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

lib/libspl/libspl/include/sys/list.h

@@ -0,0 +1,67 @@
+/*
+ * 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	_SYS_LIST_H
+#define	_SYS_LIST_H
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <sys/list_impl.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+typedef struct list_node list_node_t;
+typedef struct list list_t;
+
+void list_create(list_t *, size_t, size_t);
+void list_destroy(list_t *);
+
+void list_insert_after(list_t *, void *, void *);
+void list_insert_before(list_t *, void *, void *);
+void list_insert_head(list_t *, void *);
+void list_insert_tail(list_t *, void *);
+void list_remove(list_t *, void *);
+void *list_remove_head(list_t *);
+void *list_remove_tail(list_t *);
+void list_move_tail(list_t *, list_t *);
+
+void *list_head(list_t *);
+void *list_tail(list_t *);
+void *list_next(list_t *, void *);
+void *list_prev(list_t *, void *);
+int list_is_empty(list_t *);
+
+void list_link_init(list_node_t *);
+void list_link_replace(list_node_t *, list_node_t *);
+
+int list_link_active(list_node_t *);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _SYS_LIST_H */

lib/libspl/libspl/include/sys/list_impl.h

@@ -0,0 +1,53 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2003 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef	_SYS_LIST_IMPL_H
+#define	_SYS_LIST_IMPL_H
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <sys/types.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+struct list_node {
+	struct list_node *list_next;
+	struct list_node *list_prev;
+};
+
+struct list {
+	size_t	list_size;
+	size_t	list_offset;
+	struct list_node list_head;
+};
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _SYS_LIST_IMPL_H */

lib/libspl/libspl/list.c

@@ -0,0 +1,245 @@
+/*
+ * 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"
+
+/*
+ * Generic doubly-linked list implementation
+ */
+
+#include <sys/list.h>
+#include <sys/list_impl.h>
+#include <sys/types.h>
+#include <sys/sysmacros.h>
+#include <sys/debug.h>
+
+#define	list_d2l(a, obj) ((list_node_t *)(((char *)obj) + (a)->list_offset))
+#define	list_object(a, node) ((void *)(((char *)node) - (a)->list_offset))
+#define	list_empty(a) ((a)->list_head.list_next == &(a)->list_head)
+
+#define	list_insert_after_node(list, node, object) {	\
+	list_node_t *lnew = list_d2l(list, object);	\
+	lnew->list_prev = (node);			\
+	lnew->list_next = (node)->list_next;		\
+	(node)->list_next->list_prev = lnew;		\
+	(node)->list_next = lnew;			\
+}
+
+#define	list_insert_before_node(list, node, object) {	\
+	list_node_t *lnew = list_d2l(list, object);	\
+	lnew->list_next = (node);			\
+	lnew->list_prev = (node)->list_prev;		\
+	(node)->list_prev->list_next = lnew;		\
+	(node)->list_prev = lnew;			\
+}
+
+#define	list_remove_node(node)					\
+	(node)->list_prev->list_next = (node)->list_next;	\
+	(node)->list_next->list_prev = (node)->list_prev;	\
+	(node)->list_next = (node)->list_prev = NULL
+
+void
+list_create(list_t *list, size_t size, size_t offset)
+{
+	ASSERT(list);
+	ASSERT(size > 0);
+	ASSERT(size >= offset + sizeof (list_node_t));
+
+	list->list_size = size;
+	list->list_offset = offset;
+	list->list_head.list_next = list->list_head.list_prev =
+	    &list->list_head;
+}
+
+void
+list_destroy(list_t *list)
+{
+	list_node_t *node = &list->list_head;
+
+	ASSERT(list);
+	ASSERT(list->list_head.list_next == node);
+	ASSERT(list->list_head.list_prev == node);
+
+	node->list_next = node->list_prev = NULL;
+}
+
+void
+list_insert_after(list_t *list, void *object, void *nobject)
+{
+	if (object == NULL) {
+		list_insert_head(list, nobject);
+	} else {
+		list_node_t *lold = list_d2l(list, object);
+		list_insert_after_node(list, lold, nobject);
+	}
+}
+
+void
+list_insert_before(list_t *list, void *object, void *nobject)
+{
+	if (object == NULL) {
+		list_insert_tail(list, nobject);
+	} else {
+		list_node_t *lold = list_d2l(list, object);
+		list_insert_before_node(list, lold, nobject);
+	}
+}
+
+void
+list_insert_head(list_t *list, void *object)
+{
+	list_node_t *lold = &list->list_head;
+	list_insert_after_node(list, lold, object);
+}
+
+void
+list_insert_tail(list_t *list, void *object)
+{
+	list_node_t *lold = &list->list_head;
+	list_insert_before_node(list, lold, object);
+}
+
+void
+list_remove(list_t *list, void *object)
+{
+	list_node_t *lold = list_d2l(list, object);
+	ASSERT(!list_empty(list));
+	ASSERT(lold->list_next != NULL);
+	list_remove_node(lold);
+}
+
+void *
+list_remove_head(list_t *list)
+{
+	list_node_t *head = list->list_head.list_next;
+	if (head == &list->list_head)
+		return (NULL);
+	list_remove_node(head);
+	return (list_object(list, head));
+}
+
+void *
+list_remove_tail(list_t *list)
+{
+	list_node_t *tail = list->list_head.list_prev;
+	if (tail == &list->list_head)
+		return (NULL);
+	list_remove_node(tail);
+	return (list_object(list, tail));
+}
+
+void *
+list_head(list_t *list)
+{
+	if (list_empty(list))
+		return (NULL);
+	return (list_object(list, list->list_head.list_next));
+}
+
+void *
+list_tail(list_t *list)
+{
+	if (list_empty(list))
+		return (NULL);
+	return (list_object(list, list->list_head.list_prev));
+}
+
+void *
+list_next(list_t *list, void *object)
+{
+	list_node_t *node = list_d2l(list, object);
+
+	if (node->list_next != &list->list_head)
+		return (list_object(list, node->list_next));
+
+	return (NULL);
+}
+
+void *
+list_prev(list_t *list, void *object)
+{
+	list_node_t *node = list_d2l(list, object);
+
+	if (node->list_prev != &list->list_head)
+		return (list_object(list, node->list_prev));
+
+	return (NULL);
+}
+
+/*
+ *  Insert src list after dst list. Empty src list thereafter.
+ */
+void
+list_move_tail(list_t *dst, list_t *src)
+{
+	list_node_t *dstnode = &dst->list_head;
+	list_node_t *srcnode = &src->list_head;
+
+	ASSERT(dst->list_size == src->list_size);
+	ASSERT(dst->list_offset == src->list_offset);
+
+	if (list_empty(src))
+		return;
+
+	dstnode->list_prev->list_next = srcnode->list_next;
+	srcnode->list_next->list_prev = dstnode->list_prev;
+	dstnode->list_prev = srcnode->list_prev;
+	srcnode->list_prev->list_next = dstnode;
+
+	/* empty src list */
+	srcnode->list_next = srcnode->list_prev = srcnode;
+}
+
+void
+list_link_replace(list_node_t *lold, list_node_t *lnew)
+{
+	ASSERT(list_link_active(lold));
+	ASSERT(!list_link_active(lnew));
+
+	lnew->list_next = lold->list_next;
+	lnew->list_prev = lold->list_prev;
+	lold->list_prev->list_next = lnew;
+	lold->list_next->list_prev = lnew;
+	lold->list_next = lold->list_prev = NULL;
+}
+
+void
+list_link_init(list_node_t *link)
+{
+	link->list_next = NULL;
+	link->list_prev = NULL;
+}
+
+int
+list_link_active(list_node_t *link)
+{
+	return (link->list_next != NULL);
+}
+
+int
+list_is_empty(list_t *list)
+{
+	return (list_empty(list));
+}

lib/libspl/libspl/mkdirp.c

@@ -0,0 +1,212 @@
+/*
+ * 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.
+ */
+
+/*	Copyright (c) 1988 AT&T	*/
+/*	  All Rights Reserved  	*/
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+/*
+ * Creates directory and it's parents if the parents do not
+ * exist yet.
+ *
+ * Returns -1 if fails for reasons other than non-existing
+ * parents.
+ * Does NOT simplify pathnames with . or .. in them.
+ */
+
+#include <sys/types.h>
+#include <libgen.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <string.h>
+#include <sys/stat.h>
+
+static char *simplify(const char *str);
+
+int
+mkdirp(const char *d, mode_t mode)
+{
+	char  *endptr, *ptr, *slash, *str;
+
+	str = simplify(d);
+
+	/* If space couldn't be allocated for the simplified names, return. */
+
+	if (str == NULL)
+		return (-1);
+
+		/* Try to make the directory */
+
+	if (mkdir(str, mode) == 0) {
+		free(str);
+		return (0);
+	}
+	if (errno != ENOENT) {
+		free(str);
+		return (-1);
+	}
+	endptr = strrchr(str, '\0');
+	slash = strrchr(str, '/');
+
+		/* Search upward for the non-existing parent */
+
+	while (slash != NULL) {
+
+		ptr = slash;
+		*ptr = '\0';
+
+			/* If reached an existing parent, break */
+
+		if (access(str, F_OK) == 0)
+			break;
+
+			/* If non-existing parent */
+
+		else {
+			slash = strrchr(str, '/');
+
+				/* If under / or current directory, make it. */
+
+			if (slash == NULL || slash == str) {
+				if (mkdir(str, mode) != 0 && errno != EEXIST) {
+					free(str);
+					return (-1);
+				}
+				break;
+			}
+		}
+	}
+
+	/* Create directories starting from upmost non-existing parent */
+
+	while ((ptr = strchr(str, '\0')) != endptr) {
+		*ptr = '/';
+		if (mkdir(str, mode) != 0 && errno != EEXIST) {
+			/*
+			 *  If the mkdir fails because str already
+			 *  exists (EEXIST), then str has the form
+			 *  "existing-dir/..", and this is really
+			 *  ok. (Remember, this loop is creating the
+			 *  portion of the path that didn't exist)
+			 */
+			free(str);
+			return (-1);
+		}
+	}
+	free(str);
+	return (0);
+}
+
+/*
+ *	simplify - given a pathname, simplify that path by removing
+ *		   duplicate contiguous slashes.
+ *
+ *		   A simplified copy of the argument is returned to the
+ *		   caller, or NULL is returned on error.
+ *
+ *		   The caller should handle error reporting based upon the
+ *		   returned vlaue, and should free the returned value,
+ *		   when appropriate.
+ */
+
+static char *
+simplify(const char *str)
+{
+	int i;
+	size_t mbPathlen;	/* length of multi-byte path */
+	size_t wcPathlen;	/* length of wide-character path */
+	wchar_t *wptr;		/* scratch pointer */
+	wchar_t *wcPath;	/* wide-character version of the path */
+	char *mbPath;		/* The copy fo the path to be returned */
+
+	/*
+	 *  bail out if there is nothing there.
+	 */
+
+	if (!str)
+		return (NULL);
+
+	/*
+	 *  Get a copy of the argument.
+	 */
+
+	if ((mbPath = strdup(str)) == NULL) {
+		return (NULL);
+	}
+
+	/*
+	 *  convert the multi-byte version of the path to a
+	 *  wide-character rendering, for doing our figuring.
+	 */
+
+	mbPathlen = strlen(mbPath);
+
+	if ((wcPath = calloc(sizeof (wchar_t), mbPathlen+1)) == NULL) {
+		free(mbPath);
+		return (NULL);
+	}
+
+	if ((wcPathlen = mbstowcs(wcPath, mbPath, mbPathlen)) == (size_t)-1) {
+		free(mbPath);
+		free(wcPath);
+		return (NULL);
+	}
+
+	/*
+	 *  remove duplicate slashes first ("//../" -> "/")
+	 */
+
+	for (wptr = wcPath, i = 0; i < wcPathlen; i++) {
+		*wptr++ = wcPath[i];
+
+		if (wcPath[i] == '/') {
+			i++;
+
+			while (wcPath[i] == '/') {
+				i++;
+			}
+
+			i--;
+		}
+	}
+
+	*wptr = '\0';
+
+	/*
+	 *  now convert back to the multi-byte format.
+	 */
+
+	if (wcstombs(mbPath, wcPath, mbPathlen) == (size_t)-1) {
+		free(mbPath);
+		free(wcPath);
+		return (NULL);
+	}
+
+	free(wcPath);
+	return (mbPath);
+}

lib/libspl/libspl/strlcat.c

@@ -0,0 +1,59 @@
+/*
+ * 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"
+
+#include "lint.h"
+#include <string.h>
+#include <sys/types.h>
+
+/*
+ * Appends src to the dstsize buffer at dst. The append will never
+ * overflow the destination buffer and the buffer will always be null
+ * terminated. Never reference beyond &dst[dstsize-1] when computing
+ * the length of the pre-existing string.
+ */
+
+size_t
+strlcat(char *dst, const char *src, size_t dstsize)
+{
+	char *df = dst;
+	size_t left = dstsize;
+	size_t l1;
+	size_t l2 = strlen(src);
+	size_t copied;
+
+	while (left-- != 0 && *df != '\0')
+		df++;
+	l1 = df - dst;
+	if (dstsize == l1)
+		return (l1 + l2);
+
+	copied = l1 + l2 >= dstsize ? dstsize - l1 - 1 : l2;
+	(void) memcpy(dst + l1, src, copied);
+	dst[l1+copied] = '\0';
+	return (l1 + l2);
+}

lib/libspl/libspl/strlcpy.c

@@ -0,0 +1,55 @@
+/*
+ * 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"
+
+#include "lint.h"
+#include <string.h>
+#include <sys/types.h>
+
+/*
+ * Copies src to the dstsize buffer at dst. The copy will never
+ * overflow the destination buffer and the buffer will always be null
+ * terminated.
+ */
+
+size_t
+strlcpy(char *dst, const char *src, size_t len)
+{
+	size_t slen = strlen(src);
+	size_t copied;
+
+	if (len == 0)
+		return (slen);
+
+	if (slen >= len)
+		copied = len - 1;
+	else
+		copied = slen;
+	(void) memcpy(dst, src, copied);
+	dst[copied] = '\0';
+	return (slen);
+}

lib/libspl/libspl/strnlen.c

@@ -0,0 +1,47 @@
+/*
+ * 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"
+
+#include "lint.h"
+#include <string.h>
+#include <sys/types.h>
+
+/*
+ * Returns the number of non-NULL bytes in string argument,
+ * but not more than maxlen.  Does not look past str + maxlen.
+ */
+size_t
+strnlen(const char *str, size_t maxlen)
+{
+	const char *ptr;
+
+	ptr = memchr(str, 0, maxlen);
+	if (ptr == NULL)
+		return (maxlen);
+
+	return (ptr - str);
+}

lib/libuutil/include/libuutil.h

@@ -0,0 +1,381 @@
+/*
+ * 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	_LIBUUTIL_H
+#define	_LIBUUTIL_H
+
+#include <sys/types.h>
+#include <stdarg.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+/*
+ * Standard flags codes.
+ */
+#define	UU_DEFAULT		0
+
+/*
+ * Standard error codes.
+ */
+#define	UU_ERROR_NONE		0	/* no error */
+#define	UU_ERROR_INVALID_ARGUMENT 1	/* invalid argument */
+#define	UU_ERROR_UNKNOWN_FLAG	2	/* passed flag invalid */
+#define	UU_ERROR_NO_MEMORY	3	/* out of memory */
+#define	UU_ERROR_CALLBACK_FAILED 4	/* callback-initiated error */
+#define	UU_ERROR_NOT_SUPPORTED	5	/* operation not supported */
+#define	UU_ERROR_EMPTY		6	/* no value provided */
+#define	UU_ERROR_UNDERFLOW	7	/* value is too small */
+#define	UU_ERROR_OVERFLOW	8	/* value is too value */
+#define	UU_ERROR_INVALID_CHAR	9	/* value contains unexpected char */
+#define	UU_ERROR_INVALID_DIGIT	10	/* value contains digit not in base */
+
+#define	UU_ERROR_SYSTEM		99	/* underlying system error */
+#define	UU_ERROR_UNKNOWN	100	/* error status not known */
+
+/*
+ * Standard program exit codes.
+ */
+#define	UU_EXIT_OK	(*(uu_exit_ok()))
+#define	UU_EXIT_FATAL	(*(uu_exit_fatal()))
+#define	UU_EXIT_USAGE	(*(uu_exit_usage()))
+
+/*
+ * Exit status profiles.
+ */
+#define	UU_PROFILE_DEFAULT	0
+#define	UU_PROFILE_LAUNCHER	1
+
+/*
+ * Error reporting functions.
+ */
+uint32_t uu_error(void);
+const char *uu_strerror(uint32_t);
+
+/*
+ * Program notification functions.
+ */
+extern void uu_alt_exit(int);
+extern const char *uu_setpname(char *);
+extern const char *uu_getpname(void);
+/*PRINTFLIKE1*/
+extern void uu_warn(const char *, ...);
+extern void uu_vwarn(const char *, va_list);
+/*PRINTFLIKE1*/
+extern void uu_die(const char *, ...) __NORETURN;
+extern void uu_vdie(const char *, va_list) __NORETURN;
+/*PRINTFLIKE2*/
+extern void uu_xdie(int, const char *, ...) __NORETURN;
+extern void uu_vxdie(int, const char *, va_list) __NORETURN;
+
+/*
+ * Exit status functions (not to be used directly)
+ */
+extern int *uu_exit_ok(void);
+extern int *uu_exit_fatal(void);
+extern int *uu_exit_usage(void);
+
+/*
+ * string->number conversions
+ */
+extern int uu_strtoint(const char *, void *, size_t, int, int64_t, int64_t);
+extern int uu_strtouint(const char *, void *, size_t, int, uint64_t, uint64_t);
+
+/*
+ * Debug print facility functions.
+ */
+typedef struct uu_dprintf uu_dprintf_t;
+
+typedef enum {
+	UU_DPRINTF_SILENT,
+	UU_DPRINTF_FATAL,
+	UU_DPRINTF_WARNING,
+	UU_DPRINTF_NOTICE,
+	UU_DPRINTF_INFO,
+	UU_DPRINTF_DEBUG
+} uu_dprintf_severity_t;
+
+extern uu_dprintf_t *uu_dprintf_create(const char *, uu_dprintf_severity_t,
+    uint_t);
+/*PRINTFLIKE3*/
+extern void uu_dprintf(uu_dprintf_t *, uu_dprintf_severity_t,
+    const char *, ...);
+extern void uu_dprintf_destroy(uu_dprintf_t *);
+extern const char *uu_dprintf_getname(uu_dprintf_t *);
+
+/*
+ * Identifier test flags and function.
+ */
+#define	UU_NAME_DOMAIN		0x1	/* allow SUNW, or com.sun, prefix */
+#define	UU_NAME_PATH		0x2	/* allow '/'-delimited paths */
+
+int uu_check_name(const char *, uint_t);
+
+/*
+ * File creation functions.
+ */
+extern int uu_open_tmp(const char *dir, uint_t uflags);
+
+/*
+ * Convenience functions.
+ */
+/*PRINTFLIKE1*/
+extern char *uu_msprintf(const char *format, ...);
+extern void *uu_zalloc(size_t);
+extern char *uu_strdup(const char *);
+extern void uu_free(void *);
+
+/*
+ * Comparison function type definition.
+ *   Developers should be careful in their use of the _private argument. If you
+ *   break interface guarantees, you get undefined behavior.
+ */
+typedef int uu_compare_fn_t(const void *__left, const void *__right,
+    void *__private);
+
+/*
+ * Walk variant flags.
+ *   A data structure need not provide support for all variants and
+ *   combinations.  Refer to the appropriate documentation.
+ */
+#define	UU_WALK_ROBUST		0x00000001	/* walk can survive removes */
+#define	UU_WALK_REVERSE		0x00000002	/* reverse walk order */
+
+#define	UU_WALK_PREORDER	0x00000010	/* walk tree in pre-order */
+#define	UU_WALK_POSTORDER	0x00000020	/* walk tree in post-order */
+
+/*
+ * Walk callback function return codes.
+ */
+#define	UU_WALK_ERROR		-1
+#define	UU_WALK_NEXT		0
+#define	UU_WALK_DONE		1
+
+/*
+ * Walk callback function type definition.
+ */
+typedef int uu_walk_fn_t(void *_elem, void *_private);
+
+/*
+ * lists: opaque structures
+ */
+typedef struct uu_list_pool uu_list_pool_t;
+typedef struct uu_list uu_list_t;
+
+typedef struct uu_list_node {
+	uintptr_t uln_opaque[2];
+} uu_list_node_t;
+
+typedef struct uu_list_walk uu_list_walk_t;
+
+typedef uintptr_t uu_list_index_t;
+
+/*
+ * lists: interface
+ *
+ * basic usage:
+ *	typedef struct foo {
+ *		...
+ *		uu_list_node_t foo_node;
+ *		...
+ *	} foo_t;
+ *
+ *	static int
+ *	foo_compare(void *l_arg, void *r_arg, void *private)
+ *	{
+ *		foo_t *l = l_arg;
+ *		foo_t *r = r_arg;
+ *
+ *		if (... l greater than r ...)
+ *			return (1);
+ *		if (... l less than r ...)
+ *			return (-1);
+ *		return (0);
+ *	}
+ *
+ *	...
+ *		// at initialization time
+ *		foo_pool = uu_list_pool_create("foo_pool",
+ *		    sizeof (foo_t), offsetof(foo_t, foo_node), foo_compare,
+ *		    debugging? 0 : UU_AVL_POOL_DEBUG);
+ *	...
+ */
+uu_list_pool_t *uu_list_pool_create(const char *, size_t, size_t,
+    uu_compare_fn_t *, uint32_t);
+#define	UU_LIST_POOL_DEBUG	0x00000001
+
+void uu_list_pool_destroy(uu_list_pool_t *);
+
+/*
+ * usage:
+ *
+ *	foo_t *a;
+ *	a = malloc(sizeof(*a));
+ *	uu_list_node_init(a, &a->foo_list, pool);
+ *	...
+ *	uu_list_node_fini(a, &a->foo_list, pool);
+ *	free(a);
+ */
+void uu_list_node_init(void *, uu_list_node_t *, uu_list_pool_t *);
+void uu_list_node_fini(void *, uu_list_node_t *, uu_list_pool_t *);
+
+uu_list_t *uu_list_create(uu_list_pool_t *, void *_parent, uint32_t);
+#define	UU_LIST_DEBUG	0x00000001
+#define	UU_LIST_SORTED	0x00000002	/* list is sorted */
+
+void uu_list_destroy(uu_list_t *);	/* list must be empty */
+
+size_t uu_list_numnodes(uu_list_t *);
+
+void *uu_list_first(uu_list_t *);
+void *uu_list_last(uu_list_t *);
+
+void *uu_list_next(uu_list_t *, void *);
+void *uu_list_prev(uu_list_t *, void *);
+
+int uu_list_walk(uu_list_t *, uu_walk_fn_t *, void *, uint32_t);
+
+uu_list_walk_t *uu_list_walk_start(uu_list_t *, uint32_t);
+void *uu_list_walk_next(uu_list_walk_t *);
+void uu_list_walk_end(uu_list_walk_t *);
+
+void *uu_list_find(uu_list_t *, void *, void *, uu_list_index_t *);
+void uu_list_insert(uu_list_t *, void *, uu_list_index_t);
+
+void *uu_list_nearest_next(uu_list_t *, uu_list_index_t);
+void *uu_list_nearest_prev(uu_list_t *, uu_list_index_t);
+
+void *uu_list_teardown(uu_list_t *, void **);
+
+void uu_list_remove(uu_list_t *, void *);
+
+/*
+ * lists: interfaces for non-sorted lists only
+ */
+int uu_list_insert_before(uu_list_t *, void *_target, void *_elem);
+int uu_list_insert_after(uu_list_t *, void *_target, void *_elem);
+
+/*
+ * avl trees: opaque structures
+ */
+typedef struct uu_avl_pool uu_avl_pool_t;
+typedef struct uu_avl uu_avl_t;
+
+typedef struct uu_avl_node {
+#ifdef _LP64
+	uintptr_t uan_opaque[3];
+#else
+	uintptr_t uan_opaque[4];
+#endif
+} uu_avl_node_t;
+
+typedef struct uu_avl_walk uu_avl_walk_t;
+
+typedef uintptr_t uu_avl_index_t;
+
+/*
+ * avl trees: interface
+ *
+ * basic usage:
+ *	typedef struct foo {
+ *		...
+ *		uu_avl_node_t foo_node;
+ *		...
+ *	} foo_t;
+ *
+ *	static int
+ *	foo_compare(void *l_arg, void *r_arg, void *private)
+ *	{
+ *		foo_t *l = l_arg;
+ *		foo_t *r = r_arg;
+ *
+ *		if (... l greater than r ...)
+ *			return (1);
+ *		if (... l less than r ...)
+ *			return (-1);
+ *		return (0);
+ *	}
+ *
+ *	...
+ *		// at initialization time
+ *		foo_pool = uu_avl_pool_create("foo_pool",
+ *		    sizeof (foo_t), offsetof(foo_t, foo_node), foo_compare,
+ *		    debugging? 0 : UU_AVL_POOL_DEBUG);
+ *	...
+ */
+uu_avl_pool_t *uu_avl_pool_create(const char *, size_t, size_t,
+    uu_compare_fn_t *, uint32_t);
+#define	UU_AVL_POOL_DEBUG	0x00000001
+
+void uu_avl_pool_destroy(uu_avl_pool_t *);
+
+/*
+ * usage:
+ *
+ *	foo_t *a;
+ *	a = malloc(sizeof(*a));
+ *	uu_avl_node_init(a, &a->foo_avl, pool);
+ *	...
+ *	uu_avl_node_fini(a, &a->foo_avl, pool);
+ *	free(a);
+ */
+void uu_avl_node_init(void *, uu_avl_node_t *, uu_avl_pool_t *);
+void uu_avl_node_fini(void *, uu_avl_node_t *, uu_avl_pool_t *);
+
+uu_avl_t *uu_avl_create(uu_avl_pool_t *, void *_parent, uint32_t);
+#define	UU_AVL_DEBUG	0x00000001
+
+void uu_avl_destroy(uu_avl_t *);	/* list must be empty */
+
+size_t uu_avl_numnodes(uu_avl_t *);
+
+void *uu_avl_first(uu_avl_t *);
+void *uu_avl_last(uu_avl_t *);
+
+void *uu_avl_next(uu_avl_t *, void *);
+void *uu_avl_prev(uu_avl_t *, void *);
+
+int uu_avl_walk(uu_avl_t *, uu_walk_fn_t *, void *, uint32_t);
+
+uu_avl_walk_t *uu_avl_walk_start(uu_avl_t *, uint32_t);
+void *uu_avl_walk_next(uu_avl_walk_t *);
+void uu_avl_walk_end(uu_avl_walk_t *);
+
+void *uu_avl_find(uu_avl_t *, void *, void *, uu_avl_index_t *);
+void uu_avl_insert(uu_avl_t *, void *, uu_avl_index_t);
+
+void *uu_avl_nearest_next(uu_avl_t *, uu_avl_index_t);
+void *uu_avl_nearest_prev(uu_avl_t *, uu_avl_index_t);
+
+void *uu_avl_teardown(uu_avl_t *, void **);
+
+void uu_avl_remove(uu_avl_t *, void *);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _LIBUUTIL_H */

lib/libuutil/include/libuutil_common.h

@@ -0,0 +1,35 @@
+/*
+ * 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	_LIBUUTIL_COMMON_H
+#define	_LIBUUTIL_COMMON_H
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <libuutil.h>
+#include <libuutil_impl.h>
+
+#endif	/* _LIBUUTIL_COMMON_H */

lib/libuutil/include/libuutil_impl.h

@@ -0,0 +1,181 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2005 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef	_LIBUUTIL_IMPL_H
+#define	_LIBUUTIL_IMPL_H
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include <libuutil.h>
+#include <pthread.h>
+
+#include <sys/avl_impl.h>
+#include <sys/byteorder.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+void uu_set_error(uint_t);
+#pragma rarely_called(uu_set_error)
+
+/*PRINTFLIKE1*/
+void uu_panic(const char *format, ...);
+#pragma rarely_called(uu_panic)
+
+struct uu_dprintf {
+	char	*uud_name;
+	uu_dprintf_severity_t uud_severity;
+	uint_t	uud_flags;
+};
+
+/*
+ * For debugging purposes, libuutil keeps around linked lists of all uu_lists
+ * and uu_avls, along with pointers to their parents.  These can cause false
+ * negatives when looking for memory leaks, so we encode the pointers by
+ * storing them with swapped endianness;  this is not perfect, but it's about
+ * the best we can do without wasting a lot of space.
+ */
+#ifdef _LP64
+#define	UU_PTR_ENCODE(ptr)		BSWAP_64((uintptr_t)(void *)(ptr))
+#else
+#define	UU_PTR_ENCODE(ptr)		BSWAP_32((uintptr_t)(void *)(ptr))
+#endif
+
+#define	UU_PTR_DECODE(ptr)		((void *)UU_PTR_ENCODE(ptr))
+
+/*
+ * uu_list structures
+ */
+typedef struct uu_list_node_impl {
+	struct uu_list_node_impl *uln_next;
+	struct uu_list_node_impl *uln_prev;
+} uu_list_node_impl_t;
+
+struct uu_list_walk {
+	uu_list_walk_t	*ulw_next;
+	uu_list_walk_t	*ulw_prev;
+
+	uu_list_t	*ulw_list;
+	int8_t		ulw_dir;
+	uint8_t		ulw_robust;
+	uu_list_node_impl_t *ulw_next_result;
+};
+
+struct uu_list {
+	uintptr_t	ul_next_enc;
+	uintptr_t	ul_prev_enc;
+
+	uu_list_pool_t	*ul_pool;
+	uintptr_t	ul_parent_enc;	/* encoded parent pointer */
+	size_t		ul_offset;
+	size_t		ul_numnodes;
+	uint8_t		ul_debug;
+	uint8_t		ul_sorted;
+	uint8_t		ul_index;	/* mark for uu_list_index_ts */
+
+	uu_list_node_impl_t ul_null_node;
+	uu_list_walk_t	ul_null_walk;	/* for robust walkers */
+};
+
+#define	UU_LIST_PTR(ptr)		((uu_list_t *)UU_PTR_DECODE(ptr))
+
+#define	UU_LIST_POOL_MAXNAME	64
+
+struct uu_list_pool {
+	uu_list_pool_t	*ulp_next;
+	uu_list_pool_t	*ulp_prev;
+
+	char		ulp_name[UU_LIST_POOL_MAXNAME];
+	size_t		ulp_nodeoffset;
+	size_t		ulp_objsize;
+	uu_compare_fn_t	*ulp_cmp;
+	uint8_t		ulp_debug;
+	uint8_t		ulp_last_index;
+	pthread_mutex_t	ulp_lock;		/* protects null_list */
+	uu_list_t	ulp_null_list;
+};
+
+/*
+ * uu_avl structures
+ */
+typedef struct avl_node		uu_avl_node_impl_t;
+
+struct uu_avl_walk {
+	uu_avl_walk_t	*uaw_next;
+	uu_avl_walk_t	*uaw_prev;
+
+	uu_avl_t	*uaw_avl;
+	void		*uaw_next_result;
+	int8_t		uaw_dir;
+	uint8_t		uaw_robust;
+};
+
+struct uu_avl {
+	uintptr_t	ua_next_enc;
+	uintptr_t	ua_prev_enc;
+
+	uu_avl_pool_t	*ua_pool;
+	uintptr_t	ua_parent_enc;
+	uint8_t		ua_debug;
+	uint8_t		ua_index;	/* mark for uu_avl_index_ts */
+
+	struct avl_tree	ua_tree;
+	uu_avl_walk_t	ua_null_walk;
+};
+
+#define	UU_AVL_PTR(x)		((uu_avl_t *)UU_PTR_DECODE(x))
+
+#define	UU_AVL_POOL_MAXNAME	64
+
+struct uu_avl_pool {
+	uu_avl_pool_t	*uap_next;
+	uu_avl_pool_t	*uap_prev;
+
+	char		uap_name[UU_AVL_POOL_MAXNAME];
+	size_t		uap_nodeoffset;
+	size_t		uap_objsize;
+	uu_compare_fn_t	*uap_cmp;
+	uint8_t		uap_debug;
+	uint8_t		uap_last_index;
+	pthread_mutex_t	uap_lock;		/* protects null_avl */
+	uu_avl_t	uap_null_avl;
+};
+
+/*
+ * atfork() handlers
+ */
+void uu_avl_lockup(void);
+void uu_avl_release(void);
+
+void uu_list_lockup(void);
+void uu_list_release(void);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _LIBUUTIL_IMPL_H */

lib/libuutil/uu_alloc.c

@@ -0,0 +1,98 @@
+/*
+ * 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 "libuutil_common.h"
+
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+void *
+uu_zalloc(size_t n)
+{
+	void *p = malloc(n);
+
+	if (p == NULL) {
+		uu_set_error(UU_ERROR_SYSTEM);
+		return (NULL);
+	}
+
+	(void) memset(p, 0, n);
+
+	return (p);
+}
+
+void
+uu_free(void *p)
+{
+	free(p);
+}
+
+char *
+uu_strdup(const char *str)
+{
+	char *buf = NULL;
+
+	if (str != NULL) {
+		size_t sz;
+
+		sz = strlen(str) + 1;
+		buf = uu_zalloc(sz);
+		if (buf != NULL)
+			(void) memcpy(buf, str, sz);
+	}
+	return (buf);
+}
+
+char *
+uu_msprintf(const char *format, ...)
+{
+	va_list args;
+	char attic[1];
+	uint_t M, m;
+	char *b;
+
+	va_start(args, format);
+	M = vsnprintf(attic, 1, format, args);
+	va_end(args);
+
+	for (;;) {
+		m = M;
+		if ((b = uu_zalloc(m + 1)) == NULL)
+			return (NULL);
+
+		va_start(args, format);
+		M = vsnprintf(b, m + 1, format, args);
+		va_end(args);
+
+		if (M == m)
+			break;		/* sizes match */
+
+		uu_free(b);
+	}
+
+	return (b);
+}

lib/libuutil/uu_avl.c

@@ -0,0 +1,569 @@
+/*
+ * 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"
+
+#include "libuutil_common.h"
+
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/avl.h>
+
+static uu_avl_pool_t	uu_null_apool = { &uu_null_apool, &uu_null_apool };
+static pthread_mutex_t	uu_apool_list_lock = PTHREAD_MUTEX_INITIALIZER;
+
+/*
+ * The index mark change on every insert and delete, to catch stale
+ * references.
+ *
+ * We leave the low bit alone, since the avl code uses it.
+ */
+#define	INDEX_MAX		(sizeof (uintptr_t) - 2)
+#define	INDEX_NEXT(m)		(((m) == INDEX_MAX)? 2 : ((m) + 2) & INDEX_MAX)
+
+#define	INDEX_DECODE(i)		((i) & ~INDEX_MAX)
+#define	INDEX_ENCODE(p, n)	(((n) & ~INDEX_MAX) | (p)->ua_index)
+#define	INDEX_VALID(p, i)	(((i) & INDEX_MAX) == (p)->ua_index)
+#define	INDEX_CHECK(i)		(((i) & INDEX_MAX) != 0)
+
+/*
+ * When an element is inactive (not in a tree), we keep a marked pointer to
+ * its containing pool in its first word, and a NULL pointer in its second.
+ *
+ * On insert, we use these to verify that it comes from the correct pool.
+ */
+#define	NODE_ARRAY(p, n)	((uintptr_t *)((uintptr_t)(n) + \
+				    (pp)->uap_nodeoffset))
+
+#define	POOL_TO_MARKER(pp) (((uintptr_t)(pp) | 1))
+
+#define	DEAD_MARKER		0xc4
+
+uu_avl_pool_t *
+uu_avl_pool_create(const char *name, size_t objsize, size_t nodeoffset,
+    uu_compare_fn_t *compare_func, uint32_t flags)
+{
+	uu_avl_pool_t *pp, *next, *prev;
+
+	if (name == NULL ||
+	    uu_check_name(name, UU_NAME_DOMAIN) == -1 ||
+	    nodeoffset + sizeof (uu_avl_node_t) > objsize ||
+	    compare_func == NULL) {
+		uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+		return (NULL);
+	}
+
+	if (flags & ~UU_AVL_POOL_DEBUG) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (NULL);
+	}
+
+	pp = uu_zalloc(sizeof (uu_avl_pool_t));
+	if (pp == NULL) {
+		uu_set_error(UU_ERROR_NO_MEMORY);
+		return (NULL);
+	}
+
+	(void) strlcpy(pp->uap_name, name, sizeof (pp->uap_name));
+	pp->uap_nodeoffset = nodeoffset;
+	pp->uap_objsize = objsize;
+	pp->uap_cmp = compare_func;
+	if (flags & UU_AVL_POOL_DEBUG)
+		pp->uap_debug = 1;
+	pp->uap_last_index = 0;
+
+	(void) pthread_mutex_init(&pp->uap_lock, NULL);
+
+	pp->uap_null_avl.ua_next_enc = UU_PTR_ENCODE(&pp->uap_null_avl);
+	pp->uap_null_avl.ua_prev_enc = UU_PTR_ENCODE(&pp->uap_null_avl);
+
+	(void) pthread_mutex_lock(&uu_apool_list_lock);
+	pp->uap_next = next = &uu_null_apool;
+	pp->uap_prev = prev = next->uap_prev;
+	next->uap_prev = pp;
+	prev->uap_next = pp;
+	(void) pthread_mutex_unlock(&uu_apool_list_lock);
+
+	return (pp);
+}
+
+void
+uu_avl_pool_destroy(uu_avl_pool_t *pp)
+{
+	if (pp->uap_debug) {
+		if (pp->uap_null_avl.ua_next_enc !=
+		    UU_PTR_ENCODE(&pp->uap_null_avl) ||
+		    pp->uap_null_avl.ua_prev_enc !=
+		    UU_PTR_ENCODE(&pp->uap_null_avl)) {
+			uu_panic("uu_avl_pool_destroy: Pool \"%.*s\" (%p) has "
+			    "outstanding avls, or is corrupt.\n",
+			    (int)sizeof (pp->uap_name), pp->uap_name,
+			    (void *)pp);
+		}
+	}
+	(void) pthread_mutex_lock(&uu_apool_list_lock);
+	pp->uap_next->uap_prev = pp->uap_prev;
+	pp->uap_prev->uap_next = pp->uap_next;
+	(void) pthread_mutex_unlock(&uu_apool_list_lock);
+	pp->uap_prev = NULL;
+	pp->uap_next = NULL;
+	uu_free(pp);
+}
+
+void
+uu_avl_node_init(void *base, uu_avl_node_t *np, uu_avl_pool_t *pp)
+{
+	uintptr_t *na = (uintptr_t *)np;
+
+	if (pp->uap_debug) {
+		uintptr_t offset = (uintptr_t)np - (uintptr_t)base;
+		if (offset + sizeof (*np) > pp->uap_objsize) {
+			uu_panic("uu_avl_node_init(%p, %p, %p (\"%s\")): "
+			    "offset %ld doesn't fit in object (size %ld)\n",
+			    base, (void *)np, (void *)pp, pp->uap_name,
+			    (long)offset, (long)pp->uap_objsize);
+		}
+		if (offset != pp->uap_nodeoffset) {
+			uu_panic("uu_avl_node_init(%p, %p, %p (\"%s\")): "
+			    "offset %ld doesn't match pool's offset (%ld)\n",
+			    base, (void *)np, (void *)pp, pp->uap_name,
+			    (long)offset, (long)pp->uap_objsize);
+		}
+	}
+
+	na[0] = POOL_TO_MARKER(pp);
+	na[1] = 0;
+}
+
+void
+uu_avl_node_fini(void *base, uu_avl_node_t *np, uu_avl_pool_t *pp)
+{
+	uintptr_t *na = (uintptr_t *)np;
+
+	if (pp->uap_debug) {
+		if (na[0] == DEAD_MARKER && na[1] == DEAD_MARKER) {
+			uu_panic("uu_avl_node_fini(%p, %p, %p (\"%s\")): "
+			    "node already finied\n",
+			    base, (void *)np, (void *)pp, pp->uap_name);
+		}
+		if (na[0] != POOL_TO_MARKER(pp) || na[1] != 0) {
+			uu_panic("uu_avl_node_fini(%p, %p, %p (\"%s\")): "
+			    "node corrupt, in tree, or in different pool\n",
+			    base, (void *)np, (void *)pp, pp->uap_name);
+		}
+	}
+
+	na[0] = DEAD_MARKER;
+	na[1] = DEAD_MARKER;
+	na[2] = DEAD_MARKER;
+}
+
+struct uu_avl_node_compare_info {
+	uu_compare_fn_t	*ac_compare;
+	void		*ac_private;
+	void		*ac_right;
+	void		*ac_found;
+};
+
+static int
+uu_avl_node_compare(const void *l, const void *r)
+{
+	struct uu_avl_node_compare_info *info =
+	    (struct uu_avl_node_compare_info *)l;
+
+	int res = info->ac_compare(r, info->ac_right, info->ac_private);
+
+	if (res == 0) {
+		if (info->ac_found == NULL)
+			info->ac_found = (void *)r;
+		return (-1);
+	}
+	if (res < 0)
+		return (1);
+	return (-1);
+}
+
+uu_avl_t *
+uu_avl_create(uu_avl_pool_t *pp, void *parent, uint32_t flags)
+{
+	uu_avl_t *ap, *next, *prev;
+
+	if (flags & ~UU_AVL_DEBUG) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (NULL);
+	}
+
+	ap = uu_zalloc(sizeof (*ap));
+	if (ap == NULL) {
+		uu_set_error(UU_ERROR_NO_MEMORY);
+		return (NULL);
+	}
+
+	ap->ua_pool = pp;
+	ap->ua_parent_enc = UU_PTR_ENCODE(parent);
+	ap->ua_debug = pp->uap_debug || (flags & UU_AVL_DEBUG);
+	ap->ua_index = (pp->uap_last_index = INDEX_NEXT(pp->uap_last_index));
+
+	avl_create(&ap->ua_tree, &uu_avl_node_compare, pp->uap_objsize,
+	    pp->uap_nodeoffset);
+
+	ap->ua_null_walk.uaw_next = &ap->ua_null_walk;
+	ap->ua_null_walk.uaw_prev = &ap->ua_null_walk;
+
+	(void) pthread_mutex_lock(&pp->uap_lock);
+	next = &pp->uap_null_avl;
+	prev = UU_PTR_DECODE(next->ua_prev_enc);
+	ap->ua_next_enc = UU_PTR_ENCODE(next);
+	ap->ua_prev_enc = UU_PTR_ENCODE(prev);
+	next->ua_prev_enc = UU_PTR_ENCODE(ap);
+	prev->ua_next_enc = UU_PTR_ENCODE(ap);
+	(void) pthread_mutex_unlock(&pp->uap_lock);
+
+	return (ap);
+}
+
+void
+uu_avl_destroy(uu_avl_t *ap)
+{
+	uu_avl_pool_t *pp = ap->ua_pool;
+
+	if (ap->ua_debug) {
+		if (avl_numnodes(&ap->ua_tree) != 0) {
+			uu_panic("uu_avl_destroy(%p): tree not empty\n",
+			    (void *)ap);
+		}
+		if (ap->ua_null_walk.uaw_next != &ap->ua_null_walk ||
+		    ap->ua_null_walk.uaw_prev != &ap->ua_null_walk) {
+			uu_panic("uu_avl_destroy(%p):  outstanding walkers\n",
+			    (void *)ap);
+		}
+	}
+	(void) pthread_mutex_lock(&pp->uap_lock);
+	UU_AVL_PTR(ap->ua_next_enc)->ua_prev_enc = ap->ua_prev_enc;
+	UU_AVL_PTR(ap->ua_prev_enc)->ua_next_enc = ap->ua_next_enc;
+	(void) pthread_mutex_unlock(&pp->uap_lock);
+	ap->ua_prev_enc = UU_PTR_ENCODE(NULL);
+	ap->ua_next_enc = UU_PTR_ENCODE(NULL);
+
+	ap->ua_pool = NULL;
+	avl_destroy(&ap->ua_tree);
+
+	uu_free(ap);
+}
+
+size_t
+uu_avl_numnodes(uu_avl_t *ap)
+{
+	return (avl_numnodes(&ap->ua_tree));
+}
+
+void *
+uu_avl_first(uu_avl_t *ap)
+{
+	return (avl_first(&ap->ua_tree));
+}
+
+void *
+uu_avl_last(uu_avl_t *ap)
+{
+	return (avl_last(&ap->ua_tree));
+}
+
+void *
+uu_avl_next(uu_avl_t *ap, void *node)
+{
+	return (AVL_NEXT(&ap->ua_tree, node));
+}
+
+void *
+uu_avl_prev(uu_avl_t *ap, void *node)
+{
+	return (AVL_PREV(&ap->ua_tree, node));
+}
+
+static void
+_avl_walk_init(uu_avl_walk_t *wp, uu_avl_t *ap, uint32_t flags)
+{
+	uu_avl_walk_t *next, *prev;
+
+	int robust = (flags & UU_WALK_ROBUST);
+	int direction = (flags & UU_WALK_REVERSE)? -1 : 1;
+
+	(void) memset(wp, 0, sizeof (*wp));
+	wp->uaw_avl = ap;
+	wp->uaw_robust = robust;
+	wp->uaw_dir = direction;
+
+	if (direction > 0)
+		wp->uaw_next_result = avl_first(&ap->ua_tree);
+	else
+		wp->uaw_next_result = avl_last(&ap->ua_tree);
+
+	if (ap->ua_debug || robust) {
+		wp->uaw_next = next = &ap->ua_null_walk;
+		wp->uaw_prev = prev = next->uaw_prev;
+		next->uaw_prev = wp;
+		prev->uaw_next = wp;
+	}
+}
+
+static void *
+_avl_walk_advance(uu_avl_walk_t *wp, uu_avl_t *ap)
+{
+	void *np = wp->uaw_next_result;
+
+	avl_tree_t *t = &ap->ua_tree;
+
+	if (np == NULL)
+		return (NULL);
+
+	wp->uaw_next_result = (wp->uaw_dir > 0)? AVL_NEXT(t, np) :
+	    AVL_PREV(t, np);
+
+	return (np);
+}
+
+static void
+_avl_walk_fini(uu_avl_walk_t *wp)
+{
+	if (wp->uaw_next != NULL) {
+		wp->uaw_next->uaw_prev = wp->uaw_prev;
+		wp->uaw_prev->uaw_next = wp->uaw_next;
+		wp->uaw_next = NULL;
+		wp->uaw_prev = NULL;
+	}
+	wp->uaw_avl = NULL;
+	wp->uaw_next_result = NULL;
+}
+
+uu_avl_walk_t *
+uu_avl_walk_start(uu_avl_t *ap, uint32_t flags)
+{
+	uu_avl_walk_t *wp;
+
+	if (flags & ~(UU_WALK_ROBUST | UU_WALK_REVERSE)) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (NULL);
+	}
+
+	wp = uu_zalloc(sizeof (*wp));
+	if (wp == NULL) {
+		uu_set_error(UU_ERROR_NO_MEMORY);
+		return (NULL);
+	}
+
+	_avl_walk_init(wp, ap, flags);
+	return (wp);
+}
+
+void *
+uu_avl_walk_next(uu_avl_walk_t *wp)
+{
+	return (_avl_walk_advance(wp, wp->uaw_avl));
+}
+
+void
+uu_avl_walk_end(uu_avl_walk_t *wp)
+{
+	_avl_walk_fini(wp);
+	uu_free(wp);
+}
+
+int
+uu_avl_walk(uu_avl_t *ap, uu_walk_fn_t *func, void *private, uint32_t flags)
+{
+	void *e;
+	uu_avl_walk_t my_walk;
+
+	int status = UU_WALK_NEXT;
+
+	if (flags & ~(UU_WALK_ROBUST | UU_WALK_REVERSE)) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (-1);
+	}
+
+	_avl_walk_init(&my_walk, ap, flags);
+	while (status == UU_WALK_NEXT &&
+	    (e = _avl_walk_advance(&my_walk, ap)) != NULL)
+		status = (*func)(e, private);
+	_avl_walk_fini(&my_walk);
+
+	if (status >= 0)
+		return (0);
+	uu_set_error(UU_ERROR_CALLBACK_FAILED);
+	return (-1);
+}
+
+void
+uu_avl_remove(uu_avl_t *ap, void *elem)
+{
+	uu_avl_walk_t *wp;
+	uu_avl_pool_t *pp = ap->ua_pool;
+	uintptr_t *na = NODE_ARRAY(pp, elem);
+
+	if (ap->ua_debug) {
+		/*
+		 * invalidate outstanding uu_avl_index_ts.
+		 */
+		ap->ua_index = INDEX_NEXT(ap->ua_index);
+	}
+
+	/*
+	 * Robust walkers most be advanced, if we are removing the node
+	 * they are currently using.  In debug mode, non-robust walkers
+	 * are also on the walker list.
+	 */
+	for (wp = ap->ua_null_walk.uaw_next; wp != &ap->ua_null_walk;
+	    wp = wp->uaw_next) {
+		if (wp->uaw_robust) {
+			if (elem == wp->uaw_next_result)
+				(void) _avl_walk_advance(wp, ap);
+		} else if (wp->uaw_next_result != NULL) {
+			uu_panic("uu_avl_remove(%p, %p): active non-robust "
+			    "walker\n", (void *)ap, elem);
+		}
+	}
+
+	avl_remove(&ap->ua_tree, elem);
+
+	na[0] = POOL_TO_MARKER(pp);
+	na[1] = 0;
+}
+
+void *
+uu_avl_teardown(uu_avl_t *ap, void **cookie)
+{
+	void *elem = avl_destroy_nodes(&ap->ua_tree, cookie);
+
+	if (elem != NULL) {
+		uu_avl_pool_t *pp = ap->ua_pool;
+		uintptr_t *na = NODE_ARRAY(pp, elem);
+
+		na[0] = POOL_TO_MARKER(pp);
+		na[1] = 0;
+	}
+	return (elem);
+}
+
+void *
+uu_avl_find(uu_avl_t *ap, void *elem, void *private, uu_avl_index_t *out)
+{
+	struct uu_avl_node_compare_info info;
+	void *result;
+
+	info.ac_compare = ap->ua_pool->uap_cmp;
+	info.ac_private = private;
+	info.ac_right = elem;
+	info.ac_found = NULL;
+
+	result = avl_find(&ap->ua_tree, &info, out);
+	if (out != NULL)
+		*out = INDEX_ENCODE(ap, *out);
+
+	if (ap->ua_debug && result != NULL)
+		uu_panic("uu_avl_find: internal error: avl_find succeeded\n");
+
+	return (info.ac_found);
+}
+
+void
+uu_avl_insert(uu_avl_t *ap, void *elem, uu_avl_index_t idx)
+{
+	if (ap->ua_debug) {
+		uu_avl_pool_t *pp = ap->ua_pool;
+		uintptr_t *na = NODE_ARRAY(pp, elem);
+
+		if (na[1] != 0)
+			uu_panic("uu_avl_insert(%p, %p, %p): node already "
+			    "in tree, or corrupt\n",
+			    (void *)ap, elem, (void *)idx);
+		if (na[0] == 0)
+			uu_panic("uu_avl_insert(%p, %p, %p): node not "
+			    "initialized\n",
+			    (void *)ap, elem, (void *)idx);
+		if (na[0] != POOL_TO_MARKER(pp))
+			uu_panic("uu_avl_insert(%p, %p, %p): node from "
+			    "other pool, or corrupt\n",
+			    (void *)ap, elem, (void *)idx);
+
+		if (!INDEX_VALID(ap, idx))
+			uu_panic("uu_avl_insert(%p, %p, %p): %s\n",
+			    (void *)ap, elem, (void *)idx,
+			    INDEX_CHECK(idx)? "outdated index" :
+			    "invalid index");
+
+		/*
+		 * invalidate outstanding uu_avl_index_ts.
+		 */
+		ap->ua_index = INDEX_NEXT(ap->ua_index);
+	}
+	avl_insert(&ap->ua_tree, elem, INDEX_DECODE(idx));
+}
+
+void *
+uu_avl_nearest_next(uu_avl_t *ap, uu_avl_index_t idx)
+{
+	if (ap->ua_debug && !INDEX_VALID(ap, idx))
+		uu_panic("uu_avl_nearest_next(%p, %p): %s\n",
+		    (void *)ap, (void *)idx, INDEX_CHECK(idx)?
+		    "outdated index" : "invalid index");
+	return (avl_nearest(&ap->ua_tree, INDEX_DECODE(idx), AVL_AFTER));
+}
+
+void *
+uu_avl_nearest_prev(uu_avl_t *ap, uu_avl_index_t idx)
+{
+	if (ap->ua_debug && !INDEX_VALID(ap, idx))
+		uu_panic("uu_avl_nearest_prev(%p, %p): %s\n",
+		    (void *)ap, (void *)idx, INDEX_CHECK(idx)?
+		    "outdated index" : "invalid index");
+	return (avl_nearest(&ap->ua_tree, INDEX_DECODE(idx), AVL_BEFORE));
+}
+
+/*
+ * called from uu_lockup() and uu_release(), as part of our fork1()-safety.
+ */
+void
+uu_avl_lockup(void)
+{
+	uu_avl_pool_t *pp;
+
+	(void) pthread_mutex_lock(&uu_apool_list_lock);
+	for (pp = uu_null_apool.uap_next; pp != &uu_null_apool;
+	    pp = pp->uap_next)
+		(void) pthread_mutex_lock(&pp->uap_lock);
+}
+
+void
+uu_avl_release(void)
+{
+	uu_avl_pool_t *pp;
+
+	for (pp = uu_null_apool.uap_next; pp != &uu_null_apool;
+	    pp = pp->uap_next)
+		(void) pthread_mutex_unlock(&pp->uap_lock);
+	(void) pthread_mutex_unlock(&uu_apool_list_lock);
+}

lib/libuutil/uu_dprintf.c

@@ -0,0 +1,128 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include "libuutil_common.h"
+
+#include <errno.h>
+#include <libintl.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <strings.h>
+
+#define	FACILITY_FMT	"%s (%s): "
+
+#if !defined(TEXT_DOMAIN)
+#define	TEXT_DOMAIN "SYS_TEST"
+#endif
+
+static const char *
+strseverity(uu_dprintf_severity_t severity)
+{
+	switch (severity) {
+	case UU_DPRINTF_SILENT:
+		return (dgettext(TEXT_DOMAIN, "silent"));
+	case UU_DPRINTF_FATAL:
+		return (dgettext(TEXT_DOMAIN, "FATAL"));
+	case UU_DPRINTF_WARNING:
+		return (dgettext(TEXT_DOMAIN, "WARNING"));
+	case UU_DPRINTF_NOTICE:
+		return (dgettext(TEXT_DOMAIN, "note"));
+	case UU_DPRINTF_INFO:
+		return (dgettext(TEXT_DOMAIN, "info"));
+	case UU_DPRINTF_DEBUG:
+		return (dgettext(TEXT_DOMAIN, "debug"));
+	default:
+		return (dgettext(TEXT_DOMAIN, "unspecified"));
+	}
+}
+
+uu_dprintf_t *
+uu_dprintf_create(const char *name, uu_dprintf_severity_t severity,
+    uint_t flags)
+{
+	uu_dprintf_t *D;
+
+	if (uu_check_name(name, UU_NAME_DOMAIN) == -1) {
+		uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+		return (NULL);
+	}
+
+	if ((D = uu_zalloc(sizeof (uu_dprintf_t))) == NULL)
+		return (NULL);
+
+	if (name != NULL) {
+		D->uud_name = strdup(name);
+		if (D->uud_name == NULL) {
+			uu_free(D);
+			return (NULL);
+		}
+	} else {
+		D->uud_name = NULL;
+	}
+
+	D->uud_severity = severity;
+	D->uud_flags = flags;
+
+	return (D);
+}
+
+/*PRINTFLIKE3*/
+void
+uu_dprintf(uu_dprintf_t *D, uu_dprintf_severity_t severity,
+    const char *format, ...)
+{
+	va_list alist;
+
+	/* XXX Assert that severity is not UU_DPRINTF_SILENT. */
+
+	if (severity > D->uud_severity)
+		return;
+
+	(void) fprintf(stderr, FACILITY_FMT, D->uud_name,
+	    strseverity(severity));
+
+	va_start(alist, format);
+	(void) vfprintf(stderr, format, alist);
+	va_end(alist);
+}
+
+void
+uu_dprintf_destroy(uu_dprintf_t *D)
+{
+	if (D->uud_name)
+		free(D->uud_name);
+
+	uu_free(D);
+}
+
+const char *
+uu_dprintf_getname(uu_dprintf_t *D)
+{
+	return (D->uud_name);
+}

lib/libuutil/uu_ident.c

@@ -0,0 +1,122 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include "libuutil_common.h"
+
+#include <string.h>
+
+/*
+ * We require names of the form:
+ *	[provider,]identifier[/[provider,]identifier]...
+ *
+ * Where provider is either a stock symbol (SUNW) or a java-style reversed
+ * domain name (com.sun).
+ *
+ * Both providers and identifiers must start with a letter, and may
+ * only contain alphanumerics, dashes, and underlines.  Providers
+ * may also contain periods.
+ *
+ * Note that we do _not_ use the macros in <ctype.h>, since they are affected
+ * by the current locale settings.
+ */
+
+#define	IS_ALPHA(c) \
+	(((c) >= 'a' && (c) <= 'z') || ((c) >= 'A' && (c) <= 'Z'))
+
+#define	IS_DIGIT(c) \
+	((c) >= '0' && (c) <= '9')
+
+static int
+is_valid_ident(const char *s, const char *e, int allowdot)
+{
+	char c;
+
+	if (s >= e)
+		return (0);		/* name is empty */
+
+	c = *s++;
+	if (!IS_ALPHA(c))
+		return (0);		/* does not start with letter */
+
+	while (s < e && (c = *s++) != 0) {
+		if (IS_ALPHA(c) || IS_DIGIT(c) || c == '-' || c == '_' ||
+		    (allowdot && c == '.'))
+			continue;
+		return (0);		/* invalid character */
+	}
+	return (1);
+}
+
+static int
+is_valid_component(const char *b, const char *e, uint_t flags)
+{
+	char *sp;
+
+	if (flags & UU_NAME_DOMAIN) {
+		sp = strchr(b, ',');
+		if (sp != NULL && sp < e) {
+			if (!is_valid_ident(b, sp, 1))
+				return (0);
+			b = sp + 1;
+		}
+	}
+
+	return (is_valid_ident(b, e, 0));
+}
+
+int
+uu_check_name(const char *name, uint_t flags)
+{
+	const char *end = name + strlen(name);
+	const char *p;
+
+	if (flags & ~(UU_NAME_DOMAIN | UU_NAME_PATH)) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (-1);
+	}
+
+	if (!(flags & UU_NAME_PATH)) {
+		if (!is_valid_component(name, end, flags))
+			goto bad;
+		return (0);
+	}
+
+	while ((p = strchr(name, '/')) != NULL) {
+		if (!is_valid_component(name, p - 1, flags))
+			goto bad;
+		name = p + 1;
+	}
+	if (!is_valid_component(name, end, flags))
+		goto bad;
+
+	return (0);
+
+bad:
+	uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+	return (-1);
+}

lib/libuutil/uu_list.c

@@ -0,0 +1,718 @@
+/*
+ * 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"
+
+#include "libuutil_common.h"
+
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/time.h>
+
+#define	ELEM_TO_NODE(lp, e) \
+	((uu_list_node_impl_t *)((uintptr_t)(e) + (lp)->ul_offset))
+
+#define	NODE_TO_ELEM(lp, n) \
+	((void *)((uintptr_t)(n) - (lp)->ul_offset))
+
+/*
+ * uu_list_index_ts define a location for insertion.  They are simply a
+ * pointer to the object after the insertion point.  We store a mark
+ * in the low-bits of the index, to help prevent mistakes.
+ *
+ * When debugging, the index mark changes on every insert and delete, to
+ * catch stale references.
+ */
+#define	INDEX_MAX		(sizeof (uintptr_t) - 1)
+#define	INDEX_NEXT(m)		(((m) == INDEX_MAX)? 1 : ((m) + 1) & INDEX_MAX)
+
+#define	INDEX_TO_NODE(i)	((uu_list_node_impl_t *)((i) & ~INDEX_MAX))
+#define	NODE_TO_INDEX(p, n)	(((uintptr_t)(n) & ~INDEX_MAX) | (p)->ul_index)
+#define	INDEX_VALID(p, i)	(((i) & INDEX_MAX) == (p)->ul_index)
+#define	INDEX_CHECK(i)		(((i) & INDEX_MAX) != 0)
+
+#define	POOL_TO_MARKER(pp) ((void *)((uintptr_t)(pp) | 1))
+
+static uu_list_pool_t	uu_null_lpool = { &uu_null_lpool, &uu_null_lpool };
+static pthread_mutex_t	uu_lpool_list_lock = PTHREAD_MUTEX_INITIALIZER;
+
+uu_list_pool_t *
+uu_list_pool_create(const char *name, size_t objsize,
+    size_t nodeoffset, uu_compare_fn_t *compare_func, uint32_t flags)
+{
+	uu_list_pool_t *pp, *next, *prev;
+
+	if (name == NULL ||
+	    uu_check_name(name, UU_NAME_DOMAIN) == -1 ||
+	    nodeoffset + sizeof (uu_list_node_t) > objsize) {
+		uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+		return (NULL);
+	}
+
+	if (flags & ~UU_LIST_POOL_DEBUG) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (NULL);
+	}
+
+	pp = uu_zalloc(sizeof (uu_list_pool_t));
+	if (pp == NULL) {
+		uu_set_error(UU_ERROR_NO_MEMORY);
+		return (NULL);
+	}
+
+	(void) strlcpy(pp->ulp_name, name, sizeof (pp->ulp_name));
+	pp->ulp_nodeoffset = nodeoffset;
+	pp->ulp_objsize = objsize;
+	pp->ulp_cmp = compare_func;
+	if (flags & UU_LIST_POOL_DEBUG)
+		pp->ulp_debug = 1;
+	pp->ulp_last_index = 0;
+
+	(void) pthread_mutex_init(&pp->ulp_lock, NULL);
+
+	pp->ulp_null_list.ul_next_enc = UU_PTR_ENCODE(&pp->ulp_null_list);
+	pp->ulp_null_list.ul_prev_enc = UU_PTR_ENCODE(&pp->ulp_null_list);
+
+	(void) pthread_mutex_lock(&uu_lpool_list_lock);
+	pp->ulp_next = next = &uu_null_lpool;
+	pp->ulp_prev = prev = next->ulp_prev;
+	next->ulp_prev = pp;
+	prev->ulp_next = pp;
+	(void) pthread_mutex_unlock(&uu_lpool_list_lock);
+
+	return (pp);
+}
+
+void
+uu_list_pool_destroy(uu_list_pool_t *pp)
+{
+	if (pp->ulp_debug) {
+		if (pp->ulp_null_list.ul_next_enc !=
+		    UU_PTR_ENCODE(&pp->ulp_null_list) ||
+		    pp->ulp_null_list.ul_prev_enc !=
+		    UU_PTR_ENCODE(&pp->ulp_null_list)) {
+			uu_panic("uu_list_pool_destroy: Pool \"%.*s\" (%p) has "
+			    "outstanding lists, or is corrupt.\n",
+			    (int)sizeof (pp->ulp_name), pp->ulp_name,
+			    (void *)pp);
+		}
+	}
+	(void) pthread_mutex_lock(&uu_lpool_list_lock);
+	pp->ulp_next->ulp_prev = pp->ulp_prev;
+	pp->ulp_prev->ulp_next = pp->ulp_next;
+	(void) pthread_mutex_unlock(&uu_lpool_list_lock);
+	pp->ulp_prev = NULL;
+	pp->ulp_next = NULL;
+	uu_free(pp);
+}
+
+void
+uu_list_node_init(void *base, uu_list_node_t *np_arg, uu_list_pool_t *pp)
+{
+	uu_list_node_impl_t *np = (uu_list_node_impl_t *)np_arg;
+
+	if (pp->ulp_debug) {
+		uintptr_t offset = (uintptr_t)np - (uintptr_t)base;
+		if (offset + sizeof (*np) > pp->ulp_objsize) {
+			uu_panic("uu_list_node_init(%p, %p, %p (\"%s\")): "
+			    "offset %ld doesn't fit in object (size %ld)\n",
+			    base, (void *)np, (void *)pp, pp->ulp_name,
+			    (long)offset, (long)pp->ulp_objsize);
+		}
+		if (offset != pp->ulp_nodeoffset) {
+			uu_panic("uu_list_node_init(%p, %p, %p (\"%s\")): "
+			    "offset %ld doesn't match pool's offset (%ld)\n",
+			    base, (void *)np, (void *)pp, pp->ulp_name,
+			    (long)offset, (long)pp->ulp_objsize);
+		}
+	}
+	np->uln_next = POOL_TO_MARKER(pp);
+	np->uln_prev = NULL;
+}
+
+void
+uu_list_node_fini(void *base, uu_list_node_t *np_arg, uu_list_pool_t *pp)
+{
+	uu_list_node_impl_t *np = (uu_list_node_impl_t *)np_arg;
+
+	if (pp->ulp_debug) {
+		if (np->uln_next == NULL &&
+		    np->uln_prev == NULL) {
+			uu_panic("uu_list_node_fini(%p, %p, %p (\"%s\")): "
+			    "node already finied\n",
+			    base, (void *)np_arg, (void *)pp, pp->ulp_name);
+		}
+		if (np->uln_next != POOL_TO_MARKER(pp) ||
+		    np->uln_prev != NULL) {
+			uu_panic("uu_list_node_fini(%p, %p, %p (\"%s\")): "
+			    "node corrupt or on list\n",
+			    base, (void *)np_arg, (void *)pp, pp->ulp_name);
+		}
+	}
+	np->uln_next = NULL;
+	np->uln_prev = NULL;
+}
+
+uu_list_t *
+uu_list_create(uu_list_pool_t *pp, void *parent, uint32_t flags)
+{
+	uu_list_t *lp, *next, *prev;
+
+	if (flags & ~(UU_LIST_DEBUG | UU_LIST_SORTED)) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (NULL);
+	}
+
+	if ((flags & UU_LIST_SORTED) && pp->ulp_cmp == NULL) {
+		if (pp->ulp_debug)
+			uu_panic("uu_list_create(%p, ...): requested "
+			    "UU_LIST_SORTED, but pool has no comparison func\n",
+			    (void *)pp);
+		uu_set_error(UU_ERROR_NOT_SUPPORTED);
+		return (NULL);
+	}
+
+	lp = uu_zalloc(sizeof (*lp));
+	if (lp == NULL) {
+		uu_set_error(UU_ERROR_NO_MEMORY);
+		return (NULL);
+	}
+
+	lp->ul_pool = pp;
+	lp->ul_parent_enc = UU_PTR_ENCODE(parent);
+	lp->ul_offset = pp->ulp_nodeoffset;
+	lp->ul_debug = pp->ulp_debug || (flags & UU_LIST_DEBUG);
+	lp->ul_sorted = (flags & UU_LIST_SORTED);
+	lp->ul_numnodes = 0;
+	lp->ul_index = (pp->ulp_last_index = INDEX_NEXT(pp->ulp_last_index));
+
+	lp->ul_null_node.uln_next = &lp->ul_null_node;
+	lp->ul_null_node.uln_prev = &lp->ul_null_node;
+
+	lp->ul_null_walk.ulw_next = &lp->ul_null_walk;
+	lp->ul_null_walk.ulw_prev = &lp->ul_null_walk;
+
+	(void) pthread_mutex_lock(&pp->ulp_lock);
+	next = &pp->ulp_null_list;
+	prev = UU_PTR_DECODE(next->ul_prev_enc);
+	lp->ul_next_enc = UU_PTR_ENCODE(next);
+	lp->ul_prev_enc = UU_PTR_ENCODE(prev);
+	next->ul_prev_enc = UU_PTR_ENCODE(lp);
+	prev->ul_next_enc = UU_PTR_ENCODE(lp);
+	(void) pthread_mutex_unlock(&pp->ulp_lock);
+
+	return (lp);
+}
+
+void
+uu_list_destroy(uu_list_t *lp)
+{
+	uu_list_pool_t *pp = lp->ul_pool;
+
+	if (lp->ul_debug) {
+		if (lp->ul_null_node.uln_next != &lp->ul_null_node ||
+		    lp->ul_null_node.uln_prev != &lp->ul_null_node) {
+			uu_panic("uu_list_destroy(%p):  list not empty\n",
+			    (void *)lp);
+		}
+		if (lp->ul_numnodes != 0) {
+			uu_panic("uu_list_destroy(%p):  numnodes is nonzero, "
+			    "but list is empty\n", (void *)lp);
+		}
+		if (lp->ul_null_walk.ulw_next != &lp->ul_null_walk ||
+		    lp->ul_null_walk.ulw_prev != &lp->ul_null_walk) {
+			uu_panic("uu_list_destroy(%p):  outstanding walkers\n",
+			    (void *)lp);
+		}
+	}
+
+	(void) pthread_mutex_lock(&pp->ulp_lock);
+	UU_LIST_PTR(lp->ul_next_enc)->ul_prev_enc = lp->ul_prev_enc;
+	UU_LIST_PTR(lp->ul_prev_enc)->ul_next_enc = lp->ul_next_enc;
+	(void) pthread_mutex_unlock(&pp->ulp_lock);
+	lp->ul_prev_enc = UU_PTR_ENCODE(NULL);
+	lp->ul_next_enc = UU_PTR_ENCODE(NULL);
+	lp->ul_pool = NULL;
+	uu_free(lp);
+}
+
+static void
+list_insert(uu_list_t *lp, uu_list_node_impl_t *np, uu_list_node_impl_t *prev,
+    uu_list_node_impl_t *next)
+{
+	if (lp->ul_debug) {
+		if (next->uln_prev != prev || prev->uln_next != next)
+			uu_panic("insert(%p): internal error: %p and %p not "
+			    "neighbors\n", (void *)lp, (void *)next,
+			    (void *)prev);
+
+		if (np->uln_next != POOL_TO_MARKER(lp->ul_pool) ||
+		    np->uln_prev != NULL) {
+			uu_panic("insert(%p): elem %p node %p corrupt, "
+			    "not initialized, or already in a list.\n",
+			    (void *)lp, NODE_TO_ELEM(lp, np), (void *)np);
+		}
+		/*
+		 * invalidate outstanding uu_list_index_ts.
+		 */
+		lp->ul_index = INDEX_NEXT(lp->ul_index);
+	}
+	np->uln_next = next;
+	np->uln_prev = prev;
+	next->uln_prev = np;
+	prev->uln_next = np;
+
+	lp->ul_numnodes++;
+}
+
+void
+uu_list_insert(uu_list_t *lp, void *elem, uu_list_index_t idx)
+{
+	uu_list_node_impl_t *np;
+
+	np = INDEX_TO_NODE(idx);
+	if (np == NULL)
+		np = &lp->ul_null_node;
+
+	if (lp->ul_debug) {
+		if (!INDEX_VALID(lp, idx))
+			uu_panic("uu_list_insert(%p, %p, %p): %s\n",
+			    (void *)lp, elem, (void *)idx,
+			    INDEX_CHECK(idx)? "outdated index" :
+			    "invalid index");
+		if (np->uln_prev == NULL)
+			uu_panic("uu_list_insert(%p, %p, %p): out-of-date "
+			    "index\n", (void *)lp, elem, (void *)idx);
+	}
+
+	list_insert(lp, ELEM_TO_NODE(lp, elem), np->uln_prev, np);
+}
+
+void *
+uu_list_find(uu_list_t *lp, void *elem, void *private, uu_list_index_t *out)
+{
+	int sorted = lp->ul_sorted;
+	uu_compare_fn_t *func = lp->ul_pool->ulp_cmp;
+	uu_list_node_impl_t *np;
+
+	if (func == NULL) {
+		if (out != NULL)
+			*out = 0;
+		uu_set_error(UU_ERROR_NOT_SUPPORTED);
+		return (NULL);
+	}
+	for (np = lp->ul_null_node.uln_next; np != &lp->ul_null_node;
+	    np = np->uln_next) {
+		void *ep = NODE_TO_ELEM(lp, np);
+		int cmp = func(ep, elem, private);
+		if (cmp == 0) {
+			if (out != NULL)
+				*out = NODE_TO_INDEX(lp, np);
+			return (ep);
+		}
+		if (sorted && cmp > 0) {
+			if (out != NULL)
+				*out = NODE_TO_INDEX(lp, np);
+			return (NULL);
+		}
+	}
+	if (out != NULL)
+		*out = NODE_TO_INDEX(lp, 0);
+	return (NULL);
+}
+
+void *
+uu_list_nearest_next(uu_list_t *lp, uu_list_index_t idx)
+{
+	uu_list_node_impl_t *np = INDEX_TO_NODE(idx);
+
+	if (np == NULL)
+		np = &lp->ul_null_node;
+
+	if (lp->ul_debug) {
+		if (!INDEX_VALID(lp, idx))
+			uu_panic("uu_list_nearest_next(%p, %p): %s\n",
+			    (void *)lp, (void *)idx,
+			    INDEX_CHECK(idx)? "outdated index" :
+			    "invalid index");
+		if (np->uln_prev == NULL)
+			uu_panic("uu_list_nearest_next(%p, %p): out-of-date "
+			    "index\n", (void *)lp, (void *)idx);
+	}
+
+	if (np == &lp->ul_null_node)
+		return (NULL);
+	else
+		return (NODE_TO_ELEM(lp, np));
+}
+
+void *
+uu_list_nearest_prev(uu_list_t *lp, uu_list_index_t idx)
+{
+	uu_list_node_impl_t *np = INDEX_TO_NODE(idx);
+
+	if (np == NULL)
+		np = &lp->ul_null_node;
+
+	if (lp->ul_debug) {
+		if (!INDEX_VALID(lp, idx))
+			uu_panic("uu_list_nearest_prev(%p, %p): %s\n",
+			    (void *)lp, (void *)idx, INDEX_CHECK(idx)?
+			    "outdated index" : "invalid index");
+		if (np->uln_prev == NULL)
+			uu_panic("uu_list_nearest_prev(%p, %p): out-of-date "
+			    "index\n", (void *)lp, (void *)idx);
+	}
+
+	if ((np = np->uln_prev) == &lp->ul_null_node)
+		return (NULL);
+	else
+		return (NODE_TO_ELEM(lp, np));
+}
+
+static void
+list_walk_init(uu_list_walk_t *wp, uu_list_t *lp, uint32_t flags)
+{
+	uu_list_walk_t *next, *prev;
+
+	int robust = (flags & UU_WALK_ROBUST);
+	int direction = (flags & UU_WALK_REVERSE)? -1 : 1;
+
+	(void) memset(wp, 0, sizeof (*wp));
+	wp->ulw_list = lp;
+	wp->ulw_robust = robust;
+	wp->ulw_dir = direction;
+	if (direction > 0)
+		wp->ulw_next_result = lp->ul_null_node.uln_next;
+	else
+		wp->ulw_next_result = lp->ul_null_node.uln_prev;
+
+	if (lp->ul_debug || robust) {
+		/*
+		 * Add this walker to the list's list of walkers so
+		 * uu_list_remove() can advance us if somebody tries to
+		 * remove ulw_next_result.
+		 */
+		wp->ulw_next = next = &lp->ul_null_walk;
+		wp->ulw_prev = prev = next->ulw_prev;
+		next->ulw_prev = wp;
+		prev->ulw_next = wp;
+	}
+}
+
+static uu_list_node_impl_t *
+list_walk_advance(uu_list_walk_t *wp, uu_list_t *lp)
+{
+	uu_list_node_impl_t *np = wp->ulw_next_result;
+	uu_list_node_impl_t *next;
+
+	if (np == &lp->ul_null_node)
+		return (NULL);
+
+	next = (wp->ulw_dir > 0)? np->uln_next : np->uln_prev;
+
+	wp->ulw_next_result = next;
+	return (np);
+}
+
+static void
+list_walk_fini(uu_list_walk_t *wp)
+{
+	/* GLXXX debugging? */
+	if (wp->ulw_next != NULL) {
+		wp->ulw_next->ulw_prev = wp->ulw_prev;
+		wp->ulw_prev->ulw_next = wp->ulw_next;
+		wp->ulw_next = NULL;
+		wp->ulw_prev = NULL;
+	}
+	wp->ulw_list = NULL;
+	wp->ulw_next_result = NULL;
+}
+
+uu_list_walk_t *
+uu_list_walk_start(uu_list_t *lp, uint32_t flags)
+{
+	uu_list_walk_t *wp;
+
+	if (flags & ~(UU_WALK_ROBUST | UU_WALK_REVERSE)) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (NULL);
+	}
+
+	wp = uu_zalloc(sizeof (*wp));
+	if (wp == NULL) {
+		uu_set_error(UU_ERROR_NO_MEMORY);
+		return (NULL);
+	}
+
+	list_walk_init(wp, lp, flags);
+	return (wp);
+}
+
+void *
+uu_list_walk_next(uu_list_walk_t *wp)
+{
+	uu_list_t *lp = wp->ulw_list;
+	uu_list_node_impl_t *np = list_walk_advance(wp, lp);
+
+	if (np == NULL)
+		return (NULL);
+
+	return (NODE_TO_ELEM(lp, np));
+}
+
+void
+uu_list_walk_end(uu_list_walk_t *wp)
+{
+	list_walk_fini(wp);
+	uu_free(wp);
+}
+
+int
+uu_list_walk(uu_list_t *lp, uu_walk_fn_t *func, void *private, uint32_t flags)
+{
+	uu_list_node_impl_t *np;
+
+	int status = UU_WALK_NEXT;
+
+	int robust = (flags & UU_WALK_ROBUST);
+	int reverse = (flags & UU_WALK_REVERSE);
+
+	if (flags & ~(UU_WALK_ROBUST | UU_WALK_REVERSE)) {
+		uu_set_error(UU_ERROR_UNKNOWN_FLAG);
+		return (-1);
+	}
+
+	if (lp->ul_debug || robust) {
+		uu_list_walk_t my_walk;
+		void *e;
+
+		list_walk_init(&my_walk, lp, flags);
+		while (status == UU_WALK_NEXT &&
+		    (e = uu_list_walk_next(&my_walk)) != NULL)
+			status = (*func)(e, private);
+		list_walk_fini(&my_walk);
+	} else {
+		if (!reverse) {
+			for (np = lp->ul_null_node.uln_next;
+			    status == UU_WALK_NEXT && np != &lp->ul_null_node;
+			    np = np->uln_next) {
+				status = (*func)(NODE_TO_ELEM(lp, np), private);
+			}
+		} else {
+			for (np = lp->ul_null_node.uln_prev;
+			    status == UU_WALK_NEXT && np != &lp->ul_null_node;
+			    np = np->uln_prev) {
+				status = (*func)(NODE_TO_ELEM(lp, np), private);
+			}
+		}
+	}
+	if (status >= 0)
+		return (0);
+	uu_set_error(UU_ERROR_CALLBACK_FAILED);
+	return (-1);
+}
+
+void
+uu_list_remove(uu_list_t *lp, void *elem)
+{
+	uu_list_node_impl_t *np = ELEM_TO_NODE(lp, elem);
+	uu_list_walk_t *wp;
+
+	if (lp->ul_debug) {
+		if (np->uln_prev == NULL)
+			uu_panic("uu_list_remove(%p, %p): elem not on list\n",
+			    (void *)lp, elem);
+		/*
+		 * invalidate outstanding uu_list_index_ts.
+		 */
+		lp->ul_index = INDEX_NEXT(lp->ul_index);
+	}
+
+	/*
+	 * robust walkers must be advanced.  In debug mode, non-robust
+	 * walkers are also on the list.  If there are any, it's an error.
+	 */
+	for (wp = lp->ul_null_walk.ulw_next; wp != &lp->ul_null_walk;
+	    wp = wp->ulw_next) {
+		if (wp->ulw_robust) {
+			if (np == wp->ulw_next_result)
+				(void) list_walk_advance(wp, lp);
+		} else if (wp->ulw_next_result != NULL) {
+			uu_panic("uu_list_remove(%p, %p): active non-robust "
+			    "walker\n", (void *)lp, elem);
+		}
+	}
+
+	np->uln_next->uln_prev = np->uln_prev;
+	np->uln_prev->uln_next = np->uln_next;
+
+	lp->ul_numnodes--;
+
+	np->uln_next = POOL_TO_MARKER(lp->ul_pool);
+	np->uln_prev = NULL;
+}
+
+void *
+uu_list_teardown(uu_list_t *lp, void **cookie)
+{
+	void *ep;
+
+	/*
+	 * XXX: disable list modification until list is empty
+	 */
+	if (lp->ul_debug && *cookie != NULL)
+		uu_panic("uu_list_teardown(%p, %p): unexpected cookie\n",
+		    (void *)lp, (void *)cookie);
+
+	ep = uu_list_first(lp);
+	if (ep)
+		uu_list_remove(lp, ep);
+	return (ep);
+}
+
+int
+uu_list_insert_before(uu_list_t *lp, void *target, void *elem)
+{
+	uu_list_node_impl_t *np = ELEM_TO_NODE(lp, target);
+
+	if (target == NULL)
+		np = &lp->ul_null_node;
+
+	if (lp->ul_debug) {
+		if (np->uln_prev == NULL)
+			uu_panic("uu_list_insert_before(%p, %p, %p): %p is "
+			    "not currently on a list\n",
+			    (void *)lp, target, elem, target);
+	}
+	if (lp->ul_sorted) {
+		if (lp->ul_debug)
+			uu_panic("uu_list_insert_before(%p, ...): list is "
+			    "UU_LIST_SORTED\n", (void *)lp);
+		uu_set_error(UU_ERROR_NOT_SUPPORTED);
+		return (-1);
+	}
+
+	list_insert(lp, ELEM_TO_NODE(lp, elem), np->uln_prev, np);
+	return (0);
+}
+
+int
+uu_list_insert_after(uu_list_t *lp, void *target, void *elem)
+{
+	uu_list_node_impl_t *np = ELEM_TO_NODE(lp, target);
+
+	if (target == NULL)
+		np = &lp->ul_null_node;
+
+	if (lp->ul_debug) {
+		if (np->uln_prev == NULL)
+			uu_panic("uu_list_insert_after(%p, %p, %p): %p is "
+			    "not currently on a list\n",
+			    (void *)lp, target, elem, target);
+	}
+	if (lp->ul_sorted) {
+		if (lp->ul_debug)
+			uu_panic("uu_list_insert_after(%p, ...): list is "
+			    "UU_LIST_SORTED\n", (void *)lp);
+		uu_set_error(UU_ERROR_NOT_SUPPORTED);
+		return (-1);
+	}
+
+	list_insert(lp, ELEM_TO_NODE(lp, elem), np, np->uln_next);
+	return (0);
+}
+
+size_t
+uu_list_numnodes(uu_list_t *lp)
+{
+	return (lp->ul_numnodes);
+}
+
+void *
+uu_list_first(uu_list_t *lp)
+{
+	uu_list_node_impl_t *n = lp->ul_null_node.uln_next;
+	if (n == &lp->ul_null_node)
+		return (NULL);
+	return (NODE_TO_ELEM(lp, n));
+}
+
+void *
+uu_list_last(uu_list_t *lp)
+{
+	uu_list_node_impl_t *n = lp->ul_null_node.uln_prev;
+	if (n == &lp->ul_null_node)
+		return (NULL);
+	return (NODE_TO_ELEM(lp, n));
+}
+
+void *
+uu_list_next(uu_list_t *lp, void *elem)
+{
+	uu_list_node_impl_t *n = ELEM_TO_NODE(lp, elem);
+
+	n = n->uln_next;
+	if (n == &lp->ul_null_node)
+		return (NULL);
+	return (NODE_TO_ELEM(lp, n));
+}
+
+void *
+uu_list_prev(uu_list_t *lp, void *elem)
+{
+	uu_list_node_impl_t *n = ELEM_TO_NODE(lp, elem);
+
+	n = n->uln_prev;
+	if (n == &lp->ul_null_node)
+		return (NULL);
+	return (NODE_TO_ELEM(lp, n));
+}
+
+/*
+ * called from uu_lockup() and uu_release(), as part of our fork1()-safety.
+ */
+void
+uu_list_lockup(void)
+{
+	uu_list_pool_t *pp;
+
+	(void) pthread_mutex_lock(&uu_lpool_list_lock);
+	for (pp = uu_null_lpool.ulp_next; pp != &uu_null_lpool;
+	    pp = pp->ulp_next)
+		(void) pthread_mutex_lock(&pp->ulp_lock);
+}
+
+void
+uu_list_release(void)
+{
+	uu_list_pool_t *pp;
+
+	for (pp = uu_null_lpool.ulp_next; pp != &uu_null_lpool;
+	    pp = pp->ulp_next)
+		(void) pthread_mutex_unlock(&pp->ulp_lock);
+	(void) pthread_mutex_unlock(&uu_lpool_list_lock);
+}

lib/libuutil/uu_misc.c

@@ -0,0 +1,255 @@
+/*
+ * 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 "libuutil_common.h"
+
+#include <assert.h>
+#include <errno.h>
+#include <libintl.h>
+#include <pthread.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/debug.h>
+#include <thread.h>
+#include <unistd.h>
+
+#if !defined(TEXT_DOMAIN)
+#define	TEXT_DOMAIN "SYS_TEST"
+#endif
+
+/*
+ * All of the old code under !defined(PTHREAD_ONCE_KEY_NP)
+ * is here to enable the building of a native version of
+ * libuutil.so when the build machine has not yet been upgraded
+ * to a version of libc that provides pthread_key_create_once_np().
+ * It should all be deleted when solaris_nevada ships.
+ * The code is not MT-safe in a relaxed memory model.
+ */
+
+#if defined(PTHREAD_ONCE_KEY_NP)
+static pthread_key_t	uu_error_key = PTHREAD_ONCE_KEY_NP;
+#else	/* PTHREAD_ONCE_KEY_NP */
+static pthread_key_t	uu_error_key = 0;
+static pthread_mutex_t	uu_key_lock = PTHREAD_MUTEX_INITIALIZER;
+#endif	/* PTHREAD_ONCE_KEY_NP */
+
+static int		uu_error_key_setup = 0;
+
+static pthread_mutex_t	uu_panic_lock = PTHREAD_MUTEX_INITIALIZER;
+/* LINTED static unused */
+static const char	*uu_panic_format;
+/* LINTED static unused */
+static va_list		uu_panic_args;
+static pthread_t	uu_panic_thread;
+
+static uint32_t		_uu_main_error;
+
+void
+uu_set_error(uint_t code)
+{
+	if (thr_main() != 0) {
+		_uu_main_error = code;
+		return;
+	}
+#if defined(PTHREAD_ONCE_KEY_NP)
+	if (pthread_key_create_once_np(&uu_error_key, NULL) != 0)
+		uu_error_key_setup = -1;
+	else
+		uu_error_key_setup = 1;
+#else	/* PTHREAD_ONCE_KEY_NP */
+	if (uu_error_key_setup == 0) {
+		(void) pthread_mutex_lock(&uu_key_lock);
+		if (uu_error_key_setup == 0) {
+			if (pthread_key_create(&uu_error_key, NULL) != 0)
+				uu_error_key_setup = -1;
+			else
+				uu_error_key_setup = 1;
+		}
+		(void) pthread_mutex_unlock(&uu_key_lock);
+	}
+#endif	/* PTHREAD_ONCE_KEY_NP */
+	if (uu_error_key_setup > 0)
+		(void) pthread_setspecific(uu_error_key,
+		    (void *)(uintptr_t)code);
+}
+
+uint32_t
+uu_error(void)
+{
+	if (thr_main() != 0)
+		return (_uu_main_error);
+
+	if (uu_error_key_setup < 0)	/* can't happen? */
+		return (UU_ERROR_UNKNOWN);
+
+	/*
+	 * Because UU_ERROR_NONE == 0, if uu_set_error() was
+	 * never called, then this will return UU_ERROR_NONE:
+	 */
+	return ((uint32_t)(uintptr_t)pthread_getspecific(uu_error_key));
+}
+
+const char *
+uu_strerror(uint32_t code)
+{
+	const char *str;
+
+	switch (code) {
+	case UU_ERROR_NONE:
+		str = dgettext(TEXT_DOMAIN, "No error");
+		break;
+
+	case UU_ERROR_INVALID_ARGUMENT:
+		str = dgettext(TEXT_DOMAIN, "Invalid argument");
+		break;
+
+	case UU_ERROR_UNKNOWN_FLAG:
+		str = dgettext(TEXT_DOMAIN, "Unknown flag passed");
+		break;
+
+	case UU_ERROR_NO_MEMORY:
+		str = dgettext(TEXT_DOMAIN, "Out of memory");
+		break;
+
+	case UU_ERROR_CALLBACK_FAILED:
+		str = dgettext(TEXT_DOMAIN, "Callback-initiated failure");
+		break;
+
+	case UU_ERROR_NOT_SUPPORTED:
+		str = dgettext(TEXT_DOMAIN, "Operation not supported");
+		break;
+
+	case UU_ERROR_EMPTY:
+		str = dgettext(TEXT_DOMAIN, "No value provided");
+		break;
+
+	case UU_ERROR_UNDERFLOW:
+		str = dgettext(TEXT_DOMAIN, "Value too small");
+		break;
+
+	case UU_ERROR_OVERFLOW:
+		str = dgettext(TEXT_DOMAIN, "Value too large");
+		break;
+
+	case UU_ERROR_INVALID_CHAR:
+		str = dgettext(TEXT_DOMAIN,
+		    "Value contains unexpected character");
+		break;
+
+	case UU_ERROR_INVALID_DIGIT:
+		str = dgettext(TEXT_DOMAIN,
+		    "Value contains digit not in base");
+		break;
+
+	case UU_ERROR_SYSTEM:
+		str = dgettext(TEXT_DOMAIN, "Underlying system error");
+		break;
+
+	case UU_ERROR_UNKNOWN:
+		str = dgettext(TEXT_DOMAIN, "Error status not known");
+		break;
+
+	default:
+		errno = ESRCH;
+		str = NULL;
+		break;
+	}
+	return (str);
+}
+
+void
+uu_panic(const char *format, ...)
+{
+	va_list args;
+
+	va_start(args, format);
+
+	(void) pthread_mutex_lock(&uu_panic_lock);
+	if (uu_panic_thread == 0) {
+		uu_panic_thread = pthread_self();
+		uu_panic_format = format;
+		va_copy(uu_panic_args, args);
+	}
+	(void) pthread_mutex_unlock(&uu_panic_lock);
+
+	(void) vfprintf(stderr, format, args);
+
+	if (uu_panic_thread == pthread_self())
+		abort();
+	else
+		for (;;)
+			(void) pause();
+}
+
+int
+assfail(const char *astring, const char *file, int line)
+{
+	__assert(astring, file, line);
+	/*NOTREACHED*/
+	return (0);
+}
+
+static void
+uu_lockup(void)
+{
+	(void) pthread_mutex_lock(&uu_panic_lock);
+#if !defined(PTHREAD_ONCE_KEY_NP)
+	(void) pthread_mutex_lock(&uu_key_lock);
+#endif
+	uu_avl_lockup();
+	uu_list_lockup();
+}
+
+static void
+uu_release(void)
+{
+	(void) pthread_mutex_unlock(&uu_panic_lock);
+#if !defined(PTHREAD_ONCE_KEY_NP)
+	(void) pthread_mutex_unlock(&uu_key_lock);
+#endif
+	uu_avl_release();
+	uu_list_release();
+}
+
+static void
+uu_release_child(void)
+{
+	uu_panic_format = NULL;
+	uu_panic_thread = 0;
+
+	uu_release();
+}
+
+#pragma init(uu_init)
+static void
+uu_init(void)
+{
+	(void) pthread_atfork(uu_lockup, uu_release, uu_release_child);
+}

lib/libuutil/uu_open.c

@@ -0,0 +1,70 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include "libuutil_common.h"
+
+#include <sys/time.h>
+
+#include <errno.h>
+#include <fcntl.h>
+#include <limits.h>
+#include <stdio.h>
+#include <unistd.h>
+
+#ifdef _LP64
+#define	TMPPATHFMT	"%s/uu%ld"
+#else /* _LP64 */
+#define	TMPPATHFMT	"%s/uu%lld"
+#endif /* _LP64 */
+
+/*ARGSUSED*/
+int
+uu_open_tmp(const char *dir, uint_t uflags)
+{
+	int f;
+	char *fname = uu_zalloc(PATH_MAX);
+
+	if (fname == NULL)
+		return (-1);
+
+	for (;;) {
+		(void) snprintf(fname, PATH_MAX, "%s/uu%lld", dir, gethrtime());
+
+		f = open(fname, O_CREAT | O_EXCL | O_RDWR, 0600);
+
+		if (f >= 0 || errno != EEXIST)
+			break;
+	}
+
+	if (f >= 0)
+		(void) unlink(fname);
+
+	uu_free(fname);
+
+	return (f);
+}

lib/libuutil/uu_pname.c

@@ -0,0 +1,207 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2005 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include "libuutil_common.h"
+
+#include <libintl.h>
+#include <limits.h>
+#include <string.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <errno.h>
+#include <wchar.h>
+#include <unistd.h>
+
+static const char PNAME_FMT[] = "%s: ";
+static const char ERRNO_FMT[] = ": %s\n";
+
+static const char *pname;
+
+static void
+uu_die_internal(int status, const char *format, va_list alist) __NORETURN;
+
+int uu_exit_ok_value = EXIT_SUCCESS;
+int uu_exit_fatal_value = EXIT_FAILURE;
+int uu_exit_usage_value = 2;
+
+int *
+uu_exit_ok(void)
+{
+	return (&uu_exit_ok_value);
+}
+
+int *
+uu_exit_fatal(void)
+{
+	return (&uu_exit_fatal_value);
+}
+
+int *
+uu_exit_usage(void)
+{
+	return (&uu_exit_usage_value);
+}
+
+void
+uu_alt_exit(int profile)
+{
+	switch (profile) {
+	case UU_PROFILE_DEFAULT:
+		uu_exit_ok_value = EXIT_SUCCESS;
+		uu_exit_fatal_value = EXIT_FAILURE;
+		uu_exit_usage_value = 2;
+		break;
+	case UU_PROFILE_LAUNCHER:
+		uu_exit_ok_value = EXIT_SUCCESS;
+		uu_exit_fatal_value = 124;
+		uu_exit_usage_value = 125;
+		break;
+	}
+}
+
+static void
+uu_warn_internal(int err, const char *format, va_list alist)
+{
+	if (pname != NULL)
+		(void) fprintf(stderr, PNAME_FMT, pname);
+
+	(void) vfprintf(stderr, format, alist);
+
+	if (strrchr(format, '\n') == NULL)
+		(void) fprintf(stderr, ERRNO_FMT, strerror(err));
+}
+
+void
+uu_vwarn(const char *format, va_list alist)
+{
+	uu_warn_internal(errno, format, alist);
+}
+
+/*PRINTFLIKE1*/
+void
+uu_warn(const char *format, ...)
+{
+	va_list alist;
+	va_start(alist, format);
+	uu_warn_internal(errno, format, alist);
+	va_end(alist);
+}
+
+static void
+uu_die_internal(int status, const char *format, va_list alist)
+{
+	uu_warn_internal(errno, format, alist);
+#ifdef DEBUG
+	{
+		char *cp;
+
+		if (!issetugid()) {
+			cp = getenv("UU_DIE_ABORTS");
+			if (cp != NULL && *cp != '\0')
+				abort();
+		}
+	}
+#endif
+	exit(status);
+}
+
+void
+uu_vdie(const char *format, va_list alist)
+{
+	uu_die_internal(UU_EXIT_FATAL, format, alist);
+}
+
+/*PRINTFLIKE1*/
+void
+uu_die(const char *format, ...)
+{
+	va_list alist;
+	va_start(alist, format);
+	uu_die_internal(UU_EXIT_FATAL, format, alist);
+	va_end(alist);
+}
+
+void
+uu_vxdie(int status, const char *format, va_list alist)
+{
+	uu_die_internal(status, format, alist);
+}
+
+/*PRINTFLIKE2*/
+void
+uu_xdie(int status, const char *format, ...)
+{
+	va_list alist;
+	va_start(alist, format);
+	uu_die_internal(status, format, alist);
+	va_end(alist);
+}
+
+const char *
+uu_setpname(char *arg0)
+{
+	/*
+	 * Having a NULL argv[0], while uncommon, is possible.  It
+	 * makes more sense to handle this event in uu_setpname rather
+	 * than in each of its consumers.
+	 */
+	if (arg0 == NULL) {
+		pname = getexecname();
+		if (pname == NULL)
+			pname = "unknown_command";
+		return (pname);
+	}
+
+	/*
+	 * Guard against '/' at end of command invocation.
+	 */
+	for (;;) {
+		char *p = strrchr(arg0, '/');
+		if (p == NULL) {
+			pname = arg0;
+			break;
+		} else {
+			if (*(p + 1) == '\0') {
+				*p = '\0';
+				continue;
+			}
+
+			pname = p + 1;
+			break;
+		}
+	}
+
+	return (pname);
+}
+
+const char *
+uu_getpname(void)
+{
+	return (pname);
+}

lib/libuutil/uu_strtoint.c

@@ -0,0 +1,300 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (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 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident	"%Z%%M%	%I%	%E% SMI"
+
+#include "libuutil_common.h"
+
+#include <limits.h>
+#include <ctype.h>
+
+#define	MAX_BASE	36
+
+#define	IS_DIGIT(x)	((x) >= '0' && (x) <= '9')
+
+#define	CTOI(x) (((x) >= '0' && (x) <= '9') ? (x) - '0' : \
+	    ((x) >= 'a' && (x) <= 'z') ? (x) + 10 - 'a' : (x) + 10 - 'A')
+
+static int
+strtoint(const char *s_arg, uint64_t *out, uint32_t base, int sign)
+{
+	const unsigned char *s = (const unsigned char *)s_arg;
+
+	uint64_t val = 0;
+	uint64_t multmax;
+
+	unsigned c, i;
+
+	int neg = 0;
+
+	int bad_digit = 0;
+	int bad_char = 0;
+	int overflow = 0;
+
+	if (s == NULL || base == 1 || base > MAX_BASE) {
+		uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+		return (-1);
+	}
+
+	while ((c = *s) != 0 && isspace(c))
+		s++;
+
+	switch (c) {
+	case '-':
+		if (!sign)
+			overflow = 1;		/* becomes underflow below */
+		neg = 1;
+		/*FALLTHRU*/
+	case '+':
+		c = *++s;
+		break;
+	default:
+		break;
+	}
+
+	if (c == '\0') {
+		uu_set_error(UU_ERROR_EMPTY);
+		return (-1);
+	}
+
+	if (base == 0) {
+		if (c != '0')
+			base = 10;
+		else if (s[1] == 'x' || s[1] == 'X')
+			base = 16;
+		else
+			base = 8;
+	}
+
+	if (base == 16 && c == '0' && (s[1] == 'x' || s[1] == 'X'))
+		c = *(s += 2);
+
+	if ((val = CTOI(c)) >= base) {
+		if (IS_DIGIT(c))
+			bad_digit = 1;
+		else
+			bad_char = 1;
+		val = 0;
+	}
+
+	multmax = (uint64_t)UINT64_MAX / (uint64_t)base;
+
+	for (c = *++s; c != '\0'; c = *++s) {
+		if ((i = CTOI(c)) >= base) {
+			if (isspace(c))
+				break;
+			if (IS_DIGIT(c))
+				bad_digit = 1;
+			else
+				bad_char = 1;
+			i = 0;
+		}
+
+		if (val > multmax)
+			overflow = 1;
+
+		val *= base;
+		if ((uint64_t)UINT64_MAX - val < (uint64_t)i)
+			overflow = 1;
+
+		val += i;
+	}
+
+	while ((c = *s) != 0) {
+		if (!isspace(c))
+			bad_char = 1;
+		s++;
+	}
+
+	if (sign) {
+		if (neg) {
+			if (val > -(uint64_t)INT64_MIN)
+				overflow = 1;
+		} else {
+			if (val > INT64_MAX)
+				overflow = 1;
+		}
+	}
+
+	if (neg)
+		val = -val;
+
+	if (bad_char | bad_digit | overflow) {
+		if (bad_char)
+			uu_set_error(UU_ERROR_INVALID_CHAR);
+		else if (bad_digit)
+			uu_set_error(UU_ERROR_INVALID_DIGIT);
+		else if (overflow) {
+			if (neg)
+				uu_set_error(UU_ERROR_UNDERFLOW);
+			else
+				uu_set_error(UU_ERROR_OVERFLOW);
+		}
+		return (-1);
+	}
+
+	*out = val;
+	return (0);
+}
+
+int
+uu_strtoint(const char *s, void *v, size_t sz, int base,
+    int64_t min, int64_t max)
+{
+	uint64_t val_u;
+	int64_t val;
+
+	if (min > max)
+		goto bad_argument;
+
+	switch (sz) {
+	case 1:
+		if (max > INT8_MAX || min < INT8_MIN)
+			goto bad_argument;
+		break;
+	case 2:
+		if (max > INT16_MAX || min < INT16_MIN)
+			goto bad_argument;
+		break;
+	case 4:
+		if (max > INT32_MAX || min < INT32_MIN)
+			goto bad_argument;
+		break;
+	case 8:
+		if (max > INT64_MAX || min < INT64_MIN)
+			goto bad_argument;
+		break;
+	default:
+		goto bad_argument;
+	}
+
+	if (min == 0 && max == 0) {
+		min = -(1ULL << (8 * sz - 1));
+		max = (1ULL << (8 * sz - 1)) - 1;
+	}
+
+	if (strtoint(s, &val_u, base, 1) == -1)
+		return (-1);
+
+	val = (int64_t)val_u;
+
+	if (val < min) {
+		uu_set_error(UU_ERROR_UNDERFLOW);
+		return (-1);
+	} else if (val > max) {
+		uu_set_error(UU_ERROR_OVERFLOW);
+		return (-1);
+	}
+
+	switch (sz) {
+	case 1:
+		*(int8_t *)v = val;
+		return (0);
+	case 2:
+		*(int16_t *)v = val;
+		return (0);
+	case 4:
+		*(int32_t *)v = val;
+		return (0);
+	case 8:
+		*(int64_t *)v = val;
+		return (0);
+	default:
+		break;		/* fall through to bad_argument */
+	}
+
+bad_argument:
+	uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+	return (-1);
+}
+
+int
+uu_strtouint(const char *s, void *v, size_t sz, int base,
+    uint64_t min, uint64_t max)
+{
+	uint64_t val;
+
+	if (min > max)
+		goto bad_argument;
+
+	switch (sz) {
+	case 1:
+		if (max > UINT8_MAX)
+			goto bad_argument;
+		break;
+	case 2:
+		if (max > UINT16_MAX)
+			goto bad_argument;
+		break;
+	case 4:
+		if (max > UINT32_MAX)
+			goto bad_argument;
+		break;
+	case 8:
+		if (max > UINT64_MAX)
+			goto bad_argument;
+		break;
+	default:
+		goto bad_argument;
+	}
+
+	if (min == 0 && max == 0) {
+		/* we have to be careful, since << can overflow */
+		max = (1ULL << (8 * sz - 1)) * 2 - 1;
+	}
+
+	if (strtoint(s, &val, base, 0) == -1)
+		return (-1);
+
+	if (val < min) {
+		uu_set_error(UU_ERROR_UNDERFLOW);
+		return (-1);
+	} else if (val > max) {
+		uu_set_error(UU_ERROR_OVERFLOW);
+		return (-1);
+	}
+
+	switch (sz) {
+	case 1:
+		*(uint8_t *)v = val;
+		return (0);
+	case 2:
+		*(uint16_t *)v = val;
+		return (0);
+	case 4:
+		*(uint32_t *)v = val;
+		return (0);
+	case 8:
+		*(uint64_t *)v = val;
+		return (0);
+	default:
+		break;		/* shouldn't happen, fall through */
+	}
+
+bad_argument:
+	uu_set_error(UU_ERROR_INVALID_ARGUMENT);
+	return (-1);
+}

lib/libzfs/include/libzfs.h

@@ -0,0 +1,570 @@
+/*
+ * 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	_LIBZFS_H
+#define	_LIBZFS_H
+
+#include <assert.h>
+#include <libnvpair.h>
+#include <sys/param.h>
+#include <sys/types.h>
+#include <sys/varargs.h>
+#include <sys/fs/zfs.h>
+#include <sys/avl.h>
+#include <ucred.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+/*
+ * Miscellaneous ZFS constants
+ */
+#define	ZFS_MAXNAMELEN		MAXNAMELEN
+#define	ZPOOL_MAXNAMELEN	MAXNAMELEN
+#define	ZFS_MAXPROPLEN		MAXPATHLEN
+#define	ZPOOL_MAXPROPLEN	MAXPATHLEN
+
+/*
+ * libzfs errors
+ */
+enum {
+	EZFS_NOMEM = 2000,	/* out of memory */
+	EZFS_BADPROP,		/* invalid property value */
+	EZFS_PROPREADONLY,	/* cannot set readonly property */
+	EZFS_PROPTYPE,		/* property does not apply to dataset type */
+	EZFS_PROPNONINHERIT,	/* property is not inheritable */
+	EZFS_PROPSPACE,		/* bad quota or reservation */
+	EZFS_BADTYPE,		/* dataset is not of appropriate type */
+	EZFS_BUSY,		/* pool or dataset is busy */
+	EZFS_EXISTS,		/* pool or dataset already exists */
+	EZFS_NOENT,		/* no such pool or dataset */
+	EZFS_BADSTREAM,		/* bad backup stream */
+	EZFS_DSREADONLY,	/* dataset is readonly */
+	EZFS_VOLTOOBIG,		/* volume is too large for 32-bit system */
+	EZFS_VOLHASDATA,	/* volume already contains data */
+	EZFS_INVALIDNAME,	/* invalid dataset name */
+	EZFS_BADRESTORE,	/* unable to restore to destination */
+	EZFS_BADBACKUP,		/* backup failed */
+	EZFS_BADTARGET,		/* bad attach/detach/replace target */
+	EZFS_NODEVICE,		/* no such device in pool */
+	EZFS_BADDEV,		/* invalid device to add */
+	EZFS_NOREPLICAS,	/* no valid replicas */
+	EZFS_RESILVERING,	/* currently resilvering */
+	EZFS_BADVERSION,	/* unsupported version */
+	EZFS_POOLUNAVAIL,	/* pool is currently unavailable */
+	EZFS_DEVOVERFLOW,	/* too many devices in one vdev */
+	EZFS_BADPATH,		/* must be an absolute path */
+	EZFS_CROSSTARGET,	/* rename or clone across pool or dataset */
+	EZFS_ZONED,		/* used improperly in local zone */
+	EZFS_MOUNTFAILED,	/* failed to mount dataset */
+	EZFS_UMOUNTFAILED,	/* failed to unmount dataset */
+	EZFS_UNSHARENFSFAILED,	/* unshare(1M) failed */
+	EZFS_SHARENFSFAILED,	/* share(1M) failed */
+	EZFS_DEVLINKS,		/* failed to create zvol links */
+	EZFS_PERM,		/* permission denied */
+	EZFS_NOSPC,		/* out of space */
+	EZFS_IO,		/* I/O error */
+	EZFS_INTR,		/* signal received */
+	EZFS_ISSPARE,		/* device is a hot spare */
+	EZFS_INVALCONFIG,	/* invalid vdev configuration */
+	EZFS_RECURSIVE,		/* recursive dependency */
+	EZFS_NOHISTORY,		/* no history object */
+	EZFS_UNSHAREISCSIFAILED, /* iscsitgtd failed request to unshare */
+	EZFS_SHAREISCSIFAILED,	/* iscsitgtd failed request to share */
+	EZFS_POOLPROPS,		/* couldn't retrieve pool props */
+	EZFS_POOL_NOTSUP,	/* ops not supported for this type of pool */
+	EZFS_POOL_INVALARG,	/* invalid argument for this pool operation */
+	EZFS_NAMETOOLONG,	/* dataset name is too long */
+	EZFS_OPENFAILED,	/* open of device failed */
+	EZFS_NOCAP,		/* couldn't get capacity */
+	EZFS_LABELFAILED,	/* write of label failed */
+	EZFS_ISCSISVCUNAVAIL,	/* iscsi service unavailable */
+	EZFS_BADWHO,		/* invalid permission who */
+	EZFS_BADPERM,		/* invalid permission */
+	EZFS_BADPERMSET,	/* invalid permission set name */
+	EZFS_NODELEGATION,	/* delegated administration is disabled */
+	EZFS_PERMRDONLY,	/* pemissions are readonly */
+	EZFS_UNSHARESMBFAILED,	/* failed to unshare over smb */
+	EZFS_SHARESMBFAILED,	/* failed to share over smb */
+	EZFS_BADCACHE,		/* bad cache file */
+	EZFS_ISL2CACHE,		/* device is for the level 2 ARC */
+	EZFS_VDEVNOTSUP,	/* unsupported vdev type */
+	EZFS_NOTSUP,		/* ops not supported on this dataset */
+	EZFS_ACTIVE_SPARE,	/* pool has active shared spare devices */
+	EZFS_UNKNOWN
+};
+
+/*
+ * The following data structures are all part
+ * of the zfs_allow_t data structure which is
+ * used for printing 'allow' permissions.
+ * It is a linked list of zfs_allow_t's which
+ * then contain avl tree's for user/group/sets/...
+ * and each one of the entries in those trees have
+ * avl tree's for the permissions they belong to and
+ * whether they are local,descendent or local+descendent
+ * permissions.  The AVL trees are used primarily for
+ * sorting purposes, but also so that we can quickly find
+ * a given user and or permission.
+ */
+typedef struct zfs_perm_node {
+	avl_node_t z_node;
+	char z_pname[MAXPATHLEN];
+} zfs_perm_node_t;
+
+typedef struct zfs_allow_node {
+	avl_node_t z_node;
+	char z_key[MAXPATHLEN];		/* name, such as joe */
+	avl_tree_t z_localdescend;	/* local+descendent perms */
+	avl_tree_t z_local;		/* local permissions */
+	avl_tree_t z_descend;		/* descendent permissions */
+} zfs_allow_node_t;
+
+typedef struct zfs_allow {
+	struct zfs_allow *z_next;
+	char z_setpoint[MAXPATHLEN];
+	avl_tree_t z_sets;
+	avl_tree_t z_crperms;
+	avl_tree_t z_user;
+	avl_tree_t z_group;
+	avl_tree_t z_everyone;
+} zfs_allow_t;
+
+/*
+ * Basic handle types
+ */
+typedef struct zfs_handle zfs_handle_t;
+typedef struct zpool_handle zpool_handle_t;
+typedef struct libzfs_handle libzfs_handle_t;
+
+/*
+ * Library initialization
+ */
+extern libzfs_handle_t *libzfs_init(void);
+extern void libzfs_fini(libzfs_handle_t *);
+
+extern libzfs_handle_t *zpool_get_handle(zpool_handle_t *);
+extern libzfs_handle_t *zfs_get_handle(zfs_handle_t *);
+
+extern void libzfs_print_on_error(libzfs_handle_t *, boolean_t);
+
+extern int libzfs_errno(libzfs_handle_t *);
+extern const char *libzfs_error_action(libzfs_handle_t *);
+extern const char *libzfs_error_description(libzfs_handle_t *);
+
+/*
+ * Basic handle functions
+ */
+extern zpool_handle_t *zpool_open(libzfs_handle_t *, const char *);
+extern zpool_handle_t *zpool_open_canfail(libzfs_handle_t *, const char *);
+extern void zpool_close(zpool_handle_t *);
+extern const char *zpool_get_name(zpool_handle_t *);
+extern int zpool_get_state(zpool_handle_t *);
+extern char *zpool_state_to_name(vdev_state_t, vdev_aux_t);
+extern void zpool_free_handles(libzfs_handle_t *);
+
+/*
+ * Iterate over all active pools in the system.
+ */
+typedef int (*zpool_iter_f)(zpool_handle_t *, void *);
+extern int zpool_iter(libzfs_handle_t *, zpool_iter_f, void *);
+
+/*
+ * Functions to create and destroy pools
+ */
+extern int zpool_create(libzfs_handle_t *, const char *, nvlist_t *,
+    nvlist_t *, nvlist_t *);
+extern int zpool_destroy(zpool_handle_t *);
+extern int zpool_add(zpool_handle_t *, nvlist_t *);
+
+/*
+ * Functions to manipulate pool and vdev state
+ */
+extern int zpool_scrub(zpool_handle_t *, pool_scrub_type_t);
+extern int zpool_clear(zpool_handle_t *, const char *);
+
+extern int zpool_vdev_online(zpool_handle_t *, const char *, int,
+    vdev_state_t *);
+extern int zpool_vdev_offline(zpool_handle_t *, const char *, boolean_t);
+extern int zpool_vdev_attach(zpool_handle_t *, const char *,
+    const char *, nvlist_t *, int);
+extern int zpool_vdev_detach(zpool_handle_t *, const char *);
+extern int zpool_vdev_remove(zpool_handle_t *, const char *);
+
+extern int zpool_vdev_fault(zpool_handle_t *, uint64_t);
+extern int zpool_vdev_degrade(zpool_handle_t *, uint64_t);
+extern int zpool_vdev_clear(zpool_handle_t *, uint64_t);
+
+extern nvlist_t *zpool_find_vdev(zpool_handle_t *, const char *, boolean_t *,
+    boolean_t *, boolean_t *);
+extern int zpool_label_disk(libzfs_handle_t *, zpool_handle_t *, char *);
+
+/*
+ * Functions to manage pool properties
+ */
+extern int zpool_set_prop(zpool_handle_t *, const char *, const char *);
+extern int zpool_get_prop(zpool_handle_t *, zpool_prop_t, char *,
+    size_t proplen, zprop_source_t *);
+extern uint64_t zpool_get_prop_int(zpool_handle_t *, zpool_prop_t,
+    zprop_source_t *);
+
+extern const char *zpool_prop_to_name(zpool_prop_t);
+extern const char *zpool_prop_values(zpool_prop_t);
+
+/*
+ * Pool health statistics.
+ */
+typedef enum {
+	/*
+	 * The following correspond to faults as defined in the (fault.fs.zfs.*)
+	 * event namespace.  Each is associated with a corresponding message ID.
+	 */
+	ZPOOL_STATUS_CORRUPT_CACHE,	/* corrupt /kernel/drv/zpool.cache */
+	ZPOOL_STATUS_MISSING_DEV_R,	/* missing device with replicas */
+	ZPOOL_STATUS_MISSING_DEV_NR,	/* missing device with no replicas */
+	ZPOOL_STATUS_CORRUPT_LABEL_R,	/* bad device label with replicas */
+	ZPOOL_STATUS_CORRUPT_LABEL_NR,	/* bad device label with no replicas */
+	ZPOOL_STATUS_BAD_GUID_SUM,	/* sum of device guids didn't match */
+	ZPOOL_STATUS_CORRUPT_POOL,	/* pool metadata is corrupted */
+	ZPOOL_STATUS_CORRUPT_DATA,	/* data errors in user (meta)data */
+	ZPOOL_STATUS_FAILING_DEV,	/* device experiencing errors */
+	ZPOOL_STATUS_VERSION_NEWER,	/* newer on-disk version */
+	ZPOOL_STATUS_HOSTID_MISMATCH,	/* last accessed by another system */
+	ZPOOL_STATUS_IO_FAILURE_WAIT,	/* failed I/O, failmode 'wait' */
+	ZPOOL_STATUS_IO_FAILURE_CONTINUE, /* failed I/O, failmode 'continue' */
+	ZPOOL_STATUS_FAULTED_DEV_R,	/* faulted device with replicas */
+	ZPOOL_STATUS_FAULTED_DEV_NR,	/* faulted device with no replicas */
+	ZPOOL_STATUS_BAD_LOG,		/* cannot read log chain(s) */
+
+	/*
+	 * The following are not faults per se, but still an error possibly
+	 * requiring administrative attention.  There is no corresponding
+	 * message ID.
+	 */
+	ZPOOL_STATUS_VERSION_OLDER,	/* older on-disk version */
+	ZPOOL_STATUS_RESILVERING,	/* device being resilvered */
+	ZPOOL_STATUS_OFFLINE_DEV,	/* device online */
+
+	/*
+	 * Finally, the following indicates a healthy pool.
+	 */
+	ZPOOL_STATUS_OK
+} zpool_status_t;
+
+extern zpool_status_t zpool_get_status(zpool_handle_t *, char **);
+extern zpool_status_t zpool_import_status(nvlist_t *, char **);
+
+/*
+ * Statistics and configuration functions.
+ */
+extern nvlist_t *zpool_get_config(zpool_handle_t *, nvlist_t **);
+extern int zpool_refresh_stats(zpool_handle_t *, boolean_t *);
+extern int zpool_get_errlog(zpool_handle_t *, nvlist_t **);
+
+/*
+ * Import and export functions
+ */
+extern int zpool_export(zpool_handle_t *, boolean_t);
+extern int zpool_import(libzfs_handle_t *, nvlist_t *, const char *,
+    char *altroot);
+extern int zpool_import_props(libzfs_handle_t *, nvlist_t *, const char *,
+    nvlist_t *, boolean_t);
+
+/*
+ * Search for pools to import
+ */
+extern nvlist_t *zpool_find_import(libzfs_handle_t *, int, char **);
+extern nvlist_t *zpool_find_import_cached(libzfs_handle_t *, const char *,
+    char *, uint64_t);
+extern nvlist_t *zpool_find_import_byname(libzfs_handle_t *, int, char **,
+    char *);
+extern nvlist_t *zpool_find_import_byguid(libzfs_handle_t *, int, char **,
+    uint64_t);
+extern nvlist_t *zpool_find_import_activeok(libzfs_handle_t *, int, char **);
+
+/*
+ * Miscellaneous pool functions
+ */
+struct zfs_cmd;
+
+extern char *zpool_vdev_name(libzfs_handle_t *, zpool_handle_t *, nvlist_t *);
+extern int zpool_upgrade(zpool_handle_t *, uint64_t);
+extern int zpool_get_history(zpool_handle_t *, nvlist_t **);
+extern void zpool_set_history_str(const char *subcommand, int argc,
+    char **argv, char *history_str);
+extern int zpool_stage_history(libzfs_handle_t *, const char *);
+extern void zpool_obj_to_path(zpool_handle_t *, uint64_t, uint64_t, char *,
+    size_t len);
+extern int zfs_ioctl(libzfs_handle_t *, int, struct zfs_cmd *);
+extern int zpool_get_physpath(zpool_handle_t *, char *);
+/*
+ * Basic handle manipulations.  These functions do not create or destroy the
+ * underlying datasets, only the references to them.
+ */
+extern zfs_handle_t *zfs_open(libzfs_handle_t *, const char *, int);
+extern void zfs_close(zfs_handle_t *);
+extern zfs_type_t zfs_get_type(const zfs_handle_t *);
+extern const char *zfs_get_name(const zfs_handle_t *);
+extern zpool_handle_t *zfs_get_pool_handle(const zfs_handle_t *);
+
+/*
+ * Property management functions.  Some functions are shared with the kernel,
+ * and are found in sys/fs/zfs.h.
+ */
+
+/*
+ * zfs dataset property management
+ */
+extern const char *zfs_prop_default_string(zfs_prop_t);
+extern uint64_t zfs_prop_default_numeric(zfs_prop_t);
+extern const char *zfs_prop_column_name(zfs_prop_t);
+extern boolean_t zfs_prop_align_right(zfs_prop_t);
+
+extern nvlist_t *zfs_valid_proplist(libzfs_handle_t *, zfs_type_t,
+    nvlist_t *, uint64_t, zfs_handle_t *, const char *);
+
+extern const char *zfs_prop_to_name(zfs_prop_t);
+extern int zfs_prop_set(zfs_handle_t *, const char *, const char *);
+extern int zfs_prop_get(zfs_handle_t *, zfs_prop_t, char *, size_t,
+    zprop_source_t *, char *, size_t, boolean_t);
+extern int zfs_prop_get_numeric(zfs_handle_t *, zfs_prop_t, uint64_t *,
+    zprop_source_t *, char *, size_t);
+extern uint64_t zfs_prop_get_int(zfs_handle_t *, zfs_prop_t);
+extern int zfs_prop_inherit(zfs_handle_t *, const char *);
+extern const char *zfs_prop_values(zfs_prop_t);
+extern int zfs_prop_is_string(zfs_prop_t prop);
+extern nvlist_t *zfs_get_user_props(zfs_handle_t *);
+
+typedef struct zprop_list {
+	int		pl_prop;
+	char		*pl_user_prop;
+	struct zprop_list *pl_next;
+	boolean_t	pl_all;
+	size_t		pl_width;
+	boolean_t	pl_fixed;
+} zprop_list_t;
+
+extern int zfs_expand_proplist(zfs_handle_t *, zprop_list_t **);
+
+#define	ZFS_MOUNTPOINT_NONE	"none"
+#define	ZFS_MOUNTPOINT_LEGACY	"legacy"
+
+/*
+ * zpool property management
+ */
+extern int zpool_expand_proplist(zpool_handle_t *, zprop_list_t **);
+extern const char *zpool_prop_default_string(zpool_prop_t);
+extern uint64_t zpool_prop_default_numeric(zpool_prop_t);
+extern const char *zpool_prop_column_name(zpool_prop_t);
+extern boolean_t zpool_prop_align_right(zpool_prop_t);
+
+/*
+ * Functions shared by zfs and zpool property management.
+ */
+extern int zprop_iter(zprop_func func, void *cb, boolean_t show_all,
+    boolean_t ordered, zfs_type_t type);
+extern int zprop_get_list(libzfs_handle_t *, char *, zprop_list_t **,
+    zfs_type_t);
+extern void zprop_free_list(zprop_list_t *);
+
+/*
+ * Functions for printing zfs or zpool properties
+ */
+typedef struct zprop_get_cbdata {
+	int cb_sources;
+	int cb_columns[4];
+	int cb_colwidths[5];
+	boolean_t cb_scripted;
+	boolean_t cb_literal;
+	boolean_t cb_first;
+	zprop_list_t *cb_proplist;
+	zfs_type_t cb_type;
+} zprop_get_cbdata_t;
+
+void zprop_print_one_property(const char *, zprop_get_cbdata_t *,
+    const char *, const char *, zprop_source_t, const char *);
+
+#define	GET_COL_NAME		1
+#define	GET_COL_PROPERTY	2
+#define	GET_COL_VALUE		3
+#define	GET_COL_SOURCE		4
+
+/*
+ * Iterator functions.
+ */
+typedef int (*zfs_iter_f)(zfs_handle_t *, void *);
+extern int zfs_iter_root(libzfs_handle_t *, zfs_iter_f, void *);
+extern int zfs_iter_children(zfs_handle_t *, zfs_iter_f, void *);
+extern int zfs_iter_dependents(zfs_handle_t *, boolean_t, zfs_iter_f, void *);
+extern int zfs_iter_filesystems(zfs_handle_t *, zfs_iter_f, void *);
+extern int zfs_iter_snapshots(zfs_handle_t *, zfs_iter_f, void *);
+
+/*
+ * Functions to create and destroy datasets.
+ */
+extern int zfs_create(libzfs_handle_t *, const char *, zfs_type_t,
+    nvlist_t *);
+extern int zfs_create_ancestors(libzfs_handle_t *, const char *);
+extern int zfs_destroy(zfs_handle_t *);
+extern int zfs_destroy_snaps(zfs_handle_t *, char *);
+extern int zfs_clone(zfs_handle_t *, const char *, nvlist_t *);
+extern int zfs_snapshot(libzfs_handle_t *, const char *, boolean_t, nvlist_t *);
+extern int zfs_rollback(zfs_handle_t *, zfs_handle_t *, boolean_t);
+extern int zfs_rename(zfs_handle_t *, const char *, boolean_t);
+extern int zfs_send(zfs_handle_t *, const char *, const char *,
+    boolean_t, boolean_t, boolean_t, boolean_t, int);
+extern int zfs_promote(zfs_handle_t *);
+
+typedef struct recvflags {
+	/* print informational messages (ie, -v was specified) */
+	int verbose : 1;
+
+	/* the destination is a prefix, not the exact fs (ie, -d) */
+	int isprefix : 1;
+
+	/* do not actually do the recv, just check if it would work (ie, -n) */
+	int dryrun : 1;
+
+	/* rollback/destroy filesystems as necessary (eg, -F) */
+	int force : 1;
+
+	/* set "canmount=off" on all modified filesystems */
+	int canmountoff : 1;
+
+	/* byteswap flag is used internally; callers need not specify */
+	int byteswap : 1;
+} recvflags_t;
+
+extern int zfs_receive(libzfs_handle_t *, const char *, recvflags_t,
+    int, avl_tree_t *);
+
+/*
+ * Miscellaneous functions.
+ */
+extern const char *zfs_type_to_name(zfs_type_t);
+extern void zfs_refresh_properties(zfs_handle_t *);
+extern int zfs_name_valid(const char *, zfs_type_t);
+extern zfs_handle_t *zfs_path_to_zhandle(libzfs_handle_t *, char *, zfs_type_t);
+extern boolean_t zfs_dataset_exists(libzfs_handle_t *, const char *,
+    zfs_type_t);
+extern int zfs_spa_version(zfs_handle_t *, int *);
+
+/*
+ * dataset permission functions.
+ */
+extern int zfs_perm_set(zfs_handle_t *, nvlist_t *);
+extern int zfs_perm_remove(zfs_handle_t *, nvlist_t *);
+extern int zfs_build_perms(zfs_handle_t *, char *, char *,
+    zfs_deleg_who_type_t, zfs_deleg_inherit_t, nvlist_t **nvlist_t);
+extern int zfs_perm_get(zfs_handle_t *, zfs_allow_t **);
+extern void zfs_free_allows(zfs_allow_t *);
+extern void zfs_deleg_permissions(void);
+
+/*
+ * Mount support functions.
+ */
+extern boolean_t is_mounted(libzfs_handle_t *, const char *special, char **);
+extern boolean_t zfs_is_mounted(zfs_handle_t *, char **);
+extern int zfs_mount(zfs_handle_t *, const char *, int);
+extern int zfs_unmount(zfs_handle_t *, const char *, int);
+extern int zfs_unmountall(zfs_handle_t *, int);
+
+/*
+ * Share support functions.
+ */
+extern boolean_t zfs_is_shared(zfs_handle_t *);
+extern int zfs_share(zfs_handle_t *);
+extern int zfs_unshare(zfs_handle_t *);
+
+/*
+ * Protocol-specific share support functions.
+ */
+extern boolean_t zfs_is_shared_nfs(zfs_handle_t *, char **);
+extern boolean_t zfs_is_shared_smb(zfs_handle_t *, char **);
+extern int zfs_share_nfs(zfs_handle_t *);
+extern int zfs_share_smb(zfs_handle_t *);
+extern int zfs_shareall(zfs_handle_t *);
+extern int zfs_unshare_nfs(zfs_handle_t *, const char *);
+extern int zfs_unshare_smb(zfs_handle_t *, const char *);
+extern int zfs_unshareall_nfs(zfs_handle_t *);
+extern int zfs_unshareall_smb(zfs_handle_t *);
+extern int zfs_unshareall_bypath(zfs_handle_t *, const char *);
+extern int zfs_unshareall(zfs_handle_t *);
+extern boolean_t zfs_is_shared_iscsi(zfs_handle_t *);
+extern int zfs_share_iscsi(zfs_handle_t *);
+extern int zfs_unshare_iscsi(zfs_handle_t *);
+extern int zfs_iscsi_perm_check(libzfs_handle_t *, char *, ucred_t *);
+extern int zfs_deleg_share_nfs(libzfs_handle_t *, char *, char *,
+    void *, void *, int, zfs_share_op_t);
+
+/*
+ * When dealing with nvlists, verify() is extremely useful
+ */
+#ifdef NDEBUG
+#define	verify(EX)	((void)(EX))
+#else
+#define	verify(EX)	assert(EX)
+#endif
+
+/*
+ * Utility function to convert a number to a human-readable form.
+ */
+extern void zfs_nicenum(uint64_t, char *, size_t);
+extern int zfs_nicestrtonum(libzfs_handle_t *, const char *, uint64_t *);
+
+/*
+ * Given a device or file, determine if it is part of a pool.
+ */
+extern int zpool_in_use(libzfs_handle_t *, int, pool_state_t *, char **,
+    boolean_t *);
+
+/*
+ * ftyp special.  Read the label from a given device.
+ */
+extern int zpool_read_label(int, nvlist_t **);
+
+/*
+ * Create and remove zvol /dev links.
+ */
+extern int zpool_create_zvol_links(zpool_handle_t *);
+extern int zpool_remove_zvol_links(zpool_handle_t *);
+
+/* is this zvol valid for use as a dump device? */
+extern int zvol_check_dump_config(char *);
+
+/*
+ * Enable and disable datasets within a pool by mounting/unmounting and
+ * sharing/unsharing them.
+ */
+extern int zpool_enable_datasets(zpool_handle_t *, const char *, int);
+extern int zpool_disable_datasets(zpool_handle_t *, boolean_t);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _LIBZFS_H */

lib/libzfs/include/libzfs_impl.h

@@ -0,0 +1,193 @@
+/*
+ * CDDL HEADER SART
+ *
+ * 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	_LIBFS_IMPL_H
+#define	_LIBFS_IMPL_H
+
+#include <sys/dmu.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zfs_acl.h>
+#include <sys/spa.h>
+#include <sys/nvpair.h>
+
+#include <libuutil.h>
+#include <libzfs.h>
+#include <libshare.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+#ifdef	VERIFY
+#undef	VERIFY
+#endif
+#define	VERIFY	verify
+
+struct libzfs_handle {
+	int libzfs_error;
+	int libzfs_fd;
+	FILE *libzfs_mnttab;
+	FILE *libzfs_sharetab;
+	zpool_handle_t *libzfs_pool_handles;
+	uu_avl_pool_t *libzfs_ns_avlpool;
+	uu_avl_t *libzfs_ns_avl;
+	uint64_t libzfs_ns_gen;
+	int libzfs_desc_active;
+	char libzfs_action[1024];
+	char libzfs_desc[1024];
+	char *libzfs_log_str;
+	int libzfs_printerr;
+	void *libzfs_sharehdl; /* libshare handle */
+	uint_t libzfs_shareflags;
+};
+#define	ZFSSHARE_MISS	0x01	/* Didn't find entry in cache */
+
+struct zfs_handle {
+	libzfs_handle_t *zfs_hdl;
+	zpool_handle_t *zpool_hdl;
+	char zfs_name[ZFS_MAXNAMELEN];
+	zfs_type_t zfs_type; /* type including snapshot */
+	zfs_type_t zfs_head_type; /* type excluding snapshot */
+	dmu_objset_stats_t zfs_dmustats;
+	nvlist_t *zfs_props;
+	nvlist_t *zfs_user_props;
+	boolean_t zfs_mntcheck;
+	char *zfs_mntopts;
+};
+
+/*
+ * This is different from checking zfs_type, because it will also catch
+ * snapshots of volumes.
+ */
+#define	ZFS_IS_VOLUME(zhp) ((zhp)->zfs_head_type == ZFS_TYPE_VOLUME)
+
+struct zpool_handle {
+	libzfs_handle_t *zpool_hdl;
+	zpool_handle_t *zpool_next;
+	char zpool_name[ZPOOL_MAXNAMELEN];
+	int zpool_state;
+	size_t zpool_config_size;
+	nvlist_t *zpool_config;
+	nvlist_t *zpool_old_config;
+	nvlist_t *zpool_props;
+	diskaddr_t zpool_start_block;
+};
+
+typedef  enum {
+	PROTO_NFS = 0,
+	PROTO_SMB = 1,
+	PROTO_END = 2
+} zfs_share_proto_t;
+
+/*
+ * The following can be used as a bitmask and any new values
+ * added must preserve that capability.
+ */
+typedef enum {
+	SHARED_NOT_SHARED = 0x0,
+	SHARED_ISCSI = 0x1,
+	SHARED_NFS = 0x2,
+	SHARED_SMB = 0x4
+} zfs_share_type_t;
+
+int zfs_error(libzfs_handle_t *, int, const char *);
+int zfs_error_fmt(libzfs_handle_t *, int, const char *, ...);
+void zfs_error_aux(libzfs_handle_t *, const char *, ...);
+void *zfs_alloc(libzfs_handle_t *, size_t);
+void *zfs_realloc(libzfs_handle_t *, void *, size_t, size_t);
+char *zfs_strdup(libzfs_handle_t *, const char *);
+int no_memory(libzfs_handle_t *);
+
+int zfs_standard_error(libzfs_handle_t *, int, const char *);
+int zfs_standard_error_fmt(libzfs_handle_t *, int, const char *, ...);
+int zpool_standard_error(libzfs_handle_t *, int, const char *);
+int zpool_standard_error_fmt(libzfs_handle_t *, int, const char *, ...);
+
+int get_dependents(libzfs_handle_t *, boolean_t, const char *, char ***,
+    size_t *);
+
+
+int zprop_parse_value(libzfs_handle_t *, nvpair_t *, int, zfs_type_t,
+    nvlist_t *, char **, uint64_t *, const char *);
+int zprop_expand_list(libzfs_handle_t *hdl, zprop_list_t **plp,
+    zfs_type_t type);
+
+/*
+ * Use this changelist_gather() flag to force attempting mounts
+ * on each change node regardless of whether or not it is currently
+ * mounted.
+ */
+#define	CL_GATHER_MOUNT_ALWAYS	1
+
+typedef struct prop_changelist prop_changelist_t;
+
+int zcmd_alloc_dst_nvlist(libzfs_handle_t *, zfs_cmd_t *, size_t);
+int zcmd_write_src_nvlist(libzfs_handle_t *, zfs_cmd_t *, nvlist_t *);
+int zcmd_write_conf_nvlist(libzfs_handle_t *, zfs_cmd_t *, nvlist_t *);
+int zcmd_expand_dst_nvlist(libzfs_handle_t *, zfs_cmd_t *);
+int zcmd_read_dst_nvlist(libzfs_handle_t *, zfs_cmd_t *, nvlist_t **);
+void zcmd_free_nvlists(zfs_cmd_t *);
+
+int changelist_prefix(prop_changelist_t *);
+int changelist_postfix(prop_changelist_t *);
+void changelist_rename(prop_changelist_t *, const char *, const char *);
+void changelist_remove(prop_changelist_t *, const char *);
+void changelist_free(prop_changelist_t *);
+prop_changelist_t *changelist_gather(zfs_handle_t *, zfs_prop_t, int, int);
+int changelist_unshare(prop_changelist_t *, zfs_share_proto_t *);
+int changelist_haszonedchild(prop_changelist_t *);
+
+void remove_mountpoint(zfs_handle_t *);
+int create_parents(libzfs_handle_t *, char *, int);
+boolean_t isa_child_of(const char *dataset, const char *parent);
+
+zfs_handle_t *make_dataset_handle(libzfs_handle_t *, const char *);
+
+int zpool_open_silent(libzfs_handle_t *, const char *, zpool_handle_t **);
+
+int zvol_create_link(libzfs_handle_t *, const char *);
+int zvol_remove_link(libzfs_handle_t *, const char *);
+int zpool_iter_zvol(zpool_handle_t *, int (*)(const char *, void *), void *);
+boolean_t zpool_name_valid(libzfs_handle_t *, boolean_t, const char *);
+
+void namespace_clear(libzfs_handle_t *);
+
+/*
+ * libshare (sharemgr) interfaces used internally.
+ */
+
+extern int zfs_init_libshare(libzfs_handle_t *, int);
+extern void zfs_uninit_libshare(libzfs_handle_t *);
+extern int zfs_parse_options(char *, zfs_share_proto_t);
+
+extern int zfs_unshare_proto(zfs_handle_t *zhp,
+    const char *, zfs_share_proto_t *);
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _LIBFS_IMPL_H */

lib/libzfs/libzfs_changelist.c

@@ -0,0 +1,713 @@
+/*
+ * 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.
+ *
+ * Portions Copyright 2007 Ramprakash Jelari
+ */
+
+#include <libintl.h>
+#include <libuutil.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <zone.h>
+
+#include <libzfs.h>
+
+#include "libzfs_impl.h"
+
+/*
+ * Structure to keep track of dataset state.  Before changing the 'sharenfs' or
+ * 'mountpoint' property, we record whether the filesystem was previously
+ * mounted/shared.  This prior state dictates whether we remount/reshare the
+ * dataset after the property has been changed.
+ *
+ * The interface consists of the following sequence of functions:
+ *
+ * 	changelist_gather()
+ * 	changelist_prefix()
+ * 	< change property >
+ * 	changelist_postfix()
+ * 	changelist_free()
+ *
+ * Other interfaces:
+ *
+ * changelist_remove() - remove a node from a gathered list
+ * changelist_rename() - renames all datasets appropriately when doing a rename
+ * changelist_unshare() - unshares all the nodes in a given changelist
+ * changelist_haszonedchild() - check if there is any child exported to
+ *				a local zone
+ */
+typedef struct prop_changenode {
+	zfs_handle_t		*cn_handle;
+	int			cn_shared;
+	int			cn_mounted;
+	int			cn_zoned;
+	boolean_t		cn_needpost;	/* is postfix() needed? */
+	uu_list_node_t		cn_listnode;
+} prop_changenode_t;
+
+struct prop_changelist {
+	zfs_prop_t		cl_prop;
+	zfs_prop_t		cl_realprop;
+	zfs_prop_t		cl_shareprop;  /* used with sharenfs/sharesmb */
+	uu_list_pool_t		*cl_pool;
+	uu_list_t		*cl_list;
+	boolean_t		cl_waslegacy;
+	boolean_t		cl_allchildren;
+	boolean_t		cl_alldependents;
+	int			cl_mflags;	/* Mount flags */
+	int			cl_gflags;	/* Gather request flags */
+	boolean_t		cl_haszonedchild;
+	boolean_t		cl_sorted;
+};
+
+/*
+ * If the property is 'mountpoint', go through and unmount filesystems as
+ * necessary.  We don't do the same for 'sharenfs', because we can just re-share
+ * with different options without interrupting service. We do handle 'sharesmb'
+ * since there may be old resource names that need to be removed.
+ */
+int
+changelist_prefix(prop_changelist_t *clp)
+{
+	prop_changenode_t *cn;
+	int ret = 0;
+
+	if (clp->cl_prop != ZFS_PROP_MOUNTPOINT &&
+	    clp->cl_prop != ZFS_PROP_SHARESMB)
+		return (0);
+
+	for (cn = uu_list_first(clp->cl_list); cn != NULL;
+	    cn = uu_list_next(clp->cl_list, cn)) {
+
+		/* if a previous loop failed, set the remaining to false */
+		if (ret == -1) {
+			cn->cn_needpost = B_FALSE;
+			continue;
+		}
+
+		/*
+		 * If we are in the global zone, but this dataset is exported
+		 * to a local zone, do nothing.
+		 */
+		if (getzoneid() == GLOBAL_ZONEID && cn->cn_zoned)
+			continue;
+
+		if (ZFS_IS_VOLUME(cn->cn_handle)) {
+			switch (clp->cl_realprop) {
+			case ZFS_PROP_NAME:
+				/*
+				 * If this was a rename, unshare the zvol, and
+				 * remove the /dev/zvol links.
+				 */
+				(void) zfs_unshare_iscsi(cn->cn_handle);
+
+				if (zvol_remove_link(cn->cn_handle->zfs_hdl,
+				    cn->cn_handle->zfs_name) != 0) {
+					ret = -1;
+					cn->cn_needpost = B_FALSE;
+					(void) zfs_share_iscsi(cn->cn_handle);
+				}
+				break;
+
+			case ZFS_PROP_VOLSIZE:
+				/*
+				 * If this was a change to the volume size, we
+				 * need to unshare and reshare the volume.
+				 */
+				(void) zfs_unshare_iscsi(cn->cn_handle);
+				break;
+			}
+		} else {
+			/*
+			 * Do the property specific processing.
+			 */
+			switch (clp->cl_prop) {
+			case ZFS_PROP_MOUNTPOINT:
+				if (zfs_unmount(cn->cn_handle, NULL,
+				    clp->cl_mflags) != 0) {
+					ret = -1;
+					cn->cn_needpost = B_FALSE;
+				}
+				break;
+			case ZFS_PROP_SHARESMB:
+				(void) zfs_unshare_smb(cn->cn_handle, NULL);
+				break;
+			}
+		}
+	}
+
+	if (ret == -1)
+		(void) changelist_postfix(clp);
+
+	return (ret);
+}
+
+/*
+ * If the property is 'mountpoint' or 'sharenfs', go through and remount and/or
+ * reshare the filesystems as necessary.  In changelist_gather() we recorded
+ * whether the filesystem was previously shared or mounted.  The action we take
+ * depends on the previous state, and whether the value was previously 'legacy'.
+ * For non-legacy properties, we only remount/reshare the filesystem if it was
+ * previously mounted/shared.  Otherwise, we always remount/reshare the
+ * filesystem.
+ */
+int
+changelist_postfix(prop_changelist_t *clp)
+{
+	prop_changenode_t *cn;
+	char shareopts[ZFS_MAXPROPLEN];
+	int errors = 0;
+	libzfs_handle_t *hdl;
+
+	/*
+	 * If we're changing the mountpoint, attempt to destroy the underlying
+	 * mountpoint.  All other datasets will have inherited from this dataset
+	 * (in which case their mountpoints exist in the filesystem in the new
+	 * location), or have explicit mountpoints set (in which case they won't
+	 * be in the changelist).
+	 */
+	if ((cn = uu_list_last(clp->cl_list)) == NULL)
+		return (0);
+
+	if (clp->cl_prop == ZFS_PROP_MOUNTPOINT)
+		remove_mountpoint(cn->cn_handle);
+
+	/*
+	 * It is possible that the changelist_prefix() used libshare
+	 * to unshare some entries. Since libshare caches data, an
+	 * attempt to reshare during postfix can fail unless libshare
+	 * is uninitialized here so that it will reinitialize later.
+	 */
+	if (cn->cn_handle != NULL) {
+		hdl = cn->cn_handle->zfs_hdl;
+		assert(hdl != NULL);
+		zfs_uninit_libshare(hdl);
+	}
+
+	/*
+	 * We walk the datasets in reverse, because we want to mount any parent
+	 * datasets before mounting the children.  We walk all datasets even if
+	 * there are errors.
+	 */
+	for (cn = uu_list_last(clp->cl_list); cn != NULL;
+	    cn = uu_list_prev(clp->cl_list, cn)) {
+
+		boolean_t sharenfs;
+		boolean_t sharesmb;
+
+		/*
+		 * If we are in the global zone, but this dataset is exported
+		 * to a local zone, do nothing.
+		 */
+		if (getzoneid() == GLOBAL_ZONEID && cn->cn_zoned)
+			continue;
+
+		/* Only do post-processing if it's required */
+		if (!cn->cn_needpost)
+			continue;
+		cn->cn_needpost = B_FALSE;
+
+		zfs_refresh_properties(cn->cn_handle);
+
+		if (ZFS_IS_VOLUME(cn->cn_handle)) {
+			/*
+			 * If we're doing a rename, recreate the /dev/zvol
+			 * links.
+			 */
+			if (clp->cl_realprop == ZFS_PROP_NAME &&
+			    zvol_create_link(cn->cn_handle->zfs_hdl,
+			    cn->cn_handle->zfs_name) != 0) {
+				errors++;
+			} else if (cn->cn_shared ||
+			    clp->cl_prop == ZFS_PROP_SHAREISCSI) {
+				if (zfs_prop_get(cn->cn_handle,
+				    ZFS_PROP_SHAREISCSI, shareopts,
+				    sizeof (shareopts), NULL, NULL, 0,
+				    B_FALSE) == 0 &&
+				    strcmp(shareopts, "off") == 0) {
+					errors +=
+					    zfs_unshare_iscsi(cn->cn_handle);
+				} else {
+					errors +=
+					    zfs_share_iscsi(cn->cn_handle);
+				}
+			}
+
+			continue;
+		}
+
+		/*
+		 * Remount if previously mounted or mountpoint was legacy,
+		 * or sharenfs or sharesmb  property is set.
+		 */
+		sharenfs = ((zfs_prop_get(cn->cn_handle, ZFS_PROP_SHARENFS,
+		    shareopts, sizeof (shareopts), NULL, NULL, 0,
+		    B_FALSE) == 0) && (strcmp(shareopts, "off") != 0));
+
+		sharesmb = ((zfs_prop_get(cn->cn_handle, ZFS_PROP_SHARESMB,
+		    shareopts, sizeof (shareopts), NULL, NULL, 0,
+		    B_FALSE) == 0) && (strcmp(shareopts, "off") != 0));
+
+		if ((cn->cn_mounted || clp->cl_waslegacy || sharenfs ||
+		    sharesmb) && !zfs_is_mounted(cn->cn_handle, NULL) &&
+		    zfs_mount(cn->cn_handle, NULL, 0) != 0)
+			errors++;
+
+		/*
+		 * We always re-share even if the filesystem is currently
+		 * shared, so that we can adopt any new options.
+		 */
+		if (sharenfs)
+			errors += zfs_share_nfs(cn->cn_handle);
+		else if (cn->cn_shared || clp->cl_waslegacy)
+			errors += zfs_unshare_nfs(cn->cn_handle, NULL);
+		if (sharesmb)
+			errors += zfs_share_smb(cn->cn_handle);
+		else if (cn->cn_shared || clp->cl_waslegacy)
+			errors += zfs_unshare_smb(cn->cn_handle, NULL);
+	}
+
+	return (errors ? -1 : 0);
+}
+
+/*
+ * Is this "dataset" a child of "parent"?
+ */
+boolean_t
+isa_child_of(const char *dataset, const char *parent)
+{
+	int len;
+
+	len = strlen(parent);
+
+	if (strncmp(dataset, parent, len) == 0 &&
+	    (dataset[len] == '@' || dataset[len] == '/' ||
+	    dataset[len] == '\0'))
+		return (B_TRUE);
+	else
+		return (B_FALSE);
+
+}
+
+/*
+ * If we rename a filesystem, child filesystem handles are no longer valid
+ * since we identify each dataset by its name in the ZFS namespace.  As a
+ * result, we have to go through and fix up all the names appropriately.  We
+ * could do this automatically if libzfs kept track of all open handles, but
+ * this is a lot less work.
+ */
+void
+changelist_rename(prop_changelist_t *clp, const char *src, const char *dst)
+{
+	prop_changenode_t *cn;
+	char newname[ZFS_MAXNAMELEN];
+
+	for (cn = uu_list_first(clp->cl_list); cn != NULL;
+	    cn = uu_list_next(clp->cl_list, cn)) {
+		/*
+		 * Do not rename a clone that's not in the source hierarchy.
+		 */
+		if (!isa_child_of(cn->cn_handle->zfs_name, src))
+			continue;
+
+		/*
+		 * Destroy the previous mountpoint if needed.
+		 */
+		remove_mountpoint(cn->cn_handle);
+
+		(void) strlcpy(newname, dst, sizeof (newname));
+		(void) strcat(newname, cn->cn_handle->zfs_name + strlen(src));
+
+		(void) strlcpy(cn->cn_handle->zfs_name, newname,
+		    sizeof (cn->cn_handle->zfs_name));
+	}
+}
+
+/*
+ * Given a gathered changelist for the 'sharenfs' or 'sharesmb' property,
+ * unshare all the datasets in the list.
+ */
+int
+changelist_unshare(prop_changelist_t *clp, zfs_share_proto_t *proto)
+{
+	prop_changenode_t *cn;
+	int ret = 0;
+
+	if (clp->cl_prop != ZFS_PROP_SHARENFS &&
+	    clp->cl_prop != ZFS_PROP_SHARESMB)
+		return (0);
+
+	for (cn = uu_list_first(clp->cl_list); cn != NULL;
+	    cn = uu_list_next(clp->cl_list, cn)) {
+		if (zfs_unshare_proto(cn->cn_handle, NULL, proto) != 0)
+			ret = -1;
+	}
+
+	return (ret);
+}
+
+/*
+ * Check if there is any child exported to a local zone in a given changelist.
+ * This information has already been recorded while gathering the changelist
+ * via changelist_gather().
+ */
+int
+changelist_haszonedchild(prop_changelist_t *clp)
+{
+	return (clp->cl_haszonedchild);
+}
+
+/*
+ * Remove a node from a gathered list.
+ */
+void
+changelist_remove(prop_changelist_t *clp, const char *name)
+{
+	prop_changenode_t *cn;
+
+	for (cn = uu_list_first(clp->cl_list); cn != NULL;
+	    cn = uu_list_next(clp->cl_list, cn)) {
+
+		if (strcmp(cn->cn_handle->zfs_name, name) == 0) {
+			uu_list_remove(clp->cl_list, cn);
+			zfs_close(cn->cn_handle);
+			free(cn);
+			return;
+		}
+	}
+}
+
+/*
+ * Release any memory associated with a changelist.
+ */
+void
+changelist_free(prop_changelist_t *clp)
+{
+	prop_changenode_t *cn;
+	void *cookie;
+
+	if (clp->cl_list) {
+		cookie = NULL;
+		while ((cn = uu_list_teardown(clp->cl_list, &cookie)) != NULL) {
+			zfs_close(cn->cn_handle);
+			free(cn);
+		}
+
+		uu_list_destroy(clp->cl_list);
+	}
+	if (clp->cl_pool)
+		uu_list_pool_destroy(clp->cl_pool);
+
+	free(clp);
+}
+
+static int
+change_one(zfs_handle_t *zhp, void *data)
+{
+	prop_changelist_t *clp = data;
+	char property[ZFS_MAXPROPLEN];
+	char where[64];
+	prop_changenode_t *cn;
+	zprop_source_t sourcetype;
+	zprop_source_t share_sourcetype;
+
+	/*
+	 * We only want to unmount/unshare those filesystems that may inherit
+	 * from the target filesystem.  If we find any filesystem with a
+	 * locally set mountpoint, we ignore any children since changing the
+	 * property will not affect them.  If this is a rename, we iterate
+	 * over all children regardless, since we need them unmounted in
+	 * order to do the rename.  Also, if this is a volume and we're doing
+	 * a rename, then always add it to the changelist.
+	 */
+
+	if (!(ZFS_IS_VOLUME(zhp) && clp->cl_realprop == ZFS_PROP_NAME) &&
+	    zfs_prop_get(zhp, clp->cl_prop, property,
+	    sizeof (property), &sourcetype, where, sizeof (where),
+	    B_FALSE) != 0) {
+		zfs_close(zhp);
+		return (0);
+	}
+
+	/*
+	 * If we are "watching" sharenfs or sharesmb
+	 * then check out the companion property which is tracked
+	 * in cl_shareprop
+	 */
+	if (clp->cl_shareprop != ZPROP_INVAL &&
+	    zfs_prop_get(zhp, clp->cl_shareprop, property,
+	    sizeof (property), &share_sourcetype, where, sizeof (where),
+	    B_FALSE) != 0) {
+		zfs_close(zhp);
+		return (0);
+	}
+
+	if (clp->cl_alldependents || clp->cl_allchildren ||
+	    sourcetype == ZPROP_SRC_DEFAULT ||
+	    sourcetype == ZPROP_SRC_INHERITED ||
+	    (clp->cl_shareprop != ZPROP_INVAL &&
+	    (share_sourcetype == ZPROP_SRC_DEFAULT ||
+	    share_sourcetype == ZPROP_SRC_INHERITED))) {
+		if ((cn = zfs_alloc(zfs_get_handle(zhp),
+		    sizeof (prop_changenode_t))) == NULL) {
+			zfs_close(zhp);
+			return (-1);
+		}
+
+		cn->cn_handle = zhp;
+		cn->cn_mounted = (clp->cl_gflags & CL_GATHER_MOUNT_ALWAYS) ||
+		    zfs_is_mounted(zhp, NULL);
+		cn->cn_shared = zfs_is_shared(zhp);
+		cn->cn_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
+		cn->cn_needpost = B_TRUE;
+
+		/* Indicate if any child is exported to a local zone. */
+		if (getzoneid() == GLOBAL_ZONEID && cn->cn_zoned)
+			clp->cl_haszonedchild = B_TRUE;
+
+		uu_list_node_init(cn, &cn->cn_listnode, clp->cl_pool);
+
+		if (clp->cl_sorted) {
+			uu_list_index_t idx;
+
+			(void) uu_list_find(clp->cl_list, cn, NULL,
+			    &idx);
+			uu_list_insert(clp->cl_list, cn, idx);
+		} else {
+			ASSERT(!clp->cl_alldependents);
+			verify(uu_list_insert_before(clp->cl_list,
+			    uu_list_first(clp->cl_list), cn) == 0);
+		}
+
+		if (!clp->cl_alldependents)
+			return (zfs_iter_children(zhp, change_one, data));
+	} else {
+		zfs_close(zhp);
+	}
+
+	return (0);
+}
+
+/*ARGSUSED*/
+static int
+compare_mountpoints(const void *a, const void *b, void *unused)
+{
+	const prop_changenode_t *ca = a;
+	const prop_changenode_t *cb = b;
+
+	char mounta[MAXPATHLEN];
+	char mountb[MAXPATHLEN];
+
+	boolean_t hasmounta, hasmountb;
+
+	/*
+	 * When unsharing or unmounting filesystems, we need to do it in
+	 * mountpoint order.  This allows the user to have a mountpoint
+	 * hierarchy that is different from the dataset hierarchy, and still
+	 * allow it to be changed.  However, if either dataset doesn't have a
+	 * mountpoint (because it is a volume or a snapshot), we place it at the
+	 * end of the list, because it doesn't affect our change at all.
+	 */
+	hasmounta = (zfs_prop_get(ca->cn_handle, ZFS_PROP_MOUNTPOINT, mounta,
+	    sizeof (mounta), NULL, NULL, 0, B_FALSE) == 0);
+	hasmountb = (zfs_prop_get(cb->cn_handle, ZFS_PROP_MOUNTPOINT, mountb,
+	    sizeof (mountb), NULL, NULL, 0, B_FALSE) == 0);
+
+	if (!hasmounta && hasmountb)
+		return (-1);
+	else if (hasmounta && !hasmountb)
+		return (1);
+	else if (!hasmounta && !hasmountb)
+		return (0);
+	else
+		return (strcmp(mountb, mounta));
+}
+
+/*
+ * Given a ZFS handle and a property, construct a complete list of datasets
+ * that need to be modified as part of this process.  For anything but the
+ * 'mountpoint' and 'sharenfs' properties, this just returns an empty list.
+ * Otherwise, we iterate over all children and look for any datasets that
+ * inherit the property.  For each such dataset, we add it to the list and
+ * mark whether it was shared beforehand.
+ */
+prop_changelist_t *
+changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
+    int mnt_flags)
+{
+	prop_changelist_t *clp;
+	prop_changenode_t *cn;
+	zfs_handle_t *temp;
+	char property[ZFS_MAXPROPLEN];
+	uu_compare_fn_t *compare = NULL;
+
+	if ((clp = zfs_alloc(zhp->zfs_hdl, sizeof (prop_changelist_t))) == NULL)
+		return (NULL);
+
+	/*
+	 * For mountpoint-related tasks, we want to sort everything by
+	 * mountpoint, so that we mount and unmount them in the appropriate
+	 * order, regardless of their position in the hierarchy.
+	 */
+	if (prop == ZFS_PROP_NAME || prop == ZFS_PROP_ZONED ||
+	    prop == ZFS_PROP_MOUNTPOINT || prop == ZFS_PROP_SHARENFS ||
+	    prop == ZFS_PROP_SHARESMB) {
+		compare = compare_mountpoints;
+		clp->cl_sorted = B_TRUE;
+	}
+
+	clp->cl_pool = uu_list_pool_create("changelist_pool",
+	    sizeof (prop_changenode_t),
+	    offsetof(prop_changenode_t, cn_listnode),
+	    compare, 0);
+	if (clp->cl_pool == NULL) {
+		assert(uu_error() == UU_ERROR_NO_MEMORY);
+		(void) zfs_error(zhp->zfs_hdl, EZFS_NOMEM, "internal error");
+		changelist_free(clp);
+		return (NULL);
+	}
+
+	clp->cl_list = uu_list_create(clp->cl_pool, NULL,
+	    clp->cl_sorted ? UU_LIST_SORTED : 0);
+	clp->cl_gflags = gather_flags;
+	clp->cl_mflags = mnt_flags;
+
+	if (clp->cl_list == NULL) {
+		assert(uu_error() == UU_ERROR_NO_MEMORY);
+		(void) zfs_error(zhp->zfs_hdl, EZFS_NOMEM, "internal error");
+		changelist_free(clp);
+		return (NULL);
+	}
+
+	/*
+	 * If this is a rename or the 'zoned' property, we pretend we're
+	 * changing the mountpoint and flag it so we can catch all children in
+	 * change_one().
+	 *
+	 * Flag cl_alldependents to catch all children plus the dependents
+	 * (clones) that are not in the hierarchy.
+	 */
+	if (prop == ZFS_PROP_NAME) {
+		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
+		clp->cl_alldependents = B_TRUE;
+	} else if (prop == ZFS_PROP_ZONED) {
+		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
+		clp->cl_allchildren = B_TRUE;
+	} else if (prop == ZFS_PROP_CANMOUNT) {
+		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
+	} else if (prop == ZFS_PROP_VOLSIZE) {
+		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
+	} else if (prop == ZFS_PROP_VERSION) {
+		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
+	} else {
+		clp->cl_prop = prop;
+	}
+	clp->cl_realprop = prop;
+
+	if (clp->cl_prop != ZFS_PROP_MOUNTPOINT &&
+	    clp->cl_prop != ZFS_PROP_SHARENFS &&
+	    clp->cl_prop != ZFS_PROP_SHARESMB &&
+	    clp->cl_prop != ZFS_PROP_SHAREISCSI)
+		return (clp);
+
+	/*
+	 * If watching SHARENFS or SHARESMB then
+	 * also watch its companion property.
+	 */
+	if (clp->cl_prop == ZFS_PROP_SHARENFS)
+		clp->cl_shareprop = ZFS_PROP_SHARESMB;
+	else if (clp->cl_prop == ZFS_PROP_SHARESMB)
+		clp->cl_shareprop = ZFS_PROP_SHARENFS;
+
+	if (clp->cl_alldependents) {
+		if (zfs_iter_dependents(zhp, B_TRUE, change_one, clp) != 0) {
+			changelist_free(clp);
+			return (NULL);
+		}
+	} else if (zfs_iter_children(zhp, change_one, clp) != 0) {
+		changelist_free(clp);
+		return (NULL);
+	}
+
+	/*
+	 * We have to re-open ourselves because we auto-close all the handles
+	 * and can't tell the difference.
+	 */
+	if ((temp = zfs_open(zhp->zfs_hdl, zfs_get_name(zhp),
+	    ZFS_TYPE_DATASET)) == NULL) {
+		changelist_free(clp);
+		return (NULL);
+	}
+
+	/*
+	 * Always add ourself to the list.  We add ourselves to the end so that
+	 * we're the last to be unmounted.
+	 */
+	if ((cn = zfs_alloc(zhp->zfs_hdl,
+	    sizeof (prop_changenode_t))) == NULL) {
+		zfs_close(temp);
+		changelist_free(clp);
+		return (NULL);
+	}
+
+	cn->cn_handle = temp;
+	cn->cn_mounted = (clp->cl_gflags & CL_GATHER_MOUNT_ALWAYS) ||
+	    zfs_is_mounted(temp, NULL);
+	cn->cn_shared = zfs_is_shared(temp);
+	cn->cn_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
+	cn->cn_needpost = B_TRUE;
+
+	uu_list_node_init(cn, &cn->cn_listnode, clp->cl_pool);
+	if (clp->cl_sorted) {
+		uu_list_index_t idx;
+		(void) uu_list_find(clp->cl_list, cn, NULL, &idx);
+		uu_list_insert(clp->cl_list, cn, idx);
+	} else {
+		verify(uu_list_insert_after(clp->cl_list,
+		    uu_list_last(clp->cl_list), cn) == 0);
+	}
+
+	/*
+	 * If the mountpoint property was previously 'legacy', or 'none',
+	 * record it as the behavior of changelist_postfix() will be different.
+	 */
+	if ((clp->cl_prop == ZFS_PROP_MOUNTPOINT) &&
+	    (zfs_prop_get(zhp, prop, property, sizeof (property),
+	    NULL, NULL, 0, B_FALSE) == 0 &&
+	    (strcmp(property, "legacy") == 0 ||
+	    strcmp(property, "none") == 0))) {
+		/*
+		 * do not automatically mount ex-legacy datasets if
+		 * we specifically set canmount to noauto
+		 */
+		if (zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) !=
+		    ZFS_CANMOUNT_NOAUTO)
+			clp->cl_waslegacy = B_TRUE;
+	}
+
+	return (clp);
+}

lib/libzfs/libzfs_config.c

@@ -0,0 +1,360 @@
+/*
+ * 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"
+
+/*
+ * The pool configuration repository is stored in /etc/zfs/zpool.cache as a
+ * single packed nvlist.  While it would be nice to just read in this
+ * file from userland, this wouldn't work from a local zone.  So we have to have
+ * a zpool ioctl to return the complete configuration for all pools.  In the
+ * global zone, this will be identical to reading the file and unpacking it in
+ * userland.
+ */
+
+#include <errno.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <stddef.h>
+#include <string.h>
+#include <unistd.h>
+#include <libintl.h>
+#include <libuutil.h>
+
+#include "libzfs_impl.h"
+
+typedef struct config_node {
+	char		*cn_name;
+	nvlist_t	*cn_config;
+	uu_avl_node_t	cn_avl;
+} config_node_t;
+
+/* ARGSUSED */
+static int
+config_node_compare(const void *a, const void *b, void *unused)
+{
+	int ret;
+
+	const config_node_t *ca = (config_node_t *)a;
+	const config_node_t *cb = (config_node_t *)b;
+
+	ret = strcmp(ca->cn_name, cb->cn_name);
+
+	if (ret < 0)
+		return (-1);
+	else if (ret > 0)
+		return (1);
+	else
+		return (0);
+}
+
+void
+namespace_clear(libzfs_handle_t *hdl)
+{
+	if (hdl->libzfs_ns_avl) {
+		config_node_t *cn;
+		void *cookie = NULL;
+
+		while ((cn = uu_avl_teardown(hdl->libzfs_ns_avl,
+		    &cookie)) != NULL) {
+			nvlist_free(cn->cn_config);
+			free(cn->cn_name);
+			free(cn);
+		}
+
+		uu_avl_destroy(hdl->libzfs_ns_avl);
+		hdl->libzfs_ns_avl = NULL;
+	}
+
+	if (hdl->libzfs_ns_avlpool) {
+		uu_avl_pool_destroy(hdl->libzfs_ns_avlpool);
+		hdl->libzfs_ns_avlpool = NULL;
+	}
+}
+
+/*
+ * Loads the pool namespace, or re-loads it if the cache has changed.
+ */
+static int
+namespace_reload(libzfs_handle_t *hdl)
+{
+	nvlist_t *config;
+	config_node_t *cn;
+	nvpair_t *elem;
+	zfs_cmd_t zc = { 0 };
+	void *cookie;
+
+	if (hdl->libzfs_ns_gen == 0) {
+		/*
+		 * This is the first time we've accessed the configuration
+		 * cache.  Initialize the AVL tree and then fall through to the
+		 * common code.
+		 */
+		if ((hdl->libzfs_ns_avlpool = uu_avl_pool_create("config_pool",
+		    sizeof (config_node_t),
+		    offsetof(config_node_t, cn_avl),
+		    config_node_compare, UU_DEFAULT)) == NULL)
+			return (no_memory(hdl));
+
+		if ((hdl->libzfs_ns_avl = uu_avl_create(hdl->libzfs_ns_avlpool,
+		    NULL, UU_DEFAULT)) == NULL)
+			return (no_memory(hdl));
+	}
+
+	if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
+		return (-1);
+
+	for (;;) {
+		zc.zc_cookie = hdl->libzfs_ns_gen;
+		if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_CONFIGS, &zc) != 0) {
+			switch (errno) {
+			case EEXIST:
+				/*
+				 * The namespace hasn't changed.
+				 */
+				zcmd_free_nvlists(&zc);
+				return (0);
+
+			case ENOMEM:
+				if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
+					zcmd_free_nvlists(&zc);
+					return (-1);
+				}
+				break;
+
+			default:
+				zcmd_free_nvlists(&zc);
+				return (zfs_standard_error(hdl, errno,
+				    dgettext(TEXT_DOMAIN, "failed to read "
+				    "pool configuration")));
+			}
+		} else {
+			hdl->libzfs_ns_gen = zc.zc_cookie;
+			break;
+		}
+	}
+
+	if (zcmd_read_dst_nvlist(hdl, &zc, &config) != 0) {
+		zcmd_free_nvlists(&zc);
+		return (-1);
+	}
+
+	zcmd_free_nvlists(&zc);
+
+	/*
+	 * Clear out any existing configuration information.
+	 */
+	cookie = NULL;
+	while ((cn = uu_avl_teardown(hdl->libzfs_ns_avl, &cookie)) != NULL) {
+		nvlist_free(cn->cn_config);
+		free(cn->cn_name);
+		free(cn);
+	}
+
+	elem = NULL;
+	while ((elem = nvlist_next_nvpair(config, elem)) != NULL) {
+		nvlist_t *child;
+		uu_avl_index_t where;
+
+		if ((cn = zfs_alloc(hdl, sizeof (config_node_t))) == NULL) {
+			nvlist_free(config);
+			return (-1);
+		}
+
+		if ((cn->cn_name = zfs_strdup(hdl,
+		    nvpair_name(elem))) == NULL) {
+			free(cn);
+			nvlist_free(config);
+			return (-1);
+		}
+
+		verify(nvpair_value_nvlist(elem, &child) == 0);
+		if (nvlist_dup(child, &cn->cn_config, 0) != 0) {
+			free(cn->cn_name);
+			free(cn);
+			nvlist_free(config);
+			return (no_memory(hdl));
+		}
+		verify(uu_avl_find(hdl->libzfs_ns_avl, cn, NULL, &where)
+		    == NULL);
+
+		uu_avl_insert(hdl->libzfs_ns_avl, cn, where);
+	}
+
+	nvlist_free(config);
+	return (0);
+}
+
+/*
+ * Retrieve the configuration for the given pool.  The configuration is a nvlist
+ * describing the vdevs, as well as the statistics associated with each one.
+ */
+nvlist_t *
+zpool_get_config(zpool_handle_t *zhp, nvlist_t **oldconfig)
+{
+	if (oldconfig)
+		*oldconfig = zhp->zpool_old_config;
+	return (zhp->zpool_config);
+}
+
+/*
+ * Refresh the vdev statistics associated with the given pool.  This is used in
+ * iostat to show configuration changes and determine the delta from the last
+ * time the function was called.  This function can fail, in case the pool has
+ * been destroyed.
+ */
+int
+zpool_refresh_stats(zpool_handle_t *zhp, boolean_t *missing)
+{
+	zfs_cmd_t zc = { 0 };
+	int error;
+	nvlist_t *config;
+	libzfs_handle_t *hdl = zhp->zpool_hdl;
+
+	*missing = B_FALSE;
+	(void) strcpy(zc.zc_name, zhp->zpool_name);
+
+	if (zhp->zpool_config_size == 0)
+		zhp->zpool_config_size = 1 << 16;
+
+	if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size) != 0)
+		return (-1);
+
+	for (;;) {
+		if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_POOL_STATS,
+		    &zc) == 0) {
+			/*
+			 * The real error is returned in the zc_cookie field.
+			 */
+			error = zc.zc_cookie;
+			break;
+		}
+
+		if (errno == ENOMEM) {
+			if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
+				zcmd_free_nvlists(&zc);
+				return (-1);
+			}
+		} else {
+			zcmd_free_nvlists(&zc);
+			if (errno == ENOENT || errno == EINVAL)
+				*missing = B_TRUE;
+			zhp->zpool_state = POOL_STATE_UNAVAIL;
+			return (0);
+		}
+	}
+
+	if (zcmd_read_dst_nvlist(hdl, &zc, &config) != 0) {
+		zcmd_free_nvlists(&zc);
+		return (-1);
+	}
+
+	zcmd_free_nvlists(&zc);
+
+	zhp->zpool_config_size = zc.zc_nvlist_dst_size;
+
+	if (zhp->zpool_config != NULL) {
+		uint64_t oldtxg, newtxg;
+
+		verify(nvlist_lookup_uint64(zhp->zpool_config,
+		    ZPOOL_CONFIG_POOL_TXG, &oldtxg) == 0);
+		verify(nvlist_lookup_uint64(config,
+		    ZPOOL_CONFIG_POOL_TXG, &newtxg) == 0);
+
+		if (zhp->zpool_old_config != NULL)
+			nvlist_free(zhp->zpool_old_config);
+
+		if (oldtxg != newtxg) {
+			nvlist_free(zhp->zpool_config);
+			zhp->zpool_old_config = NULL;
+		} else {
+			zhp->zpool_old_config = zhp->zpool_config;
+		}
+	}
+
+	zhp->zpool_config = config;
+	if (error)
+		zhp->zpool_state = POOL_STATE_UNAVAIL;
+	else
+		zhp->zpool_state = POOL_STATE_ACTIVE;
+
+	return (0);
+}
+
+/*
+ * Iterate over all pools in the system.
+ */
+int
+zpool_iter(libzfs_handle_t *hdl, zpool_iter_f func, void *data)
+{
+	config_node_t *cn;
+	zpool_handle_t *zhp;
+	int ret;
+
+	if (namespace_reload(hdl) != 0)
+		return (-1);
+
+	for (cn = uu_avl_first(hdl->libzfs_ns_avl); cn != NULL;
+	    cn = uu_avl_next(hdl->libzfs_ns_avl, cn)) {
+
+		if (zpool_open_silent(hdl, cn->cn_name, &zhp) != 0)
+			return (-1);
+
+		if (zhp == NULL)
+			continue;
+
+		if ((ret = func(zhp, data)) != 0)
+			return (ret);
+	}
+
+	return (0);
+}
+
+/*
+ * Iterate over root datasets, calling the given function for each.  The zfs
+ * handle passed each time must be explicitly closed by the callback.
+ */
+int
+zfs_iter_root(libzfs_handle_t *hdl, zfs_iter_f func, void *data)
+{
+	config_node_t *cn;
+	zfs_handle_t *zhp;
+	int ret;
+
+	if (namespace_reload(hdl) != 0)
+		return (-1);
+
+	for (cn = uu_avl_first(hdl->libzfs_ns_avl); cn != NULL;
+	    cn = uu_avl_next(hdl->libzfs_ns_avl, cn)) {
+
+		if ((zhp = make_dataset_handle(hdl, cn->cn_name)) == NULL)
+			continue;
+
+		if ((ret = func(zhp, data)) != 0)
+			return (ret);
+	}
+
+	return (0);
+}

lib/libzfs/libzfs_graph.c

@@ -0,0 +1,662 @@
+/*
+ * 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"
+
+/*
+ * Iterate over all children of the current object.  This includes the normal
+ * dataset hierarchy, but also arbitrary hierarchies due to clones.  We want to
+ * walk all datasets in the pool, and construct a directed graph of the form:
+ *
+ * 			home
+ *                        |
+ *                   +----+----+
+ *                   |         |
+ *                   v         v             ws
+ *                  bar       baz             |
+ *                             |              |
+ *                             v              v
+ *                          @yesterday ----> foo
+ *
+ * In order to construct this graph, we have to walk every dataset in the pool,
+ * because the clone parent is stored as a property of the child, not the
+ * parent.  The parent only keeps track of the number of clones.
+ *
+ * In the normal case (without clones) this would be rather expensive.  To avoid
+ * unnecessary computation, we first try a walk of the subtree hierarchy
+ * starting from the initial node.  At each dataset, we construct a node in the
+ * graph and an edge leading from its parent.  If we don't see any snapshots
+ * with a non-zero clone count, then we are finished.
+ *
+ * If we do find a cloned snapshot, then we finish the walk of the current
+ * subtree, but indicate that we need to do a complete walk.  We then perform a
+ * global walk of all datasets, avoiding the subtree we already processed.
+ *
+ * At the end of this, we'll end up with a directed graph of all relevant (and
+ * possible some irrelevant) datasets in the system.  We need to both find our
+ * limiting subgraph and determine a safe ordering in which to destroy the
+ * datasets.  We do a topological ordering of our graph starting at our target
+ * dataset, and then walk the results in reverse.
+ *
+ * It's possible for the graph to have cycles if, for example, the user renames
+ * a clone to be the parent of its origin snapshot.  The user can request to
+ * generate an error in this case, or ignore the cycle and continue.
+ *
+ * When removing datasets, we want to destroy the snapshots in chronological
+ * order (because this is the most efficient method).  In order to accomplish
+ * this, we store the creation transaction group with each vertex and keep each
+ * vertex's edges sorted according to this value.  The topological sort will
+ * automatically walk the snapshots in the correct order.
+ */
+
+#include <assert.h>
+#include <libintl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <strings.h>
+#include <unistd.h>
+
+#include <libzfs.h>
+
+#include "libzfs_impl.h"
+#include "zfs_namecheck.h"
+
+#define	MIN_EDGECOUNT	4
+
+/*
+ * Vertex structure.  Indexed by dataset name, this structure maintains a list
+ * of edges to other vertices.
+ */
+struct zfs_edge;
+typedef struct zfs_vertex {
+	char			zv_dataset[ZFS_MAXNAMELEN];
+	struct zfs_vertex	*zv_next;
+	int			zv_visited;
+	uint64_t		zv_txg;
+	struct zfs_edge		**zv_edges;
+	int			zv_edgecount;
+	int			zv_edgealloc;
+} zfs_vertex_t;
+
+enum {
+	VISIT_SEEN = 1,
+	VISIT_SORT_PRE,
+	VISIT_SORT_POST
+};
+
+/*
+ * Edge structure.  Simply maintains a pointer to the destination vertex.  There
+ * is no need to store the source vertex, since we only use edges in the context
+ * of the source vertex.
+ */
+typedef struct zfs_edge {
+	zfs_vertex_t		*ze_dest;
+	struct zfs_edge		*ze_next;
+} zfs_edge_t;
+
+#define	ZFS_GRAPH_SIZE		1027	/* this could be dynamic some day */
+
+/*
+ * Graph structure.  Vertices are maintained in a hash indexed by dataset name.
+ */
+typedef struct zfs_graph {
+	zfs_vertex_t		**zg_hash;
+	size_t			zg_size;
+	size_t			zg_nvertex;
+	const char		*zg_root;
+	int			zg_clone_count;
+} zfs_graph_t;
+
+/*
+ * Allocate a new edge pointing to the target vertex.
+ */
+static zfs_edge_t *
+zfs_edge_create(libzfs_handle_t *hdl, zfs_vertex_t *dest)
+{
+	zfs_edge_t *zep = zfs_alloc(hdl, sizeof (zfs_edge_t));
+
+	if (zep == NULL)
+		return (NULL);
+
+	zep->ze_dest = dest;
+
+	return (zep);
+}
+
+/*
+ * Destroy an edge.
+ */
+static void
+zfs_edge_destroy(zfs_edge_t *zep)
+{
+	free(zep);
+}
+
+/*
+ * Allocate a new vertex with the given name.
+ */
+static zfs_vertex_t *
+zfs_vertex_create(libzfs_handle_t *hdl, const char *dataset)
+{
+	zfs_vertex_t *zvp = zfs_alloc(hdl, sizeof (zfs_vertex_t));
+
+	if (zvp == NULL)
+		return (NULL);
+
+	assert(strlen(dataset) < ZFS_MAXNAMELEN);
+
+	(void) strlcpy(zvp->zv_dataset, dataset, sizeof (zvp->zv_dataset));
+
+	if ((zvp->zv_edges = zfs_alloc(hdl,
+	    MIN_EDGECOUNT * sizeof (void *))) == NULL) {
+		free(zvp);
+		return (NULL);
+	}
+
+	zvp->zv_edgealloc = MIN_EDGECOUNT;
+
+	return (zvp);
+}
+
+/*
+ * Destroy a vertex.  Frees up any associated edges.
+ */
+static void
+zfs_vertex_destroy(zfs_vertex_t *zvp)
+{
+	int i;
+
+	for (i = 0; i < zvp->zv_edgecount; i++)
+		zfs_edge_destroy(zvp->zv_edges[i]);
+
+	free(zvp->zv_edges);
+	free(zvp);
+}
+
+/*
+ * Given a vertex, add an edge to the destination vertex.
+ */
+static int
+zfs_vertex_add_edge(libzfs_handle_t *hdl, zfs_vertex_t *zvp,
+    zfs_vertex_t *dest)
+{
+	zfs_edge_t *zep = zfs_edge_create(hdl, dest);
+
+	if (zep == NULL)
+		return (-1);
+
+	if (zvp->zv_edgecount == zvp->zv_edgealloc) {
+		void *ptr;
+
+		if ((ptr = zfs_realloc(hdl, zvp->zv_edges,
+		    zvp->zv_edgealloc * sizeof (void *),
+		    zvp->zv_edgealloc * 2 * sizeof (void *))) == NULL)
+			return (-1);
+
+		zvp->zv_edges = ptr;
+		zvp->zv_edgealloc *= 2;
+	}
+
+	zvp->zv_edges[zvp->zv_edgecount++] = zep;
+
+	return (0);
+}
+
+static int
+zfs_edge_compare(const void *a, const void *b)
+{
+	const zfs_edge_t *ea = *((zfs_edge_t **)a);
+	const zfs_edge_t *eb = *((zfs_edge_t **)b);
+
+	if (ea->ze_dest->zv_txg < eb->ze_dest->zv_txg)
+		return (-1);
+	if (ea->ze_dest->zv_txg > eb->ze_dest->zv_txg)
+		return (1);
+	return (0);
+}
+
+/*
+ * Sort the given vertex edges according to the creation txg of each vertex.
+ */
+static void
+zfs_vertex_sort_edges(zfs_vertex_t *zvp)
+{
+	if (zvp->zv_edgecount == 0)
+		return;
+
+	qsort(zvp->zv_edges, zvp->zv_edgecount, sizeof (void *),
+	    zfs_edge_compare);
+}
+
+/*
+ * Construct a new graph object.  We allow the size to be specified as a
+ * parameter so in the future we can size the hash according to the number of
+ * datasets in the pool.
+ */
+static zfs_graph_t *
+zfs_graph_create(libzfs_handle_t *hdl, const char *dataset, size_t size)
+{
+	zfs_graph_t *zgp = zfs_alloc(hdl, sizeof (zfs_graph_t));
+
+	if (zgp == NULL)
+		return (NULL);
+
+	zgp->zg_size = size;
+	if ((zgp->zg_hash = zfs_alloc(hdl,
+	    size * sizeof (zfs_vertex_t *))) == NULL) {
+		free(zgp);
+		return (NULL);
+	}
+
+	zgp->zg_root = dataset;
+	zgp->zg_clone_count = 0;
+
+	return (zgp);
+}
+
+/*
+ * Destroy a graph object.  We have to iterate over all the hash chains,
+ * destroying each vertex in the process.
+ */
+static void
+zfs_graph_destroy(zfs_graph_t *zgp)
+{
+	int i;
+	zfs_vertex_t *current, *next;
+
+	for (i = 0; i < zgp->zg_size; i++) {
+		current = zgp->zg_hash[i];
+		while (current != NULL) {
+			next = current->zv_next;
+			zfs_vertex_destroy(current);
+			current = next;
+		}
+	}
+
+	free(zgp->zg_hash);
+	free(zgp);
+}
+
+/*
+ * Graph hash function.  Classic bernstein k=33 hash function, taken from
+ * usr/src/cmd/sgs/tools/common/strhash.c
+ */
+static size_t
+zfs_graph_hash(zfs_graph_t *zgp, const char *str)
+{
+	size_t hash = 5381;
+	int c;
+
+	while ((c = *str++) != 0)
+		hash = ((hash << 5) + hash) + c; /* hash * 33 + c */
+
+	return (hash % zgp->zg_size);
+}
+
+/*
+ * Given a dataset name, finds the associated vertex, creating it if necessary.
+ */
+static zfs_vertex_t *
+zfs_graph_lookup(libzfs_handle_t *hdl, zfs_graph_t *zgp, const char *dataset,
+    uint64_t txg)
+{
+	size_t idx = zfs_graph_hash(zgp, dataset);
+	zfs_vertex_t *zvp;
+
+	for (zvp = zgp->zg_hash[idx]; zvp != NULL; zvp = zvp->zv_next) {
+		if (strcmp(zvp->zv_dataset, dataset) == 0) {
+			if (zvp->zv_txg == 0)
+				zvp->zv_txg = txg;
+			return (zvp);
+		}
+	}
+
+	if ((zvp = zfs_vertex_create(hdl, dataset)) == NULL)
+		return (NULL);
+
+	zvp->zv_next = zgp->zg_hash[idx];
+	zvp->zv_txg = txg;
+	zgp->zg_hash[idx] = zvp;
+	zgp->zg_nvertex++;
+
+	return (zvp);
+}
+
+/*
+ * Given two dataset names, create an edge between them.  For the source vertex,
+ * mark 'zv_visited' to indicate that we have seen this vertex, and not simply
+ * created it as a destination of another edge.  If 'dest' is NULL, then this
+ * is an individual vertex (i.e. the starting vertex), so don't add an edge.
+ */
+static int
+zfs_graph_add(libzfs_handle_t *hdl, zfs_graph_t *zgp, const char *source,
+    const char *dest, uint64_t txg)
+{
+	zfs_vertex_t *svp, *dvp;
+
+	if ((svp = zfs_graph_lookup(hdl, zgp, source, 0)) == NULL)
+		return (-1);
+	svp->zv_visited = VISIT_SEEN;
+	if (dest != NULL) {
+		dvp = zfs_graph_lookup(hdl, zgp, dest, txg);
+		if (dvp == NULL)
+			return (-1);
+		if (zfs_vertex_add_edge(hdl, svp, dvp) != 0)
+			return (-1);
+	}
+
+	return (0);
+}
+
+/*
+ * Iterate over all children of the given dataset, adding any vertices
+ * as necessary.  Returns -1 if there was an error, or 0 otherwise.
+ * This is a simple recursive algorithm - the ZFS namespace typically
+ * is very flat.  We manually invoke the necessary ioctl() calls to
+ * avoid the overhead and additional semantics of zfs_open().
+ */
+static int
+iterate_children(libzfs_handle_t *hdl, zfs_graph_t *zgp, const char *dataset)
+{
+	zfs_cmd_t zc = { 0 };
+	zfs_vertex_t *zvp;
+
+	/*
+	 * Look up the source vertex, and avoid it if we've seen it before.
+	 */
+	zvp = zfs_graph_lookup(hdl, zgp, dataset, 0);
+	if (zvp == NULL)
+		return (-1);
+	if (zvp->zv_visited == VISIT_SEEN)
+		return (0);
+
+	/*
+	 * Iterate over all children
+	 */
+	for ((void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
+	    ioctl(hdl->libzfs_fd, ZFS_IOC_DATASET_LIST_NEXT, &zc) == 0;
+	    (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name))) {
+
+		/*
+		 * Ignore private dataset names.
+		 */
+		if (dataset_name_hidden(zc.zc_name))
+			continue;
+
+		/*
+		 * Get statistics for this dataset, to determine the type of the
+		 * dataset and clone statistics.  If this fails, the dataset has
+		 * since been removed, and we're pretty much screwed anyway.
+		 */
+		zc.zc_objset_stats.dds_origin[0] = '\0';
+		if (ioctl(hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0)
+			continue;
+
+		if (zc.zc_objset_stats.dds_origin[0] != '\0') {
+			if (zfs_graph_add(hdl, zgp,
+			    zc.zc_objset_stats.dds_origin, zc.zc_name,
+			    zc.zc_objset_stats.dds_creation_txg) != 0)
+				return (-1);
+			/*
+			 * Count origins only if they are contained in the graph
+			 */
+			if (isa_child_of(zc.zc_objset_stats.dds_origin,
+			    zgp->zg_root))
+				zgp->zg_clone_count--;
+		}
+
+		/*
+		 * Add an edge between the parent and the child.
+		 */
+		if (zfs_graph_add(hdl, zgp, dataset, zc.zc_name,
+		    zc.zc_objset_stats.dds_creation_txg) != 0)
+			return (-1);
+
+		/*
+		 * Recursively visit child
+		 */
+		if (iterate_children(hdl, zgp, zc.zc_name))
+			return (-1);
+	}
+
+	/*
+	 * Now iterate over all snapshots.
+	 */
+	bzero(&zc, sizeof (zc));
+
+	for ((void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
+	    ioctl(hdl->libzfs_fd, ZFS_IOC_SNAPSHOT_LIST_NEXT, &zc) == 0;
+	    (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name))) {
+
+		/*
+		 * Get statistics for this dataset, to determine the type of the
+		 * dataset and clone statistics.  If this fails, the dataset has
+		 * since been removed, and we're pretty much screwed anyway.
+		 */
+		if (ioctl(hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0)
+			continue;
+
+		/*
+		 * Add an edge between the parent and the child.
+		 */
+		if (zfs_graph_add(hdl, zgp, dataset, zc.zc_name,
+		    zc.zc_objset_stats.dds_creation_txg) != 0)
+			return (-1);
+
+		zgp->zg_clone_count += zc.zc_objset_stats.dds_num_clones;
+	}
+
+	zvp->zv_visited = VISIT_SEEN;
+
+	return (0);
+}
+
+/*
+ * Returns false if there are no snapshots with dependent clones in this
+ * subtree or if all of those clones are also in this subtree.  Returns
+ * true if there is an error or there are external dependents.
+ */
+static boolean_t
+external_dependents(libzfs_handle_t *hdl, zfs_graph_t *zgp, const char *dataset)
+{
+	zfs_cmd_t zc = { 0 };
+
+	/*
+	 * Check whether this dataset is a clone or has clones since
+	 * iterate_children() only checks the children.
+	 */
+	(void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
+	if (ioctl(hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0)
+		return (B_TRUE);
+
+	if (zc.zc_objset_stats.dds_origin[0] != '\0') {
+		if (zfs_graph_add(hdl, zgp,
+		    zc.zc_objset_stats.dds_origin, zc.zc_name,
+		    zc.zc_objset_stats.dds_creation_txg) != 0)
+			return (B_TRUE);
+		if (isa_child_of(zc.zc_objset_stats.dds_origin, dataset))
+			zgp->zg_clone_count--;
+	}
+
+	if ((zc.zc_objset_stats.dds_num_clones) ||
+	    iterate_children(hdl, zgp, dataset))
+		return (B_TRUE);
+
+	return (zgp->zg_clone_count != 0);
+}
+
+/*
+ * Construct a complete graph of all necessary vertices.  First, iterate over
+ * only our object's children.  If no cloned snapshots are found, or all of
+ * the cloned snapshots are in this subtree then return a graph of the subtree.
+ * Otherwise, start at the root of the pool and iterate over all datasets.
+ */
+static zfs_graph_t *
+construct_graph(libzfs_handle_t *hdl, const char *dataset)
+{
+	zfs_graph_t *zgp = zfs_graph_create(hdl, dataset, ZFS_GRAPH_SIZE);
+	int ret = 0;
+
+	if (zgp == NULL)
+		return (zgp);
+
+	if ((strchr(dataset, '/') == NULL) ||
+	    (external_dependents(hdl, zgp, dataset))) {
+		/*
+		 * Determine pool name and try again.
+		 */
+		int len = strcspn(dataset, "/@") + 1;
+		char *pool = zfs_alloc(hdl, len);
+
+		if (pool == NULL) {
+			zfs_graph_destroy(zgp);
+			return (NULL);
+		}
+		(void) strlcpy(pool, dataset, len);
+
+		if (iterate_children(hdl, zgp, pool) == -1 ||
+		    zfs_graph_add(hdl, zgp, pool, NULL, 0) != 0) {
+			free(pool);
+			zfs_graph_destroy(zgp);
+			return (NULL);
+		}
+		free(pool);
+	}
+
+	if (ret == -1 || zfs_graph_add(hdl, zgp, dataset, NULL, 0) != 0) {
+		zfs_graph_destroy(zgp);
+		return (NULL);
+	}
+
+	return (zgp);
+}
+
+/*
+ * Given a graph, do a recursive topological sort into the given array.  This is
+ * really just a depth first search, so that the deepest nodes appear first.
+ * hijack the 'zv_visited' marker to avoid visiting the same vertex twice.
+ */
+static int
+topo_sort(libzfs_handle_t *hdl, boolean_t allowrecursion, char **result,
+    size_t *idx, zfs_vertex_t *zgv)
+{
+	int i;
+
+	if (zgv->zv_visited == VISIT_SORT_PRE && !allowrecursion) {
+		/*
+		 * If we've already seen this vertex as part of our depth-first
+		 * search, then we have a cyclic dependency, and we must return
+		 * an error.
+		 */
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+		    "recursive dependency at '%s'"),
+		    zgv->zv_dataset);
+		return (zfs_error(hdl, EZFS_RECURSIVE,
+		    dgettext(TEXT_DOMAIN,
+		    "cannot determine dependent datasets")));
+	} else if (zgv->zv_visited >= VISIT_SORT_PRE) {
+		/*
+		 * If we've already processed this as part of the topological
+		 * sort, then don't bother doing so again.
+		 */
+		return (0);
+	}
+
+	zgv->zv_visited = VISIT_SORT_PRE;
+
+	/* avoid doing a search if we don't have to */
+	zfs_vertex_sort_edges(zgv);
+	for (i = 0; i < zgv->zv_edgecount; i++) {
+		if (topo_sort(hdl, allowrecursion, result, idx,
+		    zgv->zv_edges[i]->ze_dest) != 0)
+			return (-1);
+	}
+
+	/* we may have visited this in the course of the above */
+	if (zgv->zv_visited == VISIT_SORT_POST)
+		return (0);
+
+	if ((result[*idx] = zfs_alloc(hdl,
+	    strlen(zgv->zv_dataset) + 1)) == NULL)
+		return (-1);
+
+	(void) strcpy(result[*idx], zgv->zv_dataset);
+	*idx += 1;
+	zgv->zv_visited = VISIT_SORT_POST;
+	return (0);
+}
+
+/*
+ * The only public interface for this file.  Do the dirty work of constructing a
+ * child list for the given object.  Construct the graph, do the toplogical
+ * sort, and then return the array of strings to the caller.
+ *
+ * The 'allowrecursion' parameter controls behavior when cycles are found.  If
+ * it is set, the the cycle is ignored and the results returned as if the cycle
+ * did not exist.  If it is not set, then the routine will generate an error if
+ * a cycle is found.
+ */
+int
+get_dependents(libzfs_handle_t *hdl, boolean_t allowrecursion,
+    const char *dataset, char ***result, size_t *count)
+{
+	zfs_graph_t *zgp;
+	zfs_vertex_t *zvp;
+
+	if ((zgp = construct_graph(hdl, dataset)) == NULL)
+		return (-1);
+
+	if ((*result = zfs_alloc(hdl,
+	    zgp->zg_nvertex * sizeof (char *))) == NULL) {
+		zfs_graph_destroy(zgp);
+		return (-1);
+	}
+
+	if ((zvp = zfs_graph_lookup(hdl, zgp, dataset, 0)) == NULL) {
+		free(*result);
+		zfs_graph_destroy(zgp);
+		return (-1);
+	}
+
+	*count = 0;
+	if (topo_sort(hdl, allowrecursion, *result, count, zvp) != 0) {
+		free(*result);
+		zfs_graph_destroy(zgp);
+		return (-1);
+	}
+
+	/*
+	 * Get rid of the last entry, which is our starting vertex and not
+	 * strictly a dependent.
+	 */
+	assert(*count > 0);
+	free((*result)[*count - 1]);
+	(*count)--;
+
+	zfs_graph_destroy(zgp);
+
+	return (0);
+}

lib/libzfs/libzfs_status.c

@@ -0,0 +1,317 @@
+/*
+ * 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.
+ */
+
+/*
+ * This file contains the functions which analyze the status of a pool.  This
+ * include both the status of an active pool, as well as the status exported
+ * pools.  Returns one of the ZPOOL_STATUS_* defines describing the status of
+ * the pool.  This status is independent (to a certain degree) from the state of
+ * the pool.  A pool's state describes only whether or not it is capable of
+ * providing the necessary fault tolerance for data.  The status describes the
+ * overall status of devices.  A pool that is online can still have a device
+ * that is experiencing errors.
+ *
+ * Only a subset of the possible faults can be detected using 'zpool status',
+ * and not all possible errors correspond to a FMA message ID.  The explanation
+ * is left up to the caller, depending on whether it is a live pool or an
+ * import.
+ */
+
+#include <libzfs.h>
+#include <string.h>
+#include <unistd.h>
+#include "libzfs_impl.h"
+
+/*
+ * Message ID table.  This must be kept in sync with the ZPOOL_STATUS_* defines
+ * in libzfs.h.  Note that there are some status results which go past the end
+ * of this table, and hence have no associated message ID.
+ */
+static char *zfs_msgid_table[] = {
+	"ZFS-8000-14",
+	"ZFS-8000-2Q",
+	"ZFS-8000-3C",
+	"ZFS-8000-4J",
+	"ZFS-8000-5E",
+	"ZFS-8000-6X",
+	"ZFS-8000-72",
+	"ZFS-8000-8A",
+	"ZFS-8000-9P",
+	"ZFS-8000-A5",
+	"ZFS-8000-EY",
+	"ZFS-8000-HC",
+	"ZFS-8000-JQ",
+	"ZFS-8000-K4",
+};
+
+#define	NMSGID	(sizeof (zfs_msgid_table) / sizeof (zfs_msgid_table[0]))
+
+/* ARGSUSED */
+static int
+vdev_missing(uint64_t state, uint64_t aux, uint64_t errs)
+{
+	return (state == VDEV_STATE_CANT_OPEN &&
+	    aux == VDEV_AUX_OPEN_FAILED);
+}
+
+/* ARGSUSED */
+static int
+vdev_faulted(uint64_t state, uint64_t aux, uint64_t errs)
+{
+	return (state == VDEV_STATE_FAULTED);
+}
+
+/* ARGSUSED */
+static int
+vdev_errors(uint64_t state, uint64_t aux, uint64_t errs)
+{
+	return (state == VDEV_STATE_DEGRADED || errs != 0);
+}
+
+/* ARGSUSED */
+static int
+vdev_broken(uint64_t state, uint64_t aux, uint64_t errs)
+{
+	return (state == VDEV_STATE_CANT_OPEN);
+}
+
+/* ARGSUSED */
+static int
+vdev_offlined(uint64_t state, uint64_t aux, uint64_t errs)
+{
+	return (state == VDEV_STATE_OFFLINE);
+}
+
+/*
+ * Detect if any leaf devices that have seen errors or could not be opened.
+ */
+static boolean_t
+find_vdev_problem(nvlist_t *vdev, int (*func)(uint64_t, uint64_t, uint64_t))
+{
+	nvlist_t **child;
+	vdev_stat_t *vs;
+	uint_t c, children;
+	char *type;
+
+	/*
+	 * Ignore problems within a 'replacing' vdev, since we're presumably in
+	 * the process of repairing any such errors, and don't want to call them
+	 * out again.  We'll pick up the fact that a resilver is happening
+	 * later.
+	 */
+	verify(nvlist_lookup_string(vdev, ZPOOL_CONFIG_TYPE, &type) == 0);
+	if (strcmp(type, VDEV_TYPE_REPLACING) == 0)
+		return (B_FALSE);
+
+	if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, &child,
+	    &children) == 0) {
+		for (c = 0; c < children; c++)
+			if (find_vdev_problem(child[c], func))
+				return (B_TRUE);
+	} else {
+		verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_STATS,
+		    (uint64_t **)&vs, &c) == 0);
+
+		if (func(vs->vs_state, vs->vs_aux,
+		    vs->vs_read_errors +
+		    vs->vs_write_errors +
+		    vs->vs_checksum_errors))
+			return (B_TRUE);
+	}
+
+	return (B_FALSE);
+}
+
+/*
+ * Active pool health status.
+ *
+ * To determine the status for a pool, we make several passes over the config,
+ * picking the most egregious error we find.  In order of importance, we do the
+ * following:
+ *
+ *	- Check for a complete and valid configuration
+ *	- Look for any faulted or missing devices in a non-replicated config
+ *	- Check for any data errors
+ *	- Check for any faulted or missing devices in a replicated config
+ *	- Look for any devices showing errors
+ *	- Check for any resilvering devices
+ *
+ * There can obviously be multiple errors within a single pool, so this routine
+ * only picks the most damaging of all the current errors to report.
+ */
+static zpool_status_t
+check_status(nvlist_t *config, boolean_t isimport)
+{
+	nvlist_t *nvroot;
+	vdev_stat_t *vs;
+	uint_t vsc;
+	uint64_t nerr;
+	uint64_t version;
+	uint64_t stateval;
+	uint64_t suspended;
+	uint64_t hostid = 0;
+
+	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
+	    &version) == 0);
+	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
+	    &nvroot) == 0);
+	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_STATS,
+	    (uint64_t **)&vs, &vsc) == 0);
+	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
+	    &stateval) == 0);
+	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
+
+	/*
+	 * Pool last accessed by another system.
+	 */
+	if (hostid != 0 && (unsigned long)hostid != gethostid() &&
+	    stateval == POOL_STATE_ACTIVE)
+		return (ZPOOL_STATUS_HOSTID_MISMATCH);
+
+	/*
+	 * Newer on-disk version.
+	 */
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    vs->vs_aux == VDEV_AUX_VERSION_NEWER)
+		return (ZPOOL_STATUS_VERSION_NEWER);
+
+	/*
+	 * Check that the config is complete.
+	 */
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    vs->vs_aux == VDEV_AUX_BAD_GUID_SUM)
+		return (ZPOOL_STATUS_BAD_GUID_SUM);
+
+	/*
+	 * Check whether the pool has suspended due to failed I/O.
+	 */
+	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED,
+	    &suspended) == 0) {
+		if (suspended == ZIO_FAILURE_MODE_CONTINUE)
+			return (ZPOOL_STATUS_IO_FAILURE_CONTINUE);
+		return (ZPOOL_STATUS_IO_FAILURE_WAIT);
+	}
+
+	/*
+	 * Could not read a log.
+	 */
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    vs->vs_aux == VDEV_AUX_BAD_LOG) {
+		return (ZPOOL_STATUS_BAD_LOG);
+	}
+
+	/*
+	 * Bad devices in non-replicated config.
+	 */
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    find_vdev_problem(nvroot, vdev_faulted))
+		return (ZPOOL_STATUS_FAULTED_DEV_NR);
+
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    find_vdev_problem(nvroot, vdev_missing))
+		return (ZPOOL_STATUS_MISSING_DEV_NR);
+
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    find_vdev_problem(nvroot, vdev_broken))
+		return (ZPOOL_STATUS_CORRUPT_LABEL_NR);
+
+	/*
+	 * Corrupted pool metadata
+	 */
+	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
+	    vs->vs_aux == VDEV_AUX_CORRUPT_DATA)
+		return (ZPOOL_STATUS_CORRUPT_POOL);
+
+	/*
+	 * Persistent data errors.
+	 */
+	if (!isimport) {
+		if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
+		    &nerr) == 0 && nerr != 0)
+			return (ZPOOL_STATUS_CORRUPT_DATA);
+	}
+
+	/*
+	 * Missing devices in a replicated config.
+	 */
+	if (find_vdev_problem(nvroot, vdev_faulted))
+		return (ZPOOL_STATUS_FAULTED_DEV_R);
+	if (find_vdev_problem(nvroot, vdev_missing))
+		return (ZPOOL_STATUS_MISSING_DEV_R);
+	if (find_vdev_problem(nvroot, vdev_broken))
+		return (ZPOOL_STATUS_CORRUPT_LABEL_R);
+
+	/*
+	 * Devices with errors
+	 */
+	if (!isimport && find_vdev_problem(nvroot, vdev_errors))
+		return (ZPOOL_STATUS_FAILING_DEV);
+
+	/*
+	 * Offlined devices
+	 */
+	if (find_vdev_problem(nvroot, vdev_offlined))
+		return (ZPOOL_STATUS_OFFLINE_DEV);
+
+	/*
+	 * Currently resilvering
+	 */
+	if (!vs->vs_scrub_complete && vs->vs_scrub_type == POOL_SCRUB_RESILVER)
+		return (ZPOOL_STATUS_RESILVERING);
+
+	/*
+	 * Outdated, but usable, version
+	 */
+	if (version < SPA_VERSION)
+		return (ZPOOL_STATUS_VERSION_OLDER);
+
+	return (ZPOOL_STATUS_OK);
+}
+
+zpool_status_t
+zpool_get_status(zpool_handle_t *zhp, char **msgid)
+{
+	zpool_status_t ret = check_status(zhp->zpool_config, B_FALSE);
+
+	if (ret >= NMSGID)
+		*msgid = NULL;
+	else
+		*msgid = zfs_msgid_table[ret];
+
+	return (ret);
+}
+
+zpool_status_t
+zpool_import_status(nvlist_t *config, char **msgid)
+{
+	zpool_status_t ret = check_status(config, B_TRUE);
+
+	if (ret >= NMSGID)
+		*msgid = NULL;
+	else
+		*msgid = zfs_msgid_table[ret];
+
+	return (ret);
+}

lib/libzpool/kernel.c

@@ -0,0 +1,885 @@
+/*
+ * 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 <assert.h>
+#include <fcntl.h>
+#include <poll.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <zlib.h>
+#include <sys/spa.h>
+#include <sys/stat.h>
+#include <sys/processor.h>
+#include <sys/zfs_context.h>
+#include <sys/zmod.h>
+#include <sys/utsname.h>
+
+/*
+ * Emulation of kernel services in userland.
+ */
+
+uint64_t physmem;
+vnode_t *rootdir = (vnode_t *)0xabcd1234;
+char hw_serial[11];
+
+struct utsname utsname = {
+	"userland", "libzpool", "1", "1", "na"
+};
+
+/*
+ * =========================================================================
+ * threads
+ * =========================================================================
+ */
+/*ARGSUSED*/
+kthread_t *
+zk_thread_create(void (*func)(), void *arg)
+{
+	thread_t tid;
+
+	VERIFY(thr_create(0, 0, (void *(*)(void *))func, arg, THR_DETACHED,
+	    &tid) == 0);
+
+	return ((void *)(uintptr_t)tid);
+}
+
+/*
+ * =========================================================================
+ * kstats
+ * =========================================================================
+ */
+/*ARGSUSED*/
+kstat_t *
+kstat_create(char *module, int instance, char *name, char *class,
+    uchar_t type, ulong_t ndata, uchar_t ks_flag)
+{
+	return (NULL);
+}
+
+/*ARGSUSED*/
+void
+kstat_install(kstat_t *ksp)
+{}
+
+/*ARGSUSED*/
+void
+kstat_delete(kstat_t *ksp)
+{}
+
+/*
+ * =========================================================================
+ * mutexes
+ * =========================================================================
+ */
+void
+zmutex_init(kmutex_t *mp)
+{
+	mp->m_owner = NULL;
+	mp->initialized = B_TRUE;
+	(void) _mutex_init(&mp->m_lock, USYNC_THREAD, NULL);
+}
+
+void
+zmutex_destroy(kmutex_t *mp)
+{
+	ASSERT(mp->initialized == B_TRUE);
+	ASSERT(mp->m_owner == NULL);
+	(void) _mutex_destroy(&(mp)->m_lock);
+	mp->m_owner = (void *)-1UL;
+	mp->initialized = B_FALSE;
+}
+
+void
+mutex_enter(kmutex_t *mp)
+{
+	ASSERT(mp->initialized == B_TRUE);
+	ASSERT(mp->m_owner != (void *)-1UL);
+	ASSERT(mp->m_owner != curthread);
+	VERIFY(mutex_lock(&mp->m_lock) == 0);
+	ASSERT(mp->m_owner == NULL);
+	mp->m_owner = curthread;
+}
+
+int
+mutex_tryenter(kmutex_t *mp)
+{
+	ASSERT(mp->initialized == B_TRUE);
+	ASSERT(mp->m_owner != (void *)-1UL);
+	if (0 == mutex_trylock(&mp->m_lock)) {
+		ASSERT(mp->m_owner == NULL);
+		mp->m_owner = curthread;
+		return (1);
+	} else {
+		return (0);
+	}
+}
+
+void
+mutex_exit(kmutex_t *mp)
+{
+	ASSERT(mp->initialized == B_TRUE);
+	ASSERT(mutex_owner(mp) == curthread);
+	mp->m_owner = NULL;
+	VERIFY(mutex_unlock(&mp->m_lock) == 0);
+}
+
+void *
+mutex_owner(kmutex_t *mp)
+{
+	ASSERT(mp->initialized == B_TRUE);
+	return (mp->m_owner);
+}
+
+/*
+ * =========================================================================
+ * rwlocks
+ * =========================================================================
+ */
+/*ARGSUSED*/
+void
+rw_init(krwlock_t *rwlp, char *name, int type, void *arg)
+{
+	rwlock_init(&rwlp->rw_lock, USYNC_THREAD, NULL);
+	rwlp->rw_owner = NULL;
+	rwlp->initialized = B_TRUE;
+}
+
+void
+rw_destroy(krwlock_t *rwlp)
+{
+	rwlock_destroy(&rwlp->rw_lock);
+	rwlp->rw_owner = (void *)-1UL;
+	rwlp->initialized = B_FALSE;
+}
+
+void
+rw_enter(krwlock_t *rwlp, krw_t rw)
+{
+	ASSERT(!RW_LOCK_HELD(rwlp));
+	ASSERT(rwlp->initialized == B_TRUE);
+	ASSERT(rwlp->rw_owner != (void *)-1UL);
+	ASSERT(rwlp->rw_owner != curthread);
+
+	if (rw == RW_READER)
+		VERIFY(rw_rdlock(&rwlp->rw_lock) == 0);
+	else
+		VERIFY(rw_wrlock(&rwlp->rw_lock) == 0);
+
+	rwlp->rw_owner = curthread;
+}
+
+void
+rw_exit(krwlock_t *rwlp)
+{
+	ASSERT(rwlp->initialized == B_TRUE);
+	ASSERT(rwlp->rw_owner != (void *)-1UL);
+
+	rwlp->rw_owner = NULL;
+	VERIFY(rw_unlock(&rwlp->rw_lock) == 0);
+}
+
+int
+rw_tryenter(krwlock_t *rwlp, krw_t rw)
+{
+	int rv;
+
+	ASSERT(rwlp->initialized == B_TRUE);
+	ASSERT(rwlp->rw_owner != (void *)-1UL);
+
+	if (rw == RW_READER)
+		rv = rw_tryrdlock(&rwlp->rw_lock);
+	else
+		rv = rw_trywrlock(&rwlp->rw_lock);
+
+	if (rv == 0) {
+		rwlp->rw_owner = curthread;
+		return (1);
+	}
+
+	return (0);
+}
+
+/*ARGSUSED*/
+int
+rw_tryupgrade(krwlock_t *rwlp)
+{
+	ASSERT(rwlp->initialized == B_TRUE);
+	ASSERT(rwlp->rw_owner != (void *)-1UL);
+
+	return (0);
+}
+
+/*
+ * =========================================================================
+ * condition variables
+ * =========================================================================
+ */
+/*ARGSUSED*/
+void
+cv_init(kcondvar_t *cv, char *name, int type, void *arg)
+{
+	VERIFY(cond_init(cv, type, NULL) == 0);
+}
+
+void
+cv_destroy(kcondvar_t *cv)
+{
+	VERIFY(cond_destroy(cv) == 0);
+}
+
+void
+cv_wait(kcondvar_t *cv, kmutex_t *mp)
+{
+	ASSERT(mutex_owner(mp) == curthread);
+	mp->m_owner = NULL;
+	int ret = cond_wait(cv, &mp->m_lock);
+	VERIFY(ret == 0 || ret == EINTR);
+	mp->m_owner = curthread;
+}
+
+clock_t
+cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
+{
+	int error;
+	timestruc_t ts;
+	clock_t delta;
+
+top:
+	delta = abstime - lbolt;
+	if (delta <= 0)
+		return (-1);
+
+	ts.tv_sec = delta / hz;
+	ts.tv_nsec = (delta % hz) * (NANOSEC / hz);
+
+	ASSERT(mutex_owner(mp) == curthread);
+	mp->m_owner = NULL;
+	error = cond_reltimedwait(cv, &mp->m_lock, &ts);
+	mp->m_owner = curthread;
+
+	if (error == ETIME)
+		return (-1);
+
+	if (error == EINTR)
+		goto top;
+
+	ASSERT(error == 0);
+
+	return (1);
+}
+
+void
+cv_signal(kcondvar_t *cv)
+{
+	VERIFY(cond_signal(cv) == 0);
+}
+
+void
+cv_broadcast(kcondvar_t *cv)
+{
+	VERIFY(cond_broadcast(cv) == 0);
+}
+
+/*
+ * =========================================================================
+ * vnode operations
+ * =========================================================================
+ */
+/*
+ * Note: for the xxxat() versions of these functions, we assume that the
+ * starting vp is always rootdir (which is true for spa_directory.c, the only
+ * ZFS consumer of these interfaces).  We assert this is true, and then emulate
+ * them by adding '/' in front of the path.
+ */
+
+/*ARGSUSED*/
+int
+vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
+{
+	int fd;
+	vnode_t *vp;
+	int old_umask;
+	char realpath[MAXPATHLEN];
+	struct stat64 st;
+
+	/*
+	 * If we're accessing a real disk from userland, we need to use
+	 * the character interface to avoid caching.  This is particularly
+	 * important if we're trying to look at a real in-kernel storage
+	 * pool from userland, e.g. via zdb, because otherwise we won't
+	 * see the changes occurring under the segmap cache.
+	 * On the other hand, the stupid character device returns zero
+	 * for its size.  So -- gag -- we open the block device to get
+	 * its size, and remember it for subsequent VOP_GETATTR().
+	 */
+	if (strncmp(path, "/dev/", 5) == 0) {
+		char *dsk;
+		fd = open64(path, O_RDONLY);
+		if (fd == -1)
+			return (errno);
+		if (fstat64(fd, &st) == -1) {
+			close(fd);
+			return (errno);
+		}
+		close(fd);
+		(void) sprintf(realpath, "%s", path);
+		dsk = strstr(path, "/dsk/");
+		if (dsk != NULL)
+			(void) sprintf(realpath + (dsk - path) + 1, "r%s",
+			    dsk + 1);
+	} else {
+		(void) sprintf(realpath, "%s", path);
+		if (!(flags & FCREAT) && stat64(realpath, &st) == -1)
+			return (errno);
+	}
+
+	if (flags & FCREAT)
+		old_umask = umask(0);
+
+	/*
+	 * The construct 'flags - FREAD' conveniently maps combinations of
+	 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
+	 */
+	fd = open64(realpath, flags - FREAD, mode);
+
+	if (flags & FCREAT)
+		(void) umask(old_umask);
+
+	if (fd == -1)
+		return (errno);
+
+	if (fstat64(fd, &st) == -1) {
+		close(fd);
+		return (errno);
+	}
+
+	(void) fcntl(fd, F_SETFD, FD_CLOEXEC);
+
+	*vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
+
+	vp->v_fd = fd;
+	vp->v_size = st.st_size;
+	vp->v_path = spa_strdup(path);
+
+	return (0);
+}
+
+/*ARGSUSED*/
+int
+vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
+    int x3, vnode_t *startvp, int fd)
+{
+	char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
+	int ret;
+
+	ASSERT(startvp == rootdir);
+	(void) sprintf(realpath, "/%s", path);
+
+	/* fd ignored for now, need if want to simulate nbmand support */
+	ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
+
+	umem_free(realpath, strlen(path) + 2);
+
+	return (ret);
+}
+
+/*ARGSUSED*/
+int
+vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
+	int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
+{
+	ssize_t iolen, split;
+
+	if (uio == UIO_READ) {
+		iolen = pread64(vp->v_fd, addr, len, offset);
+	} else {
+		/*
+		 * To simulate partial disk writes, we split writes into two
+		 * system calls so that the process can be killed in between.
+		 */
+		split = (len > 0 ? rand() % len : 0);
+		iolen = pwrite64(vp->v_fd, addr, split, offset);
+		iolen += pwrite64(vp->v_fd, (char *)addr + split,
+		    len - split, offset + split);
+	}
+
+	if (iolen == -1)
+		return (errno);
+	if (residp)
+		*residp = len - iolen;
+	else if (iolen != len)
+		return (EIO);
+	return (0);
+}
+
+void
+vn_close(vnode_t *vp)
+{
+	close(vp->v_fd);
+	spa_strfree(vp->v_path);
+	umem_free(vp, sizeof (vnode_t));
+}
+
+#ifdef ZFS_DEBUG
+
+/*
+ * =========================================================================
+ * Figure out which debugging statements to print
+ * =========================================================================
+ */
+
+static char *dprintf_string;
+static int dprintf_print_all;
+
+int
+dprintf_find_string(const char *string)
+{
+	char *tmp_str = dprintf_string;
+	int len = strlen(string);
+
+	/*
+	 * Find out if this is a string we want to print.
+	 * String format: file1.c,function_name1,file2.c,file3.c
+	 */
+
+	while (tmp_str != NULL) {
+		if (strncmp(tmp_str, string, len) == 0 &&
+		    (tmp_str[len] == ',' || tmp_str[len] == '\0'))
+			return (1);
+		tmp_str = strchr(tmp_str, ',');
+		if (tmp_str != NULL)
+			tmp_str++; /* Get rid of , */
+	}
+	return (0);
+}
+
+void
+dprintf_setup(int *argc, char **argv)
+{
+	int i, j;
+
+	/*
+	 * Debugging can be specified two ways: by setting the
+	 * environment variable ZFS_DEBUG, or by including a
+	 * "debug=..."  argument on the command line.  The command
+	 * line setting overrides the environment variable.
+	 */
+
+	for (i = 1; i < *argc; i++) {
+		int len = strlen("debug=");
+		/* First look for a command line argument */
+		if (strncmp("debug=", argv[i], len) == 0) {
+			dprintf_string = argv[i] + len;
+			/* Remove from args */
+			for (j = i; j < *argc; j++)
+				argv[j] = argv[j+1];
+			argv[j] = NULL;
+			(*argc)--;
+		}
+	}
+
+	if (dprintf_string == NULL) {
+		/* Look for ZFS_DEBUG environment variable */
+		dprintf_string = getenv("ZFS_DEBUG");
+	}
+
+	/*
+	 * Are we just turning on all debugging?
+	 */
+	if (dprintf_find_string("on"))
+		dprintf_print_all = 1;
+}
+
+/*
+ * =========================================================================
+ * debug printfs
+ * =========================================================================
+ */
+void
+__dprintf(const char *file, const char *func, int line, const char *fmt, ...)
+{
+	const char *newfile;
+	va_list adx;
+
+	/*
+	 * Get rid of annoying "../common/" prefix to filename.
+	 */
+	newfile = strrchr(file, '/');
+	if (newfile != NULL) {
+		newfile = newfile + 1; /* Get rid of leading / */
+	} else {
+		newfile = file;
+	}
+
+	if (dprintf_print_all ||
+	    dprintf_find_string(newfile) ||
+	    dprintf_find_string(func)) {
+		/* Print out just the function name if requested */
+		flockfile(stdout);
+		if (dprintf_find_string("pid"))
+			(void) printf("%d ", getpid());
+		if (dprintf_find_string("tid"))
+			(void) printf("%u ", thr_self());
+		if (dprintf_find_string("cpu"))
+			(void) printf("%u ", getcpuid());
+		if (dprintf_find_string("time"))
+			(void) printf("%llu ", gethrtime());
+		if (dprintf_find_string("long"))
+			(void) printf("%s, line %d: ", newfile, line);
+		(void) printf("%s: ", func);
+		va_start(adx, fmt);
+		(void) vprintf(fmt, adx);
+		va_end(adx);
+		funlockfile(stdout);
+	}
+}
+
+#endif /* ZFS_DEBUG */
+
+/*
+ * =========================================================================
+ * cmn_err() and panic()
+ * =========================================================================
+ */
+static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
+static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
+
+void
+vpanic(const char *fmt, va_list adx)
+{
+	(void) fprintf(stderr, "error: ");
+	(void) vfprintf(stderr, fmt, adx);
+	(void) fprintf(stderr, "\n");
+
+	abort();	/* think of it as a "user-level crash dump" */
+}
+
+void
+panic(const char *fmt, ...)
+{
+	va_list adx;
+
+	va_start(adx, fmt);
+	vpanic(fmt, adx);
+	va_end(adx);
+}
+
+void
+vcmn_err(int ce, const char *fmt, va_list adx)
+{
+	if (ce == CE_PANIC)
+		vpanic(fmt, adx);
+	if (ce != CE_NOTE) {	/* suppress noise in userland stress testing */
+		(void) fprintf(stderr, "%s", ce_prefix[ce]);
+		(void) vfprintf(stderr, fmt, adx);
+		(void) fprintf(stderr, "%s", ce_suffix[ce]);
+	}
+}
+
+/*PRINTFLIKE2*/
+void
+cmn_err(int ce, const char *fmt, ...)
+{
+	va_list adx;
+
+	va_start(adx, fmt);
+	vcmn_err(ce, fmt, adx);
+	va_end(adx);
+}
+
+/*
+ * =========================================================================
+ * kobj interfaces
+ * =========================================================================
+ */
+struct _buf *
+kobj_open_file(char *name)
+{
+	struct _buf *file;
+	vnode_t *vp;
+
+	/* set vp as the _fd field of the file */
+	if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
+	    -1) != 0)
+		return ((void *)-1UL);
+
+	file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
+	file->_fd = (intptr_t)vp;
+	return (file);
+}
+
+int
+kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
+{
+	ssize_t resid;
+
+	vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
+	    UIO_SYSSPACE, 0, 0, 0, &resid);
+
+	return (size - resid);
+}
+
+void
+kobj_close_file(struct _buf *file)
+{
+	vn_close((vnode_t *)file->_fd);
+	umem_free(file, sizeof (struct _buf));
+}
+
+int
+kobj_get_filesize(struct _buf *file, uint64_t *size)
+{
+	struct stat64 st;
+	vnode_t *vp = (vnode_t *)file->_fd;
+
+	if (fstat64(vp->v_fd, &st) == -1) {
+		vn_close(vp);
+		return (errno);
+	}
+	*size = st.st_size;
+	return (0);
+}
+
+/*
+ * =========================================================================
+ * misc routines
+ * =========================================================================
+ */
+
+void
+delay(clock_t ticks)
+{
+	poll(0, 0, ticks * (1000 / hz));
+}
+
+/*
+ * Find highest one bit set.
+ *	Returns bit number + 1 of highest bit that is set, otherwise returns 0.
+ * High order bit is 31 (or 63 in _LP64 kernel).
+ */
+int
+highbit(ulong_t i)
+{
+	register int h = 1;
+
+	if (i == 0)
+		return (0);
+#ifdef _LP64
+	if (i & 0xffffffff00000000ul) {
+		h += 32; i >>= 32;
+	}
+#endif
+	if (i & 0xffff0000) {
+		h += 16; i >>= 16;
+	}
+	if (i & 0xff00) {
+		h += 8; i >>= 8;
+	}
+	if (i & 0xf0) {
+		h += 4; i >>= 4;
+	}
+	if (i & 0xc) {
+		h += 2; i >>= 2;
+	}
+	if (i & 0x2) {
+		h += 1;
+	}
+	return (h);
+}
+
+static int random_fd = -1, urandom_fd = -1;
+
+static int
+random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
+{
+	size_t resid = len;
+	ssize_t bytes;
+
+	ASSERT(fd != -1);
+
+	while (resid != 0) {
+		bytes = read(fd, ptr, resid);
+		ASSERT3S(bytes, >=, 0);
+		ptr += bytes;
+		resid -= bytes;
+	}
+
+	return (0);
+}
+
+int
+random_get_bytes(uint8_t *ptr, size_t len)
+{
+	return (random_get_bytes_common(ptr, len, random_fd));
+}
+
+int
+random_get_pseudo_bytes(uint8_t *ptr, size_t len)
+{
+	return (random_get_bytes_common(ptr, len, urandom_fd));
+}
+
+int
+ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
+{
+	char *end;
+
+	*result = strtoul(hw_serial, &end, base);
+	if (*result == 0)
+		return (errno);
+	return (0);
+}
+
+/*
+ * =========================================================================
+ * kernel emulation setup & teardown
+ * =========================================================================
+ */
+static int
+umem_out_of_memory(void)
+{
+	char errmsg[] = "out of memory -- generating core dump\n";
+
+	write(fileno(stderr), errmsg, sizeof (errmsg));
+	abort();
+	return (0);
+}
+
+void
+kernel_init(int mode)
+{
+	umem_nofail_callback(umem_out_of_memory);
+
+	physmem = sysconf(_SC_PHYS_PAGES);
+
+	dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
+	    (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
+
+	snprintf(hw_serial, sizeof (hw_serial), "%ld", gethostid());
+
+	VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1);
+	VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1);
+
+	system_taskq_init();
+
+	spa_init(mode);
+}
+
+void
+kernel_fini(void)
+{
+	spa_fini();
+
+	close(random_fd);
+	close(urandom_fd);
+
+	random_fd = -1;
+	urandom_fd = -1;
+}
+
+int
+z_uncompress(void *dst, size_t *dstlen, const void *src, size_t srclen)
+{
+	int ret;
+	uLongf len = *dstlen;
+
+	if ((ret = uncompress(dst, &len, src, srclen)) == Z_OK)
+		*dstlen = (size_t)len;
+
+	return (ret);
+}
+
+int
+z_compress_level(void *dst, size_t *dstlen, const void *src, size_t srclen,
+    int level)
+{
+	int ret;
+	uLongf len = *dstlen;
+
+	if ((ret = compress2(dst, &len, src, srclen, level)) == Z_OK)
+		*dstlen = (size_t)len;
+
+	return (ret);
+}
+
+uid_t
+crgetuid(cred_t *cr)
+{
+	return (0);
+}
+
+gid_t
+crgetgid(cred_t *cr)
+{
+	return (0);
+}
+
+int
+crgetngroups(cred_t *cr)
+{
+	return (0);
+}
+
+gid_t *
+crgetgroups(cred_t *cr)
+{
+	return (NULL);
+}
+
+int
+zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
+{
+	return (0);
+}
+
+int
+zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
+{
+	return (0);
+}
+
+int
+zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
+{
+	return (0);
+}
+
+ksiddomain_t *
+ksid_lookupdomain(const char *dom)
+{
+	ksiddomain_t *kd;
+
+	kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
+	kd->kd_name = spa_strdup(dom);
+	return (kd);
+}
+
+void
+ksiddomain_rele(ksiddomain_t *ksid)
+{
+	spa_strfree(ksid->kd_name);
+	umem_free(ksid, sizeof (ksiddomain_t));
+}

lib/libzpool/taskq.c

@@ -0,0 +1,261 @@
+/*
+ * 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 <sys/zfs_context.h>
+
+int taskq_now;
+taskq_t *system_taskq;
+
+typedef struct task {
+	struct task	*task_next;
+	struct task	*task_prev;
+	task_func_t	*task_func;
+	void		*task_arg;
+} task_t;
+
+#define	TASKQ_ACTIVE	0x00010000
+
+struct taskq {
+	kmutex_t	tq_lock;
+	krwlock_t	tq_threadlock;
+	kcondvar_t	tq_dispatch_cv;
+	kcondvar_t	tq_wait_cv;
+	thread_t	*tq_threadlist;
+	int		tq_flags;
+	int		tq_active;
+	int		tq_nthreads;
+	int		tq_nalloc;
+	int		tq_minalloc;
+	int		tq_maxalloc;
+	task_t		*tq_freelist;
+	task_t		tq_task;
+};
+
+static task_t *
+task_alloc(taskq_t *tq, int tqflags)
+{
+	task_t *t;
+
+	if ((t = tq->tq_freelist) != NULL && tq->tq_nalloc >= tq->tq_minalloc) {
+		tq->tq_freelist = t->task_next;
+	} else {
+		mutex_exit(&tq->tq_lock);
+		if (tq->tq_nalloc >= tq->tq_maxalloc) {
+			if (!(tqflags & KM_SLEEP)) {
+				mutex_enter(&tq->tq_lock);
+				return (NULL);
+			}
+			/*
+			 * We don't want to exceed tq_maxalloc, but we can't
+			 * wait for other tasks to complete (and thus free up
+			 * task structures) without risking deadlock with
+			 * the caller.  So, we just delay for one second
+			 * to throttle the allocation rate.
+			 */
+			delay(hz);
+		}
+		t = kmem_alloc(sizeof (task_t), tqflags);
+		mutex_enter(&tq->tq_lock);
+		if (t != NULL)
+			tq->tq_nalloc++;
+	}
+	return (t);
+}
+
+static void
+task_free(taskq_t *tq, task_t *t)
+{
+	if (tq->tq_nalloc <= tq->tq_minalloc) {
+		t->task_next = tq->tq_freelist;
+		tq->tq_freelist = t;
+	} else {
+		tq->tq_nalloc--;
+		mutex_exit(&tq->tq_lock);
+		kmem_free(t, sizeof (task_t));
+		mutex_enter(&tq->tq_lock);
+	}
+}
+
+taskqid_t
+taskq_dispatch(taskq_t *tq, task_func_t func, void *arg, uint_t tqflags)
+{
+	task_t *t;
+
+	if (taskq_now) {
+		func(arg);
+		return (1);
+	}
+
+	mutex_enter(&tq->tq_lock);
+	ASSERT(tq->tq_flags & TASKQ_ACTIVE);
+	if ((t = task_alloc(tq, tqflags)) == NULL) {
+		mutex_exit(&tq->tq_lock);
+		return (0);
+	}
+	t->task_next = &tq->tq_task;
+	t->task_prev = tq->tq_task.task_prev;
+	t->task_next->task_prev = t;
+	t->task_prev->task_next = t;
+	t->task_func = func;
+	t->task_arg = arg;
+	cv_signal(&tq->tq_dispatch_cv);
+	mutex_exit(&tq->tq_lock);
+	return (1);
+}
+
+void
+taskq_wait(taskq_t *tq)
+{
+	mutex_enter(&tq->tq_lock);
+	while (tq->tq_task.task_next != &tq->tq_task || tq->tq_active != 0)
+		cv_wait(&tq->tq_wait_cv, &tq->tq_lock);
+	mutex_exit(&tq->tq_lock);
+}
+
+static void *
+taskq_thread(void *arg)
+{
+	taskq_t *tq = arg;
+	task_t *t;
+
+	mutex_enter(&tq->tq_lock);
+	while (tq->tq_flags & TASKQ_ACTIVE) {
+		if ((t = tq->tq_task.task_next) == &tq->tq_task) {
+			if (--tq->tq_active == 0)
+				cv_broadcast(&tq->tq_wait_cv);
+			cv_wait(&tq->tq_dispatch_cv, &tq->tq_lock);
+			tq->tq_active++;
+			continue;
+		}
+		t->task_prev->task_next = t->task_next;
+		t->task_next->task_prev = t->task_prev;
+		mutex_exit(&tq->tq_lock);
+
+		rw_enter(&tq->tq_threadlock, RW_READER);
+		t->task_func(t->task_arg);
+		rw_exit(&tq->tq_threadlock);
+
+		mutex_enter(&tq->tq_lock);
+		task_free(tq, t);
+	}
+	tq->tq_nthreads--;
+	cv_broadcast(&tq->tq_wait_cv);
+	mutex_exit(&tq->tq_lock);
+	return (NULL);
+}
+
+/*ARGSUSED*/
+taskq_t *
+taskq_create(const char *name, int nthreads, pri_t pri,
+	int minalloc, int maxalloc, uint_t flags)
+{
+	taskq_t *tq = kmem_zalloc(sizeof (taskq_t), KM_SLEEP);
+	int t;
+
+	rw_init(&tq->tq_threadlock, NULL, RW_DEFAULT, NULL);
+	mutex_init(&tq->tq_lock, NULL, MUTEX_DEFAULT, NULL);
+	cv_init(&tq->tq_dispatch_cv, NULL, CV_DEFAULT, NULL);
+	cv_init(&tq->tq_wait_cv, NULL, CV_DEFAULT, NULL);
+	tq->tq_flags = flags | TASKQ_ACTIVE;
+	tq->tq_active = nthreads;
+	tq->tq_nthreads = nthreads;
+	tq->tq_minalloc = minalloc;
+	tq->tq_maxalloc = maxalloc;
+	tq->tq_task.task_next = &tq->tq_task;
+	tq->tq_task.task_prev = &tq->tq_task;
+	tq->tq_threadlist = kmem_alloc(nthreads * sizeof (thread_t), KM_SLEEP);
+
+	if (flags & TASKQ_PREPOPULATE) {
+		mutex_enter(&tq->tq_lock);
+		while (minalloc-- > 0)
+			task_free(tq, task_alloc(tq, KM_SLEEP));
+		mutex_exit(&tq->tq_lock);
+	}
+
+	for (t = 0; t < nthreads; t++)
+		(void) thr_create(0, 0, taskq_thread,
+		    tq, THR_BOUND, &tq->tq_threadlist[t]);
+
+	return (tq);
+}
+
+void
+taskq_destroy(taskq_t *tq)
+{
+	int t;
+	int nthreads = tq->tq_nthreads;
+
+	taskq_wait(tq);
+
+	mutex_enter(&tq->tq_lock);
+
+	tq->tq_flags &= ~TASKQ_ACTIVE;
+	cv_broadcast(&tq->tq_dispatch_cv);
+
+	while (tq->tq_nthreads != 0)
+		cv_wait(&tq->tq_wait_cv, &tq->tq_lock);
+
+	tq->tq_minalloc = 0;
+	while (tq->tq_nalloc != 0) {
+		ASSERT(tq->tq_freelist != NULL);
+		task_free(tq, task_alloc(tq, KM_SLEEP));
+	}
+
+	mutex_exit(&tq->tq_lock);
+
+	for (t = 0; t < nthreads; t++)
+		(void) thr_join(tq->tq_threadlist[t], NULL, NULL);
+
+	kmem_free(tq->tq_threadlist, nthreads * sizeof (thread_t));
+
+	rw_destroy(&tq->tq_threadlock);
+	mutex_destroy(&tq->tq_lock);
+	cv_destroy(&tq->tq_dispatch_cv);
+	cv_destroy(&tq->tq_wait_cv);
+
+	kmem_free(tq, sizeof (taskq_t));
+}
+
+int
+taskq_member(taskq_t *tq, void *t)
+{
+	int i;
+
+	if (taskq_now)
+		return (1);
+
+	for (i = 0; i < tq->tq_nthreads; i++)
+		if (tq->tq_threadlist[i] == (thread_t)(uintptr_t)t)
+			return (1);
+
+	return (0);
+}
+
+void
+system_taskq_init(void)
+{
+	system_taskq = taskq_create("system_taskq", 64, minclsyspri, 4, 512,
+	    TASKQ_DYNAMIC | TASKQ_PREPOPULATE);
+}

lib/libzpool/util.c

@@ -0,0 +1,156 @@
+/*
+ * 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 <assert.h>
+#include <sys/zfs_context.h>
+#include <sys/avl.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/spa.h>
+#include <sys/fs/zfs.h>
+#include <sys/refcount.h>
+
+/*
+ * Routines needed by more than one client of libzpool.
+ */
+
+void
+nicenum(uint64_t num, char *buf)
+{
+	uint64_t n = num;
+	int index = 0;
+	char u;
+
+	while (n >= 1024) {
+		n = (n + (1024 / 2)) / 1024; /* Round up or down */
+		index++;
+	}
+
+	u = " KMGTPE"[index];
+
+	if (index == 0) {
+		(void) sprintf(buf, "%llu", (u_longlong_t)n);
+	} else if (n < 10 && (num & (num - 1)) != 0) {
+		(void) sprintf(buf, "%.2f%c",
+		    (double)num / (1ULL << 10 * index), u);
+	} else if (n < 100 && (num & (num - 1)) != 0) {
+		(void) sprintf(buf, "%.1f%c",
+		    (double)num / (1ULL << 10 * index), u);
+	} else {
+		(void) sprintf(buf, "%llu%c", (u_longlong_t)n, u);
+	}
+}
+
+static void
+show_vdev_stats(const char *desc, const char *ctype, nvlist_t *nv, int indent)
+{
+	vdev_stat_t *vs;
+	vdev_stat_t v0 = { 0 };
+	uint64_t sec;
+	uint64_t is_log = 0;
+	nvlist_t **child;
+	uint_t c, children;
+	char used[6], avail[6];
+	char rops[6], wops[6], rbytes[6], wbytes[6], rerr[6], werr[6], cerr[6];
+	char *prefix = "";
+
+	if (indent == 0 && desc != NULL) {
+		(void) printf("                           "
+		    " capacity   operations   bandwidth  ---- errors ----\n");
+		(void) printf("description                "
+		    "used avail  read write  read write  read write cksum\n");
+	}
+
+	if (desc != NULL) {
+		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log);
+
+		if (is_log)
+			prefix = "log ";
+
+		if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
+		    (uint64_t **)&vs, &c) != 0)
+			vs = &v0;
+
+		sec = MAX(1, vs->vs_timestamp / NANOSEC);
+
+		nicenum(vs->vs_alloc, used);
+		nicenum(vs->vs_space - vs->vs_alloc, avail);
+		nicenum(vs->vs_ops[ZIO_TYPE_READ] / sec, rops);
+		nicenum(vs->vs_ops[ZIO_TYPE_WRITE] / sec, wops);
+		nicenum(vs->vs_bytes[ZIO_TYPE_READ] / sec, rbytes);
+		nicenum(vs->vs_bytes[ZIO_TYPE_WRITE] / sec, wbytes);
+		nicenum(vs->vs_read_errors, rerr);
+		nicenum(vs->vs_write_errors, werr);
+		nicenum(vs->vs_checksum_errors, cerr);
+
+		(void) printf("%*s%s%*s%*s%*s %5s %5s %5s %5s %5s %5s %5s\n",
+		    indent, "",
+		    prefix,
+		    indent + strlen(prefix) - 25 - (vs->vs_space ? 0 : 12),
+		    desc,
+		    vs->vs_space ? 6 : 0, vs->vs_space ? used : "",
+		    vs->vs_space ? 6 : 0, vs->vs_space ? avail : "",
+		    rops, wops, rbytes, wbytes, rerr, werr, cerr);
+	}
+
+	if (nvlist_lookup_nvlist_array(nv, ctype, &child, &children) != 0)
+		return;
+
+	for (c = 0; c < children; c++) {
+		nvlist_t *cnv = child[c];
+		char *cname, *tname;
+		uint64_t np;
+		if (nvlist_lookup_string(cnv, ZPOOL_CONFIG_PATH, &cname) &&
+		    nvlist_lookup_string(cnv, ZPOOL_CONFIG_TYPE, &cname))
+			cname = "<unknown>";
+		tname = calloc(1, strlen(cname) + 2);
+		(void) strcpy(tname, cname);
+		if (nvlist_lookup_uint64(cnv, ZPOOL_CONFIG_NPARITY, &np) == 0)
+			tname[strlen(tname)] = '0' + np;
+		show_vdev_stats(tname, ctype, cnv, indent + 2);
+		free(tname);
+	}
+}
+
+void
+show_pool_stats(spa_t *spa)
+{
+	nvlist_t *config, *nvroot;
+	char *name;
+
+	VERIFY(spa_get_stats(spa_name(spa), &config, NULL, 0) == 0);
+
+	VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
+	    &nvroot) == 0);
+	VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
+	    &name) == 0);
+
+	show_vdev_stats(name, ZPOOL_CONFIG_CHILDREN, nvroot, 0);
+	show_vdev_stats(NULL, ZPOOL_CONFIG_L2CACHE, nvroot, 0);
+	show_vdev_stats(NULL, ZPOOL_CONFIG_SPARES, nvroot, 0);
+
+	nvlist_free(config);
+}