath, strerror(errno));
+		return (-1);
+	}
+
+	if ((fp = fopen(MNTTAB, "r")) == NULL) {
+		(void) fprintf(stderr, "cannot open /etc/mnttab\n");
+		return (-1);
+	}
+
+	match = 0;
+	while (getextmntent(fp, &mp, sizeof (mp)) == 0) {
+		if (makedev(mp.mnt_major, mp.mnt_minor) == statbuf->st_dev) {
+			match = 1;
+			break;
+		}
+	}
+
+	if (!match) {
+		(void) fprintf(stderr, "cannot find mountpoint for '%s'\n",
+		    fullpath);
+		return (-1);
+	}
+
+	if (strcmp(mp.mnt_fstype, MNTTYPE_ZFS) != 0) {
+		(void) fprintf(stderr, "invalid path '%s': not a ZFS "
+		    "filesystem\n", fullpath);
+		return (-1);
+	}
+
+	if (strncmp(fullpath, mp.mnt_mountp, strlen(mp.mnt_mountp)) != 0) {
+		(void) fprintf(stderr, "invalid path '%s': mountpoint "
+		    "doesn't match path\n", fullpath);
+		return (-1);
+	}
+
+	(void) strcpy(dataset, mp.mnt_special);
+
+	rel = fullpath + strlen(mp.mnt_mountp);
+	if (rel[0] == '/')
+		rel++;
+	(void) strcpy(relpath, rel);
+
+	return (0);
+}
+
+/*
+ * Convert from a (dataset, path) pair into a (objset, object) pair.  Note that
+ * we grab the object number from the inode number, since looking this up via
+ * libzpool is a real pain.
+ */
+/* ARGSUSED */
+static int
+object_from_path(const char *dataset, const char *path, struct stat64 *statbuf,
+    zinject_record_t *record)
+{
+	objset_t *os;
+	int err;
+
+	/*
+	 * Before doing any libzpool operations, call sync() to ensure that the
+	 * on-disk state is consistent with the in-core state.
+	 */
+	sync();
+
+	if ((err = dmu_objset_open(dataset, DMU_OST_ZFS,
+	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0) {
+		(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
+		    dataset, strerror(err));
+		return (-1);
+	}
+
+	record->zi_objset = dmu_objset_id(os);
+	record->zi_object = statbuf->st_ino;
+
+	dmu_objset_close(os);
+
+	return (0);
+}
+
+/*
+ * Calculate the real range based on the type, level, and range given.
+ */
+static int
+calculate_range(const char *dataset, err_type_t type, int level, char *range,
+    zinject_record_t *record)
+{
+	objset_t *os = NULL;
+	dnode_t *dn = NULL;
+	int err;
+	int ret = -1;
+
+	/*
+	 * Determine the numeric range from the string.
+	 */
+	if (range == NULL) {
+		/*
+		 * If range is unspecified, set the range to [0,-1], which
+		 * indicates that the whole object should be treated as an
+		 * error.
+		 */
+		record->zi_start = 0;
+		record->zi_end = -1ULL;
+	} else {
+		char *end;
+
+		/* XXX add support for suffixes */
+		record->zi_start = strtoull(range, &end, 10);
+
+
+		if (*end == '\0')
+			record->zi_end = record->zi_start + 1;
+		else if (*end == ',')
+			record->zi_end = strtoull(end + 1, &end, 10);
+
+		if (*end != '\0') {
+			(void) fprintf(stderr, "invalid range '%s': must be "
+			    "a numeric range of the form 'start[,end]'\n",
+			    range);
+			goto out;
+		}
+	}
+
+	switch (type) {
+	case TYPE_DATA:
+		break;
+
+	case TYPE_DNODE:
+		/*
+		 * If this is a request to inject faults into the dnode, then we
+		 * must translate the current (objset,object) pair into an
+		 * offset within the metadnode for the objset.  Specifying any
+		 * kind of range with type 'dnode' is illegal.
+		 */
+		if (range != NULL) {
+			(void) fprintf(stderr, "range cannot be specified when "
+			    "type is 'dnode'\n");
+			goto out;
+		}
+
+		record->zi_start = record->zi_object * sizeof (dnode_phys_t);
+		record->zi_end = record->zi_start + sizeof (dnode_phys_t);
+		record->zi_object = 0;
+		break;
+	}
+
+	/*
+	 * Get the dnode associated with object, so we can calculate the block
+	 * size.
+	 */
+	if ((err = dmu_objset_open(dataset, DMU_OST_ANY,
+	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0) {
+		(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
+		    dataset, strerror(err));
+		goto out;
+	}
+
+	if (record->zi_object == 0) {
+		dn = os->os->os_meta_dnode;
+	} else {
+		err = dnode_hold(os->os, record->zi_object, FTAG, &dn);
+		if (err != 0) {
+			(void) fprintf(stderr, "failed to hold dnode "
+			    "for object %llu\n",
+			    (u_longlong_t)record->zi_object);
+			goto out;
+		}
+	}
+
+
+	ziprintf("data shift: %d\n", (int)dn->dn_datablkshift);
+	ziprintf(" ind shift: %d\n", (int)dn->dn_indblkshift);
+
+	/*
+	 * Translate range into block IDs.
+	 */
+	if (record->zi_start != 0 || record->zi_end != -1ULL) {
+		record->zi_start >>= dn->dn_datablkshift;
+		record->zi_end >>= dn->dn_datablkshift;
+	}
+
+	/*
+	 * Check level, and then translate level 0 blkids into ranges
+	 * appropriate for level of indirection.
+	 */
+	record->zi_level = level;
+	if (level > 0) {
+		ziprintf("level 0 blkid range: [%llu, %llu]\n",
+		    record->zi_start, record->zi_end);
+
+		if (level >= dn->dn_nlevels) {
+			(void) fprintf(stderr, "level %d exceeds max level "
+			    "of object (%d)\n", level, dn->dn_nlevels - 1);
+			goto out;
+		}
+
+		if (record->zi_start != 0 || record->zi_end != 0) {
+			int shift = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+
+			for (; level > 0; level--) {
+				record->zi_start >>= shift;
+				record->zi_end >>= shift;
+			}
+		}
+	}
+
+	ret = 0;
+out:
+	if (dn) {
+		if (dn != os->os->os_meta_dnode)
+			dnode_rele(dn, FTAG);
+	}
+	if (os)
+		dmu_objset_close(os);
+
+	return (ret);
+}
+
+int
+translate_record(err_type_t type, const char *object, const char *range,
+    int level, zinject_record_t *record, char *poolname, char *dataset)
+{
+	char path[MAXPATHLEN];
+	char *slash;
+	struct stat64 statbuf;
+	int ret = -1;
+
+	kernel_init(FREAD);
+
+	debug = (getenv("ZINJECT_DEBUG") != NULL);
+
+	ziprintf("translating: %s\n", object);
+
+	if (MOS_TYPE(type)) {
+		/*
+		 * MOS objects are treated specially.
+		 */
+		switch (type) {
+		case TYPE_MOS:
+			record->zi_type = 0;
+			break;
+		case TYPE_MOSDIR:
+			record->zi_type = DMU_OT_OBJECT_DIRECTORY;
+			break;
+		case TYPE_METASLAB:
+			record->zi_type = DMU_OT_OBJECT_ARRAY;
+			break;
+		case TYPE_CONFIG:
+			record->zi_type = DMU_OT_PACKED_NVLIST;
+			break;
+		case TYPE_BPLIST:
+			record->zi_type = DMU_OT_BPLIST;
+			break;
+		case TYPE_SPACEMAP:
+			record->zi_type = DMU_OT_SPACE_MAP;
+			break;
+		case TYPE_ERRLOG:
+			record->zi_type = DMU_OT_ERROR_LOG;
+			break;
+		}
+
+		dataset[0] = '\0';
+		(void) strcpy(poolname, object);
+		return (0);
+	}
+
+	/*
+	 * Convert a full path into a (dataset, file) pair.
+	 */
+	if (parse_pathname(object, dataset, path, &statbuf) != 0)
+		goto err;
+
+	ziprintf("   dataset: %s\n", dataset);
+	ziprintf("      path: %s\n", path);
+
+	/*
+	 * Convert (dataset, file) into (objset, object)
+	 */
+	if (object_from_path(dataset, path, &statbuf, record) != 0)
+		goto err;
+
+	ziprintf("raw objset: %llu\n", record->zi_objset);
+	ziprintf("raw object: %llu\n", record->zi_object);
+
+	/*
+	 * For the given object, calculate the real (type, level, range)
+	 */
+	if (calculate_range(dataset, type, level, (char *)range, record) != 0)
+		goto err;
+
+	ziprintf("    objset: %llu\n", record->zi_objset);
+	ziprintf("    object: %llu\n", record->zi_object);
+	if (record->zi_start == 0 &&
+	    record->zi_end == -1ULL)
+		ziprintf("     range: all\n");
+	else
+		ziprintf("     range: [%llu, %llu]\n", record->zi_start,
+		    record->zi_end);
+
+	/*
+	 * Copy the pool name
+	 */
+	(void) strcpy(poolname, dataset);
+	if ((slash = strchr(poolname, '/')) != NULL)
+		*slash = '\0';
+
+	ret = 0;
+
+err:
+	kernel_fini();
+	return (ret);
+}
+
+int
+translate_raw(const char *str, zinject_record_t *record)
+{
+	/*
+	 * A raw bookmark of the form objset:object:level:blkid, where each
+	 * number is a hexidecimal value.
+	 */
+	if (sscanf(str, "%llx:%llx:%x:%llx", (u_longlong_t *)&record->zi_objset,
+	    (u_longlong_t *)&record->zi_object, &record->zi_level,
+	    (u_longlong_t *)&record->zi_start) != 4) {
+		(void) fprintf(stderr, "bad raw spec '%s': must be of the form "
+		    "'objset:object:level:blkid'\n", str);
+		return (-1);
+	}
+
+	record->zi_end = record->zi_start;
+
+	return (0);
+}
+
+int
+translate_device(const char *pool, const char *device, err_type_t label_type,
+    zinject_record_t *record)
+{
+	char *end;
+	zpool_handle_t *zhp;
+	nvlist_t *tgt;
+	boolean_t isspare, iscache;
+
+	/*
+	 * Given a device name or GUID, create an appropriate injection record
+	 * with zi_guid set.
+	 */
+	if ((zhp = zpool_open(g_zfs, pool)) == NULL)
+		return (-1);
+
+	record->zi_guid = strtoull(device, &end, 16);
+	if (record->zi_guid == 0 || *end != '\0') {
+		tgt = zpool_find_vdev(zhp, device, &isspare, &iscache, NULL);
+
+		if (tgt == NULL) {
+			(void) fprintf(stderr, "cannot find device '%s' in "
+			    "pool '%s'\n", device, pool);
+			return (-1);
+		}
+
+		verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
+		    &record->zi_guid) == 0);
+	}
+
+	switch (label_type) {
+	case TYPE_LABEL_UBERBLOCK:
+		record->zi_start = offsetof(vdev_label_t, vl_uberblock[0]);
+		record->zi_end = record->zi_start + VDEV_UBERBLOCK_RING - 1;
+		break;
+	case TYPE_LABEL_NVLIST:
+		record->zi_start = offsetof(vdev_label_t, vl_vdev_phys);
+		record->zi_end = record->zi_start + VDEV_PHYS_SIZE - 1;
+		break;
+	}
+	return (0);
+}

cmd/zinject/zinject.c

@@ -0,0 +1,768 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+/*
+ * ZFS Fault Injector
+ *
+ * This userland component takes a set of options and uses libzpool to translate
+ * from a user-visible object type and name to an internal representation.
+ * There are two basic types of faults: device faults and data faults.
+ *
+ *
+ * DEVICE FAULTS
+ *
+ * Errors can be injected into a particular vdev using the '-d' option.  This
+ * option takes a path or vdev GUID to uniquely identify the device within a
+ * pool.  There are two types of errors that can be injected, EIO and ENXIO,
+ * that can be controlled through the '-e' option.  The default is ENXIO.  For
+ * EIO failures, any attempt to read data from the device will return EIO, but
+ * subsequent attempt to reopen the device will succeed.  For ENXIO failures,
+ * any attempt to read from the device will return EIO, but any attempt to
+ * reopen the device will also return ENXIO.
+ * For label faults, the -L option must be specified. This allows faults
+ * to be injected into either the nvlist or uberblock region of all the labels
+ * for the specified device.
+ *
+ * This form of the command looks like:
+ *
+ * 	zinject -d device [-e errno] [-L <uber | nvlist>] pool
+ *
+ *
+ * DATA FAULTS
+ *
+ * We begin with a tuple of the form:
+ *
+ * 	<type,level,range,object>
+ *
+ * 	type	A string describing the type of data to target.  Each type
+ * 		implicitly describes how to interpret 'object'. Currently,
+ * 		the following values are supported:
+ *
+ * 		data		User data for a file
+ * 		dnode		Dnode for a file or directory
+ *
+ *		The following MOS objects are special.  Instead of injecting
+ *		errors on a particular object or blkid, we inject errors across
+ *		all objects of the given type.
+ *
+ * 		mos		Any data in the MOS
+ * 		mosdir		object directory
+ * 		config		pool configuration
+ * 		bplist		blkptr list
+ * 		spacemap	spacemap
+ * 		metaslab	metaslab
+ * 		errlog		persistent error log
+ *
+ * 	level	Object level.  Defaults to '0', not applicable to all types.  If
+ * 		a range is given, this corresponds to the indirect block
+ * 		corresponding to the specific range.
+ *
+ *	range	A numerical range [start,end) within the object.  Defaults to
+ *		the full size of the file.
+ *
+ * 	object	A string describing the logical location of the object.  For
+ * 		files and directories (currently the only supported types),
+ * 		this is the path of the object on disk.
+ *
+ * This is translated, via libzpool, into the following internal representation:
+ *
+ * 	<type,objset,object,level,range>
+ *
+ * These types should be self-explanatory.  This tuple is then passed to the
+ * kernel via a special ioctl() to initiate fault injection for the given
+ * object.  Note that 'type' is not strictly necessary for fault injection, but
+ * is used when translating existing faults into a human-readable string.
+ *
+ *
+ * The command itself takes one of the forms:
+ *
+ * 	zinject
+ * 	zinject <-a | -u pool>
+ * 	zinject -c <id|all>
+ * 	zinject [-q] <-t type> [-f freq] [-u] [-a] [-m] [-e errno] [-l level]
+ *	    [-r range] <object>
+ * 	zinject [-f freq] [-a] [-m] [-u] -b objset:object:level:start:end pool
+ *
+ * With no arguments, the command prints all currently registered injection
+ * handlers, with their numeric identifiers.
+ *
+ * The '-c' option will clear the given handler, or all handlers if 'all' is
+ * specified.
+ *
+ * The '-e' option takes a string describing the errno to simulate.  This must
+ * be either 'io' or 'checksum'.  In most cases this will result in the same
+ * behavior, but RAID-Z will produce a different set of ereports for this
+ * situation.
+ *
+ * The '-a', '-u', and '-m' flags toggle internal flush behavior.  If '-a' is
+ * specified, then the ARC cache is flushed appropriately.  If '-u' is
+ * specified, then the underlying SPA is unloaded.  Either of these flags can be
+ * specified independently of any other handlers.  The '-m' flag automatically
+ * does an unmount and remount of the underlying dataset to aid in flushing the
+ * cache.
+ *
+ * The '-f' flag controls the frequency of errors injected, expressed as a
+ * integer percentage between 1 and 100.  The default is 100.
+ *
+ * The this form is responsible for actually injecting the handler into the
+ * framework.  It takes the arguments described above, translates them to the
+ * internal tuple using libzpool, and then issues an ioctl() to register the
+ * handler.
+ *
+ * The final form can target a specific bookmark, regardless of whether a
+ * human-readable interface has been designed.  It allows developers to specify
+ * a particular block by number.
+ */
+
+#include <errno.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <strings.h>
+#include <unistd.h>
+
+#include <sys/fs/zfs.h>
+#include <sys/mount.h>
+
+#include <libzfs.h>
+
+#undef verify	/* both libzfs.h and zfs_context.h want to define this */
+
+#include "zinject.h"
+
+libzfs_handle_t *g_zfs;
+int zfs_fd;
+
+#define	ECKSUM	EBADE
+
+static const char *errtable[TYPE_INVAL] = {
+	"data",
+	"dnode",
+	"mos",
+	"mosdir",
+	"metaslab",
+	"config",
+	"bplist",
+	"spacemap",
+	"errlog",
+	"uber",
+	"nvlist"
+};
+
+static err_type_t
+name_to_type(const char *arg)
+{
+	int i;
+	for (i = 0; i < TYPE_INVAL; i++)
+		if (strcmp(errtable[i], arg) == 0)
+			return (i);
+
+	return (TYPE_INVAL);
+}
+
+static const char *
+type_to_name(uint64_t type)
+{
+	switch (type) {
+	case DMU_OT_OBJECT_DIRECTORY:
+		return ("mosdir");
+	case DMU_OT_OBJECT_ARRAY:
+		return ("metaslab");
+	case DMU_OT_PACKED_NVLIST:
+		return ("config");
+	case DMU_OT_BPLIST:
+		return ("bplist");
+	case DMU_OT_SPACE_MAP:
+		return ("spacemap");
+	case DMU_OT_ERROR_LOG:
+		return ("errlog");
+	default:
+		return ("-");
+	}
+}
+
+
+/*
+ * Print usage message.
+ */
+void
+usage(void)
+{
+	(void) printf(
+	    "usage:\n"
+	    "\n"
+	    "\tzinject\n"
+	    "\n"
+	    "\t\tList all active injection records.\n"
+	    "\n"
+	    "\tzinject -c <id|all>\n"
+	    "\n"
+	    "\t\tClear the particular record (if given a numeric ID), or\n"
+	    "\t\tall records if 'all' is specificed.\n"
+	    "\n"
+	    "\tzinject -d device [-e errno] [-L <nvlist|uber>] pool\n"
+	    "\t\tInject a fault into a particular device or the device's\n"
+	    "\t\tlabel.  Label injection can either be 'nvlist' or 'uber'.\n"
+	    "\t\t'errno' can either be 'nxio' (the default) or 'io'.\n"
+	    "\n"
+	    "\tzinject -b objset:object:level:blkid pool\n"
+	    "\n"
+	    "\t\tInject an error into pool 'pool' with the numeric bookmark\n"
+	    "\t\tspecified by the remaining tuple.  Each number is in\n"
+	    "\t\thexidecimal, and only one block can be specified.\n"
+	    "\n"
+	    "\tzinject [-q] <-t type> [-e errno] [-l level] [-r range]\n"
+	    "\t    [-a] [-m] [-u] [-f freq] <object>\n"
+	    "\n"
+	    "\t\tInject an error into the object specified by the '-t' option\n"
+	    "\t\tand the object descriptor.  The 'object' parameter is\n"
+	    "\t\tinterperted depending on the '-t' option.\n"
+	    "\n"
+	    "\t\t-q\tQuiet mode.  Only print out the handler number added.\n"
+	    "\t\t-e\tInject a specific error.  Must be either 'io' or\n"
+	    "\t\t\t'checksum'.  Default is 'io'.\n"
+	    "\t\t-l\tInject error at a particular block level. Default is "
+	    "0.\n"
+	    "\t\t-m\tAutomatically remount underlying filesystem.\n"
+	    "\t\t-r\tInject error over a particular logical range of an\n"
+	    "\t\t\tobject.  Will be translated to the appropriate blkid\n"
+	    "\t\t\trange according to the object's properties.\n"
+	    "\t\t-a\tFlush the ARC cache.  Can be specified without any\n"
+	    "\t\t\tassociated object.\n"
+	    "\t\t-u\tUnload the associated pool.  Can be specified with only\n"
+	    "\t\t\ta pool object.\n"
+	    "\t\t-f\tOnly inject errors a fraction of the time.  Expressed as\n"
+	    "\t\t\ta percentage between 1 and 100.\n"
+	    "\n"
+	    "\t-t data\t\tInject an error into the plain file contents of a\n"
+	    "\t\t\tfile.  The object must be specified as a complete path\n"
+	    "\t\t\tto a file on a ZFS filesystem.\n"
+	    "\n"
+	    "\t-t dnode\tInject an error into the metadnode in the block\n"
+	    "\t\t\tcorresponding to the dnode for a file or directory.  The\n"
+	    "\t\t\t'-r' option is incompatible with this mode.  The object\n"
+	    "\t\t\tis specified as a complete path to a file or directory\n"
+	    "\t\t\ton a ZFS filesystem.\n"
+	    "\n"
+	    "\t-t <mos>\tInject errors into the MOS for objects of the given\n"
+	    "\t\t\ttype.  Valid types are: mos, mosdir, config, bplist,\n"
+	    "\t\t\tspacemap, metaslab, errlog.  The only valid <object> is\n"
+	    "\t\t\tthe poolname.\n");
+}
+
+static int
+iter_handlers(int (*func)(int, const char *, zinject_record_t *, void *),
+    void *data)
+{
+	zfs_cmd_t zc;
+	int ret;
+
+	zc.zc_guid = 0;
+
+	while (ioctl(zfs_fd, ZFS_IOC_INJECT_LIST_NEXT, &zc) == 0)
+		if ((ret = func((int)zc.zc_guid, zc.zc_name,
+		    &zc.zc_inject_record, data)) != 0)
+			return (ret);
+
+	return (0);
+}
+
+static int
+print_data_handler(int id, const char *pool, zinject_record_t *record,
+    void *data)
+{
+	int *count = data;
+
+	if (record->zi_guid != 0)
+		return (0);
+
+	if (*count == 0) {
+		(void) printf("%3s  %-15s  %-6s  %-6s  %-8s  %3s  %-15s\n",
+		    "ID", "POOL", "OBJSET", "OBJECT", "TYPE", "LVL",  "RANGE");
+		(void) printf("---  ---------------  ------  "
+		    "------  --------  ---  ---------------\n");
+	}
+
+	*count += 1;
+
+	(void) printf("%3d  %-15s  %-6llu  %-6llu  %-8s  %3d  ", id, pool,
+	    (u_longlong_t)record->zi_objset, (u_longlong_t)record->zi_object,
+	    type_to_name(record->zi_type), record->zi_level);
+
+	if (record->zi_start == 0 &&
+	    record->zi_end == -1ULL)
+		(void) printf("all\n");
+	else
+		(void) printf("[%llu, %llu]\n", (u_longlong_t)record->zi_start,
+		    (u_longlong_t)record->zi_end);
+
+	return (0);
+}
+
+static int
+print_device_handler(int id, const char *pool, zinject_record_t *record,
+    void *data)
+{
+	int *count = data;
+
+	if (record->zi_guid == 0)
+		return (0);
+
+	if (*count == 0) {
+		(void) printf("%3s  %-15s  %s\n", "ID", "POOL", "GUID");
+		(void) printf("---  ---------------  ----------------\n");
+	}
+
+	*count += 1;
+
+	(void) printf("%3d  %-15s  %llx\n", id, pool,
+	    (u_longlong_t)record->zi_guid);
+
+	return (0);
+}
+
+/*
+ * Print all registered error handlers.  Returns the number of handlers
+ * registered.
+ */
+static int
+print_all_handlers(void)
+{
+	int count = 0;
+
+	(void) iter_handlers(print_device_handler, &count);
+	(void) printf("\n");
+	count = 0;
+	(void) iter_handlers(print_data_handler, &count);
+
+	return (count);
+}
+
+/* ARGSUSED */
+static int
+cancel_one_handler(int id, const char *pool, zinject_record_t *record,
+    void *data)
+{
+	zfs_cmd_t zc;
+
+	zc.zc_guid = (uint64_t)id;
+
+	if (ioctl(zfs_fd, ZFS_IOC_CLEAR_FAULT, &zc) != 0) {
+		(void) fprintf(stderr, "failed to remove handler %d: %s\n",
+		    id, strerror(errno));
+		return (1);
+	}
+
+	return (0);
+}
+
+/*
+ * Remove all fault injection handlers.
+ */
+static int
+cancel_all_handlers(void)
+{
+	int ret = iter_handlers(cancel_one_handler, NULL);
+
+	(void) printf("removed all registered handlers\n");
+
+	return (ret);
+}
+
+/*
+ * Remove a specific fault injection handler.
+ */
+static int
+cancel_handler(int id)
+{
+	zfs_cmd_t zc;
+
+	zc.zc_guid = (uint64_t)id;
+
+	if (ioctl(zfs_fd, ZFS_IOC_CLEAR_FAULT, &zc) != 0) {
+		(void) fprintf(stderr, "failed to remove handler %d: %s\n",
+		    id, strerror(errno));
+		return (1);
+	}
+
+	(void) printf("removed handler %d\n", id);
+
+	return (0);
+}
+
+/*
+ * Register a new fault injection handler.
+ */
+static int
+register_handler(const char *pool, int flags, zinject_record_t *record,
+    int quiet)
+{
+	zfs_cmd_t zc;
+
+	(void) strcpy(zc.zc_name, pool);
+	zc.zc_inject_record = *record;
+	zc.zc_guid = flags;
+
+	if (ioctl(zfs_fd, ZFS_IOC_INJECT_FAULT, &zc) != 0) {
+		(void) fprintf(stderr, "failed to add handler: %s\n",
+		    strerror(errno));
+		return (1);
+	}
+
+	if (flags & ZINJECT_NULL)
+		return (0);
+
+	if (quiet) {
+		(void) printf("%llu\n", (u_longlong_t)zc.zc_guid);
+	} else {
+		(void) printf("Added handler %llu with the following "
+		    "properties:\n", (u_longlong_t)zc.zc_guid);
+		(void) printf("  pool: %s\n", pool);
+		if (record->zi_guid) {
+			(void) printf("  vdev: %llx\n",
+			    (u_longlong_t)record->zi_guid);
+		} else {
+			(void) printf("objset: %llu\n",
+			    (u_longlong_t)record->zi_objset);
+			(void) printf("object: %llu\n",
+			    (u_longlong_t)record->zi_object);
+			(void) printf("  type: %llu\n",
+			    (u_longlong_t)record->zi_type);
+			(void) printf(" level: %d\n", record->zi_level);
+			if (record->zi_start == 0 &&
+			    record->zi_end == -1ULL)
+				(void) printf(" range: all\n");
+			else
+				(void) printf(" range: [%llu, %llu)\n",
+				    (u_longlong_t)record->zi_start,
+				    (u_longlong_t)record->zi_end);
+		}
+	}
+
+	return (0);
+}
+
+int
+main(int argc, char **argv)
+{
+	int c;
+	char *range = NULL;
+	char *cancel = NULL;
+	char *end;
+	char *raw = NULL;
+	char *device = NULL;
+	int level = 0;
+	int quiet = 0;
+	int error = 0;
+	int domount = 0;
+	err_type_t type = TYPE_INVAL;
+	err_type_t label = TYPE_INVAL;
+	zinject_record_t record = { 0 };
+	char pool[MAXNAMELEN];
+	char dataset[MAXNAMELEN];
+	zfs_handle_t *zhp;
+	int ret;
+	int flags = 0;
+
+	if ((g_zfs = libzfs_init()) == NULL) {
+		(void) fprintf(stderr, "internal error: failed to "
+		    "initialize ZFS library\n");
+		return (1);
+	}
+
+	libzfs_print_on_error(g_zfs, B_TRUE);
+
+	if ((zfs_fd = open(ZFS_DEV, O_RDWR)) < 0) {
+		(void) fprintf(stderr, "failed to open ZFS device\n");
+		return (1);
+	}
+
+	if (argc == 1) {
+		/*
+		 * No arguments.  Print the available handlers.  If there are no
+		 * available handlers, direct the user to '-h' for help
+		 * information.
+		 */
+		if (print_all_handlers() == 0) {
+			(void) printf("No handlers registered.\n");
+			(void) printf("Run 'zinject -h' for usage "
+			    "information.\n");
+		}
+
+		return (0);
+	}
+
+	while ((c = getopt(argc, argv, ":ab:d:f:qhc:t:l:mr:e:uL:")) != -1) {
+		switch (c) {
+		case 'a':
+			flags |= ZINJECT_FLUSH_ARC;
+			break;
+		case 'b':
+			raw = optarg;
+			break;
+		case 'c':
+			cancel = optarg;
+			break;
+		case 'd':
+			device = optarg;
+			break;
+		case 'e':
+			if (strcasecmp(optarg, "io") == 0) {
+				error = EIO;
+			} else if (strcasecmp(optarg, "checksum") == 0) {
+				error = ECKSUM;
+			} else if (strcasecmp(optarg, "nxio") == 0) {
+				error = ENXIO;
+			} else {
+				(void) fprintf(stderr, "invalid error type "
+				    "'%s': must be 'io', 'checksum' or "
+				    "'nxio'\n", optarg);
+				usage();
+				return (1);
+			}
+			break;
+		case 'f':
+			record.zi_freq = atoi(optarg);
+			if (record.zi_freq < 1 || record.zi_freq > 100) {
+				(void) fprintf(stderr, "frequency range must "
+				    "be in the range (0, 100]\n");
+				return (1);
+			}
+			break;
+		case 'h':
+			usage();
+			return (0);
+		case 'l':
+			level = (int)strtol(optarg, &end, 10);
+			if (*end != '\0') {
+				(void) fprintf(stderr, "invalid level '%s': "
+				    "must be an integer\n", optarg);
+				usage();
+				return (1);
+			}
+			break;
+		case 'm':
+			domount = 1;
+			break;
+		case 'q':
+			quiet = 1;
+			break;
+		case 'r':
+			range = optarg;
+			break;
+		case 't':
+			if ((type = name_to_type(optarg)) == TYPE_INVAL &&
+			    !MOS_TYPE(type)) {
+				(void) fprintf(stderr, "invalid type '%s'\n",
+				    optarg);
+				usage();
+				return (1);
+			}
+			break;
+		case 'u':
+			flags |= ZINJECT_UNLOAD_SPA;
+			break;
+		case 'L':
+			if ((label = name_to_type(optarg)) == TYPE_INVAL &&
+			    !LABEL_TYPE(type)) {
+				(void) fprintf(stderr, "invalid label type "
+				    "'%s'\n", optarg);
+				usage();
+				return (1);
+			}
+			break;
+		case ':':
+			(void) fprintf(stderr, "option -%c requires an "
+			    "operand\n", optopt);
+			usage();
+			return (1);
+		case '?':
+			(void) fprintf(stderr, "invalid option '%c'\n",
+			    optopt);
+			usage();
+			return (2);
+		}
+	}
+
+	argc -= optind;
+	argv += optind;
+
+	if (cancel != NULL) {
+		/*
+		 * '-c' is invalid with any other options.
+		 */
+		if (raw != NULL || range != NULL || type != TYPE_INVAL ||
+		    level != 0) {
+			(void) fprintf(stderr, "cancel (-c) incompatible with "
+			    "any other options\n");
+			usage();
+			return (2);
+		}
+		if (argc != 0) {
+			(void) fprintf(stderr, "extraneous argument to '-c'\n");
+			usage();
+			return (2);
+		}
+
+		if (strcmp(cancel, "all") == 0) {
+			return (cancel_all_handlers());
+		} else {
+			int id = (int)strtol(cancel, &end, 10);
+			if (*end != '\0') {
+				(void) fprintf(stderr, "invalid handle id '%s':"
+				    " must be an integer or 'all'\n", cancel);
+				usage();
+				return (1);
+			}
+			return (cancel_handler(id));
+		}
+	}
+
+	if (device != NULL) {
+		/*
+		 * Device (-d) injection uses a completely different mechanism
+		 * for doing injection, so handle it separately here.
+		 */
+		if (raw != NULL || range != NULL || type != TYPE_INVAL ||
+		    level != 0) {
+			(void) fprintf(stderr, "device (-d) incompatible with "
+			    "data error injection\n");
+			usage();
+			return (2);
+		}
+
+		if (argc != 1) {
+			(void) fprintf(stderr, "device (-d) injection requires "
+			    "a single pool name\n");
+			usage();
+			return (2);
+		}
+
+		(void) strcpy(pool, argv[0]);
+		dataset[0] = '\0';
+
+		if (error == ECKSUM) {
+			(void) fprintf(stderr, "device error type must be "
+			    "'io' or 'nxio'\n");
+			return (1);
+		}
+
+		if (translate_device(pool, device, label, &record) != 0)
+			return (1);
+		if (!error)
+			error = ENXIO;
+	} else if (raw != NULL) {
+		if (range != NULL || type != TYPE_INVAL || level != 0) {
+			(void) fprintf(stderr, "raw (-b) format with "
+			    "any other options\n");
+			usage();
+			return (2);
+		}
+
+		if (argc != 1) {
+			(void) fprintf(stderr, "raw (-b) format expects a "
+			    "single pool name\n");
+			usage();
+			return (2);
+		}
+
+		(void) strcpy(pool, argv[0]);
+		dataset[0] = '\0';
+
+		if (error == ENXIO) {
+			(void) fprintf(stderr, "data error type must be "
+			    "'checksum' or 'io'\n");
+			return (1);
+		}
+
+		if (translate_raw(raw, &record) != 0)
+			return (1);
+		if (!error)
+			error = EIO;
+	} else if (type == TYPE_INVAL) {
+		if (flags == 0) {
+			(void) fprintf(stderr, "at least one of '-b', '-d', "
+			    "'-t', '-a', or '-u' must be specified\n");
+			usage();
+			return (2);
+		}
+
+		if (argc == 1 && (flags & ZINJECT_UNLOAD_SPA)) {
+			(void) strcpy(pool, argv[0]);
+			dataset[0] = '\0';
+		} else if (argc != 0) {
+			(void) fprintf(stderr, "extraneous argument for "
+			    "'-f'\n");
+			usage();
+			return (2);
+		}
+
+		flags |= ZINJECT_NULL;
+	} else {
+		if (argc != 1) {
+			(void) fprintf(stderr, "missing object\n");
+			usage();
+			return (2);
+		}
+
+		if (error == ENXIO) {
+			(void) fprintf(stderr, "data error type must be "
+			    "'checksum' or 'io'\n");
+			return (1);
+		}
+
+		if (translate_record(type, argv[0], range, level, &record, pool,
+		    dataset) != 0)
+			return (1);
+		if (!error)
+			error = EIO;
+	}
+
+	/*
+	 * If this is pool-wide metadata, unmount everything.  The ioctl() will
+	 * unload the pool, so that we trigger spa-wide reopen of metadata next
+	 * time we access the pool.
+	 */
+	if (dataset[0] != '\0' && domount) {
+		if ((zhp = zfs_open(g_zfs, dataset, ZFS_TYPE_DATASET)) == NULL)
+			return (1);
+
+		if (zfs_unmount(zhp, NULL, 0) != 0)
+			return (1);
+	}
+
+	record.zi_error = error;
+
+	ret = register_handler(pool, flags, &record, quiet);
+
+	if (dataset[0] != '\0' && domount)
+		ret = (zfs_mount(zhp, NULL, 0) != 0);
+
+	libzfs_fini(g_zfs);
+
+	return (ret);
+}

cmd/zinject/zinject.h

@@ -0,0 +1,71 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef	_ZINJECT_H
+#define	_ZINJECT_H
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <sys/zfs_ioctl.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+typedef enum {
+	TYPE_DATA,		/* plain file contents		*/
+	TYPE_DNODE,		/* metadnode contents		*/
+	TYPE_MOS,		/* all MOS data			*/
+	TYPE_MOSDIR,		/* MOS object directory		*/
+	TYPE_METASLAB,		/* metaslab objects		*/
+	TYPE_CONFIG,		/* MOS config			*/
+	TYPE_BPLIST,		/* block pointer list		*/
+	TYPE_SPACEMAP,		/* space map objects		*/
+	TYPE_ERRLOG,		/* persistent error log		*/
+	TYPE_LABEL_UBERBLOCK,	/* label specific uberblock	*/
+	TYPE_LABEL_NVLIST,	/* label specific nvlist	*/
+	TYPE_INVAL
+} err_type_t;
+
+#define	MOS_TYPE(t)	\
+	((t) >= TYPE_MOS && (t) < TYPE_LABEL_UBERBLOCK)
+
+#define	LABEL_TYPE(t)	\
+	((t) >= TYPE_LABEL_UBERBLOCK && (t) < TYPE_INVAL)
+
+int translate_record(err_type_t type, const char *object, const char *range,
+    int level, zinject_record_t *record, char *poolname, char *dataset);
+int translate_raw(const char *raw, zinject_record_t *record);
+int translate_device(const char *pool, const char *device,
+    err_type_t label_type, zinject_record_t *record);
+void usage(void);
+
+extern libzfs_handle_t *g_zfs;
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _ZINJECT_H */

cmd/zpool/zpool_iter.c

@@ -0,0 +1,252 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <libintl.h>
+#include <libuutil.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <strings.h>
+
+#include <libzfs.h>
+
+#include "zpool_util.h"
+
+/*
+ * Private interface for iterating over pools specified on the command line.
+ * Most consumers will call for_each_pool, but in order to support iostat, we
+ * allow fined grained control through the zpool_list_t interface.
+ */
+
+typedef struct zpool_node {
+	zpool_handle_t	*zn_handle;
+	uu_avl_node_t	zn_avlnode;
+	int		zn_mark;
+} zpool_node_t;
+
+struct zpool_list {
+	boolean_t	zl_findall;
+	uu_avl_t	*zl_avl;
+	uu_avl_pool_t	*zl_pool;
+	zprop_list_t	**zl_proplist;
+};
+
+/* ARGSUSED */
+static int
+zpool_compare(const void *larg, const void *rarg, void *unused)
+{
+	zpool_handle_t *l = ((zpool_node_t *)larg)->zn_handle;
+	zpool_handle_t *r = ((zpool_node_t *)rarg)->zn_handle;
+	const char *lname = zpool_get_name(l);
+	const char *rname = zpool_get_name(r);
+
+	return (strcmp(lname, rname));
+}
+
+/*
+ * Callback function for pool_list_get().  Adds the given pool to the AVL tree
+ * of known pools.
+ */
+static int
+add_pool(zpool_handle_t *zhp, void *data)
+{
+	zpool_list_t *zlp = data;
+	zpool_node_t *node = safe_malloc(sizeof (zpool_node_t));
+	uu_avl_index_t idx;
+
+	node->zn_handle = zhp;
+	uu_avl_node_init(node, &node->zn_avlnode, zlp->zl_pool);
+	if (uu_avl_find(zlp->zl_avl, node, NULL, &idx) == NULL) {
+		if (zlp->zl_proplist &&
+		    zpool_expand_proplist(zhp, zlp->zl_proplist) != 0) {
+			zpool_close(zhp);
+			free(node);
+			return (-1);
+		}
+		uu_avl_insert(zlp->zl_avl, node, idx);
+	} else {
+		zpool_close(zhp);
+		free(node);
+		return (-1);
+	}
+
+	return (0);
+}
+
+/*
+ * Create a list of pools based on the given arguments.  If we're given no
+ * arguments, then iterate over all pools in the system and add them to the AVL
+ * tree.  Otherwise, add only those pool explicitly specified on the command
+ * line.
+ */
+zpool_list_t *
+pool_list_get(int argc, char **argv, zprop_list_t **proplist, int *err)
+{
+	zpool_list_t *zlp;
+
+	zlp = safe_malloc(sizeof (zpool_list_t));
+
+	zlp->zl_pool = uu_avl_pool_create("zfs_pool", sizeof (zpool_node_t),
+	    offsetof(zpool_node_t, zn_avlnode), zpool_compare, UU_DEFAULT);
+
+	if (zlp->zl_pool == NULL)
+		zpool_no_memory();
+
+	if ((zlp->zl_avl = uu_avl_create(zlp->zl_pool, NULL,
+	    UU_DEFAULT)) == NULL)
+		zpool_no_memory();
+
+	zlp->zl_proplist = proplist;
+
+	if (argc == 0) {
+		(void) zpool_iter(g_zfs, add_pool, zlp);
+		zlp->zl_findall = B_TRUE;
+	} else {
+		int i;
+
+		for (i = 0; i < argc; i++) {
+			zpool_handle_t *zhp;
+
+			if (zhp = zpool_open_canfail(g_zfs, argv[i])) {
+				if (add_pool(zhp, zlp) != 0)
+					*err = B_TRUE;
+			} else {
+				*err = B_TRUE;
+			}
+		}
+	}
+
+	return (zlp);
+}
+
+/*
+ * Search for any new pools, adding them to the list.  We only add pools when no
+ * options were given on the command line.  Otherwise, we keep the list fixed as
+ * those that were explicitly specified.
+ */
+void
+pool_list_update(zpool_list_t *zlp)
+{
+	if (zlp->zl_findall)
+		(void) zpool_iter(g_zfs, add_pool, zlp);
+}
+
+/*
+ * Iterate over all pools in the list, executing the callback for each
+ */
+int
+pool_list_iter(zpool_list_t *zlp, int unavail, zpool_iter_f func,
+    void *data)
+{
+	zpool_node_t *node, *next_node;
+	int ret = 0;
+
+	for (node = uu_avl_first(zlp->zl_avl); node != NULL; node = next_node) {
+		next_node = uu_avl_next(zlp->zl_avl, node);
+		if (zpool_get_state(node->zn_handle) != POOL_STATE_UNAVAIL ||
+		    unavail)
+			ret |= func(node->zn_handle, data);
+	}
+
+	return (ret);
+}
+
+/*
+ * Remove the given pool from the list.  When running iostat, we want to remove
+ * those pools that no longer exist.
+ */
+void
+pool_list_remove(zpool_list_t *zlp, zpool_handle_t *zhp)
+{
+	zpool_node_t search, *node;
+
+	search.zn_handle = zhp;
+	if ((node = uu_avl_find(zlp->zl_avl, &search, NULL, NULL)) != NULL) {
+		uu_avl_remove(zlp->zl_avl, node);
+		zpool_close(node->zn_handle);
+		free(node);
+	}
+}
+
+/*
+ * Free all the handles associated with this list.
+ */
+void
+pool_list_free(zpool_list_t *zlp)
+{
+	uu_avl_walk_t *walk;
+	zpool_node_t *node;
+
+	if ((walk = uu_avl_walk_start(zlp->zl_avl, UU_WALK_ROBUST)) == NULL) {
+		(void) fprintf(stderr,
+		    gettext("internal error: out of memory"));
+		exit(1);
+	}
+
+	while ((node = uu_avl_walk_next(walk)) != NULL) {
+		uu_avl_remove(zlp->zl_avl, node);
+		zpool_close(node->zn_handle);
+		free(node);
+	}
+
+	uu_avl_walk_end(walk);
+	uu_avl_destroy(zlp->zl_avl);
+	uu_avl_pool_destroy(zlp->zl_pool);
+
+	free(zlp);
+}
+
+/*
+ * Returns the number of elements in the pool list.
+ */
+int
+pool_list_count(zpool_list_t *zlp)
+{
+	return (uu_avl_numnodes(zlp->zl_avl));
+}
+
+/*
+ * High level function which iterates over all pools given on the command line,
+ * using the pool_list_* interfaces.
+ */
+int
+for_each_pool(int argc, char **argv, boolean_t unavail,
+    zprop_list_t **proplist, zpool_iter_f func, void *data)
+{
+	zpool_list_t *list;
+	int ret = 0;
+
+	if ((list = pool_list_get(argc, argv, proplist, &ret)) == NULL)
+		return (1);
+
+	if (pool_list_iter(list, unavail, func, data) != 0)
+		ret = 1;
+
+	pool_list_free(list);
+
+	return (ret);
+}

cmd/zpool/zpool_util.c

@@ -0,0 +1,104 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <errno.h>
+#include <libgen.h>
+#include <libintl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <strings.h>
+
+#include "zpool_util.h"
+
+/*
+ * Utility function to guarantee malloc() success.
+ */
+void *
+safe_malloc(size_t size)
+{
+	void *data;
+
+	if ((data = calloc(1, size)) == NULL) {
+		(void) fprintf(stderr, "internal error: out of memory\n");
+		exit(1);
+	}
+
+	return (data);
+}
+
+/*
+ * Same as above, but for strdup()
+ */
+char *
+safe_strdup(const char *str)
+{
+	char *ret;
+
+	if ((ret = strdup(str)) == NULL) {
+		(void) fprintf(stderr, "internal error: out of memory\n");
+		exit(1);
+	}
+
+	return (ret);
+}
+
+/*
+ * Display an out of memory error message and abort the current program.
+ */
+void
+zpool_no_memory(void)
+{
+	assert(errno == ENOMEM);
+	(void) fprintf(stderr,
+	    gettext("internal error: out of memory\n"));
+	exit(1);
+}
+
+/*
+ * Return the number of logs in supplied nvlist
+ */
+uint_t
+num_logs(nvlist_t *nv)
+{
+	uint_t nlogs = 0;
+	uint_t c, children;
+	nvlist_t **child;
+
+	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) != 0)
+		return (0);
+
+	for (c = 0; c < children; c++) {
+		uint64_t is_log = B_FALSE;
+
+		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
+		    &is_log);
+		if (is_log)
+			nlogs++;
+	}
+	return (nlogs);
+}

cmd/zpool/zpool_util.h

@@ -0,0 +1,72 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef	ZPOOL_UTIL_H
+#define	ZPOOL_UTIL_H
+
+#include <libnvpair.h>
+#include <libzfs.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+/*
+ * Basic utility functions
+ */
+void *safe_malloc(size_t);
+char *safe_strdup(const char *);
+void zpool_no_memory(void);
+uint_t num_logs(nvlist_t *nv);
+
+/*
+ * Virtual device functions
+ */
+
+nvlist_t *make_root_vdev(zpool_handle_t *zhp, int force, int check_rep,
+    boolean_t isreplace, boolean_t dryrun, int argc, char **argv);
+
+/*
+ * Pool list functions
+ */
+int for_each_pool(int, char **, boolean_t unavail, zprop_list_t **,
+    zpool_iter_f, void *);
+
+typedef struct zpool_list zpool_list_t;
+
+zpool_list_t *pool_list_get(int, char **, zprop_list_t **, int *);
+void pool_list_update(zpool_list_t *);
+int pool_list_iter(zpool_list_t *, int unavail, zpool_iter_f, void *);
+void pool_list_free(zpool_list_t *);
+int pool_list_count(zpool_list_t *);
+void pool_list_remove(zpool_list_t *, zpool_handle_t *);
+
+libzfs_handle_t *g_zfs;
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* ZPOOL_UTIL_H */