1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22/*
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2011, 2020 by Delphix. All rights reserved.
26 * Copyright (c) 2012 by Frederik Wessels. All rights reserved.
27 * Copyright (c) 2012 by Cyril Plisko. All rights reserved.
28 * Copyright (c) 2013 by Prasad Joshi (sTec). All rights reserved.
29 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
30 * Copyright (c) 2017 Datto Inc.
31 * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
32 * Copyright (c) 2017, Intel Corporation.
33 * Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
34 * Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
35 * Copyright [2021] Hewlett Packard Enterprise Development LP
36 */
37
38#include <assert.h>
39#include <ctype.h>
40#include <dirent.h>
41#include <errno.h>
42#include <fcntl.h>
43#include <getopt.h>
44#include <libgen.h>
45#include <libintl.h>
46#include <libuutil.h>
47#include <locale.h>
48#include <pthread.h>
49#include <stdio.h>
50#include <stdlib.h>
51#include <string.h>
52#include <strings.h>
53#include <time.h>
54#include <unistd.h>
55#include <pwd.h>
56#include <zone.h>
57#include <sys/wait.h>
58#include <zfs_prop.h>
59#include <sys/fs/zfs.h>
60#include <sys/stat.h>
61#include <sys/systeminfo.h>
62#include <sys/fm/fs/zfs.h>
63#include <sys/fm/util.h>
64#include <sys/fm/protocol.h>
65#include <sys/zfs_ioctl.h>
66#include <sys/mount.h>
67#include <sys/sysmacros.h>
68
69#include <math.h>
70
71#include <libzfs.h>
72#include <libzutil.h>
73
74#include "zpool_util.h"
75#include "zfs_comutil.h"
76#include "zfeature_common.h"
77
78#include "statcommon.h"
79
80libzfs_handle_t *g_zfs;
81
82static int zpool_do_create(int, char **);
83static int zpool_do_destroy(int, char **);
84
85static int zpool_do_add(int, char **);
86static int zpool_do_remove(int, char **);
87static int zpool_do_labelclear(int, char **);
88
89static int zpool_do_checkpoint(int, char **);
90
91static int zpool_do_list(int, char **);
92static int zpool_do_iostat(int, char **);
93static int zpool_do_status(int, char **);
94
95static int zpool_do_online(int, char **);
96static int zpool_do_offline(int, char **);
97static int zpool_do_clear(int, char **);
98static int zpool_do_reopen(int, char **);
99
100static int zpool_do_reguid(int, char **);
101
102static int zpool_do_attach(int, char **);
103static int zpool_do_detach(int, char **);
104static int zpool_do_replace(int, char **);
105static int zpool_do_split(int, char **);
106
107static int zpool_do_initialize(int, char **);
108static int zpool_do_scrub(int, char **);
109static int zpool_do_resilver(int, char **);
110static int zpool_do_trim(int, char **);
111
112static int zpool_do_import(int, char **);
113static int zpool_do_export(int, char **);
114
115static int zpool_do_upgrade(int, char **);
116
117static int zpool_do_history(int, char **);
118static int zpool_do_events(int, char **);
119
120static int zpool_do_get(int, char **);
121static int zpool_do_set(int, char **);
122
123static int zpool_do_sync(int, char **);
124
125static int zpool_do_version(int, char **);
126
127static int zpool_do_wait(int, char **);
128
129static zpool_compat_status_t zpool_do_load_compat(
130    const char *, boolean_t *);
131
132/*
133 * These libumem hooks provide a reasonable set of defaults for the allocator's
134 * debugging facilities.
135 */
136
137#ifdef DEBUG
138const char *
139_umem_debug_init(void)
140{
141	return ("default,verbose"); /* $UMEM_DEBUG setting */
142}
143
144const char *
145_umem_logging_init(void)
146{
147	return ("fail,contents"); /* $UMEM_LOGGING setting */
148}
149#endif
150
151typedef enum {
152	HELP_ADD,
153	HELP_ATTACH,
154	HELP_CLEAR,
155	HELP_CREATE,
156	HELP_CHECKPOINT,
157	HELP_DESTROY,
158	HELP_DETACH,
159	HELP_EXPORT,
160	HELP_HISTORY,
161	HELP_IMPORT,
162	HELP_IOSTAT,
163	HELP_LABELCLEAR,
164	HELP_LIST,
165	HELP_OFFLINE,
166	HELP_ONLINE,
167	HELP_REPLACE,
168	HELP_REMOVE,
169	HELP_INITIALIZE,
170	HELP_SCRUB,
171	HELP_RESILVER,
172	HELP_TRIM,
173	HELP_STATUS,
174	HELP_UPGRADE,
175	HELP_EVENTS,
176	HELP_GET,
177	HELP_SET,
178	HELP_SPLIT,
179	HELP_SYNC,
180	HELP_REGUID,
181	HELP_REOPEN,
182	HELP_VERSION,
183	HELP_WAIT
184} zpool_help_t;
185
186
187/*
188 * Flags for stats to display with "zpool iostats"
189 */
190enum iostat_type {
191	IOS_DEFAULT = 0,
192	IOS_LATENCY = 1,
193	IOS_QUEUES = 2,
194	IOS_L_HISTO = 3,
195	IOS_RQ_HISTO = 4,
196	IOS_COUNT,	/* always last element */
197};
198
199/* iostat_type entries as bitmasks */
200#define	IOS_DEFAULT_M	(1ULL << IOS_DEFAULT)
201#define	IOS_LATENCY_M	(1ULL << IOS_LATENCY)
202#define	IOS_QUEUES_M	(1ULL << IOS_QUEUES)
203#define	IOS_L_HISTO_M	(1ULL << IOS_L_HISTO)
204#define	IOS_RQ_HISTO_M	(1ULL << IOS_RQ_HISTO)
205
206/* Mask of all the histo bits */
207#define	IOS_ANYHISTO_M (IOS_L_HISTO_M | IOS_RQ_HISTO_M)
208
209/*
210 * Lookup table for iostat flags to nvlist names.  Basically a list
211 * of all the nvlists a flag requires.  Also specifies the order in
212 * which data gets printed in zpool iostat.
213 */
214static const char *vsx_type_to_nvlist[IOS_COUNT][13] = {
215	[IOS_L_HISTO] = {
216	    ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
217	    ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
218	    ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
219	    ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
220	    ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
221	    ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
222	    ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
223	    ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
224	    ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
225	    ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
226	    NULL},
227	[IOS_LATENCY] = {
228	    ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
229	    ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
230	    ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
231	    ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
232	    ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
233	    NULL},
234	[IOS_QUEUES] = {
235	    ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
236	    ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
237	    ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
238	    ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
239	    ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
240	    ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
241	    NULL},
242	[IOS_RQ_HISTO] = {
243	    ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO,
244	    ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO,
245	    ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO,
246	    ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO,
247	    ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO,
248	    ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO,
249	    ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO,
250	    ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO,
251	    ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO,
252	    ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO,
253	    ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO,
254	    ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO,
255	    NULL},
256};
257
258
259/*
260 * Given a cb->cb_flags with a histogram bit set, return the iostat_type.
261 * Right now, only one histo bit is ever set at one time, so we can
262 * just do a highbit64(a)
263 */
264#define	IOS_HISTO_IDX(a)	(highbit64(a & IOS_ANYHISTO_M) - 1)
265
266typedef struct zpool_command {
267	const char	*name;
268	int		(*func)(int, char **);
269	zpool_help_t	usage;
270} zpool_command_t;
271
272/*
273 * Master command table.  Each ZFS command has a name, associated function, and
274 * usage message.  The usage messages need to be internationalized, so we have
275 * to have a function to return the usage message based on a command index.
276 *
277 * These commands are organized according to how they are displayed in the usage
278 * message.  An empty command (one with a NULL name) indicates an empty line in
279 * the generic usage message.
280 */
281static zpool_command_t command_table[] = {
282	{ "version",	zpool_do_version,	HELP_VERSION		},
283	{ NULL },
284	{ "create",	zpool_do_create,	HELP_CREATE		},
285	{ "destroy",	zpool_do_destroy,	HELP_DESTROY		},
286	{ NULL },
287	{ "add",	zpool_do_add,		HELP_ADD		},
288	{ "remove",	zpool_do_remove,	HELP_REMOVE		},
289	{ NULL },
290	{ "labelclear",	zpool_do_labelclear,	HELP_LABELCLEAR		},
291	{ NULL },
292	{ "checkpoint",	zpool_do_checkpoint,	HELP_CHECKPOINT		},
293	{ NULL },
294	{ "list",	zpool_do_list,		HELP_LIST		},
295	{ "iostat",	zpool_do_iostat,	HELP_IOSTAT		},
296	{ "status",	zpool_do_status,	HELP_STATUS		},
297	{ NULL },
298	{ "online",	zpool_do_online,	HELP_ONLINE		},
299	{ "offline",	zpool_do_offline,	HELP_OFFLINE		},
300	{ "clear",	zpool_do_clear,		HELP_CLEAR		},
301	{ "reopen",	zpool_do_reopen,	HELP_REOPEN		},
302	{ NULL },
303	{ "attach",	zpool_do_attach,	HELP_ATTACH		},
304	{ "detach",	zpool_do_detach,	HELP_DETACH		},
305	{ "replace",	zpool_do_replace,	HELP_REPLACE		},
306	{ "split",	zpool_do_split,		HELP_SPLIT		},
307	{ NULL },
308	{ "initialize",	zpool_do_initialize,	HELP_INITIALIZE		},
309	{ "resilver",	zpool_do_resilver,	HELP_RESILVER		},
310	{ "scrub",	zpool_do_scrub,		HELP_SCRUB		},
311	{ "trim",	zpool_do_trim,		HELP_TRIM		},
312	{ NULL },
313	{ "import",	zpool_do_import,	HELP_IMPORT		},
314	{ "export",	zpool_do_export,	HELP_EXPORT		},
315	{ "upgrade",	zpool_do_upgrade,	HELP_UPGRADE		},
316	{ "reguid",	zpool_do_reguid,	HELP_REGUID		},
317	{ NULL },
318	{ "history",	zpool_do_history,	HELP_HISTORY		},
319	{ "events",	zpool_do_events,	HELP_EVENTS		},
320	{ NULL },
321	{ "get",	zpool_do_get,		HELP_GET		},
322	{ "set",	zpool_do_set,		HELP_SET		},
323	{ "sync",	zpool_do_sync,		HELP_SYNC		},
324	{ NULL },
325	{ "wait",	zpool_do_wait,		HELP_WAIT		},
326};
327
328#define	NCOMMAND	(ARRAY_SIZE(command_table))
329
330#define	VDEV_ALLOC_CLASS_LOGS	"logs"
331
332static zpool_command_t *current_command;
333static char history_str[HIS_MAX_RECORD_LEN];
334static boolean_t log_history = B_TRUE;
335static uint_t timestamp_fmt = NODATE;
336
337static const char *
338get_usage(zpool_help_t idx)
339{
340	switch (idx) {
341	case HELP_ADD:
342		return (gettext("\tadd [-fgLnP] [-o property=value] "
343		    "<pool> <vdev> ...\n"));
344	case HELP_ATTACH:
345		return (gettext("\tattach [-fsw] [-o property=value] "
346		    "<pool> <device> <new-device>\n"));
347	case HELP_CLEAR:
348		return (gettext("\tclear [-nF] <pool> [device]\n"));
349	case HELP_CREATE:
350		return (gettext("\tcreate [-fnd] [-o property=value] ... \n"
351		    "\t    [-O file-system-property=value] ... \n"
352		    "\t    [-m mountpoint] [-R root] <pool> <vdev> ...\n"));
353	case HELP_CHECKPOINT:
354		return (gettext("\tcheckpoint [-d [-w]] <pool> ...\n"));
355	case HELP_DESTROY:
356		return (gettext("\tdestroy [-f] <pool>\n"));
357	case HELP_DETACH:
358		return (gettext("\tdetach <pool> <device>\n"));
359	case HELP_EXPORT:
360		return (gettext("\texport [-af] <pool> ...\n"));
361	case HELP_HISTORY:
362		return (gettext("\thistory [-il] [<pool>] ...\n"));
363	case HELP_IMPORT:
364		return (gettext("\timport [-d dir] [-D]\n"
365		    "\timport [-o mntopts] [-o property=value] ... \n"
366		    "\t    [-d dir | -c cachefile] [-D] [-l] [-f] [-m] [-N] "
367		    "[-R root] [-F [-n]] -a\n"
368		    "\timport [-o mntopts] [-o property=value] ... \n"
369		    "\t    [-d dir | -c cachefile] [-D] [-l] [-f] [-m] [-N] "
370		    "[-R root] [-F [-n]]\n"
371		    "\t    [--rewind-to-checkpoint] <pool | id> [newpool]\n"));
372	case HELP_IOSTAT:
373		return (gettext("\tiostat [[[-c [script1,script2,...]"
374		    "[-lq]]|[-rw]] [-T d | u] [-ghHLpPvy]\n"
375		    "\t    [[pool ...]|[pool vdev ...]|[vdev ...]]"
376		    " [[-n] interval [count]]\n"));
377	case HELP_LABELCLEAR:
378		return (gettext("\tlabelclear [-f] <vdev>\n"));
379	case HELP_LIST:
380		return (gettext("\tlist [-gHLpPv] [-o property[,...]] "
381		    "[-T d|u] [pool] ... \n"
382		    "\t    [interval [count]]\n"));
383	case HELP_OFFLINE:
384		return (gettext("\toffline [-f] [-t] <pool> <device> ...\n"));
385	case HELP_ONLINE:
386		return (gettext("\tonline [-e] <pool> <device> ...\n"));
387	case HELP_REPLACE:
388		return (gettext("\treplace [-fsw] [-o property=value] "
389		    "<pool> <device> [new-device]\n"));
390	case HELP_REMOVE:
391		return (gettext("\tremove [-npsw] <pool> <device> ...\n"));
392	case HELP_REOPEN:
393		return (gettext("\treopen [-n] <pool>\n"));
394	case HELP_INITIALIZE:
395		return (gettext("\tinitialize [-c | -s] [-w] <pool> "
396		    "[<device> ...]\n"));
397	case HELP_SCRUB:
398		return (gettext("\tscrub [-s | -p] [-w] <pool> ...\n"));
399	case HELP_RESILVER:
400		return (gettext("\tresilver <pool> ...\n"));
401	case HELP_TRIM:
402		return (gettext("\ttrim [-dw] [-r <rate>] [-c | -s] <pool> "
403		    "[<device> ...]\n"));
404	case HELP_STATUS:
405		return (gettext("\tstatus [-c [script1,script2,...]] "
406		    "[-igLpPstvxD]  [-T d|u] [pool] ... \n"
407		    "\t    [interval [count]]\n"));
408	case HELP_UPGRADE:
409		return (gettext("\tupgrade\n"
410		    "\tupgrade -v\n"
411		    "\tupgrade [-V version] <-a | pool ...>\n"));
412	case HELP_EVENTS:
413		return (gettext("\tevents [-vHf [pool] | -c]\n"));
414	case HELP_GET:
415		return (gettext("\tget [-Hp] [-o \"all\" | field[,...]] "
416		    "<\"all\" | property[,...]> <pool> ...\n"));
417	case HELP_SET:
418		return (gettext("\tset <property=value> <pool> \n"));
419	case HELP_SPLIT:
420		return (gettext("\tsplit [-gLnPl] [-R altroot] [-o mntopts]\n"
421		    "\t    [-o property=value] <pool> <newpool> "
422		    "[<device> ...]\n"));
423	case HELP_REGUID:
424		return (gettext("\treguid <pool>\n"));
425	case HELP_SYNC:
426		return (gettext("\tsync [pool] ...\n"));
427	case HELP_VERSION:
428		return (gettext("\tversion\n"));
429	case HELP_WAIT:
430		return (gettext("\twait [-Hp] [-T d|u] [-t <activity>[,...]] "
431		    "<pool> [interval]\n"));
432	}
433
434	abort();
435	/* NOTREACHED */
436}
437
438static void
439zpool_collect_leaves(zpool_handle_t *zhp, nvlist_t *nvroot, nvlist_t *res)
440{
441	uint_t children = 0;
442	nvlist_t **child;
443	uint_t i;
444
445	(void) nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
446	    &child, &children);
447
448	if (children == 0) {
449		char *path = zpool_vdev_name(g_zfs, zhp, nvroot,
450		    VDEV_NAME_PATH);
451
452		if (strcmp(path, VDEV_TYPE_INDIRECT) != 0 &&
453		    strcmp(path, VDEV_TYPE_HOLE) != 0)
454			fnvlist_add_boolean(res, path);
455
456		free(path);
457		return;
458	}
459
460	for (i = 0; i < children; i++) {
461		zpool_collect_leaves(zhp, child[i], res);
462	}
463}
464
465/*
466 * Callback routine that will print out a pool property value.
467 */
468static int
469print_prop_cb(int prop, void *cb)
470{
471	FILE *fp = cb;
472
473	(void) fprintf(fp, "\t%-19s  ", zpool_prop_to_name(prop));
474
475	if (zpool_prop_readonly(prop))
476		(void) fprintf(fp, "  NO   ");
477	else
478		(void) fprintf(fp, " YES   ");
479
480	if (zpool_prop_values(prop) == NULL)
481		(void) fprintf(fp, "-\n");
482	else
483		(void) fprintf(fp, "%s\n", zpool_prop_values(prop));
484
485	return (ZPROP_CONT);
486}
487
488/*
489 * Display usage message.  If we're inside a command, display only the usage for
490 * that command.  Otherwise, iterate over the entire command table and display
491 * a complete usage message.
492 */
493static void
494usage(boolean_t requested)
495{
496	FILE *fp = requested ? stdout : stderr;
497
498	if (current_command == NULL) {
499		int i;
500
501		(void) fprintf(fp, gettext("usage: zpool command args ...\n"));
502		(void) fprintf(fp,
503		    gettext("where 'command' is one of the following:\n\n"));
504
505		for (i = 0; i < NCOMMAND; i++) {
506			if (command_table[i].name == NULL)
507				(void) fprintf(fp, "\n");
508			else
509				(void) fprintf(fp, "%s",
510				    get_usage(command_table[i].usage));
511		}
512	} else {
513		(void) fprintf(fp, gettext("usage:\n"));
514		(void) fprintf(fp, "%s", get_usage(current_command->usage));
515	}
516
517	if (current_command != NULL &&
518	    ((strcmp(current_command->name, "set") == 0) ||
519	    (strcmp(current_command->name, "get") == 0) ||
520	    (strcmp(current_command->name, "list") == 0))) {
521
522		(void) fprintf(fp,
523		    gettext("\nthe following properties are supported:\n"));
524
525		(void) fprintf(fp, "\n\t%-19s  %s   %s\n\n",
526		    "PROPERTY", "EDIT", "VALUES");
527
528		/* Iterate over all properties */
529		(void) zprop_iter(print_prop_cb, fp, B_FALSE, B_TRUE,
530		    ZFS_TYPE_POOL);
531
532		(void) fprintf(fp, "\t%-19s   ", "feature@...");
533		(void) fprintf(fp, "YES   disabled | enabled | active\n");
534
535		(void) fprintf(fp, gettext("\nThe feature@ properties must be "
536		    "appended with a feature name.\nSee zpool-features(7).\n"));
537	}
538
539	/*
540	 * See comments at end of main().
541	 */
542	if (getenv("ZFS_ABORT") != NULL) {
543		(void) printf("dumping core by request\n");
544		abort();
545	}
546
547	exit(requested ? 0 : 2);
548}
549
550/*
551 * zpool initialize [-c | -s] [-w] <pool> [<vdev> ...]
552 * Initialize all unused blocks in the specified vdevs, or all vdevs in the pool
553 * if none specified.
554 *
555 *	-c	Cancel. Ends active initializing.
556 *	-s	Suspend. Initializing can then be restarted with no flags.
557 *	-w	Wait. Blocks until initializing has completed.
558 */
559int
560zpool_do_initialize(int argc, char **argv)
561{
562	int c;
563	char *poolname;
564	zpool_handle_t *zhp;
565	nvlist_t *vdevs;
566	int err = 0;
567	boolean_t wait = B_FALSE;
568
569	struct option long_options[] = {
570		{"cancel",	no_argument,		NULL, 'c'},
571		{"suspend",	no_argument,		NULL, 's'},
572		{"wait",	no_argument,		NULL, 'w'},
573		{0, 0, 0, 0}
574	};
575
576	pool_initialize_func_t cmd_type = POOL_INITIALIZE_START;
577	while ((c = getopt_long(argc, argv, "csw", long_options, NULL)) != -1) {
578		switch (c) {
579		case 'c':
580			if (cmd_type != POOL_INITIALIZE_START &&
581			    cmd_type != POOL_INITIALIZE_CANCEL) {
582				(void) fprintf(stderr, gettext("-c cannot be "
583				    "combined with other options\n"));
584				usage(B_FALSE);
585			}
586			cmd_type = POOL_INITIALIZE_CANCEL;
587			break;
588		case 's':
589			if (cmd_type != POOL_INITIALIZE_START &&
590			    cmd_type != POOL_INITIALIZE_SUSPEND) {
591				(void) fprintf(stderr, gettext("-s cannot be "
592				    "combined with other options\n"));
593				usage(B_FALSE);
594			}
595			cmd_type = POOL_INITIALIZE_SUSPEND;
596			break;
597		case 'w':
598			wait = B_TRUE;
599			break;
600		case '?':
601			if (optopt != 0) {
602				(void) fprintf(stderr,
603				    gettext("invalid option '%c'\n"), optopt);
604			} else {
605				(void) fprintf(stderr,
606				    gettext("invalid option '%s'\n"),
607				    argv[optind - 1]);
608			}
609			usage(B_FALSE);
610		}
611	}
612
613	argc -= optind;
614	argv += optind;
615
616	if (argc < 1) {
617		(void) fprintf(stderr, gettext("missing pool name argument\n"));
618		usage(B_FALSE);
619		return (-1);
620	}
621
622	if (wait && (cmd_type != POOL_INITIALIZE_START)) {
623		(void) fprintf(stderr, gettext("-w cannot be used with -c or "
624		    "-s\n"));
625		usage(B_FALSE);
626	}
627
628	poolname = argv[0];
629	zhp = zpool_open(g_zfs, poolname);
630	if (zhp == NULL)
631		return (-1);
632
633	vdevs = fnvlist_alloc();
634	if (argc == 1) {
635		/* no individual leaf vdevs specified, so add them all */
636		nvlist_t *config = zpool_get_config(zhp, NULL);
637		nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
638		    ZPOOL_CONFIG_VDEV_TREE);
639		zpool_collect_leaves(zhp, nvroot, vdevs);
640	} else {
641		for (int i = 1; i < argc; i++) {
642			fnvlist_add_boolean(vdevs, argv[i]);
643		}
644	}
645
646	if (wait)
647		err = zpool_initialize_wait(zhp, cmd_type, vdevs);
648	else
649		err = zpool_initialize(zhp, cmd_type, vdevs);
650
651	fnvlist_free(vdevs);
652	zpool_close(zhp);
653
654	return (err);
655}
656
657/*
658 * print a pool vdev config for dry runs
659 */
660static void
661print_vdev_tree(zpool_handle_t *zhp, const char *name, nvlist_t *nv, int indent,
662    const char *match, int name_flags)
663{
664	nvlist_t **child;
665	uint_t c, children;
666	char *vname;
667	boolean_t printed = B_FALSE;
668
669	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
670	    &child, &children) != 0) {
671		if (name != NULL)
672			(void) printf("\t%*s%s\n", indent, "", name);
673		return;
674	}
675
676	for (c = 0; c < children; c++) {
677		uint64_t is_log = B_FALSE, is_hole = B_FALSE;
678		char *class = "";
679
680		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
681		    &is_hole);
682
683		if (is_hole == B_TRUE) {
684			continue;
685		}
686
687		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
688		    &is_log);
689		if (is_log)
690			class = VDEV_ALLOC_BIAS_LOG;
691		(void) nvlist_lookup_string(child[c],
692		    ZPOOL_CONFIG_ALLOCATION_BIAS, &class);
693		if (strcmp(match, class) != 0)
694			continue;
695
696		if (!printed && name != NULL) {
697			(void) printf("\t%*s%s\n", indent, "", name);
698			printed = B_TRUE;
699		}
700		vname = zpool_vdev_name(g_zfs, zhp, child[c], name_flags);
701		print_vdev_tree(zhp, vname, child[c], indent + 2, "",
702		    name_flags);
703		free(vname);
704	}
705}
706
707/*
708 * Print the list of l2cache devices for dry runs.
709 */
710static void
711print_cache_list(nvlist_t *nv, int indent)
712{
713	nvlist_t **child;
714	uint_t c, children;
715
716	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
717	    &child, &children) == 0 && children > 0) {
718		(void) printf("\t%*s%s\n", indent, "", "cache");
719	} else {
720		return;
721	}
722	for (c = 0; c < children; c++) {
723		char *vname;
724
725		vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
726		(void) printf("\t%*s%s\n", indent + 2, "", vname);
727		free(vname);
728	}
729}
730
731/*
732 * Print the list of spares for dry runs.
733 */
734static void
735print_spare_list(nvlist_t *nv, int indent)
736{
737	nvlist_t **child;
738	uint_t c, children;
739
740	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
741	    &child, &children) == 0 && children > 0) {
742		(void) printf("\t%*s%s\n", indent, "", "spares");
743	} else {
744		return;
745	}
746	for (c = 0; c < children; c++) {
747		char *vname;
748
749		vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
750		(void) printf("\t%*s%s\n", indent + 2, "", vname);
751		free(vname);
752	}
753}
754
755static boolean_t
756prop_list_contains_feature(nvlist_t *proplist)
757{
758	nvpair_t *nvp;
759	for (nvp = nvlist_next_nvpair(proplist, NULL); NULL != nvp;
760	    nvp = nvlist_next_nvpair(proplist, nvp)) {
761		if (zpool_prop_feature(nvpair_name(nvp)))
762			return (B_TRUE);
763	}
764	return (B_FALSE);
765}
766
767/*
768 * Add a property pair (name, string-value) into a property nvlist.
769 */
770static int
771add_prop_list(const char *propname, char *propval, nvlist_t **props,
772    boolean_t poolprop)
773{
774	zpool_prop_t prop = ZPOOL_PROP_INVAL;
775	nvlist_t *proplist;
776	const char *normnm;
777	char *strval;
778
779	if (*props == NULL &&
780	    nvlist_alloc(props, NV_UNIQUE_NAME, 0) != 0) {
781		(void) fprintf(stderr,
782		    gettext("internal error: out of memory\n"));
783		return (1);
784	}
785
786	proplist = *props;
787
788	if (poolprop) {
789		const char *vname = zpool_prop_to_name(ZPOOL_PROP_VERSION);
790		const char *cname =
791		    zpool_prop_to_name(ZPOOL_PROP_COMPATIBILITY);
792
793		if ((prop = zpool_name_to_prop(propname)) == ZPOOL_PROP_INVAL &&
794		    !zpool_prop_feature(propname)) {
795			(void) fprintf(stderr, gettext("property '%s' is "
796			    "not a valid pool property\n"), propname);
797			return (2);
798		}
799
800		/*
801		 * feature@ properties and version should not be specified
802		 * at the same time.
803		 */
804		if ((prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname) &&
805		    nvlist_exists(proplist, vname)) ||
806		    (prop == ZPOOL_PROP_VERSION &&
807		    prop_list_contains_feature(proplist))) {
808			(void) fprintf(stderr, gettext("'feature@' and "
809			    "'version' properties cannot be specified "
810			    "together\n"));
811			return (2);
812		}
813
814		/*
815		 * if version is specified, only "legacy" compatibility
816		 * may be requested
817		 */
818		if ((prop == ZPOOL_PROP_COMPATIBILITY &&
819		    strcmp(propval, ZPOOL_COMPAT_LEGACY) != 0 &&
820		    nvlist_exists(proplist, vname)) ||
821		    (prop == ZPOOL_PROP_VERSION &&
822		    nvlist_exists(proplist, cname) &&
823		    strcmp(fnvlist_lookup_string(proplist, cname),
824		    ZPOOL_COMPAT_LEGACY) != 0)) {
825			(void) fprintf(stderr, gettext("when 'version' is "
826			    "specified, the 'compatibility' feature may only "
827			    "be set to '" ZPOOL_COMPAT_LEGACY "'\n"));
828			return (2);
829		}
830
831		if (zpool_prop_feature(propname))
832			normnm = propname;
833		else
834			normnm = zpool_prop_to_name(prop);
835	} else {
836		zfs_prop_t fsprop = zfs_name_to_prop(propname);
837
838		if (zfs_prop_valid_for_type(fsprop, ZFS_TYPE_FILESYSTEM,
839		    B_FALSE)) {
840			normnm = zfs_prop_to_name(fsprop);
841		} else if (zfs_prop_user(propname) ||
842		    zfs_prop_userquota(propname)) {
843			normnm = propname;
844		} else {
845			(void) fprintf(stderr, gettext("property '%s' is "
846			    "not a valid filesystem property\n"), propname);
847			return (2);
848		}
849	}
850
851	if (nvlist_lookup_string(proplist, normnm, &strval) == 0 &&
852	    prop != ZPOOL_PROP_CACHEFILE) {
853		(void) fprintf(stderr, gettext("property '%s' "
854		    "specified multiple times\n"), propname);
855		return (2);
856	}
857
858	if (nvlist_add_string(proplist, normnm, propval) != 0) {
859		(void) fprintf(stderr, gettext("internal "
860		    "error: out of memory\n"));
861		return (1);
862	}
863
864	return (0);
865}
866
867/*
868 * Set a default property pair (name, string-value) in a property nvlist
869 */
870static int
871add_prop_list_default(const char *propname, char *propval, nvlist_t **props,
872    boolean_t poolprop)
873{
874	char *pval;
875
876	if (nvlist_lookup_string(*props, propname, &pval) == 0)
877		return (0);
878
879	return (add_prop_list(propname, propval, props, B_TRUE));
880}
881
882/*
883 * zpool add [-fgLnP] [-o property=value] <pool> <vdev> ...
884 *
885 *	-f	Force addition of devices, even if they appear in use
886 *	-g	Display guid for individual vdev name.
887 *	-L	Follow links when resolving vdev path name.
888 *	-n	Do not add the devices, but display the resulting layout if
889 *		they were to be added.
890 *	-o	Set property=value.
891 *	-P	Display full path for vdev name.
892 *
893 * Adds the given vdevs to 'pool'.  As with create, the bulk of this work is
894 * handled by make_root_vdev(), which constructs the nvlist needed to pass to
895 * libzfs.
896 */
897int
898zpool_do_add(int argc, char **argv)
899{
900	boolean_t force = B_FALSE;
901	boolean_t dryrun = B_FALSE;
902	int name_flags = 0;
903	int c;
904	nvlist_t *nvroot;
905	char *poolname;
906	int ret;
907	zpool_handle_t *zhp;
908	nvlist_t *config;
909	nvlist_t *props = NULL;
910	char *propval;
911
912	/* check options */
913	while ((c = getopt(argc, argv, "fgLno:P")) != -1) {
914		switch (c) {
915		case 'f':
916			force = B_TRUE;
917			break;
918		case 'g':
919			name_flags |= VDEV_NAME_GUID;
920			break;
921		case 'L':
922			name_flags |= VDEV_NAME_FOLLOW_LINKS;
923			break;
924		case 'n':
925			dryrun = B_TRUE;
926			break;
927		case 'o':
928			if ((propval = strchr(optarg, '=')) == NULL) {
929				(void) fprintf(stderr, gettext("missing "
930				    "'=' for -o option\n"));
931				usage(B_FALSE);
932			}
933			*propval = '\0';
934			propval++;
935
936			if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) ||
937			    (add_prop_list(optarg, propval, &props, B_TRUE)))
938				usage(B_FALSE);
939			break;
940		case 'P':
941			name_flags |= VDEV_NAME_PATH;
942			break;
943		case '?':
944			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
945			    optopt);
946			usage(B_FALSE);
947		}
948	}
949
950	argc -= optind;
951	argv += optind;
952
953	/* get pool name and check number of arguments */
954	if (argc < 1) {
955		(void) fprintf(stderr, gettext("missing pool name argument\n"));
956		usage(B_FALSE);
957	}
958	if (argc < 2) {
959		(void) fprintf(stderr, gettext("missing vdev specification\n"));
960		usage(B_FALSE);
961	}
962
963	poolname = argv[0];
964
965	argc--;
966	argv++;
967
968	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
969		return (1);
970
971	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
972		(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
973		    poolname);
974		zpool_close(zhp);
975		return (1);
976	}
977
978	/* unless manually specified use "ashift" pool property (if set) */
979	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
980		int intval;
981		zprop_source_t src;
982		char strval[ZPOOL_MAXPROPLEN];
983
984		intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
985		if (src != ZPROP_SRC_DEFAULT) {
986			(void) sprintf(strval, "%" PRId32, intval);
987			verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
988			    &props, B_TRUE) == 0);
989		}
990	}
991
992	/* pass off to make_root_vdev for processing */
993	nvroot = make_root_vdev(zhp, props, force, !force, B_FALSE, dryrun,
994	    argc, argv);
995	if (nvroot == NULL) {
996		zpool_close(zhp);
997		return (1);
998	}
999
1000	if (dryrun) {
1001		nvlist_t *poolnvroot;
1002		nvlist_t **l2child, **sparechild;
1003		uint_t l2children, sparechildren, c;
1004		char *vname;
1005		boolean_t hadcache = B_FALSE, hadspare = B_FALSE;
1006
1007		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
1008		    &poolnvroot) == 0);
1009
1010		(void) printf(gettext("would update '%s' to the following "
1011		    "configuration:\n\n"), zpool_get_name(zhp));
1012
1013		/* print original main pool and new tree */
1014		print_vdev_tree(zhp, poolname, poolnvroot, 0, "",
1015		    name_flags | VDEV_NAME_TYPE_ID);
1016		print_vdev_tree(zhp, NULL, nvroot, 0, "", name_flags);
1017
1018		/* print other classes: 'dedup', 'special', and 'log' */
1019		if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_DEDUP)) {
1020			print_vdev_tree(zhp, "dedup", poolnvroot, 0,
1021			    VDEV_ALLOC_BIAS_DEDUP, name_flags);
1022			print_vdev_tree(zhp, NULL, nvroot, 0,
1023			    VDEV_ALLOC_BIAS_DEDUP, name_flags);
1024		} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_DEDUP)) {
1025			print_vdev_tree(zhp, "dedup", nvroot, 0,
1026			    VDEV_ALLOC_BIAS_DEDUP, name_flags);
1027		}
1028
1029		if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
1030			print_vdev_tree(zhp, "special", poolnvroot, 0,
1031			    VDEV_ALLOC_BIAS_SPECIAL, name_flags);
1032			print_vdev_tree(zhp, NULL, nvroot, 0,
1033			    VDEV_ALLOC_BIAS_SPECIAL, name_flags);
1034		} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
1035			print_vdev_tree(zhp, "special", nvroot, 0,
1036			    VDEV_ALLOC_BIAS_SPECIAL, name_flags);
1037		}
1038
1039		if (num_logs(poolnvroot) > 0) {
1040			print_vdev_tree(zhp, "logs", poolnvroot, 0,
1041			    VDEV_ALLOC_BIAS_LOG, name_flags);
1042			print_vdev_tree(zhp, NULL, nvroot, 0,
1043			    VDEV_ALLOC_BIAS_LOG, name_flags);
1044		} else if (num_logs(nvroot) > 0) {
1045			print_vdev_tree(zhp, "logs", nvroot, 0,
1046			    VDEV_ALLOC_BIAS_LOG, name_flags);
1047		}
1048
1049		/* Do the same for the caches */
1050		if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_L2CACHE,
1051		    &l2child, &l2children) == 0 && l2children) {
1052			hadcache = B_TRUE;
1053			(void) printf(gettext("\tcache\n"));
1054			for (c = 0; c < l2children; c++) {
1055				vname = zpool_vdev_name(g_zfs, NULL,
1056				    l2child[c], name_flags);
1057				(void) printf("\t  %s\n", vname);
1058				free(vname);
1059			}
1060		}
1061		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1062		    &l2child, &l2children) == 0 && l2children) {
1063			if (!hadcache)
1064				(void) printf(gettext("\tcache\n"));
1065			for (c = 0; c < l2children; c++) {
1066				vname = zpool_vdev_name(g_zfs, NULL,
1067				    l2child[c], name_flags);
1068				(void) printf("\t  %s\n", vname);
1069				free(vname);
1070			}
1071		}
1072		/* And finally the spares */
1073		if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_SPARES,
1074		    &sparechild, &sparechildren) == 0 && sparechildren > 0) {
1075			hadspare = B_TRUE;
1076			(void) printf(gettext("\tspares\n"));
1077			for (c = 0; c < sparechildren; c++) {
1078				vname = zpool_vdev_name(g_zfs, NULL,
1079				    sparechild[c], name_flags);
1080				(void) printf("\t  %s\n", vname);
1081				free(vname);
1082			}
1083		}
1084		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1085		    &sparechild, &sparechildren) == 0 && sparechildren > 0) {
1086			if (!hadspare)
1087				(void) printf(gettext("\tspares\n"));
1088			for (c = 0; c < sparechildren; c++) {
1089				vname = zpool_vdev_name(g_zfs, NULL,
1090				    sparechild[c], name_flags);
1091				(void) printf("\t  %s\n", vname);
1092				free(vname);
1093			}
1094		}
1095
1096		ret = 0;
1097	} else {
1098		ret = (zpool_add(zhp, nvroot) != 0);
1099	}
1100
1101	nvlist_free(props);
1102	nvlist_free(nvroot);
1103	zpool_close(zhp);
1104
1105	return (ret);
1106}
1107
1108/*
1109 * zpool remove [-npsw] <pool> <vdev> ...
1110 *
1111 * Removes the given vdev from the pool.
1112 */
1113int
1114zpool_do_remove(int argc, char **argv)
1115{
1116	char *poolname;
1117	int i, ret = 0;
1118	zpool_handle_t *zhp = NULL;
1119	boolean_t stop = B_FALSE;
1120	int c;
1121	boolean_t noop = B_FALSE;
1122	boolean_t parsable = B_FALSE;
1123	boolean_t wait = B_FALSE;
1124
1125	/* check options */
1126	while ((c = getopt(argc, argv, "npsw")) != -1) {
1127		switch (c) {
1128		case 'n':
1129			noop = B_TRUE;
1130			break;
1131		case 'p':
1132			parsable = B_TRUE;
1133			break;
1134		case 's':
1135			stop = B_TRUE;
1136			break;
1137		case 'w':
1138			wait = B_TRUE;
1139			break;
1140		case '?':
1141			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1142			    optopt);
1143			usage(B_FALSE);
1144		}
1145	}
1146
1147	argc -= optind;
1148	argv += optind;
1149
1150	/* get pool name and check number of arguments */
1151	if (argc < 1) {
1152		(void) fprintf(stderr, gettext("missing pool name argument\n"));
1153		usage(B_FALSE);
1154	}
1155
1156	poolname = argv[0];
1157
1158	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
1159		return (1);
1160
1161	if (stop && noop) {
1162		(void) fprintf(stderr, gettext("stop request ignored\n"));
1163		return (0);
1164	}
1165
1166	if (stop) {
1167		if (argc > 1) {
1168			(void) fprintf(stderr, gettext("too many arguments\n"));
1169			usage(B_FALSE);
1170		}
1171		if (zpool_vdev_remove_cancel(zhp) != 0)
1172			ret = 1;
1173		if (wait) {
1174			(void) fprintf(stderr, gettext("invalid option "
1175			    "combination: -w cannot be used with -s\n"));
1176			usage(B_FALSE);
1177		}
1178	} else {
1179		if (argc < 2) {
1180			(void) fprintf(stderr, gettext("missing device\n"));
1181			usage(B_FALSE);
1182		}
1183
1184		for (i = 1; i < argc; i++) {
1185			if (noop) {
1186				uint64_t size;
1187
1188				if (zpool_vdev_indirect_size(zhp, argv[i],
1189				    &size) != 0) {
1190					ret = 1;
1191					break;
1192				}
1193				if (parsable) {
1194					(void) printf("%s %llu\n",
1195					    argv[i], (unsigned long long)size);
1196				} else {
1197					char valstr[32];
1198					zfs_nicenum(size, valstr,
1199					    sizeof (valstr));
1200					(void) printf("Memory that will be "
1201					    "used after removing %s: %s\n",
1202					    argv[i], valstr);
1203				}
1204			} else {
1205				if (zpool_vdev_remove(zhp, argv[i]) != 0)
1206					ret = 1;
1207			}
1208		}
1209
1210		if (ret == 0 && wait)
1211			ret = zpool_wait(zhp, ZPOOL_WAIT_REMOVE);
1212	}
1213	zpool_close(zhp);
1214
1215	return (ret);
1216}
1217
1218/*
1219 * zpool labelclear [-f] <vdev>
1220 *
1221 *	-f	Force clearing the label for the vdevs which are members of
1222 *		the exported or foreign pools.
1223 *
1224 * Verifies that the vdev is not active and zeros out the label information
1225 * on the device.
1226 */
1227int
1228zpool_do_labelclear(int argc, char **argv)
1229{
1230	char vdev[MAXPATHLEN];
1231	char *name = NULL;
1232	struct stat st;
1233	int c, fd = -1, ret = 0;
1234	nvlist_t *config;
1235	pool_state_t state;
1236	boolean_t inuse = B_FALSE;
1237	boolean_t force = B_FALSE;
1238
1239	/* check options */
1240	while ((c = getopt(argc, argv, "f")) != -1) {
1241		switch (c) {
1242		case 'f':
1243			force = B_TRUE;
1244			break;
1245		default:
1246			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1247			    optopt);
1248			usage(B_FALSE);
1249		}
1250	}
1251
1252	argc -= optind;
1253	argv += optind;
1254
1255	/* get vdev name */
1256	if (argc < 1) {
1257		(void) fprintf(stderr, gettext("missing vdev name\n"));
1258		usage(B_FALSE);
1259	}
1260	if (argc > 1) {
1261		(void) fprintf(stderr, gettext("too many arguments\n"));
1262		usage(B_FALSE);
1263	}
1264
1265	/*
1266	 * Check if we were given absolute path and use it as is.
1267	 * Otherwise if the provided vdev name doesn't point to a file,
1268	 * try prepending expected disk paths and partition numbers.
1269	 */
1270	(void) strlcpy(vdev, argv[0], sizeof (vdev));
1271	if (vdev[0] != '/' && stat(vdev, &st) != 0) {
1272		int error;
1273
1274		error = zfs_resolve_shortname(argv[0], vdev, MAXPATHLEN);
1275		if (error == 0 && zfs_dev_is_whole_disk(vdev)) {
1276			if (zfs_append_partition(vdev, MAXPATHLEN) == -1)
1277				error = ENOENT;
1278		}
1279
1280		if (error || (stat(vdev, &st) != 0)) {
1281			(void) fprintf(stderr, gettext(
1282			    "failed to find device %s, try specifying absolute "
1283			    "path instead\n"), argv[0]);
1284			return (1);
1285		}
1286	}
1287
1288	if ((fd = open(vdev, O_RDWR)) < 0) {
1289		(void) fprintf(stderr, gettext("failed to open %s: %s\n"),
1290		    vdev, strerror(errno));
1291		return (1);
1292	}
1293
1294	/*
1295	 * Flush all dirty pages for the block device.  This should not be
1296	 * fatal when the device does not support BLKFLSBUF as would be the
1297	 * case for a file vdev.
1298	 */
1299	if ((zfs_dev_flush(fd) != 0) && (errno != ENOTTY))
1300		(void) fprintf(stderr, gettext("failed to invalidate "
1301		    "cache for %s: %s\n"), vdev, strerror(errno));
1302
1303	if (zpool_read_label(fd, &config, NULL) != 0) {
1304		(void) fprintf(stderr,
1305		    gettext("failed to read label from %s\n"), vdev);
1306		ret = 1;
1307		goto errout;
1308	}
1309	nvlist_free(config);
1310
1311	ret = zpool_in_use(g_zfs, fd, &state, &name, &inuse);
1312	if (ret != 0) {
1313		(void) fprintf(stderr,
1314		    gettext("failed to check state for %s\n"), vdev);
1315		ret = 1;
1316		goto errout;
1317	}
1318
1319	if (!inuse)
1320		goto wipe_label;
1321
1322	switch (state) {
1323	default:
1324	case POOL_STATE_ACTIVE:
1325	case POOL_STATE_SPARE:
1326	case POOL_STATE_L2CACHE:
1327		(void) fprintf(stderr, gettext(
1328		    "%s is a member (%s) of pool \"%s\"\n"),
1329		    vdev, zpool_pool_state_to_name(state), name);
1330		ret = 1;
1331		goto errout;
1332
1333	case POOL_STATE_EXPORTED:
1334		if (force)
1335			break;
1336		(void) fprintf(stderr, gettext(
1337		    "use '-f' to override the following error:\n"
1338		    "%s is a member of exported pool \"%s\"\n"),
1339		    vdev, name);
1340		ret = 1;
1341		goto errout;
1342
1343	case POOL_STATE_POTENTIALLY_ACTIVE:
1344		if (force)
1345			break;
1346		(void) fprintf(stderr, gettext(
1347		    "use '-f' to override the following error:\n"
1348		    "%s is a member of potentially active pool \"%s\"\n"),
1349		    vdev, name);
1350		ret = 1;
1351		goto errout;
1352
1353	case POOL_STATE_DESTROYED:
1354		/* inuse should never be set for a destroyed pool */
1355		assert(0);
1356		break;
1357	}
1358
1359wipe_label:
1360	ret = zpool_clear_label(fd);
1361	if (ret != 0) {
1362		(void) fprintf(stderr,
1363		    gettext("failed to clear label for %s\n"), vdev);
1364	}
1365
1366errout:
1367	free(name);
1368	(void) close(fd);
1369
1370	return (ret);
1371}
1372
1373/*
1374 * zpool create [-fnd] [-o property=value] ...
1375 *		[-O file-system-property=value] ...
1376 *		[-R root] [-m mountpoint] <pool> <dev> ...
1377 *
1378 *	-f	Force creation, even if devices appear in use
1379 *	-n	Do not create the pool, but display the resulting layout if it
1380 *		were to be created.
1381 *      -R	Create a pool under an alternate root
1382 *      -m	Set default mountpoint for the root dataset.  By default it's
1383 *		'/<pool>'
1384 *	-o	Set property=value.
1385 *	-o	Set feature@feature=enabled|disabled.
1386 *	-d	Don't automatically enable all supported pool features
1387 *		(individual features can be enabled with -o).
1388 *	-O	Set fsproperty=value in the pool's root file system
1389 *
1390 * Creates the named pool according to the given vdev specification.  The
1391 * bulk of the vdev processing is done in make_root_vdev() in zpool_vdev.c.
1392 * Once we get the nvlist back from make_root_vdev(), we either print out the
1393 * contents (if '-n' was specified), or pass it to libzfs to do the creation.
1394 */
1395int
1396zpool_do_create(int argc, char **argv)
1397{
1398	boolean_t force = B_FALSE;
1399	boolean_t dryrun = B_FALSE;
1400	boolean_t enable_pool_features = B_TRUE;
1401
1402	int c;
1403	nvlist_t *nvroot = NULL;
1404	char *poolname;
1405	char *tname = NULL;
1406	int ret = 1;
1407	char *altroot = NULL;
1408	char *compat = NULL;
1409	char *mountpoint = NULL;
1410	nvlist_t *fsprops = NULL;
1411	nvlist_t *props = NULL;
1412	char *propval;
1413
1414	/* check options */
1415	while ((c = getopt(argc, argv, ":fndR:m:o:O:t:")) != -1) {
1416		switch (c) {
1417		case 'f':
1418			force = B_TRUE;
1419			break;
1420		case 'n':
1421			dryrun = B_TRUE;
1422			break;
1423		case 'd':
1424			enable_pool_features = B_FALSE;
1425			break;
1426		case 'R':
1427			altroot = optarg;
1428			if (add_prop_list(zpool_prop_to_name(
1429			    ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
1430				goto errout;
1431			if (add_prop_list_default(zpool_prop_to_name(
1432			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
1433				goto errout;
1434			break;
1435		case 'm':
1436			/* Equivalent to -O mountpoint=optarg */
1437			mountpoint = optarg;
1438			break;
1439		case 'o':
1440			if ((propval = strchr(optarg, '=')) == NULL) {
1441				(void) fprintf(stderr, gettext("missing "
1442				    "'=' for -o option\n"));
1443				goto errout;
1444			}
1445			*propval = '\0';
1446			propval++;
1447
1448			if (add_prop_list(optarg, propval, &props, B_TRUE))
1449				goto errout;
1450
1451			/*
1452			 * If the user is creating a pool that doesn't support
1453			 * feature flags, don't enable any features.
1454			 */
1455			if (zpool_name_to_prop(optarg) == ZPOOL_PROP_VERSION) {
1456				char *end;
1457				u_longlong_t ver;
1458
1459				ver = strtoull(propval, &end, 10);
1460				if (*end == '\0' &&
1461				    ver < SPA_VERSION_FEATURES) {
1462					enable_pool_features = B_FALSE;
1463				}
1464			}
1465			if (zpool_name_to_prop(optarg) == ZPOOL_PROP_ALTROOT)
1466				altroot = propval;
1467			if (zpool_name_to_prop(optarg) ==
1468			    ZPOOL_PROP_COMPATIBILITY)
1469				compat = propval;
1470			break;
1471		case 'O':
1472			if ((propval = strchr(optarg, '=')) == NULL) {
1473				(void) fprintf(stderr, gettext("missing "
1474				    "'=' for -O option\n"));
1475				goto errout;
1476			}
1477			*propval = '\0';
1478			propval++;
1479
1480			/*
1481			 * Mountpoints are checked and then added later.
1482			 * Uniquely among properties, they can be specified
1483			 * more than once, to avoid conflict with -m.
1484			 */
1485			if (0 == strcmp(optarg,
1486			    zfs_prop_to_name(ZFS_PROP_MOUNTPOINT))) {
1487				mountpoint = propval;
1488			} else if (add_prop_list(optarg, propval, &fsprops,
1489			    B_FALSE)) {
1490				goto errout;
1491			}
1492			break;
1493		case 't':
1494			/*
1495			 * Sanity check temporary pool name.
1496			 */
1497			if (strchr(optarg, '/') != NULL) {
1498				(void) fprintf(stderr, gettext("cannot create "
1499				    "'%s': invalid character '/' in temporary "
1500				    "name\n"), optarg);
1501				(void) fprintf(stderr, gettext("use 'zfs "
1502				    "create' to create a dataset\n"));
1503				goto errout;
1504			}
1505
1506			if (add_prop_list(zpool_prop_to_name(
1507			    ZPOOL_PROP_TNAME), optarg, &props, B_TRUE))
1508				goto errout;
1509			if (add_prop_list_default(zpool_prop_to_name(
1510			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
1511				goto errout;
1512			tname = optarg;
1513			break;
1514		case ':':
1515			(void) fprintf(stderr, gettext("missing argument for "
1516			    "'%c' option\n"), optopt);
1517			goto badusage;
1518		case '?':
1519			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1520			    optopt);
1521			goto badusage;
1522		}
1523	}
1524
1525	argc -= optind;
1526	argv += optind;
1527
1528	/* get pool name and check number of arguments */
1529	if (argc < 1) {
1530		(void) fprintf(stderr, gettext("missing pool name argument\n"));
1531		goto badusage;
1532	}
1533	if (argc < 2) {
1534		(void) fprintf(stderr, gettext("missing vdev specification\n"));
1535		goto badusage;
1536	}
1537
1538	poolname = argv[0];
1539
1540	/*
1541	 * As a special case, check for use of '/' in the name, and direct the
1542	 * user to use 'zfs create' instead.
1543	 */
1544	if (strchr(poolname, '/') != NULL) {
1545		(void) fprintf(stderr, gettext("cannot create '%s': invalid "
1546		    "character '/' in pool name\n"), poolname);
1547		(void) fprintf(stderr, gettext("use 'zfs create' to "
1548		    "create a dataset\n"));
1549		goto errout;
1550	}
1551
1552	/* pass off to make_root_vdev for bulk processing */
1553	nvroot = make_root_vdev(NULL, props, force, !force, B_FALSE, dryrun,
1554	    argc - 1, argv + 1);
1555	if (nvroot == NULL)
1556		goto errout;
1557
1558	/* make_root_vdev() allows 0 toplevel children if there are spares */
1559	if (!zfs_allocatable_devs(nvroot)) {
1560		(void) fprintf(stderr, gettext("invalid vdev "
1561		    "specification: at least one toplevel vdev must be "
1562		    "specified\n"));
1563		goto errout;
1564	}
1565
1566	if (altroot != NULL && altroot[0] != '/') {
1567		(void) fprintf(stderr, gettext("invalid alternate root '%s': "
1568		    "must be an absolute path\n"), altroot);
1569		goto errout;
1570	}
1571
1572	/*
1573	 * Check the validity of the mountpoint and direct the user to use the
1574	 * '-m' mountpoint option if it looks like its in use.
1575	 */
1576	if (mountpoint == NULL ||
1577	    (strcmp(mountpoint, ZFS_MOUNTPOINT_LEGACY) != 0 &&
1578	    strcmp(mountpoint, ZFS_MOUNTPOINT_NONE) != 0)) {
1579		char buf[MAXPATHLEN];
1580		DIR *dirp;
1581
1582		if (mountpoint && mountpoint[0] != '/') {
1583			(void) fprintf(stderr, gettext("invalid mountpoint "
1584			    "'%s': must be an absolute path, 'legacy', or "
1585			    "'none'\n"), mountpoint);
1586			goto errout;
1587		}
1588
1589		if (mountpoint == NULL) {
1590			if (altroot != NULL)
1591				(void) snprintf(buf, sizeof (buf), "%s/%s",
1592				    altroot, poolname);
1593			else
1594				(void) snprintf(buf, sizeof (buf), "/%s",
1595				    poolname);
1596		} else {
1597			if (altroot != NULL)
1598				(void) snprintf(buf, sizeof (buf), "%s%s",
1599				    altroot, mountpoint);
1600			else
1601				(void) snprintf(buf, sizeof (buf), "%s",
1602				    mountpoint);
1603		}
1604
1605		if ((dirp = opendir(buf)) == NULL && errno != ENOENT) {
1606			(void) fprintf(stderr, gettext("mountpoint '%s' : "
1607			    "%s\n"), buf, strerror(errno));
1608			(void) fprintf(stderr, gettext("use '-m' "
1609			    "option to provide a different default\n"));
1610			goto errout;
1611		} else if (dirp) {
1612			int count = 0;
1613
1614			while (count < 3 && readdir(dirp) != NULL)
1615				count++;
1616			(void) closedir(dirp);
1617
1618			if (count > 2) {
1619				(void) fprintf(stderr, gettext("mountpoint "
1620				    "'%s' exists and is not empty\n"), buf);
1621				(void) fprintf(stderr, gettext("use '-m' "
1622				    "option to provide a "
1623				    "different default\n"));
1624				goto errout;
1625			}
1626		}
1627	}
1628
1629	/*
1630	 * Now that the mountpoint's validity has been checked, ensure that
1631	 * the property is set appropriately prior to creating the pool.
1632	 */
1633	if (mountpoint != NULL) {
1634		ret = add_prop_list(zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
1635		    mountpoint, &fsprops, B_FALSE);
1636		if (ret != 0)
1637			goto errout;
1638	}
1639
1640	ret = 1;
1641	if (dryrun) {
1642		/*
1643		 * For a dry run invocation, print out a basic message and run
1644		 * through all the vdevs in the list and print out in an
1645		 * appropriate hierarchy.
1646		 */
1647		(void) printf(gettext("would create '%s' with the "
1648		    "following layout:\n\n"), poolname);
1649
1650		print_vdev_tree(NULL, poolname, nvroot, 0, "", 0);
1651		print_vdev_tree(NULL, "dedup", nvroot, 0,
1652		    VDEV_ALLOC_BIAS_DEDUP, 0);
1653		print_vdev_tree(NULL, "special", nvroot, 0,
1654		    VDEV_ALLOC_BIAS_SPECIAL, 0);
1655		print_vdev_tree(NULL, "logs", nvroot, 0,
1656		    VDEV_ALLOC_BIAS_LOG, 0);
1657		print_cache_list(nvroot, 0);
1658		print_spare_list(nvroot, 0);
1659
1660		ret = 0;
1661	} else {
1662		/*
1663		 * Load in feature set.
1664		 * Note: if compatibility property not given, we'll have
1665		 * NULL, which means 'all features'.
1666		 */
1667		boolean_t requested_features[SPA_FEATURES];
1668		if (zpool_do_load_compat(compat, requested_features) !=
1669		    ZPOOL_COMPATIBILITY_OK)
1670			goto errout;
1671
1672		/*
1673		 * props contains list of features to enable.
1674		 * For each feature:
1675		 *  - remove it if feature@name=disabled
1676		 *  - leave it there if feature@name=enabled
1677		 *  - add it if:
1678		 *    - enable_pool_features (ie: no '-d' or '-o version')
1679		 *    - it's supported by the kernel module
1680		 *    - it's in the requested feature set
1681		 *  - warn if it's enabled but not in compat
1682		 */
1683		for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
1684			char propname[MAXPATHLEN];
1685			char *propval;
1686			zfeature_info_t *feat = &spa_feature_table[i];
1687
1688			(void) snprintf(propname, sizeof (propname),
1689			    "feature@%s", feat->fi_uname);
1690
1691			if (!nvlist_lookup_string(props, propname, &propval)) {
1692				if (strcmp(propval, ZFS_FEATURE_DISABLED) == 0)
1693					(void) nvlist_remove_all(props,
1694					    propname);
1695				if (strcmp(propval,
1696				    ZFS_FEATURE_ENABLED) == 0 &&
1697				    !requested_features[i])
1698					(void) fprintf(stderr, gettext(
1699					    "Warning: feature \"%s\" enabled "
1700					    "but is not in specified "
1701					    "'compatibility' feature set.\n"),
1702					    feat->fi_uname);
1703			} else if (
1704			    enable_pool_features &&
1705			    feat->fi_zfs_mod_supported &&
1706			    requested_features[i]) {
1707				ret = add_prop_list(propname,
1708				    ZFS_FEATURE_ENABLED, &props, B_TRUE);
1709				if (ret != 0)
1710					goto errout;
1711			}
1712		}
1713
1714		ret = 1;
1715		if (zpool_create(g_zfs, poolname,
1716		    nvroot, props, fsprops) == 0) {
1717			zfs_handle_t *pool = zfs_open(g_zfs,
1718			    tname ? tname : poolname, ZFS_TYPE_FILESYSTEM);
1719			if (pool != NULL) {
1720				if (zfs_mount(pool, NULL, 0) == 0) {
1721					ret = zfs_shareall(pool);
1722					zfs_commit_all_shares();
1723				}
1724				zfs_close(pool);
1725			}
1726		} else if (libzfs_errno(g_zfs) == EZFS_INVALIDNAME) {
1727			(void) fprintf(stderr, gettext("pool name may have "
1728			    "been omitted\n"));
1729		}
1730	}
1731
1732errout:
1733	nvlist_free(nvroot);
1734	nvlist_free(fsprops);
1735	nvlist_free(props);
1736	return (ret);
1737badusage:
1738	nvlist_free(fsprops);
1739	nvlist_free(props);
1740	usage(B_FALSE);
1741	return (2);
1742}
1743
1744/*
1745 * zpool destroy <pool>
1746 *
1747 * 	-f	Forcefully unmount any datasets
1748 *
1749 * Destroy the given pool.  Automatically unmounts any datasets in the pool.
1750 */
1751int
1752zpool_do_destroy(int argc, char **argv)
1753{
1754	boolean_t force = B_FALSE;
1755	int c;
1756	char *pool;
1757	zpool_handle_t *zhp;
1758	int ret;
1759
1760	/* check options */
1761	while ((c = getopt(argc, argv, "f")) != -1) {
1762		switch (c) {
1763		case 'f':
1764			force = B_TRUE;
1765			break;
1766		case '?':
1767			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1768			    optopt);
1769			usage(B_FALSE);
1770		}
1771	}
1772
1773	argc -= optind;
1774	argv += optind;
1775
1776	/* check arguments */
1777	if (argc < 1) {
1778		(void) fprintf(stderr, gettext("missing pool argument\n"));
1779		usage(B_FALSE);
1780	}
1781	if (argc > 1) {
1782		(void) fprintf(stderr, gettext("too many arguments\n"));
1783		usage(B_FALSE);
1784	}
1785
1786	pool = argv[0];
1787
1788	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
1789		/*
1790		 * As a special case, check for use of '/' in the name, and
1791		 * direct the user to use 'zfs destroy' instead.
1792		 */
1793		if (strchr(pool, '/') != NULL)
1794			(void) fprintf(stderr, gettext("use 'zfs destroy' to "
1795			    "destroy a dataset\n"));
1796		return (1);
1797	}
1798
1799	if (zpool_disable_datasets(zhp, force) != 0) {
1800		(void) fprintf(stderr, gettext("could not destroy '%s': "
1801		    "could not unmount datasets\n"), zpool_get_name(zhp));
1802		zpool_close(zhp);
1803		return (1);
1804	}
1805
1806	/* The history must be logged as part of the export */
1807	log_history = B_FALSE;
1808
1809	ret = (zpool_destroy(zhp, history_str) != 0);
1810
1811	zpool_close(zhp);
1812
1813	return (ret);
1814}
1815
1816typedef struct export_cbdata {
1817	boolean_t force;
1818	boolean_t hardforce;
1819} export_cbdata_t;
1820
1821/*
1822 * Export one pool
1823 */
1824static int
1825zpool_export_one(zpool_handle_t *zhp, void *data)
1826{
1827	export_cbdata_t *cb = data;
1828
1829	if (zpool_disable_datasets(zhp, cb->force) != 0)
1830		return (1);
1831
1832	/* The history must be logged as part of the export */
1833	log_history = B_FALSE;
1834
1835	if (cb->hardforce) {
1836		if (zpool_export_force(zhp, history_str) != 0)
1837			return (1);
1838	} else if (zpool_export(zhp, cb->force, history_str) != 0) {
1839		return (1);
1840	}
1841
1842	return (0);
1843}
1844
1845/*
1846 * zpool export [-f] <pool> ...
1847 *
1848 *	-a	Export all pools
1849 *	-f	Forcefully unmount datasets
1850 *
1851 * Export the given pools.  By default, the command will attempt to cleanly
1852 * unmount any active datasets within the pool.  If the '-f' flag is specified,
1853 * then the datasets will be forcefully unmounted.
1854 */
1855int
1856zpool_do_export(int argc, char **argv)
1857{
1858	export_cbdata_t cb;
1859	boolean_t do_all = B_FALSE;
1860	boolean_t force = B_FALSE;
1861	boolean_t hardforce = B_FALSE;
1862	int c, ret;
1863
1864	/* check options */
1865	while ((c = getopt(argc, argv, "afF")) != -1) {
1866		switch (c) {
1867		case 'a':
1868			do_all = B_TRUE;
1869			break;
1870		case 'f':
1871			force = B_TRUE;
1872			break;
1873		case 'F':
1874			hardforce = B_TRUE;
1875			break;
1876		case '?':
1877			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1878			    optopt);
1879			usage(B_FALSE);
1880		}
1881	}
1882
1883	cb.force = force;
1884	cb.hardforce = hardforce;
1885	argc -= optind;
1886	argv += optind;
1887
1888	if (do_all) {
1889		if (argc != 0) {
1890			(void) fprintf(stderr, gettext("too many arguments\n"));
1891			usage(B_FALSE);
1892		}
1893
1894		return (for_each_pool(argc, argv, B_TRUE, NULL,
1895		    B_FALSE, zpool_export_one, &cb));
1896	}
1897
1898	/* check arguments */
1899	if (argc < 1) {
1900		(void) fprintf(stderr, gettext("missing pool argument\n"));
1901		usage(B_FALSE);
1902	}
1903
1904	ret = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE, zpool_export_one,
1905	    &cb);
1906
1907	return (ret);
1908}
1909
1910/*
1911 * Given a vdev configuration, determine the maximum width needed for the device
1912 * name column.
1913 */
1914static int
1915max_width(zpool_handle_t *zhp, nvlist_t *nv, int depth, int max,
1916    int name_flags)
1917{
1918	char *name;
1919	nvlist_t **child;
1920	uint_t c, children;
1921	int ret;
1922
1923	name = zpool_vdev_name(g_zfs, zhp, nv, name_flags);
1924	if (strlen(name) + depth > max)
1925		max = strlen(name) + depth;
1926
1927	free(name);
1928
1929	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1930	    &child, &children) == 0) {
1931		for (c = 0; c < children; c++)
1932			if ((ret = max_width(zhp, child[c], depth + 2,
1933			    max, name_flags)) > max)
1934				max = ret;
1935	}
1936
1937	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1938	    &child, &children) == 0) {
1939		for (c = 0; c < children; c++)
1940			if ((ret = max_width(zhp, child[c], depth + 2,
1941			    max, name_flags)) > max)
1942				max = ret;
1943	}
1944
1945	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1946	    &child, &children) == 0) {
1947		for (c = 0; c < children; c++)
1948			if ((ret = max_width(zhp, child[c], depth + 2,
1949			    max, name_flags)) > max)
1950				max = ret;
1951	}
1952
1953	return (max);
1954}
1955
1956typedef struct spare_cbdata {
1957	uint64_t	cb_guid;
1958	zpool_handle_t	*cb_zhp;
1959} spare_cbdata_t;
1960
1961static boolean_t
1962find_vdev(nvlist_t *nv, uint64_t search)
1963{
1964	uint64_t guid;
1965	nvlist_t **child;
1966	uint_t c, children;
1967
1968	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
1969	    search == guid)
1970		return (B_TRUE);
1971
1972	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1973	    &child, &children) == 0) {
1974		for (c = 0; c < children; c++)
1975			if (find_vdev(child[c], search))
1976				return (B_TRUE);
1977	}
1978
1979	return (B_FALSE);
1980}
1981
1982static int
1983find_spare(zpool_handle_t *zhp, void *data)
1984{
1985	spare_cbdata_t *cbp = data;
1986	nvlist_t *config, *nvroot;
1987
1988	config = zpool_get_config(zhp, NULL);
1989	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
1990	    &nvroot) == 0);
1991
1992	if (find_vdev(nvroot, cbp->cb_guid)) {
1993		cbp->cb_zhp = zhp;
1994		return (1);
1995	}
1996
1997	zpool_close(zhp);
1998	return (0);
1999}
2000
2001typedef struct status_cbdata {
2002	int		cb_count;
2003	int		cb_name_flags;
2004	int		cb_namewidth;
2005	boolean_t	cb_allpools;
2006	boolean_t	cb_verbose;
2007	boolean_t	cb_literal;
2008	boolean_t	cb_explain;
2009	boolean_t	cb_first;
2010	boolean_t	cb_dedup_stats;
2011	boolean_t	cb_print_status;
2012	boolean_t	cb_print_slow_ios;
2013	boolean_t	cb_print_vdev_init;
2014	boolean_t	cb_print_vdev_trim;
2015	vdev_cmd_data_list_t	*vcdl;
2016} status_cbdata_t;
2017
2018/* Return 1 if string is NULL, empty, or whitespace; return 0 otherwise. */
2019static int
2020is_blank_str(char *str)
2021{
2022	while (str != NULL && *str != '\0') {
2023		if (!isblank(*str))
2024			return (0);
2025		str++;
2026	}
2027	return (1);
2028}
2029
2030/* Print command output lines for specific vdev in a specific pool */
2031static void
2032zpool_print_cmd(vdev_cmd_data_list_t *vcdl, const char *pool, char *path)
2033{
2034	vdev_cmd_data_t *data;
2035	int i, j;
2036	char *val;
2037
2038	for (i = 0; i < vcdl->count; i++) {
2039		if ((strcmp(vcdl->data[i].path, path) != 0) ||
2040		    (strcmp(vcdl->data[i].pool, pool) != 0)) {
2041			/* Not the vdev we're looking for */
2042			continue;
2043		}
2044
2045		data = &vcdl->data[i];
2046		/* Print out all the output values for this vdev */
2047		for (j = 0; j < vcdl->uniq_cols_cnt; j++) {
2048			val = NULL;
2049			/* Does this vdev have values for this column? */
2050			for (int k = 0; k < data->cols_cnt; k++) {
2051				if (strcmp(data->cols[k],
2052				    vcdl->uniq_cols[j]) == 0) {
2053					/* yes it does, record the value */
2054					val = data->lines[k];
2055					break;
2056				}
2057			}
2058			/*
2059			 * Mark empty values with dashes to make output
2060			 * awk-able.
2061			 */
2062			if (val == NULL || is_blank_str(val))
2063				val = "-";
2064
2065			printf("%*s", vcdl->uniq_cols_width[j], val);
2066			if (j < vcdl->uniq_cols_cnt - 1)
2067				printf("  ");
2068		}
2069
2070		/* Print out any values that aren't in a column at the end */
2071		for (j = data->cols_cnt; j < data->lines_cnt; j++) {
2072			/* Did we have any columns?  If so print a spacer. */
2073			if (vcdl->uniq_cols_cnt > 0)
2074				printf("  ");
2075
2076			val = data->lines[j];
2077			printf("%s", val ? val : "");
2078		}
2079		break;
2080	}
2081}
2082
2083/*
2084 * Print vdev initialization status for leaves
2085 */
2086static void
2087print_status_initialize(vdev_stat_t *vs, boolean_t verbose)
2088{
2089	if (verbose) {
2090		if ((vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE ||
2091		    vs->vs_initialize_state == VDEV_INITIALIZE_SUSPENDED ||
2092		    vs->vs_initialize_state == VDEV_INITIALIZE_COMPLETE) &&
2093		    !vs->vs_scan_removing) {
2094			char zbuf[1024];
2095			char tbuf[256];
2096			struct tm zaction_ts;
2097
2098			time_t t = vs->vs_initialize_action_time;
2099			int initialize_pct = 100;
2100			if (vs->vs_initialize_state !=
2101			    VDEV_INITIALIZE_COMPLETE) {
2102				initialize_pct = (vs->vs_initialize_bytes_done *
2103				    100 / (vs->vs_initialize_bytes_est + 1));
2104			}
2105
2106			(void) localtime_r(&t, &zaction_ts);
2107			(void) strftime(tbuf, sizeof (tbuf), "%c", &zaction_ts);
2108
2109			switch (vs->vs_initialize_state) {
2110			case VDEV_INITIALIZE_SUSPENDED:
2111				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2112				    gettext("suspended, started at"), tbuf);
2113				break;
2114			case VDEV_INITIALIZE_ACTIVE:
2115				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2116				    gettext("started at"), tbuf);
2117				break;
2118			case VDEV_INITIALIZE_COMPLETE:
2119				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2120				    gettext("completed at"), tbuf);
2121				break;
2122			}
2123
2124			(void) printf(gettext("  (%d%% initialized%s)"),
2125			    initialize_pct, zbuf);
2126		} else {
2127			(void) printf(gettext("  (uninitialized)"));
2128		}
2129	} else if (vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE) {
2130		(void) printf(gettext("  (initializing)"));
2131	}
2132}
2133
2134/*
2135 * Print vdev TRIM status for leaves
2136 */
2137static void
2138print_status_trim(vdev_stat_t *vs, boolean_t verbose)
2139{
2140	if (verbose) {
2141		if ((vs->vs_trim_state == VDEV_TRIM_ACTIVE ||
2142		    vs->vs_trim_state == VDEV_TRIM_SUSPENDED ||
2143		    vs->vs_trim_state == VDEV_TRIM_COMPLETE) &&
2144		    !vs->vs_scan_removing) {
2145			char zbuf[1024];
2146			char tbuf[256];
2147			struct tm zaction_ts;
2148
2149			time_t t = vs->vs_trim_action_time;
2150			int trim_pct = 100;
2151			if (vs->vs_trim_state != VDEV_TRIM_COMPLETE) {
2152				trim_pct = (vs->vs_trim_bytes_done *
2153				    100 / (vs->vs_trim_bytes_est + 1));
2154			}
2155
2156			(void) localtime_r(&t, &zaction_ts);
2157			(void) strftime(tbuf, sizeof (tbuf), "%c", &zaction_ts);
2158
2159			switch (vs->vs_trim_state) {
2160			case VDEV_TRIM_SUSPENDED:
2161				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2162				    gettext("suspended, started at"), tbuf);
2163				break;
2164			case VDEV_TRIM_ACTIVE:
2165				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2166				    gettext("started at"), tbuf);
2167				break;
2168			case VDEV_TRIM_COMPLETE:
2169				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2170				    gettext("completed at"), tbuf);
2171				break;
2172			}
2173
2174			(void) printf(gettext("  (%d%% trimmed%s)"),
2175			    trim_pct, zbuf);
2176		} else if (vs->vs_trim_notsup) {
2177			(void) printf(gettext("  (trim unsupported)"));
2178		} else {
2179			(void) printf(gettext("  (untrimmed)"));
2180		}
2181	} else if (vs->vs_trim_state == VDEV_TRIM_ACTIVE) {
2182		(void) printf(gettext("  (trimming)"));
2183	}
2184}
2185
2186/*
2187 * Return the color associated with a health string.  This includes returning
2188 * NULL for no color change.
2189 */
2190static char *
2191health_str_to_color(const char *health)
2192{
2193	if (strcmp(health, gettext("FAULTED")) == 0 ||
2194	    strcmp(health, gettext("SUSPENDED")) == 0 ||
2195	    strcmp(health, gettext("UNAVAIL")) == 0) {
2196		return (ANSI_RED);
2197	}
2198
2199	if (strcmp(health, gettext("OFFLINE")) == 0 ||
2200	    strcmp(health, gettext("DEGRADED")) == 0 ||
2201	    strcmp(health, gettext("REMOVED")) == 0) {
2202		return (ANSI_YELLOW);
2203	}
2204
2205	return (NULL);
2206}
2207
2208/*
2209 * Print out configuration state as requested by status_callback.
2210 */
2211static void
2212print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
2213    nvlist_t *nv, int depth, boolean_t isspare, vdev_rebuild_stat_t *vrs)
2214{
2215	nvlist_t **child, *root;
2216	uint_t c, i, vsc, children;
2217	pool_scan_stat_t *ps = NULL;
2218	vdev_stat_t *vs;
2219	char rbuf[6], wbuf[6], cbuf[6];
2220	char *vname;
2221	uint64_t notpresent;
2222	spare_cbdata_t spare_cb;
2223	const char *state;
2224	char *type;
2225	char *path = NULL;
2226	char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL;
2227
2228	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2229	    &child, &children) != 0)
2230		children = 0;
2231
2232	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
2233	    (uint64_t **)&vs, &vsc) == 0);
2234
2235	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
2236
2237	if (strcmp(type, VDEV_TYPE_INDIRECT) == 0)
2238		return;
2239
2240	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
2241
2242	if (isspare) {
2243		/*
2244		 * For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
2245		 * online drives.
2246		 */
2247		if (vs->vs_aux == VDEV_AUX_SPARED)
2248			state = gettext("INUSE");
2249		else if (vs->vs_state == VDEV_STATE_HEALTHY)
2250			state = gettext("AVAIL");
2251	}
2252
2253	printf_color(health_str_to_color(state),
2254	    "\t%*s%-*s  %-8s", depth, "", cb->cb_namewidth - depth,
2255	    name, state);
2256
2257	if (!isspare) {
2258		if (vs->vs_read_errors)
2259			rcolor = ANSI_RED;
2260
2261		if (vs->vs_write_errors)
2262			wcolor = ANSI_RED;
2263
2264		if (vs->vs_checksum_errors)
2265			ccolor = ANSI_RED;
2266
2267		if (cb->cb_literal) {
2268			printf(" ");
2269			printf_color(rcolor, "%5llu",
2270			    (u_longlong_t)vs->vs_read_errors);
2271			printf(" ");
2272			printf_color(wcolor, "%5llu",
2273			    (u_longlong_t)vs->vs_write_errors);
2274			printf(" ");
2275			printf_color(ccolor, "%5llu",
2276			    (u_longlong_t)vs->vs_checksum_errors);
2277		} else {
2278			zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
2279			zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
2280			zfs_nicenum(vs->vs_checksum_errors, cbuf,
2281			    sizeof (cbuf));
2282			printf(" ");
2283			printf_color(rcolor, "%5s", rbuf);
2284			printf(" ");
2285			printf_color(wcolor, "%5s", wbuf);
2286			printf(" ");
2287			printf_color(ccolor, "%5s", cbuf);
2288		}
2289		if (cb->cb_print_slow_ios) {
2290			if (children == 0)  {
2291				/* Only leafs vdevs have slow IOs */
2292				zfs_nicenum(vs->vs_slow_ios, rbuf,
2293				    sizeof (rbuf));
2294			} else {
2295				snprintf(rbuf, sizeof (rbuf), "-");
2296			}
2297
2298			if (cb->cb_literal)
2299				printf(" %5llu", (u_longlong_t)vs->vs_slow_ios);
2300			else
2301				printf(" %5s", rbuf);
2302		}
2303	}
2304
2305	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
2306	    &notpresent) == 0) {
2307		verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
2308		(void) printf("  %s %s", gettext("was"), path);
2309	} else if (vs->vs_aux != 0) {
2310		(void) printf("  ");
2311		color_start(ANSI_RED);
2312		switch (vs->vs_aux) {
2313		case VDEV_AUX_OPEN_FAILED:
2314			(void) printf(gettext("cannot open"));
2315			break;
2316
2317		case VDEV_AUX_BAD_GUID_SUM:
2318			(void) printf(gettext("missing device"));
2319			break;
2320
2321		case VDEV_AUX_NO_REPLICAS:
2322			(void) printf(gettext("insufficient replicas"));
2323			break;
2324
2325		case VDEV_AUX_VERSION_NEWER:
2326			(void) printf(gettext("newer version"));
2327			break;
2328
2329		case VDEV_AUX_UNSUP_FEAT:
2330			(void) printf(gettext("unsupported feature(s)"));
2331			break;
2332
2333		case VDEV_AUX_ASHIFT_TOO_BIG:
2334			(void) printf(gettext("unsupported minimum blocksize"));
2335			break;
2336
2337		case VDEV_AUX_SPARED:
2338			verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2339			    &spare_cb.cb_guid) == 0);
2340			if (zpool_iter(g_zfs, find_spare, &spare_cb) == 1) {
2341				if (strcmp(zpool_get_name(spare_cb.cb_zhp),
2342				    zpool_get_name(zhp)) == 0)
2343					(void) printf(gettext("currently in "
2344					    "use"));
2345				else
2346					(void) printf(gettext("in use by "
2347					    "pool '%s'"),
2348					    zpool_get_name(spare_cb.cb_zhp));
2349				zpool_close(spare_cb.cb_zhp);
2350			} else {
2351				(void) printf(gettext("currently in use"));
2352			}
2353			break;
2354
2355		case VDEV_AUX_ERR_EXCEEDED:
2356			(void) printf(gettext("too many errors"));
2357			break;
2358
2359		case VDEV_AUX_IO_FAILURE:
2360			(void) printf(gettext("experienced I/O failures"));
2361			break;
2362
2363		case VDEV_AUX_BAD_LOG:
2364			(void) printf(gettext("bad intent log"));
2365			break;
2366
2367		case VDEV_AUX_EXTERNAL:
2368			(void) printf(gettext("external device fault"));
2369			break;
2370
2371		case VDEV_AUX_SPLIT_POOL:
2372			(void) printf(gettext("split into new pool"));
2373			break;
2374
2375		case VDEV_AUX_ACTIVE:
2376			(void) printf(gettext("currently in use"));
2377			break;
2378
2379		case VDEV_AUX_CHILDREN_OFFLINE:
2380			(void) printf(gettext("all children offline"));
2381			break;
2382
2383		case VDEV_AUX_BAD_LABEL:
2384			(void) printf(gettext("invalid label"));
2385			break;
2386
2387		default:
2388			(void) printf(gettext("corrupted data"));
2389			break;
2390		}
2391		color_end();
2392	} else if (children == 0 && !isspare &&
2393	    getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") == NULL &&
2394	    VDEV_STAT_VALID(vs_physical_ashift, vsc) &&
2395	    vs->vs_configured_ashift < vs->vs_physical_ashift) {
2396		(void) printf(
2397		    gettext("  block size: %dB configured, %dB native"),
2398		    1 << vs->vs_configured_ashift, 1 << vs->vs_physical_ashift);
2399	}
2400
2401	/* The root vdev has the scrub/resilver stats */
2402	root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2403	    ZPOOL_CONFIG_VDEV_TREE);
2404	(void) nvlist_lookup_uint64_array(root, ZPOOL_CONFIG_SCAN_STATS,
2405	    (uint64_t **)&ps, &c);
2406
2407	if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0) {
2408		if (vs->vs_scan_processed != 0) {
2409			(void) printf(gettext("  (%s)"),
2410			    (ps->pss_func == POOL_SCAN_RESILVER) ?
2411			    "resilvering" : "repairing");
2412		} else if (vs->vs_resilver_deferred) {
2413			(void) printf(gettext("  (awaiting resilver)"));
2414		}
2415	}
2416
2417	/* The top-level vdevs have the rebuild stats */
2418	if (vrs != NULL && vrs->vrs_state == VDEV_REBUILD_ACTIVE &&
2419	    children == 0) {
2420		if (vs->vs_rebuild_processed != 0) {
2421			(void) printf(gettext("  (resilvering)"));
2422		}
2423	}
2424
2425	if (cb->vcdl != NULL) {
2426		if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
2427			printf("  ");
2428			zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
2429		}
2430	}
2431
2432	/* Display vdev initialization and trim status for leaves. */
2433	if (children == 0) {
2434		print_status_initialize(vs, cb->cb_print_vdev_init);
2435		print_status_trim(vs, cb->cb_print_vdev_trim);
2436	}
2437
2438	(void) printf("\n");
2439
2440	for (c = 0; c < children; c++) {
2441		uint64_t islog = B_FALSE, ishole = B_FALSE;
2442
2443		/* Don't print logs or holes here */
2444		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2445		    &islog);
2446		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2447		    &ishole);
2448		if (islog || ishole)
2449			continue;
2450		/* Only print normal classes here */
2451		if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
2452			continue;
2453
2454		/* Provide vdev_rebuild_stats to children if available */
2455		if (vrs == NULL) {
2456			(void) nvlist_lookup_uint64_array(nv,
2457			    ZPOOL_CONFIG_REBUILD_STATS,
2458			    (uint64_t **)&vrs, &i);
2459		}
2460
2461		vname = zpool_vdev_name(g_zfs, zhp, child[c],
2462		    cb->cb_name_flags | VDEV_NAME_TYPE_ID);
2463		print_status_config(zhp, cb, vname, child[c], depth + 2,
2464		    isspare, vrs);
2465		free(vname);
2466	}
2467}
2468
2469/*
2470 * Print the configuration of an exported pool.  Iterate over all vdevs in the
2471 * pool, printing out the name and status for each one.
2472 */
2473static void
2474print_import_config(status_cbdata_t *cb, const char *name, nvlist_t *nv,
2475    int depth)
2476{
2477	nvlist_t **child;
2478	uint_t c, children;
2479	vdev_stat_t *vs;
2480	char *type, *vname;
2481
2482	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
2483	if (strcmp(type, VDEV_TYPE_MISSING) == 0 ||
2484	    strcmp(type, VDEV_TYPE_HOLE) == 0)
2485		return;
2486
2487	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
2488	    (uint64_t **)&vs, &c) == 0);
2489
2490	(void) printf("\t%*s%-*s", depth, "", cb->cb_namewidth - depth, name);
2491	(void) printf("  %s", zpool_state_to_name(vs->vs_state, vs->vs_aux));
2492
2493	if (vs->vs_aux != 0) {
2494		(void) printf("  ");
2495
2496		switch (vs->vs_aux) {
2497		case VDEV_AUX_OPEN_FAILED:
2498			(void) printf(gettext("cannot open"));
2499			break;
2500
2501		case VDEV_AUX_BAD_GUID_SUM:
2502			(void) printf(gettext("missing device"));
2503			break;
2504
2505		case VDEV_AUX_NO_REPLICAS:
2506			(void) printf(gettext("insufficient replicas"));
2507			break;
2508
2509		case VDEV_AUX_VERSION_NEWER:
2510			(void) printf(gettext("newer version"));
2511			break;
2512
2513		case VDEV_AUX_UNSUP_FEAT:
2514			(void) printf(gettext("unsupported feature(s)"));
2515			break;
2516
2517		case VDEV_AUX_ERR_EXCEEDED:
2518			(void) printf(gettext("too many errors"));
2519			break;
2520
2521		case VDEV_AUX_ACTIVE:
2522			(void) printf(gettext("currently in use"));
2523			break;
2524
2525		case VDEV_AUX_CHILDREN_OFFLINE:
2526			(void) printf(gettext("all children offline"));
2527			break;
2528
2529		case VDEV_AUX_BAD_LABEL:
2530			(void) printf(gettext("invalid label"));
2531			break;
2532
2533		default:
2534			(void) printf(gettext("corrupted data"));
2535			break;
2536		}
2537	}
2538	(void) printf("\n");
2539
2540	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2541	    &child, &children) != 0)
2542		return;
2543
2544	for (c = 0; c < children; c++) {
2545		uint64_t is_log = B_FALSE;
2546
2547		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2548		    &is_log);
2549		if (is_log)
2550			continue;
2551		if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
2552			continue;
2553
2554		vname = zpool_vdev_name(g_zfs, NULL, child[c],
2555		    cb->cb_name_flags | VDEV_NAME_TYPE_ID);
2556		print_import_config(cb, vname, child[c], depth + 2);
2557		free(vname);
2558	}
2559
2560	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2561	    &child, &children) == 0) {
2562		(void) printf(gettext("\tcache\n"));
2563		for (c = 0; c < children; c++) {
2564			vname = zpool_vdev_name(g_zfs, NULL, child[c],
2565			    cb->cb_name_flags);
2566			(void) printf("\t  %s\n", vname);
2567			free(vname);
2568		}
2569	}
2570
2571	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2572	    &child, &children) == 0) {
2573		(void) printf(gettext("\tspares\n"));
2574		for (c = 0; c < children; c++) {
2575			vname = zpool_vdev_name(g_zfs, NULL, child[c],
2576			    cb->cb_name_flags);
2577			(void) printf("\t  %s\n", vname);
2578			free(vname);
2579		}
2580	}
2581}
2582
2583/*
2584 * Print specialized class vdevs.
2585 *
2586 * These are recorded as top level vdevs in the main pool child array
2587 * but with "is_log" set to 1 or an "alloc_bias" string. We use either
2588 * print_status_config() or print_import_config() to print the top level
2589 * class vdevs then any of their children (eg mirrored slogs) are printed
2590 * recursively - which works because only the top level vdev is marked.
2591 */
2592static void
2593print_class_vdevs(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t *nv,
2594    const char *class)
2595{
2596	uint_t c, children;
2597	nvlist_t **child;
2598	boolean_t printed = B_FALSE;
2599
2600	assert(zhp != NULL || !cb->cb_verbose);
2601
2602	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child,
2603	    &children) != 0)
2604		return;
2605
2606	for (c = 0; c < children; c++) {
2607		uint64_t is_log = B_FALSE;
2608		char *bias = NULL;
2609		char *type = NULL;
2610
2611		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2612		    &is_log);
2613
2614		if (is_log) {
2615			bias = VDEV_ALLOC_CLASS_LOGS;
2616		} else {
2617			(void) nvlist_lookup_string(child[c],
2618			    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
2619			(void) nvlist_lookup_string(child[c],
2620			    ZPOOL_CONFIG_TYPE, &type);
2621		}
2622
2623		if (bias == NULL || strcmp(bias, class) != 0)
2624			continue;
2625		if (!is_log && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
2626			continue;
2627
2628		if (!printed) {
2629			(void) printf("\t%s\t\n", gettext(class));
2630			printed = B_TRUE;
2631		}
2632
2633		char *name = zpool_vdev_name(g_zfs, zhp, child[c],
2634		    cb->cb_name_flags | VDEV_NAME_TYPE_ID);
2635		if (cb->cb_print_status)
2636			print_status_config(zhp, cb, name, child[c], 2,
2637			    B_FALSE, NULL);
2638		else
2639			print_import_config(cb, name, child[c], 2);
2640		free(name);
2641	}
2642}
2643
2644/*
2645 * Display the status for the given pool.
2646 */
2647static int
2648show_import(nvlist_t *config, boolean_t report_error)
2649{
2650	uint64_t pool_state;
2651	vdev_stat_t *vs;
2652	char *name;
2653	uint64_t guid;
2654	uint64_t hostid = 0;
2655	char *msgid;
2656	char *hostname = "unknown";
2657	nvlist_t *nvroot, *nvinfo;
2658	zpool_status_t reason;
2659	zpool_errata_t errata;
2660	const char *health;
2661	uint_t vsc;
2662	char *comment;
2663	status_cbdata_t cb = { 0 };
2664
2665	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
2666	    &name) == 0);
2667	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
2668	    &guid) == 0);
2669	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
2670	    &pool_state) == 0);
2671	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2672	    &nvroot) == 0);
2673
2674	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
2675	    (uint64_t **)&vs, &vsc) == 0);
2676	health = zpool_state_to_name(vs->vs_state, vs->vs_aux);
2677
2678	reason = zpool_import_status(config, &msgid, &errata);
2679
2680	/*
2681	 * If we're importing using a cachefile, then we won't report any
2682	 * errors unless we are in the scan phase of the import.
2683	 */
2684	if (reason != ZPOOL_STATUS_OK && !report_error)
2685		return (reason);
2686
2687	(void) printf(gettext("   pool: %s\n"), name);
2688	(void) printf(gettext("     id: %llu\n"), (u_longlong_t)guid);
2689	(void) printf(gettext("  state: %s"), health);
2690	if (pool_state == POOL_STATE_DESTROYED)
2691		(void) printf(gettext(" (DESTROYED)"));
2692	(void) printf("\n");
2693
2694	switch (reason) {
2695	case ZPOOL_STATUS_MISSING_DEV_R:
2696	case ZPOOL_STATUS_MISSING_DEV_NR:
2697	case ZPOOL_STATUS_BAD_GUID_SUM:
2698		printf_color(ANSI_BOLD, gettext("status: "));
2699		printf_color(ANSI_YELLOW, gettext("One or more devices are "
2700		    "missing from the system.\n"));
2701		break;
2702
2703	case ZPOOL_STATUS_CORRUPT_LABEL_R:
2704	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
2705		printf_color(ANSI_BOLD, gettext("status: "));
2706		printf_color(ANSI_YELLOW, gettext("One or more devices contains"
2707		    " corrupted data.\n"));
2708		break;
2709
2710	case ZPOOL_STATUS_CORRUPT_DATA:
2711		(void) printf(
2712		    gettext(" status: The pool data is corrupted.\n"));
2713		break;
2714
2715	case ZPOOL_STATUS_OFFLINE_DEV:
2716		printf_color(ANSI_BOLD, gettext("status: "));
2717		printf_color(ANSI_YELLOW, gettext("One or more devices "
2718		    "are offlined.\n"));
2719		break;
2720
2721	case ZPOOL_STATUS_CORRUPT_POOL:
2722		printf_color(ANSI_BOLD, gettext("status: "));
2723		printf_color(ANSI_YELLOW, gettext("The pool metadata is "
2724		    "corrupted.\n"));
2725		break;
2726
2727	case ZPOOL_STATUS_VERSION_OLDER:
2728		printf_color(ANSI_BOLD, gettext("status: "));
2729		printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
2730		    "a legacy on-disk version.\n"));
2731		break;
2732
2733	case ZPOOL_STATUS_VERSION_NEWER:
2734		printf_color(ANSI_BOLD, gettext("status: "));
2735		printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
2736		    "an incompatible version.\n"));
2737		break;
2738
2739	case ZPOOL_STATUS_FEAT_DISABLED:
2740		printf_color(ANSI_BOLD, gettext("status: "));
2741		printf_color(ANSI_YELLOW, gettext("Some supported "
2742		    "features are not enabled on the pool.\n\t"
2743		    "(Note that they may be intentionally disabled "
2744		    "if the\n\t'compatibility' property is set.)\n"));
2745		break;
2746
2747	case ZPOOL_STATUS_COMPATIBILITY_ERR:
2748		printf_color(ANSI_BOLD, gettext("status: "));
2749		printf_color(ANSI_YELLOW, gettext("Error reading or parsing "
2750		    "the file(s) indicated by the 'compatibility'\n"
2751		    "property.\n"));
2752		break;
2753
2754	case ZPOOL_STATUS_INCOMPATIBLE_FEAT:
2755		printf_color(ANSI_BOLD, gettext("status: "));
2756		printf_color(ANSI_YELLOW, gettext("One or more features "
2757		    "are enabled on the pool despite not being\n"
2758		    "requested by the 'compatibility' property.\n"));
2759		break;
2760
2761	case ZPOOL_STATUS_UNSUP_FEAT_READ:
2762		printf_color(ANSI_BOLD, gettext("status: "));
2763		printf_color(ANSI_YELLOW, gettext("The pool uses the following "
2764		    "feature(s) not supported on this system:\n"));
2765		color_start(ANSI_YELLOW);
2766		zpool_print_unsup_feat(config);
2767		color_end();
2768		break;
2769
2770	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
2771		printf_color(ANSI_BOLD, gettext("status: "));
2772		printf_color(ANSI_YELLOW, gettext("The pool can only be "
2773		    "accessed in read-only mode on this system. It\n\tcannot be"
2774		    " accessed in read-write mode because it uses the "
2775		    "following\n\tfeature(s) not supported on this system:\n"));
2776		color_start(ANSI_YELLOW);
2777		zpool_print_unsup_feat(config);
2778		color_end();
2779		break;
2780
2781	case ZPOOL_STATUS_HOSTID_ACTIVE:
2782		printf_color(ANSI_BOLD, gettext("status: "));
2783		printf_color(ANSI_YELLOW, gettext("The pool is currently "
2784		    "imported by another system.\n"));
2785		break;
2786
2787	case ZPOOL_STATUS_HOSTID_REQUIRED:
2788		printf_color(ANSI_BOLD, gettext("status: "));
2789		printf_color(ANSI_YELLOW, gettext("The pool has the "
2790		    "multihost property on.  It cannot\n\tbe safely imported "
2791		    "when the system hostid is not set.\n"));
2792		break;
2793
2794	case ZPOOL_STATUS_HOSTID_MISMATCH:
2795		printf_color(ANSI_BOLD, gettext("status: "));
2796		printf_color(ANSI_YELLOW, gettext("The pool was last accessed "
2797		    "by another system.\n"));
2798		break;
2799
2800	case ZPOOL_STATUS_FAULTED_DEV_R:
2801	case ZPOOL_STATUS_FAULTED_DEV_NR:
2802		printf_color(ANSI_BOLD, gettext("status: "));
2803		printf_color(ANSI_YELLOW, gettext("One or more devices are "
2804		    "faulted.\n"));
2805		break;
2806
2807	case ZPOOL_STATUS_BAD_LOG:
2808		printf_color(ANSI_BOLD, gettext("status: "));
2809		printf_color(ANSI_YELLOW, gettext("An intent log record cannot "
2810		    "be read.\n"));
2811		break;
2812
2813	case ZPOOL_STATUS_RESILVERING:
2814	case ZPOOL_STATUS_REBUILDING:
2815		printf_color(ANSI_BOLD, gettext("status: "));
2816		printf_color(ANSI_YELLOW, gettext("One or more devices were "
2817		    "being resilvered.\n"));
2818		break;
2819
2820	case ZPOOL_STATUS_ERRATA:
2821		printf_color(ANSI_BOLD, gettext("status: "));
2822		printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
2823		    errata);
2824		break;
2825
2826	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
2827		printf_color(ANSI_BOLD, gettext("status: "));
2828		printf_color(ANSI_YELLOW, gettext("One or more devices are "
2829		    "configured to use a non-native block size.\n"
2830		    "\tExpect reduced performance.\n"));
2831		break;
2832
2833	default:
2834		/*
2835		 * No other status can be seen when importing pools.
2836		 */
2837		assert(reason == ZPOOL_STATUS_OK);
2838	}
2839
2840	/*
2841	 * Print out an action according to the overall state of the pool.
2842	 */
2843	if (vs->vs_state == VDEV_STATE_HEALTHY) {
2844		if (reason == ZPOOL_STATUS_VERSION_OLDER ||
2845		    reason == ZPOOL_STATUS_FEAT_DISABLED) {
2846			(void) printf(gettext(" action: The pool can be "
2847			    "imported using its name or numeric identifier, "
2848			    "though\n\tsome features will not be available "
2849			    "without an explicit 'zpool upgrade'.\n"));
2850		} else if (reason == ZPOOL_STATUS_COMPATIBILITY_ERR) {
2851			(void) printf(gettext(" action: The pool can be "
2852			    "imported using its name or numeric\n\tidentifier, "
2853			    "though the file(s) indicated by its "
2854			    "'compatibility'\n\tproperty cannot be parsed at "
2855			    "this time.\n"));
2856		} else if (reason == ZPOOL_STATUS_HOSTID_MISMATCH) {
2857			(void) printf(gettext(" action: The pool can be "
2858			    "imported using its name or numeric "
2859			    "identifier and\n\tthe '-f' flag.\n"));
2860		} else if (reason == ZPOOL_STATUS_ERRATA) {
2861			switch (errata) {
2862			case ZPOOL_ERRATA_NONE:
2863				break;
2864
2865			case ZPOOL_ERRATA_ZOL_2094_SCRUB:
2866				(void) printf(gettext(" action: The pool can "
2867				    "be imported using its name or numeric "
2868				    "identifier,\n\thowever there is a compat"
2869				    "ibility issue which should be corrected"
2870				    "\n\tby running 'zpool scrub'\n"));
2871				break;
2872
2873			case ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY:
2874				(void) printf(gettext(" action: The pool can"
2875				    "not be imported with this version of ZFS "
2876				    "due to\n\tan active asynchronous destroy. "
2877				    "Revert to an earlier version\n\tand "
2878				    "allow the destroy to complete before "
2879				    "updating.\n"));
2880				break;
2881
2882			case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
2883				(void) printf(gettext(" action: Existing "
2884				    "encrypted datasets contain an on-disk "
2885				    "incompatibility, which\n\tneeds to be "
2886				    "corrected. Backup these datasets to new "
2887				    "encrypted datasets\n\tand destroy the "
2888				    "old ones.\n"));
2889				break;
2890
2891			case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
2892				(void) printf(gettext(" action: Existing "
2893				    "encrypted snapshots and bookmarks contain "
2894				    "an on-disk\n\tincompatibility. This may "
2895				    "cause on-disk corruption if they are used"
2896				    "\n\twith 'zfs recv'. To correct the "
2897				    "issue, enable the bookmark_v2 feature.\n\t"
2898				    "No additional action is needed if there "
2899				    "are no encrypted snapshots or\n\t"
2900				    "bookmarks. If preserving the encrypted "
2901				    "snapshots and bookmarks is\n\trequired, "
2902				    "use a non-raw send to backup and restore "
2903				    "them. Alternately,\n\tthey may be removed"
2904				    " to resolve the incompatibility.\n"));
2905				break;
2906			default:
2907				/*
2908				 * All errata must contain an action message.
2909				 */
2910				assert(0);
2911			}
2912		} else {
2913			(void) printf(gettext(" action: The pool can be "
2914			    "imported using its name or numeric "
2915			    "identifier.\n"));
2916		}
2917	} else if (vs->vs_state == VDEV_STATE_DEGRADED) {
2918		(void) printf(gettext(" action: The pool can be imported "
2919		    "despite missing or damaged devices.  The\n\tfault "
2920		    "tolerance of the pool may be compromised if imported.\n"));
2921	} else {
2922		switch (reason) {
2923		case ZPOOL_STATUS_VERSION_NEWER:
2924			(void) printf(gettext(" action: The pool cannot be "
2925			    "imported.  Access the pool on a system running "
2926			    "newer\n\tsoftware, or recreate the pool from "
2927			    "backup.\n"));
2928			break;
2929		case ZPOOL_STATUS_UNSUP_FEAT_READ:
2930			printf_color(ANSI_BOLD, gettext("action: "));
2931			printf_color(ANSI_YELLOW, gettext("The pool cannot be "
2932			    "imported. Access the pool on a system that "
2933			    "supports\n\tthe required feature(s), or recreate "
2934			    "the pool from backup.\n"));
2935			break;
2936		case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
2937			printf_color(ANSI_BOLD, gettext("action: "));
2938			printf_color(ANSI_YELLOW, gettext("The pool cannot be "
2939			    "imported in read-write mode. Import the pool "
2940			    "with\n"
2941			    "\t\"-o readonly=on\", access the pool on a system "
2942			    "that supports the\n\trequired feature(s), or "
2943			    "recreate the pool from backup.\n"));
2944			break;
2945		case ZPOOL_STATUS_MISSING_DEV_R:
2946		case ZPOOL_STATUS_MISSING_DEV_NR:
2947		case ZPOOL_STATUS_BAD_GUID_SUM:
2948			(void) printf(gettext(" action: The pool cannot be "
2949			    "imported. Attach the missing\n\tdevices and try "
2950			    "again.\n"));
2951			break;
2952		case ZPOOL_STATUS_HOSTID_ACTIVE:
2953			VERIFY0(nvlist_lookup_nvlist(config,
2954			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo));
2955
2956			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
2957				hostname = fnvlist_lookup_string(nvinfo,
2958				    ZPOOL_CONFIG_MMP_HOSTNAME);
2959
2960			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
2961				hostid = fnvlist_lookup_uint64(nvinfo,
2962				    ZPOOL_CONFIG_MMP_HOSTID);
2963
2964			(void) printf(gettext(" action: The pool must be "
2965			    "exported from %s (hostid=%lx)\n\tbefore it "
2966			    "can be safely imported.\n"), hostname,
2967			    (unsigned long) hostid);
2968			break;
2969		case ZPOOL_STATUS_HOSTID_REQUIRED:
2970			(void) printf(gettext(" action: Set a unique system "
2971			    "hostid with the zgenhostid(8) command.\n"));
2972			break;
2973		default:
2974			(void) printf(gettext(" action: The pool cannot be "
2975			    "imported due to damaged devices or data.\n"));
2976		}
2977	}
2978
2979	/* Print the comment attached to the pool. */
2980	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
2981		(void) printf(gettext("comment: %s\n"), comment);
2982
2983	/*
2984	 * If the state is "closed" or "can't open", and the aux state
2985	 * is "corrupt data":
2986	 */
2987	if (((vs->vs_state == VDEV_STATE_CLOSED) ||
2988	    (vs->vs_state == VDEV_STATE_CANT_OPEN)) &&
2989	    (vs->vs_aux == VDEV_AUX_CORRUPT_DATA)) {
2990		if (pool_state == POOL_STATE_DESTROYED)
2991			(void) printf(gettext("\tThe pool was destroyed, "
2992			    "but can be imported using the '-Df' flags.\n"));
2993		else if (pool_state != POOL_STATE_EXPORTED)
2994			(void) printf(gettext("\tThe pool may be active on "
2995			    "another system, but can be imported using\n\t"
2996			    "the '-f' flag.\n"));
2997	}
2998
2999	if (msgid != NULL) {
3000		(void) printf(gettext(
3001		    "   see: https://openzfs.github.io/openzfs-docs/msg/%s\n"),
3002		    msgid);
3003	}
3004
3005	(void) printf(gettext(" config:\n\n"));
3006
3007	cb.cb_namewidth = max_width(NULL, nvroot, 0, strlen(name),
3008	    VDEV_NAME_TYPE_ID);
3009	if (cb.cb_namewidth < 10)
3010		cb.cb_namewidth = 10;
3011
3012	print_import_config(&cb, name, nvroot, 0);
3013
3014	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_DEDUP);
3015	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
3016	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_CLASS_LOGS);
3017
3018	if (reason == ZPOOL_STATUS_BAD_GUID_SUM) {
3019		(void) printf(gettext("\n\tAdditional devices are known to "
3020		    "be part of this pool, though their\n\texact "
3021		    "configuration cannot be determined.\n"));
3022	}
3023	return (0);
3024}
3025
3026static boolean_t
3027zfs_force_import_required(nvlist_t *config)
3028{
3029	uint64_t state;
3030	uint64_t hostid = 0;
3031	nvlist_t *nvinfo;
3032
3033	state = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE);
3034	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
3035
3036	if (state != POOL_STATE_EXPORTED && hostid != get_system_hostid())
3037		return (B_TRUE);
3038
3039	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
3040	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE)) {
3041		mmp_state_t mmp_state = fnvlist_lookup_uint64(nvinfo,
3042		    ZPOOL_CONFIG_MMP_STATE);
3043
3044		if (mmp_state != MMP_STATE_INACTIVE)
3045			return (B_TRUE);
3046	}
3047
3048	return (B_FALSE);
3049}
3050
3051/*
3052 * Perform the import for the given configuration.  This passes the heavy
3053 * lifting off to zpool_import_props(), and then mounts the datasets contained
3054 * within the pool.
3055 */
3056static int
3057do_import(nvlist_t *config, const char *newname, const char *mntopts,
3058    nvlist_t *props, int flags)
3059{
3060	int ret = 0;
3061	zpool_handle_t *zhp;
3062	char *name;
3063	uint64_t version;
3064
3065	name = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);
3066	version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);
3067
3068	if (!SPA_VERSION_IS_SUPPORTED(version)) {
3069		(void) fprintf(stderr, gettext("cannot import '%s': pool "
3070		    "is formatted using an unsupported ZFS version\n"), name);
3071		return (1);
3072	} else if (zfs_force_import_required(config) &&
3073	    !(flags & ZFS_IMPORT_ANY_HOST)) {
3074		mmp_state_t mmp_state = MMP_STATE_INACTIVE;
3075		nvlist_t *nvinfo;
3076
3077		nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
3078		if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE))
3079			mmp_state = fnvlist_lookup_uint64(nvinfo,
3080			    ZPOOL_CONFIG_MMP_STATE);
3081
3082		if (mmp_state == MMP_STATE_ACTIVE) {
3083			char *hostname = "<unknown>";
3084			uint64_t hostid = 0;
3085
3086			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
3087				hostname = fnvlist_lookup_string(nvinfo,
3088				    ZPOOL_CONFIG_MMP_HOSTNAME);
3089
3090			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
3091				hostid = fnvlist_lookup_uint64(nvinfo,
3092				    ZPOOL_CONFIG_MMP_HOSTID);
3093
3094			(void) fprintf(stderr, gettext("cannot import '%s': "
3095			    "pool is imported on %s (hostid: "
3096			    "0x%lx)\nExport the pool on the other system, "
3097			    "then run 'zpool import'.\n"),
3098			    name, hostname, (unsigned long) hostid);
3099		} else if (mmp_state == MMP_STATE_NO_HOSTID) {
3100			(void) fprintf(stderr, gettext("Cannot import '%s': "
3101			    "pool has the multihost property on and the\n"
3102			    "system's hostid is not set. Set a unique hostid "
3103			    "with the zgenhostid(8) command.\n"), name);
3104		} else {
3105			char *hostname = "<unknown>";
3106			uint64_t timestamp = 0;
3107			uint64_t hostid = 0;
3108
3109			if (nvlist_exists(config, ZPOOL_CONFIG_HOSTNAME))
3110				hostname = fnvlist_lookup_string(config,
3111				    ZPOOL_CONFIG_HOSTNAME);
3112
3113			if (nvlist_exists(config, ZPOOL_CONFIG_TIMESTAMP))
3114				timestamp = fnvlist_lookup_uint64(config,
3115				    ZPOOL_CONFIG_TIMESTAMP);
3116
3117			if (nvlist_exists(config, ZPOOL_CONFIG_HOSTID))
3118				hostid = fnvlist_lookup_uint64(config,
3119				    ZPOOL_CONFIG_HOSTID);
3120
3121			(void) fprintf(stderr, gettext("cannot import '%s': "
3122			    "pool was previously in use from another system.\n"
3123			    "Last accessed by %s (hostid=%lx) at %s"
3124			    "The pool can be imported, use 'zpool import -f' "
3125			    "to import the pool.\n"), name, hostname,
3126			    (unsigned long)hostid, ctime((time_t *)&timestamp));
3127		}
3128
3129		return (1);
3130	}
3131
3132	if (zpool_import_props(g_zfs, config, newname, props, flags) != 0)
3133		return (1);
3134
3135	if (newname != NULL)
3136		name = (char *)newname;
3137
3138	if ((zhp = zpool_open_canfail(g_zfs, name)) == NULL)
3139		return (1);
3140
3141	/*
3142	 * Loading keys is best effort. We don't want to return immediately
3143	 * if it fails but we do want to give the error to the caller.
3144	 */
3145	if (flags & ZFS_IMPORT_LOAD_KEYS) {
3146		ret = zfs_crypto_attempt_load_keys(g_zfs, name);
3147		if (ret != 0)
3148			ret = 1;
3149	}
3150
3151	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
3152	    !(flags & ZFS_IMPORT_ONLY) &&
3153	    zpool_enable_datasets(zhp, mntopts, 0) != 0) {
3154		zpool_close(zhp);
3155		return (1);
3156	}
3157
3158	zpool_close(zhp);
3159	return (ret);
3160}
3161
3162static int
3163import_pools(nvlist_t *pools, nvlist_t *props, char *mntopts, int flags,
3164    char *orig_name, char *new_name,
3165    boolean_t do_destroyed, boolean_t pool_specified, boolean_t do_all,
3166    importargs_t *import)
3167{
3168	nvlist_t *config = NULL;
3169	nvlist_t *found_config = NULL;
3170	uint64_t pool_state;
3171
3172	/*
3173	 * At this point we have a list of import candidate configs. Even if
3174	 * we were searching by pool name or guid, we still need to
3175	 * post-process the list to deal with pool state and possible
3176	 * duplicate names.
3177	 */
3178	int err = 0;
3179	nvpair_t *elem = NULL;
3180	boolean_t first = B_TRUE;
3181	while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
3182
3183		verify(nvpair_value_nvlist(elem, &config) == 0);
3184
3185		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
3186		    &pool_state) == 0);
3187		if (!do_destroyed && pool_state == POOL_STATE_DESTROYED)
3188			continue;
3189		if (do_destroyed && pool_state != POOL_STATE_DESTROYED)
3190			continue;
3191
3192		verify(nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
3193		    import->policy) == 0);
3194
3195		if (!pool_specified) {
3196			if (first)
3197				first = B_FALSE;
3198			else if (!do_all)
3199				(void) printf("\n");
3200
3201			if (do_all) {
3202				err |= do_import(config, NULL, mntopts,
3203				    props, flags);
3204			} else {
3205				/*
3206				 * If we're importing from cachefile, then
3207				 * we don't want to report errors until we
3208				 * are in the scan phase of the import. If
3209				 * we get an error, then we return that error
3210				 * to invoke the scan phase.
3211				 */
3212				if (import->cachefile && !import->scan)
3213					err = show_import(config, B_FALSE);
3214				else
3215					(void) show_import(config, B_TRUE);
3216			}
3217		} else if (import->poolname != NULL) {
3218			char *name;
3219
3220			/*
3221			 * We are searching for a pool based on name.
3222			 */
3223			verify(nvlist_lookup_string(config,
3224			    ZPOOL_CONFIG_POOL_NAME, &name) == 0);
3225
3226			if (strcmp(name, import->poolname) == 0) {
3227				if (found_config != NULL) {
3228					(void) fprintf(stderr, gettext(
3229					    "cannot import '%s': more than "
3230					    "one matching pool\n"),
3231					    import->poolname);
3232					(void) fprintf(stderr, gettext(
3233					    "import by numeric ID instead\n"));
3234					err = B_TRUE;
3235				}
3236				found_config = config;
3237			}
3238		} else {
3239			uint64_t guid;
3240
3241			/*
3242			 * Search for a pool by guid.
3243			 */
3244			verify(nvlist_lookup_uint64(config,
3245			    ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
3246
3247			if (guid == import->guid)
3248				found_config = config;
3249		}
3250	}
3251
3252	/*
3253	 * If we were searching for a specific pool, verify that we found a
3254	 * pool, and then do the import.
3255	 */
3256	if (pool_specified && err == 0) {
3257		if (found_config == NULL) {
3258			(void) fprintf(stderr, gettext("cannot import '%s': "
3259			    "no such pool available\n"), orig_name);
3260			err = B_TRUE;
3261		} else {
3262			err |= do_import(found_config, new_name,
3263			    mntopts, props, flags);
3264		}
3265	}
3266
3267	/*
3268	 * If we were just looking for pools, report an error if none were
3269	 * found.
3270	 */
3271	if (!pool_specified && first)
3272		(void) fprintf(stderr,
3273		    gettext("no pools available to import\n"));
3274	return (err);
3275}
3276
3277typedef struct target_exists_args {
3278	const char	*poolname;
3279	uint64_t	poolguid;
3280} target_exists_args_t;
3281
3282static int
3283name_or_guid_exists(zpool_handle_t *zhp, void *data)
3284{
3285	target_exists_args_t *args = data;
3286	nvlist_t *config = zpool_get_config(zhp, NULL);
3287	int found = 0;
3288
3289	if (config == NULL)
3290		return (0);
3291
3292	if (args->poolname != NULL) {
3293		char *pool_name;
3294
3295		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
3296		    &pool_name) == 0);
3297		if (strcmp(pool_name, args->poolname) == 0)
3298			found = 1;
3299	} else {
3300		uint64_t pool_guid;
3301
3302		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
3303		    &pool_guid) == 0);
3304		if (pool_guid == args->poolguid)
3305			found = 1;
3306	}
3307	zpool_close(zhp);
3308
3309	return (found);
3310}
3311/*
3312 * zpool checkpoint <pool>
3313 *       checkpoint --discard <pool>
3314 *
3315 *       -d         Discard the checkpoint from a checkpointed
3316 *       --discard  pool.
3317 *
3318 *       -w         Wait for discarding a checkpoint to complete.
3319 *       --wait
3320 *
3321 * Checkpoints the specified pool, by taking a "snapshot" of its
3322 * current state. A pool can only have one checkpoint at a time.
3323 */
3324int
3325zpool_do_checkpoint(int argc, char **argv)
3326{
3327	boolean_t discard, wait;
3328	char *pool;
3329	zpool_handle_t *zhp;
3330	int c, err;
3331
3332	struct option long_options[] = {
3333		{"discard", no_argument, NULL, 'd'},
3334		{"wait", no_argument, NULL, 'w'},
3335		{0, 0, 0, 0}
3336	};
3337
3338	discard = B_FALSE;
3339	wait = B_FALSE;
3340	while ((c = getopt_long(argc, argv, ":dw", long_options, NULL)) != -1) {
3341		switch (c) {
3342		case 'd':
3343			discard = B_TRUE;
3344			break;
3345		case 'w':
3346			wait = B_TRUE;
3347			break;
3348		case '?':
3349			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
3350			    optopt);
3351			usage(B_FALSE);
3352		}
3353	}
3354
3355	if (wait && !discard) {
3356		(void) fprintf(stderr, gettext("--wait only valid when "
3357		    "--discard also specified\n"));
3358		usage(B_FALSE);
3359	}
3360
3361	argc -= optind;
3362	argv += optind;
3363
3364	if (argc < 1) {
3365		(void) fprintf(stderr, gettext("missing pool argument\n"));
3366		usage(B_FALSE);
3367	}
3368
3369	if (argc > 1) {
3370		(void) fprintf(stderr, gettext("too many arguments\n"));
3371		usage(B_FALSE);
3372	}
3373
3374	pool = argv[0];
3375
3376	if ((zhp = zpool_open(g_zfs, pool)) == NULL) {
3377		/* As a special case, check for use of '/' in the name */
3378		if (strchr(pool, '/') != NULL)
3379			(void) fprintf(stderr, gettext("'zpool checkpoint' "
3380			    "doesn't work on datasets. To save the state "
3381			    "of a dataset from a specific point in time "
3382			    "please use 'zfs snapshot'\n"));
3383		return (1);
3384	}
3385
3386	if (discard) {
3387		err = (zpool_discard_checkpoint(zhp) != 0);
3388		if (err == 0 && wait)
3389			err = zpool_wait(zhp, ZPOOL_WAIT_CKPT_DISCARD);
3390	} else {
3391		err = (zpool_checkpoint(zhp) != 0);
3392	}
3393
3394	zpool_close(zhp);
3395
3396	return (err);
3397}
3398
3399#define	CHECKPOINT_OPT	1024
3400
3401/*
3402 * zpool import [-d dir] [-D]
3403 *       import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
3404 *              [-d dir | -c cachefile | -s] [-f] -a
3405 *       import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
3406 *              [-d dir | -c cachefile | -s] [-f] [-n] [-F] <pool | id>
3407 *              [newpool]
3408 *
3409 *	-c	Read pool information from a cachefile instead of searching
3410 *		devices. If importing from a cachefile config fails, then
3411 *		fallback to searching for devices only in the directories that
3412 *		exist in the cachefile.
3413 *
3414 *	-d	Scan in a specific directory, other than /dev/.  More than
3415 *		one directory can be specified using multiple '-d' options.
3416 *
3417 *	-D	Scan for previously destroyed pools or import all or only
3418 *		specified destroyed pools.
3419 *
3420 *	-R	Temporarily import the pool, with all mountpoints relative to
3421 *		the given root.  The pool will remain exported when the machine
3422 *		is rebooted.
3423 *
3424 *	-V	Import even in the presence of faulted vdevs.  This is an
3425 *		intentionally undocumented option for testing purposes, and
3426 *		treats the pool configuration as complete, leaving any bad
3427 *		vdevs in the FAULTED state. In other words, it does verbatim
3428 *		import.
3429 *
3430 *	-f	Force import, even if it appears that the pool is active.
3431 *
3432 *	-F	Attempt rewind if necessary.
3433 *
3434 *	-n	See if rewind would work, but don't actually rewind.
3435 *
3436 *	-N	Import the pool but don't mount datasets.
3437 *
3438 *	-T	Specify a starting txg to use for import. This option is
3439 *		intentionally undocumented option for testing purposes.
3440 *
3441 *	-a	Import all pools found.
3442 *
3443 *	-l	Load encryption keys while importing.
3444 *
3445 *	-o	Set property=value and/or temporary mount options (without '=').
3446 *
3447 *	-s	Scan using the default search path, the libblkid cache will
3448 *		not be consulted.
3449 *
3450 *	--rewind-to-checkpoint
3451 *		Import the pool and revert back to the checkpoint.
3452 *
3453 * The import command scans for pools to import, and import pools based on pool
3454 * name and GUID.  The pool can also be renamed as part of the import process.
3455 */
3456int
3457zpool_do_import(int argc, char **argv)
3458{
3459	char **searchdirs = NULL;
3460	char *env, *envdup = NULL;
3461	int nsearch = 0;
3462	int c;
3463	int err = 0;
3464	nvlist_t *pools = NULL;
3465	boolean_t do_all = B_FALSE;
3466	boolean_t do_destroyed = B_FALSE;
3467	char *mntopts = NULL;
3468	uint64_t searchguid = 0;
3469	char *searchname = NULL;
3470	char *propval;
3471	nvlist_t *policy = NULL;
3472	nvlist_t *props = NULL;
3473	int flags = ZFS_IMPORT_NORMAL;
3474	uint32_t rewind_policy = ZPOOL_NO_REWIND;
3475	boolean_t dryrun = B_FALSE;
3476	boolean_t do_rewind = B_FALSE;
3477	boolean_t xtreme_rewind = B_FALSE;
3478	boolean_t do_scan = B_FALSE;
3479	boolean_t pool_exists = B_FALSE;
3480	boolean_t pool_specified = B_FALSE;
3481	uint64_t txg = -1ULL;
3482	char *cachefile = NULL;
3483	importargs_t idata = { 0 };
3484	char *endptr;
3485
3486	struct option long_options[] = {
3487		{"rewind-to-checkpoint", no_argument, NULL, CHECKPOINT_OPT},
3488		{0, 0, 0, 0}
3489	};
3490
3491	/* check options */
3492	while ((c = getopt_long(argc, argv, ":aCc:d:DEfFlmnNo:R:stT:VX",
3493	    long_options, NULL)) != -1) {
3494		switch (c) {
3495		case 'a':
3496			do_all = B_TRUE;
3497			break;
3498		case 'c':
3499			cachefile = optarg;
3500			break;
3501		case 'd':
3502			if (searchdirs == NULL) {
3503				searchdirs = safe_malloc(sizeof (char *));
3504			} else {
3505				char **tmp = safe_malloc((nsearch + 1) *
3506				    sizeof (char *));
3507				bcopy(searchdirs, tmp, nsearch *
3508				    sizeof (char *));
3509				free(searchdirs);
3510				searchdirs = tmp;
3511			}
3512			searchdirs[nsearch++] = optarg;
3513			break;
3514		case 'D':
3515			do_destroyed = B_TRUE;
3516			break;
3517		case 'f':
3518			flags |= ZFS_IMPORT_ANY_HOST;
3519			break;
3520		case 'F':
3521			do_rewind = B_TRUE;
3522			break;
3523		case 'l':
3524			flags |= ZFS_IMPORT_LOAD_KEYS;
3525			break;
3526		case 'm':
3527			flags |= ZFS_IMPORT_MISSING_LOG;
3528			break;
3529		case 'n':
3530			dryrun = B_TRUE;
3531			break;
3532		case 'N':
3533			flags |= ZFS_IMPORT_ONLY;
3534			break;
3535		case 'o':
3536			if ((propval = strchr(optarg, '=')) != NULL) {
3537				*propval = '\0';
3538				propval++;
3539				if (add_prop_list(optarg, propval,
3540				    &props, B_TRUE))
3541					goto error;
3542			} else {
3543				mntopts = optarg;
3544			}
3545			break;
3546		case 'R':
3547			if (add_prop_list(zpool_prop_to_name(
3548			    ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
3549				goto error;
3550			if (add_prop_list_default(zpool_prop_to_name(
3551			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
3552				goto error;
3553			break;
3554		case 's':
3555			do_scan = B_TRUE;
3556			break;
3557		case 't':
3558			flags |= ZFS_IMPORT_TEMP_NAME;
3559			if (add_prop_list_default(zpool_prop_to_name(
3560			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
3561				goto error;
3562			break;
3563
3564		case 'T':
3565			errno = 0;
3566			txg = strtoull(optarg, &endptr, 0);
3567			if (errno != 0 || *endptr != '\0') {
3568				(void) fprintf(stderr,
3569				    gettext("invalid txg value\n"));
3570				usage(B_FALSE);
3571			}
3572			rewind_policy = ZPOOL_DO_REWIND | ZPOOL_EXTREME_REWIND;
3573			break;
3574		case 'V':
3575			flags |= ZFS_IMPORT_VERBATIM;
3576			break;
3577		case 'X':
3578			xtreme_rewind = B_TRUE;
3579			break;
3580		case CHECKPOINT_OPT:
3581			flags |= ZFS_IMPORT_CHECKPOINT;
3582			break;
3583		case ':':
3584			(void) fprintf(stderr, gettext("missing argument for "
3585			    "'%c' option\n"), optopt);
3586			usage(B_FALSE);
3587			break;
3588		case '?':
3589			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
3590			    optopt);
3591			usage(B_FALSE);
3592		}
3593	}
3594
3595	argc -= optind;
3596	argv += optind;
3597
3598	if (cachefile && nsearch != 0) {
3599		(void) fprintf(stderr, gettext("-c is incompatible with -d\n"));
3600		usage(B_FALSE);
3601	}
3602
3603	if (cachefile && do_scan) {
3604		(void) fprintf(stderr, gettext("-c is incompatible with -s\n"));
3605		usage(B_FALSE);
3606	}
3607
3608	if ((flags & ZFS_IMPORT_LOAD_KEYS) && (flags & ZFS_IMPORT_ONLY)) {
3609		(void) fprintf(stderr, gettext("-l is incompatible with -N\n"));
3610		usage(B_FALSE);
3611	}
3612
3613	if ((flags & ZFS_IMPORT_LOAD_KEYS) && !do_all && argc == 0) {
3614		(void) fprintf(stderr, gettext("-l is only meaningful during "
3615		    "an import\n"));
3616		usage(B_FALSE);
3617	}
3618
3619	if ((dryrun || xtreme_rewind) && !do_rewind) {
3620		(void) fprintf(stderr,
3621		    gettext("-n or -X only meaningful with -F\n"));
3622		usage(B_FALSE);
3623	}
3624	if (dryrun)
3625		rewind_policy = ZPOOL_TRY_REWIND;
3626	else if (do_rewind)
3627		rewind_policy = ZPOOL_DO_REWIND;
3628	if (xtreme_rewind)
3629		rewind_policy |= ZPOOL_EXTREME_REWIND;
3630
3631	/* In the future, we can capture further policy and include it here */
3632	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 ||
3633	    nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, txg) != 0 ||
3634	    nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
3635	    rewind_policy) != 0)
3636		goto error;
3637
3638	/* check argument count */
3639	if (do_all) {
3640		if (argc != 0) {
3641			(void) fprintf(stderr, gettext("too many arguments\n"));
3642			usage(B_FALSE);
3643		}
3644	} else {
3645		if (argc > 2) {
3646			(void) fprintf(stderr, gettext("too many arguments\n"));
3647			usage(B_FALSE);
3648		}
3649	}
3650
3651	/*
3652	 * Check for the effective uid.  We do this explicitly here because
3653	 * otherwise any attempt to discover pools will silently fail.
3654	 */
3655	if (argc == 0 && geteuid() != 0) {
3656		(void) fprintf(stderr, gettext("cannot "
3657		    "discover pools: permission denied\n"));
3658		if (searchdirs != NULL)
3659			free(searchdirs);
3660
3661		nvlist_free(props);
3662		nvlist_free(policy);
3663		return (1);
3664	}
3665
3666	/*
3667	 * Depending on the arguments given, we do one of the following:
3668	 *
3669	 *	<none>	Iterate through all pools and display information about
3670	 *		each one.
3671	 *
3672	 *	-a	Iterate through all pools and try to import each one.
3673	 *
3674	 *	<id>	Find the pool that corresponds to the given GUID/pool
3675	 *		name and import that one.
3676	 *
3677	 *	-D	Above options applies only to destroyed pools.
3678	 */
3679	if (argc != 0) {
3680		char *endptr;
3681
3682		errno = 0;
3683		searchguid = strtoull(argv[0], &endptr, 10);
3684		if (errno != 0 || *endptr != '\0') {
3685			searchname = argv[0];
3686			searchguid = 0;
3687		}
3688		pool_specified = B_TRUE;
3689
3690		/*
3691		 * User specified a name or guid.  Ensure it's unique.
3692		 */
3693		target_exists_args_t search = {searchname, searchguid};
3694		pool_exists = zpool_iter(g_zfs, name_or_guid_exists, &search);
3695	}
3696
3697	/*
3698	 * Check the environment for the preferred search path.
3699	 */
3700	if ((searchdirs == NULL) && (env = getenv("ZPOOL_IMPORT_PATH"))) {
3701		char *dir;
3702
3703		envdup = strdup(env);
3704
3705		dir = strtok(envdup, ":");
3706		while (dir != NULL) {
3707			if (searchdirs == NULL) {
3708				searchdirs = safe_malloc(sizeof (char *));
3709			} else {
3710				char **tmp = safe_malloc((nsearch + 1) *
3711				    sizeof (char *));
3712				bcopy(searchdirs, tmp, nsearch *
3713				    sizeof (char *));
3714				free(searchdirs);
3715				searchdirs = tmp;
3716			}
3717			searchdirs[nsearch++] = dir;
3718			dir = strtok(NULL, ":");
3719		}
3720	}
3721
3722	idata.path = searchdirs;
3723	idata.paths = nsearch;
3724	idata.poolname = searchname;
3725	idata.guid = searchguid;
3726	idata.cachefile = cachefile;
3727	idata.scan = do_scan;
3728	idata.policy = policy;
3729
3730	pools = zpool_search_import(g_zfs, &idata, &libzfs_config_ops);
3731
3732	if (pools != NULL && pool_exists &&
3733	    (argc == 1 || strcmp(argv[0], argv[1]) == 0)) {
3734		(void) fprintf(stderr, gettext("cannot import '%s': "
3735		    "a pool with that name already exists\n"),
3736		    argv[0]);
3737		(void) fprintf(stderr, gettext("use the form '%s "
3738		    "<pool | id> <newpool>' to give it a new name\n"),
3739		    "zpool import");
3740		err = 1;
3741	} else if (pools == NULL && pool_exists) {
3742		(void) fprintf(stderr, gettext("cannot import '%s': "
3743		    "a pool with that name is already created/imported,\n"),
3744		    argv[0]);
3745		(void) fprintf(stderr, gettext("and no additional pools "
3746		    "with that name were found\n"));
3747		err = 1;
3748	} else if (pools == NULL) {
3749		if (argc != 0) {
3750			(void) fprintf(stderr, gettext("cannot import '%s': "
3751			    "no such pool available\n"), argv[0]);
3752		}
3753		err = 1;
3754	}
3755
3756	if (err == 1) {
3757		if (searchdirs != NULL)
3758			free(searchdirs);
3759		if (envdup != NULL)
3760			free(envdup);
3761		nvlist_free(policy);
3762		nvlist_free(pools);
3763		nvlist_free(props);
3764		return (1);
3765	}
3766
3767	err = import_pools(pools, props, mntopts, flags, argv[0],
3768	    argc == 1 ? NULL : argv[1], do_destroyed, pool_specified,
3769	    do_all, &idata);
3770
3771	/*
3772	 * If we're using the cachefile and we failed to import, then
3773	 * fallback to scanning the directory for pools that match
3774	 * those in the cachefile.
3775	 */
3776	if (err != 0 && cachefile != NULL) {
3777		(void) printf(gettext("cachefile import failed, retrying\n"));
3778
3779		/*
3780		 * We use the scan flag to gather the directories that exist
3781		 * in the cachefile. If we need to fallback to searching for
3782		 * the pool config, we will only search devices in these
3783		 * directories.
3784		 */
3785		idata.scan = B_TRUE;
3786		nvlist_free(pools);
3787		pools = zpool_search_import(g_zfs, &idata, &libzfs_config_ops);
3788
3789		err = import_pools(pools, props, mntopts, flags, argv[0],
3790		    argc == 1 ? NULL : argv[1], do_destroyed, pool_specified,
3791		    do_all, &idata);
3792	}
3793
3794error:
3795	nvlist_free(props);
3796	nvlist_free(pools);
3797	nvlist_free(policy);
3798	if (searchdirs != NULL)
3799		free(searchdirs);
3800	if (envdup != NULL)
3801		free(envdup);
3802
3803	return (err ? 1 : 0);
3804}
3805
3806/*
3807 * zpool sync [-f] [pool] ...
3808 *
3809 * -f (undocumented) force uberblock (and config including zpool cache file)
3810 *    update.
3811 *
3812 * Sync the specified pool(s).
3813 * Without arguments "zpool sync" will sync all pools.
3814 * This command initiates TXG sync(s) and will return after the TXG(s) commit.
3815 *
3816 */
3817static int
3818zpool_do_sync(int argc, char **argv)
3819{
3820	int ret;
3821	boolean_t force = B_FALSE;
3822
3823	/* check options */
3824	while ((ret  = getopt(argc, argv, "f")) != -1) {
3825		switch (ret) {
3826		case 'f':
3827			force = B_TRUE;
3828			break;
3829		case '?':
3830			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
3831			    optopt);
3832			usage(B_FALSE);
3833		}
3834	}
3835
3836	argc -= optind;
3837	argv += optind;
3838
3839	/* if argc == 0 we will execute zpool_sync_one on all pools */
3840	ret = for_each_pool(argc, argv, B_FALSE, NULL, B_FALSE, zpool_sync_one,
3841	    &force);
3842
3843	return (ret);
3844}
3845
3846typedef struct iostat_cbdata {
3847	uint64_t cb_flags;
3848	int cb_name_flags;
3849	int cb_namewidth;
3850	int cb_iteration;
3851	char **cb_vdev_names; /* Only show these vdevs */
3852	unsigned int cb_vdev_names_count;
3853	boolean_t cb_verbose;
3854	boolean_t cb_literal;
3855	boolean_t cb_scripted;
3856	zpool_list_t *cb_list;
3857	vdev_cmd_data_list_t *vcdl;
3858} iostat_cbdata_t;
3859
3860/*  iostat labels */
3861typedef struct name_and_columns {
3862	const char *name;	/* Column name */
3863	unsigned int columns;	/* Center name to this number of columns */
3864} name_and_columns_t;
3865
3866#define	IOSTAT_MAX_LABELS	13	/* Max number of labels on one line */
3867
3868static const name_and_columns_t iostat_top_labels[][IOSTAT_MAX_LABELS] =
3869{
3870	[IOS_DEFAULT] = {{"capacity", 2}, {"operations", 2}, {"bandwidth", 2},
3871	    {NULL}},
3872	[IOS_LATENCY] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
3873	    {"asyncq_wait", 2}, {"scrub", 1}, {"trim", 1}, {NULL}},
3874	[IOS_QUEUES] = {{"syncq_read", 2}, {"syncq_write", 2},
3875	    {"asyncq_read", 2}, {"asyncq_write", 2}, {"scrubq_read", 2},
3876	    {"trimq_write", 2}, {NULL}},
3877	[IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
3878	    {"asyncq_wait", 2}, {NULL}},
3879	[IOS_RQ_HISTO] = {{"sync_read", 2}, {"sync_write", 2},
3880	    {"async_read", 2}, {"async_write", 2}, {"scrub", 2},
3881	    {"trim", 2}, {NULL}},
3882};
3883
3884/* Shorthand - if "columns" field not set, default to 1 column */
3885static const name_and_columns_t iostat_bottom_labels[][IOSTAT_MAX_LABELS] =
3886{
3887	[IOS_DEFAULT] = {{"alloc"}, {"free"}, {"read"}, {"write"}, {"read"},
3888	    {"write"}, {NULL}},
3889	[IOS_LATENCY] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
3890	    {"write"}, {"read"}, {"write"}, {"wait"}, {"wait"}, {NULL}},
3891	[IOS_QUEUES] = {{"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"},
3892	    {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"},
3893	    {"pend"}, {"activ"}, {NULL}},
3894	[IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
3895	    {"write"}, {"read"}, {"write"}, {"scrub"}, {"trim"}, {NULL}},
3896	[IOS_RQ_HISTO] = {{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
3897	    {"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {NULL}},
3898};
3899
3900static const char *histo_to_title[] = {
3901	[IOS_L_HISTO] = "latency",
3902	[IOS_RQ_HISTO] = "req_size",
3903};
3904
3905/*
3906 * Return the number of labels in a null-terminated name_and_columns_t
3907 * array.
3908 *
3909 */
3910static unsigned int
3911label_array_len(const name_and_columns_t *labels)
3912{
3913	int i = 0;
3914
3915	while (labels[i].name)
3916		i++;
3917
3918	return (i);
3919}
3920
3921/*
3922 * Return the number of strings in a null-terminated string array.
3923 * For example:
3924 *
3925 *     const char foo[] = {"bar", "baz", NULL}
3926 *
3927 * returns 2
3928 */
3929static uint64_t
3930str_array_len(const char *array[])
3931{
3932	uint64_t i = 0;
3933	while (array[i])
3934		i++;
3935
3936	return (i);
3937}
3938
3939
3940/*
3941 * Return a default column width for default/latency/queue columns. This does
3942 * not include histograms, which have their columns autosized.
3943 */
3944static unsigned int
3945default_column_width(iostat_cbdata_t *cb, enum iostat_type type)
3946{
3947	unsigned long column_width = 5; /* Normal niceprint */
3948	static unsigned long widths[] = {
3949		/*
3950		 * Choose some sane default column sizes for printing the
3951		 * raw numbers.
3952		 */
3953		[IOS_DEFAULT] = 15, /* 1PB capacity */
3954		[IOS_LATENCY] = 10, /* 1B ns = 10sec */
3955		[IOS_QUEUES] = 6,   /* 1M queue entries */
3956		[IOS_L_HISTO] = 10, /* 1B ns = 10sec */
3957		[IOS_RQ_HISTO] = 6, /* 1M queue entries */
3958	};
3959
3960	if (cb->cb_literal)
3961		column_width = widths[type];
3962
3963	return (column_width);
3964}
3965
3966/*
3967 * Print the column labels, i.e:
3968 *
3969 *   capacity     operations     bandwidth
3970 * alloc   free   read  write   read  write  ...
3971 *
3972 * If force_column_width is set, use it for the column width.  If not set, use
3973 * the default column width.
3974 */
3975static void
3976print_iostat_labels(iostat_cbdata_t *cb, unsigned int force_column_width,
3977    const name_and_columns_t labels[][IOSTAT_MAX_LABELS])
3978{
3979	int i, idx, s;
3980	int text_start, rw_column_width, spaces_to_end;
3981	uint64_t flags = cb->cb_flags;
3982	uint64_t f;
3983	unsigned int column_width = force_column_width;
3984
3985	/* For each bit set in flags */
3986	for (f = flags; f; f &= ~(1ULL << idx)) {
3987		idx = lowbit64(f) - 1;
3988		if (!force_column_width)
3989			column_width = default_column_width(cb, idx);
3990		/* Print our top labels centered over "read  write" label. */
3991		for (i = 0; i < label_array_len(labels[idx]); i++) {
3992			const char *name = labels[idx][i].name;
3993			/*
3994			 * We treat labels[][].columns == 0 as shorthand
3995			 * for one column.  It makes writing out the label
3996			 * tables more concise.
3997			 */
3998			unsigned int columns = MAX(1, labels[idx][i].columns);
3999			unsigned int slen = strlen(name);
4000
4001			rw_column_width = (column_width * columns) +
4002			    (2 * (columns - 1));
4003
4004			text_start = (int)((rw_column_width) / columns -
4005			    slen / columns);
4006			if (text_start < 0)
4007				text_start = 0;
4008
4009			printf("  ");	/* Two spaces between columns */
4010
4011			/* Space from beginning of column to label */
4012			for (s = 0; s < text_start; s++)
4013				printf(" ");
4014
4015			printf("%s", name);
4016
4017			/* Print space after label to end of column */
4018			spaces_to_end = rw_column_width - text_start - slen;
4019			if (spaces_to_end < 0)
4020				spaces_to_end = 0;
4021
4022			for (s = 0; s < spaces_to_end; s++)
4023				printf(" ");
4024		}
4025	}
4026}
4027
4028
4029/*
4030 * print_cmd_columns - Print custom column titles from -c
4031 *
4032 * If the user specified the "zpool status|iostat -c" then print their custom
4033 * column titles in the header.  For example, print_cmd_columns() would print
4034 * the "  col1  col2" part of this:
4035 *
4036 * $ zpool iostat -vc 'echo col1=val1; echo col2=val2'
4037 * ...
4038 *	      capacity     operations     bandwidth
4039 * pool        alloc   free   read  write   read  write  col1  col2
4040 * ----------  -----  -----  -----  -----  -----  -----  ----  ----
4041 * mypool       269K  1008M      0      0    107    946
4042 *   mirror     269K  1008M      0      0    107    946
4043 *     sdb         -      -      0      0    102    473  val1  val2
4044 *     sdc         -      -      0      0      5    473  val1  val2
4045 * ----------  -----  -----  -----  -----  -----  -----  ----  ----
4046 */
4047static void
4048print_cmd_columns(vdev_cmd_data_list_t *vcdl, int use_dashes)
4049{
4050	int i, j;
4051	vdev_cmd_data_t *data = &vcdl->data[0];
4052
4053	if (vcdl->count == 0 || data == NULL)
4054		return;
4055
4056	/*
4057	 * Each vdev cmd should have the same column names unless the user did
4058	 * something weird with their cmd.  Just take the column names from the
4059	 * first vdev and assume it works for all of them.
4060	 */
4061	for (i = 0; i < vcdl->uniq_cols_cnt; i++) {
4062		printf("  ");
4063		if (use_dashes) {
4064			for (j = 0; j < vcdl->uniq_cols_width[i]; j++)
4065				printf("-");
4066		} else {
4067			printf_color(ANSI_BOLD, "%*s", vcdl->uniq_cols_width[i],
4068			    vcdl->uniq_cols[i]);
4069		}
4070	}
4071}
4072
4073
4074/*
4075 * Utility function to print out a line of dashes like:
4076 *
4077 * 	--------------------------------  -----  -----  -----  -----  -----
4078 *
4079 * ...or a dashed named-row line like:
4080 *
4081 * 	logs                                  -      -      -      -      -
4082 *
4083 * @cb:				iostat data
4084 *
4085 * @force_column_width		If non-zero, use the value as the column width.
4086 * 				Otherwise use the default column widths.
4087 *
4088 * @name:			Print a dashed named-row line starting
4089 * 				with @name.  Otherwise, print a regular
4090 * 				dashed line.
4091 */
4092static void
4093print_iostat_dashes(iostat_cbdata_t *cb, unsigned int force_column_width,
4094    const char *name)
4095{
4096	int i;
4097	unsigned int namewidth;
4098	uint64_t flags = cb->cb_flags;
4099	uint64_t f;
4100	int idx;
4101	const name_and_columns_t *labels;
4102	const char *title;
4103
4104
4105	if (cb->cb_flags & IOS_ANYHISTO_M) {
4106		title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
4107	} else if (cb->cb_vdev_names_count) {
4108		title = "vdev";
4109	} else  {
4110		title = "pool";
4111	}
4112
4113	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
4114	    name ? strlen(name) : 0);
4115
4116
4117	if (name) {
4118		printf("%-*s", namewidth, name);
4119	} else {
4120		for (i = 0; i < namewidth; i++)
4121			(void) printf("-");
4122	}
4123
4124	/* For each bit in flags */
4125	for (f = flags; f; f &= ~(1ULL << idx)) {
4126		unsigned int column_width;
4127		idx = lowbit64(f) - 1;
4128		if (force_column_width)
4129			column_width = force_column_width;
4130		else
4131			column_width = default_column_width(cb, idx);
4132
4133		labels = iostat_bottom_labels[idx];
4134		for (i = 0; i < label_array_len(labels); i++) {
4135			if (name)
4136				printf("  %*s-", column_width - 1, " ");
4137			else
4138				printf("  %.*s", column_width,
4139				    "--------------------");
4140		}
4141	}
4142}
4143
4144
4145static void
4146print_iostat_separator_impl(iostat_cbdata_t *cb,
4147    unsigned int force_column_width)
4148{
4149	print_iostat_dashes(cb, force_column_width, NULL);
4150}
4151
4152static void
4153print_iostat_separator(iostat_cbdata_t *cb)
4154{
4155	print_iostat_separator_impl(cb, 0);
4156}
4157
4158static void
4159print_iostat_header_impl(iostat_cbdata_t *cb, unsigned int force_column_width,
4160    const char *histo_vdev_name)
4161{
4162	unsigned int namewidth;
4163	const char *title;
4164
4165	if (cb->cb_flags & IOS_ANYHISTO_M) {
4166		title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
4167	} else if (cb->cb_vdev_names_count) {
4168		title = "vdev";
4169	} else  {
4170		title = "pool";
4171	}
4172
4173	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
4174	    histo_vdev_name ? strlen(histo_vdev_name) : 0);
4175
4176	if (histo_vdev_name)
4177		printf("%-*s", namewidth, histo_vdev_name);
4178	else
4179		printf("%*s", namewidth, "");
4180
4181
4182	print_iostat_labels(cb, force_column_width, iostat_top_labels);
4183	printf("\n");
4184
4185	printf("%-*s", namewidth, title);
4186
4187	print_iostat_labels(cb, force_column_width, iostat_bottom_labels);
4188	if (cb->vcdl != NULL)
4189		print_cmd_columns(cb->vcdl, 0);
4190
4191	printf("\n");
4192
4193	print_iostat_separator_impl(cb, force_column_width);
4194
4195	if (cb->vcdl != NULL)
4196		print_cmd_columns(cb->vcdl, 1);
4197
4198	printf("\n");
4199}
4200
4201static void
4202print_iostat_header(iostat_cbdata_t *cb)
4203{
4204	print_iostat_header_impl(cb, 0, NULL);
4205}
4206
4207
4208/*
4209 * Display a single statistic.
4210 */
4211static void
4212print_one_stat(uint64_t value, enum zfs_nicenum_format format,
4213    unsigned int column_size, boolean_t scripted)
4214{
4215	char buf[64];
4216
4217	zfs_nicenum_format(value, buf, sizeof (buf), format);
4218
4219	if (scripted)
4220		printf("\t%s", buf);
4221	else
4222		printf("  %*s", column_size, buf);
4223}
4224
4225/*
4226 * Calculate the default vdev stats
4227 *
4228 * Subtract oldvs from newvs, apply a scaling factor, and save the resulting
4229 * stats into calcvs.
4230 */
4231static void
4232calc_default_iostats(vdev_stat_t *oldvs, vdev_stat_t *newvs,
4233    vdev_stat_t *calcvs)
4234{
4235	int i;
4236
4237	memcpy(calcvs, newvs, sizeof (*calcvs));
4238	for (i = 0; i < ARRAY_SIZE(calcvs->vs_ops); i++)
4239		calcvs->vs_ops[i] = (newvs->vs_ops[i] - oldvs->vs_ops[i]);
4240
4241	for (i = 0; i < ARRAY_SIZE(calcvs->vs_bytes); i++)
4242		calcvs->vs_bytes[i] = (newvs->vs_bytes[i] - oldvs->vs_bytes[i]);
4243}
4244
4245/*
4246 * Internal representation of the extended iostats data.
4247 *
4248 * The extended iostat stats are exported in nvlists as either uint64_t arrays
4249 * or single uint64_t's.  We make both look like arrays to make them easier
4250 * to process.  In order to make single uint64_t's look like arrays, we set
4251 * __data to the stat data, and then set *data = &__data with count = 1.  Then,
4252 * we can just use *data and count.
4253 */
4254struct stat_array {
4255	uint64_t *data;
4256	uint_t count;	/* Number of entries in data[] */
4257	uint64_t __data; /* Only used when data is a single uint64_t */
4258};
4259
4260static uint64_t
4261stat_histo_max(struct stat_array *nva, unsigned int len)
4262{
4263	uint64_t max = 0;
4264	int i;
4265	for (i = 0; i < len; i++)
4266		max = MAX(max, array64_max(nva[i].data, nva[i].count));
4267
4268	return (max);
4269}
4270
4271/*
4272 * Helper function to lookup a uint64_t array or uint64_t value and store its
4273 * data as a stat_array.  If the nvpair is a single uint64_t value, then we make
4274 * it look like a one element array to make it easier to process.
4275 */
4276static int
4277nvpair64_to_stat_array(nvlist_t *nvl, const char *name,
4278    struct stat_array *nva)
4279{
4280	nvpair_t *tmp;
4281	int ret;
4282
4283	verify(nvlist_lookup_nvpair(nvl, name, &tmp) == 0);
4284	switch (nvpair_type(tmp)) {
4285	case DATA_TYPE_UINT64_ARRAY:
4286		ret = nvpair_value_uint64_array(tmp, &nva->data, &nva->count);
4287		break;
4288	case DATA_TYPE_UINT64:
4289		ret = nvpair_value_uint64(tmp, &nva->__data);
4290		nva->data = &nva->__data;
4291		nva->count = 1;
4292		break;
4293	default:
4294		/* Not a uint64_t */
4295		ret = EINVAL;
4296		break;
4297	}
4298
4299	return (ret);
4300}
4301
4302/*
4303 * Given a list of nvlist names, look up the extended stats in newnv and oldnv,
4304 * subtract them, and return the results in a newly allocated stat_array.
4305 * You must free the returned array after you are done with it with
4306 * free_calc_stats().
4307 *
4308 * Additionally, you can set "oldnv" to NULL if you simply want the newnv
4309 * values.
4310 */
4311static struct stat_array *
4312calc_and_alloc_stats_ex(const char **names, unsigned int len, nvlist_t *oldnv,
4313    nvlist_t *newnv)
4314{
4315	nvlist_t *oldnvx = NULL, *newnvx;
4316	struct stat_array *oldnva, *newnva, *calcnva;
4317	int i, j;
4318	unsigned int alloc_size = (sizeof (struct stat_array)) * len;
4319
4320	/* Extract our extended stats nvlist from the main list */
4321	verify(nvlist_lookup_nvlist(newnv, ZPOOL_CONFIG_VDEV_STATS_EX,
4322	    &newnvx) == 0);
4323	if (oldnv) {
4324		verify(nvlist_lookup_nvlist(oldnv, ZPOOL_CONFIG_VDEV_STATS_EX,
4325		    &oldnvx) == 0);
4326	}
4327
4328	newnva = safe_malloc(alloc_size);
4329	oldnva = safe_malloc(alloc_size);
4330	calcnva = safe_malloc(alloc_size);
4331
4332	for (j = 0; j < len; j++) {
4333		verify(nvpair64_to_stat_array(newnvx, names[j],
4334		    &newnva[j]) == 0);
4335		calcnva[j].count = newnva[j].count;
4336		alloc_size = calcnva[j].count * sizeof (calcnva[j].data[0]);
4337		calcnva[j].data = safe_malloc(alloc_size);
4338		memcpy(calcnva[j].data, newnva[j].data, alloc_size);
4339
4340		if (oldnvx) {
4341			verify(nvpair64_to_stat_array(oldnvx, names[j],
4342			    &oldnva[j]) == 0);
4343			for (i = 0; i < oldnva[j].count; i++)
4344				calcnva[j].data[i] -= oldnva[j].data[i];
4345		}
4346	}
4347	free(newnva);
4348	free(oldnva);
4349	return (calcnva);
4350}
4351
4352static void
4353free_calc_stats(struct stat_array *nva, unsigned int len)
4354{
4355	int i;
4356	for (i = 0; i < len; i++)
4357		free(nva[i].data);
4358
4359	free(nva);
4360}
4361
4362static void
4363print_iostat_histo(struct stat_array *nva, unsigned int len,
4364    iostat_cbdata_t *cb, unsigned int column_width, unsigned int namewidth,
4365    double scale)
4366{
4367	int i, j;
4368	char buf[6];
4369	uint64_t val;
4370	enum zfs_nicenum_format format;
4371	unsigned int buckets;
4372	unsigned int start_bucket;
4373
4374	if (cb->cb_literal)
4375		format = ZFS_NICENUM_RAW;
4376	else
4377		format = ZFS_NICENUM_1024;
4378
4379	/* All these histos are the same size, so just use nva[0].count */
4380	buckets = nva[0].count;
4381
4382	if (cb->cb_flags & IOS_RQ_HISTO_M) {
4383		/* Start at 512 - req size should never be lower than this */
4384		start_bucket = 9;
4385	} else {
4386		start_bucket = 0;
4387	}
4388
4389	for (j = start_bucket; j < buckets; j++) {
4390		/* Print histogram bucket label */
4391		if (cb->cb_flags & IOS_L_HISTO_M) {
4392			/* Ending range of this bucket */
4393			val = (1UL << (j + 1)) - 1;
4394			zfs_nicetime(val, buf, sizeof (buf));
4395		} else {
4396			/* Request size (starting range of bucket) */
4397			val = (1UL << j);
4398			zfs_nicenum(val, buf, sizeof (buf));
4399		}
4400
4401		if (cb->cb_scripted)
4402			printf("%llu", (u_longlong_t)val);
4403		else
4404			printf("%-*s", namewidth, buf);
4405
4406		/* Print the values on the line */
4407		for (i = 0; i < len; i++) {
4408			print_one_stat(nva[i].data[j] * scale, format,
4409			    column_width, cb->cb_scripted);
4410		}
4411		printf("\n");
4412	}
4413}
4414
4415static void
4416print_solid_separator(unsigned int length)
4417{
4418	while (length--)
4419		printf("-");
4420	printf("\n");
4421}
4422
4423static void
4424print_iostat_histos(iostat_cbdata_t *cb, nvlist_t *oldnv,
4425    nvlist_t *newnv, double scale, const char *name)
4426{
4427	unsigned int column_width;
4428	unsigned int namewidth;
4429	unsigned int entire_width;
4430	enum iostat_type type;
4431	struct stat_array *nva;
4432	const char **names;
4433	unsigned int names_len;
4434
4435	/* What type of histo are we? */
4436	type = IOS_HISTO_IDX(cb->cb_flags);
4437
4438	/* Get NULL-terminated array of nvlist names for our histo */
4439	names = vsx_type_to_nvlist[type];
4440	names_len = str_array_len(names); /* num of names */
4441
4442	nva = calc_and_alloc_stats_ex(names, names_len, oldnv, newnv);
4443
4444	if (cb->cb_literal) {
4445		column_width = MAX(5,
4446		    (unsigned int) log10(stat_histo_max(nva, names_len)) + 1);
4447	} else {
4448		column_width = 5;
4449	}
4450
4451	namewidth = MAX(cb->cb_namewidth,
4452	    strlen(histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]));
4453
4454	/*
4455	 * Calculate the entire line width of what we're printing.  The
4456	 * +2 is for the two spaces between columns:
4457	 */
4458	/*	 read  write				*/
4459	/*	-----  -----				*/
4460	/*	|___|  <---------- column_width		*/
4461	/*						*/
4462	/*	|__________|  <--- entire_width		*/
4463	/*						*/
4464	entire_width = namewidth + (column_width + 2) *
4465	    label_array_len(iostat_bottom_labels[type]);
4466
4467	if (cb->cb_scripted)
4468		printf("%s\n", name);
4469	else
4470		print_iostat_header_impl(cb, column_width, name);
4471
4472	print_iostat_histo(nva, names_len, cb, column_width,
4473	    namewidth, scale);
4474
4475	free_calc_stats(nva, names_len);
4476	if (!cb->cb_scripted)
4477		print_solid_separator(entire_width);
4478}
4479
4480/*
4481 * Calculate the average latency of a power-of-two latency histogram
4482 */
4483static uint64_t
4484single_histo_average(uint64_t *histo, unsigned int buckets)
4485{
4486	int i;
4487	uint64_t count = 0, total = 0;
4488
4489	for (i = 0; i < buckets; i++) {
4490		/*
4491		 * Our buckets are power-of-two latency ranges.  Use the
4492		 * midpoint latency of each bucket to calculate the average.
4493		 * For example:
4494		 *
4495		 * Bucket          Midpoint
4496		 * 8ns-15ns:       12ns
4497		 * 16ns-31ns:      24ns
4498		 * ...
4499		 */
4500		if (histo[i] != 0) {
4501			total += histo[i] * (((1UL << i) + ((1UL << i)/2)));
4502			count += histo[i];
4503		}
4504	}
4505
4506	/* Prevent divide by zero */
4507	return (count == 0 ? 0 : total / count);
4508}
4509
4510static void
4511print_iostat_queues(iostat_cbdata_t *cb, nvlist_t *oldnv,
4512    nvlist_t *newnv)
4513{
4514	int i;
4515	uint64_t val;
4516	const char *names[] = {
4517		ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE,
4518		ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
4519		ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE,
4520		ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
4521		ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE,
4522		ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
4523		ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE,
4524		ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
4525		ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE,
4526		ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
4527		ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE,
4528		ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
4529	};
4530
4531	struct stat_array *nva;
4532
4533	unsigned int column_width = default_column_width(cb, IOS_QUEUES);
4534	enum zfs_nicenum_format format;
4535
4536	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), NULL, newnv);
4537
4538	if (cb->cb_literal)
4539		format = ZFS_NICENUM_RAW;
4540	else
4541		format = ZFS_NICENUM_1024;
4542
4543	for (i = 0; i < ARRAY_SIZE(names); i++) {
4544		val = nva[i].data[0];
4545		print_one_stat(val, format, column_width, cb->cb_scripted);
4546	}
4547
4548	free_calc_stats(nva, ARRAY_SIZE(names));
4549}
4550
4551static void
4552print_iostat_latency(iostat_cbdata_t *cb, nvlist_t *oldnv,
4553    nvlist_t *newnv)
4554{
4555	int i;
4556	uint64_t val;
4557	const char *names[] = {
4558		ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
4559		ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
4560		ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
4561		ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
4562		ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
4563		ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
4564		ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
4565		ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
4566		ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
4567		ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
4568	};
4569	struct stat_array *nva;
4570
4571	unsigned int column_width = default_column_width(cb, IOS_LATENCY);
4572	enum zfs_nicenum_format format;
4573
4574	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), oldnv, newnv);
4575
4576	if (cb->cb_literal)
4577		format = ZFS_NICENUM_RAWTIME;
4578	else
4579		format = ZFS_NICENUM_TIME;
4580
4581	/* Print our avg latencies on the line */
4582	for (i = 0; i < ARRAY_SIZE(names); i++) {
4583		/* Compute average latency for a latency histo */
4584		val = single_histo_average(nva[i].data, nva[i].count);
4585		print_one_stat(val, format, column_width, cb->cb_scripted);
4586	}
4587	free_calc_stats(nva, ARRAY_SIZE(names));
4588}
4589
4590/*
4591 * Print default statistics (capacity/operations/bandwidth)
4592 */
4593static void
4594print_iostat_default(vdev_stat_t *vs, iostat_cbdata_t *cb, double scale)
4595{
4596	unsigned int column_width = default_column_width(cb, IOS_DEFAULT);
4597	enum zfs_nicenum_format format;
4598	char na;	/* char to print for "not applicable" values */
4599
4600	if (cb->cb_literal) {
4601		format = ZFS_NICENUM_RAW;
4602		na = '0';
4603	} else {
4604		format = ZFS_NICENUM_1024;
4605		na = '-';
4606	}
4607
4608	/* only toplevel vdevs have capacity stats */
4609	if (vs->vs_space == 0) {
4610		if (cb->cb_scripted)
4611			printf("\t%c\t%c", na, na);
4612		else
4613			printf("  %*c  %*c", column_width, na, column_width,
4614			    na);
4615	} else {
4616		print_one_stat(vs->vs_alloc, format, column_width,
4617		    cb->cb_scripted);
4618		print_one_stat(vs->vs_space - vs->vs_alloc, format,
4619		    column_width, cb->cb_scripted);
4620	}
4621
4622	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_READ] * scale),
4623	    format, column_width, cb->cb_scripted);
4624	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_WRITE] * scale),
4625	    format, column_width, cb->cb_scripted);
4626	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_READ] * scale),
4627	    format, column_width, cb->cb_scripted);
4628	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_WRITE] * scale),
4629	    format, column_width, cb->cb_scripted);
4630}
4631
4632static const char *class_name[] = {
4633	VDEV_ALLOC_BIAS_DEDUP,
4634	VDEV_ALLOC_BIAS_SPECIAL,
4635	VDEV_ALLOC_CLASS_LOGS
4636};
4637
4638/*
4639 * Print out all the statistics for the given vdev.  This can either be the
4640 * toplevel configuration, or called recursively.  If 'name' is NULL, then this
4641 * is a verbose output, and we don't want to display the toplevel pool stats.
4642 *
4643 * Returns the number of stat lines printed.
4644 */
4645static unsigned int
4646print_vdev_stats(zpool_handle_t *zhp, const char *name, nvlist_t *oldnv,
4647    nvlist_t *newnv, iostat_cbdata_t *cb, int depth)
4648{
4649	nvlist_t **oldchild, **newchild;
4650	uint_t c, children, oldchildren;
4651	vdev_stat_t *oldvs, *newvs, *calcvs;
4652	vdev_stat_t zerovs = { 0 };
4653	char *vname;
4654	int i;
4655	int ret = 0;
4656	uint64_t tdelta;
4657	double scale;
4658
4659	if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
4660		return (ret);
4661
4662	calcvs = safe_malloc(sizeof (*calcvs));
4663
4664	if (oldnv != NULL) {
4665		verify(nvlist_lookup_uint64_array(oldnv,
4666		    ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&oldvs, &c) == 0);
4667	} else {
4668		oldvs = &zerovs;
4669	}
4670
4671	/* Do we only want to see a specific vdev? */
4672	for (i = 0; i < cb->cb_vdev_names_count; i++) {
4673		/* Yes we do.  Is this the vdev? */
4674		if (strcmp(name, cb->cb_vdev_names[i]) == 0) {
4675			/*
4676			 * This is our vdev.  Since it is the only vdev we
4677			 * will be displaying, make depth = 0 so that it
4678			 * doesn't get indented.
4679			 */
4680			depth = 0;
4681			break;
4682		}
4683	}
4684
4685	if (cb->cb_vdev_names_count && (i == cb->cb_vdev_names_count)) {
4686		/* Couldn't match the name */
4687		goto children;
4688	}
4689
4690
4691	verify(nvlist_lookup_uint64_array(newnv, ZPOOL_CONFIG_VDEV_STATS,
4692	    (uint64_t **)&newvs, &c) == 0);
4693
4694	/*
4695	 * Print the vdev name unless it's is a histogram.  Histograms
4696	 * display the vdev name in the header itself.
4697	 */
4698	if (!(cb->cb_flags & IOS_ANYHISTO_M)) {
4699		if (cb->cb_scripted) {
4700			printf("%s", name);
4701		} else {
4702			if (strlen(name) + depth > cb->cb_namewidth)
4703				(void) printf("%*s%s", depth, "", name);
4704			else
4705				(void) printf("%*s%s%*s", depth, "", name,
4706				    (int)(cb->cb_namewidth - strlen(name) -
4707				    depth), "");
4708		}
4709	}
4710
4711	/* Calculate our scaling factor */
4712	tdelta = newvs->vs_timestamp - oldvs->vs_timestamp;
4713	if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_ANYHISTO_M)) {
4714		/*
4715		 * If we specify printing histograms with no time interval, then
4716		 * print the histogram numbers over the entire lifetime of the
4717		 * vdev.
4718		 */
4719		scale = 1;
4720	} else {
4721		if (tdelta == 0)
4722			scale = 1.0;
4723		else
4724			scale = (double)NANOSEC / tdelta;
4725	}
4726
4727	if (cb->cb_flags & IOS_DEFAULT_M) {
4728		calc_default_iostats(oldvs, newvs, calcvs);
4729		print_iostat_default(calcvs, cb, scale);
4730	}
4731	if (cb->cb_flags & IOS_LATENCY_M)
4732		print_iostat_latency(cb, oldnv, newnv);
4733	if (cb->cb_flags & IOS_QUEUES_M)
4734		print_iostat_queues(cb, oldnv, newnv);
4735	if (cb->cb_flags & IOS_ANYHISTO_M) {
4736		printf("\n");
4737		print_iostat_histos(cb, oldnv, newnv, scale, name);
4738	}
4739
4740	if (cb->vcdl != NULL) {
4741		char *path;
4742		if (nvlist_lookup_string(newnv, ZPOOL_CONFIG_PATH,
4743		    &path) == 0) {
4744			printf("  ");
4745			zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
4746		}
4747	}
4748
4749	if (!(cb->cb_flags & IOS_ANYHISTO_M))
4750		printf("\n");
4751
4752	ret++;
4753
4754children:
4755
4756	free(calcvs);
4757
4758	if (!cb->cb_verbose)
4759		return (ret);
4760
4761	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_CHILDREN,
4762	    &newchild, &children) != 0)
4763		return (ret);
4764
4765	if (oldnv) {
4766		if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_CHILDREN,
4767		    &oldchild, &oldchildren) != 0)
4768			return (ret);
4769
4770		children = MIN(oldchildren, children);
4771	}
4772
4773	/*
4774	 * print normal top-level devices
4775	 */
4776	for (c = 0; c < children; c++) {
4777		uint64_t ishole = B_FALSE, islog = B_FALSE;
4778
4779		(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_HOLE,
4780		    &ishole);
4781
4782		(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_LOG,
4783		    &islog);
4784
4785		if (ishole || islog)
4786			continue;
4787
4788		if (nvlist_exists(newchild[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
4789			continue;
4790
4791		vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
4792		    cb->cb_name_flags);
4793		ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL,
4794		    newchild[c], cb, depth + 2);
4795		free(vname);
4796	}
4797
4798	/*
4799	 * print all other top-level devices
4800	 */
4801	for (uint_t n = 0; n < 3; n++) {
4802		boolean_t printed = B_FALSE;
4803
4804		for (c = 0; c < children; c++) {
4805			uint64_t islog = B_FALSE;
4806			char *bias = NULL;
4807			char *type = NULL;
4808
4809			(void) nvlist_lookup_uint64(newchild[c],
4810			    ZPOOL_CONFIG_IS_LOG, &islog);
4811			if (islog) {
4812				bias = VDEV_ALLOC_CLASS_LOGS;
4813			} else {
4814				(void) nvlist_lookup_string(newchild[c],
4815				    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
4816				(void) nvlist_lookup_string(newchild[c],
4817				    ZPOOL_CONFIG_TYPE, &type);
4818			}
4819			if (bias == NULL || strcmp(bias, class_name[n]) != 0)
4820				continue;
4821			if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
4822				continue;
4823
4824			if (!printed) {
4825				if ((!(cb->cb_flags & IOS_ANYHISTO_M)) &&
4826				    !cb->cb_scripted && !cb->cb_vdev_names) {
4827					print_iostat_dashes(cb, 0,
4828					    class_name[n]);
4829				}
4830				printf("\n");
4831				printed = B_TRUE;
4832			}
4833
4834			vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
4835			    cb->cb_name_flags);
4836			ret += print_vdev_stats(zhp, vname, oldnv ?
4837			    oldchild[c] : NULL, newchild[c], cb, depth + 2);
4838			free(vname);
4839		}
4840	}
4841
4842	/*
4843	 * Include level 2 ARC devices in iostat output
4844	 */
4845	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_L2CACHE,
4846	    &newchild, &children) != 0)
4847		return (ret);
4848
4849	if (oldnv) {
4850		if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_L2CACHE,
4851		    &oldchild, &oldchildren) != 0)
4852			return (ret);
4853
4854		children = MIN(oldchildren, children);
4855	}
4856
4857	if (children > 0) {
4858		if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted &&
4859		    !cb->cb_vdev_names) {
4860			print_iostat_dashes(cb, 0, "cache");
4861		}
4862		printf("\n");
4863
4864		for (c = 0; c < children; c++) {
4865			vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
4866			    cb->cb_name_flags);
4867			ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c]
4868			    : NULL, newchild[c], cb, depth + 2);
4869			free(vname);
4870		}
4871	}
4872
4873	return (ret);
4874}
4875
4876static int
4877refresh_iostat(zpool_handle_t *zhp, void *data)
4878{
4879	iostat_cbdata_t *cb = data;
4880	boolean_t missing;
4881
4882	/*
4883	 * If the pool has disappeared, remove it from the list and continue.
4884	 */
4885	if (zpool_refresh_stats(zhp, &missing) != 0)
4886		return (-1);
4887
4888	if (missing)
4889		pool_list_remove(cb->cb_list, zhp);
4890
4891	return (0);
4892}
4893
4894/*
4895 * Callback to print out the iostats for the given pool.
4896 */
4897static int
4898print_iostat(zpool_handle_t *zhp, void *data)
4899{
4900	iostat_cbdata_t *cb = data;
4901	nvlist_t *oldconfig, *newconfig;
4902	nvlist_t *oldnvroot, *newnvroot;
4903	int ret;
4904
4905	newconfig = zpool_get_config(zhp, &oldconfig);
4906
4907	if (cb->cb_iteration == 1)
4908		oldconfig = NULL;
4909
4910	verify(nvlist_lookup_nvlist(newconfig, ZPOOL_CONFIG_VDEV_TREE,
4911	    &newnvroot) == 0);
4912
4913	if (oldconfig == NULL)
4914		oldnvroot = NULL;
4915	else
4916		verify(nvlist_lookup_nvlist(oldconfig, ZPOOL_CONFIG_VDEV_TREE,
4917		    &oldnvroot) == 0);
4918
4919	ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot,
4920	    cb, 0);
4921	if ((ret != 0) && !(cb->cb_flags & IOS_ANYHISTO_M) &&
4922	    !cb->cb_scripted && cb->cb_verbose && !cb->cb_vdev_names_count) {
4923		print_iostat_separator(cb);
4924		if (cb->vcdl != NULL) {
4925			print_cmd_columns(cb->vcdl, 1);
4926		}
4927		printf("\n");
4928	}
4929
4930	return (ret);
4931}
4932
4933static int
4934get_columns(void)
4935{
4936	struct winsize ws;
4937	int columns = 80;
4938	int error;
4939
4940	if (isatty(STDOUT_FILENO)) {
4941		error = ioctl(STDOUT_FILENO, TIOCGWINSZ, &ws);
4942		if (error == 0)
4943			columns = ws.ws_col;
4944	} else {
4945		columns = 999;
4946	}
4947
4948	return (columns);
4949}
4950
4951/*
4952 * Return the required length of the pool/vdev name column.  The minimum
4953 * allowed width and output formatting flags must be provided.
4954 */
4955static int
4956get_namewidth(zpool_handle_t *zhp, int min_width, int flags, boolean_t verbose)
4957{
4958	nvlist_t *config, *nvroot;
4959	int width = min_width;
4960
4961	if ((config = zpool_get_config(zhp, NULL)) != NULL) {
4962		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4963		    &nvroot) == 0);
4964		unsigned int poolname_len = strlen(zpool_get_name(zhp));
4965		if (verbose == B_FALSE) {
4966			width = MAX(poolname_len, min_width);
4967		} else {
4968			width = MAX(poolname_len,
4969			    max_width(zhp, nvroot, 0, min_width, flags));
4970		}
4971	}
4972
4973	return (width);
4974}
4975
4976/*
4977 * Parse the input string, get the 'interval' and 'count' value if there is one.
4978 */
4979static void
4980get_interval_count(int *argcp, char **argv, float *iv,
4981    unsigned long *cnt)
4982{
4983	float interval = 0;
4984	unsigned long count = 0;
4985	int argc = *argcp;
4986
4987	/*
4988	 * Determine if the last argument is an integer or a pool name
4989	 */
4990	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
4991		char *end;
4992
4993		errno = 0;
4994		interval = strtof(argv[argc - 1], &end);
4995
4996		if (*end == '\0' && errno == 0) {
4997			if (interval == 0) {
4998				(void) fprintf(stderr, gettext(
4999				    "interval cannot be zero\n"));
5000				usage(B_FALSE);
5001			}
5002			/*
5003			 * Ignore the last parameter
5004			 */
5005			argc--;
5006		} else {
5007			/*
5008			 * If this is not a valid number, just plow on.  The
5009			 * user will get a more informative error message later
5010			 * on.
5011			 */
5012			interval = 0;
5013		}
5014	}
5015
5016	/*
5017	 * If the last argument is also an integer, then we have both a count
5018	 * and an interval.
5019	 */
5020	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
5021		char *end;
5022
5023		errno = 0;
5024		count = interval;
5025		interval = strtof(argv[argc - 1], &end);
5026
5027		if (*end == '\0' && errno == 0) {
5028			if (interval == 0) {
5029				(void) fprintf(stderr, gettext(
5030				    "interval cannot be zero\n"));
5031				usage(B_FALSE);
5032			}
5033
5034			/*
5035			 * Ignore the last parameter
5036			 */
5037			argc--;
5038		} else {
5039			interval = 0;
5040		}
5041	}
5042
5043	*iv = interval;
5044	*cnt = count;
5045	*argcp = argc;
5046}
5047
5048static void
5049get_timestamp_arg(char c)
5050{
5051	if (c == 'u')
5052		timestamp_fmt = UDATE;
5053	else if (c == 'd')
5054		timestamp_fmt = DDATE;
5055	else
5056		usage(B_FALSE);
5057}
5058
5059/*
5060 * Return stat flags that are supported by all pools by both the module and
5061 * zpool iostat.  "*data" should be initialized to all 0xFFs before running.
5062 * It will get ANDed down until only the flags that are supported on all pools
5063 * remain.
5064 */
5065static int
5066get_stat_flags_cb(zpool_handle_t *zhp, void *data)
5067{
5068	uint64_t *mask = data;
5069	nvlist_t *config, *nvroot, *nvx;
5070	uint64_t flags = 0;
5071	int i, j;
5072
5073	config = zpool_get_config(zhp, NULL);
5074	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
5075	    &nvroot) == 0);
5076
5077	/* Default stats are always supported, but for completeness.. */
5078	if (nvlist_exists(nvroot, ZPOOL_CONFIG_VDEV_STATS))
5079		flags |= IOS_DEFAULT_M;
5080
5081	/* Get our extended stats nvlist from the main list */
5082	if (nvlist_lookup_nvlist(nvroot, ZPOOL_CONFIG_VDEV_STATS_EX,
5083	    &nvx) != 0) {
5084		/*
5085		 * No extended stats; they're probably running an older
5086		 * module.  No big deal, we support that too.
5087		 */
5088		goto end;
5089	}
5090
5091	/* For each extended stat, make sure all its nvpairs are supported */
5092	for (j = 0; j < ARRAY_SIZE(vsx_type_to_nvlist); j++) {
5093		if (!vsx_type_to_nvlist[j][0])
5094			continue;
5095
5096		/* Start off by assuming the flag is supported, then check */
5097		flags |= (1ULL << j);
5098		for (i = 0; vsx_type_to_nvlist[j][i]; i++) {
5099			if (!nvlist_exists(nvx, vsx_type_to_nvlist[j][i])) {
5100				/* flag isn't supported */
5101				flags = flags & ~(1ULL  << j);
5102				break;
5103			}
5104		}
5105	}
5106end:
5107	*mask = *mask & flags;
5108	return (0);
5109}
5110
5111/*
5112 * Return a bitmask of stats that are supported on all pools by both the module
5113 * and zpool iostat.
5114 */
5115static uint64_t
5116get_stat_flags(zpool_list_t *list)
5117{
5118	uint64_t mask = -1;
5119
5120	/*
5121	 * get_stat_flags_cb() will lop off bits from "mask" until only the
5122	 * flags that are supported on all pools remain.
5123	 */
5124	pool_list_iter(list, B_FALSE, get_stat_flags_cb, &mask);
5125	return (mask);
5126}
5127
5128/*
5129 * Return 1 if cb_data->cb_vdev_names[0] is this vdev's name, 0 otherwise.
5130 */
5131static int
5132is_vdev_cb(zpool_handle_t *zhp, nvlist_t *nv, void *cb_data)
5133{
5134	iostat_cbdata_t *cb = cb_data;
5135	char *name = NULL;
5136	int ret = 0;
5137
5138	name = zpool_vdev_name(g_zfs, zhp, nv, cb->cb_name_flags);
5139
5140	if (strcmp(name, cb->cb_vdev_names[0]) == 0)
5141		ret = 1; /* match */
5142	free(name);
5143
5144	return (ret);
5145}
5146
5147/*
5148 * Returns 1 if cb_data->cb_vdev_names[0] is a vdev name, 0 otherwise.
5149 */
5150static int
5151is_vdev(zpool_handle_t *zhp, void *cb_data)
5152{
5153	return (for_each_vdev(zhp, is_vdev_cb, cb_data));
5154}
5155
5156/*
5157 * Check if vdevs are in a pool
5158 *
5159 * Return 1 if all argv[] strings are vdev names in pool "pool_name". Otherwise
5160 * return 0.  If pool_name is NULL, then search all pools.
5161 */
5162static int
5163are_vdevs_in_pool(int argc, char **argv, char *pool_name,
5164    iostat_cbdata_t *cb)
5165{
5166	char **tmp_name;
5167	int ret = 0;
5168	int i;
5169	int pool_count = 0;
5170
5171	if ((argc == 0) || !*argv)
5172		return (0);
5173
5174	if (pool_name)
5175		pool_count = 1;
5176
5177	/* Temporarily hijack cb_vdev_names for a second... */
5178	tmp_name = cb->cb_vdev_names;
5179
5180	/* Go though our list of prospective vdev names */
5181	for (i = 0; i < argc; i++) {
5182		cb->cb_vdev_names = argv + i;
5183
5184		/* Is this name a vdev in our pools? */
5185		ret = for_each_pool(pool_count, &pool_name, B_TRUE, NULL,
5186		    B_FALSE, is_vdev, cb);
5187		if (!ret) {
5188			/* No match */
5189			break;
5190		}
5191	}
5192
5193	cb->cb_vdev_names = tmp_name;
5194
5195	return (ret);
5196}
5197
5198static int
5199is_pool_cb(zpool_handle_t *zhp, void *data)
5200{
5201	char *name = data;
5202	if (strcmp(name, zpool_get_name(zhp)) == 0)
5203		return (1);
5204
5205	return (0);
5206}
5207
5208/*
5209 * Do we have a pool named *name?  If so, return 1, otherwise 0.
5210 */
5211static int
5212is_pool(char *name)
5213{
5214	return (for_each_pool(0, NULL, B_TRUE, NULL, B_FALSE, is_pool_cb,
5215	    name));
5216}
5217
5218/* Are all our argv[] strings pool names?  If so return 1, 0 otherwise. */
5219static int
5220are_all_pools(int argc, char **argv)
5221{
5222	if ((argc == 0) || !*argv)
5223		return (0);
5224
5225	while (--argc >= 0)
5226		if (!is_pool(argv[argc]))
5227			return (0);
5228
5229	return (1);
5230}
5231
5232/*
5233 * Helper function to print out vdev/pool names we can't resolve.  Used for an
5234 * error message.
5235 */
5236static void
5237error_list_unresolved_vdevs(int argc, char **argv, char *pool_name,
5238    iostat_cbdata_t *cb)
5239{
5240	int i;
5241	char *name;
5242	char *str;
5243	for (i = 0; i < argc; i++) {
5244		name = argv[i];
5245
5246		if (is_pool(name))
5247			str = gettext("pool");
5248		else if (are_vdevs_in_pool(1, &name, pool_name, cb))
5249			str = gettext("vdev in this pool");
5250		else if (are_vdevs_in_pool(1, &name, NULL, cb))
5251			str = gettext("vdev in another pool");
5252		else
5253			str = gettext("unknown");
5254
5255		fprintf(stderr, "\t%s (%s)\n", name, str);
5256	}
5257}
5258
5259/*
5260 * Same as get_interval_count(), but with additional checks to not misinterpret
5261 * guids as interval/count values.  Assumes VDEV_NAME_GUID is set in
5262 * cb.cb_name_flags.
5263 */
5264static void
5265get_interval_count_filter_guids(int *argc, char **argv, float *interval,
5266    unsigned long *count, iostat_cbdata_t *cb)
5267{
5268	char **tmpargv = argv;
5269	int argc_for_interval = 0;
5270
5271	/* Is the last arg an interval value?  Or a guid? */
5272	if (*argc >= 1 && !are_vdevs_in_pool(1, &argv[*argc - 1], NULL, cb)) {
5273		/*
5274		 * The last arg is not a guid, so it's probably an
5275		 * interval value.
5276		 */
5277		argc_for_interval++;
5278
5279		if (*argc >= 2 &&
5280		    !are_vdevs_in_pool(1, &argv[*argc - 2], NULL, cb)) {
5281			/*
5282			 * The 2nd to last arg is not a guid, so it's probably
5283			 * an interval value.
5284			 */
5285			argc_for_interval++;
5286		}
5287	}
5288
5289	/* Point to our list of possible intervals */
5290	tmpargv = &argv[*argc - argc_for_interval];
5291
5292	*argc = *argc - argc_for_interval;
5293	get_interval_count(&argc_for_interval, tmpargv,
5294	    interval, count);
5295}
5296
5297/*
5298 * Floating point sleep().  Allows you to pass in a floating point value for
5299 * seconds.
5300 */
5301static void
5302fsleep(float sec)
5303{
5304	struct timespec req;
5305	req.tv_sec = floor(sec);
5306	req.tv_nsec = (sec - (float)req.tv_sec) * NANOSEC;
5307	nanosleep(&req, NULL);
5308}
5309
5310/*
5311 * Terminal height, in rows. Returns -1 if stdout is not connected to a TTY or
5312 * if we were unable to determine its size.
5313 */
5314static int
5315terminal_height(void)
5316{
5317	struct winsize win;
5318
5319	if (isatty(STDOUT_FILENO) == 0)
5320		return (-1);
5321
5322	if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) != -1 && win.ws_row > 0)
5323		return (win.ws_row);
5324
5325	return (-1);
5326}
5327
5328/*
5329 * Run one of the zpool status/iostat -c scripts with the help (-h) option and
5330 * print the result.
5331 *
5332 * name:	Short name of the script ('iostat').
5333 * path:	Full path to the script ('/usr/local/etc/zfs/zpool.d/iostat');
5334 */
5335static void
5336print_zpool_script_help(char *name, char *path)
5337{
5338	char *argv[] = {path, "-h", NULL};
5339	char **lines = NULL;
5340	int lines_cnt = 0;
5341	int rc;
5342
5343	rc = libzfs_run_process_get_stdout_nopath(path, argv, NULL, &lines,
5344	    &lines_cnt);
5345	if (rc != 0 || lines == NULL || lines_cnt <= 0) {
5346		if (lines != NULL)
5347			libzfs_free_str_array(lines, lines_cnt);
5348		return;
5349	}
5350
5351	for (int i = 0; i < lines_cnt; i++)
5352		if (!is_blank_str(lines[i]))
5353			printf("  %-14s  %s\n", name, lines[i]);
5354
5355	libzfs_free_str_array(lines, lines_cnt);
5356}
5357
5358/*
5359 * Go though the zpool status/iostat -c scripts in the user's path, run their
5360 * help option (-h), and print out the results.
5361 */
5362static void
5363print_zpool_dir_scripts(char *dirpath)
5364{
5365	DIR *dir;
5366	struct dirent *ent;
5367	char fullpath[MAXPATHLEN];
5368	struct stat dir_stat;
5369
5370	if ((dir = opendir(dirpath)) != NULL) {
5371		/* print all the files and directories within directory */
5372		while ((ent = readdir(dir)) != NULL) {
5373			sprintf(fullpath, "%s/%s", dirpath, ent->d_name);
5374
5375			/* Print the scripts */
5376			if (stat(fullpath, &dir_stat) == 0)
5377				if (dir_stat.st_mode & S_IXUSR &&
5378				    S_ISREG(dir_stat.st_mode))
5379					print_zpool_script_help(ent->d_name,
5380					    fullpath);
5381		}
5382		closedir(dir);
5383	}
5384}
5385
5386/*
5387 * Print out help text for all zpool status/iostat -c scripts.
5388 */
5389static void
5390print_zpool_script_list(char *subcommand)
5391{
5392	char *dir, *sp;
5393
5394	printf(gettext("Available 'zpool %s -c' commands:\n"), subcommand);
5395
5396	sp = zpool_get_cmd_search_path();
5397	if (sp == NULL)
5398		return;
5399
5400	dir = strtok(sp, ":");
5401	while (dir != NULL) {
5402		print_zpool_dir_scripts(dir);
5403		dir = strtok(NULL, ":");
5404	}
5405
5406	free(sp);
5407}
5408
5409/*
5410 * Set the minimum pool/vdev name column width.  The width must be at least 10,
5411 * but may be as large as the column width - 42 so it still fits on one line.
5412 * NOTE: 42 is the width of the default capacity/operations/bandwidth output
5413 */
5414static int
5415get_namewidth_iostat(zpool_handle_t *zhp, void *data)
5416{
5417	iostat_cbdata_t *cb = data;
5418	int width, available_width;
5419
5420	/*
5421	 * get_namewidth() returns the maximum width of any name in that column
5422	 * for any pool/vdev/device line that will be output.
5423	 */
5424	width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags,
5425	    cb->cb_verbose);
5426
5427	/*
5428	 * The width we are calculating is the width of the header and also the
5429	 * padding width for names that are less than maximum width.  The stats
5430	 * take up 42 characters, so the width available for names is:
5431	 */
5432	available_width = get_columns() - 42;
5433
5434	/*
5435	 * If the maximum width fits on a screen, then great!  Make everything
5436	 * line up by justifying all lines to the same width.  If that max
5437	 * width is larger than what's available, the name plus stats won't fit
5438	 * on one line, and justifying to that width would cause every line to
5439	 * wrap on the screen.  We only want lines with long names to wrap.
5440	 * Limit the padding to what won't wrap.
5441	 */
5442	if (width > available_width)
5443		width = available_width;
5444
5445	/*
5446	 * And regardless of whatever the screen width is (get_columns can
5447	 * return 0 if the width is not known or less than 42 for a narrow
5448	 * terminal) have the width be a minimum of 10.
5449	 */
5450	if (width < 10)
5451		width = 10;
5452
5453	/* Save the calculated width */
5454	cb->cb_namewidth = width;
5455
5456	return (0);
5457}
5458
5459/*
5460 * zpool iostat [[-c [script1,script2,...]] [-lq]|[-rw]] [-ghHLpPvy] [-n name]
5461 *              [-T d|u] [[ pool ...]|[pool vdev ...]|[vdev ...]]
5462 *              [interval [count]]
5463 *
5464 *	-c CMD  For each vdev, run command CMD
5465 *	-g	Display guid for individual vdev name.
5466 *	-L	Follow links when resolving vdev path name.
5467 *	-P	Display full path for vdev name.
5468 *	-v	Display statistics for individual vdevs
5469 *	-h	Display help
5470 *	-p	Display values in parsable (exact) format.
5471 *	-H	Scripted mode.  Don't display headers, and separate properties
5472 *		by a single tab.
5473 *	-l	Display average latency
5474 *	-q	Display queue depths
5475 *	-w	Display latency histograms
5476 *	-r	Display request size histogram
5477 *	-T	Display a timestamp in date(1) or Unix format
5478 *	-n	Only print headers once
5479 *
5480 * This command can be tricky because we want to be able to deal with pool
5481 * creation/destruction as well as vdev configuration changes.  The bulk of this
5482 * processing is handled by the pool_list_* routines in zpool_iter.c.  We rely
5483 * on pool_list_update() to detect the addition of new pools.  Configuration
5484 * changes are all handled within libzfs.
5485 */
5486int
5487zpool_do_iostat(int argc, char **argv)
5488{
5489	int c;
5490	int ret;
5491	int npools;
5492	float interval = 0;
5493	unsigned long count = 0;
5494	int winheight = 24;
5495	zpool_list_t *list;
5496	boolean_t verbose = B_FALSE;
5497	boolean_t latency = B_FALSE, l_histo = B_FALSE, rq_histo = B_FALSE;
5498	boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE;
5499	boolean_t omit_since_boot = B_FALSE;
5500	boolean_t guid = B_FALSE;
5501	boolean_t follow_links = B_FALSE;
5502	boolean_t full_name = B_FALSE;
5503	boolean_t headers_once = B_FALSE;
5504	iostat_cbdata_t cb = { 0 };
5505	char *cmd = NULL;
5506
5507	/* Used for printing error message */
5508	const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q',
5509	    [IOS_L_HISTO] = 'w', [IOS_RQ_HISTO] = 'r'};
5510
5511	uint64_t unsupported_flags;
5512
5513	/* check options */
5514	while ((c = getopt(argc, argv, "c:gLPT:vyhplqrwnH")) != -1) {
5515		switch (c) {
5516		case 'c':
5517			if (cmd != NULL) {
5518				fprintf(stderr,
5519				    gettext("Can't set -c flag twice\n"));
5520				exit(1);
5521			}
5522
5523			if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
5524			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
5525				fprintf(stderr, gettext(
5526				    "Can't run -c, disabled by "
5527				    "ZPOOL_SCRIPTS_ENABLED.\n"));
5528				exit(1);
5529			}
5530
5531			if ((getuid() <= 0 || geteuid() <= 0) &&
5532			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
5533				fprintf(stderr, gettext(
5534				    "Can't run -c with root privileges "
5535				    "unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
5536				exit(1);
5537			}
5538			cmd = optarg;
5539			verbose = B_TRUE;
5540			break;
5541		case 'g':
5542			guid = B_TRUE;
5543			break;
5544		case 'L':
5545			follow_links = B_TRUE;
5546			break;
5547		case 'P':
5548			full_name = B_TRUE;
5549			break;
5550		case 'T':
5551			get_timestamp_arg(*optarg);
5552			break;
5553		case 'v':
5554			verbose = B_TRUE;
5555			break;
5556		case 'p':
5557			parsable = B_TRUE;
5558			break;
5559		case 'l':
5560			latency = B_TRUE;
5561			break;
5562		case 'q':
5563			queues = B_TRUE;
5564			break;
5565		case 'H':
5566			scripted = B_TRUE;
5567			break;
5568		case 'w':
5569			l_histo = B_TRUE;
5570			break;
5571		case 'r':
5572			rq_histo = B_TRUE;
5573			break;
5574		case 'y':
5575			omit_since_boot = B_TRUE;
5576			break;
5577		case 'n':
5578			headers_once = B_TRUE;
5579			break;
5580		case 'h':
5581			usage(B_FALSE);
5582			break;
5583		case '?':
5584			if (optopt == 'c') {
5585				print_zpool_script_list("iostat");
5586				exit(0);
5587			} else {
5588				fprintf(stderr,
5589				    gettext("invalid option '%c'\n"), optopt);
5590			}
5591			usage(B_FALSE);
5592		}
5593	}
5594
5595	argc -= optind;
5596	argv += optind;
5597
5598	cb.cb_literal = parsable;
5599	cb.cb_scripted = scripted;
5600
5601	if (guid)
5602		cb.cb_name_flags |= VDEV_NAME_GUID;
5603	if (follow_links)
5604		cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS;
5605	if (full_name)
5606		cb.cb_name_flags |= VDEV_NAME_PATH;
5607	cb.cb_iteration = 0;
5608	cb.cb_namewidth = 0;
5609	cb.cb_verbose = verbose;
5610
5611	/* Get our interval and count values (if any) */
5612	if (guid) {
5613		get_interval_count_filter_guids(&argc, argv, &interval,
5614		    &count, &cb);
5615	} else {
5616		get_interval_count(&argc, argv, &interval, &count);
5617	}
5618
5619	if (argc == 0) {
5620		/* No args, so just print the defaults. */
5621	} else if (are_all_pools(argc, argv)) {
5622		/* All the args are pool names */
5623	} else if (are_vdevs_in_pool(argc, argv, NULL, &cb)) {
5624		/* All the args are vdevs */
5625		cb.cb_vdev_names = argv;
5626		cb.cb_vdev_names_count = argc;
5627		argc = 0; /* No pools to process */
5628	} else if (are_all_pools(1, argv)) {
5629		/* The first arg is a pool name */
5630		if (are_vdevs_in_pool(argc - 1, argv + 1, argv[0], &cb)) {
5631			/* ...and the rest are vdev names */
5632			cb.cb_vdev_names = argv + 1;
5633			cb.cb_vdev_names_count = argc - 1;
5634			argc = 1; /* One pool to process */
5635		} else {
5636			fprintf(stderr, gettext("Expected either a list of "));
5637			fprintf(stderr, gettext("pools, or list of vdevs in"));
5638			fprintf(stderr, " \"%s\", ", argv[0]);
5639			fprintf(stderr, gettext("but got:\n"));
5640			error_list_unresolved_vdevs(argc - 1, argv + 1,
5641			    argv[0], &cb);
5642			fprintf(stderr, "\n");
5643			usage(B_FALSE);
5644			return (1);
5645		}
5646	} else {
5647		/*
5648		 * The args don't make sense. The first arg isn't a pool name,
5649		 * nor are all the args vdevs.
5650		 */
5651		fprintf(stderr, gettext("Unable to parse pools/vdevs list.\n"));
5652		fprintf(stderr, "\n");
5653		return (1);
5654	}
5655
5656	if (cb.cb_vdev_names_count != 0) {
5657		/*
5658		 * If user specified vdevs, it implies verbose.
5659		 */
5660		cb.cb_verbose = B_TRUE;
5661	}
5662
5663	/*
5664	 * Construct the list of all interesting pools.
5665	 */
5666	ret = 0;
5667	if ((list = pool_list_get(argc, argv, NULL, parsable, &ret)) == NULL)
5668		return (1);
5669
5670	if (pool_list_count(list) == 0 && argc != 0) {
5671		pool_list_free(list);
5672		return (1);
5673	}
5674
5675	if (pool_list_count(list) == 0 && interval == 0) {
5676		pool_list_free(list);
5677		(void) fprintf(stderr, gettext("no pools available\n"));
5678		return (1);
5679	}
5680
5681	if ((l_histo || rq_histo) && (cmd != NULL || latency || queues)) {
5682		pool_list_free(list);
5683		(void) fprintf(stderr,
5684		    gettext("[-r|-w] isn't allowed with [-c|-l|-q]\n"));
5685		usage(B_FALSE);
5686		return (1);
5687	}
5688
5689	if (l_histo && rq_histo) {
5690		pool_list_free(list);
5691		(void) fprintf(stderr,
5692		    gettext("Only one of [-r|-w] can be passed at a time\n"));
5693		usage(B_FALSE);
5694		return (1);
5695	}
5696
5697	/*
5698	 * Enter the main iostat loop.
5699	 */
5700	cb.cb_list = list;
5701
5702	if (l_histo) {
5703		/*
5704		 * Histograms tables look out of place when you try to display
5705		 * them with the other stats, so make a rule that you can only
5706		 * print histograms by themselves.
5707		 */
5708		cb.cb_flags = IOS_L_HISTO_M;
5709	} else if (rq_histo) {
5710		cb.cb_flags = IOS_RQ_HISTO_M;
5711	} else {
5712		cb.cb_flags = IOS_DEFAULT_M;
5713		if (latency)
5714			cb.cb_flags |= IOS_LATENCY_M;
5715		if (queues)
5716			cb.cb_flags |= IOS_QUEUES_M;
5717	}
5718
5719	/*
5720	 * See if the module supports all the stats we want to display.
5721	 */
5722	unsupported_flags = cb.cb_flags & ~get_stat_flags(list);
5723	if (unsupported_flags) {
5724		uint64_t f;
5725		int idx;
5726		fprintf(stderr,
5727		    gettext("The loaded zfs module doesn't support:"));
5728
5729		/* for each bit set in unsupported_flags */
5730		for (f = unsupported_flags; f; f &= ~(1ULL << idx)) {
5731			idx = lowbit64(f) - 1;
5732			fprintf(stderr, " -%c", flag_to_arg[idx]);
5733		}
5734
5735		fprintf(stderr, ".  Try running a newer module.\n");
5736		pool_list_free(list);
5737
5738		return (1);
5739	}
5740
5741	for (;;) {
5742		if ((npools = pool_list_count(list)) == 0)
5743			(void) fprintf(stderr, gettext("no pools available\n"));
5744		else {
5745			/*
5746			 * If this is the first iteration and -y was supplied
5747			 * we skip any printing.
5748			 */
5749			boolean_t skip = (omit_since_boot &&
5750			    cb.cb_iteration == 0);
5751
5752			/*
5753			 * Refresh all statistics.  This is done as an
5754			 * explicit step before calculating the maximum name
5755			 * width, so that any * configuration changes are
5756			 * properly accounted for.
5757			 */
5758			(void) pool_list_iter(list, B_FALSE, refresh_iostat,
5759			    &cb);
5760
5761			/*
5762			 * Iterate over all pools to determine the maximum width
5763			 * for the pool / device name column across all pools.
5764			 */
5765			cb.cb_namewidth = 0;
5766			(void) pool_list_iter(list, B_FALSE,
5767			    get_namewidth_iostat, &cb);
5768
5769			if (timestamp_fmt != NODATE)
5770				print_timestamp(timestamp_fmt);
5771
5772			if (cmd != NULL && cb.cb_verbose &&
5773			    !(cb.cb_flags & IOS_ANYHISTO_M)) {
5774				cb.vcdl = all_pools_for_each_vdev_run(argc,
5775				    argv, cmd, g_zfs, cb.cb_vdev_names,
5776				    cb.cb_vdev_names_count, cb.cb_name_flags);
5777			} else {
5778				cb.vcdl = NULL;
5779			}
5780
5781
5782			/*
5783			 * Check terminal size so we can print headers
5784			 * even when terminal window has its height
5785			 * changed.
5786			 */
5787			winheight = terminal_height();
5788			/*
5789			 * Are we connected to TTY? If not, headers_once
5790			 * should be true, to avoid breaking scripts.
5791			 */
5792			if (winheight < 0)
5793				headers_once = B_TRUE;
5794
5795			/*
5796			 * If it's the first time and we're not skipping it,
5797			 * or either skip or verbose mode, print the header.
5798			 *
5799			 * The histogram code explicitly prints its header on
5800			 * every vdev, so skip this for histograms.
5801			 */
5802			if (((++cb.cb_iteration == 1 && !skip) ||
5803			    (skip != verbose) ||
5804			    (!headers_once &&
5805			    (cb.cb_iteration % winheight) == 0)) &&
5806			    (!(cb.cb_flags & IOS_ANYHISTO_M)) &&
5807			    !cb.cb_scripted)
5808				print_iostat_header(&cb);
5809
5810			if (skip) {
5811				(void) fsleep(interval);
5812				continue;
5813			}
5814
5815			pool_list_iter(list, B_FALSE, print_iostat, &cb);
5816
5817			/*
5818			 * If there's more than one pool, and we're not in
5819			 * verbose mode (which prints a separator for us),
5820			 * then print a separator.
5821			 *
5822			 * In addition, if we're printing specific vdevs then
5823			 * we also want an ending separator.
5824			 */
5825			if (((npools > 1 && !verbose &&
5826			    !(cb.cb_flags & IOS_ANYHISTO_M)) ||
5827			    (!(cb.cb_flags & IOS_ANYHISTO_M) &&
5828			    cb.cb_vdev_names_count)) &&
5829			    !cb.cb_scripted) {
5830				print_iostat_separator(&cb);
5831				if (cb.vcdl != NULL)
5832					print_cmd_columns(cb.vcdl, 1);
5833				printf("\n");
5834			}
5835
5836			if (cb.vcdl != NULL)
5837				free_vdev_cmd_data_list(cb.vcdl);
5838
5839		}
5840
5841		/*
5842		 * Flush the output so that redirection to a file isn't buffered
5843		 * indefinitely.
5844		 */
5845		(void) fflush(stdout);
5846
5847		if (interval == 0)
5848			break;
5849
5850		if (count != 0 && --count == 0)
5851			break;
5852
5853		(void) fsleep(interval);
5854	}
5855
5856	pool_list_free(list);
5857
5858	return (ret);
5859}
5860
5861typedef struct list_cbdata {
5862	boolean_t	cb_verbose;
5863	int		cb_name_flags;
5864	int		cb_namewidth;
5865	boolean_t	cb_scripted;
5866	zprop_list_t	*cb_proplist;
5867	boolean_t	cb_literal;
5868} list_cbdata_t;
5869
5870
5871/*
5872 * Given a list of columns to display, output appropriate headers for each one.
5873 */
5874static void
5875print_header(list_cbdata_t *cb)
5876{
5877	zprop_list_t *pl = cb->cb_proplist;
5878	char headerbuf[ZPOOL_MAXPROPLEN];
5879	const char *header;
5880	boolean_t first = B_TRUE;
5881	boolean_t right_justify;
5882	size_t width = 0;
5883
5884	for (; pl != NULL; pl = pl->pl_next) {
5885		width = pl->pl_width;
5886		if (first && cb->cb_verbose) {
5887			/*
5888			 * Reset the width to accommodate the verbose listing
5889			 * of devices.
5890			 */
5891			width = cb->cb_namewidth;
5892		}
5893
5894		if (!first)
5895			(void) printf("  ");
5896		else
5897			first = B_FALSE;
5898
5899		right_justify = B_FALSE;
5900		if (pl->pl_prop != ZPROP_INVAL) {
5901			header = zpool_prop_column_name(pl->pl_prop);
5902			right_justify = zpool_prop_align_right(pl->pl_prop);
5903		} else {
5904			int i;
5905
5906			for (i = 0; pl->pl_user_prop[i] != '\0'; i++)
5907				headerbuf[i] = toupper(pl->pl_user_prop[i]);
5908			headerbuf[i] = '\0';
5909			header = headerbuf;
5910		}
5911
5912		if (pl->pl_next == NULL && !right_justify)
5913			(void) printf("%s", header);
5914		else if (right_justify)
5915			(void) printf("%*s", (int)width, header);
5916		else
5917			(void) printf("%-*s", (int)width, header);
5918	}
5919
5920	(void) printf("\n");
5921}
5922
5923/*
5924 * Given a pool and a list of properties, print out all the properties according
5925 * to the described layout. Used by zpool_do_list().
5926 */
5927static void
5928print_pool(zpool_handle_t *zhp, list_cbdata_t *cb)
5929{
5930	zprop_list_t *pl = cb->cb_proplist;
5931	boolean_t first = B_TRUE;
5932	char property[ZPOOL_MAXPROPLEN];
5933	char *propstr;
5934	boolean_t right_justify;
5935	size_t width;
5936
5937	for (; pl != NULL; pl = pl->pl_next) {
5938
5939		width = pl->pl_width;
5940		if (first && cb->cb_verbose) {
5941			/*
5942			 * Reset the width to accommodate the verbose listing
5943			 * of devices.
5944			 */
5945			width = cb->cb_namewidth;
5946		}
5947
5948		if (!first) {
5949			if (cb->cb_scripted)
5950				(void) printf("\t");
5951			else
5952				(void) printf("  ");
5953		} else {
5954			first = B_FALSE;
5955		}
5956
5957		right_justify = B_FALSE;
5958		if (pl->pl_prop != ZPROP_INVAL) {
5959			if (zpool_get_prop(zhp, pl->pl_prop, property,
5960			    sizeof (property), NULL, cb->cb_literal) != 0)
5961				propstr = "-";
5962			else
5963				propstr = property;
5964
5965			right_justify = zpool_prop_align_right(pl->pl_prop);
5966		} else if ((zpool_prop_feature(pl->pl_user_prop) ||
5967		    zpool_prop_unsupported(pl->pl_user_prop)) &&
5968		    zpool_prop_get_feature(zhp, pl->pl_user_prop, property,
5969		    sizeof (property)) == 0) {
5970			propstr = property;
5971		} else {
5972			propstr = "-";
5973		}
5974
5975
5976		/*
5977		 * If this is being called in scripted mode, or if this is the
5978		 * last column and it is left-justified, don't include a width
5979		 * format specifier.
5980		 */
5981		if (cb->cb_scripted || (pl->pl_next == NULL && !right_justify))
5982			(void) printf("%s", propstr);
5983		else if (right_justify)
5984			(void) printf("%*s", (int)width, propstr);
5985		else
5986			(void) printf("%-*s", (int)width, propstr);
5987	}
5988
5989	(void) printf("\n");
5990}
5991
5992static void
5993print_one_column(zpool_prop_t prop, uint64_t value, const char *str,
5994    boolean_t scripted, boolean_t valid, enum zfs_nicenum_format format)
5995{
5996	char propval[64];
5997	boolean_t fixed;
5998	size_t width = zprop_width(prop, &fixed, ZFS_TYPE_POOL);
5999
6000	switch (prop) {
6001	case ZPOOL_PROP_EXPANDSZ:
6002	case ZPOOL_PROP_CHECKPOINT:
6003	case ZPOOL_PROP_DEDUPRATIO:
6004		if (value == 0)
6005			(void) strlcpy(propval, "-", sizeof (propval));
6006		else
6007			zfs_nicenum_format(value, propval, sizeof (propval),
6008			    format);
6009		break;
6010	case ZPOOL_PROP_FRAGMENTATION:
6011		if (value == ZFS_FRAG_INVALID) {
6012			(void) strlcpy(propval, "-", sizeof (propval));
6013		} else if (format == ZFS_NICENUM_RAW) {
6014			(void) snprintf(propval, sizeof (propval), "%llu",
6015			    (unsigned long long)value);
6016		} else {
6017			(void) snprintf(propval, sizeof (propval), "%llu%%",
6018			    (unsigned long long)value);
6019		}
6020		break;
6021	case ZPOOL_PROP_CAPACITY:
6022		/* capacity value is in parts-per-10,000 (aka permyriad) */
6023		if (format == ZFS_NICENUM_RAW)
6024			(void) snprintf(propval, sizeof (propval), "%llu",
6025			    (unsigned long long)value / 100);
6026		else
6027			(void) snprintf(propval, sizeof (propval),
6028			    value < 1000 ? "%1.2f%%" : value < 10000 ?
6029			    "%2.1f%%" : "%3.0f%%", value / 100.0);
6030		break;
6031	case ZPOOL_PROP_HEALTH:
6032		width = 8;
6033		(void) strlcpy(propval, str, sizeof (propval));
6034		break;
6035	default:
6036		zfs_nicenum_format(value, propval, sizeof (propval), format);
6037	}
6038
6039	if (!valid)
6040		(void) strlcpy(propval, "-", sizeof (propval));
6041
6042	if (scripted)
6043		(void) printf("\t%s", propval);
6044	else
6045		(void) printf("  %*s", (int)width, propval);
6046}
6047
6048/*
6049 * print static default line per vdev
6050 * not compatible with '-o' <proplist> option
6051 */
6052static void
6053print_list_stats(zpool_handle_t *zhp, const char *name, nvlist_t *nv,
6054    list_cbdata_t *cb, int depth, boolean_t isspare)
6055{
6056	nvlist_t **child;
6057	vdev_stat_t *vs;
6058	uint_t c, children;
6059	char *vname;
6060	boolean_t scripted = cb->cb_scripted;
6061	uint64_t islog = B_FALSE;
6062	char *dashes = "%-*s      -      -      -        -         "
6063	    "-      -      -      -  -\n";
6064
6065	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
6066	    (uint64_t **)&vs, &c) == 0);
6067
6068	if (name != NULL) {
6069		boolean_t toplevel = (vs->vs_space != 0);
6070		uint64_t cap;
6071		enum zfs_nicenum_format format;
6072		const char *state;
6073
6074		if (cb->cb_literal)
6075			format = ZFS_NICENUM_RAW;
6076		else
6077			format = ZFS_NICENUM_1024;
6078
6079		if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
6080			return;
6081
6082		if (scripted)
6083			(void) printf("\t%s", name);
6084		else if (strlen(name) + depth > cb->cb_namewidth)
6085			(void) printf("%*s%s", depth, "", name);
6086		else
6087			(void) printf("%*s%s%*s", depth, "", name,
6088			    (int)(cb->cb_namewidth - strlen(name) - depth), "");
6089
6090		/*
6091		 * Print the properties for the individual vdevs. Some
6092		 * properties are only applicable to toplevel vdevs. The
6093		 * 'toplevel' boolean value is passed to the print_one_column()
6094		 * to indicate that the value is valid.
6095		 */
6096		print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL, scripted,
6097		    toplevel, format);
6098		print_one_column(ZPOOL_PROP_ALLOCATED, vs->vs_alloc, NULL,
6099		    scripted, toplevel, format);
6100		print_one_column(ZPOOL_PROP_FREE, vs->vs_space - vs->vs_alloc,
6101		    NULL, scripted, toplevel, format);
6102		print_one_column(ZPOOL_PROP_CHECKPOINT,
6103		    vs->vs_checkpoint_space, NULL, scripted, toplevel, format);
6104		print_one_column(ZPOOL_PROP_EXPANDSZ, vs->vs_esize, NULL,
6105		    scripted, B_TRUE, format);
6106		print_one_column(ZPOOL_PROP_FRAGMENTATION,
6107		    vs->vs_fragmentation, NULL, scripted,
6108		    (vs->vs_fragmentation != ZFS_FRAG_INVALID && toplevel),
6109		    format);
6110		cap = (vs->vs_space == 0) ? 0 :
6111		    (vs->vs_alloc * 10000 / vs->vs_space);
6112		print_one_column(ZPOOL_PROP_CAPACITY, cap, NULL,
6113		    scripted, toplevel, format);
6114		print_one_column(ZPOOL_PROP_DEDUPRATIO, 0, NULL,
6115		    scripted, toplevel, format);
6116		state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
6117		if (isspare) {
6118			if (vs->vs_aux == VDEV_AUX_SPARED)
6119				state = "INUSE";
6120			else if (vs->vs_state == VDEV_STATE_HEALTHY)
6121				state = "AVAIL";
6122		}
6123		print_one_column(ZPOOL_PROP_HEALTH, 0, state, scripted,
6124		    B_TRUE, format);
6125		(void) printf("\n");
6126	}
6127
6128	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
6129	    &child, &children) != 0)
6130		return;
6131
6132	/* list the normal vdevs first */
6133	for (c = 0; c < children; c++) {
6134		uint64_t ishole = B_FALSE;
6135
6136		if (nvlist_lookup_uint64(child[c],
6137		    ZPOOL_CONFIG_IS_HOLE, &ishole) == 0 && ishole)
6138			continue;
6139
6140		if (nvlist_lookup_uint64(child[c],
6141		    ZPOOL_CONFIG_IS_LOG, &islog) == 0 && islog)
6142			continue;
6143
6144		if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
6145			continue;
6146
6147		vname = zpool_vdev_name(g_zfs, zhp, child[c],
6148		    cb->cb_name_flags);
6149		print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE);
6150		free(vname);
6151	}
6152
6153	/* list the classes: 'logs', 'dedup', and 'special' */
6154	for (uint_t n = 0; n < 3; n++) {
6155		boolean_t printed = B_FALSE;
6156
6157		for (c = 0; c < children; c++) {
6158			char *bias = NULL;
6159			char *type = NULL;
6160
6161			if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
6162			    &islog) == 0 && islog) {
6163				bias = VDEV_ALLOC_CLASS_LOGS;
6164			} else {
6165				(void) nvlist_lookup_string(child[c],
6166				    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
6167				(void) nvlist_lookup_string(child[c],
6168				    ZPOOL_CONFIG_TYPE, &type);
6169			}
6170			if (bias == NULL || strcmp(bias, class_name[n]) != 0)
6171				continue;
6172			if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
6173				continue;
6174
6175			if (!printed) {
6176				/* LINTED E_SEC_PRINTF_VAR_FMT */
6177				(void) printf(dashes, cb->cb_namewidth,
6178				    class_name[n]);
6179				printed = B_TRUE;
6180			}
6181			vname = zpool_vdev_name(g_zfs, zhp, child[c],
6182			    cb->cb_name_flags);
6183			print_list_stats(zhp, vname, child[c], cb, depth + 2,
6184			    B_FALSE);
6185			free(vname);
6186		}
6187	}
6188
6189	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
6190	    &child, &children) == 0 && children > 0) {
6191		/* LINTED E_SEC_PRINTF_VAR_FMT */
6192		(void) printf(dashes, cb->cb_namewidth, "cache");
6193		for (c = 0; c < children; c++) {
6194			vname = zpool_vdev_name(g_zfs, zhp, child[c],
6195			    cb->cb_name_flags);
6196			print_list_stats(zhp, vname, child[c], cb, depth + 2,
6197			    B_FALSE);
6198			free(vname);
6199		}
6200	}
6201
6202	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child,
6203	    &children) == 0 && children > 0) {
6204		/* LINTED E_SEC_PRINTF_VAR_FMT */
6205		(void) printf(dashes, cb->cb_namewidth, "spare");
6206		for (c = 0; c < children; c++) {
6207			vname = zpool_vdev_name(g_zfs, zhp, child[c],
6208			    cb->cb_name_flags);
6209			print_list_stats(zhp, vname, child[c], cb, depth + 2,
6210			    B_TRUE);
6211			free(vname);
6212		}
6213	}
6214}
6215
6216/*
6217 * Generic callback function to list a pool.
6218 */
6219static int
6220list_callback(zpool_handle_t *zhp, void *data)
6221{
6222	list_cbdata_t *cbp = data;
6223
6224	print_pool(zhp, cbp);
6225
6226	if (cbp->cb_verbose) {
6227		nvlist_t *config, *nvroot;
6228
6229		config = zpool_get_config(zhp, NULL);
6230		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
6231		    &nvroot) == 0);
6232		print_list_stats(zhp, NULL, nvroot, cbp, 0, B_FALSE);
6233	}
6234
6235	return (0);
6236}
6237
6238/*
6239 * Set the minimum pool/vdev name column width.  The width must be at least 9,
6240 * but may be as large as needed.
6241 */
6242static int
6243get_namewidth_list(zpool_handle_t *zhp, void *data)
6244{
6245	list_cbdata_t *cb = data;
6246	int width;
6247
6248	width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags,
6249	    cb->cb_verbose);
6250
6251	if (width < 9)
6252		width = 9;
6253
6254	cb->cb_namewidth = width;
6255
6256	return (0);
6257}
6258
6259/*
6260 * zpool list [-gHLpP] [-o prop[,prop]*] [-T d|u] [pool] ... [interval [count]]
6261 *
6262 *	-g	Display guid for individual vdev name.
6263 *	-H	Scripted mode.  Don't display headers, and separate properties
6264 *		by a single tab.
6265 *	-L	Follow links when resolving vdev path name.
6266 *	-o	List of properties to display.  Defaults to
6267 *		"name,size,allocated,free,expandsize,fragmentation,capacity,"
6268 *		"dedupratio,health,altroot"
6269 *	-p	Display values in parsable (exact) format.
6270 *	-P	Display full path for vdev name.
6271 *	-T	Display a timestamp in date(1) or Unix format
6272 *
6273 * List all pools in the system, whether or not they're healthy.  Output space
6274 * statistics for each one, as well as health status summary.
6275 */
6276int
6277zpool_do_list(int argc, char **argv)
6278{
6279	int c;
6280	int ret = 0;
6281	list_cbdata_t cb = { 0 };
6282	static char default_props[] =
6283	    "name,size,allocated,free,checkpoint,expandsize,fragmentation,"
6284	    "capacity,dedupratio,health,altroot";
6285	char *props = default_props;
6286	float interval = 0;
6287	unsigned long count = 0;
6288	zpool_list_t *list;
6289	boolean_t first = B_TRUE;
6290
6291	/* check options */
6292	while ((c = getopt(argc, argv, ":gHLo:pPT:v")) != -1) {
6293		switch (c) {
6294		case 'g':
6295			cb.cb_name_flags |= VDEV_NAME_GUID;
6296			break;
6297		case 'H':
6298			cb.cb_scripted = B_TRUE;
6299			break;
6300		case 'L':
6301			cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS;
6302			break;
6303		case 'o':
6304			props = optarg;
6305			break;
6306		case 'P':
6307			cb.cb_name_flags |= VDEV_NAME_PATH;
6308			break;
6309		case 'p':
6310			cb.cb_literal = B_TRUE;
6311			break;
6312		case 'T':
6313			get_timestamp_arg(*optarg);
6314			break;
6315		case 'v':
6316			cb.cb_verbose = B_TRUE;
6317			cb.cb_namewidth = 8;	/* 8 until precalc is avail */
6318			break;
6319		case ':':
6320			(void) fprintf(stderr, gettext("missing argument for "
6321			    "'%c' option\n"), optopt);
6322			usage(B_FALSE);
6323			break;
6324		case '?':
6325			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6326			    optopt);
6327			usage(B_FALSE);
6328		}
6329	}
6330
6331	argc -= optind;
6332	argv += optind;
6333
6334	get_interval_count(&argc, argv, &interval, &count);
6335
6336	if (zprop_get_list(g_zfs, props, &cb.cb_proplist, ZFS_TYPE_POOL) != 0)
6337		usage(B_FALSE);
6338
6339	for (;;) {
6340		if ((list = pool_list_get(argc, argv, &cb.cb_proplist,
6341		    cb.cb_literal, &ret)) == NULL)
6342			return (1);
6343
6344		if (pool_list_count(list) == 0)
6345			break;
6346
6347		cb.cb_namewidth = 0;
6348		(void) pool_list_iter(list, B_FALSE, get_namewidth_list, &cb);
6349
6350		if (timestamp_fmt != NODATE)
6351			print_timestamp(timestamp_fmt);
6352
6353		if (!cb.cb_scripted && (first || cb.cb_verbose)) {
6354			print_header(&cb);
6355			first = B_FALSE;
6356		}
6357		ret = pool_list_iter(list, B_TRUE, list_callback, &cb);
6358
6359		if (interval == 0)
6360			break;
6361
6362		if (count != 0 && --count == 0)
6363			break;
6364
6365		pool_list_free(list);
6366		(void) fsleep(interval);
6367	}
6368
6369	if (argc == 0 && !cb.cb_scripted && pool_list_count(list) == 0) {
6370		(void) printf(gettext("no pools available\n"));
6371		ret = 0;
6372	}
6373
6374	pool_list_free(list);
6375	zprop_free_list(cb.cb_proplist);
6376	return (ret);
6377}
6378
6379static int
6380zpool_do_attach_or_replace(int argc, char **argv, int replacing)
6381{
6382	boolean_t force = B_FALSE;
6383	boolean_t rebuild = B_FALSE;
6384	boolean_t wait = B_FALSE;
6385	int c;
6386	nvlist_t *nvroot;
6387	char *poolname, *old_disk, *new_disk;
6388	zpool_handle_t *zhp;
6389	nvlist_t *props = NULL;
6390	char *propval;
6391	int ret;
6392
6393	/* check options */
6394	while ((c = getopt(argc, argv, "fo:sw")) != -1) {
6395		switch (c) {
6396		case 'f':
6397			force = B_TRUE;
6398			break;
6399		case 'o':
6400			if ((propval = strchr(optarg, '=')) == NULL) {
6401				(void) fprintf(stderr, gettext("missing "
6402				    "'=' for -o option\n"));
6403				usage(B_FALSE);
6404			}
6405			*propval = '\0';
6406			propval++;
6407
6408			if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) ||
6409			    (add_prop_list(optarg, propval, &props, B_TRUE)))
6410				usage(B_FALSE);
6411			break;
6412		case 's':
6413			rebuild = B_TRUE;
6414			break;
6415		case 'w':
6416			wait = B_TRUE;
6417			break;
6418		case '?':
6419			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6420			    optopt);
6421			usage(B_FALSE);
6422		}
6423	}
6424
6425	argc -= optind;
6426	argv += optind;
6427
6428	/* get pool name and check number of arguments */
6429	if (argc < 1) {
6430		(void) fprintf(stderr, gettext("missing pool name argument\n"));
6431		usage(B_FALSE);
6432	}
6433
6434	poolname = argv[0];
6435
6436	if (argc < 2) {
6437		(void) fprintf(stderr,
6438		    gettext("missing <device> specification\n"));
6439		usage(B_FALSE);
6440	}
6441
6442	old_disk = argv[1];
6443
6444	if (argc < 3) {
6445		if (!replacing) {
6446			(void) fprintf(stderr,
6447			    gettext("missing <new_device> specification\n"));
6448			usage(B_FALSE);
6449		}
6450		new_disk = old_disk;
6451		argc -= 1;
6452		argv += 1;
6453	} else {
6454		new_disk = argv[2];
6455		argc -= 2;
6456		argv += 2;
6457	}
6458
6459	if (argc > 1) {
6460		(void) fprintf(stderr, gettext("too many arguments\n"));
6461		usage(B_FALSE);
6462	}
6463
6464	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
6465		nvlist_free(props);
6466		return (1);
6467	}
6468
6469	if (zpool_get_config(zhp, NULL) == NULL) {
6470		(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
6471		    poolname);
6472		zpool_close(zhp);
6473		nvlist_free(props);
6474		return (1);
6475	}
6476
6477	/* unless manually specified use "ashift" pool property (if set) */
6478	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
6479		int intval;
6480		zprop_source_t src;
6481		char strval[ZPOOL_MAXPROPLEN];
6482
6483		intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
6484		if (src != ZPROP_SRC_DEFAULT) {
6485			(void) sprintf(strval, "%" PRId32, intval);
6486			verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
6487			    &props, B_TRUE) == 0);
6488		}
6489	}
6490
6491	nvroot = make_root_vdev(zhp, props, force, B_FALSE, replacing, B_FALSE,
6492	    argc, argv);
6493	if (nvroot == NULL) {
6494		zpool_close(zhp);
6495		nvlist_free(props);
6496		return (1);
6497	}
6498
6499	ret = zpool_vdev_attach(zhp, old_disk, new_disk, nvroot, replacing,
6500	    rebuild);
6501
6502	if (ret == 0 && wait)
6503		ret = zpool_wait(zhp,
6504		    replacing ? ZPOOL_WAIT_REPLACE : ZPOOL_WAIT_RESILVER);
6505
6506	nvlist_free(props);
6507	nvlist_free(nvroot);
6508	zpool_close(zhp);
6509
6510	return (ret);
6511}
6512
6513/*
6514 * zpool replace [-fsw] [-o property=value] <pool> <device> <new_device>
6515 *
6516 *	-f	Force attach, even if <new_device> appears to be in use.
6517 *	-s	Use sequential instead of healing reconstruction for resilver.
6518 *	-o	Set property=value.
6519 *	-w	Wait for replacing to complete before returning
6520 *
6521 * Replace <device> with <new_device>.
6522 */
6523/* ARGSUSED */
6524int
6525zpool_do_replace(int argc, char **argv)
6526{
6527	return (zpool_do_attach_or_replace(argc, argv, B_TRUE));
6528}
6529
6530/*
6531 * zpool attach [-fsw] [-o property=value] <pool> <device> <new_device>
6532 *
6533 *	-f	Force attach, even if <new_device> appears to be in use.
6534 *	-s	Use sequential instead of healing reconstruction for resilver.
6535 *	-o	Set property=value.
6536 *	-w	Wait for resilvering to complete before returning
6537 *
6538 * Attach <new_device> to the mirror containing <device>.  If <device> is not
6539 * part of a mirror, then <device> will be transformed into a mirror of
6540 * <device> and <new_device>.  In either case, <new_device> will begin life
6541 * with a DTL of [0, now], and will immediately begin to resilver itself.
6542 */
6543int
6544zpool_do_attach(int argc, char **argv)
6545{
6546	return (zpool_do_attach_or_replace(argc, argv, B_FALSE));
6547}
6548
6549/*
6550 * zpool detach [-f] <pool> <device>
6551 *
6552 *	-f	Force detach of <device>, even if DTLs argue against it
6553 *		(not supported yet)
6554 *
6555 * Detach a device from a mirror.  The operation will be refused if <device>
6556 * is the last device in the mirror, or if the DTLs indicate that this device
6557 * has the only valid copy of some data.
6558 */
6559/* ARGSUSED */
6560int
6561zpool_do_detach(int argc, char **argv)
6562{
6563	int c;
6564	char *poolname, *path;
6565	zpool_handle_t *zhp;
6566	int ret;
6567
6568	/* check options */
6569	while ((c = getopt(argc, argv, "")) != -1) {
6570		switch (c) {
6571		case '?':
6572			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6573			    optopt);
6574			usage(B_FALSE);
6575		}
6576	}
6577
6578	argc -= optind;
6579	argv += optind;
6580
6581	/* get pool name and check number of arguments */
6582	if (argc < 1) {
6583		(void) fprintf(stderr, gettext("missing pool name argument\n"));
6584		usage(B_FALSE);
6585	}
6586
6587	if (argc < 2) {
6588		(void) fprintf(stderr,
6589		    gettext("missing <device> specification\n"));
6590		usage(B_FALSE);
6591	}
6592
6593	poolname = argv[0];
6594	path = argv[1];
6595
6596	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6597		return (1);
6598
6599	ret = zpool_vdev_detach(zhp, path);
6600
6601	zpool_close(zhp);
6602
6603	return (ret);
6604}
6605
6606/*
6607 * zpool split [-gLnP] [-o prop=val] ...
6608 *		[-o mntopt] ...
6609 *		[-R altroot] <pool> <newpool> [<device> ...]
6610 *
6611 *	-g      Display guid for individual vdev name.
6612 *	-L	Follow links when resolving vdev path name.
6613 *	-n	Do not split the pool, but display the resulting layout if
6614 *		it were to be split.
6615 *	-o	Set property=value, or set mount options.
6616 *	-P	Display full path for vdev name.
6617 *	-R	Mount the split-off pool under an alternate root.
6618 *	-l	Load encryption keys while importing.
6619 *
6620 * Splits the named pool and gives it the new pool name.  Devices to be split
6621 * off may be listed, provided that no more than one device is specified
6622 * per top-level vdev mirror.  The newly split pool is left in an exported
6623 * state unless -R is specified.
6624 *
6625 * Restrictions: the top-level of the pool pool must only be made up of
6626 * mirrors; all devices in the pool must be healthy; no device may be
6627 * undergoing a resilvering operation.
6628 */
6629int
6630zpool_do_split(int argc, char **argv)
6631{
6632	char *srcpool, *newpool, *propval;
6633	char *mntopts = NULL;
6634	splitflags_t flags;
6635	int c, ret = 0;
6636	boolean_t loadkeys = B_FALSE;
6637	zpool_handle_t *zhp;
6638	nvlist_t *config, *props = NULL;
6639
6640	flags.dryrun = B_FALSE;
6641	flags.import = B_FALSE;
6642	flags.name_flags = 0;
6643
6644	/* check options */
6645	while ((c = getopt(argc, argv, ":gLR:lno:P")) != -1) {
6646		switch (c) {
6647		case 'g':
6648			flags.name_flags |= VDEV_NAME_GUID;
6649			break;
6650		case 'L':
6651			flags.name_flags |= VDEV_NAME_FOLLOW_LINKS;
6652			break;
6653		case 'R':
6654			flags.import = B_TRUE;
6655			if (add_prop_list(
6656			    zpool_prop_to_name(ZPOOL_PROP_ALTROOT), optarg,
6657			    &props, B_TRUE) != 0) {
6658				nvlist_free(props);
6659				usage(B_FALSE);
6660			}
6661			break;
6662		case 'l':
6663			loadkeys = B_TRUE;
6664			break;
6665		case 'n':
6666			flags.dryrun = B_TRUE;
6667			break;
6668		case 'o':
6669			if ((propval = strchr(optarg, '=')) != NULL) {
6670				*propval = '\0';
6671				propval++;
6672				if (add_prop_list(optarg, propval,
6673				    &props, B_TRUE) != 0) {
6674					nvlist_free(props);
6675					usage(B_FALSE);
6676				}
6677			} else {
6678				mntopts = optarg;
6679			}
6680			break;
6681		case 'P':
6682			flags.name_flags |= VDEV_NAME_PATH;
6683			break;
6684		case ':':
6685			(void) fprintf(stderr, gettext("missing argument for "
6686			    "'%c' option\n"), optopt);
6687			usage(B_FALSE);
6688			break;
6689		case '?':
6690			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6691			    optopt);
6692			usage(B_FALSE);
6693			break;
6694		}
6695	}
6696
6697	if (!flags.import && mntopts != NULL) {
6698		(void) fprintf(stderr, gettext("setting mntopts is only "
6699		    "valid when importing the pool\n"));
6700		usage(B_FALSE);
6701	}
6702
6703	if (!flags.import && loadkeys) {
6704		(void) fprintf(stderr, gettext("loading keys is only "
6705		    "valid when importing the pool\n"));
6706		usage(B_FALSE);
6707	}
6708
6709	argc -= optind;
6710	argv += optind;
6711
6712	if (argc < 1) {
6713		(void) fprintf(stderr, gettext("Missing pool name\n"));
6714		usage(B_FALSE);
6715	}
6716	if (argc < 2) {
6717		(void) fprintf(stderr, gettext("Missing new pool name\n"));
6718		usage(B_FALSE);
6719	}
6720
6721	srcpool = argv[0];
6722	newpool = argv[1];
6723
6724	argc -= 2;
6725	argv += 2;
6726
6727	if ((zhp = zpool_open(g_zfs, srcpool)) == NULL) {
6728		nvlist_free(props);
6729		return (1);
6730	}
6731
6732	config = split_mirror_vdev(zhp, newpool, props, flags, argc, argv);
6733	if (config == NULL) {
6734		ret = 1;
6735	} else {
6736		if (flags.dryrun) {
6737			(void) printf(gettext("would create '%s' with the "
6738			    "following layout:\n\n"), newpool);
6739			print_vdev_tree(NULL, newpool, config, 0, "",
6740			    flags.name_flags);
6741			print_vdev_tree(NULL, "dedup", config, 0,
6742			    VDEV_ALLOC_BIAS_DEDUP, 0);
6743			print_vdev_tree(NULL, "special", config, 0,
6744			    VDEV_ALLOC_BIAS_SPECIAL, 0);
6745		}
6746	}
6747
6748	zpool_close(zhp);
6749
6750	if (ret != 0 || flags.dryrun || !flags.import) {
6751		nvlist_free(config);
6752		nvlist_free(props);
6753		return (ret);
6754	}
6755
6756	/*
6757	 * The split was successful. Now we need to open the new
6758	 * pool and import it.
6759	 */
6760	if ((zhp = zpool_open_canfail(g_zfs, newpool)) == NULL) {
6761		nvlist_free(config);
6762		nvlist_free(props);
6763		return (1);
6764	}
6765
6766	if (loadkeys) {
6767		ret = zfs_crypto_attempt_load_keys(g_zfs, newpool);
6768		if (ret != 0)
6769			ret = 1;
6770	}
6771
6772	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
6773	    zpool_enable_datasets(zhp, mntopts, 0) != 0) {
6774		ret = 1;
6775		(void) fprintf(stderr, gettext("Split was successful, but "
6776		    "the datasets could not all be mounted\n"));
6777		(void) fprintf(stderr, gettext("Try doing '%s' with a "
6778		    "different altroot\n"), "zpool import");
6779	}
6780	zpool_close(zhp);
6781	nvlist_free(config);
6782	nvlist_free(props);
6783
6784	return (ret);
6785}
6786
6787
6788
6789/*
6790 * zpool online <pool> <device> ...
6791 */
6792int
6793zpool_do_online(int argc, char **argv)
6794{
6795	int c, i;
6796	char *poolname;
6797	zpool_handle_t *zhp;
6798	int ret = 0;
6799	vdev_state_t newstate;
6800	int flags = 0;
6801
6802	/* check options */
6803	while ((c = getopt(argc, argv, "e")) != -1) {
6804		switch (c) {
6805		case 'e':
6806			flags |= ZFS_ONLINE_EXPAND;
6807			break;
6808		case '?':
6809			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6810			    optopt);
6811			usage(B_FALSE);
6812		}
6813	}
6814
6815	argc -= optind;
6816	argv += optind;
6817
6818	/* get pool name and check number of arguments */
6819	if (argc < 1) {
6820		(void) fprintf(stderr, gettext("missing pool name\n"));
6821		usage(B_FALSE);
6822	}
6823	if (argc < 2) {
6824		(void) fprintf(stderr, gettext("missing device name\n"));
6825		usage(B_FALSE);
6826	}
6827
6828	poolname = argv[0];
6829
6830	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6831		return (1);
6832
6833	for (i = 1; i < argc; i++) {
6834		if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
6835			if (newstate != VDEV_STATE_HEALTHY) {
6836				(void) printf(gettext("warning: device '%s' "
6837				    "onlined, but remains in faulted state\n"),
6838				    argv[i]);
6839				if (newstate == VDEV_STATE_FAULTED)
6840					(void) printf(gettext("use 'zpool "
6841					    "clear' to restore a faulted "
6842					    "device\n"));
6843				else
6844					(void) printf(gettext("use 'zpool "
6845					    "replace' to replace devices "
6846					    "that are no longer present\n"));
6847			}
6848		} else {
6849			ret = 1;
6850		}
6851	}
6852
6853	zpool_close(zhp);
6854
6855	return (ret);
6856}
6857
6858/*
6859 * zpool offline [-ft] <pool> <device> ...
6860 *
6861 *	-f	Force the device into a faulted state.
6862 *
6863 *	-t	Only take the device off-line temporarily.  The offline/faulted
6864 *		state will not be persistent across reboots.
6865 */
6866/* ARGSUSED */
6867int
6868zpool_do_offline(int argc, char **argv)
6869{
6870	int c, i;
6871	char *poolname;
6872	zpool_handle_t *zhp;
6873	int ret = 0;
6874	boolean_t istmp = B_FALSE;
6875	boolean_t fault = B_FALSE;
6876
6877	/* check options */
6878	while ((c = getopt(argc, argv, "ft")) != -1) {
6879		switch (c) {
6880		case 'f':
6881			fault = B_TRUE;
6882			break;
6883		case 't':
6884			istmp = B_TRUE;
6885			break;
6886		case '?':
6887			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6888			    optopt);
6889			usage(B_FALSE);
6890		}
6891	}
6892
6893	argc -= optind;
6894	argv += optind;
6895
6896	/* get pool name and check number of arguments */
6897	if (argc < 1) {
6898		(void) fprintf(stderr, gettext("missing pool name\n"));
6899		usage(B_FALSE);
6900	}
6901	if (argc < 2) {
6902		(void) fprintf(stderr, gettext("missing device name\n"));
6903		usage(B_FALSE);
6904	}
6905
6906	poolname = argv[0];
6907
6908	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6909		return (1);
6910
6911	for (i = 1; i < argc; i++) {
6912		if (fault) {
6913			uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]);
6914			vdev_aux_t aux;
6915			if (istmp == B_FALSE) {
6916				/* Force the fault to persist across imports */
6917				aux = VDEV_AUX_EXTERNAL_PERSIST;
6918			} else {
6919				aux = VDEV_AUX_EXTERNAL;
6920			}
6921
6922			if (guid == 0 || zpool_vdev_fault(zhp, guid, aux) != 0)
6923				ret = 1;
6924		} else {
6925			if (zpool_vdev_offline(zhp, argv[i], istmp) != 0)
6926				ret = 1;
6927		}
6928	}
6929
6930	zpool_close(zhp);
6931
6932	return (ret);
6933}
6934
6935/*
6936 * zpool clear <pool> [device]
6937 *
6938 * Clear all errors associated with a pool or a particular device.
6939 */
6940int
6941zpool_do_clear(int argc, char **argv)
6942{
6943	int c;
6944	int ret = 0;
6945	boolean_t dryrun = B_FALSE;
6946	boolean_t do_rewind = B_FALSE;
6947	boolean_t xtreme_rewind = B_FALSE;
6948	uint32_t rewind_policy = ZPOOL_NO_REWIND;
6949	nvlist_t *policy = NULL;
6950	zpool_handle_t *zhp;
6951	char *pool, *device;
6952
6953	/* check options */
6954	while ((c = getopt(argc, argv, "FnX")) != -1) {
6955		switch (c) {
6956		case 'F':
6957			do_rewind = B_TRUE;
6958			break;
6959		case 'n':
6960			dryrun = B_TRUE;
6961			break;
6962		case 'X':
6963			xtreme_rewind = B_TRUE;
6964			break;
6965		case '?':
6966			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6967			    optopt);
6968			usage(B_FALSE);
6969		}
6970	}
6971
6972	argc -= optind;
6973	argv += optind;
6974
6975	if (argc < 1) {
6976		(void) fprintf(stderr, gettext("missing pool name\n"));
6977		usage(B_FALSE);
6978	}
6979
6980	if (argc > 2) {
6981		(void) fprintf(stderr, gettext("too many arguments\n"));
6982		usage(B_FALSE);
6983	}
6984
6985	if ((dryrun || xtreme_rewind) && !do_rewind) {
6986		(void) fprintf(stderr,
6987		    gettext("-n or -X only meaningful with -F\n"));
6988		usage(B_FALSE);
6989	}
6990	if (dryrun)
6991		rewind_policy = ZPOOL_TRY_REWIND;
6992	else if (do_rewind)
6993		rewind_policy = ZPOOL_DO_REWIND;
6994	if (xtreme_rewind)
6995		rewind_policy |= ZPOOL_EXTREME_REWIND;
6996
6997	/* In future, further rewind policy choices can be passed along here */
6998	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 ||
6999	    nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
7000	    rewind_policy) != 0) {
7001		return (1);
7002	}
7003
7004	pool = argv[0];
7005	device = argc == 2 ? argv[1] : NULL;
7006
7007	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
7008		nvlist_free(policy);
7009		return (1);
7010	}
7011
7012	if (zpool_clear(zhp, device, policy) != 0)
7013		ret = 1;
7014
7015	zpool_close(zhp);
7016
7017	nvlist_free(policy);
7018
7019	return (ret);
7020}
7021
7022/*
7023 * zpool reguid <pool>
7024 */
7025int
7026zpool_do_reguid(int argc, char **argv)
7027{
7028	int c;
7029	char *poolname;
7030	zpool_handle_t *zhp;
7031	int ret = 0;
7032
7033	/* check options */
7034	while ((c = getopt(argc, argv, "")) != -1) {
7035		switch (c) {
7036		case '?':
7037			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
7038			    optopt);
7039			usage(B_FALSE);
7040		}
7041	}
7042
7043	argc -= optind;
7044	argv += optind;
7045
7046	/* get pool name and check number of arguments */
7047	if (argc < 1) {
7048		(void) fprintf(stderr, gettext("missing pool name\n"));
7049		usage(B_FALSE);
7050	}
7051
7052	if (argc > 1) {
7053		(void) fprintf(stderr, gettext("too many arguments\n"));
7054		usage(B_FALSE);
7055	}
7056
7057	poolname = argv[0];
7058	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
7059		return (1);
7060
7061	ret = zpool_reguid(zhp);
7062
7063	zpool_close(zhp);
7064	return (ret);
7065}
7066
7067
7068/*
7069 * zpool reopen <pool>
7070 *
7071 * Reopen the pool so that the kernel can update the sizes of all vdevs.
7072 */
7073int
7074zpool_do_reopen(int argc, char **argv)
7075{
7076	int c;
7077	int ret = 0;
7078	boolean_t scrub_restart = B_TRUE;
7079
7080	/* check options */
7081	while ((c = getopt(argc, argv, "n")) != -1) {
7082		switch (c) {
7083		case 'n':
7084			scrub_restart = B_FALSE;
7085			break;
7086		case '?':
7087			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
7088			    optopt);
7089			usage(B_FALSE);
7090		}
7091	}
7092
7093	argc -= optind;
7094	argv += optind;
7095
7096	/* if argc == 0 we will execute zpool_reopen_one on all pools */
7097	ret = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE, zpool_reopen_one,
7098	    &scrub_restart);
7099
7100	return (ret);
7101}
7102
7103typedef struct scrub_cbdata {
7104	int	cb_type;
7105	pool_scrub_cmd_t cb_scrub_cmd;
7106} scrub_cbdata_t;
7107
7108static boolean_t
7109zpool_has_checkpoint(zpool_handle_t *zhp)
7110{
7111	nvlist_t *config, *nvroot;
7112
7113	config = zpool_get_config(zhp, NULL);
7114
7115	if (config != NULL) {
7116		pool_checkpoint_stat_t *pcs = NULL;
7117		uint_t c;
7118
7119		nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
7120		(void) nvlist_lookup_uint64_array(nvroot,
7121		    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
7122
7123		if (pcs == NULL || pcs->pcs_state == CS_NONE)
7124			return (B_FALSE);
7125
7126		assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS ||
7127		    pcs->pcs_state == CS_CHECKPOINT_DISCARDING);
7128		return (B_TRUE);
7129	}
7130
7131	return (B_FALSE);
7132}
7133
7134static int
7135scrub_callback(zpool_handle_t *zhp, void *data)
7136{
7137	scrub_cbdata_t *cb = data;
7138	int err;
7139
7140	/*
7141	 * Ignore faulted pools.
7142	 */
7143	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
7144		(void) fprintf(stderr, gettext("cannot scan '%s': pool is "
7145		    "currently unavailable\n"), zpool_get_name(zhp));
7146		return (1);
7147	}
7148
7149	err = zpool_scan(zhp, cb->cb_type, cb->cb_scrub_cmd);
7150
7151	if (err == 0 && zpool_has_checkpoint(zhp) &&
7152	    cb->cb_type == POOL_SCAN_SCRUB) {
7153		(void) printf(gettext("warning: will not scrub state that "
7154		    "belongs to the checkpoint of pool '%s'\n"),
7155		    zpool_get_name(zhp));
7156	}
7157
7158	return (err != 0);
7159}
7160
7161static int
7162wait_callback(zpool_handle_t *zhp, void *data)
7163{
7164	zpool_wait_activity_t *act = data;
7165	return (zpool_wait(zhp, *act));
7166}
7167
7168/*
7169 * zpool scrub [-s | -p] [-w] <pool> ...
7170 *
7171 *	-s	Stop.  Stops any in-progress scrub.
7172 *	-p	Pause. Pause in-progress scrub.
7173 *	-w	Wait.  Blocks until scrub has completed.
7174 */
7175int
7176zpool_do_scrub(int argc, char **argv)
7177{
7178	int c;
7179	scrub_cbdata_t cb;
7180	boolean_t wait = B_FALSE;
7181	int error;
7182
7183	cb.cb_type = POOL_SCAN_SCRUB;
7184	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;
7185
7186	/* check options */
7187	while ((c = getopt(argc, argv, "spw")) != -1) {
7188		switch (c) {
7189		case 's':
7190			cb.cb_type = POOL_SCAN_NONE;
7191			break;
7192		case 'p':
7193			cb.cb_scrub_cmd = POOL_SCRUB_PAUSE;
7194			break;
7195		case 'w':
7196			wait = B_TRUE;
7197			break;
7198		case '?':
7199			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
7200			    optopt);
7201			usage(B_FALSE);
7202		}
7203	}
7204
7205	if (cb.cb_type == POOL_SCAN_NONE &&
7206	    cb.cb_scrub_cmd == POOL_SCRUB_PAUSE) {
7207		(void) fprintf(stderr, gettext("invalid option combination: "
7208		    "-s and -p are mutually exclusive\n"));
7209		usage(B_FALSE);
7210	}
7211
7212	if (wait && (cb.cb_type == POOL_SCAN_NONE ||
7213	    cb.cb_scrub_cmd == POOL_SCRUB_PAUSE)) {
7214		(void) fprintf(stderr, gettext("invalid option combination: "
7215		    "-w cannot be used with -p or -s\n"));
7216		usage(B_FALSE);
7217	}
7218
7219	argc -= optind;
7220	argv += optind;
7221
7222	if (argc < 1) {
7223		(void) fprintf(stderr, gettext("missing pool name argument\n"));
7224		usage(B_FALSE);
7225	}
7226
7227	error = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE,
7228	    scrub_callback, &cb);
7229
7230	if (wait && !error) {
7231		zpool_wait_activity_t act = ZPOOL_WAIT_SCRUB;
7232		error = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE,
7233		    wait_callback, &act);
7234	}
7235
7236	return (error);
7237}
7238
7239/*
7240 * zpool resilver <pool> ...
7241 *
7242 *	Restarts any in-progress resilver
7243 */
7244int
7245zpool_do_resilver(int argc, char **argv)
7246{
7247	int c;
7248	scrub_cbdata_t cb;
7249
7250	cb.cb_type = POOL_SCAN_RESILVER;
7251	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;
7252
7253	/* check options */
7254	while ((c = getopt(argc, argv, "")) != -1) {
7255		switch (c) {
7256		case '?':
7257			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
7258			    optopt);
7259			usage(B_FALSE);
7260		}
7261	}
7262
7263	argc -= optind;
7264	argv += optind;
7265
7266	if (argc < 1) {
7267		(void) fprintf(stderr, gettext("missing pool name argument\n"));
7268		usage(B_FALSE);
7269	}
7270
7271	return (for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE,
7272	    scrub_callback, &cb));
7273}
7274
7275/*
7276 * zpool trim [-d] [-r <rate>] [-c | -s] <pool> [<device> ...]
7277 *
7278 *	-c		Cancel. Ends any in-progress trim.
7279 *	-d		Secure trim.  Requires kernel and device support.
7280 *	-r <rate>	Sets the TRIM rate in bytes (per second). Supports
7281 *			adding a multiplier suffix such as 'k' or 'm'.
7282 *	-s		Suspend. TRIM can then be restarted with no flags.
7283 *	-w		Wait. Blocks until trimming has completed.
7284 */
7285int
7286zpool_do_trim(int argc, char **argv)
7287{
7288	struct option long_options[] = {
7289		{"cancel",	no_argument,		NULL,	'c'},
7290		{"secure",	no_argument,		NULL,	'd'},
7291		{"rate",	required_argument,	NULL,	'r'},
7292		{"suspend",	no_argument,		NULL,	's'},
7293		{"wait",	no_argument,		NULL,	'w'},
7294		{0, 0, 0, 0}
7295	};
7296
7297	pool_trim_func_t cmd_type = POOL_TRIM_START;
7298	uint64_t rate = 0;
7299	boolean_t secure = B_FALSE;
7300	boolean_t wait = B_FALSE;
7301
7302	int c;
7303	while ((c = getopt_long(argc, argv, "cdr:sw", long_options, NULL))
7304	    != -1) {
7305		switch (c) {
7306		case 'c':
7307			if (cmd_type != POOL_TRIM_START &&
7308			    cmd_type != POOL_TRIM_CANCEL) {
7309				(void) fprintf(stderr, gettext("-c cannot be "
7310				    "combined with other options\n"));
7311				usage(B_FALSE);
7312			}
7313			cmd_type = POOL_TRIM_CANCEL;
7314			break;
7315		case 'd':
7316			if (cmd_type != POOL_TRIM_START) {
7317				(void) fprintf(stderr, gettext("-d cannot be "
7318				    "combined with the -c or -s options\n"));
7319				usage(B_FALSE);
7320			}
7321			secure = B_TRUE;
7322			break;
7323		case 'r':
7324			if (cmd_type != POOL_TRIM_START) {
7325				(void) fprintf(stderr, gettext("-r cannot be "
7326				    "combined with the -c or -s options\n"));
7327				usage(B_FALSE);
7328			}
7329			if (zfs_nicestrtonum(NULL, optarg, &rate) == -1) {
7330				(void) fprintf(stderr,
7331				    gettext("invalid value for rate\n"));
7332				usage(B_FALSE);
7333			}
7334			break;
7335		case 's':
7336			if (cmd_type != POOL_TRIM_START &&
7337			    cmd_type != POOL_TRIM_SUSPEND) {
7338				(void) fprintf(stderr, gettext("-s cannot be "
7339				    "combined with other options\n"));
7340				usage(B_FALSE);
7341			}
7342			cmd_type = POOL_TRIM_SUSPEND;
7343			break;
7344		case 'w':
7345			wait = B_TRUE;
7346			break;
7347		case '?':
7348			if (optopt != 0) {
7349				(void) fprintf(stderr,
7350				    gettext("invalid option '%c'\n"), optopt);
7351			} else {
7352				(void) fprintf(stderr,
7353				    gettext("invalid option '%s'\n"),
7354				    argv[optind - 1]);
7355			}
7356			usage(B_FALSE);
7357		}
7358	}
7359
7360	argc -= optind;
7361	argv += optind;
7362
7363	if (argc < 1) {
7364		(void) fprintf(stderr, gettext("missing pool name argument\n"));
7365		usage(B_FALSE);
7366		return (-1);
7367	}
7368
7369	if (wait && (cmd_type != POOL_TRIM_START)) {
7370		(void) fprintf(stderr, gettext("-w cannot be used with -c or "
7371		    "-s\n"));
7372		usage(B_FALSE);
7373	}
7374
7375	char *poolname = argv[0];
7376	zpool_handle_t *zhp = zpool_open(g_zfs, poolname);
7377	if (zhp == NULL)
7378		return (-1);
7379
7380	trimflags_t trim_flags = {
7381		.secure = secure,
7382		.rate = rate,
7383		.wait = wait,
7384	};
7385
7386	nvlist_t *vdevs = fnvlist_alloc();
7387	if (argc == 1) {
7388		/* no individual leaf vdevs specified, so add them all */
7389		nvlist_t *config = zpool_get_config(zhp, NULL);
7390		nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
7391		    ZPOOL_CONFIG_VDEV_TREE);
7392		zpool_collect_leaves(zhp, nvroot, vdevs);
7393		trim_flags.fullpool = B_TRUE;
7394	} else {
7395		trim_flags.fullpool = B_FALSE;
7396		for (int i = 1; i < argc; i++) {
7397			fnvlist_add_boolean(vdevs, argv[i]);
7398		}
7399	}
7400
7401	int error = zpool_trim(zhp, cmd_type, vdevs, &trim_flags);
7402
7403	fnvlist_free(vdevs);
7404	zpool_close(zhp);
7405
7406	return (error);
7407}
7408
7409/*
7410 * Converts a total number of seconds to a human readable string broken
7411 * down in to days/hours/minutes/seconds.
7412 */
7413static void
7414secs_to_dhms(uint64_t total, char *buf)
7415{
7416	uint64_t days = total / 60 / 60 / 24;
7417	uint64_t hours = (total / 60 / 60) % 24;
7418	uint64_t mins = (total / 60) % 60;
7419	uint64_t secs = (total % 60);
7420
7421	if (days > 0) {
7422		(void) sprintf(buf, "%llu days %02llu:%02llu:%02llu",
7423		    (u_longlong_t)days, (u_longlong_t)hours,
7424		    (u_longlong_t)mins, (u_longlong_t)secs);
7425	} else {
7426		(void) sprintf(buf, "%02llu:%02llu:%02llu",
7427		    (u_longlong_t)hours, (u_longlong_t)mins,
7428		    (u_longlong_t)secs);
7429	}
7430}
7431
7432/*
7433 * Print out detailed scrub status.
7434 */
7435static void
7436print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
7437{
7438	time_t start, end, pause;
7439	uint64_t pass_scanned, scanned, pass_issued, issued, total;
7440	uint64_t elapsed, scan_rate, issue_rate;
7441	double fraction_done;
7442	char processed_buf[7], scanned_buf[7], issued_buf[7], total_buf[7];
7443	char srate_buf[7], irate_buf[7], time_buf[32];
7444
7445	printf("  ");
7446	printf_color(ANSI_BOLD, gettext("scan:"));
7447	printf(" ");
7448
7449	/* If there's never been a scan, there's not much to say. */
7450	if (ps == NULL || ps->pss_func == POOL_SCAN_NONE ||
7451	    ps->pss_func >= POOL_SCAN_FUNCS) {
7452		(void) printf(gettext("none requested\n"));
7453		return;
7454	}
7455
7456	start = ps->pss_start_time;
7457	end = ps->pss_end_time;
7458	pause = ps->pss_pass_scrub_pause;
7459
7460	zfs_nicebytes(ps->pss_processed, processed_buf, sizeof (processed_buf));
7461
7462	assert(ps->pss_func == POOL_SCAN_SCRUB ||
7463	    ps->pss_func == POOL_SCAN_RESILVER);
7464
7465	/* Scan is finished or canceled. */
7466	if (ps->pss_state == DSS_FINISHED) {
7467		secs_to_dhms(end - start, time_buf);
7468
7469		if (ps->pss_func == POOL_SCAN_SCRUB) {
7470			(void) printf(gettext("scrub repaired %s "
7471			    "in %s with %llu errors on %s"), processed_buf,
7472			    time_buf, (u_longlong_t)ps->pss_errors,
7473			    ctime(&end));
7474		} else if (ps->pss_func == POOL_SCAN_RESILVER) {
7475			(void) printf(gettext("resilvered %s "
7476			    "in %s with %llu errors on %s"), processed_buf,
7477			    time_buf, (u_longlong_t)ps->pss_errors,
7478			    ctime(&end));
7479		}
7480		return;
7481	} else if (ps->pss_state == DSS_CANCELED) {
7482		if (ps->pss_func == POOL_SCAN_SCRUB) {
7483			(void) printf(gettext("scrub canceled on %s"),
7484			    ctime(&end));
7485		} else if (ps->pss_func == POOL_SCAN_RESILVER) {
7486			(void) printf(gettext("resilver canceled on %s"),
7487			    ctime(&end));
7488		}
7489		return;
7490	}
7491
7492	assert(ps->pss_state == DSS_SCANNING);
7493
7494	/* Scan is in progress. Resilvers can't be paused. */
7495	if (ps->pss_func == POOL_SCAN_SCRUB) {
7496		if (pause == 0) {
7497			(void) printf(gettext("scrub in progress since %s"),
7498			    ctime(&start));
7499		} else {
7500			(void) printf(gettext("scrub paused since %s"),
7501			    ctime(&pause));
7502			(void) printf(gettext("\tscrub started on %s"),
7503			    ctime(&start));
7504		}
7505	} else if (ps->pss_func == POOL_SCAN_RESILVER) {
7506		(void) printf(gettext("resilver in progress since %s"),
7507		    ctime(&start));
7508	}
7509
7510	scanned = ps->pss_examined;
7511	pass_scanned = ps->pss_pass_exam;
7512	issued = ps->pss_issued;
7513	pass_issued = ps->pss_pass_issued;
7514	total = ps->pss_to_examine;
7515
7516	/* we are only done with a block once we have issued the IO for it */
7517	fraction_done = (double)issued / total;
7518
7519	/* elapsed time for this pass, rounding up to 1 if it's 0 */
7520	elapsed = time(NULL) - ps->pss_pass_start;
7521	elapsed -= ps->pss_pass_scrub_spent_paused;
7522	elapsed = (elapsed != 0) ? elapsed : 1;
7523
7524	scan_rate = pass_scanned / elapsed;
7525	issue_rate = pass_issued / elapsed;
7526	uint64_t total_secs_left = (issue_rate != 0 && total >= issued) ?
7527	    ((total - issued) / issue_rate) : UINT64_MAX;
7528	secs_to_dhms(total_secs_left, time_buf);
7529
7530	/* format all of the numbers we will be reporting */
7531	zfs_nicebytes(scanned, scanned_buf, sizeof (scanned_buf));
7532	zfs_nicebytes(issued, issued_buf, sizeof (issued_buf));
7533	zfs_nicebytes(total, total_buf, sizeof (total_buf));
7534	zfs_nicebytes(scan_rate, srate_buf, sizeof (srate_buf));
7535	zfs_nicebytes(issue_rate, irate_buf, sizeof (irate_buf));
7536
7537	/* do not print estimated time if we have a paused scrub */
7538	if (pause == 0) {
7539		(void) printf(gettext("\t%s scanned at %s/s, "
7540		    "%s issued at %s/s, %s total\n"),
7541		    scanned_buf, srate_buf, issued_buf, irate_buf, total_buf);
7542	} else {
7543		(void) printf(gettext("\t%s scanned, %s issued, %s total\n"),
7544		    scanned_buf, issued_buf, total_buf);
7545	}
7546
7547	if (ps->pss_func == POOL_SCAN_RESILVER) {
7548		(void) printf(gettext("\t%s resilvered, %.2f%% done"),
7549		    processed_buf, 100 * fraction_done);
7550	} else if (ps->pss_func == POOL_SCAN_SCRUB) {
7551		(void) printf(gettext("\t%s repaired, %.2f%% done"),
7552		    processed_buf, 100 * fraction_done);
7553	}
7554
7555	if (pause == 0) {
7556		if (total_secs_left != UINT64_MAX &&
7557		    issue_rate >= 10 * 1024 * 1024) {
7558			(void) printf(gettext(", %s to go\n"), time_buf);
7559		} else {
7560			(void) printf(gettext(", no estimated "
7561			    "completion time\n"));
7562		}
7563	} else {
7564		(void) printf(gettext("\n"));
7565	}
7566}
7567
7568static void
7569print_rebuild_status_impl(vdev_rebuild_stat_t *vrs, char *vdev_name)
7570{
7571	if (vrs == NULL || vrs->vrs_state == VDEV_REBUILD_NONE)
7572		return;
7573
7574	printf("  ");
7575	printf_color(ANSI_BOLD, gettext("scan:"));
7576	printf(" ");
7577
7578	uint64_t bytes_scanned = vrs->vrs_bytes_scanned;
7579	uint64_t bytes_issued = vrs->vrs_bytes_issued;
7580	uint64_t bytes_rebuilt = vrs->vrs_bytes_rebuilt;
7581	uint64_t bytes_est = vrs->vrs_bytes_est;
7582	uint64_t scan_rate = (vrs->vrs_pass_bytes_scanned /
7583	    (vrs->vrs_pass_time_ms + 1)) * 1000;
7584	uint64_t issue_rate = (vrs->vrs_pass_bytes_issued /
7585	    (vrs->vrs_pass_time_ms + 1)) * 1000;
7586	double scan_pct = MIN((double)bytes_scanned * 100 /
7587	    (bytes_est + 1), 100);
7588
7589	/* Format all of the numbers we will be reporting */
7590	char bytes_scanned_buf[7], bytes_issued_buf[7];
7591	char bytes_rebuilt_buf[7], bytes_est_buf[7];
7592	char scan_rate_buf[7], issue_rate_buf[7], time_buf[32];
7593	zfs_nicebytes(bytes_scanned, bytes_scanned_buf,
7594	    sizeof (bytes_scanned_buf));
7595	zfs_nicebytes(bytes_issued, bytes_issued_buf,
7596	    sizeof (bytes_issued_buf));
7597	zfs_nicebytes(bytes_rebuilt, bytes_rebuilt_buf,
7598	    sizeof (bytes_rebuilt_buf));
7599	zfs_nicebytes(bytes_est, bytes_est_buf, sizeof (bytes_est_buf));
7600	zfs_nicebytes(scan_rate, scan_rate_buf, sizeof (scan_rate_buf));
7601	zfs_nicebytes(issue_rate, issue_rate_buf, sizeof (issue_rate_buf));
7602
7603	time_t start = vrs->vrs_start_time;
7604	time_t end = vrs->vrs_end_time;
7605
7606	/* Rebuild is finished or canceled. */
7607	if (vrs->vrs_state == VDEV_REBUILD_COMPLETE) {
7608		secs_to_dhms(vrs->vrs_scan_time_ms / 1000, time_buf);
7609		(void) printf(gettext("resilvered (%s) %s in %s "
7610		    "with %llu errors on %s"), vdev_name, bytes_rebuilt_buf,
7611		    time_buf, (u_longlong_t)vrs->vrs_errors, ctime(&end));
7612		return;
7613	} else if (vrs->vrs_state == VDEV_REBUILD_CANCELED) {
7614		(void) printf(gettext("resilver (%s) canceled on %s"),
7615		    vdev_name, ctime(&end));
7616		return;
7617	} else if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
7618		(void) printf(gettext("resilver (%s) in progress since %s"),
7619		    vdev_name, ctime(&start));
7620	}
7621
7622	assert(vrs->vrs_state == VDEV_REBUILD_ACTIVE);
7623
7624	secs_to_dhms(MAX((int64_t)bytes_est - (int64_t)bytes_scanned, 0) /
7625	    MAX(scan_rate, 1), time_buf);
7626
7627	(void) printf(gettext("\t%s scanned at %s/s, %s issued %s/s, "
7628	    "%s total\n"), bytes_scanned_buf, scan_rate_buf,
7629	    bytes_issued_buf, issue_rate_buf, bytes_est_buf);
7630	(void) printf(gettext("\t%s resilvered, %.2f%% done"),
7631	    bytes_rebuilt_buf, scan_pct);
7632
7633	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
7634		if (scan_rate >= 10 * 1024 * 1024) {
7635			(void) printf(gettext(", %s to go\n"), time_buf);
7636		} else {
7637			(void) printf(gettext(", no estimated "
7638			    "completion time\n"));
7639		}
7640	} else {
7641		(void) printf(gettext("\n"));
7642	}
7643}
7644
7645/*
7646 * Print rebuild status for top-level vdevs.
7647 */
7648static void
7649print_rebuild_status(zpool_handle_t *zhp, nvlist_t *nvroot)
7650{
7651	nvlist_t **child;
7652	uint_t children;
7653
7654	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
7655	    &child, &children) != 0)
7656		children = 0;
7657
7658	for (uint_t c = 0; c < children; c++) {
7659		vdev_rebuild_stat_t *vrs;
7660		uint_t i;
7661
7662		if (nvlist_lookup_uint64_array(child[c],
7663		    ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {
7664			char *name = zpool_vdev_name(g_zfs, zhp,
7665			    child[c], VDEV_NAME_TYPE_ID);
7666			print_rebuild_status_impl(vrs, name);
7667			free(name);
7668		}
7669	}
7670}
7671
7672/*
7673 * As we don't scrub checkpointed blocks, we want to warn the user that we
7674 * skipped scanning some blocks if a checkpoint exists or existed at any
7675 * time during the scan.  If a sequential instead of healing reconstruction
7676 * was performed then the blocks were reconstructed.  However, their checksums
7677 * have not been verified so we still print the warning.
7678 */
7679static void
7680print_checkpoint_scan_warning(pool_scan_stat_t *ps, pool_checkpoint_stat_t *pcs)
7681{
7682	if (ps == NULL || pcs == NULL)
7683		return;
7684
7685	if (pcs->pcs_state == CS_NONE ||
7686	    pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
7687		return;
7688
7689	assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS);
7690
7691	if (ps->pss_state == DSS_NONE)
7692		return;
7693
7694	if ((ps->pss_state == DSS_FINISHED || ps->pss_state == DSS_CANCELED) &&
7695	    ps->pss_end_time < pcs->pcs_start_time)
7696		return;
7697
7698	if (ps->pss_state == DSS_FINISHED || ps->pss_state == DSS_CANCELED) {
7699		(void) printf(gettext("    scan warning: skipped blocks "
7700		    "that are only referenced by the checkpoint.\n"));
7701	} else {
7702		assert(ps->pss_state == DSS_SCANNING);
7703		(void) printf(gettext("    scan warning: skipping blocks "
7704		    "that are only referenced by the checkpoint.\n"));
7705	}
7706}
7707
7708/*
7709 * Returns B_TRUE if there is an active rebuild in progress.  Otherwise,
7710 * B_FALSE is returned and 'rebuild_end_time' is set to the end time for
7711 * the last completed (or cancelled) rebuild.
7712 */
7713static boolean_t
7714check_rebuilding(nvlist_t *nvroot, uint64_t *rebuild_end_time)
7715{
7716	nvlist_t **child;
7717	uint_t children;
7718	boolean_t rebuilding = B_FALSE;
7719	uint64_t end_time = 0;
7720
7721	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
7722	    &child, &children) != 0)
7723		children = 0;
7724
7725	for (uint_t c = 0; c < children; c++) {
7726		vdev_rebuild_stat_t *vrs;
7727		uint_t i;
7728
7729		if (nvlist_lookup_uint64_array(child[c],
7730		    ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {
7731
7732			if (vrs->vrs_end_time > end_time)
7733				end_time = vrs->vrs_end_time;
7734
7735			if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
7736				rebuilding = B_TRUE;
7737				end_time = 0;
7738				break;
7739			}
7740		}
7741	}
7742
7743	if (rebuild_end_time != NULL)
7744		*rebuild_end_time = end_time;
7745
7746	return (rebuilding);
7747}
7748
7749/*
7750 * Print the scan status.
7751 */
7752static void
7753print_scan_status(zpool_handle_t *zhp, nvlist_t *nvroot)
7754{
7755	uint64_t rebuild_end_time = 0, resilver_end_time = 0;
7756	boolean_t have_resilver = B_FALSE, have_scrub = B_FALSE;
7757	boolean_t active_resilver = B_FALSE;
7758	pool_checkpoint_stat_t *pcs = NULL;
7759	pool_scan_stat_t *ps = NULL;
7760	uint_t c;
7761
7762	if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
7763	    (uint64_t **)&ps, &c) == 0) {
7764		if (ps->pss_func == POOL_SCAN_RESILVER) {
7765			resilver_end_time = ps->pss_end_time;
7766			active_resilver = (ps->pss_state == DSS_SCANNING);
7767		}
7768
7769		have_resilver = (ps->pss_func == POOL_SCAN_RESILVER);
7770		have_scrub = (ps->pss_func == POOL_SCAN_SCRUB);
7771	}
7772
7773	boolean_t active_rebuild = check_rebuilding(nvroot, &rebuild_end_time);
7774	boolean_t have_rebuild = (active_rebuild || (rebuild_end_time > 0));
7775
7776	/* Always print the scrub status when available. */
7777	if (have_scrub)
7778		print_scan_scrub_resilver_status(ps);
7779
7780	/*
7781	 * When there is an active resilver or rebuild print its status.
7782	 * Otherwise print the status of the last resilver or rebuild.
7783	 */
7784	if (active_resilver || (!active_rebuild && have_resilver &&
7785	    resilver_end_time && resilver_end_time > rebuild_end_time)) {
7786		print_scan_scrub_resilver_status(ps);
7787	} else if (active_rebuild || (!active_resilver && have_rebuild &&
7788	    rebuild_end_time && rebuild_end_time > resilver_end_time)) {
7789		print_rebuild_status(zhp, nvroot);
7790	}
7791
7792	(void) nvlist_lookup_uint64_array(nvroot,
7793	    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
7794	print_checkpoint_scan_warning(ps, pcs);
7795}
7796
7797/*
7798 * Print out detailed removal status.
7799 */
7800static void
7801print_removal_status(zpool_handle_t *zhp, pool_removal_stat_t *prs)
7802{
7803	char copied_buf[7], examined_buf[7], total_buf[7], rate_buf[7];
7804	time_t start, end;
7805	nvlist_t *config, *nvroot;
7806	nvlist_t **child;
7807	uint_t children;
7808	char *vdev_name;
7809
7810	if (prs == NULL || prs->prs_state == DSS_NONE)
7811		return;
7812
7813	/*
7814	 * Determine name of vdev.
7815	 */
7816	config = zpool_get_config(zhp, NULL);
7817	nvroot = fnvlist_lookup_nvlist(config,
7818	    ZPOOL_CONFIG_VDEV_TREE);
7819	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
7820	    &child, &children) == 0);
7821	assert(prs->prs_removing_vdev < children);
7822	vdev_name = zpool_vdev_name(g_zfs, zhp,
7823	    child[prs->prs_removing_vdev], B_TRUE);
7824
7825	printf_color(ANSI_BOLD, gettext("remove: "));
7826
7827	start = prs->prs_start_time;
7828	end = prs->prs_end_time;
7829	zfs_nicenum(prs->prs_copied, copied_buf, sizeof (copied_buf));
7830
7831	/*
7832	 * Removal is finished or canceled.
7833	 */
7834	if (prs->prs_state == DSS_FINISHED) {
7835		uint64_t minutes_taken = (end - start) / 60;
7836
7837		(void) printf(gettext("Removal of vdev %llu copied %s "
7838		    "in %lluh%um, completed on %s"),
7839		    (longlong_t)prs->prs_removing_vdev,
7840		    copied_buf,
7841		    (u_longlong_t)(minutes_taken / 60),
7842		    (uint_t)(minutes_taken % 60),
7843		    ctime((time_t *)&end));
7844	} else if (prs->prs_state == DSS_CANCELED) {
7845		(void) printf(gettext("Removal of %s canceled on %s"),
7846		    vdev_name, ctime(&end));
7847	} else {
7848		uint64_t copied, total, elapsed, mins_left, hours_left;
7849		double fraction_done;
7850		uint_t rate;
7851
7852		assert(prs->prs_state == DSS_SCANNING);
7853
7854		/*
7855		 * Removal is in progress.
7856		 */
7857		(void) printf(gettext(
7858		    "Evacuation of %s in progress since %s"),
7859		    vdev_name, ctime(&start));
7860
7861		copied = prs->prs_copied > 0 ? prs->prs_copied : 1;
7862		total = prs->prs_to_copy;
7863		fraction_done = (double)copied / total;
7864
7865		/* elapsed time for this pass */
7866		elapsed = time(NULL) - prs->prs_start_time;
7867		elapsed = elapsed > 0 ? elapsed : 1;
7868		rate = copied / elapsed;
7869		rate = rate > 0 ? rate : 1;
7870		mins_left = ((total - copied) / rate) / 60;
7871		hours_left = mins_left / 60;
7872
7873		zfs_nicenum(copied, examined_buf, sizeof (examined_buf));
7874		zfs_nicenum(total, total_buf, sizeof (total_buf));
7875		zfs_nicenum(rate, rate_buf, sizeof (rate_buf));
7876
7877		/*
7878		 * do not print estimated time if hours_left is more than
7879		 * 30 days
7880		 */
7881		(void) printf(gettext(
7882		    "\t%s copied out of %s at %s/s, %.2f%% done"),
7883		    examined_buf, total_buf, rate_buf, 100 * fraction_done);
7884		if (hours_left < (30 * 24)) {
7885			(void) printf(gettext(", %lluh%um to go\n"),
7886			    (u_longlong_t)hours_left, (uint_t)(mins_left % 60));
7887		} else {
7888			(void) printf(gettext(
7889			    ", (copy is slow, no estimated time)\n"));
7890		}
7891	}
7892	free(vdev_name);
7893
7894	if (prs->prs_mapping_memory > 0) {
7895		char mem_buf[7];
7896		zfs_nicenum(prs->prs_mapping_memory, mem_buf, sizeof (mem_buf));
7897		(void) printf(gettext(
7898		    "\t%s memory used for removed device mappings\n"),
7899		    mem_buf);
7900	}
7901}
7902
7903static void
7904print_checkpoint_status(pool_checkpoint_stat_t *pcs)
7905{
7906	time_t start;
7907	char space_buf[7];
7908
7909	if (pcs == NULL || pcs->pcs_state == CS_NONE)
7910		return;
7911
7912	(void) printf(gettext("checkpoint: "));
7913
7914	start = pcs->pcs_start_time;
7915	zfs_nicenum(pcs->pcs_space, space_buf, sizeof (space_buf));
7916
7917	if (pcs->pcs_state == CS_CHECKPOINT_EXISTS) {
7918		char *date = ctime(&start);
7919
7920		/*
7921		 * ctime() adds a newline at the end of the generated
7922		 * string, thus the weird format specifier and the
7923		 * strlen() call used to chop it off from the output.
7924		 */
7925		(void) printf(gettext("created %.*s, consumes %s\n"),
7926		    (int)(strlen(date) - 1), date, space_buf);
7927		return;
7928	}
7929
7930	assert(pcs->pcs_state == CS_CHECKPOINT_DISCARDING);
7931
7932	(void) printf(gettext("discarding, %s remaining.\n"),
7933	    space_buf);
7934}
7935
7936static void
7937print_error_log(zpool_handle_t *zhp)
7938{
7939	nvlist_t *nverrlist = NULL;
7940	nvpair_t *elem;
7941	char *pathname;
7942	size_t len = MAXPATHLEN * 2;
7943
7944	if (zpool_get_errlog(zhp, &nverrlist) != 0)
7945		return;
7946
7947	(void) printf("errors: Permanent errors have been "
7948	    "detected in the following files:\n\n");
7949
7950	pathname = safe_malloc(len);
7951	elem = NULL;
7952	while ((elem = nvlist_next_nvpair(nverrlist, elem)) != NULL) {
7953		nvlist_t *nv;
7954		uint64_t dsobj, obj;
7955
7956		verify(nvpair_value_nvlist(elem, &nv) == 0);
7957		verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_DATASET,
7958		    &dsobj) == 0);
7959		verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_OBJECT,
7960		    &obj) == 0);
7961		zpool_obj_to_path(zhp, dsobj, obj, pathname, len);
7962		(void) printf("%7s %s\n", "", pathname);
7963	}
7964	free(pathname);
7965	nvlist_free(nverrlist);
7966}
7967
7968static void
7969print_spares(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t **spares,
7970    uint_t nspares)
7971{
7972	uint_t i;
7973	char *name;
7974
7975	if (nspares == 0)
7976		return;
7977
7978	(void) printf(gettext("\tspares\n"));
7979
7980	for (i = 0; i < nspares; i++) {
7981		name = zpool_vdev_name(g_zfs, zhp, spares[i],
7982		    cb->cb_name_flags);
7983		print_status_config(zhp, cb, name, spares[i], 2, B_TRUE, NULL);
7984		free(name);
7985	}
7986}
7987
7988static void
7989print_l2cache(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t **l2cache,
7990    uint_t nl2cache)
7991{
7992	uint_t i;
7993	char *name;
7994
7995	if (nl2cache == 0)
7996		return;
7997
7998	(void) printf(gettext("\tcache\n"));
7999
8000	for (i = 0; i < nl2cache; i++) {
8001		name = zpool_vdev_name(g_zfs, zhp, l2cache[i],
8002		    cb->cb_name_flags);
8003		print_status_config(zhp, cb, name, l2cache[i], 2,
8004		    B_FALSE, NULL);
8005		free(name);
8006	}
8007}
8008
8009static void
8010print_dedup_stats(nvlist_t *config)
8011{
8012	ddt_histogram_t *ddh;
8013	ddt_stat_t *dds;
8014	ddt_object_t *ddo;
8015	uint_t c;
8016	char dspace[6], mspace[6];
8017
8018	/*
8019	 * If the pool was faulted then we may not have been able to
8020	 * obtain the config. Otherwise, if we have anything in the dedup
8021	 * table continue processing the stats.
8022	 */
8023	if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_OBJ_STATS,
8024	    (uint64_t **)&ddo, &c) != 0)
8025		return;
8026
8027	(void) printf("\n");
8028	(void) printf(gettext(" dedup: "));
8029	if (ddo->ddo_count == 0) {
8030		(void) printf(gettext("no DDT entries\n"));
8031		return;
8032	}
8033
8034	zfs_nicebytes(ddo->ddo_dspace, dspace, sizeof (dspace));
8035	zfs_nicebytes(ddo->ddo_mspace, mspace, sizeof (mspace));
8036	(void) printf("DDT entries %llu, size %s on disk, %s in core\n",
8037	    (u_longlong_t)ddo->ddo_count,
8038	    dspace,
8039	    mspace);
8040
8041	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_STATS,
8042	    (uint64_t **)&dds, &c) == 0);
8043	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_HISTOGRAM,
8044	    (uint64_t **)&ddh, &c) == 0);
8045	zpool_dump_ddt(dds, ddh);
8046}
8047
8048/*
8049 * Display a summary of pool status.  Displays a summary such as:
8050 *
8051 *        pool: tank
8052 *	status: DEGRADED
8053 *	reason: One or more devices ...
8054 *         see: https://openzfs.github.io/openzfs-docs/msg/ZFS-xxxx-01
8055 *	config:
8056 *		mirror		DEGRADED
8057 *                c1t0d0	OK
8058 *                c2t0d0	UNAVAIL
8059 *
8060 * When given the '-v' option, we print out the complete config.  If the '-e'
8061 * option is specified, then we print out error rate information as well.
8062 */
8063static int
8064status_callback(zpool_handle_t *zhp, void *data)
8065{
8066	status_cbdata_t *cbp = data;
8067	nvlist_t *config, *nvroot;
8068	char *msgid;
8069	zpool_status_t reason;
8070	zpool_errata_t errata;
8071	const char *health;
8072	uint_t c;
8073	vdev_stat_t *vs;
8074
8075	config = zpool_get_config(zhp, NULL);
8076	reason = zpool_get_status(zhp, &msgid, &errata);
8077
8078	cbp->cb_count++;
8079
8080	/*
8081	 * If we were given 'zpool status -x', only report those pools with
8082	 * problems.
8083	 */
8084	if (cbp->cb_explain &&
8085	    (reason == ZPOOL_STATUS_OK ||
8086	    reason == ZPOOL_STATUS_VERSION_OLDER ||
8087	    reason == ZPOOL_STATUS_FEAT_DISABLED ||
8088	    reason == ZPOOL_STATUS_COMPATIBILITY_ERR ||
8089	    reason == ZPOOL_STATUS_INCOMPATIBLE_FEAT)) {
8090		if (!cbp->cb_allpools) {
8091			(void) printf(gettext("pool '%s' is healthy\n"),
8092			    zpool_get_name(zhp));
8093			if (cbp->cb_first)
8094				cbp->cb_first = B_FALSE;
8095		}
8096		return (0);
8097	}
8098
8099	if (cbp->cb_first)
8100		cbp->cb_first = B_FALSE;
8101	else
8102		(void) printf("\n");
8103
8104	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
8105	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
8106	    (uint64_t **)&vs, &c) == 0);
8107
8108	health = zpool_get_state_str(zhp);
8109
8110	printf("  ");
8111	printf_color(ANSI_BOLD, gettext("pool:"));
8112	printf(" %s\n", zpool_get_name(zhp));
8113	printf(" ");
8114	printf_color(ANSI_BOLD, gettext("state: "));
8115
8116	printf_color(health_str_to_color(health), "%s", health);
8117
8118	printf("\n");
8119
8120	switch (reason) {
8121	case ZPOOL_STATUS_MISSING_DEV_R:
8122		printf_color(ANSI_BOLD, gettext("status: "));
8123		printf_color(ANSI_YELLOW, gettext("One or more devices could "
8124		    "not be opened.  Sufficient replicas exist for\n\tthe pool "
8125		    "to continue functioning in a degraded state.\n"));
8126		printf_color(ANSI_BOLD, gettext("action: "));
8127		printf_color(ANSI_YELLOW, gettext("Attach the missing device "
8128		    "and online it using 'zpool online'.\n"));
8129		break;
8130
8131	case ZPOOL_STATUS_MISSING_DEV_NR:
8132		printf_color(ANSI_BOLD, gettext("status: "));
8133		printf_color(ANSI_YELLOW, gettext("One or more devices could "
8134		    "not be opened.  There are insufficient\n\treplicas for the"
8135		    " pool to continue functioning.\n"));
8136		printf_color(ANSI_BOLD, gettext("action: "));
8137		printf_color(ANSI_YELLOW, gettext("Attach the missing device "
8138		    "and online it using 'zpool online'.\n"));
8139		break;
8140
8141	case ZPOOL_STATUS_CORRUPT_LABEL_R:
8142		printf_color(ANSI_BOLD, gettext("status: "));
8143		printf_color(ANSI_YELLOW, gettext("One or more devices could "
8144		    "not be used because the label is missing or\n\tinvalid.  "
8145		    "Sufficient replicas exist for the pool to continue\n\t"
8146		    "functioning in a degraded state.\n"));
8147		printf_color(ANSI_BOLD, gettext("action: "));
8148		printf_color(ANSI_YELLOW, gettext("Replace the device using "
8149		    "'zpool replace'.\n"));
8150		break;
8151
8152	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
8153		printf_color(ANSI_BOLD, gettext("status: "));
8154		printf_color(ANSI_YELLOW, gettext("One or more devices could "
8155		    "not be used because the label is missing \n\tor invalid.  "
8156		    "There are insufficient replicas for the pool to "
8157		    "continue\n\tfunctioning.\n"));
8158		zpool_explain_recover(zpool_get_handle(zhp),
8159		    zpool_get_name(zhp), reason, config);
8160		break;
8161
8162	case ZPOOL_STATUS_FAILING_DEV:
8163		printf_color(ANSI_BOLD, gettext("status: "));
8164		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8165		    "experienced an unrecoverable error.  An\n\tattempt was "
8166		    "made to correct the error.  Applications are "
8167		    "unaffected.\n"));
8168		printf_color(ANSI_BOLD, gettext("action: "));
8169			printf_color(ANSI_YELLOW, gettext("Determine if the "
8170		    "device needs to be replaced, and clear the errors\n\tusing"
8171		    " 'zpool clear' or replace the device with 'zpool "
8172		    "replace'.\n"));
8173		break;
8174
8175	case ZPOOL_STATUS_OFFLINE_DEV:
8176		printf_color(ANSI_BOLD, gettext("status: "));
8177		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8178		    "been taken offline by the administrator.\n\tSufficient "
8179		    "replicas exist for the pool to continue functioning in "
8180		    "a\n\tdegraded state.\n"));
8181		printf_color(ANSI_BOLD, gettext("action: "));
8182		printf_color(ANSI_YELLOW, gettext("Online the device "
8183		    "using 'zpool online' or replace the device with\n\t'zpool "
8184		    "replace'.\n"));
8185		break;
8186
8187	case ZPOOL_STATUS_REMOVED_DEV:
8188		printf_color(ANSI_BOLD, gettext("status: "));
8189		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8190		    "been removed by the administrator.\n\tSufficient "
8191		    "replicas exist for the pool to continue functioning in "
8192		    "a\n\tdegraded state.\n"));
8193		printf_color(ANSI_BOLD, gettext("action: "));
8194		printf_color(ANSI_YELLOW, gettext("Online the device "
8195		    "using zpool online' or replace the device with\n\t'zpool "
8196		    "replace'.\n"));
8197		break;
8198
8199	case ZPOOL_STATUS_RESILVERING:
8200	case ZPOOL_STATUS_REBUILDING:
8201		printf_color(ANSI_BOLD, gettext("status: "));
8202		printf_color(ANSI_YELLOW, gettext("One or more devices is "
8203		    "currently being resilvered.  The pool will\n\tcontinue "
8204		    "to function, possibly in a degraded state.\n"));
8205		printf_color(ANSI_BOLD, gettext("action: "));
8206		printf_color(ANSI_YELLOW, gettext("Wait for the resilver to "
8207		    "complete.\n"));
8208		break;
8209
8210	case ZPOOL_STATUS_REBUILD_SCRUB:
8211		printf_color(ANSI_BOLD, gettext("status: "));
8212		printf_color(ANSI_YELLOW, gettext("One or more devices have "
8213		    "been sequentially resilvered, scrubbing\n\tthe pool "
8214		    "is recommended.\n"));
8215		printf_color(ANSI_BOLD, gettext("action: "));
8216		printf_color(ANSI_YELLOW, gettext("Use 'zpool scrub' to "
8217		    "verify all data checksums.\n"));
8218		break;
8219
8220	case ZPOOL_STATUS_CORRUPT_DATA:
8221		printf_color(ANSI_BOLD, gettext("status: "));
8222		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8223		    "experienced an error resulting in data\n\tcorruption.  "
8224		    "Applications may be affected.\n"));
8225		printf_color(ANSI_BOLD, gettext("action: "));
8226		printf_color(ANSI_YELLOW, gettext("Restore the file in question"
8227		    " if possible.  Otherwise restore the\n\tentire pool from "
8228		    "backup.\n"));
8229		break;
8230
8231	case ZPOOL_STATUS_CORRUPT_POOL:
8232		printf_color(ANSI_BOLD, gettext("status: "));
8233		printf_color(ANSI_YELLOW, gettext("The pool metadata is "
8234		    "corrupted and the pool cannot be opened.\n"));
8235		zpool_explain_recover(zpool_get_handle(zhp),
8236		    zpool_get_name(zhp), reason, config);
8237		break;
8238
8239	case ZPOOL_STATUS_VERSION_OLDER:
8240		printf_color(ANSI_BOLD, gettext("status: "));
8241		printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
8242		    "a legacy on-disk format.  The pool can\n\tstill be used, "
8243		    "but some features are unavailable.\n"));
8244		printf_color(ANSI_BOLD, gettext("action: "));
8245		printf_color(ANSI_YELLOW, gettext("Upgrade the pool using "
8246		    "'zpool upgrade'.  Once this is done, the\n\tpool will no "
8247		    "longer be accessible on software that does not support\n\t"
8248		    "feature flags.\n"));
8249		break;
8250
8251	case ZPOOL_STATUS_VERSION_NEWER:
8252		printf_color(ANSI_BOLD, gettext("status: "));
8253		printf_color(ANSI_YELLOW, gettext("The pool has been upgraded "
8254		    "to a newer, incompatible on-disk version.\n\tThe pool "
8255		    "cannot be accessed on this system.\n"));
8256		printf_color(ANSI_BOLD, gettext("action: "));
8257		printf_color(ANSI_YELLOW, gettext("Access the pool from a "
8258		    "system running more recent software, or\n\trestore the "
8259		    "pool from backup.\n"));
8260		break;
8261
8262	case ZPOOL_STATUS_FEAT_DISABLED:
8263		printf_color(ANSI_BOLD, gettext("status: "));
8264		printf_color(ANSI_YELLOW, gettext("Some supported and "
8265		    "requested features are not enabled on the pool.\n\t"
8266		    "The pool can still be used, but some features are "
8267		    "unavailable.\n"));
8268		printf_color(ANSI_BOLD, gettext("action: "));
8269		printf_color(ANSI_YELLOW, gettext("Enable all features using "
8270		    "'zpool upgrade'. Once this is done,\n\tthe pool may no "
8271		    "longer be accessible by software that does not support\n\t"
8272		    "the features. See zpool-features(7) for details.\n"));
8273		break;
8274
8275	case ZPOOL_STATUS_COMPATIBILITY_ERR:
8276		printf_color(ANSI_BOLD, gettext("status: "));
8277		printf_color(ANSI_YELLOW, gettext("This pool has a "
8278		    "compatibility list specified, but it could not be\n\t"
8279		    "read/parsed at this time. The pool can still be used, "
8280		    "but this\n\tshould be investigated.\n"));
8281		printf_color(ANSI_BOLD, gettext("action: "));
8282		printf_color(ANSI_YELLOW, gettext("Check the value of the "
8283		    "'compatibility' property against the\n\t"
8284		    "appropriate file in " ZPOOL_SYSCONF_COMPAT_D " or "
8285		    ZPOOL_DATA_COMPAT_D ".\n"));
8286		break;
8287
8288	case ZPOOL_STATUS_INCOMPATIBLE_FEAT:
8289		printf_color(ANSI_BOLD, gettext("status: "));
8290		printf_color(ANSI_YELLOW, gettext("One or more features "
8291		    "are enabled on the pool despite not being\n\t"
8292		    "requested by the 'compatibility' property.\n"));
8293		printf_color(ANSI_BOLD, gettext("action: "));
8294		printf_color(ANSI_YELLOW, gettext("Consider setting "
8295		    "'compatibility' to an appropriate value, or\n\t"
8296		    "adding needed features to the relevant file in\n\t"
8297		    ZPOOL_SYSCONF_COMPAT_D " or " ZPOOL_DATA_COMPAT_D ".\n"));
8298		break;
8299
8300	case ZPOOL_STATUS_UNSUP_FEAT_READ:
8301		printf_color(ANSI_BOLD, gettext("status: "));
8302		printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
8303		    "on this system because it uses the\n\tfollowing feature(s)"
8304		    " not supported on this system:\n"));
8305		zpool_print_unsup_feat(config);
8306		(void) printf("\n");
8307		printf_color(ANSI_BOLD, gettext("action: "));
8308		printf_color(ANSI_YELLOW, gettext("Access the pool from a "
8309		    "system that supports the required feature(s),\n\tor "
8310		    "restore the pool from backup.\n"));
8311		break;
8312
8313	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
8314		printf_color(ANSI_BOLD, gettext("status: "));
8315		printf_color(ANSI_YELLOW, gettext("The pool can only be "
8316		    "accessed in read-only mode on this system. It\n\tcannot be"
8317		    " accessed in read-write mode because it uses the "
8318		    "following\n\tfeature(s) not supported on this system:\n"));
8319		zpool_print_unsup_feat(config);
8320		(void) printf("\n");
8321		printf_color(ANSI_BOLD, gettext("action: "));
8322		printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
8323		    "in read-write mode. Import the pool with\n"
8324		    "\t\"-o readonly=on\", access the pool from a system that "
8325		    "supports the\n\trequired feature(s), or restore the "
8326		    "pool from backup.\n"));
8327		break;
8328
8329	case ZPOOL_STATUS_FAULTED_DEV_R:
8330		printf_color(ANSI_BOLD, gettext("status: "));
8331		printf_color(ANSI_YELLOW, gettext("One or more devices are "
8332		    "faulted in response to persistent errors.\n\tSufficient "
8333		    "replicas exist for the pool to continue functioning "
8334		    "in a\n\tdegraded state.\n"));
8335		printf_color(ANSI_BOLD, gettext("action: "));
8336		printf_color(ANSI_YELLOW, gettext("Replace the faulted device, "
8337		    "or use 'zpool clear' to mark the device\n\trepaired.\n"));
8338		break;
8339
8340	case ZPOOL_STATUS_FAULTED_DEV_NR:
8341		printf_color(ANSI_BOLD, gettext("status: "));
8342		printf_color(ANSI_YELLOW, gettext("One or more devices are "
8343		    "faulted in response to persistent errors.  There are "
8344		    "insufficient replicas for the pool to\n\tcontinue "
8345		    "functioning.\n"));
8346		printf_color(ANSI_BOLD, gettext("action: "));
8347		printf_color(ANSI_YELLOW, gettext("Destroy and re-create the "
8348		    "pool from a backup source.  Manually marking the device\n"
8349		    "\trepaired using 'zpool clear' may allow some data "
8350		    "to be recovered.\n"));
8351		break;
8352
8353	case ZPOOL_STATUS_IO_FAILURE_MMP:
8354		printf_color(ANSI_BOLD, gettext("status: "));
8355		printf_color(ANSI_YELLOW, gettext("The pool is suspended "
8356		    "because multihost writes failed or were delayed;\n\t"
8357		    "another system could import the pool undetected.\n"));
8358		printf_color(ANSI_BOLD, gettext("action: "));
8359		printf_color(ANSI_YELLOW, gettext("Make sure the pool's devices"
8360		    " are connected, then reboot your system and\n\timport the "
8361		    "pool.\n"));
8362		break;
8363
8364	case ZPOOL_STATUS_IO_FAILURE_WAIT:
8365	case ZPOOL_STATUS_IO_FAILURE_CONTINUE:
8366		printf_color(ANSI_BOLD, gettext("status: "));
8367		printf_color(ANSI_YELLOW, gettext("One or more devices are "
8368		    "faulted in response to IO failures.\n"));
8369		printf_color(ANSI_BOLD, gettext("action: "));
8370		printf_color(ANSI_YELLOW, gettext("Make sure the affected "
8371		    "devices are connected, then run 'zpool clear'.\n"));
8372		break;
8373
8374	case ZPOOL_STATUS_BAD_LOG:
8375		printf_color(ANSI_BOLD, gettext("status: "));
8376		printf_color(ANSI_YELLOW, gettext("An intent log record "
8377		    "could not be read.\n"
8378		    "\tWaiting for administrator intervention to fix the "
8379		    "faulted pool.\n"));
8380		printf_color(ANSI_BOLD, gettext("action: "));
8381		printf_color(ANSI_YELLOW, gettext("Either restore the affected "
8382		    "device(s) and run 'zpool online',\n"
8383		    "\tor ignore the intent log records by running "
8384		    "'zpool clear'.\n"));
8385		break;
8386
8387	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
8388		(void) printf(gettext("status: One or more devices are "
8389		    "configured to use a non-native block size.\n"
8390		    "\tExpect reduced performance.\n"));
8391		(void) printf(gettext("action: Replace affected devices with "
8392		    "devices that support the\n\tconfigured block size, or "
8393		    "migrate data to a properly configured\n\tpool.\n"));
8394		break;
8395
8396	case ZPOOL_STATUS_HOSTID_MISMATCH:
8397		printf_color(ANSI_BOLD, gettext("status: "));
8398		printf_color(ANSI_YELLOW, gettext("Mismatch between pool hostid"
8399		    " and system hostid on imported pool.\n\tThis pool was "
8400		    "previously imported into a system with a different "
8401		    "hostid,\n\tand then was verbatim imported into this "
8402		    "system.\n"));
8403		printf_color(ANSI_BOLD, gettext("action: "));
8404		printf_color(ANSI_YELLOW, gettext("Export this pool on all "
8405		    "systems on which it is imported.\n"
8406		    "\tThen import it to correct the mismatch.\n"));
8407		break;
8408
8409	case ZPOOL_STATUS_ERRATA:
8410		printf_color(ANSI_BOLD, gettext("status: "));
8411		printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
8412		    errata);
8413
8414		switch (errata) {
8415		case ZPOOL_ERRATA_NONE:
8416			break;
8417
8418		case ZPOOL_ERRATA_ZOL_2094_SCRUB:
8419			printf_color(ANSI_BOLD, gettext("action: "));
8420			printf_color(ANSI_YELLOW, gettext("To correct the issue"
8421			    " run 'zpool scrub'.\n"));
8422			break;
8423
8424		case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
8425			(void) printf(gettext("\tExisting encrypted datasets "
8426			    "contain an on-disk incompatibility\n\twhich "
8427			    "needs to be corrected.\n"));
8428			printf_color(ANSI_BOLD, gettext("action: "));
8429			printf_color(ANSI_YELLOW, gettext("To correct the issue"
8430			    " backup existing encrypted datasets to new\n\t"
8431			    "encrypted datasets and destroy the old ones. "
8432			    "'zfs mount -o ro' can\n\tbe used to temporarily "
8433			    "mount existing encrypted datasets readonly.\n"));
8434			break;
8435
8436		case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
8437			(void) printf(gettext("\tExisting encrypted snapshots "
8438			    "and bookmarks contain an on-disk\n\tincompat"
8439			    "ibility. This may cause on-disk corruption if "
8440			    "they are used\n\twith 'zfs recv'.\n"));
8441			printf_color(ANSI_BOLD, gettext("action: "));
8442			printf_color(ANSI_YELLOW, gettext("To correct the"
8443			    "issue, enable the bookmark_v2 feature. No "
8444			    "additional\n\taction is needed if there are no "
8445			    "encrypted snapshots or bookmarks.\n\tIf preserving"
8446			    "the encrypted snapshots and bookmarks is required,"
8447			    " use\n\ta non-raw send to backup and restore them."
8448			    " Alternately, they may be\n\tremoved to resolve "
8449			    "the incompatibility.\n"));
8450			break;
8451
8452		default:
8453			/*
8454			 * All errata which allow the pool to be imported
8455			 * must contain an action message.
8456			 */
8457			assert(0);
8458		}
8459		break;
8460
8461	default:
8462		/*
8463		 * The remaining errors can't actually be generated, yet.
8464		 */
8465		assert(reason == ZPOOL_STATUS_OK);
8466	}
8467
8468	if (msgid != NULL) {
8469		printf("   ");
8470		printf_color(ANSI_BOLD, gettext("see:"));
8471		printf(gettext(
8472		    " https://openzfs.github.io/openzfs-docs/msg/%s\n"),
8473		    msgid);
8474	}
8475
8476	if (config != NULL) {
8477		uint64_t nerr;
8478		nvlist_t **spares, **l2cache;
8479		uint_t nspares, nl2cache;
8480		pool_checkpoint_stat_t *pcs = NULL;
8481		pool_removal_stat_t *prs = NULL;
8482
8483		print_scan_status(zhp, nvroot);
8484
8485		(void) nvlist_lookup_uint64_array(nvroot,
8486		    ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
8487		print_removal_status(zhp, prs);
8488
8489		(void) nvlist_lookup_uint64_array(nvroot,
8490		    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
8491		print_checkpoint_status(pcs);
8492
8493		cbp->cb_namewidth = max_width(zhp, nvroot, 0, 0,
8494		    cbp->cb_name_flags | VDEV_NAME_TYPE_ID);
8495		if (cbp->cb_namewidth < 10)
8496			cbp->cb_namewidth = 10;
8497
8498		color_start(ANSI_BOLD);
8499		(void) printf(gettext("config:\n\n"));
8500		(void) printf(gettext("\t%-*s  %-8s %5s %5s %5s"),
8501		    cbp->cb_namewidth, "NAME", "STATE", "READ", "WRITE",
8502		    "CKSUM");
8503		color_end();
8504
8505		if (cbp->cb_print_slow_ios) {
8506			printf_color(ANSI_BOLD, " %5s", gettext("SLOW"));
8507		}
8508
8509		if (cbp->vcdl != NULL)
8510			print_cmd_columns(cbp->vcdl, 0);
8511
8512		printf("\n");
8513
8514		print_status_config(zhp, cbp, zpool_get_name(zhp), nvroot, 0,
8515		    B_FALSE, NULL);
8516
8517		print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_DEDUP);
8518		print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
8519		print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_CLASS_LOGS);
8520
8521		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
8522		    &l2cache, &nl2cache) == 0)
8523			print_l2cache(zhp, cbp, l2cache, nl2cache);
8524
8525		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
8526		    &spares, &nspares) == 0)
8527			print_spares(zhp, cbp, spares, nspares);
8528
8529		if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
8530		    &nerr) == 0) {
8531			nvlist_t *nverrlist = NULL;
8532
8533			/*
8534			 * If the approximate error count is small, get a
8535			 * precise count by fetching the entire log and
8536			 * uniquifying the results.
8537			 */
8538			if (nerr > 0 && nerr < 100 && !cbp->cb_verbose &&
8539			    zpool_get_errlog(zhp, &nverrlist) == 0) {
8540				nvpair_t *elem;
8541
8542				elem = NULL;
8543				nerr = 0;
8544				while ((elem = nvlist_next_nvpair(nverrlist,
8545				    elem)) != NULL) {
8546					nerr++;
8547				}
8548			}
8549			nvlist_free(nverrlist);
8550
8551			(void) printf("\n");
8552
8553			if (nerr == 0)
8554				(void) printf(gettext("errors: No known data "
8555				    "errors\n"));
8556			else if (!cbp->cb_verbose)
8557				(void) printf(gettext("errors: %llu data "
8558				    "errors, use '-v' for a list\n"),
8559				    (u_longlong_t)nerr);
8560			else
8561				print_error_log(zhp);
8562		}
8563
8564		if (cbp->cb_dedup_stats)
8565			print_dedup_stats(config);
8566	} else {
8567		(void) printf(gettext("config: The configuration cannot be "
8568		    "determined.\n"));
8569	}
8570
8571	return (0);
8572}
8573
8574/*
8575 * zpool status [-c [script1,script2,...]] [-igLpPstvx] [-T d|u] [pool] ...
8576 *              [interval [count]]
8577 *
8578 *	-c CMD	For each vdev, run command CMD
8579 *	-i	Display vdev initialization status.
8580 *	-g	Display guid for individual vdev name.
8581 *	-L	Follow links when resolving vdev path name.
8582 *	-p	Display values in parsable (exact) format.
8583 *	-P	Display full path for vdev name.
8584 *	-s	Display slow IOs column.
8585 *	-v	Display complete error logs
8586 *	-x	Display only pools with potential problems
8587 *	-D	Display dedup status (undocumented)
8588 *	-t	Display vdev TRIM status.
8589 *	-T	Display a timestamp in date(1) or Unix format
8590 *
8591 * Describes the health status of all pools or some subset.
8592 */
8593int
8594zpool_do_status(int argc, char **argv)
8595{
8596	int c;
8597	int ret;
8598	float interval = 0;
8599	unsigned long count = 0;
8600	status_cbdata_t cb = { 0 };
8601	char *cmd = NULL;
8602
8603	/* check options */
8604	while ((c = getopt(argc, argv, "c:igLpPsvxDtT:")) != -1) {
8605		switch (c) {
8606		case 'c':
8607			if (cmd != NULL) {
8608				fprintf(stderr,
8609				    gettext("Can't set -c flag twice\n"));
8610				exit(1);
8611			}
8612
8613			if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
8614			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
8615				fprintf(stderr, gettext(
8616				    "Can't run -c, disabled by "
8617				    "ZPOOL_SCRIPTS_ENABLED.\n"));
8618				exit(1);
8619			}
8620
8621			if ((getuid() <= 0 || geteuid() <= 0) &&
8622			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
8623				fprintf(stderr, gettext(
8624				    "Can't run -c with root privileges "
8625				    "unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
8626				exit(1);
8627			}
8628			cmd = optarg;
8629			break;
8630		case 'i':
8631			cb.cb_print_vdev_init = B_TRUE;
8632			break;
8633		case 'g':
8634			cb.cb_name_flags |= VDEV_NAME_GUID;
8635			break;
8636		case 'L':
8637			cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS;
8638			break;
8639		case 'p':
8640			cb.cb_literal = B_TRUE;
8641			break;
8642		case 'P':
8643			cb.cb_name_flags |= VDEV_NAME_PATH;
8644			break;
8645		case 's':
8646			cb.cb_print_slow_ios = B_TRUE;
8647			break;
8648		case 'v':
8649			cb.cb_verbose = B_TRUE;
8650			break;
8651		case 'x':
8652			cb.cb_explain = B_TRUE;
8653			break;
8654		case 'D':
8655			cb.cb_dedup_stats = B_TRUE;
8656			break;
8657		case 't':
8658			cb.cb_print_vdev_trim = B_TRUE;
8659			break;
8660		case 'T':
8661			get_timestamp_arg(*optarg);
8662			break;
8663		case '?':
8664			if (optopt == 'c') {
8665				print_zpool_script_list("status");
8666				exit(0);
8667			} else {
8668				fprintf(stderr,
8669				    gettext("invalid option '%c'\n"), optopt);
8670			}
8671			usage(B_FALSE);
8672		}
8673	}
8674
8675	argc -= optind;
8676	argv += optind;
8677
8678	get_interval_count(&argc, argv, &interval, &count);
8679
8680	if (argc == 0)
8681		cb.cb_allpools = B_TRUE;
8682
8683	cb.cb_first = B_TRUE;
8684	cb.cb_print_status = B_TRUE;
8685
8686	for (;;) {
8687		if (timestamp_fmt != NODATE)
8688			print_timestamp(timestamp_fmt);
8689
8690		if (cmd != NULL)
8691			cb.vcdl = all_pools_for_each_vdev_run(argc, argv, cmd,
8692			    NULL, NULL, 0, 0);
8693
8694		ret = for_each_pool(argc, argv, B_TRUE, NULL, cb.cb_literal,
8695		    status_callback, &cb);
8696
8697		if (cb.vcdl != NULL)
8698			free_vdev_cmd_data_list(cb.vcdl);
8699
8700		if (argc == 0 && cb.cb_count == 0)
8701			(void) fprintf(stderr, gettext("no pools available\n"));
8702		else if (cb.cb_explain && cb.cb_first && cb.cb_allpools)
8703			(void) printf(gettext("all pools are healthy\n"));
8704
8705		if (ret != 0)
8706			return (ret);
8707
8708		if (interval == 0)
8709			break;
8710
8711		if (count != 0 && --count == 0)
8712			break;
8713
8714		(void) fsleep(interval);
8715	}
8716
8717	return (0);
8718}
8719
8720typedef struct upgrade_cbdata {
8721	int	cb_first;
8722	int	cb_argc;
8723	uint64_t cb_version;
8724	char	**cb_argv;
8725} upgrade_cbdata_t;
8726
8727static int
8728check_unsupp_fs(zfs_handle_t *zhp, void *unsupp_fs)
8729{
8730	int zfs_version = (int)zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
8731	int *count = (int *)unsupp_fs;
8732
8733	if (zfs_version > ZPL_VERSION) {
8734		(void) printf(gettext("%s (v%d) is not supported by this "
8735		    "implementation of ZFS.\n"),
8736		    zfs_get_name(zhp), zfs_version);
8737		(*count)++;
8738	}
8739
8740	zfs_iter_filesystems(zhp, check_unsupp_fs, unsupp_fs);
8741
8742	zfs_close(zhp);
8743
8744	return (0);
8745}
8746
8747static int
8748upgrade_version(zpool_handle_t *zhp, uint64_t version)
8749{
8750	int ret;
8751	nvlist_t *config;
8752	uint64_t oldversion;
8753	int unsupp_fs = 0;
8754
8755	config = zpool_get_config(zhp, NULL);
8756	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8757	    &oldversion) == 0);
8758
8759	char compat[ZFS_MAXPROPLEN];
8760	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat,
8761	    ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
8762		compat[0] = '\0';
8763
8764	assert(SPA_VERSION_IS_SUPPORTED(oldversion));
8765	assert(oldversion < version);
8766
8767	ret = zfs_iter_root(zpool_get_handle(zhp), check_unsupp_fs, &unsupp_fs);
8768	if (ret != 0)
8769		return (ret);
8770
8771	if (unsupp_fs) {
8772		(void) fprintf(stderr, gettext("Upgrade not performed due "
8773		    "to %d unsupported filesystems (max v%d).\n"),
8774		    unsupp_fs, (int)ZPL_VERSION);
8775		return (1);
8776	}
8777
8778	if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) {
8779		(void) fprintf(stderr, gettext("Upgrade not performed because "
8780		    "'compatibility' property set to '"
8781		    ZPOOL_COMPAT_LEGACY "'.\n"));
8782		return (1);
8783	}
8784
8785	ret = zpool_upgrade(zhp, version);
8786	if (ret != 0)
8787		return (ret);
8788
8789	if (version >= SPA_VERSION_FEATURES) {
8790		(void) printf(gettext("Successfully upgraded "
8791		    "'%s' from version %llu to feature flags.\n"),
8792		    zpool_get_name(zhp), (u_longlong_t)oldversion);
8793	} else {
8794		(void) printf(gettext("Successfully upgraded "
8795		    "'%s' from version %llu to version %llu.\n"),
8796		    zpool_get_name(zhp), (u_longlong_t)oldversion,
8797		    (u_longlong_t)version);
8798	}
8799
8800	return (0);
8801}
8802
8803static int
8804upgrade_enable_all(zpool_handle_t *zhp, int *countp)
8805{
8806	int i, ret, count;
8807	boolean_t firstff = B_TRUE;
8808	nvlist_t *enabled = zpool_get_features(zhp);
8809
8810	char compat[ZFS_MAXPROPLEN];
8811	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat,
8812	    ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
8813		compat[0] = '\0';
8814
8815	boolean_t requested_features[SPA_FEATURES];
8816	if (zpool_do_load_compat(compat, requested_features) !=
8817	    ZPOOL_COMPATIBILITY_OK)
8818		return (-1);
8819
8820	count = 0;
8821	for (i = 0; i < SPA_FEATURES; i++) {
8822		const char *fname = spa_feature_table[i].fi_uname;
8823		const char *fguid = spa_feature_table[i].fi_guid;
8824
8825		if (!spa_feature_table[i].fi_zfs_mod_supported)
8826			continue;
8827
8828		if (!nvlist_exists(enabled, fguid) && requested_features[i]) {
8829			char *propname;
8830			verify(-1 != asprintf(&propname, "feature@%s", fname));
8831			ret = zpool_set_prop(zhp, propname,
8832			    ZFS_FEATURE_ENABLED);
8833			if (ret != 0) {
8834				free(propname);
8835				return (ret);
8836			}
8837			count++;
8838
8839			if (firstff) {
8840				(void) printf(gettext("Enabled the "
8841				    "following features on '%s':\n"),
8842				    zpool_get_name(zhp));
8843				firstff = B_FALSE;
8844			}
8845			(void) printf(gettext("  %s\n"), fname);
8846			free(propname);
8847		}
8848	}
8849
8850	if (countp != NULL)
8851		*countp = count;
8852	return (0);
8853}
8854
8855static int
8856upgrade_cb(zpool_handle_t *zhp, void *arg)
8857{
8858	upgrade_cbdata_t *cbp = arg;
8859	nvlist_t *config;
8860	uint64_t version;
8861	boolean_t modified_pool = B_FALSE;
8862	int ret;
8863
8864	config = zpool_get_config(zhp, NULL);
8865	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8866	    &version) == 0);
8867
8868	assert(SPA_VERSION_IS_SUPPORTED(version));
8869
8870	if (version < cbp->cb_version) {
8871		cbp->cb_first = B_FALSE;
8872		ret = upgrade_version(zhp, cbp->cb_version);
8873		if (ret != 0)
8874			return (ret);
8875		modified_pool = B_TRUE;
8876
8877		/*
8878		 * If they did "zpool upgrade -a", then we could
8879		 * be doing ioctls to different pools.  We need
8880		 * to log this history once to each pool, and bypass
8881		 * the normal history logging that happens in main().
8882		 */
8883		(void) zpool_log_history(g_zfs, history_str);
8884		log_history = B_FALSE;
8885	}
8886
8887	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
8888		int count;
8889		ret = upgrade_enable_all(zhp, &count);
8890		if (ret != 0)
8891			return (ret);
8892
8893		if (count > 0) {
8894			cbp->cb_first = B_FALSE;
8895			modified_pool = B_TRUE;
8896		}
8897	}
8898
8899	if (modified_pool) {
8900		(void) printf("\n");
8901		(void) after_zpool_upgrade(zhp);
8902	}
8903
8904	return (0);
8905}
8906
8907static int
8908upgrade_list_older_cb(zpool_handle_t *zhp, void *arg)
8909{
8910	upgrade_cbdata_t *cbp = arg;
8911	nvlist_t *config;
8912	uint64_t version;
8913
8914	config = zpool_get_config(zhp, NULL);
8915	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8916	    &version) == 0);
8917
8918	assert(SPA_VERSION_IS_SUPPORTED(version));
8919
8920	if (version < SPA_VERSION_FEATURES) {
8921		if (cbp->cb_first) {
8922			(void) printf(gettext("The following pools are "
8923			    "formatted with legacy version numbers and can\n"
8924			    "be upgraded to use feature flags.  After "
8925			    "being upgraded, these pools\nwill no "
8926			    "longer be accessible by software that does not "
8927			    "support feature\nflags.\n\n"
8928			    "Note that setting a pool's 'compatibility' "
8929			    "feature to '" ZPOOL_COMPAT_LEGACY "' will\n"
8930			    "inhibit upgrades.\n\n"));
8931			(void) printf(gettext("VER  POOL\n"));
8932			(void) printf(gettext("---  ------------\n"));
8933			cbp->cb_first = B_FALSE;
8934		}
8935
8936		(void) printf("%2llu   %s\n", (u_longlong_t)version,
8937		    zpool_get_name(zhp));
8938	}
8939
8940	return (0);
8941}
8942
8943static int
8944upgrade_list_disabled_cb(zpool_handle_t *zhp, void *arg)
8945{
8946	upgrade_cbdata_t *cbp = arg;
8947	nvlist_t *config;
8948	uint64_t version;
8949
8950	config = zpool_get_config(zhp, NULL);
8951	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8952	    &version) == 0);
8953
8954	if (version >= SPA_VERSION_FEATURES) {
8955		int i;
8956		boolean_t poolfirst = B_TRUE;
8957		nvlist_t *enabled = zpool_get_features(zhp);
8958
8959		for (i = 0; i < SPA_FEATURES; i++) {
8960			const char *fguid = spa_feature_table[i].fi_guid;
8961			const char *fname = spa_feature_table[i].fi_uname;
8962
8963			if (!spa_feature_table[i].fi_zfs_mod_supported)
8964				continue;
8965
8966			if (!nvlist_exists(enabled, fguid)) {
8967				if (cbp->cb_first) {
8968					(void) printf(gettext("\nSome "
8969					    "supported features are not "
8970					    "enabled on the following pools. "
8971					    "Once a\nfeature is enabled the "
8972					    "pool may become incompatible with "
8973					    "software\nthat does not support "
8974					    "the feature. See "
8975					    "zpool-features(7) for "
8976					    "details.\n\n"
8977					    "Note that the pool "
8978					    "'compatibility' feature can be "
8979					    "used to inhibit\nfeature "
8980					    "upgrades.\n\n"));
8981					(void) printf(gettext("POOL  "
8982					    "FEATURE\n"));
8983					(void) printf(gettext("------"
8984					    "---------\n"));
8985					cbp->cb_first = B_FALSE;
8986				}
8987
8988				if (poolfirst) {
8989					(void) printf(gettext("%s\n"),
8990					    zpool_get_name(zhp));
8991					poolfirst = B_FALSE;
8992				}
8993
8994				(void) printf(gettext("      %s\n"), fname);
8995			}
8996			/*
8997			 * If they did "zpool upgrade -a", then we could
8998			 * be doing ioctls to different pools.  We need
8999			 * to log this history once to each pool, and bypass
9000			 * the normal history logging that happens in main().
9001			 */
9002			(void) zpool_log_history(g_zfs, history_str);
9003			log_history = B_FALSE;
9004		}
9005	}
9006
9007	return (0);
9008}
9009
9010/* ARGSUSED */
9011static int
9012upgrade_one(zpool_handle_t *zhp, void *data)
9013{
9014	boolean_t modified_pool = B_FALSE;
9015	upgrade_cbdata_t *cbp = data;
9016	uint64_t cur_version;
9017	int ret;
9018
9019	if (strcmp("log", zpool_get_name(zhp)) == 0) {
9020		(void) fprintf(stderr, gettext("'log' is now a reserved word\n"
9021		    "Pool 'log' must be renamed using export and import"
9022		    " to upgrade.\n"));
9023		return (1);
9024	}
9025
9026	cur_version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
9027	if (cur_version > cbp->cb_version) {
9028		(void) printf(gettext("Pool '%s' is already formatted "
9029		    "using more current version '%llu'.\n\n"),
9030		    zpool_get_name(zhp), (u_longlong_t)cur_version);
9031		return (0);
9032	}
9033
9034	if (cbp->cb_version != SPA_VERSION && cur_version == cbp->cb_version) {
9035		(void) printf(gettext("Pool '%s' is already formatted "
9036		    "using version %llu.\n\n"), zpool_get_name(zhp),
9037		    (u_longlong_t)cbp->cb_version);
9038		return (0);
9039	}
9040
9041	if (cur_version != cbp->cb_version) {
9042		modified_pool = B_TRUE;
9043		ret = upgrade_version(zhp, cbp->cb_version);
9044		if (ret != 0)
9045			return (ret);
9046	}
9047
9048	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
9049		int count = 0;
9050		ret = upgrade_enable_all(zhp, &count);
9051		if (ret != 0)
9052			return (ret);
9053
9054		if (count != 0) {
9055			modified_pool = B_TRUE;
9056		} else if (cur_version == SPA_VERSION) {
9057			(void) printf(gettext("Pool '%s' already has all "
9058			    "supported and requested features enabled.\n"),
9059			    zpool_get_name(zhp));
9060		}
9061	}
9062
9063	if (modified_pool) {
9064		(void) printf("\n");
9065		(void) after_zpool_upgrade(zhp);
9066	}
9067
9068	return (0);
9069}
9070
9071/*
9072 * zpool upgrade
9073 * zpool upgrade -v
9074 * zpool upgrade [-V version] <-a | pool ...>
9075 *
9076 * With no arguments, display downrev'd ZFS pool available for upgrade.
9077 * Individual pools can be upgraded by specifying the pool, and '-a' will
9078 * upgrade all pools.
9079 */
9080int
9081zpool_do_upgrade(int argc, char **argv)
9082{
9083	int c;
9084	upgrade_cbdata_t cb = { 0 };
9085	int ret = 0;
9086	boolean_t showversions = B_FALSE;
9087	boolean_t upgradeall = B_FALSE;
9088	char *end;
9089
9090
9091	/* check options */
9092	while ((c = getopt(argc, argv, ":avV:")) != -1) {
9093		switch (c) {
9094		case 'a':
9095			upgradeall = B_TRUE;
9096			break;
9097		case 'v':
9098			showversions = B_TRUE;
9099			break;
9100		case 'V':
9101			cb.cb_version = strtoll(optarg, &end, 10);
9102			if (*end != '\0' ||
9103			    !SPA_VERSION_IS_SUPPORTED(cb.cb_version)) {
9104				(void) fprintf(stderr,
9105				    gettext("invalid version '%s'\n"), optarg);
9106				usage(B_FALSE);
9107			}
9108			break;
9109		case ':':
9110			(void) fprintf(stderr, gettext("missing argument for "
9111			    "'%c' option\n"), optopt);
9112			usage(B_FALSE);
9113			break;
9114		case '?':
9115			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9116			    optopt);
9117			usage(B_FALSE);
9118		}
9119	}
9120
9121	cb.cb_argc = argc;
9122	cb.cb_argv = argv;
9123	argc -= optind;
9124	argv += optind;
9125
9126	if (cb.cb_version == 0) {
9127		cb.cb_version = SPA_VERSION;
9128	} else if (!upgradeall && argc == 0) {
9129		(void) fprintf(stderr, gettext("-V option is "
9130		    "incompatible with other arguments\n"));
9131		usage(B_FALSE);
9132	}
9133
9134	if (showversions) {
9135		if (upgradeall || argc != 0) {
9136			(void) fprintf(stderr, gettext("-v option is "
9137			    "incompatible with other arguments\n"));
9138			usage(B_FALSE);
9139		}
9140	} else if (upgradeall) {
9141		if (argc != 0) {
9142			(void) fprintf(stderr, gettext("-a option should not "
9143			    "be used along with a pool name\n"));
9144			usage(B_FALSE);
9145		}
9146	}
9147
9148	(void) printf(gettext("This system supports ZFS pool feature "
9149	    "flags.\n\n"));
9150	if (showversions) {
9151		int i;
9152
9153		(void) printf(gettext("The following features are "
9154		    "supported:\n\n"));
9155		(void) printf(gettext("FEAT DESCRIPTION\n"));
9156		(void) printf("----------------------------------------------"
9157		    "---------------\n");
9158		for (i = 0; i < SPA_FEATURES; i++) {
9159			zfeature_info_t *fi = &spa_feature_table[i];
9160			if (!fi->fi_zfs_mod_supported)
9161				continue;
9162			const char *ro =
9163			    (fi->fi_flags & ZFEATURE_FLAG_READONLY_COMPAT) ?
9164			    " (read-only compatible)" : "";
9165
9166			(void) printf("%-37s%s\n", fi->fi_uname, ro);
9167			(void) printf("     %s\n", fi->fi_desc);
9168		}
9169		(void) printf("\n");
9170
9171		(void) printf(gettext("The following legacy versions are also "
9172		    "supported:\n\n"));
9173		(void) printf(gettext("VER  DESCRIPTION\n"));
9174		(void) printf("---  -----------------------------------------"
9175		    "---------------\n");
9176		(void) printf(gettext(" 1   Initial ZFS version\n"));
9177		(void) printf(gettext(" 2   Ditto blocks "
9178		    "(replicated metadata)\n"));
9179		(void) printf(gettext(" 3   Hot spares and double parity "
9180		    "RAID-Z\n"));
9181		(void) printf(gettext(" 4   zpool history\n"));
9182		(void) printf(gettext(" 5   Compression using the gzip "
9183		    "algorithm\n"));
9184		(void) printf(gettext(" 6   bootfs pool property\n"));
9185		(void) printf(gettext(" 7   Separate intent log devices\n"));
9186		(void) printf(gettext(" 8   Delegated administration\n"));
9187		(void) printf(gettext(" 9   refquota and refreservation "
9188		    "properties\n"));
9189		(void) printf(gettext(" 10  Cache devices\n"));
9190		(void) printf(gettext(" 11  Improved scrub performance\n"));
9191		(void) printf(gettext(" 12  Snapshot properties\n"));
9192		(void) printf(gettext(" 13  snapused property\n"));
9193		(void) printf(gettext(" 14  passthrough-x aclinherit\n"));
9194		(void) printf(gettext(" 15  user/group space accounting\n"));
9195		(void) printf(gettext(" 16  stmf property support\n"));
9196		(void) printf(gettext(" 17  Triple-parity RAID-Z\n"));
9197		(void) printf(gettext(" 18  Snapshot user holds\n"));
9198		(void) printf(gettext(" 19  Log device removal\n"));
9199		(void) printf(gettext(" 20  Compression using zle "
9200		    "(zero-length encoding)\n"));
9201		(void) printf(gettext(" 21  Deduplication\n"));
9202		(void) printf(gettext(" 22  Received properties\n"));
9203		(void) printf(gettext(" 23  Slim ZIL\n"));
9204		(void) printf(gettext(" 24  System attributes\n"));
9205		(void) printf(gettext(" 25  Improved scrub stats\n"));
9206		(void) printf(gettext(" 26  Improved snapshot deletion "
9207		    "performance\n"));
9208		(void) printf(gettext(" 27  Improved snapshot creation "
9209		    "performance\n"));
9210		(void) printf(gettext(" 28  Multiple vdev replacements\n"));
9211		(void) printf(gettext("\nFor more information on a particular "
9212		    "version, including supported releases,\n"));
9213		(void) printf(gettext("see the ZFS Administration Guide.\n\n"));
9214	} else if (argc == 0 && upgradeall) {
9215		cb.cb_first = B_TRUE;
9216		ret = zpool_iter(g_zfs, upgrade_cb, &cb);
9217		if (ret == 0 && cb.cb_first) {
9218			if (cb.cb_version == SPA_VERSION) {
9219				(void) printf(gettext("All pools are already "
9220				    "formatted using feature flags.\n\n"));
9221				(void) printf(gettext("Every feature flags "
9222				    "pool already has all supported and "
9223				    "requested features enabled.\n"));
9224			} else {
9225				(void) printf(gettext("All pools are already "
9226				    "formatted with version %llu or higher.\n"),
9227				    (u_longlong_t)cb.cb_version);
9228			}
9229		}
9230	} else if (argc == 0) {
9231		cb.cb_first = B_TRUE;
9232		ret = zpool_iter(g_zfs, upgrade_list_older_cb, &cb);
9233		assert(ret == 0);
9234
9235		if (cb.cb_first) {
9236			(void) printf(gettext("All pools are formatted "
9237			    "using feature flags.\n\n"));
9238		} else {
9239			(void) printf(gettext("\nUse 'zpool upgrade -v' "
9240			    "for a list of available legacy versions.\n"));
9241		}
9242
9243		cb.cb_first = B_TRUE;
9244		ret = zpool_iter(g_zfs, upgrade_list_disabled_cb, &cb);
9245		assert(ret == 0);
9246
9247		if (cb.cb_first) {
9248			(void) printf(gettext("Every feature flags pool has "
9249			    "all supported and requested features enabled.\n"));
9250		} else {
9251			(void) printf(gettext("\n"));
9252		}
9253	} else {
9254		ret = for_each_pool(argc, argv, B_FALSE, NULL, B_FALSE,
9255		    upgrade_one, &cb);
9256	}
9257
9258	return (ret);
9259}
9260
9261typedef struct hist_cbdata {
9262	boolean_t first;
9263	boolean_t longfmt;
9264	boolean_t internal;
9265} hist_cbdata_t;
9266
9267static void
9268print_history_records(nvlist_t *nvhis, hist_cbdata_t *cb)
9269{
9270	nvlist_t **records;
9271	uint_t numrecords;
9272	int i;
9273
9274	verify(nvlist_lookup_nvlist_array(nvhis, ZPOOL_HIST_RECORD,
9275	    &records, &numrecords) == 0);
9276	for (i = 0; i < numrecords; i++) {
9277		nvlist_t *rec = records[i];
9278		char tbuf[64] = "";
9279
9280		if (nvlist_exists(rec, ZPOOL_HIST_TIME)) {
9281			time_t tsec;
9282			struct tm t;
9283
9284			tsec = fnvlist_lookup_uint64(records[i],
9285			    ZPOOL_HIST_TIME);
9286			(void) localtime_r(&tsec, &t);
9287			(void) strftime(tbuf, sizeof (tbuf), "%F.%T", &t);
9288		}
9289
9290		if (nvlist_exists(rec, ZPOOL_HIST_ELAPSED_NS)) {
9291			uint64_t elapsed_ns = fnvlist_lookup_int64(records[i],
9292			    ZPOOL_HIST_ELAPSED_NS);
9293			(void) snprintf(tbuf + strlen(tbuf),
9294			    sizeof (tbuf) - strlen(tbuf),
9295			    " (%lldms)", (long long)elapsed_ns / 1000 / 1000);
9296		}
9297
9298		if (nvlist_exists(rec, ZPOOL_HIST_CMD)) {
9299			(void) printf("%s %s", tbuf,
9300			    fnvlist_lookup_string(rec, ZPOOL_HIST_CMD));
9301		} else if (nvlist_exists(rec, ZPOOL_HIST_INT_EVENT)) {
9302			int ievent =
9303			    fnvlist_lookup_uint64(rec, ZPOOL_HIST_INT_EVENT);
9304			if (!cb->internal)
9305				continue;
9306			if (ievent >= ZFS_NUM_LEGACY_HISTORY_EVENTS) {
9307				(void) printf("%s unrecognized record:\n",
9308				    tbuf);
9309				dump_nvlist(rec, 4);
9310				continue;
9311			}
9312			(void) printf("%s [internal %s txg:%lld] %s", tbuf,
9313			    zfs_history_event_names[ievent],
9314			    (longlong_t)fnvlist_lookup_uint64(
9315			    rec, ZPOOL_HIST_TXG),
9316			    fnvlist_lookup_string(rec, ZPOOL_HIST_INT_STR));
9317		} else if (nvlist_exists(rec, ZPOOL_HIST_INT_NAME)) {
9318			if (!cb->internal)
9319				continue;
9320			(void) printf("%s [txg:%lld] %s", tbuf,
9321			    (longlong_t)fnvlist_lookup_uint64(
9322			    rec, ZPOOL_HIST_TXG),
9323			    fnvlist_lookup_string(rec, ZPOOL_HIST_INT_NAME));
9324			if (nvlist_exists(rec, ZPOOL_HIST_DSNAME)) {
9325				(void) printf(" %s (%llu)",
9326				    fnvlist_lookup_string(rec,
9327				    ZPOOL_HIST_DSNAME),
9328				    (u_longlong_t)fnvlist_lookup_uint64(rec,
9329				    ZPOOL_HIST_DSID));
9330			}
9331			(void) printf(" %s", fnvlist_lookup_string(rec,
9332			    ZPOOL_HIST_INT_STR));
9333		} else if (nvlist_exists(rec, ZPOOL_HIST_IOCTL)) {
9334			if (!cb->internal)
9335				continue;
9336			(void) printf("%s ioctl %s\n", tbuf,
9337			    fnvlist_lookup_string(rec, ZPOOL_HIST_IOCTL));
9338			if (nvlist_exists(rec, ZPOOL_HIST_INPUT_NVL)) {
9339				(void) printf("    input:\n");
9340				dump_nvlist(fnvlist_lookup_nvlist(rec,
9341				    ZPOOL_HIST_INPUT_NVL), 8);
9342			}
9343			if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_NVL)) {
9344				(void) printf("    output:\n");
9345				dump_nvlist(fnvlist_lookup_nvlist(rec,
9346				    ZPOOL_HIST_OUTPUT_NVL), 8);
9347			}
9348			if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_SIZE)) {
9349				(void) printf("    output nvlist omitted; "
9350				    "original size: %lldKB\n",
9351				    (longlong_t)fnvlist_lookup_int64(rec,
9352				    ZPOOL_HIST_OUTPUT_SIZE) / 1024);
9353			}
9354			if (nvlist_exists(rec, ZPOOL_HIST_ERRNO)) {
9355				(void) printf("    errno: %lld\n",
9356				    (longlong_t)fnvlist_lookup_int64(rec,
9357				    ZPOOL_HIST_ERRNO));
9358			}
9359		} else {
9360			if (!cb->internal)
9361				continue;
9362			(void) printf("%s unrecognized record:\n", tbuf);
9363			dump_nvlist(rec, 4);
9364		}
9365
9366		if (!cb->longfmt) {
9367			(void) printf("\n");
9368			continue;
9369		}
9370		(void) printf(" [");
9371		if (nvlist_exists(rec, ZPOOL_HIST_WHO)) {
9372			uid_t who = fnvlist_lookup_uint64(rec, ZPOOL_HIST_WHO);
9373			struct passwd *pwd = getpwuid(who);
9374			(void) printf("user %d ", (int)who);
9375			if (pwd != NULL)
9376				(void) printf("(%s) ", pwd->pw_name);
9377		}
9378		if (nvlist_exists(rec, ZPOOL_HIST_HOST)) {
9379			(void) printf("on %s",
9380			    fnvlist_lookup_string(rec, ZPOOL_HIST_HOST));
9381		}
9382		if (nvlist_exists(rec, ZPOOL_HIST_ZONE)) {
9383			(void) printf(":%s",
9384			    fnvlist_lookup_string(rec, ZPOOL_HIST_ZONE));
9385		}
9386
9387		(void) printf("]");
9388		(void) printf("\n");
9389	}
9390}
9391
9392/*
9393 * Print out the command history for a specific pool.
9394 */
9395static int
9396get_history_one(zpool_handle_t *zhp, void *data)
9397{
9398	nvlist_t *nvhis;
9399	int ret;
9400	hist_cbdata_t *cb = (hist_cbdata_t *)data;
9401	uint64_t off = 0;
9402	boolean_t eof = B_FALSE;
9403
9404	cb->first = B_FALSE;
9405
9406	(void) printf(gettext("History for '%s':\n"), zpool_get_name(zhp));
9407
9408	while (!eof) {
9409		if ((ret = zpool_get_history(zhp, &nvhis, &off, &eof)) != 0)
9410			return (ret);
9411
9412		print_history_records(nvhis, cb);
9413		nvlist_free(nvhis);
9414	}
9415	(void) printf("\n");
9416
9417	return (ret);
9418}
9419
9420/*
9421 * zpool history <pool>
9422 *
9423 * Displays the history of commands that modified pools.
9424 */
9425int
9426zpool_do_history(int argc, char **argv)
9427{
9428	hist_cbdata_t cbdata = { 0 };
9429	int ret;
9430	int c;
9431
9432	cbdata.first = B_TRUE;
9433	/* check options */
9434	while ((c = getopt(argc, argv, "li")) != -1) {
9435		switch (c) {
9436		case 'l':
9437			cbdata.longfmt = B_TRUE;
9438			break;
9439		case 'i':
9440			cbdata.internal = B_TRUE;
9441			break;
9442		case '?':
9443			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9444			    optopt);
9445			usage(B_FALSE);
9446		}
9447	}
9448	argc -= optind;
9449	argv += optind;
9450
9451	ret = for_each_pool(argc, argv, B_FALSE, NULL, B_FALSE, get_history_one,
9452	    &cbdata);
9453
9454	if (argc == 0 && cbdata.first == B_TRUE) {
9455		(void) fprintf(stderr, gettext("no pools available\n"));
9456		return (0);
9457	}
9458
9459	return (ret);
9460}
9461
9462typedef struct ev_opts {
9463	int verbose;
9464	int scripted;
9465	int follow;
9466	int clear;
9467	char poolname[ZFS_MAX_DATASET_NAME_LEN];
9468} ev_opts_t;
9469
9470static void
9471zpool_do_events_short(nvlist_t *nvl, ev_opts_t *opts)
9472{
9473	char ctime_str[26], str[32], *ptr;
9474	int64_t *tv;
9475	uint_t n;
9476
9477	verify(nvlist_lookup_int64_array(nvl, FM_EREPORT_TIME, &tv, &n) == 0);
9478	memset(str, ' ', 32);
9479	(void) ctime_r((const time_t *)&tv[0], ctime_str);
9480	(void) memcpy(str, ctime_str+4,  6);		/* 'Jun 30' */
9481	(void) memcpy(str+7, ctime_str+20, 4);		/* '1993' */
9482	(void) memcpy(str+12, ctime_str+11, 8);		/* '21:49:08' */
9483	(void) sprintf(str+20, ".%09lld", (longlong_t)tv[1]); /* '.123456789' */
9484	if (opts->scripted)
9485		(void) printf(gettext("%s\t"), str);
9486	else
9487		(void) printf(gettext("%s "), str);
9488
9489	verify(nvlist_lookup_string(nvl, FM_CLASS, &ptr) == 0);
9490	(void) printf(gettext("%s\n"), ptr);
9491}
9492
9493static void
9494zpool_do_events_nvprint(nvlist_t *nvl, int depth)
9495{
9496	nvpair_t *nvp;
9497
9498	for (nvp = nvlist_next_nvpair(nvl, NULL);
9499	    nvp != NULL; nvp = nvlist_next_nvpair(nvl, nvp)) {
9500
9501		data_type_t type = nvpair_type(nvp);
9502		const char *name = nvpair_name(nvp);
9503
9504		boolean_t b;
9505		uint8_t i8;
9506		uint16_t i16;
9507		uint32_t i32;
9508		uint64_t i64;
9509		char *str;
9510		nvlist_t *cnv;
9511
9512		printf(gettext("%*s%s = "), depth, "", name);
9513
9514		switch (type) {
9515		case DATA_TYPE_BOOLEAN:
9516			printf(gettext("%s"), "1");
9517			break;
9518
9519		case DATA_TYPE_BOOLEAN_VALUE:
9520			(void) nvpair_value_boolean_value(nvp, &b);
9521			printf(gettext("%s"), b ? "1" : "0");
9522			break;
9523
9524		case DATA_TYPE_BYTE:
9525			(void) nvpair_value_byte(nvp, &i8);
9526			printf(gettext("0x%x"), i8);
9527			break;
9528
9529		case DATA_TYPE_INT8:
9530			(void) nvpair_value_int8(nvp, (void *)&i8);
9531			printf(gettext("0x%x"), i8);
9532			break;
9533
9534		case DATA_TYPE_UINT8:
9535			(void) nvpair_value_uint8(nvp, &i8);
9536			printf(gettext("0x%x"), i8);
9537			break;
9538
9539		case DATA_TYPE_INT16:
9540			(void) nvpair_value_int16(nvp, (void *)&i16);
9541			printf(gettext("0x%x"), i16);
9542			break;
9543
9544		case DATA_TYPE_UINT16:
9545			(void) nvpair_value_uint16(nvp, &i16);
9546			printf(gettext("0x%x"), i16);
9547			break;
9548
9549		case DATA_TYPE_INT32:
9550			(void) nvpair_value_int32(nvp, (void *)&i32);
9551			printf(gettext("0x%x"), i32);
9552			break;
9553
9554		case DATA_TYPE_UINT32:
9555			(void) nvpair_value_uint32(nvp, &i32);
9556			printf(gettext("0x%x"), i32);
9557			break;
9558
9559		case DATA_TYPE_INT64:
9560			(void) nvpair_value_int64(nvp, (void *)&i64);
9561			printf(gettext("0x%llx"), (u_longlong_t)i64);
9562			break;
9563
9564		case DATA_TYPE_UINT64:
9565			(void) nvpair_value_uint64(nvp, &i64);
9566			/*
9567			 * translate vdev state values to readable
9568			 * strings to aide zpool events consumers
9569			 */
9570			if (strcmp(name,
9571			    FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE) == 0 ||
9572			    strcmp(name,
9573			    FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE) == 0) {
9574				printf(gettext("\"%s\" (0x%llx)"),
9575				    zpool_state_to_name(i64, VDEV_AUX_NONE),
9576				    (u_longlong_t)i64);
9577			} else {
9578				printf(gettext("0x%llx"), (u_longlong_t)i64);
9579			}
9580			break;
9581
9582		case DATA_TYPE_HRTIME:
9583			(void) nvpair_value_hrtime(nvp, (void *)&i64);
9584			printf(gettext("0x%llx"), (u_longlong_t)i64);
9585			break;
9586
9587		case DATA_TYPE_STRING:
9588			(void) nvpair_value_string(nvp, &str);
9589			printf(gettext("\"%s\""), str ? str : "<NULL>");
9590			break;
9591
9592		case DATA_TYPE_NVLIST:
9593			printf(gettext("(embedded nvlist)\n"));
9594			(void) nvpair_value_nvlist(nvp, &cnv);
9595			zpool_do_events_nvprint(cnv, depth + 8);
9596			printf(gettext("%*s(end %s)"), depth, "", name);
9597			break;
9598
9599		case DATA_TYPE_NVLIST_ARRAY: {
9600			nvlist_t **val;
9601			uint_t i, nelem;
9602
9603			(void) nvpair_value_nvlist_array(nvp, &val, &nelem);
9604			printf(gettext("(%d embedded nvlists)\n"), nelem);
9605			for (i = 0; i < nelem; i++) {
9606				printf(gettext("%*s%s[%d] = %s\n"),
9607				    depth, "", name, i, "(embedded nvlist)");
9608				zpool_do_events_nvprint(val[i], depth + 8);
9609				printf(gettext("%*s(end %s[%i])\n"),
9610				    depth, "", name, i);
9611			}
9612			printf(gettext("%*s(end %s)\n"), depth, "", name);
9613			}
9614			break;
9615
9616		case DATA_TYPE_INT8_ARRAY: {
9617			int8_t *val;
9618			uint_t i, nelem;
9619
9620			(void) nvpair_value_int8_array(nvp, &val, &nelem);
9621			for (i = 0; i < nelem; i++)
9622				printf(gettext("0x%x "), val[i]);
9623
9624			break;
9625			}
9626
9627		case DATA_TYPE_UINT8_ARRAY: {
9628			uint8_t *val;
9629			uint_t i, nelem;
9630
9631			(void) nvpair_value_uint8_array(nvp, &val, &nelem);
9632			for (i = 0; i < nelem; i++)
9633				printf(gettext("0x%x "), val[i]);
9634
9635			break;
9636			}
9637
9638		case DATA_TYPE_INT16_ARRAY: {
9639			int16_t *val;
9640			uint_t i, nelem;
9641
9642			(void) nvpair_value_int16_array(nvp, &val, &nelem);
9643			for (i = 0; i < nelem; i++)
9644				printf(gettext("0x%x "), val[i]);
9645
9646			break;
9647			}
9648
9649		case DATA_TYPE_UINT16_ARRAY: {
9650			uint16_t *val;
9651			uint_t i, nelem;
9652
9653			(void) nvpair_value_uint16_array(nvp, &val, &nelem);
9654			for (i = 0; i < nelem; i++)
9655				printf(gettext("0x%x "), val[i]);
9656
9657			break;
9658			}
9659
9660		case DATA_TYPE_INT32_ARRAY: {
9661			int32_t *val;
9662			uint_t i, nelem;
9663
9664			(void) nvpair_value_int32_array(nvp, &val, &nelem);
9665			for (i = 0; i < nelem; i++)
9666				printf(gettext("0x%x "), val[i]);
9667
9668			break;
9669			}
9670
9671		case DATA_TYPE_UINT32_ARRAY: {
9672			uint32_t *val;
9673			uint_t i, nelem;
9674
9675			(void) nvpair_value_uint32_array(nvp, &val, &nelem);
9676			for (i = 0; i < nelem; i++)
9677				printf(gettext("0x%x "), val[i]);
9678
9679			break;
9680			}
9681
9682		case DATA_TYPE_INT64_ARRAY: {
9683			int64_t *val;
9684			uint_t i, nelem;
9685
9686			(void) nvpair_value_int64_array(nvp, &val, &nelem);
9687			for (i = 0; i < nelem; i++)
9688				printf(gettext("0x%llx "),
9689				    (u_longlong_t)val[i]);
9690
9691			break;
9692			}
9693
9694		case DATA_TYPE_UINT64_ARRAY: {
9695			uint64_t *val;
9696			uint_t i, nelem;
9697
9698			(void) nvpair_value_uint64_array(nvp, &val, &nelem);
9699			for (i = 0; i < nelem; i++)
9700				printf(gettext("0x%llx "),
9701				    (u_longlong_t)val[i]);
9702
9703			break;
9704			}
9705
9706		case DATA_TYPE_STRING_ARRAY: {
9707			char **str;
9708			uint_t i, nelem;
9709
9710			(void) nvpair_value_string_array(nvp, &str, &nelem);
9711			for (i = 0; i < nelem; i++)
9712				printf(gettext("\"%s\" "),
9713				    str[i] ? str[i] : "<NULL>");
9714
9715			break;
9716			}
9717
9718		case DATA_TYPE_BOOLEAN_ARRAY:
9719		case DATA_TYPE_BYTE_ARRAY:
9720		case DATA_TYPE_DOUBLE:
9721		case DATA_TYPE_DONTCARE:
9722		case DATA_TYPE_UNKNOWN:
9723			printf(gettext("<unknown>"));
9724			break;
9725		}
9726
9727		printf(gettext("\n"));
9728	}
9729}
9730
9731static int
9732zpool_do_events_next(ev_opts_t *opts)
9733{
9734	nvlist_t *nvl;
9735	int zevent_fd, ret, dropped;
9736	char *pool;
9737
9738	zevent_fd = open(ZFS_DEV, O_RDWR);
9739	VERIFY(zevent_fd >= 0);
9740
9741	if (!opts->scripted)
9742		(void) printf(gettext("%-30s %s\n"), "TIME", "CLASS");
9743
9744	while (1) {
9745		ret = zpool_events_next(g_zfs, &nvl, &dropped,
9746		    (opts->follow ? ZEVENT_NONE : ZEVENT_NONBLOCK), zevent_fd);
9747		if (ret || nvl == NULL)
9748			break;
9749
9750		if (dropped > 0)
9751			(void) printf(gettext("dropped %d events\n"), dropped);
9752
9753		if (strlen(opts->poolname) > 0 &&
9754		    nvlist_lookup_string(nvl, FM_FMRI_ZFS_POOL, &pool) == 0 &&
9755		    strcmp(opts->poolname, pool) != 0)
9756			continue;
9757
9758		zpool_do_events_short(nvl, opts);
9759
9760		if (opts->verbose) {
9761			zpool_do_events_nvprint(nvl, 8);
9762			printf(gettext("\n"));
9763		}
9764		(void) fflush(stdout);
9765
9766		nvlist_free(nvl);
9767	}
9768
9769	VERIFY(0 == close(zevent_fd));
9770
9771	return (ret);
9772}
9773
9774static int
9775zpool_do_events_clear(ev_opts_t *opts)
9776{
9777	int count, ret;
9778
9779	ret = zpool_events_clear(g_zfs, &count);
9780	if (!ret)
9781		(void) printf(gettext("cleared %d events\n"), count);
9782
9783	return (ret);
9784}
9785
9786/*
9787 * zpool events [-vHf [pool] | -c]
9788 *
9789 * Displays events logs by ZFS.
9790 */
9791int
9792zpool_do_events(int argc, char **argv)
9793{
9794	ev_opts_t opts = { 0 };
9795	int ret;
9796	int c;
9797
9798	/* check options */
9799	while ((c = getopt(argc, argv, "vHfc")) != -1) {
9800		switch (c) {
9801		case 'v':
9802			opts.verbose = 1;
9803			break;
9804		case 'H':
9805			opts.scripted = 1;
9806			break;
9807		case 'f':
9808			opts.follow = 1;
9809			break;
9810		case 'c':
9811			opts.clear = 1;
9812			break;
9813		case '?':
9814			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9815			    optopt);
9816			usage(B_FALSE);
9817		}
9818	}
9819	argc -= optind;
9820	argv += optind;
9821
9822	if (argc > 1) {
9823		(void) fprintf(stderr, gettext("too many arguments\n"));
9824		usage(B_FALSE);
9825	} else if (argc == 1) {
9826		(void) strlcpy(opts.poolname, argv[0], sizeof (opts.poolname));
9827		if (!zfs_name_valid(opts.poolname, ZFS_TYPE_POOL)) {
9828			(void) fprintf(stderr,
9829			    gettext("invalid pool name '%s'\n"), opts.poolname);
9830			usage(B_FALSE);
9831		}
9832	}
9833
9834	if ((argc == 1 || opts.verbose || opts.scripted || opts.follow) &&
9835	    opts.clear) {
9836		(void) fprintf(stderr,
9837		    gettext("invalid options combined with -c\n"));
9838		usage(B_FALSE);
9839	}
9840
9841	if (opts.clear)
9842		ret = zpool_do_events_clear(&opts);
9843	else
9844		ret = zpool_do_events_next(&opts);
9845
9846	return (ret);
9847}
9848
9849static int
9850get_callback(zpool_handle_t *zhp, void *data)
9851{
9852	zprop_get_cbdata_t *cbp = (zprop_get_cbdata_t *)data;
9853	char value[MAXNAMELEN];
9854	zprop_source_t srctype;
9855	zprop_list_t *pl;
9856
9857	for (pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) {
9858
9859		/*
9860		 * Skip the special fake placeholder. This will also skip
9861		 * over the name property when 'all' is specified.
9862		 */
9863		if (pl->pl_prop == ZPOOL_PROP_NAME &&
9864		    pl == cbp->cb_proplist)
9865			continue;
9866
9867		if (pl->pl_prop == ZPROP_INVAL &&
9868		    (zpool_prop_feature(pl->pl_user_prop) ||
9869		    zpool_prop_unsupported(pl->pl_user_prop))) {
9870			srctype = ZPROP_SRC_LOCAL;
9871
9872			if (zpool_prop_get_feature(zhp, pl->pl_user_prop,
9873			    value, sizeof (value)) == 0) {
9874				zprop_print_one_property(zpool_get_name(zhp),
9875				    cbp, pl->pl_user_prop, value, srctype,
9876				    NULL, NULL);
9877			}
9878		} else {
9879			if (zpool_get_prop(zhp, pl->pl_prop, value,
9880			    sizeof (value), &srctype, cbp->cb_literal) != 0)
9881				continue;
9882
9883			zprop_print_one_property(zpool_get_name(zhp), cbp,
9884			    zpool_prop_to_name(pl->pl_prop), value, srctype,
9885			    NULL, NULL);
9886		}
9887	}
9888	return (0);
9889}
9890
9891/*
9892 * zpool get [-Hp] [-o "all" | field[,...]] <"all" | property[,...]> <pool> ...
9893 *
9894 *	-H	Scripted mode.  Don't display headers, and separate properties
9895 *		by a single tab.
9896 *	-o	List of columns to display.  Defaults to
9897 *		"name,property,value,source".
9898 * 	-p	Display values in parsable (exact) format.
9899 *
9900 * Get properties of pools in the system. Output space statistics
9901 * for each one as well as other attributes.
9902 */
9903int
9904zpool_do_get(int argc, char **argv)
9905{
9906	zprop_get_cbdata_t cb = { 0 };
9907	zprop_list_t fake_name = { 0 };
9908	int ret;
9909	int c, i;
9910	char *value;
9911
9912	cb.cb_first = B_TRUE;
9913
9914	/*
9915	 * Set up default columns and sources.
9916	 */
9917	cb.cb_sources = ZPROP_SRC_ALL;
9918	cb.cb_columns[0] = GET_COL_NAME;
9919	cb.cb_columns[1] = GET_COL_PROPERTY;
9920	cb.cb_columns[2] = GET_COL_VALUE;
9921	cb.cb_columns[3] = GET_COL_SOURCE;
9922	cb.cb_type = ZFS_TYPE_POOL;
9923
9924	/* check options */
9925	while ((c = getopt(argc, argv, ":Hpo:")) != -1) {
9926		switch (c) {
9927		case 'p':
9928			cb.cb_literal = B_TRUE;
9929			break;
9930		case 'H':
9931			cb.cb_scripted = B_TRUE;
9932			break;
9933		case 'o':
9934			bzero(&cb.cb_columns, sizeof (cb.cb_columns));
9935			i = 0;
9936			while (*optarg != '\0') {
9937				static char *col_subopts[] =
9938				{ "name", "property", "value", "source",
9939				"all", NULL };
9940
9941				if (i == ZFS_GET_NCOLS) {
9942					(void) fprintf(stderr, gettext("too "
9943					"many fields given to -o "
9944					"option\n"));
9945					usage(B_FALSE);
9946				}
9947
9948				switch (getsubopt(&optarg, col_subopts,
9949				    &value)) {
9950				case 0:
9951					cb.cb_columns[i++] = GET_COL_NAME;
9952					break;
9953				case 1:
9954					cb.cb_columns[i++] = GET_COL_PROPERTY;
9955					break;
9956				case 2:
9957					cb.cb_columns[i++] = GET_COL_VALUE;
9958					break;
9959				case 3:
9960					cb.cb_columns[i++] = GET_COL_SOURCE;
9961					break;
9962				case 4:
9963					if (i > 0) {
9964						(void) fprintf(stderr,
9965						    gettext("\"all\" conflicts "
9966						    "with specific fields "
9967						    "given to -o option\n"));
9968						usage(B_FALSE);
9969					}
9970					cb.cb_columns[0] = GET_COL_NAME;
9971					cb.cb_columns[1] = GET_COL_PROPERTY;
9972					cb.cb_columns[2] = GET_COL_VALUE;
9973					cb.cb_columns[3] = GET_COL_SOURCE;
9974					i = ZFS_GET_NCOLS;
9975					break;
9976				default:
9977					(void) fprintf(stderr,
9978					    gettext("invalid column name "
9979					    "'%s'\n"), value);
9980					usage(B_FALSE);
9981				}
9982			}
9983			break;
9984		case '?':
9985			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9986			    optopt);
9987			usage(B_FALSE);
9988		}
9989	}
9990
9991	argc -= optind;
9992	argv += optind;
9993
9994	if (argc < 1) {
9995		(void) fprintf(stderr, gettext("missing property "
9996		    "argument\n"));
9997		usage(B_FALSE);
9998	}
9999
10000	if (zprop_get_list(g_zfs, argv[0], &cb.cb_proplist,
10001	    ZFS_TYPE_POOL) != 0)
10002		usage(B_FALSE);
10003
10004	argc--;
10005	argv++;
10006
10007	if (cb.cb_proplist != NULL) {
10008		fake_name.pl_prop = ZPOOL_PROP_NAME;
10009		fake_name.pl_width = strlen(gettext("NAME"));
10010		fake_name.pl_next = cb.cb_proplist;
10011		cb.cb_proplist = &fake_name;
10012	}
10013
10014	ret = for_each_pool(argc, argv, B_TRUE, &cb.cb_proplist, cb.cb_literal,
10015	    get_callback, &cb);
10016
10017	if (cb.cb_proplist == &fake_name)
10018		zprop_free_list(fake_name.pl_next);
10019	else
10020		zprop_free_list(cb.cb_proplist);
10021
10022	return (ret);
10023}
10024
10025typedef struct set_cbdata {
10026	char *cb_propname;
10027	char *cb_value;
10028	boolean_t cb_any_successful;
10029} set_cbdata_t;
10030
10031static int
10032set_callback(zpool_handle_t *zhp, void *data)
10033{
10034	int error;
10035	set_cbdata_t *cb = (set_cbdata_t *)data;
10036
10037	/* Check if we have out-of-bounds features */
10038	if (strcmp(cb->cb_propname, ZPOOL_CONFIG_COMPATIBILITY) == 0) {
10039		boolean_t features[SPA_FEATURES];
10040		if (zpool_do_load_compat(cb->cb_value, features) !=
10041		    ZPOOL_COMPATIBILITY_OK)
10042			return (-1);
10043
10044		nvlist_t *enabled = zpool_get_features(zhp);
10045		spa_feature_t i;
10046		for (i = 0; i < SPA_FEATURES; i++) {
10047			const char *fguid = spa_feature_table[i].fi_guid;
10048			if (nvlist_exists(enabled, fguid) && !features[i])
10049				break;
10050		}
10051		if (i < SPA_FEATURES)
10052			(void) fprintf(stderr, gettext("Warning: one or "
10053			    "more features already enabled on pool '%s'\n"
10054			    "are not present in this compatibility set.\n"),
10055			    zpool_get_name(zhp));
10056	}
10057
10058	/* if we're setting a feature, check it's in compatibility set */
10059	if (zpool_prop_feature(cb->cb_propname) &&
10060	    strcmp(cb->cb_value, ZFS_FEATURE_ENABLED) == 0) {
10061		char *fname = strchr(cb->cb_propname, '@') + 1;
10062		spa_feature_t f;
10063
10064		if (zfeature_lookup_name(fname, &f) == 0) {
10065			char compat[ZFS_MAXPROPLEN];
10066			if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY,
10067			    compat, ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
10068				compat[0] = '\0';
10069
10070			boolean_t features[SPA_FEATURES];
10071			if (zpool_do_load_compat(compat, features) !=
10072			    ZPOOL_COMPATIBILITY_OK) {
10073				(void) fprintf(stderr, gettext("Error: "
10074				    "cannot enable feature '%s' on pool '%s'\n"
10075				    "because the pool's 'compatibility' "
10076				    "property cannot be parsed.\n"),
10077				    fname, zpool_get_name(zhp));
10078				return (-1);
10079			}
10080
10081			if (!features[f]) {
10082				(void) fprintf(stderr, gettext("Error: "
10083				    "cannot enable feature '%s' on pool '%s'\n"
10084				    "as it is not specified in this pool's "
10085				    "current compatibility set.\n"
10086				    "Consider setting 'compatibility' to a "
10087				    "less restrictive set, or to 'off'.\n"),
10088				    fname, zpool_get_name(zhp));
10089				return (-1);
10090			}
10091		}
10092	}
10093
10094	error = zpool_set_prop(zhp, cb->cb_propname, cb->cb_value);
10095
10096	if (!error)
10097		cb->cb_any_successful = B_TRUE;
10098
10099	return (error);
10100}
10101
10102int
10103zpool_do_set(int argc, char **argv)
10104{
10105	set_cbdata_t cb = { 0 };
10106	int error;
10107
10108	if (argc > 1 && argv[1][0] == '-') {
10109		(void) fprintf(stderr, gettext("invalid option '%c'\n"),
10110		    argv[1][1]);
10111		usage(B_FALSE);
10112	}
10113
10114	if (argc < 2) {
10115		(void) fprintf(stderr, gettext("missing property=value "
10116		    "argument\n"));
10117		usage(B_FALSE);
10118	}
10119
10120	if (argc < 3) {
10121		(void) fprintf(stderr, gettext("missing pool name\n"));
10122		usage(B_FALSE);
10123	}
10124
10125	if (argc > 3) {
10126		(void) fprintf(stderr, gettext("too many pool names\n"));
10127		usage(B_FALSE);
10128	}
10129
10130	cb.cb_propname = argv[1];
10131	cb.cb_value = strchr(cb.cb_propname, '=');
10132	if (cb.cb_value == NULL) {
10133		(void) fprintf(stderr, gettext("missing value in "
10134		    "property=value argument\n"));
10135		usage(B_FALSE);
10136	}
10137
10138	*(cb.cb_value) = '\0';
10139	cb.cb_value++;
10140
10141	error = for_each_pool(argc - 2, argv + 2, B_TRUE, NULL, B_FALSE,
10142	    set_callback, &cb);
10143
10144	return (error);
10145}
10146
10147/* Add up the total number of bytes left to initialize/trim across all vdevs */
10148static uint64_t
10149vdev_activity_remaining(nvlist_t *nv, zpool_wait_activity_t activity)
10150{
10151	uint64_t bytes_remaining;
10152	nvlist_t **child;
10153	uint_t c, children;
10154	vdev_stat_t *vs;
10155
10156	assert(activity == ZPOOL_WAIT_INITIALIZE ||
10157	    activity == ZPOOL_WAIT_TRIM);
10158
10159	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
10160	    (uint64_t **)&vs, &c) == 0);
10161
10162	if (activity == ZPOOL_WAIT_INITIALIZE &&
10163	    vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE)
10164		bytes_remaining = vs->vs_initialize_bytes_est -
10165		    vs->vs_initialize_bytes_done;
10166	else if (activity == ZPOOL_WAIT_TRIM &&
10167	    vs->vs_trim_state == VDEV_TRIM_ACTIVE)
10168		bytes_remaining = vs->vs_trim_bytes_est -
10169		    vs->vs_trim_bytes_done;
10170	else
10171		bytes_remaining = 0;
10172
10173	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
10174	    &child, &children) != 0)
10175		children = 0;
10176
10177	for (c = 0; c < children; c++)
10178		bytes_remaining += vdev_activity_remaining(child[c], activity);
10179
10180	return (bytes_remaining);
10181}
10182
10183/* Add up the total number of bytes left to rebuild across top-level vdevs */
10184static uint64_t
10185vdev_activity_top_remaining(nvlist_t *nv)
10186{
10187	uint64_t bytes_remaining = 0;
10188	nvlist_t **child;
10189	uint_t children;
10190	int error;
10191
10192	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
10193	    &child, &children) != 0)
10194		children = 0;
10195
10196	for (uint_t c = 0; c < children; c++) {
10197		vdev_rebuild_stat_t *vrs;
10198		uint_t i;
10199
10200		error = nvlist_lookup_uint64_array(child[c],
10201		    ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i);
10202		if (error == 0) {
10203			if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
10204				bytes_remaining += (vrs->vrs_bytes_est -
10205				    vrs->vrs_bytes_rebuilt);
10206			}
10207		}
10208	}
10209
10210	return (bytes_remaining);
10211}
10212
10213/* Whether any vdevs are 'spare' or 'replacing' vdevs */
10214static boolean_t
10215vdev_any_spare_replacing(nvlist_t *nv)
10216{
10217	nvlist_t **child;
10218	uint_t c, children;
10219	char *vdev_type;
10220
10221	(void) nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &vdev_type);
10222
10223	if (strcmp(vdev_type, VDEV_TYPE_REPLACING) == 0 ||
10224	    strcmp(vdev_type, VDEV_TYPE_SPARE) == 0 ||
10225	    strcmp(vdev_type, VDEV_TYPE_DRAID_SPARE) == 0) {
10226		return (B_TRUE);
10227	}
10228
10229	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
10230	    &child, &children) != 0)
10231		children = 0;
10232
10233	for (c = 0; c < children; c++) {
10234		if (vdev_any_spare_replacing(child[c]))
10235			return (B_TRUE);
10236	}
10237
10238	return (B_FALSE);
10239}
10240
10241typedef struct wait_data {
10242	char *wd_poolname;
10243	boolean_t wd_scripted;
10244	boolean_t wd_exact;
10245	boolean_t wd_headers_once;
10246	boolean_t wd_should_exit;
10247	/* Which activities to wait for */
10248	boolean_t wd_enabled[ZPOOL_WAIT_NUM_ACTIVITIES];
10249	float wd_interval;
10250	pthread_cond_t wd_cv;
10251	pthread_mutex_t wd_mutex;
10252} wait_data_t;
10253
10254/*
10255 * Print to stdout a single line, containing one column for each activity that
10256 * we are waiting for specifying how many bytes of work are left for that
10257 * activity.
10258 */
10259static void
10260print_wait_status_row(wait_data_t *wd, zpool_handle_t *zhp, int row)
10261{
10262	nvlist_t *config, *nvroot;
10263	uint_t c;
10264	int i;
10265	pool_checkpoint_stat_t *pcs = NULL;
10266	pool_scan_stat_t *pss = NULL;
10267	pool_removal_stat_t *prs = NULL;
10268	char *headers[] = {"DISCARD", "FREE", "INITIALIZE", "REPLACE",
10269	    "REMOVE", "RESILVER", "SCRUB", "TRIM"};
10270	int col_widths[ZPOOL_WAIT_NUM_ACTIVITIES];
10271
10272	/* Calculate the width of each column */
10273	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10274		/*
10275		 * Make sure we have enough space in the col for pretty-printed
10276		 * numbers and for the column header, and then leave a couple
10277		 * spaces between cols for readability.
10278		 */
10279		col_widths[i] = MAX(strlen(headers[i]), 6) + 2;
10280	}
10281
10282	/* Print header if appropriate */
10283	int term_height = terminal_height();
10284	boolean_t reprint_header = (!wd->wd_headers_once && term_height > 0 &&
10285	    row % (term_height-1) == 0);
10286	if (!wd->wd_scripted && (row == 0 || reprint_header)) {
10287		for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10288			if (wd->wd_enabled[i])
10289				(void) printf("%*s", col_widths[i], headers[i]);
10290		}
10291		(void) printf("\n");
10292	}
10293
10294	/* Bytes of work remaining in each activity */
10295	int64_t bytes_rem[ZPOOL_WAIT_NUM_ACTIVITIES] = {0};
10296
10297	bytes_rem[ZPOOL_WAIT_FREE] =
10298	    zpool_get_prop_int(zhp, ZPOOL_PROP_FREEING, NULL);
10299
10300	config = zpool_get_config(zhp, NULL);
10301	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
10302
10303	(void) nvlist_lookup_uint64_array(nvroot,
10304	    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
10305	if (pcs != NULL && pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
10306		bytes_rem[ZPOOL_WAIT_CKPT_DISCARD] = pcs->pcs_space;
10307
10308	(void) nvlist_lookup_uint64_array(nvroot,
10309	    ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
10310	if (prs != NULL && prs->prs_state == DSS_SCANNING)
10311		bytes_rem[ZPOOL_WAIT_REMOVE] = prs->prs_to_copy -
10312		    prs->prs_copied;
10313
10314	(void) nvlist_lookup_uint64_array(nvroot,
10315	    ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&pss, &c);
10316	if (pss != NULL && pss->pss_state == DSS_SCANNING &&
10317	    pss->pss_pass_scrub_pause == 0) {
10318		int64_t rem = pss->pss_to_examine - pss->pss_issued;
10319		if (pss->pss_func == POOL_SCAN_SCRUB)
10320			bytes_rem[ZPOOL_WAIT_SCRUB] = rem;
10321		else
10322			bytes_rem[ZPOOL_WAIT_RESILVER] = rem;
10323	} else if (check_rebuilding(nvroot, NULL)) {
10324		bytes_rem[ZPOOL_WAIT_RESILVER] =
10325		    vdev_activity_top_remaining(nvroot);
10326	}
10327
10328	bytes_rem[ZPOOL_WAIT_INITIALIZE] =
10329	    vdev_activity_remaining(nvroot, ZPOOL_WAIT_INITIALIZE);
10330	bytes_rem[ZPOOL_WAIT_TRIM] =
10331	    vdev_activity_remaining(nvroot, ZPOOL_WAIT_TRIM);
10332
10333	/*
10334	 * A replace finishes after resilvering finishes, so the amount of work
10335	 * left for a replace is the same as for resilvering.
10336	 *
10337	 * It isn't quite correct to say that if we have any 'spare' or
10338	 * 'replacing' vdevs and a resilver is happening, then a replace is in
10339	 * progress, like we do here. When a hot spare is used, the faulted vdev
10340	 * is not removed after the hot spare is resilvered, so parent 'spare'
10341	 * vdev is not removed either. So we could have a 'spare' vdev, but be
10342	 * resilvering for a different reason. However, we use it as a heuristic
10343	 * because we don't have access to the DTLs, which could tell us whether
10344	 * or not we have really finished resilvering a hot spare.
10345	 */
10346	if (vdev_any_spare_replacing(nvroot))
10347		bytes_rem[ZPOOL_WAIT_REPLACE] =  bytes_rem[ZPOOL_WAIT_RESILVER];
10348
10349	if (timestamp_fmt != NODATE)
10350		print_timestamp(timestamp_fmt);
10351
10352	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10353		char buf[64];
10354		if (!wd->wd_enabled[i])
10355			continue;
10356
10357		if (wd->wd_exact)
10358			(void) snprintf(buf, sizeof (buf), "%" PRIi64,
10359			    bytes_rem[i]);
10360		else
10361			zfs_nicenum(bytes_rem[i], buf, sizeof (buf));
10362
10363		if (wd->wd_scripted)
10364			(void) printf(i == 0 ? "%s" : "\t%s", buf);
10365		else
10366			(void) printf(" %*s", col_widths[i] - 1, buf);
10367	}
10368	(void) printf("\n");
10369	(void) fflush(stdout);
10370}
10371
10372static void *
10373wait_status_thread(void *arg)
10374{
10375	wait_data_t *wd = (wait_data_t *)arg;
10376	zpool_handle_t *zhp;
10377
10378	if ((zhp = zpool_open(g_zfs, wd->wd_poolname)) == NULL)
10379		return (void *)(1);
10380
10381	for (int row = 0; ; row++) {
10382		boolean_t missing;
10383		struct timespec timeout;
10384		int ret = 0;
10385		(void) clock_gettime(CLOCK_REALTIME, &timeout);
10386
10387		if (zpool_refresh_stats(zhp, &missing) != 0 || missing ||
10388		    zpool_props_refresh(zhp) != 0) {
10389			zpool_close(zhp);
10390			return (void *)(uintptr_t)(missing ? 0 : 1);
10391		}
10392
10393		print_wait_status_row(wd, zhp, row);
10394
10395		timeout.tv_sec += floor(wd->wd_interval);
10396		long nanos = timeout.tv_nsec +
10397		    (wd->wd_interval - floor(wd->wd_interval)) * NANOSEC;
10398		if (nanos >= NANOSEC) {
10399			timeout.tv_sec++;
10400			timeout.tv_nsec = nanos - NANOSEC;
10401		} else {
10402			timeout.tv_nsec = nanos;
10403		}
10404		pthread_mutex_lock(&wd->wd_mutex);
10405		if (!wd->wd_should_exit)
10406			ret = pthread_cond_timedwait(&wd->wd_cv, &wd->wd_mutex,
10407			    &timeout);
10408		pthread_mutex_unlock(&wd->wd_mutex);
10409		if (ret == 0) {
10410			break; /* signaled by main thread */
10411		} else if (ret != ETIMEDOUT) {
10412			(void) fprintf(stderr, gettext("pthread_cond_timedwait "
10413			    "failed: %s\n"), strerror(ret));
10414			zpool_close(zhp);
10415			return (void *)(uintptr_t)(1);
10416		}
10417	}
10418
10419	zpool_close(zhp);
10420	return (void *)(0);
10421}
10422
10423int
10424zpool_do_wait(int argc, char **argv)
10425{
10426	boolean_t verbose = B_FALSE;
10427	int c;
10428	char *value;
10429	int i;
10430	unsigned long count;
10431	pthread_t status_thr;
10432	int error = 0;
10433	zpool_handle_t *zhp;
10434
10435	wait_data_t wd;
10436	wd.wd_scripted = B_FALSE;
10437	wd.wd_exact = B_FALSE;
10438	wd.wd_headers_once = B_FALSE;
10439	wd.wd_should_exit = B_FALSE;
10440
10441	pthread_mutex_init(&wd.wd_mutex, NULL);
10442	pthread_cond_init(&wd.wd_cv, NULL);
10443
10444	/* By default, wait for all types of activity. */
10445	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++)
10446		wd.wd_enabled[i] = B_TRUE;
10447
10448	while ((c = getopt(argc, argv, "HpT:t:")) != -1) {
10449		switch (c) {
10450		case 'H':
10451			wd.wd_scripted = B_TRUE;
10452			break;
10453		case 'n':
10454			wd.wd_headers_once = B_TRUE;
10455			break;
10456		case 'p':
10457			wd.wd_exact = B_TRUE;
10458			break;
10459		case 'T':
10460			get_timestamp_arg(*optarg);
10461			break;
10462		case 't':
10463		{
10464			static char *col_subopts[] = { "discard", "free",
10465			    "initialize", "replace", "remove", "resilver",
10466			    "scrub", "trim", NULL };
10467
10468			/* Reset activities array */
10469			bzero(&wd.wd_enabled, sizeof (wd.wd_enabled));
10470			while (*optarg != '\0') {
10471				int activity = getsubopt(&optarg, col_subopts,
10472				    &value);
10473
10474				if (activity < 0) {
10475					(void) fprintf(stderr,
10476					    gettext("invalid activity '%s'\n"),
10477					    value);
10478					usage(B_FALSE);
10479				}
10480
10481				wd.wd_enabled[activity] = B_TRUE;
10482			}
10483			break;
10484		}
10485		case '?':
10486			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
10487			    optopt);
10488			usage(B_FALSE);
10489		}
10490	}
10491
10492	argc -= optind;
10493	argv += optind;
10494
10495	get_interval_count(&argc, argv, &wd.wd_interval, &count);
10496	if (count != 0) {
10497		/* This subcmd only accepts an interval, not a count */
10498		(void) fprintf(stderr, gettext("too many arguments\n"));
10499		usage(B_FALSE);
10500	}
10501
10502	if (wd.wd_interval != 0)
10503		verbose = B_TRUE;
10504
10505	if (argc < 1) {
10506		(void) fprintf(stderr, gettext("missing 'pool' argument\n"));
10507		usage(B_FALSE);
10508	}
10509	if (argc > 1) {
10510		(void) fprintf(stderr, gettext("too many arguments\n"));
10511		usage(B_FALSE);
10512	}
10513
10514	wd.wd_poolname = argv[0];
10515
10516	if ((zhp = zpool_open(g_zfs, wd.wd_poolname)) == NULL)
10517		return (1);
10518
10519	if (verbose) {
10520		/*
10521		 * We use a separate thread for printing status updates because
10522		 * the main thread will call lzc_wait(), which blocks as long
10523		 * as an activity is in progress, which can be a long time.
10524		 */
10525		if (pthread_create(&status_thr, NULL, wait_status_thread, &wd)
10526		    != 0) {
10527			(void) fprintf(stderr, gettext("failed to create status"
10528			    "thread: %s\n"), strerror(errno));
10529			zpool_close(zhp);
10530			return (1);
10531		}
10532	}
10533
10534	/*
10535	 * Loop over all activities that we are supposed to wait for until none
10536	 * of them are in progress. Note that this means we can end up waiting
10537	 * for more activities to complete than just those that were in progress
10538	 * when we began waiting; if an activity we are interested in begins
10539	 * while we are waiting for another activity, we will wait for both to
10540	 * complete before exiting.
10541	 */
10542	for (;;) {
10543		boolean_t missing = B_FALSE;
10544		boolean_t any_waited = B_FALSE;
10545
10546		for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10547			boolean_t waited;
10548
10549			if (!wd.wd_enabled[i])
10550				continue;
10551
10552			error = zpool_wait_status(zhp, i, &missing, &waited);
10553			if (error != 0 || missing)
10554				break;
10555
10556			any_waited = (any_waited || waited);
10557		}
10558
10559		if (error != 0 || missing || !any_waited)
10560			break;
10561	}
10562
10563	zpool_close(zhp);
10564
10565	if (verbose) {
10566		uintptr_t status;
10567		pthread_mutex_lock(&wd.wd_mutex);
10568		wd.wd_should_exit = B_TRUE;
10569		pthread_cond_signal(&wd.wd_cv);
10570		pthread_mutex_unlock(&wd.wd_mutex);
10571		(void) pthread_join(status_thr, (void *)&status);
10572		if (status != 0)
10573			error = status;
10574	}
10575
10576	pthread_mutex_destroy(&wd.wd_mutex);
10577	pthread_cond_destroy(&wd.wd_cv);
10578	return (error);
10579}
10580
10581static int
10582find_command_idx(char *command, int *idx)
10583{
10584	int i;
10585
10586	for (i = 0; i < NCOMMAND; i++) {
10587		if (command_table[i].name == NULL)
10588			continue;
10589
10590		if (strcmp(command, command_table[i].name) == 0) {
10591			*idx = i;
10592			return (0);
10593		}
10594	}
10595	return (1);
10596}
10597
10598/*
10599 * Display version message
10600 */
10601static int
10602zpool_do_version(int argc, char **argv)
10603{
10604	if (zfs_version_print() == -1)
10605		return (1);
10606
10607	return (0);
10608}
10609
10610/*
10611 * Do zpool_load_compat() and print error message on failure
10612 */
10613static zpool_compat_status_t
10614zpool_do_load_compat(const char *compat, boolean_t *list)
10615{
10616	char report[1024];
10617
10618	zpool_compat_status_t ret;
10619
10620	ret = zpool_load_compat(compat, list, report, 1024);
10621	switch (ret) {
10622
10623	case ZPOOL_COMPATIBILITY_OK:
10624		break;
10625
10626	case ZPOOL_COMPATIBILITY_NOFILES:
10627	case ZPOOL_COMPATIBILITY_BADFILE:
10628	case ZPOOL_COMPATIBILITY_BADTOKEN:
10629		(void) fprintf(stderr, "Error: %s\n", report);
10630		break;
10631
10632	case ZPOOL_COMPATIBILITY_WARNTOKEN:
10633		(void) fprintf(stderr, "Warning: %s\n", report);
10634		ret = ZPOOL_COMPATIBILITY_OK;
10635		break;
10636	}
10637	return (ret);
10638}
10639
10640int
10641main(int argc, char **argv)
10642{
10643	int ret = 0;
10644	int i = 0;
10645	char *cmdname;
10646	char **newargv;
10647
10648	(void) setlocale(LC_ALL, "");
10649	(void) setlocale(LC_NUMERIC, "C");
10650	(void) textdomain(TEXT_DOMAIN);
10651	srand(time(NULL));
10652
10653	opterr = 0;
10654
10655	/*
10656	 * Make sure the user has specified some command.
10657	 */
10658	if (argc < 2) {
10659		(void) fprintf(stderr, gettext("missing command\n"));
10660		usage(B_FALSE);
10661	}
10662
10663	cmdname = argv[1];
10664
10665	/*
10666	 * Special case '-?'
10667	 */
10668	if ((strcmp(cmdname, "-?") == 0) || strcmp(cmdname, "--help") == 0)
10669		usage(B_TRUE);
10670
10671	/*
10672	 * Special case '-V|--version'
10673	 */
10674	if ((strcmp(cmdname, "-V") == 0) || (strcmp(cmdname, "--version") == 0))
10675		return (zpool_do_version(argc, argv));
10676
10677	if ((g_zfs = libzfs_init()) == NULL) {
10678		(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
10679		return (1);
10680	}
10681
10682	libzfs_print_on_error(g_zfs, B_TRUE);
10683
10684	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));
10685
10686	/*
10687	 * Many commands modify input strings for string parsing reasons.
10688	 * We create a copy to protect the original argv.
10689	 */
10690	newargv = malloc((argc + 1) * sizeof (newargv[0]));
10691	for (i = 0; i < argc; i++)
10692		newargv[i] = strdup(argv[i]);
10693	newargv[argc] = NULL;
10694
10695	/*
10696	 * Run the appropriate command.
10697	 */
10698	if (find_command_idx(cmdname, &i) == 0) {
10699		current_command = &command_table[i];
10700		ret = command_table[i].func(argc - 1, newargv + 1);
10701	} else if (strchr(cmdname, '=')) {
10702		verify(find_command_idx("set", &i) == 0);
10703		current_command = &command_table[i];
10704		ret = command_table[i].func(argc, newargv);
10705	} else if (strcmp(cmdname, "freeze") == 0 && argc == 3) {
10706		/*
10707		 * 'freeze' is a vile debugging abomination, so we treat
10708		 * it as such.
10709		 */
10710		zfs_cmd_t zc = {"\0"};
10711
10712		(void) strlcpy(zc.zc_name, argv[2], sizeof (zc.zc_name));
10713		ret = zfs_ioctl(g_zfs, ZFS_IOC_POOL_FREEZE, &zc);
10714		if (ret != 0) {
10715			(void) fprintf(stderr,
10716			gettext("failed to freeze pool: %d\n"), errno);
10717			ret = 1;
10718		}
10719
10720		log_history = 0;
10721	} else {
10722		(void) fprintf(stderr, gettext("unrecognized "
10723		    "command '%s'\n"), cmdname);
10724		usage(B_FALSE);
10725		ret = 1;
10726	}
10727
10728	for (i = 0; i < argc; i++)
10729		free(newargv[i]);
10730	free(newargv);
10731
10732	if (ret == 0 && log_history)
10733		(void) zpool_log_history(g_zfs, history_str);
10734
10735	libzfs_fini(g_zfs);
10736
10737	/*
10738	 * The 'ZFS_ABORT' environment variable causes us to dump core on exit
10739	 * for the purposes of running ::findleaks.
10740	 */
10741	if (getenv("ZFS_ABORT") != NULL) {
10742		(void) printf("dumping core by request\n");
10743		abort();
10744	}
10745
10746	return (ret);
10747}
10748