xref: /freebsd/sys/contrib/openzfs/cmd/zpool/zpool_main.c (revision c1a3d7f20696ab5b72eee45863f3e04410d81fc8)
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  */
35 
36 #include <assert.h>
37 #include <ctype.h>
38 #include <dirent.h>
39 #include <errno.h>
40 #include <fcntl.h>
41 #include <getopt.h>
42 #include <libgen.h>
43 #include <libintl.h>
44 #include <libuutil.h>
45 #include <locale.h>
46 #include <pthread.h>
47 #include <stdio.h>
48 #include <stdlib.h>
49 #include <string.h>
50 #include <strings.h>
51 #include <time.h>
52 #include <unistd.h>
53 #include <pwd.h>
54 #include <zone.h>
55 #include <sys/wait.h>
56 #include <zfs_prop.h>
57 #include <sys/fs/zfs.h>
58 #include <sys/stat.h>
59 #include <sys/systeminfo.h>
60 #include <sys/fm/fs/zfs.h>
61 #include <sys/fm/util.h>
62 #include <sys/fm/protocol.h>
63 #include <sys/zfs_ioctl.h>
64 #include <sys/mount.h>
65 #include <sys/sysmacros.h>
66 
67 #include <math.h>
68 
69 #include <libzfs.h>
70 #include <libzutil.h>
71 
72 #include "zpool_util.h"
73 #include "zfs_comutil.h"
74 #include "zfeature_common.h"
75 
76 #include "statcommon.h"
77 
78 libzfs_handle_t *g_zfs;
79 
80 static int zpool_do_create(int, char **);
81 static int zpool_do_destroy(int, char **);
82 
83 static int zpool_do_add(int, char **);
84 static int zpool_do_remove(int, char **);
85 static int zpool_do_labelclear(int, char **);
86 
87 static int zpool_do_checkpoint(int, char **);
88 
89 static int zpool_do_list(int, char **);
90 static int zpool_do_iostat(int, char **);
91 static int zpool_do_status(int, char **);
92 
93 static int zpool_do_online(int, char **);
94 static int zpool_do_offline(int, char **);
95 static int zpool_do_clear(int, char **);
96 static int zpool_do_reopen(int, char **);
97 
98 static int zpool_do_reguid(int, char **);
99 
100 static int zpool_do_attach(int, char **);
101 static int zpool_do_detach(int, char **);
102 static int zpool_do_replace(int, char **);
103 static int zpool_do_split(int, char **);
104 
105 static int zpool_do_initialize(int, char **);
106 static int zpool_do_scrub(int, char **);
107 static int zpool_do_resilver(int, char **);
108 static int zpool_do_trim(int, char **);
109 
110 static int zpool_do_import(int, char **);
111 static int zpool_do_export(int, char **);
112 
113 static int zpool_do_upgrade(int, char **);
114 
115 static int zpool_do_history(int, char **);
116 static int zpool_do_events(int, char **);
117 
118 static int zpool_do_get(int, char **);
119 static int zpool_do_set(int, char **);
120 
121 static int zpool_do_sync(int, char **);
122 
123 static int zpool_do_version(int, char **);
124 
125 static int zpool_do_wait(int, char **);
126 
127 /*
128  * These libumem hooks provide a reasonable set of defaults for the allocator's
129  * debugging facilities.
130  */
131 
132 #ifdef DEBUG
133 const char *
134 _umem_debug_init(void)
135 {
136 	return ("default,verbose"); /* $UMEM_DEBUG setting */
137 }
138 
139 const char *
140 _umem_logging_init(void)
141 {
142 	return ("fail,contents"); /* $UMEM_LOGGING setting */
143 }
144 #endif
145 
146 typedef enum {
147 	HELP_ADD,
148 	HELP_ATTACH,
149 	HELP_CLEAR,
150 	HELP_CREATE,
151 	HELP_CHECKPOINT,
152 	HELP_DESTROY,
153 	HELP_DETACH,
154 	HELP_EXPORT,
155 	HELP_HISTORY,
156 	HELP_IMPORT,
157 	HELP_IOSTAT,
158 	HELP_LABELCLEAR,
159 	HELP_LIST,
160 	HELP_OFFLINE,
161 	HELP_ONLINE,
162 	HELP_REPLACE,
163 	HELP_REMOVE,
164 	HELP_INITIALIZE,
165 	HELP_SCRUB,
166 	HELP_RESILVER,
167 	HELP_TRIM,
168 	HELP_STATUS,
169 	HELP_UPGRADE,
170 	HELP_EVENTS,
171 	HELP_GET,
172 	HELP_SET,
173 	HELP_SPLIT,
174 	HELP_SYNC,
175 	HELP_REGUID,
176 	HELP_REOPEN,
177 	HELP_VERSION,
178 	HELP_WAIT
179 } zpool_help_t;
180 
181 
182 /*
183  * Flags for stats to display with "zpool iostats"
184  */
185 enum iostat_type {
186 	IOS_DEFAULT = 0,
187 	IOS_LATENCY = 1,
188 	IOS_QUEUES = 2,
189 	IOS_L_HISTO = 3,
190 	IOS_RQ_HISTO = 4,
191 	IOS_COUNT,	/* always last element */
192 };
193 
194 /* iostat_type entries as bitmasks */
195 #define	IOS_DEFAULT_M	(1ULL << IOS_DEFAULT)
196 #define	IOS_LATENCY_M	(1ULL << IOS_LATENCY)
197 #define	IOS_QUEUES_M	(1ULL << IOS_QUEUES)
198 #define	IOS_L_HISTO_M	(1ULL << IOS_L_HISTO)
199 #define	IOS_RQ_HISTO_M	(1ULL << IOS_RQ_HISTO)
200 
201 /* Mask of all the histo bits */
202 #define	IOS_ANYHISTO_M (IOS_L_HISTO_M | IOS_RQ_HISTO_M)
203 
204 /*
205  * Lookup table for iostat flags to nvlist names.  Basically a list
206  * of all the nvlists a flag requires.  Also specifies the order in
207  * which data gets printed in zpool iostat.
208  */
209 static const char *vsx_type_to_nvlist[IOS_COUNT][13] = {
210 	[IOS_L_HISTO] = {
211 	    ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
212 	    ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
213 	    ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
214 	    ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
215 	    ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
216 	    ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
217 	    ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
218 	    ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
219 	    ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
220 	    ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
221 	    NULL},
222 	[IOS_LATENCY] = {
223 	    ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
224 	    ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
225 	    ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
226 	    ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
227 	    ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
228 	    NULL},
229 	[IOS_QUEUES] = {
230 	    ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
231 	    ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
232 	    ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
233 	    ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
234 	    ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
235 	    ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
236 	    NULL},
237 	[IOS_RQ_HISTO] = {
238 	    ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO,
239 	    ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO,
240 	    ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO,
241 	    ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO,
242 	    ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO,
243 	    ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO,
244 	    ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO,
245 	    ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO,
246 	    ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO,
247 	    ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO,
248 	    ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO,
249 	    ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO,
250 	    NULL},
251 };
252 
253 
254 /*
255  * Given a cb->cb_flags with a histogram bit set, return the iostat_type.
256  * Right now, only one histo bit is ever set at one time, so we can
257  * just do a highbit64(a)
258  */
259 #define	IOS_HISTO_IDX(a)	(highbit64(a & IOS_ANYHISTO_M) - 1)
260 
261 typedef struct zpool_command {
262 	const char	*name;
263 	int		(*func)(int, char **);
264 	zpool_help_t	usage;
265 } zpool_command_t;
266 
267 /*
268  * Master command table.  Each ZFS command has a name, associated function, and
269  * usage message.  The usage messages need to be internationalized, so we have
270  * to have a function to return the usage message based on a command index.
271  *
272  * These commands are organized according to how they are displayed in the usage
273  * message.  An empty command (one with a NULL name) indicates an empty line in
274  * the generic usage message.
275  */
276 static zpool_command_t command_table[] = {
277 	{ "version",	zpool_do_version,	HELP_VERSION		},
278 	{ NULL },
279 	{ "create",	zpool_do_create,	HELP_CREATE		},
280 	{ "destroy",	zpool_do_destroy,	HELP_DESTROY		},
281 	{ NULL },
282 	{ "add",	zpool_do_add,		HELP_ADD		},
283 	{ "remove",	zpool_do_remove,	HELP_REMOVE		},
284 	{ NULL },
285 	{ "labelclear",	zpool_do_labelclear,	HELP_LABELCLEAR		},
286 	{ NULL },
287 	{ "checkpoint",	zpool_do_checkpoint,	HELP_CHECKPOINT		},
288 	{ NULL },
289 	{ "list",	zpool_do_list,		HELP_LIST		},
290 	{ "iostat",	zpool_do_iostat,	HELP_IOSTAT		},
291 	{ "status",	zpool_do_status,	HELP_STATUS		},
292 	{ NULL },
293 	{ "online",	zpool_do_online,	HELP_ONLINE		},
294 	{ "offline",	zpool_do_offline,	HELP_OFFLINE		},
295 	{ "clear",	zpool_do_clear,		HELP_CLEAR		},
296 	{ "reopen",	zpool_do_reopen,	HELP_REOPEN		},
297 	{ NULL },
298 	{ "attach",	zpool_do_attach,	HELP_ATTACH		},
299 	{ "detach",	zpool_do_detach,	HELP_DETACH		},
300 	{ "replace",	zpool_do_replace,	HELP_REPLACE		},
301 	{ "split",	zpool_do_split,		HELP_SPLIT		},
302 	{ NULL },
303 	{ "initialize",	zpool_do_initialize,	HELP_INITIALIZE		},
304 	{ "resilver",	zpool_do_resilver,	HELP_RESILVER		},
305 	{ "scrub",	zpool_do_scrub,		HELP_SCRUB		},
306 	{ "trim",	zpool_do_trim,		HELP_TRIM		},
307 	{ NULL },
308 	{ "import",	zpool_do_import,	HELP_IMPORT		},
309 	{ "export",	zpool_do_export,	HELP_EXPORT		},
310 	{ "upgrade",	zpool_do_upgrade,	HELP_UPGRADE		},
311 	{ "reguid",	zpool_do_reguid,	HELP_REGUID		},
312 	{ NULL },
313 	{ "history",	zpool_do_history,	HELP_HISTORY		},
314 	{ "events",	zpool_do_events,	HELP_EVENTS		},
315 	{ NULL },
316 	{ "get",	zpool_do_get,		HELP_GET		},
317 	{ "set",	zpool_do_set,		HELP_SET		},
318 	{ "sync",	zpool_do_sync,		HELP_SYNC		},
319 	{ NULL },
320 	{ "wait",	zpool_do_wait,		HELP_WAIT		},
321 };
322 
323 #define	NCOMMAND	(ARRAY_SIZE(command_table))
324 
325 #define	VDEV_ALLOC_CLASS_LOGS	"logs"
326 
327 static zpool_command_t *current_command;
328 static char history_str[HIS_MAX_RECORD_LEN];
329 static boolean_t log_history = B_TRUE;
330 static uint_t timestamp_fmt = NODATE;
331 
332 static const char *
333 get_usage(zpool_help_t idx)
334 {
335 	switch (idx) {
336 	case HELP_ADD:
337 		return (gettext("\tadd [-fgLnP] [-o property=value] "
338 		    "<pool> <vdev> ...\n"));
339 	case HELP_ATTACH:
340 		return (gettext("\tattach [-fsw] [-o property=value] "
341 		    "<pool> <device> <new-device>\n"));
342 	case HELP_CLEAR:
343 		return (gettext("\tclear [-nF] <pool> [device]\n"));
344 	case HELP_CREATE:
345 		return (gettext("\tcreate [-fnd] [-o property=value] ... \n"
346 		    "\t    [-O file-system-property=value] ... \n"
347 		    "\t    [-m mountpoint] [-R root] <pool> <vdev> ...\n"));
348 	case HELP_CHECKPOINT:
349 		return (gettext("\tcheckpoint [-d [-w]] <pool> ...\n"));
350 	case HELP_DESTROY:
351 		return (gettext("\tdestroy [-f] <pool>\n"));
352 	case HELP_DETACH:
353 		return (gettext("\tdetach <pool> <device>\n"));
354 	case HELP_EXPORT:
355 		return (gettext("\texport [-af] <pool> ...\n"));
356 	case HELP_HISTORY:
357 		return (gettext("\thistory [-il] [<pool>] ...\n"));
358 	case HELP_IMPORT:
359 		return (gettext("\timport [-d dir] [-D]\n"
360 		    "\timport [-o mntopts] [-o property=value] ... \n"
361 		    "\t    [-d dir | -c cachefile] [-D] [-l] [-f] [-m] [-N] "
362 		    "[-R root] [-F [-n]] -a\n"
363 		    "\timport [-o mntopts] [-o property=value] ... \n"
364 		    "\t    [-d dir | -c cachefile] [-D] [-l] [-f] [-m] [-N] "
365 		    "[-R root] [-F [-n]]\n"
366 		    "\t    [--rewind-to-checkpoint] <pool | id> [newpool]\n"));
367 	case HELP_IOSTAT:
368 		return (gettext("\tiostat [[[-c [script1,script2,...]"
369 		    "[-lq]]|[-rw]] [-T d | u] [-ghHLpPvy]\n"
370 		    "\t    [[pool ...]|[pool vdev ...]|[vdev ...]]"
371 		    " [[-n] interval [count]]\n"));
372 	case HELP_LABELCLEAR:
373 		return (gettext("\tlabelclear [-f] <vdev>\n"));
374 	case HELP_LIST:
375 		return (gettext("\tlist [-gHLpPv] [-o property[,...]] "
376 		    "[-T d|u] [pool] ... \n"
377 		    "\t    [interval [count]]\n"));
378 	case HELP_OFFLINE:
379 		return (gettext("\toffline [-f] [-t] <pool> <device> ...\n"));
380 	case HELP_ONLINE:
381 		return (gettext("\tonline [-e] <pool> <device> ...\n"));
382 	case HELP_REPLACE:
383 		return (gettext("\treplace [-fsw] [-o property=value] "
384 		    "<pool> <device> [new-device]\n"));
385 	case HELP_REMOVE:
386 		return (gettext("\tremove [-npsw] <pool> <device> ...\n"));
387 	case HELP_REOPEN:
388 		return (gettext("\treopen [-n] <pool>\n"));
389 	case HELP_INITIALIZE:
390 		return (gettext("\tinitialize [-c | -s] [-w] <pool> "
391 		    "[<device> ...]\n"));
392 	case HELP_SCRUB:
393 		return (gettext("\tscrub [-s | -p] [-w] <pool> ...\n"));
394 	case HELP_RESILVER:
395 		return (gettext("\tresilver <pool> ...\n"));
396 	case HELP_TRIM:
397 		return (gettext("\ttrim [-dw] [-r <rate>] [-c | -s] <pool> "
398 		    "[<device> ...]\n"));
399 	case HELP_STATUS:
400 		return (gettext("\tstatus [-c [script1,script2,...]] "
401 		    "[-igLpPstvxD]  [-T d|u] [pool] ... \n"
402 		    "\t    [interval [count]]\n"));
403 	case HELP_UPGRADE:
404 		return (gettext("\tupgrade\n"
405 		    "\tupgrade -v\n"
406 		    "\tupgrade [-V version] <-a | pool ...>\n"));
407 	case HELP_EVENTS:
408 		return (gettext("\tevents [-vHf [pool] | -c]\n"));
409 	case HELP_GET:
410 		return (gettext("\tget [-Hp] [-o \"all\" | field[,...]] "
411 		    "<\"all\" | property[,...]> <pool> ...\n"));
412 	case HELP_SET:
413 		return (gettext("\tset <property=value> <pool> \n"));
414 	case HELP_SPLIT:
415 		return (gettext("\tsplit [-gLnPl] [-R altroot] [-o mntopts]\n"
416 		    "\t    [-o property=value] <pool> <newpool> "
417 		    "[<device> ...]\n"));
418 	case HELP_REGUID:
419 		return (gettext("\treguid <pool>\n"));
420 	case HELP_SYNC:
421 		return (gettext("\tsync [pool] ...\n"));
422 	case HELP_VERSION:
423 		return (gettext("\tversion\n"));
424 	case HELP_WAIT:
425 		return (gettext("\twait [-Hp] [-T d|u] [-t <activity>[,...]] "
426 		    "<pool> [interval]\n"));
427 	}
428 
429 	abort();
430 	/* NOTREACHED */
431 }
432 
433 static void
434 zpool_collect_leaves(zpool_handle_t *zhp, nvlist_t *nvroot, nvlist_t *res)
435 {
436 	uint_t children = 0;
437 	nvlist_t **child;
438 	uint_t i;
439 
440 	(void) nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
441 	    &child, &children);
442 
443 	if (children == 0) {
444 		char *path = zpool_vdev_name(g_zfs, zhp, nvroot,
445 		    VDEV_NAME_PATH);
446 
447 		if (strcmp(path, VDEV_TYPE_INDIRECT) != 0 &&
448 		    strcmp(path, VDEV_TYPE_HOLE) != 0)
449 			fnvlist_add_boolean(res, path);
450 
451 		free(path);
452 		return;
453 	}
454 
455 	for (i = 0; i < children; i++) {
456 		zpool_collect_leaves(zhp, child[i], res);
457 	}
458 }
459 
460 /*
461  * Callback routine that will print out a pool property value.
462  */
463 static int
464 print_prop_cb(int prop, void *cb)
465 {
466 	FILE *fp = cb;
467 
468 	(void) fprintf(fp, "\t%-19s  ", zpool_prop_to_name(prop));
469 
470 	if (zpool_prop_readonly(prop))
471 		(void) fprintf(fp, "  NO   ");
472 	else
473 		(void) fprintf(fp, " YES   ");
474 
475 	if (zpool_prop_values(prop) == NULL)
476 		(void) fprintf(fp, "-\n");
477 	else
478 		(void) fprintf(fp, "%s\n", zpool_prop_values(prop));
479 
480 	return (ZPROP_CONT);
481 }
482 
483 /*
484  * Display usage message.  If we're inside a command, display only the usage for
485  * that command.  Otherwise, iterate over the entire command table and display
486  * a complete usage message.
487  */
488 static void
489 usage(boolean_t requested)
490 {
491 	FILE *fp = requested ? stdout : stderr;
492 
493 	if (current_command == NULL) {
494 		int i;
495 
496 		(void) fprintf(fp, gettext("usage: zpool command args ...\n"));
497 		(void) fprintf(fp,
498 		    gettext("where 'command' is one of the following:\n\n"));
499 
500 		for (i = 0; i < NCOMMAND; i++) {
501 			if (command_table[i].name == NULL)
502 				(void) fprintf(fp, "\n");
503 			else
504 				(void) fprintf(fp, "%s",
505 				    get_usage(command_table[i].usage));
506 		}
507 	} else {
508 		(void) fprintf(fp, gettext("usage:\n"));
509 		(void) fprintf(fp, "%s", get_usage(current_command->usage));
510 	}
511 
512 	if (current_command != NULL &&
513 	    ((strcmp(current_command->name, "set") == 0) ||
514 	    (strcmp(current_command->name, "get") == 0) ||
515 	    (strcmp(current_command->name, "list") == 0))) {
516 
517 		(void) fprintf(fp,
518 		    gettext("\nthe following properties are supported:\n"));
519 
520 		(void) fprintf(fp, "\n\t%-19s  %s   %s\n\n",
521 		    "PROPERTY", "EDIT", "VALUES");
522 
523 		/* Iterate over all properties */
524 		(void) zprop_iter(print_prop_cb, fp, B_FALSE, B_TRUE,
525 		    ZFS_TYPE_POOL);
526 
527 		(void) fprintf(fp, "\t%-19s   ", "feature@...");
528 		(void) fprintf(fp, "YES   disabled | enabled | active\n");
529 
530 		(void) fprintf(fp, gettext("\nThe feature@ properties must be "
531 		    "appended with a feature name.\nSee zpool-features(5).\n"));
532 	}
533 
534 	/*
535 	 * See comments at end of main().
536 	 */
537 	if (getenv("ZFS_ABORT") != NULL) {
538 		(void) printf("dumping core by request\n");
539 		abort();
540 	}
541 
542 	exit(requested ? 0 : 2);
543 }
544 
545 /*
546  * zpool initialize [-c | -s] [-w] <pool> [<vdev> ...]
547  * Initialize all unused blocks in the specified vdevs, or all vdevs in the pool
548  * if none specified.
549  *
550  *	-c	Cancel. Ends active initializing.
551  *	-s	Suspend. Initializing can then be restarted with no flags.
552  *	-w	Wait. Blocks until initializing has completed.
553  */
554 int
555 zpool_do_initialize(int argc, char **argv)
556 {
557 	int c;
558 	char *poolname;
559 	zpool_handle_t *zhp;
560 	nvlist_t *vdevs;
561 	int err = 0;
562 	boolean_t wait = B_FALSE;
563 
564 	struct option long_options[] = {
565 		{"cancel",	no_argument,		NULL, 'c'},
566 		{"suspend",	no_argument,		NULL, 's'},
567 		{"wait",	no_argument,		NULL, 'w'},
568 		{0, 0, 0, 0}
569 	};
570 
571 	pool_initialize_func_t cmd_type = POOL_INITIALIZE_START;
572 	while ((c = getopt_long(argc, argv, "csw", long_options, NULL)) != -1) {
573 		switch (c) {
574 		case 'c':
575 			if (cmd_type != POOL_INITIALIZE_START &&
576 			    cmd_type != POOL_INITIALIZE_CANCEL) {
577 				(void) fprintf(stderr, gettext("-c cannot be "
578 				    "combined with other options\n"));
579 				usage(B_FALSE);
580 			}
581 			cmd_type = POOL_INITIALIZE_CANCEL;
582 			break;
583 		case 's':
584 			if (cmd_type != POOL_INITIALIZE_START &&
585 			    cmd_type != POOL_INITIALIZE_SUSPEND) {
586 				(void) fprintf(stderr, gettext("-s cannot be "
587 				    "combined with other options\n"));
588 				usage(B_FALSE);
589 			}
590 			cmd_type = POOL_INITIALIZE_SUSPEND;
591 			break;
592 		case 'w':
593 			wait = B_TRUE;
594 			break;
595 		case '?':
596 			if (optopt != 0) {
597 				(void) fprintf(stderr,
598 				    gettext("invalid option '%c'\n"), optopt);
599 			} else {
600 				(void) fprintf(stderr,
601 				    gettext("invalid option '%s'\n"),
602 				    argv[optind - 1]);
603 			}
604 			usage(B_FALSE);
605 		}
606 	}
607 
608 	argc -= optind;
609 	argv += optind;
610 
611 	if (argc < 1) {
612 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
613 		usage(B_FALSE);
614 		return (-1);
615 	}
616 
617 	if (wait && (cmd_type != POOL_INITIALIZE_START)) {
618 		(void) fprintf(stderr, gettext("-w cannot be used with -c or "
619 		    "-s\n"));
620 		usage(B_FALSE);
621 	}
622 
623 	poolname = argv[0];
624 	zhp = zpool_open(g_zfs, poolname);
625 	if (zhp == NULL)
626 		return (-1);
627 
628 	vdevs = fnvlist_alloc();
629 	if (argc == 1) {
630 		/* no individual leaf vdevs specified, so add them all */
631 		nvlist_t *config = zpool_get_config(zhp, NULL);
632 		nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
633 		    ZPOOL_CONFIG_VDEV_TREE);
634 		zpool_collect_leaves(zhp, nvroot, vdevs);
635 	} else {
636 		for (int i = 1; i < argc; i++) {
637 			fnvlist_add_boolean(vdevs, argv[i]);
638 		}
639 	}
640 
641 	if (wait)
642 		err = zpool_initialize_wait(zhp, cmd_type, vdevs);
643 	else
644 		err = zpool_initialize(zhp, cmd_type, vdevs);
645 
646 	fnvlist_free(vdevs);
647 	zpool_close(zhp);
648 
649 	return (err);
650 }
651 
652 /*
653  * print a pool vdev config for dry runs
654  */
655 static void
656 print_vdev_tree(zpool_handle_t *zhp, const char *name, nvlist_t *nv, int indent,
657     const char *match, int name_flags)
658 {
659 	nvlist_t **child;
660 	uint_t c, children;
661 	char *vname;
662 	boolean_t printed = B_FALSE;
663 
664 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
665 	    &child, &children) != 0) {
666 		if (name != NULL)
667 			(void) printf("\t%*s%s\n", indent, "", name);
668 		return;
669 	}
670 
671 	for (c = 0; c < children; c++) {
672 		uint64_t is_log = B_FALSE, is_hole = B_FALSE;
673 		char *class = "";
674 
675 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
676 		    &is_hole);
677 
678 		if (is_hole == B_TRUE) {
679 			continue;
680 		}
681 
682 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
683 		    &is_log);
684 		if (is_log)
685 			class = VDEV_ALLOC_BIAS_LOG;
686 		(void) nvlist_lookup_string(child[c],
687 		    ZPOOL_CONFIG_ALLOCATION_BIAS, &class);
688 		if (strcmp(match, class) != 0)
689 			continue;
690 
691 		if (!printed && name != NULL) {
692 			(void) printf("\t%*s%s\n", indent, "", name);
693 			printed = B_TRUE;
694 		}
695 		vname = zpool_vdev_name(g_zfs, zhp, child[c], name_flags);
696 		print_vdev_tree(zhp, vname, child[c], indent + 2, "",
697 		    name_flags);
698 		free(vname);
699 	}
700 }
701 
702 /*
703  * Print the list of l2cache devices for dry runs.
704  */
705 static void
706 print_cache_list(nvlist_t *nv, int indent)
707 {
708 	nvlist_t **child;
709 	uint_t c, children;
710 
711 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
712 	    &child, &children) == 0 && children > 0) {
713 		(void) printf("\t%*s%s\n", indent, "", "cache");
714 	} else {
715 		return;
716 	}
717 	for (c = 0; c < children; c++) {
718 		char *vname;
719 
720 		vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
721 		(void) printf("\t%*s%s\n", indent + 2, "", vname);
722 		free(vname);
723 	}
724 }
725 
726 /*
727  * Print the list of spares for dry runs.
728  */
729 static void
730 print_spare_list(nvlist_t *nv, int indent)
731 {
732 	nvlist_t **child;
733 	uint_t c, children;
734 
735 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
736 	    &child, &children) == 0 && children > 0) {
737 		(void) printf("\t%*s%s\n", indent, "", "spares");
738 	} else {
739 		return;
740 	}
741 	for (c = 0; c < children; c++) {
742 		char *vname;
743 
744 		vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
745 		(void) printf("\t%*s%s\n", indent + 2, "", vname);
746 		free(vname);
747 	}
748 }
749 
750 static boolean_t
751 prop_list_contains_feature(nvlist_t *proplist)
752 {
753 	nvpair_t *nvp;
754 	for (nvp = nvlist_next_nvpair(proplist, NULL); NULL != nvp;
755 	    nvp = nvlist_next_nvpair(proplist, nvp)) {
756 		if (zpool_prop_feature(nvpair_name(nvp)))
757 			return (B_TRUE);
758 	}
759 	return (B_FALSE);
760 }
761 
762 /*
763  * Add a property pair (name, string-value) into a property nvlist.
764  */
765 static int
766 add_prop_list(const char *propname, char *propval, nvlist_t **props,
767     boolean_t poolprop)
768 {
769 	zpool_prop_t prop = ZPOOL_PROP_INVAL;
770 	nvlist_t *proplist;
771 	const char *normnm;
772 	char *strval;
773 
774 	if (*props == NULL &&
775 	    nvlist_alloc(props, NV_UNIQUE_NAME, 0) != 0) {
776 		(void) fprintf(stderr,
777 		    gettext("internal error: out of memory\n"));
778 		return (1);
779 	}
780 
781 	proplist = *props;
782 
783 	if (poolprop) {
784 		const char *vname = zpool_prop_to_name(ZPOOL_PROP_VERSION);
785 
786 		if ((prop = zpool_name_to_prop(propname)) == ZPOOL_PROP_INVAL &&
787 		    !zpool_prop_feature(propname)) {
788 			(void) fprintf(stderr, gettext("property '%s' is "
789 			    "not a valid pool property\n"), propname);
790 			return (2);
791 		}
792 
793 		/*
794 		 * feature@ properties and version should not be specified
795 		 * at the same time.
796 		 */
797 		if ((prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname) &&
798 		    nvlist_exists(proplist, vname)) ||
799 		    (prop == ZPOOL_PROP_VERSION &&
800 		    prop_list_contains_feature(proplist))) {
801 			(void) fprintf(stderr, gettext("'feature@' and "
802 			    "'version' properties cannot be specified "
803 			    "together\n"));
804 			return (2);
805 		}
806 
807 
808 		if (zpool_prop_feature(propname))
809 			normnm = propname;
810 		else
811 			normnm = zpool_prop_to_name(prop);
812 	} else {
813 		zfs_prop_t fsprop = zfs_name_to_prop(propname);
814 
815 		if (zfs_prop_valid_for_type(fsprop, ZFS_TYPE_FILESYSTEM,
816 		    B_FALSE)) {
817 			normnm = zfs_prop_to_name(fsprop);
818 		} else if (zfs_prop_user(propname) ||
819 		    zfs_prop_userquota(propname)) {
820 			normnm = propname;
821 		} else {
822 			(void) fprintf(stderr, gettext("property '%s' is "
823 			    "not a valid filesystem property\n"), propname);
824 			return (2);
825 		}
826 	}
827 
828 	if (nvlist_lookup_string(proplist, normnm, &strval) == 0 &&
829 	    prop != ZPOOL_PROP_CACHEFILE) {
830 		(void) fprintf(stderr, gettext("property '%s' "
831 		    "specified multiple times\n"), propname);
832 		return (2);
833 	}
834 
835 	if (nvlist_add_string(proplist, normnm, propval) != 0) {
836 		(void) fprintf(stderr, gettext("internal "
837 		    "error: out of memory\n"));
838 		return (1);
839 	}
840 
841 	return (0);
842 }
843 
844 /*
845  * Set a default property pair (name, string-value) in a property nvlist
846  */
847 static int
848 add_prop_list_default(const char *propname, char *propval, nvlist_t **props,
849     boolean_t poolprop)
850 {
851 	char *pval;
852 
853 	if (nvlist_lookup_string(*props, propname, &pval) == 0)
854 		return (0);
855 
856 	return (add_prop_list(propname, propval, props, B_TRUE));
857 }
858 
859 /*
860  * zpool add [-fgLnP] [-o property=value] <pool> <vdev> ...
861  *
862  *	-f	Force addition of devices, even if they appear in use
863  *	-g	Display guid for individual vdev name.
864  *	-L	Follow links when resolving vdev path name.
865  *	-n	Do not add the devices, but display the resulting layout if
866  *		they were to be added.
867  *	-o	Set property=value.
868  *	-P	Display full path for vdev name.
869  *
870  * Adds the given vdevs to 'pool'.  As with create, the bulk of this work is
871  * handled by make_root_vdev(), which constructs the nvlist needed to pass to
872  * libzfs.
873  */
874 int
875 zpool_do_add(int argc, char **argv)
876 {
877 	boolean_t force = B_FALSE;
878 	boolean_t dryrun = B_FALSE;
879 	int name_flags = 0;
880 	int c;
881 	nvlist_t *nvroot;
882 	char *poolname;
883 	int ret;
884 	zpool_handle_t *zhp;
885 	nvlist_t *config;
886 	nvlist_t *props = NULL;
887 	char *propval;
888 
889 	/* check options */
890 	while ((c = getopt(argc, argv, "fgLno:P")) != -1) {
891 		switch (c) {
892 		case 'f':
893 			force = B_TRUE;
894 			break;
895 		case 'g':
896 			name_flags |= VDEV_NAME_GUID;
897 			break;
898 		case 'L':
899 			name_flags |= VDEV_NAME_FOLLOW_LINKS;
900 			break;
901 		case 'n':
902 			dryrun = B_TRUE;
903 			break;
904 		case 'o':
905 			if ((propval = strchr(optarg, '=')) == NULL) {
906 				(void) fprintf(stderr, gettext("missing "
907 				    "'=' for -o option\n"));
908 				usage(B_FALSE);
909 			}
910 			*propval = '\0';
911 			propval++;
912 
913 			if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) ||
914 			    (add_prop_list(optarg, propval, &props, B_TRUE)))
915 				usage(B_FALSE);
916 			break;
917 		case 'P':
918 			name_flags |= VDEV_NAME_PATH;
919 			break;
920 		case '?':
921 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
922 			    optopt);
923 			usage(B_FALSE);
924 		}
925 	}
926 
927 	argc -= optind;
928 	argv += optind;
929 
930 	/* get pool name and check number of arguments */
931 	if (argc < 1) {
932 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
933 		usage(B_FALSE);
934 	}
935 	if (argc < 2) {
936 		(void) fprintf(stderr, gettext("missing vdev specification\n"));
937 		usage(B_FALSE);
938 	}
939 
940 	poolname = argv[0];
941 
942 	argc--;
943 	argv++;
944 
945 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
946 		return (1);
947 
948 	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
949 		(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
950 		    poolname);
951 		zpool_close(zhp);
952 		return (1);
953 	}
954 
955 	/* unless manually specified use "ashift" pool property (if set) */
956 	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
957 		int intval;
958 		zprop_source_t src;
959 		char strval[ZPOOL_MAXPROPLEN];
960 
961 		intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
962 		if (src != ZPROP_SRC_DEFAULT) {
963 			(void) sprintf(strval, "%" PRId32, intval);
964 			verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
965 			    &props, B_TRUE) == 0);
966 		}
967 	}
968 
969 	/* pass off to make_root_vdev for processing */
970 	nvroot = make_root_vdev(zhp, props, force, !force, B_FALSE, dryrun,
971 	    argc, argv);
972 	if (nvroot == NULL) {
973 		zpool_close(zhp);
974 		return (1);
975 	}
976 
977 	if (dryrun) {
978 		nvlist_t *poolnvroot;
979 		nvlist_t **l2child, **sparechild;
980 		uint_t l2children, sparechildren, c;
981 		char *vname;
982 		boolean_t hadcache = B_FALSE, hadspare = B_FALSE;
983 
984 		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
985 		    &poolnvroot) == 0);
986 
987 		(void) printf(gettext("would update '%s' to the following "
988 		    "configuration:\n\n"), zpool_get_name(zhp));
989 
990 		/* print original main pool and new tree */
991 		print_vdev_tree(zhp, poolname, poolnvroot, 0, "",
992 		    name_flags | VDEV_NAME_TYPE_ID);
993 		print_vdev_tree(zhp, NULL, nvroot, 0, "", name_flags);
994 
995 		/* print other classes: 'dedup', 'special', and 'log' */
996 		if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_DEDUP)) {
997 			print_vdev_tree(zhp, "dedup", poolnvroot, 0,
998 			    VDEV_ALLOC_BIAS_DEDUP, name_flags);
999 			print_vdev_tree(zhp, NULL, nvroot, 0,
1000 			    VDEV_ALLOC_BIAS_DEDUP, name_flags);
1001 		} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_DEDUP)) {
1002 			print_vdev_tree(zhp, "dedup", nvroot, 0,
1003 			    VDEV_ALLOC_BIAS_DEDUP, name_flags);
1004 		}
1005 
1006 		if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
1007 			print_vdev_tree(zhp, "special", poolnvroot, 0,
1008 			    VDEV_ALLOC_BIAS_SPECIAL, name_flags);
1009 			print_vdev_tree(zhp, NULL, nvroot, 0,
1010 			    VDEV_ALLOC_BIAS_SPECIAL, name_flags);
1011 		} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
1012 			print_vdev_tree(zhp, "special", nvroot, 0,
1013 			    VDEV_ALLOC_BIAS_SPECIAL, name_flags);
1014 		}
1015 
1016 		if (num_logs(poolnvroot) > 0) {
1017 			print_vdev_tree(zhp, "logs", poolnvroot, 0,
1018 			    VDEV_ALLOC_BIAS_LOG, name_flags);
1019 			print_vdev_tree(zhp, NULL, nvroot, 0,
1020 			    VDEV_ALLOC_BIAS_LOG, name_flags);
1021 		} else if (num_logs(nvroot) > 0) {
1022 			print_vdev_tree(zhp, "logs", nvroot, 0,
1023 			    VDEV_ALLOC_BIAS_LOG, name_flags);
1024 		}
1025 
1026 		/* Do the same for the caches */
1027 		if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_L2CACHE,
1028 		    &l2child, &l2children) == 0 && l2children) {
1029 			hadcache = B_TRUE;
1030 			(void) printf(gettext("\tcache\n"));
1031 			for (c = 0; c < l2children; c++) {
1032 				vname = zpool_vdev_name(g_zfs, NULL,
1033 				    l2child[c], name_flags);
1034 				(void) printf("\t  %s\n", vname);
1035 				free(vname);
1036 			}
1037 		}
1038 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1039 		    &l2child, &l2children) == 0 && l2children) {
1040 			if (!hadcache)
1041 				(void) printf(gettext("\tcache\n"));
1042 			for (c = 0; c < l2children; c++) {
1043 				vname = zpool_vdev_name(g_zfs, NULL,
1044 				    l2child[c], name_flags);
1045 				(void) printf("\t  %s\n", vname);
1046 				free(vname);
1047 			}
1048 		}
1049 		/* And finaly the spares */
1050 		if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_SPARES,
1051 		    &sparechild, &sparechildren) == 0 && sparechildren > 0) {
1052 			hadspare = B_TRUE;
1053 			(void) printf(gettext("\tspares\n"));
1054 			for (c = 0; c < sparechildren; c++) {
1055 				vname = zpool_vdev_name(g_zfs, NULL,
1056 				    sparechild[c], name_flags);
1057 				(void) printf("\t  %s\n", vname);
1058 				free(vname);
1059 			}
1060 		}
1061 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1062 		    &sparechild, &sparechildren) == 0 && sparechildren > 0) {
1063 			if (!hadspare)
1064 				(void) printf(gettext("\tspares\n"));
1065 			for (c = 0; c < sparechildren; c++) {
1066 				vname = zpool_vdev_name(g_zfs, NULL,
1067 				    sparechild[c], name_flags);
1068 				(void) printf("\t  %s\n", vname);
1069 				free(vname);
1070 			}
1071 		}
1072 
1073 		ret = 0;
1074 	} else {
1075 		ret = (zpool_add(zhp, nvroot) != 0);
1076 	}
1077 
1078 	nvlist_free(props);
1079 	nvlist_free(nvroot);
1080 	zpool_close(zhp);
1081 
1082 	return (ret);
1083 }
1084 
1085 /*
1086  * zpool remove [-npsw] <pool> <vdev> ...
1087  *
1088  * Removes the given vdev from the pool.
1089  */
1090 int
1091 zpool_do_remove(int argc, char **argv)
1092 {
1093 	char *poolname;
1094 	int i, ret = 0;
1095 	zpool_handle_t *zhp = NULL;
1096 	boolean_t stop = B_FALSE;
1097 	int c;
1098 	boolean_t noop = B_FALSE;
1099 	boolean_t parsable = B_FALSE;
1100 	boolean_t wait = B_FALSE;
1101 
1102 	/* check options */
1103 	while ((c = getopt(argc, argv, "npsw")) != -1) {
1104 		switch (c) {
1105 		case 'n':
1106 			noop = B_TRUE;
1107 			break;
1108 		case 'p':
1109 			parsable = B_TRUE;
1110 			break;
1111 		case 's':
1112 			stop = B_TRUE;
1113 			break;
1114 		case 'w':
1115 			wait = B_TRUE;
1116 			break;
1117 		case '?':
1118 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1119 			    optopt);
1120 			usage(B_FALSE);
1121 		}
1122 	}
1123 
1124 	argc -= optind;
1125 	argv += optind;
1126 
1127 	/* get pool name and check number of arguments */
1128 	if (argc < 1) {
1129 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
1130 		usage(B_FALSE);
1131 	}
1132 
1133 	poolname = argv[0];
1134 
1135 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
1136 		return (1);
1137 
1138 	if (stop && noop) {
1139 		(void) fprintf(stderr, gettext("stop request ignored\n"));
1140 		return (0);
1141 	}
1142 
1143 	if (stop) {
1144 		if (argc > 1) {
1145 			(void) fprintf(stderr, gettext("too many arguments\n"));
1146 			usage(B_FALSE);
1147 		}
1148 		if (zpool_vdev_remove_cancel(zhp) != 0)
1149 			ret = 1;
1150 		if (wait) {
1151 			(void) fprintf(stderr, gettext("invalid option "
1152 			    "combination: -w cannot be used with -s\n"));
1153 			usage(B_FALSE);
1154 		}
1155 	} else {
1156 		if (argc < 2) {
1157 			(void) fprintf(stderr, gettext("missing device\n"));
1158 			usage(B_FALSE);
1159 		}
1160 
1161 		for (i = 1; i < argc; i++) {
1162 			if (noop) {
1163 				uint64_t size;
1164 
1165 				if (zpool_vdev_indirect_size(zhp, argv[i],
1166 				    &size) != 0) {
1167 					ret = 1;
1168 					break;
1169 				}
1170 				if (parsable) {
1171 					(void) printf("%s %llu\n",
1172 					    argv[i], (unsigned long long)size);
1173 				} else {
1174 					char valstr[32];
1175 					zfs_nicenum(size, valstr,
1176 					    sizeof (valstr));
1177 					(void) printf("Memory that will be "
1178 					    "used after removing %s: %s\n",
1179 					    argv[i], valstr);
1180 				}
1181 			} else {
1182 				if (zpool_vdev_remove(zhp, argv[i]) != 0)
1183 					ret = 1;
1184 			}
1185 		}
1186 
1187 		if (ret == 0 && wait)
1188 			ret = zpool_wait(zhp, ZPOOL_WAIT_REMOVE);
1189 	}
1190 	zpool_close(zhp);
1191 
1192 	return (ret);
1193 }
1194 
1195 /*
1196  * zpool labelclear [-f] <vdev>
1197  *
1198  *	-f	Force clearing the label for the vdevs which are members of
1199  *		the exported or foreign pools.
1200  *
1201  * Verifies that the vdev is not active and zeros out the label information
1202  * on the device.
1203  */
1204 int
1205 zpool_do_labelclear(int argc, char **argv)
1206 {
1207 	char vdev[MAXPATHLEN];
1208 	char *name = NULL;
1209 	struct stat st;
1210 	int c, fd = -1, ret = 0;
1211 	nvlist_t *config;
1212 	pool_state_t state;
1213 	boolean_t inuse = B_FALSE;
1214 	boolean_t force = B_FALSE;
1215 
1216 	/* check options */
1217 	while ((c = getopt(argc, argv, "f")) != -1) {
1218 		switch (c) {
1219 		case 'f':
1220 			force = B_TRUE;
1221 			break;
1222 		default:
1223 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1224 			    optopt);
1225 			usage(B_FALSE);
1226 		}
1227 	}
1228 
1229 	argc -= optind;
1230 	argv += optind;
1231 
1232 	/* get vdev name */
1233 	if (argc < 1) {
1234 		(void) fprintf(stderr, gettext("missing vdev name\n"));
1235 		usage(B_FALSE);
1236 	}
1237 	if (argc > 1) {
1238 		(void) fprintf(stderr, gettext("too many arguments\n"));
1239 		usage(B_FALSE);
1240 	}
1241 
1242 	/*
1243 	 * Check if we were given absolute path and use it as is.
1244 	 * Otherwise if the provided vdev name doesn't point to a file,
1245 	 * try prepending expected disk paths and partition numbers.
1246 	 */
1247 	(void) strlcpy(vdev, argv[0], sizeof (vdev));
1248 	if (vdev[0] != '/' && stat(vdev, &st) != 0) {
1249 		int error;
1250 
1251 		error = zfs_resolve_shortname(argv[0], vdev, MAXPATHLEN);
1252 		if (error == 0 && zfs_dev_is_whole_disk(vdev)) {
1253 			if (zfs_append_partition(vdev, MAXPATHLEN) == -1)
1254 				error = ENOENT;
1255 		}
1256 
1257 		if (error || (stat(vdev, &st) != 0)) {
1258 			(void) fprintf(stderr, gettext(
1259 			    "failed to find device %s, try specifying absolute "
1260 			    "path instead\n"), argv[0]);
1261 			return (1);
1262 		}
1263 	}
1264 
1265 	if ((fd = open(vdev, O_RDWR)) < 0) {
1266 		(void) fprintf(stderr, gettext("failed to open %s: %s\n"),
1267 		    vdev, strerror(errno));
1268 		return (1);
1269 	}
1270 
1271 	/*
1272 	 * Flush all dirty pages for the block device.  This should not be
1273 	 * fatal when the device does not support BLKFLSBUF as would be the
1274 	 * case for a file vdev.
1275 	 */
1276 	if ((zfs_dev_flush(fd) != 0) && (errno != ENOTTY))
1277 		(void) fprintf(stderr, gettext("failed to invalidate "
1278 		    "cache for %s: %s\n"), vdev, strerror(errno));
1279 
1280 	if (zpool_read_label(fd, &config, NULL) != 0) {
1281 		(void) fprintf(stderr,
1282 		    gettext("failed to read label from %s\n"), vdev);
1283 		ret = 1;
1284 		goto errout;
1285 	}
1286 	nvlist_free(config);
1287 
1288 	ret = zpool_in_use(g_zfs, fd, &state, &name, &inuse);
1289 	if (ret != 0) {
1290 		(void) fprintf(stderr,
1291 		    gettext("failed to check state for %s\n"), vdev);
1292 		ret = 1;
1293 		goto errout;
1294 	}
1295 
1296 	if (!inuse)
1297 		goto wipe_label;
1298 
1299 	switch (state) {
1300 	default:
1301 	case POOL_STATE_ACTIVE:
1302 	case POOL_STATE_SPARE:
1303 	case POOL_STATE_L2CACHE:
1304 		(void) fprintf(stderr, gettext(
1305 		    "%s is a member (%s) of pool \"%s\"\n"),
1306 		    vdev, zpool_pool_state_to_name(state), name);
1307 		ret = 1;
1308 		goto errout;
1309 
1310 	case POOL_STATE_EXPORTED:
1311 		if (force)
1312 			break;
1313 		(void) fprintf(stderr, gettext(
1314 		    "use '-f' to override the following error:\n"
1315 		    "%s is a member of exported pool \"%s\"\n"),
1316 		    vdev, name);
1317 		ret = 1;
1318 		goto errout;
1319 
1320 	case POOL_STATE_POTENTIALLY_ACTIVE:
1321 		if (force)
1322 			break;
1323 		(void) fprintf(stderr, gettext(
1324 		    "use '-f' to override the following error:\n"
1325 		    "%s is a member of potentially active pool \"%s\"\n"),
1326 		    vdev, name);
1327 		ret = 1;
1328 		goto errout;
1329 
1330 	case POOL_STATE_DESTROYED:
1331 		/* inuse should never be set for a destroyed pool */
1332 		assert(0);
1333 		break;
1334 	}
1335 
1336 wipe_label:
1337 	ret = zpool_clear_label(fd);
1338 	if (ret != 0) {
1339 		(void) fprintf(stderr,
1340 		    gettext("failed to clear label for %s\n"), vdev);
1341 	}
1342 
1343 errout:
1344 	free(name);
1345 	(void) close(fd);
1346 
1347 	return (ret);
1348 }
1349 
1350 /*
1351  * zpool create [-fnd] [-o property=value] ...
1352  *		[-O file-system-property=value] ...
1353  *		[-R root] [-m mountpoint] <pool> <dev> ...
1354  *
1355  *	-f	Force creation, even if devices appear in use
1356  *	-n	Do not create the pool, but display the resulting layout if it
1357  *		were to be created.
1358  *      -R	Create a pool under an alternate root
1359  *      -m	Set default mountpoint for the root dataset.  By default it's
1360  *		'/<pool>'
1361  *	-o	Set property=value.
1362  *	-o	Set feature@feature=enabled|disabled.
1363  *	-d	Don't automatically enable all supported pool features
1364  *		(individual features can be enabled with -o).
1365  *	-O	Set fsproperty=value in the pool's root file system
1366  *
1367  * Creates the named pool according to the given vdev specification.  The
1368  * bulk of the vdev processing is done in make_root_vdev() in zpool_vdev.c.
1369  * Once we get the nvlist back from make_root_vdev(), we either print out the
1370  * contents (if '-n' was specified), or pass it to libzfs to do the creation.
1371  */
1372 int
1373 zpool_do_create(int argc, char **argv)
1374 {
1375 	boolean_t force = B_FALSE;
1376 	boolean_t dryrun = B_FALSE;
1377 	boolean_t enable_all_pool_feat = B_TRUE;
1378 	int c;
1379 	nvlist_t *nvroot = NULL;
1380 	char *poolname;
1381 	char *tname = NULL;
1382 	int ret = 1;
1383 	char *altroot = NULL;
1384 	char *mountpoint = NULL;
1385 	nvlist_t *fsprops = NULL;
1386 	nvlist_t *props = NULL;
1387 	char *propval;
1388 
1389 	/* check options */
1390 	while ((c = getopt(argc, argv, ":fndR:m:o:O:t:")) != -1) {
1391 		switch (c) {
1392 		case 'f':
1393 			force = B_TRUE;
1394 			break;
1395 		case 'n':
1396 			dryrun = B_TRUE;
1397 			break;
1398 		case 'd':
1399 			enable_all_pool_feat = B_FALSE;
1400 			break;
1401 		case 'R':
1402 			altroot = optarg;
1403 			if (add_prop_list(zpool_prop_to_name(
1404 			    ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
1405 				goto errout;
1406 			if (add_prop_list_default(zpool_prop_to_name(
1407 			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
1408 				goto errout;
1409 			break;
1410 		case 'm':
1411 			/* Equivalent to -O mountpoint=optarg */
1412 			mountpoint = optarg;
1413 			break;
1414 		case 'o':
1415 			if ((propval = strchr(optarg, '=')) == NULL) {
1416 				(void) fprintf(stderr, gettext("missing "
1417 				    "'=' for -o option\n"));
1418 				goto errout;
1419 			}
1420 			*propval = '\0';
1421 			propval++;
1422 
1423 			if (add_prop_list(optarg, propval, &props, B_TRUE))
1424 				goto errout;
1425 
1426 			/*
1427 			 * If the user is creating a pool that doesn't support
1428 			 * feature flags, don't enable any features.
1429 			 */
1430 			if (zpool_name_to_prop(optarg) == ZPOOL_PROP_VERSION) {
1431 				char *end;
1432 				u_longlong_t ver;
1433 
1434 				ver = strtoull(propval, &end, 10);
1435 				if (*end == '\0' &&
1436 				    ver < SPA_VERSION_FEATURES) {
1437 					enable_all_pool_feat = B_FALSE;
1438 				}
1439 			}
1440 			if (zpool_name_to_prop(optarg) == ZPOOL_PROP_ALTROOT)
1441 				altroot = propval;
1442 			break;
1443 		case 'O':
1444 			if ((propval = strchr(optarg, '=')) == NULL) {
1445 				(void) fprintf(stderr, gettext("missing "
1446 				    "'=' for -O option\n"));
1447 				goto errout;
1448 			}
1449 			*propval = '\0';
1450 			propval++;
1451 
1452 			/*
1453 			 * Mountpoints are checked and then added later.
1454 			 * Uniquely among properties, they can be specified
1455 			 * more than once, to avoid conflict with -m.
1456 			 */
1457 			if (0 == strcmp(optarg,
1458 			    zfs_prop_to_name(ZFS_PROP_MOUNTPOINT))) {
1459 				mountpoint = propval;
1460 			} else if (add_prop_list(optarg, propval, &fsprops,
1461 			    B_FALSE)) {
1462 				goto errout;
1463 			}
1464 			break;
1465 		case 't':
1466 			/*
1467 			 * Sanity check temporary pool name.
1468 			 */
1469 			if (strchr(optarg, '/') != NULL) {
1470 				(void) fprintf(stderr, gettext("cannot create "
1471 				    "'%s': invalid character '/' in temporary "
1472 				    "name\n"), optarg);
1473 				(void) fprintf(stderr, gettext("use 'zfs "
1474 				    "create' to create a dataset\n"));
1475 				goto errout;
1476 			}
1477 
1478 			if (add_prop_list(zpool_prop_to_name(
1479 			    ZPOOL_PROP_TNAME), optarg, &props, B_TRUE))
1480 				goto errout;
1481 			if (add_prop_list_default(zpool_prop_to_name(
1482 			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
1483 				goto errout;
1484 			tname = optarg;
1485 			break;
1486 		case ':':
1487 			(void) fprintf(stderr, gettext("missing argument for "
1488 			    "'%c' option\n"), optopt);
1489 			goto badusage;
1490 		case '?':
1491 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1492 			    optopt);
1493 			goto badusage;
1494 		}
1495 	}
1496 
1497 	argc -= optind;
1498 	argv += optind;
1499 
1500 	/* get pool name and check number of arguments */
1501 	if (argc < 1) {
1502 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
1503 		goto badusage;
1504 	}
1505 	if (argc < 2) {
1506 		(void) fprintf(stderr, gettext("missing vdev specification\n"));
1507 		goto badusage;
1508 	}
1509 
1510 	poolname = argv[0];
1511 
1512 	/*
1513 	 * As a special case, check for use of '/' in the name, and direct the
1514 	 * user to use 'zfs create' instead.
1515 	 */
1516 	if (strchr(poolname, '/') != NULL) {
1517 		(void) fprintf(stderr, gettext("cannot create '%s': invalid "
1518 		    "character '/' in pool name\n"), poolname);
1519 		(void) fprintf(stderr, gettext("use 'zfs create' to "
1520 		    "create a dataset\n"));
1521 		goto errout;
1522 	}
1523 
1524 	/* pass off to make_root_vdev for bulk processing */
1525 	nvroot = make_root_vdev(NULL, props, force, !force, B_FALSE, dryrun,
1526 	    argc - 1, argv + 1);
1527 	if (nvroot == NULL)
1528 		goto errout;
1529 
1530 	/* make_root_vdev() allows 0 toplevel children if there are spares */
1531 	if (!zfs_allocatable_devs(nvroot)) {
1532 		(void) fprintf(stderr, gettext("invalid vdev "
1533 		    "specification: at least one toplevel vdev must be "
1534 		    "specified\n"));
1535 		goto errout;
1536 	}
1537 
1538 	if (altroot != NULL && altroot[0] != '/') {
1539 		(void) fprintf(stderr, gettext("invalid alternate root '%s': "
1540 		    "must be an absolute path\n"), altroot);
1541 		goto errout;
1542 	}
1543 
1544 	/*
1545 	 * Check the validity of the mountpoint and direct the user to use the
1546 	 * '-m' mountpoint option if it looks like its in use.
1547 	 */
1548 	if (mountpoint == NULL ||
1549 	    (strcmp(mountpoint, ZFS_MOUNTPOINT_LEGACY) != 0 &&
1550 	    strcmp(mountpoint, ZFS_MOUNTPOINT_NONE) != 0)) {
1551 		char buf[MAXPATHLEN];
1552 		DIR *dirp;
1553 
1554 		if (mountpoint && mountpoint[0] != '/') {
1555 			(void) fprintf(stderr, gettext("invalid mountpoint "
1556 			    "'%s': must be an absolute path, 'legacy', or "
1557 			    "'none'\n"), mountpoint);
1558 			goto errout;
1559 		}
1560 
1561 		if (mountpoint == NULL) {
1562 			if (altroot != NULL)
1563 				(void) snprintf(buf, sizeof (buf), "%s/%s",
1564 				    altroot, poolname);
1565 			else
1566 				(void) snprintf(buf, sizeof (buf), "/%s",
1567 				    poolname);
1568 		} else {
1569 			if (altroot != NULL)
1570 				(void) snprintf(buf, sizeof (buf), "%s%s",
1571 				    altroot, mountpoint);
1572 			else
1573 				(void) snprintf(buf, sizeof (buf), "%s",
1574 				    mountpoint);
1575 		}
1576 
1577 		if ((dirp = opendir(buf)) == NULL && errno != ENOENT) {
1578 			(void) fprintf(stderr, gettext("mountpoint '%s' : "
1579 			    "%s\n"), buf, strerror(errno));
1580 			(void) fprintf(stderr, gettext("use '-m' "
1581 			    "option to provide a different default\n"));
1582 			goto errout;
1583 		} else if (dirp) {
1584 			int count = 0;
1585 
1586 			while (count < 3 && readdir(dirp) != NULL)
1587 				count++;
1588 			(void) closedir(dirp);
1589 
1590 			if (count > 2) {
1591 				(void) fprintf(stderr, gettext("mountpoint "
1592 				    "'%s' exists and is not empty\n"), buf);
1593 				(void) fprintf(stderr, gettext("use '-m' "
1594 				    "option to provide a "
1595 				    "different default\n"));
1596 				goto errout;
1597 			}
1598 		}
1599 	}
1600 
1601 	/*
1602 	 * Now that the mountpoint's validity has been checked, ensure that
1603 	 * the property is set appropriately prior to creating the pool.
1604 	 */
1605 	if (mountpoint != NULL) {
1606 		ret = add_prop_list(zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
1607 		    mountpoint, &fsprops, B_FALSE);
1608 		if (ret != 0)
1609 			goto errout;
1610 	}
1611 
1612 	ret = 1;
1613 	if (dryrun) {
1614 		/*
1615 		 * For a dry run invocation, print out a basic message and run
1616 		 * through all the vdevs in the list and print out in an
1617 		 * appropriate hierarchy.
1618 		 */
1619 		(void) printf(gettext("would create '%s' with the "
1620 		    "following layout:\n\n"), poolname);
1621 
1622 		print_vdev_tree(NULL, poolname, nvroot, 0, "", 0);
1623 		print_vdev_tree(NULL, "dedup", nvroot, 0,
1624 		    VDEV_ALLOC_BIAS_DEDUP, 0);
1625 		print_vdev_tree(NULL, "special", nvroot, 0,
1626 		    VDEV_ALLOC_BIAS_SPECIAL, 0);
1627 		print_vdev_tree(NULL, "logs", nvroot, 0,
1628 		    VDEV_ALLOC_BIAS_LOG, 0);
1629 		print_cache_list(nvroot, 0);
1630 		print_spare_list(nvroot, 0);
1631 
1632 		ret = 0;
1633 	} else {
1634 		/*
1635 		 * Hand off to libzfs.
1636 		 */
1637 		spa_feature_t i;
1638 		for (i = 0; i < SPA_FEATURES; i++) {
1639 			char propname[MAXPATHLEN];
1640 			char *propval;
1641 			zfeature_info_t *feat = &spa_feature_table[i];
1642 
1643 			(void) snprintf(propname, sizeof (propname),
1644 			    "feature@%s", feat->fi_uname);
1645 
1646 			/*
1647 			 * Only features contained in props will be enabled:
1648 			 * remove from the nvlist every ZFS_FEATURE_DISABLED
1649 			 * value and add every missing ZFS_FEATURE_ENABLED if
1650 			 * enable_all_pool_feat is set.
1651 			 */
1652 			if (!nvlist_lookup_string(props, propname, &propval)) {
1653 				if (strcmp(propval, ZFS_FEATURE_DISABLED) == 0)
1654 					(void) nvlist_remove_all(props,
1655 					    propname);
1656 			} else if (enable_all_pool_feat) {
1657 				ret = add_prop_list(propname,
1658 				    ZFS_FEATURE_ENABLED, &props, B_TRUE);
1659 				if (ret != 0)
1660 					goto errout;
1661 			}
1662 		}
1663 
1664 		ret = 1;
1665 		if (zpool_create(g_zfs, poolname,
1666 		    nvroot, props, fsprops) == 0) {
1667 			zfs_handle_t *pool = zfs_open(g_zfs,
1668 			    tname ? tname : poolname, ZFS_TYPE_FILESYSTEM);
1669 			if (pool != NULL) {
1670 				if (zfs_mount(pool, NULL, 0) == 0) {
1671 					ret = zfs_shareall(pool);
1672 					zfs_commit_all_shares();
1673 				}
1674 				zfs_close(pool);
1675 			}
1676 		} else if (libzfs_errno(g_zfs) == EZFS_INVALIDNAME) {
1677 			(void) fprintf(stderr, gettext("pool name may have "
1678 			    "been omitted\n"));
1679 		}
1680 	}
1681 
1682 errout:
1683 	nvlist_free(nvroot);
1684 	nvlist_free(fsprops);
1685 	nvlist_free(props);
1686 	return (ret);
1687 badusage:
1688 	nvlist_free(fsprops);
1689 	nvlist_free(props);
1690 	usage(B_FALSE);
1691 	return (2);
1692 }
1693 
1694 /*
1695  * zpool destroy <pool>
1696  *
1697  * 	-f	Forcefully unmount any datasets
1698  *
1699  * Destroy the given pool.  Automatically unmounts any datasets in the pool.
1700  */
1701 int
1702 zpool_do_destroy(int argc, char **argv)
1703 {
1704 	boolean_t force = B_FALSE;
1705 	int c;
1706 	char *pool;
1707 	zpool_handle_t *zhp;
1708 	int ret;
1709 
1710 	/* check options */
1711 	while ((c = getopt(argc, argv, "f")) != -1) {
1712 		switch (c) {
1713 		case 'f':
1714 			force = B_TRUE;
1715 			break;
1716 		case '?':
1717 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1718 			    optopt);
1719 			usage(B_FALSE);
1720 		}
1721 	}
1722 
1723 	argc -= optind;
1724 	argv += optind;
1725 
1726 	/* check arguments */
1727 	if (argc < 1) {
1728 		(void) fprintf(stderr, gettext("missing pool argument\n"));
1729 		usage(B_FALSE);
1730 	}
1731 	if (argc > 1) {
1732 		(void) fprintf(stderr, gettext("too many arguments\n"));
1733 		usage(B_FALSE);
1734 	}
1735 
1736 	pool = argv[0];
1737 
1738 	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
1739 		/*
1740 		 * As a special case, check for use of '/' in the name, and
1741 		 * direct the user to use 'zfs destroy' instead.
1742 		 */
1743 		if (strchr(pool, '/') != NULL)
1744 			(void) fprintf(stderr, gettext("use 'zfs destroy' to "
1745 			    "destroy a dataset\n"));
1746 		return (1);
1747 	}
1748 
1749 	if (zpool_disable_datasets(zhp, force) != 0) {
1750 		(void) fprintf(stderr, gettext("could not destroy '%s': "
1751 		    "could not unmount datasets\n"), zpool_get_name(zhp));
1752 		zpool_close(zhp);
1753 		return (1);
1754 	}
1755 
1756 	/* The history must be logged as part of the export */
1757 	log_history = B_FALSE;
1758 
1759 	ret = (zpool_destroy(zhp, history_str) != 0);
1760 
1761 	zpool_close(zhp);
1762 
1763 	return (ret);
1764 }
1765 
1766 typedef struct export_cbdata {
1767 	boolean_t force;
1768 	boolean_t hardforce;
1769 } export_cbdata_t;
1770 
1771 /*
1772  * Export one pool
1773  */
1774 static int
1775 zpool_export_one(zpool_handle_t *zhp, void *data)
1776 {
1777 	export_cbdata_t *cb = data;
1778 
1779 	if (zpool_disable_datasets(zhp, cb->force) != 0)
1780 		return (1);
1781 
1782 	/* The history must be logged as part of the export */
1783 	log_history = B_FALSE;
1784 
1785 	if (cb->hardforce) {
1786 		if (zpool_export_force(zhp, history_str) != 0)
1787 			return (1);
1788 	} else if (zpool_export(zhp, cb->force, history_str) != 0) {
1789 		return (1);
1790 	}
1791 
1792 	return (0);
1793 }
1794 
1795 /*
1796  * zpool export [-f] <pool> ...
1797  *
1798  *	-a	Export all pools
1799  *	-f	Forcefully unmount datasets
1800  *
1801  * Export the given pools.  By default, the command will attempt to cleanly
1802  * unmount any active datasets within the pool.  If the '-f' flag is specified,
1803  * then the datasets will be forcefully unmounted.
1804  */
1805 int
1806 zpool_do_export(int argc, char **argv)
1807 {
1808 	export_cbdata_t cb;
1809 	boolean_t do_all = B_FALSE;
1810 	boolean_t force = B_FALSE;
1811 	boolean_t hardforce = B_FALSE;
1812 	int c, ret;
1813 
1814 	/* check options */
1815 	while ((c = getopt(argc, argv, "afF")) != -1) {
1816 		switch (c) {
1817 		case 'a':
1818 			do_all = B_TRUE;
1819 			break;
1820 		case 'f':
1821 			force = B_TRUE;
1822 			break;
1823 		case 'F':
1824 			hardforce = B_TRUE;
1825 			break;
1826 		case '?':
1827 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
1828 			    optopt);
1829 			usage(B_FALSE);
1830 		}
1831 	}
1832 
1833 	cb.force = force;
1834 	cb.hardforce = hardforce;
1835 	argc -= optind;
1836 	argv += optind;
1837 
1838 	if (do_all) {
1839 		if (argc != 0) {
1840 			(void) fprintf(stderr, gettext("too many arguments\n"));
1841 			usage(B_FALSE);
1842 		}
1843 
1844 		return (for_each_pool(argc, argv, B_TRUE, NULL,
1845 		    B_FALSE, zpool_export_one, &cb));
1846 	}
1847 
1848 	/* check arguments */
1849 	if (argc < 1) {
1850 		(void) fprintf(stderr, gettext("missing pool argument\n"));
1851 		usage(B_FALSE);
1852 	}
1853 
1854 	ret = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE, zpool_export_one,
1855 	    &cb);
1856 
1857 	return (ret);
1858 }
1859 
1860 /*
1861  * Given a vdev configuration, determine the maximum width needed for the device
1862  * name column.
1863  */
1864 static int
1865 max_width(zpool_handle_t *zhp, nvlist_t *nv, int depth, int max,
1866     int name_flags)
1867 {
1868 	char *name;
1869 	nvlist_t **child;
1870 	uint_t c, children;
1871 	int ret;
1872 
1873 	name = zpool_vdev_name(g_zfs, zhp, nv, name_flags);
1874 	if (strlen(name) + depth > max)
1875 		max = strlen(name) + depth;
1876 
1877 	free(name);
1878 
1879 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1880 	    &child, &children) == 0) {
1881 		for (c = 0; c < children; c++)
1882 			if ((ret = max_width(zhp, child[c], depth + 2,
1883 			    max, name_flags)) > max)
1884 				max = ret;
1885 	}
1886 
1887 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1888 	    &child, &children) == 0) {
1889 		for (c = 0; c < children; c++)
1890 			if ((ret = max_width(zhp, child[c], depth + 2,
1891 			    max, name_flags)) > max)
1892 				max = ret;
1893 	}
1894 
1895 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1896 	    &child, &children) == 0) {
1897 		for (c = 0; c < children; c++)
1898 			if ((ret = max_width(zhp, child[c], depth + 2,
1899 			    max, name_flags)) > max)
1900 				max = ret;
1901 	}
1902 
1903 	return (max);
1904 }
1905 
1906 typedef struct spare_cbdata {
1907 	uint64_t	cb_guid;
1908 	zpool_handle_t	*cb_zhp;
1909 } spare_cbdata_t;
1910 
1911 static boolean_t
1912 find_vdev(nvlist_t *nv, uint64_t search)
1913 {
1914 	uint64_t guid;
1915 	nvlist_t **child;
1916 	uint_t c, children;
1917 
1918 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
1919 	    search == guid)
1920 		return (B_TRUE);
1921 
1922 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1923 	    &child, &children) == 0) {
1924 		for (c = 0; c < children; c++)
1925 			if (find_vdev(child[c], search))
1926 				return (B_TRUE);
1927 	}
1928 
1929 	return (B_FALSE);
1930 }
1931 
1932 static int
1933 find_spare(zpool_handle_t *zhp, void *data)
1934 {
1935 	spare_cbdata_t *cbp = data;
1936 	nvlist_t *config, *nvroot;
1937 
1938 	config = zpool_get_config(zhp, NULL);
1939 	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
1940 	    &nvroot) == 0);
1941 
1942 	if (find_vdev(nvroot, cbp->cb_guid)) {
1943 		cbp->cb_zhp = zhp;
1944 		return (1);
1945 	}
1946 
1947 	zpool_close(zhp);
1948 	return (0);
1949 }
1950 
1951 typedef struct status_cbdata {
1952 	int		cb_count;
1953 	int		cb_name_flags;
1954 	int		cb_namewidth;
1955 	boolean_t	cb_allpools;
1956 	boolean_t	cb_verbose;
1957 	boolean_t	cb_literal;
1958 	boolean_t	cb_explain;
1959 	boolean_t	cb_first;
1960 	boolean_t	cb_dedup_stats;
1961 	boolean_t	cb_print_status;
1962 	boolean_t	cb_print_slow_ios;
1963 	boolean_t	cb_print_vdev_init;
1964 	boolean_t	cb_print_vdev_trim;
1965 	vdev_cmd_data_list_t	*vcdl;
1966 } status_cbdata_t;
1967 
1968 /* Return 1 if string is NULL, empty, or whitespace; return 0 otherwise. */
1969 static int
1970 is_blank_str(char *str)
1971 {
1972 	while (str != NULL && *str != '\0') {
1973 		if (!isblank(*str))
1974 			return (0);
1975 		str++;
1976 	}
1977 	return (1);
1978 }
1979 
1980 /* Print command output lines for specific vdev in a specific pool */
1981 static void
1982 zpool_print_cmd(vdev_cmd_data_list_t *vcdl, const char *pool, char *path)
1983 {
1984 	vdev_cmd_data_t *data;
1985 	int i, j;
1986 	char *val;
1987 
1988 	for (i = 0; i < vcdl->count; i++) {
1989 		if ((strcmp(vcdl->data[i].path, path) != 0) ||
1990 		    (strcmp(vcdl->data[i].pool, pool) != 0)) {
1991 			/* Not the vdev we're looking for */
1992 			continue;
1993 		}
1994 
1995 		data = &vcdl->data[i];
1996 		/* Print out all the output values for this vdev */
1997 		for (j = 0; j < vcdl->uniq_cols_cnt; j++) {
1998 			val = NULL;
1999 			/* Does this vdev have values for this column? */
2000 			for (int k = 0; k < data->cols_cnt; k++) {
2001 				if (strcmp(data->cols[k],
2002 				    vcdl->uniq_cols[j]) == 0) {
2003 					/* yes it does, record the value */
2004 					val = data->lines[k];
2005 					break;
2006 				}
2007 			}
2008 			/*
2009 			 * Mark empty values with dashes to make output
2010 			 * awk-able.
2011 			 */
2012 			if (is_blank_str(val))
2013 				val = "-";
2014 
2015 			printf("%*s", vcdl->uniq_cols_width[j], val);
2016 			if (j < vcdl->uniq_cols_cnt - 1)
2017 				printf("  ");
2018 		}
2019 
2020 		/* Print out any values that aren't in a column at the end */
2021 		for (j = data->cols_cnt; j < data->lines_cnt; j++) {
2022 			/* Did we have any columns?  If so print a spacer. */
2023 			if (vcdl->uniq_cols_cnt > 0)
2024 				printf("  ");
2025 
2026 			val = data->lines[j];
2027 			printf("%s", val ? val : "");
2028 		}
2029 		break;
2030 	}
2031 }
2032 
2033 /*
2034  * Print vdev initialization status for leaves
2035  */
2036 static void
2037 print_status_initialize(vdev_stat_t *vs, boolean_t verbose)
2038 {
2039 	if (verbose) {
2040 		if ((vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE ||
2041 		    vs->vs_initialize_state == VDEV_INITIALIZE_SUSPENDED ||
2042 		    vs->vs_initialize_state == VDEV_INITIALIZE_COMPLETE) &&
2043 		    !vs->vs_scan_removing) {
2044 			char zbuf[1024];
2045 			char tbuf[256];
2046 			struct tm zaction_ts;
2047 
2048 			time_t t = vs->vs_initialize_action_time;
2049 			int initialize_pct = 100;
2050 			if (vs->vs_initialize_state !=
2051 			    VDEV_INITIALIZE_COMPLETE) {
2052 				initialize_pct = (vs->vs_initialize_bytes_done *
2053 				    100 / (vs->vs_initialize_bytes_est + 1));
2054 			}
2055 
2056 			(void) localtime_r(&t, &zaction_ts);
2057 			(void) strftime(tbuf, sizeof (tbuf), "%c", &zaction_ts);
2058 
2059 			switch (vs->vs_initialize_state) {
2060 			case VDEV_INITIALIZE_SUSPENDED:
2061 				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2062 				    gettext("suspended, started at"), tbuf);
2063 				break;
2064 			case VDEV_INITIALIZE_ACTIVE:
2065 				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2066 				    gettext("started at"), tbuf);
2067 				break;
2068 			case VDEV_INITIALIZE_COMPLETE:
2069 				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2070 				    gettext("completed at"), tbuf);
2071 				break;
2072 			}
2073 
2074 			(void) printf(gettext("  (%d%% initialized%s)"),
2075 			    initialize_pct, zbuf);
2076 		} else {
2077 			(void) printf(gettext("  (uninitialized)"));
2078 		}
2079 	} else if (vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE) {
2080 		(void) printf(gettext("  (initializing)"));
2081 	}
2082 }
2083 
2084 /*
2085  * Print vdev TRIM status for leaves
2086  */
2087 static void
2088 print_status_trim(vdev_stat_t *vs, boolean_t verbose)
2089 {
2090 	if (verbose) {
2091 		if ((vs->vs_trim_state == VDEV_TRIM_ACTIVE ||
2092 		    vs->vs_trim_state == VDEV_TRIM_SUSPENDED ||
2093 		    vs->vs_trim_state == VDEV_TRIM_COMPLETE) &&
2094 		    !vs->vs_scan_removing) {
2095 			char zbuf[1024];
2096 			char tbuf[256];
2097 			struct tm zaction_ts;
2098 
2099 			time_t t = vs->vs_trim_action_time;
2100 			int trim_pct = 100;
2101 			if (vs->vs_trim_state != VDEV_TRIM_COMPLETE) {
2102 				trim_pct = (vs->vs_trim_bytes_done *
2103 				    100 / (vs->vs_trim_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_trim_state) {
2110 			case VDEV_TRIM_SUSPENDED:
2111 				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2112 				    gettext("suspended, started at"), tbuf);
2113 				break;
2114 			case VDEV_TRIM_ACTIVE:
2115 				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2116 				    gettext("started at"), tbuf);
2117 				break;
2118 			case VDEV_TRIM_COMPLETE:
2119 				(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
2120 				    gettext("completed at"), tbuf);
2121 				break;
2122 			}
2123 
2124 			(void) printf(gettext("  (%d%% trimmed%s)"),
2125 			    trim_pct, zbuf);
2126 		} else if (vs->vs_trim_notsup) {
2127 			(void) printf(gettext("  (trim unsupported)"));
2128 		} else {
2129 			(void) printf(gettext("  (untrimmed)"));
2130 		}
2131 	} else if (vs->vs_trim_state == VDEV_TRIM_ACTIVE) {
2132 		(void) printf(gettext("  (trimming)"));
2133 	}
2134 }
2135 
2136 /*
2137  * Return the color associated with a health string.  This includes returning
2138  * NULL for no color change.
2139  */
2140 static char *
2141 health_str_to_color(const char *health)
2142 {
2143 	if (strcmp(health, gettext("FAULTED")) == 0 ||
2144 	    strcmp(health, gettext("SUSPENDED")) == 0 ||
2145 	    strcmp(health, gettext("UNAVAIL")) == 0) {
2146 		return (ANSI_RED);
2147 	}
2148 
2149 	if (strcmp(health, gettext("OFFLINE")) == 0 ||
2150 	    strcmp(health, gettext("DEGRADED")) == 0 ||
2151 	    strcmp(health, gettext("REMOVED")) == 0) {
2152 		return (ANSI_YELLOW);
2153 	}
2154 
2155 	return (NULL);
2156 }
2157 
2158 /*
2159  * Print out configuration state as requested by status_callback.
2160  */
2161 static void
2162 print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
2163     nvlist_t *nv, int depth, boolean_t isspare, vdev_rebuild_stat_t *vrs)
2164 {
2165 	nvlist_t **child, *root;
2166 	uint_t c, i, vsc, children;
2167 	pool_scan_stat_t *ps = NULL;
2168 	vdev_stat_t *vs;
2169 	char rbuf[6], wbuf[6], cbuf[6];
2170 	char *vname;
2171 	uint64_t notpresent;
2172 	spare_cbdata_t spare_cb;
2173 	const char *state;
2174 	char *type;
2175 	char *path = NULL;
2176 	char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL;
2177 
2178 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2179 	    &child, &children) != 0)
2180 		children = 0;
2181 
2182 	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
2183 	    (uint64_t **)&vs, &vsc) == 0);
2184 
2185 	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
2186 
2187 	if (strcmp(type, VDEV_TYPE_INDIRECT) == 0)
2188 		return;
2189 
2190 	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
2191 
2192 	if (isspare) {
2193 		/*
2194 		 * For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
2195 		 * online drives.
2196 		 */
2197 		if (vs->vs_aux == VDEV_AUX_SPARED)
2198 			state = gettext("INUSE");
2199 		else if (vs->vs_state == VDEV_STATE_HEALTHY)
2200 			state = gettext("AVAIL");
2201 	}
2202 
2203 	printf_color(health_str_to_color(state),
2204 	    "\t%*s%-*s  %-8s", depth, "", cb->cb_namewidth - depth,
2205 	    name, state);
2206 
2207 	if (!isspare) {
2208 		if (vs->vs_read_errors)
2209 			rcolor = ANSI_RED;
2210 
2211 		if (vs->vs_write_errors)
2212 			wcolor = ANSI_RED;
2213 
2214 		if (vs->vs_checksum_errors)
2215 			ccolor = ANSI_RED;
2216 
2217 		if (cb->cb_literal) {
2218 			printf(" ");
2219 			printf_color(rcolor, "%5llu",
2220 			    (u_longlong_t)vs->vs_read_errors);
2221 			printf(" ");
2222 			printf_color(wcolor, "%5llu",
2223 			    (u_longlong_t)vs->vs_write_errors);
2224 			printf(" ");
2225 			printf_color(ccolor, "%5llu",
2226 			    (u_longlong_t)vs->vs_checksum_errors);
2227 		} else {
2228 			zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
2229 			zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
2230 			zfs_nicenum(vs->vs_checksum_errors, cbuf,
2231 			    sizeof (cbuf));
2232 			printf(" ");
2233 			printf_color(rcolor, "%5s", rbuf);
2234 			printf(" ");
2235 			printf_color(wcolor, "%5s", wbuf);
2236 			printf(" ");
2237 			printf_color(ccolor, "%5s", cbuf);
2238 		}
2239 		if (cb->cb_print_slow_ios) {
2240 			if (children == 0)  {
2241 				/* Only leafs vdevs have slow IOs */
2242 				zfs_nicenum(vs->vs_slow_ios, rbuf,
2243 				    sizeof (rbuf));
2244 			} else {
2245 				snprintf(rbuf, sizeof (rbuf), "-");
2246 			}
2247 
2248 			if (cb->cb_literal)
2249 				printf(" %5llu", (u_longlong_t)vs->vs_slow_ios);
2250 			else
2251 				printf(" %5s", rbuf);
2252 		}
2253 	}
2254 
2255 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
2256 	    &notpresent) == 0) {
2257 		verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
2258 		(void) printf("  %s %s", gettext("was"), path);
2259 	} else if (vs->vs_aux != 0) {
2260 		(void) printf("  ");
2261 		color_start(ANSI_RED);
2262 		switch (vs->vs_aux) {
2263 		case VDEV_AUX_OPEN_FAILED:
2264 			(void) printf(gettext("cannot open"));
2265 			break;
2266 
2267 		case VDEV_AUX_BAD_GUID_SUM:
2268 			(void) printf(gettext("missing device"));
2269 			break;
2270 
2271 		case VDEV_AUX_NO_REPLICAS:
2272 			(void) printf(gettext("insufficient replicas"));
2273 			break;
2274 
2275 		case VDEV_AUX_VERSION_NEWER:
2276 			(void) printf(gettext("newer version"));
2277 			break;
2278 
2279 		case VDEV_AUX_UNSUP_FEAT:
2280 			(void) printf(gettext("unsupported feature(s)"));
2281 			break;
2282 
2283 		case VDEV_AUX_ASHIFT_TOO_BIG:
2284 			(void) printf(gettext("unsupported minimum blocksize"));
2285 			break;
2286 
2287 		case VDEV_AUX_SPARED:
2288 			verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2289 			    &spare_cb.cb_guid) == 0);
2290 			if (zpool_iter(g_zfs, find_spare, &spare_cb) == 1) {
2291 				if (strcmp(zpool_get_name(spare_cb.cb_zhp),
2292 				    zpool_get_name(zhp)) == 0)
2293 					(void) printf(gettext("currently in "
2294 					    "use"));
2295 				else
2296 					(void) printf(gettext("in use by "
2297 					    "pool '%s'"),
2298 					    zpool_get_name(spare_cb.cb_zhp));
2299 				zpool_close(spare_cb.cb_zhp);
2300 			} else {
2301 				(void) printf(gettext("currently in use"));
2302 			}
2303 			break;
2304 
2305 		case VDEV_AUX_ERR_EXCEEDED:
2306 			(void) printf(gettext("too many errors"));
2307 			break;
2308 
2309 		case VDEV_AUX_IO_FAILURE:
2310 			(void) printf(gettext("experienced I/O failures"));
2311 			break;
2312 
2313 		case VDEV_AUX_BAD_LOG:
2314 			(void) printf(gettext("bad intent log"));
2315 			break;
2316 
2317 		case VDEV_AUX_EXTERNAL:
2318 			(void) printf(gettext("external device fault"));
2319 			break;
2320 
2321 		case VDEV_AUX_SPLIT_POOL:
2322 			(void) printf(gettext("split into new pool"));
2323 			break;
2324 
2325 		case VDEV_AUX_ACTIVE:
2326 			(void) printf(gettext("currently in use"));
2327 			break;
2328 
2329 		case VDEV_AUX_CHILDREN_OFFLINE:
2330 			(void) printf(gettext("all children offline"));
2331 			break;
2332 
2333 		default:
2334 			(void) printf(gettext("corrupted data"));
2335 			break;
2336 		}
2337 		color_end();
2338 	} else if (children == 0 && !isspare &&
2339 	    getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") == NULL &&
2340 	    VDEV_STAT_VALID(vs_physical_ashift, vsc) &&
2341 	    vs->vs_configured_ashift < vs->vs_physical_ashift) {
2342 		(void) printf(
2343 		    gettext("  block size: %dB configured, %dB native"),
2344 		    1 << vs->vs_configured_ashift, 1 << vs->vs_physical_ashift);
2345 	}
2346 
2347 	/* The root vdev has the scrub/resilver stats */
2348 	root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2349 	    ZPOOL_CONFIG_VDEV_TREE);
2350 	(void) nvlist_lookup_uint64_array(root, ZPOOL_CONFIG_SCAN_STATS,
2351 	    (uint64_t **)&ps, &c);
2352 
2353 	if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0) {
2354 		if (vs->vs_scan_processed != 0) {
2355 			(void) printf(gettext("  (%s)"),
2356 			    (ps->pss_func == POOL_SCAN_RESILVER) ?
2357 			    "resilvering" : "repairing");
2358 		} else if (vs->vs_resilver_deferred) {
2359 			(void) printf(gettext("  (awaiting resilver)"));
2360 		}
2361 	}
2362 
2363 	/* The top-level vdevs have the rebuild stats */
2364 	if (vrs != NULL && vrs->vrs_state == VDEV_REBUILD_ACTIVE &&
2365 	    children == 0) {
2366 		if (vs->vs_rebuild_processed != 0) {
2367 			(void) printf(gettext("  (resilvering)"));
2368 		}
2369 	}
2370 
2371 	if (cb->vcdl != NULL) {
2372 		if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
2373 			printf("  ");
2374 			zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
2375 		}
2376 	}
2377 
2378 	/* Display vdev initialization and trim status for leaves. */
2379 	if (children == 0) {
2380 		print_status_initialize(vs, cb->cb_print_vdev_init);
2381 		print_status_trim(vs, cb->cb_print_vdev_trim);
2382 	}
2383 
2384 	(void) printf("\n");
2385 
2386 	for (c = 0; c < children; c++) {
2387 		uint64_t islog = B_FALSE, ishole = B_FALSE;
2388 
2389 		/* Don't print logs or holes here */
2390 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2391 		    &islog);
2392 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2393 		    &ishole);
2394 		if (islog || ishole)
2395 			continue;
2396 		/* Only print normal classes here */
2397 		if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
2398 			continue;
2399 
2400 		/* Provide vdev_rebuild_stats to children if available */
2401 		if (vrs == NULL) {
2402 			(void) nvlist_lookup_uint64_array(nv,
2403 			    ZPOOL_CONFIG_REBUILD_STATS,
2404 			    (uint64_t **)&vrs, &i);
2405 		}
2406 
2407 		vname = zpool_vdev_name(g_zfs, zhp, child[c],
2408 		    cb->cb_name_flags | VDEV_NAME_TYPE_ID);
2409 		print_status_config(zhp, cb, vname, child[c], depth + 2,
2410 		    isspare, vrs);
2411 		free(vname);
2412 	}
2413 }
2414 
2415 /*
2416  * Print the configuration of an exported pool.  Iterate over all vdevs in the
2417  * pool, printing out the name and status for each one.
2418  */
2419 static void
2420 print_import_config(status_cbdata_t *cb, const char *name, nvlist_t *nv,
2421     int depth)
2422 {
2423 	nvlist_t **child;
2424 	uint_t c, children;
2425 	vdev_stat_t *vs;
2426 	char *type, *vname;
2427 
2428 	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
2429 	if (strcmp(type, VDEV_TYPE_MISSING) == 0 ||
2430 	    strcmp(type, VDEV_TYPE_HOLE) == 0)
2431 		return;
2432 
2433 	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
2434 	    (uint64_t **)&vs, &c) == 0);
2435 
2436 	(void) printf("\t%*s%-*s", depth, "", cb->cb_namewidth - depth, name);
2437 	(void) printf("  %s", zpool_state_to_name(vs->vs_state, vs->vs_aux));
2438 
2439 	if (vs->vs_aux != 0) {
2440 		(void) printf("  ");
2441 
2442 		switch (vs->vs_aux) {
2443 		case VDEV_AUX_OPEN_FAILED:
2444 			(void) printf(gettext("cannot open"));
2445 			break;
2446 
2447 		case VDEV_AUX_BAD_GUID_SUM:
2448 			(void) printf(gettext("missing device"));
2449 			break;
2450 
2451 		case VDEV_AUX_NO_REPLICAS:
2452 			(void) printf(gettext("insufficient replicas"));
2453 			break;
2454 
2455 		case VDEV_AUX_VERSION_NEWER:
2456 			(void) printf(gettext("newer version"));
2457 			break;
2458 
2459 		case VDEV_AUX_UNSUP_FEAT:
2460 			(void) printf(gettext("unsupported feature(s)"));
2461 			break;
2462 
2463 		case VDEV_AUX_ERR_EXCEEDED:
2464 			(void) printf(gettext("too many errors"));
2465 			break;
2466 
2467 		case VDEV_AUX_ACTIVE:
2468 			(void) printf(gettext("currently in use"));
2469 			break;
2470 
2471 		case VDEV_AUX_CHILDREN_OFFLINE:
2472 			(void) printf(gettext("all children offline"));
2473 			break;
2474 
2475 		default:
2476 			(void) printf(gettext("corrupted data"));
2477 			break;
2478 		}
2479 	}
2480 	(void) printf("\n");
2481 
2482 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2483 	    &child, &children) != 0)
2484 		return;
2485 
2486 	for (c = 0; c < children; c++) {
2487 		uint64_t is_log = B_FALSE;
2488 
2489 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2490 		    &is_log);
2491 		if (is_log)
2492 			continue;
2493 		if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
2494 			continue;
2495 
2496 		vname = zpool_vdev_name(g_zfs, NULL, child[c],
2497 		    cb->cb_name_flags | VDEV_NAME_TYPE_ID);
2498 		print_import_config(cb, vname, child[c], depth + 2);
2499 		free(vname);
2500 	}
2501 
2502 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2503 	    &child, &children) == 0) {
2504 		(void) printf(gettext("\tcache\n"));
2505 		for (c = 0; c < children; c++) {
2506 			vname = zpool_vdev_name(g_zfs, NULL, child[c],
2507 			    cb->cb_name_flags);
2508 			(void) printf("\t  %s\n", vname);
2509 			free(vname);
2510 		}
2511 	}
2512 
2513 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2514 	    &child, &children) == 0) {
2515 		(void) printf(gettext("\tspares\n"));
2516 		for (c = 0; c < children; c++) {
2517 			vname = zpool_vdev_name(g_zfs, NULL, child[c],
2518 			    cb->cb_name_flags);
2519 			(void) printf("\t  %s\n", vname);
2520 			free(vname);
2521 		}
2522 	}
2523 }
2524 
2525 /*
2526  * Print specialized class vdevs.
2527  *
2528  * These are recorded as top level vdevs in the main pool child array
2529  * but with "is_log" set to 1 or an "alloc_bias" string. We use either
2530  * print_status_config() or print_import_config() to print the top level
2531  * class vdevs then any of their children (eg mirrored slogs) are printed
2532  * recursively - which works because only the top level vdev is marked.
2533  */
2534 static void
2535 print_class_vdevs(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t *nv,
2536     const char *class)
2537 {
2538 	uint_t c, children;
2539 	nvlist_t **child;
2540 	boolean_t printed = B_FALSE;
2541 
2542 	assert(zhp != NULL || !cb->cb_verbose);
2543 
2544 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child,
2545 	    &children) != 0)
2546 		return;
2547 
2548 	for (c = 0; c < children; c++) {
2549 		uint64_t is_log = B_FALSE;
2550 		char *bias = NULL;
2551 		char *type = NULL;
2552 
2553 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2554 		    &is_log);
2555 
2556 		if (is_log) {
2557 			bias = VDEV_ALLOC_CLASS_LOGS;
2558 		} else {
2559 			(void) nvlist_lookup_string(child[c],
2560 			    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
2561 			(void) nvlist_lookup_string(child[c],
2562 			    ZPOOL_CONFIG_TYPE, &type);
2563 		}
2564 
2565 		if (bias == NULL || strcmp(bias, class) != 0)
2566 			continue;
2567 		if (!is_log && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
2568 			continue;
2569 
2570 		if (!printed) {
2571 			(void) printf("\t%s\t\n", gettext(class));
2572 			printed = B_TRUE;
2573 		}
2574 
2575 		char *name = zpool_vdev_name(g_zfs, zhp, child[c],
2576 		    cb->cb_name_flags | VDEV_NAME_TYPE_ID);
2577 		if (cb->cb_print_status)
2578 			print_status_config(zhp, cb, name, child[c], 2,
2579 			    B_FALSE, NULL);
2580 		else
2581 			print_import_config(cb, name, child[c], 2);
2582 		free(name);
2583 	}
2584 }
2585 
2586 /*
2587  * Display the status for the given pool.
2588  */
2589 static void
2590 show_import(nvlist_t *config)
2591 {
2592 	uint64_t pool_state;
2593 	vdev_stat_t *vs;
2594 	char *name;
2595 	uint64_t guid;
2596 	uint64_t hostid = 0;
2597 	char *msgid;
2598 	char *hostname = "unknown";
2599 	nvlist_t *nvroot, *nvinfo;
2600 	zpool_status_t reason;
2601 	zpool_errata_t errata;
2602 	const char *health;
2603 	uint_t vsc;
2604 	char *comment;
2605 	status_cbdata_t cb = { 0 };
2606 
2607 	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
2608 	    &name) == 0);
2609 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
2610 	    &guid) == 0);
2611 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
2612 	    &pool_state) == 0);
2613 	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2614 	    &nvroot) == 0);
2615 
2616 	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
2617 	    (uint64_t **)&vs, &vsc) == 0);
2618 	health = zpool_state_to_name(vs->vs_state, vs->vs_aux);
2619 
2620 	reason = zpool_import_status(config, &msgid, &errata);
2621 
2622 	(void) printf(gettext("   pool: %s\n"), name);
2623 	(void) printf(gettext("     id: %llu\n"), (u_longlong_t)guid);
2624 	(void) printf(gettext("  state: %s"), health);
2625 	if (pool_state == POOL_STATE_DESTROYED)
2626 		(void) printf(gettext(" (DESTROYED)"));
2627 	(void) printf("\n");
2628 
2629 	switch (reason) {
2630 	case ZPOOL_STATUS_MISSING_DEV_R:
2631 	case ZPOOL_STATUS_MISSING_DEV_NR:
2632 	case ZPOOL_STATUS_BAD_GUID_SUM:
2633 		printf_color(ANSI_BOLD, gettext("status: "));
2634 		printf_color(ANSI_YELLOW, gettext("One or more devices are "
2635 		    "missing from the system.\n"));
2636 		break;
2637 
2638 	case ZPOOL_STATUS_CORRUPT_LABEL_R:
2639 	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
2640 		printf_color(ANSI_BOLD, gettext("status: "));
2641 		printf_color(ANSI_YELLOW, gettext("One or more devices contains"
2642 		    " corrupted data.\n"));
2643 		break;
2644 
2645 	case ZPOOL_STATUS_CORRUPT_DATA:
2646 		(void) printf(
2647 		    gettext(" status: The pool data is corrupted.\n"));
2648 		break;
2649 
2650 	case ZPOOL_STATUS_OFFLINE_DEV:
2651 		printf_color(ANSI_BOLD, gettext("status: "));
2652 		printf_color(ANSI_YELLOW, gettext("One or more devices "
2653 		    "are offlined.\n"));
2654 		break;
2655 
2656 	case ZPOOL_STATUS_CORRUPT_POOL:
2657 		printf_color(ANSI_BOLD, gettext("status: "));
2658 		printf_color(ANSI_YELLOW, gettext("The pool metadata is "
2659 		    "corrupted.\n"));
2660 		break;
2661 
2662 	case ZPOOL_STATUS_VERSION_OLDER:
2663 		printf_color(ANSI_BOLD, gettext("status: "));
2664 		printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
2665 		    "a legacy on-disk version.\n"));
2666 		break;
2667 
2668 	case ZPOOL_STATUS_VERSION_NEWER:
2669 		printf_color(ANSI_BOLD, gettext("status: "));
2670 		printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
2671 		    "an incompatible version.\n"));
2672 		break;
2673 
2674 	case ZPOOL_STATUS_FEAT_DISABLED:
2675 		printf_color(ANSI_BOLD, gettext("status: "));
2676 		printf_color(ANSI_YELLOW, gettext("Some supported features are "
2677 		    "not enabled on the pool.\n"));
2678 		break;
2679 
2680 	case ZPOOL_STATUS_UNSUP_FEAT_READ:
2681 		printf_color(ANSI_BOLD, gettext("status: "));
2682 		printf_color(ANSI_YELLOW, gettext("The pool uses the following "
2683 		    "feature(s) not supported on this system:\n"));
2684 		color_start(ANSI_YELLOW);
2685 		zpool_print_unsup_feat(config);
2686 		color_end();
2687 		break;
2688 
2689 	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
2690 		printf_color(ANSI_BOLD, gettext("status: "));
2691 		printf_color(ANSI_YELLOW, gettext("The pool can only be "
2692 		    "accessed in read-only mode on this system. It\n\tcannot be"
2693 		    " accessed in read-write mode because it uses the "
2694 		    "following\n\tfeature(s) not supported on this system:\n"));
2695 		color_start(ANSI_YELLOW);
2696 		zpool_print_unsup_feat(config);
2697 		color_end();
2698 		break;
2699 
2700 	case ZPOOL_STATUS_HOSTID_ACTIVE:
2701 		printf_color(ANSI_BOLD, gettext("status: "));
2702 		printf_color(ANSI_YELLOW, gettext("The pool is currently "
2703 		    "imported by another system.\n"));
2704 		break;
2705 
2706 	case ZPOOL_STATUS_HOSTID_REQUIRED:
2707 		printf_color(ANSI_BOLD, gettext("status: "));
2708 		printf_color(ANSI_YELLOW, gettext("The pool has the "
2709 		    "multihost property on.  It cannot\n\tbe safely imported "
2710 		    "when the system hostid is not set.\n"));
2711 		break;
2712 
2713 	case ZPOOL_STATUS_HOSTID_MISMATCH:
2714 		printf_color(ANSI_BOLD, gettext("status: "));
2715 		printf_color(ANSI_YELLOW, gettext("The pool was last accessed "
2716 		    "by another system.\n"));
2717 		break;
2718 
2719 	case ZPOOL_STATUS_FAULTED_DEV_R:
2720 	case ZPOOL_STATUS_FAULTED_DEV_NR:
2721 		printf_color(ANSI_BOLD, gettext("status: "));
2722 		printf_color(ANSI_YELLOW, gettext("One or more devices are "
2723 		    "faulted.\n"));
2724 		break;
2725 
2726 	case ZPOOL_STATUS_BAD_LOG:
2727 		printf_color(ANSI_BOLD, gettext("status: "));
2728 		printf_color(ANSI_YELLOW, gettext("An intent log record cannot "
2729 		    "be read.\n"));
2730 		break;
2731 
2732 	case ZPOOL_STATUS_RESILVERING:
2733 	case ZPOOL_STATUS_REBUILDING:
2734 		printf_color(ANSI_BOLD, gettext("status: "));
2735 		printf_color(ANSI_YELLOW, gettext("One or more devices were "
2736 		    "being resilvered.\n"));
2737 		break;
2738 
2739 	case ZPOOL_STATUS_ERRATA:
2740 		printf_color(ANSI_BOLD, gettext("status: "));
2741 		printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
2742 		    errata);
2743 		break;
2744 
2745 	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
2746 		printf_color(ANSI_BOLD, gettext("status: "));
2747 		printf_color(ANSI_YELLOW, gettext("One or more devices are "
2748 		    "configured to use a non-native block size.\n"
2749 		    "\tExpect reduced performance.\n"));
2750 		break;
2751 
2752 	default:
2753 		/*
2754 		 * No other status can be seen when importing pools.
2755 		 */
2756 		assert(reason == ZPOOL_STATUS_OK);
2757 	}
2758 
2759 	/*
2760 	 * Print out an action according to the overall state of the pool.
2761 	 */
2762 	if (vs->vs_state == VDEV_STATE_HEALTHY) {
2763 		if (reason == ZPOOL_STATUS_VERSION_OLDER ||
2764 		    reason == ZPOOL_STATUS_FEAT_DISABLED) {
2765 			(void) printf(gettext(" action: The pool can be "
2766 			    "imported using its name or numeric identifier, "
2767 			    "though\n\tsome features will not be available "
2768 			    "without an explicit 'zpool upgrade'.\n"));
2769 		} else if (reason == ZPOOL_STATUS_HOSTID_MISMATCH) {
2770 			(void) printf(gettext(" action: The pool can be "
2771 			    "imported using its name or numeric "
2772 			    "identifier and\n\tthe '-f' flag.\n"));
2773 		} else if (reason == ZPOOL_STATUS_ERRATA) {
2774 			switch (errata) {
2775 			case ZPOOL_ERRATA_NONE:
2776 				break;
2777 
2778 			case ZPOOL_ERRATA_ZOL_2094_SCRUB:
2779 				(void) printf(gettext(" action: The pool can "
2780 				    "be imported using its name or numeric "
2781 				    "identifier,\n\thowever there is a compat"
2782 				    "ibility issue which should be corrected"
2783 				    "\n\tby running 'zpool scrub'\n"));
2784 				break;
2785 
2786 			case ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY:
2787 				(void) printf(gettext(" action: The pool can"
2788 				    "not be imported with this version of ZFS "
2789 				    "due to\n\tan active asynchronous destroy. "
2790 				    "Revert to an earlier version\n\tand "
2791 				    "allow the destroy to complete before "
2792 				    "updating.\n"));
2793 				break;
2794 
2795 			case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
2796 				(void) printf(gettext(" action: Existing "
2797 				    "encrypted datasets contain an on-disk "
2798 				    "incompatibility, which\n\tneeds to be "
2799 				    "corrected. Backup these datasets to new "
2800 				    "encrypted datasets\n\tand destroy the "
2801 				    "old ones.\n"));
2802 				break;
2803 
2804 			case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
2805 				(void) printf(gettext(" action: Existing "
2806 				    "encrypted snapshots and bookmarks contain "
2807 				    "an on-disk\n\tincompatibility. This may "
2808 				    "cause on-disk corruption if they are used"
2809 				    "\n\twith 'zfs recv'. To correct the "
2810 				    "issue, enable the bookmark_v2 feature.\n\t"
2811 				    "No additional action is needed if there "
2812 				    "are no encrypted snapshots or\n\t"
2813 				    "bookmarks. If preserving the encrypted "
2814 				    "snapshots and bookmarks is\n\trequired, "
2815 				    "use a non-raw send to backup and restore "
2816 				    "them. Alternately,\n\tthey may be removed"
2817 				    " to resolve the incompatibility.\n"));
2818 				break;
2819 			default:
2820 				/*
2821 				 * All errata must contain an action message.
2822 				 */
2823 				assert(0);
2824 			}
2825 		} else {
2826 			(void) printf(gettext(" action: The pool can be "
2827 			    "imported using its name or numeric "
2828 			    "identifier.\n"));
2829 		}
2830 	} else if (vs->vs_state == VDEV_STATE_DEGRADED) {
2831 		(void) printf(gettext(" action: The pool can be imported "
2832 		    "despite missing or damaged devices.  The\n\tfault "
2833 		    "tolerance of the pool may be compromised if imported.\n"));
2834 	} else {
2835 		switch (reason) {
2836 		case ZPOOL_STATUS_VERSION_NEWER:
2837 			(void) printf(gettext(" action: The pool cannot be "
2838 			    "imported.  Access the pool on a system running "
2839 			    "newer\n\tsoftware, or recreate the pool from "
2840 			    "backup.\n"));
2841 			break;
2842 		case ZPOOL_STATUS_UNSUP_FEAT_READ:
2843 			printf_color(ANSI_BOLD, gettext("action: "));
2844 			printf_color(ANSI_YELLOW, gettext("The pool cannot be "
2845 			    "imported. Access the pool on a system that "
2846 			    "supports\n\tthe required feature(s), or recreate "
2847 			    "the pool from backup.\n"));
2848 			break;
2849 		case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
2850 			printf_color(ANSI_BOLD, gettext("action: "));
2851 			printf_color(ANSI_YELLOW, gettext("The pool cannot be "
2852 			    "imported in read-write mode. Import the pool "
2853 			    "with\n"
2854 			    "\t\"-o readonly=on\", access the pool on a system "
2855 			    "that supports the\n\trequired feature(s), or "
2856 			    "recreate the pool from backup.\n"));
2857 			break;
2858 		case ZPOOL_STATUS_MISSING_DEV_R:
2859 		case ZPOOL_STATUS_MISSING_DEV_NR:
2860 		case ZPOOL_STATUS_BAD_GUID_SUM:
2861 			(void) printf(gettext(" action: The pool cannot be "
2862 			    "imported. Attach the missing\n\tdevices and try "
2863 			    "again.\n"));
2864 			break;
2865 		case ZPOOL_STATUS_HOSTID_ACTIVE:
2866 			VERIFY0(nvlist_lookup_nvlist(config,
2867 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo));
2868 
2869 			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
2870 				hostname = fnvlist_lookup_string(nvinfo,
2871 				    ZPOOL_CONFIG_MMP_HOSTNAME);
2872 
2873 			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
2874 				hostid = fnvlist_lookup_uint64(nvinfo,
2875 				    ZPOOL_CONFIG_MMP_HOSTID);
2876 
2877 			(void) printf(gettext(" action: The pool must be "
2878 			    "exported from %s (hostid=%lx)\n\tbefore it "
2879 			    "can be safely imported.\n"), hostname,
2880 			    (unsigned long) hostid);
2881 			break;
2882 		case ZPOOL_STATUS_HOSTID_REQUIRED:
2883 			(void) printf(gettext(" action: Set a unique system "
2884 			    "hostid with the zgenhostid(8) command.\n"));
2885 			break;
2886 		default:
2887 			(void) printf(gettext(" action: The pool cannot be "
2888 			    "imported due to damaged devices or data.\n"));
2889 		}
2890 	}
2891 
2892 	/* Print the comment attached to the pool. */
2893 	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
2894 		(void) printf(gettext("comment: %s\n"), comment);
2895 
2896 	/*
2897 	 * If the state is "closed" or "can't open", and the aux state
2898 	 * is "corrupt data":
2899 	 */
2900 	if (((vs->vs_state == VDEV_STATE_CLOSED) ||
2901 	    (vs->vs_state == VDEV_STATE_CANT_OPEN)) &&
2902 	    (vs->vs_aux == VDEV_AUX_CORRUPT_DATA)) {
2903 		if (pool_state == POOL_STATE_DESTROYED)
2904 			(void) printf(gettext("\tThe pool was destroyed, "
2905 			    "but can be imported using the '-Df' flags.\n"));
2906 		else if (pool_state != POOL_STATE_EXPORTED)
2907 			(void) printf(gettext("\tThe pool may be active on "
2908 			    "another system, but can be imported using\n\t"
2909 			    "the '-f' flag.\n"));
2910 	}
2911 
2912 	if (msgid != NULL) {
2913 		(void) printf(gettext(
2914 		    "   see: https://openzfs.github.io/openzfs-docs/msg/%s\n"),
2915 		    msgid);
2916 	}
2917 
2918 	(void) printf(gettext(" config:\n\n"));
2919 
2920 	cb.cb_namewidth = max_width(NULL, nvroot, 0, strlen(name),
2921 	    VDEV_NAME_TYPE_ID);
2922 	if (cb.cb_namewidth < 10)
2923 		cb.cb_namewidth = 10;
2924 
2925 	print_import_config(&cb, name, nvroot, 0);
2926 
2927 	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_DEDUP);
2928 	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
2929 	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_CLASS_LOGS);
2930 
2931 	if (reason == ZPOOL_STATUS_BAD_GUID_SUM) {
2932 		(void) printf(gettext("\n\tAdditional devices are known to "
2933 		    "be part of this pool, though their\n\texact "
2934 		    "configuration cannot be determined.\n"));
2935 	}
2936 }
2937 
2938 static boolean_t
2939 zfs_force_import_required(nvlist_t *config)
2940 {
2941 	uint64_t state;
2942 	uint64_t hostid = 0;
2943 	nvlist_t *nvinfo;
2944 
2945 	state = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE);
2946 	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
2947 
2948 	if (state != POOL_STATE_EXPORTED && hostid != get_system_hostid())
2949 		return (B_TRUE);
2950 
2951 	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
2952 	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE)) {
2953 		mmp_state_t mmp_state = fnvlist_lookup_uint64(nvinfo,
2954 		    ZPOOL_CONFIG_MMP_STATE);
2955 
2956 		if (mmp_state != MMP_STATE_INACTIVE)
2957 			return (B_TRUE);
2958 	}
2959 
2960 	return (B_FALSE);
2961 }
2962 
2963 /*
2964  * Perform the import for the given configuration.  This passes the heavy
2965  * lifting off to zpool_import_props(), and then mounts the datasets contained
2966  * within the pool.
2967  */
2968 static int
2969 do_import(nvlist_t *config, const char *newname, const char *mntopts,
2970     nvlist_t *props, int flags)
2971 {
2972 	int ret = 0;
2973 	zpool_handle_t *zhp;
2974 	char *name;
2975 	uint64_t version;
2976 
2977 	name = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);
2978 	version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);
2979 
2980 	if (!SPA_VERSION_IS_SUPPORTED(version)) {
2981 		(void) fprintf(stderr, gettext("cannot import '%s': pool "
2982 		    "is formatted using an unsupported ZFS version\n"), name);
2983 		return (1);
2984 	} else if (zfs_force_import_required(config) &&
2985 	    !(flags & ZFS_IMPORT_ANY_HOST)) {
2986 		mmp_state_t mmp_state = MMP_STATE_INACTIVE;
2987 		nvlist_t *nvinfo;
2988 
2989 		nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
2990 		if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE))
2991 			mmp_state = fnvlist_lookup_uint64(nvinfo,
2992 			    ZPOOL_CONFIG_MMP_STATE);
2993 
2994 		if (mmp_state == MMP_STATE_ACTIVE) {
2995 			char *hostname = "<unknown>";
2996 			uint64_t hostid = 0;
2997 
2998 			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
2999 				hostname = fnvlist_lookup_string(nvinfo,
3000 				    ZPOOL_CONFIG_MMP_HOSTNAME);
3001 
3002 			if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
3003 				hostid = fnvlist_lookup_uint64(nvinfo,
3004 				    ZPOOL_CONFIG_MMP_HOSTID);
3005 
3006 			(void) fprintf(stderr, gettext("cannot import '%s': "
3007 			    "pool is imported on %s (hostid: "
3008 			    "0x%lx)\nExport the pool on the other system, "
3009 			    "then run 'zpool import'.\n"),
3010 			    name, hostname, (unsigned long) hostid);
3011 		} else if (mmp_state == MMP_STATE_NO_HOSTID) {
3012 			(void) fprintf(stderr, gettext("Cannot import '%s': "
3013 			    "pool has the multihost property on and the\n"
3014 			    "system's hostid is not set. Set a unique hostid "
3015 			    "with the zgenhostid(8) command.\n"), name);
3016 		} else {
3017 			char *hostname = "<unknown>";
3018 			uint64_t timestamp = 0;
3019 			uint64_t hostid = 0;
3020 
3021 			if (nvlist_exists(config, ZPOOL_CONFIG_HOSTNAME))
3022 				hostname = fnvlist_lookup_string(config,
3023 				    ZPOOL_CONFIG_HOSTNAME);
3024 
3025 			if (nvlist_exists(config, ZPOOL_CONFIG_TIMESTAMP))
3026 				timestamp = fnvlist_lookup_uint64(config,
3027 				    ZPOOL_CONFIG_TIMESTAMP);
3028 
3029 			if (nvlist_exists(config, ZPOOL_CONFIG_HOSTID))
3030 				hostid = fnvlist_lookup_uint64(config,
3031 				    ZPOOL_CONFIG_HOSTID);
3032 
3033 			(void) fprintf(stderr, gettext("cannot import '%s': "
3034 			    "pool was previously in use from another system.\n"
3035 			    "Last accessed by %s (hostid=%lx) at %s"
3036 			    "The pool can be imported, use 'zpool import -f' "
3037 			    "to import the pool.\n"), name, hostname,
3038 			    (unsigned long)hostid, ctime((time_t *)&timestamp));
3039 		}
3040 
3041 		return (1);
3042 	}
3043 
3044 	if (zpool_import_props(g_zfs, config, newname, props, flags) != 0)
3045 		return (1);
3046 
3047 	if (newname != NULL)
3048 		name = (char *)newname;
3049 
3050 	if ((zhp = zpool_open_canfail(g_zfs, name)) == NULL)
3051 		return (1);
3052 
3053 	/*
3054 	 * Loading keys is best effort. We don't want to return immediately
3055 	 * if it fails but we do want to give the error to the caller.
3056 	 */
3057 	if (flags & ZFS_IMPORT_LOAD_KEYS) {
3058 		ret = zfs_crypto_attempt_load_keys(g_zfs, name);
3059 		if (ret != 0)
3060 			ret = 1;
3061 	}
3062 
3063 	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
3064 	    !(flags & ZFS_IMPORT_ONLY) &&
3065 	    zpool_enable_datasets(zhp, mntopts, 0) != 0) {
3066 		zpool_close(zhp);
3067 		return (1);
3068 	}
3069 
3070 	zpool_close(zhp);
3071 	return (ret);
3072 }
3073 
3074 typedef struct target_exists_args {
3075 	const char	*poolname;
3076 	uint64_t	poolguid;
3077 } target_exists_args_t;
3078 
3079 static int
3080 name_or_guid_exists(zpool_handle_t *zhp, void *data)
3081 {
3082 	target_exists_args_t *args = data;
3083 	nvlist_t *config = zpool_get_config(zhp, NULL);
3084 	int found = 0;
3085 
3086 	if (config == NULL)
3087 		return (0);
3088 
3089 	if (args->poolname != NULL) {
3090 		char *pool_name;
3091 
3092 		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
3093 		    &pool_name) == 0);
3094 		if (strcmp(pool_name, args->poolname) == 0)
3095 			found = 1;
3096 	} else {
3097 		uint64_t pool_guid;
3098 
3099 		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
3100 		    &pool_guid) == 0);
3101 		if (pool_guid == args->poolguid)
3102 			found = 1;
3103 	}
3104 	zpool_close(zhp);
3105 
3106 	return (found);
3107 }
3108 /*
3109  * zpool checkpoint <pool>
3110  *       checkpoint --discard <pool>
3111  *
3112  *       -d         Discard the checkpoint from a checkpointed
3113  *       --discard  pool.
3114  *
3115  *       -w         Wait for discarding a checkpoint to complete.
3116  *       --wait
3117  *
3118  * Checkpoints the specified pool, by taking a "snapshot" of its
3119  * current state. A pool can only have one checkpoint at a time.
3120  */
3121 int
3122 zpool_do_checkpoint(int argc, char **argv)
3123 {
3124 	boolean_t discard, wait;
3125 	char *pool;
3126 	zpool_handle_t *zhp;
3127 	int c, err;
3128 
3129 	struct option long_options[] = {
3130 		{"discard", no_argument, NULL, 'd'},
3131 		{"wait", no_argument, NULL, 'w'},
3132 		{0, 0, 0, 0}
3133 	};
3134 
3135 	discard = B_FALSE;
3136 	wait = B_FALSE;
3137 	while ((c = getopt_long(argc, argv, ":dw", long_options, NULL)) != -1) {
3138 		switch (c) {
3139 		case 'd':
3140 			discard = B_TRUE;
3141 			break;
3142 		case 'w':
3143 			wait = B_TRUE;
3144 			break;
3145 		case '?':
3146 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
3147 			    optopt);
3148 			usage(B_FALSE);
3149 		}
3150 	}
3151 
3152 	if (wait && !discard) {
3153 		(void) fprintf(stderr, gettext("--wait only valid when "
3154 		    "--discard also specified\n"));
3155 		usage(B_FALSE);
3156 	}
3157 
3158 	argc -= optind;
3159 	argv += optind;
3160 
3161 	if (argc < 1) {
3162 		(void) fprintf(stderr, gettext("missing pool argument\n"));
3163 		usage(B_FALSE);
3164 	}
3165 
3166 	if (argc > 1) {
3167 		(void) fprintf(stderr, gettext("too many arguments\n"));
3168 		usage(B_FALSE);
3169 	}
3170 
3171 	pool = argv[0];
3172 
3173 	if ((zhp = zpool_open(g_zfs, pool)) == NULL) {
3174 		/* As a special case, check for use of '/' in the name */
3175 		if (strchr(pool, '/') != NULL)
3176 			(void) fprintf(stderr, gettext("'zpool checkpoint' "
3177 			    "doesn't work on datasets. To save the state "
3178 			    "of a dataset from a specific point in time "
3179 			    "please use 'zfs snapshot'\n"));
3180 		return (1);
3181 	}
3182 
3183 	if (discard) {
3184 		err = (zpool_discard_checkpoint(zhp) != 0);
3185 		if (err == 0 && wait)
3186 			err = zpool_wait(zhp, ZPOOL_WAIT_CKPT_DISCARD);
3187 	} else {
3188 		err = (zpool_checkpoint(zhp) != 0);
3189 	}
3190 
3191 	zpool_close(zhp);
3192 
3193 	return (err);
3194 }
3195 
3196 #define	CHECKPOINT_OPT	1024
3197 
3198 /*
3199  * zpool import [-d dir] [-D]
3200  *       import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
3201  *              [-d dir | -c cachefile] [-f] -a
3202  *       import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
3203  *              [-d dir | -c cachefile] [-f] [-n] [-F] <pool | id> [newpool]
3204  *
3205  *	 -c	Read pool information from a cachefile instead of searching
3206  *		devices.
3207  *
3208  *       -d	Scan in a specific directory, other than /dev/.  More than
3209  *		one directory can be specified using multiple '-d' options.
3210  *
3211  *       -D     Scan for previously destroyed pools or import all or only
3212  *              specified destroyed pools.
3213  *
3214  *       -R	Temporarily import the pool, with all mountpoints relative to
3215  *		the given root.  The pool will remain exported when the machine
3216  *		is rebooted.
3217  *
3218  *       -V	Import even in the presence of faulted vdevs.  This is an
3219  *       	intentionally undocumented option for testing purposes, and
3220  *       	treats the pool configuration as complete, leaving any bad
3221  *		vdevs in the FAULTED state. In other words, it does verbatim
3222  *		import.
3223  *
3224  *       -f	Force import, even if it appears that the pool is active.
3225  *
3226  *       -F     Attempt rewind if necessary.
3227  *
3228  *       -n     See if rewind would work, but don't actually rewind.
3229  *
3230  *       -N     Import the pool but don't mount datasets.
3231  *
3232  *       -T     Specify a starting txg to use for import. This option is
3233  *       	intentionally undocumented option for testing purposes.
3234  *
3235  *       -a	Import all pools found.
3236  *
3237  *       -l	Load encryption keys while importing.
3238  *
3239  *       -o	Set property=value and/or temporary mount options (without '=').
3240  *
3241  *	 -s	Scan using the default search path, the libblkid cache will
3242  *	        not be consulted.
3243  *
3244  *       --rewind-to-checkpoint
3245  *       	Import the pool and revert back to the checkpoint.
3246  *
3247  * The import command scans for pools to import, and import pools based on pool
3248  * name and GUID.  The pool can also be renamed as part of the import process.
3249  */
3250 int
3251 zpool_do_import(int argc, char **argv)
3252 {
3253 	char **searchdirs = NULL;
3254 	char *env, *envdup = NULL;
3255 	int nsearch = 0;
3256 	int c;
3257 	int err = 0;
3258 	nvlist_t *pools = NULL;
3259 	boolean_t do_all = B_FALSE;
3260 	boolean_t do_destroyed = B_FALSE;
3261 	char *mntopts = NULL;
3262 	nvpair_t *elem;
3263 	nvlist_t *config;
3264 	uint64_t searchguid = 0;
3265 	char *searchname = NULL;
3266 	char *propval;
3267 	nvlist_t *found_config;
3268 	nvlist_t *policy = NULL;
3269 	nvlist_t *props = NULL;
3270 	boolean_t first;
3271 	int flags = ZFS_IMPORT_NORMAL;
3272 	uint32_t rewind_policy = ZPOOL_NO_REWIND;
3273 	boolean_t dryrun = B_FALSE;
3274 	boolean_t do_rewind = B_FALSE;
3275 	boolean_t xtreme_rewind = B_FALSE;
3276 	boolean_t do_scan = B_FALSE;
3277 	boolean_t pool_exists = B_FALSE;
3278 	uint64_t pool_state, txg = -1ULL;
3279 	char *cachefile = NULL;
3280 	importargs_t idata = { 0 };
3281 	char *endptr;
3282 
3283 	struct option long_options[] = {
3284 		{"rewind-to-checkpoint", no_argument, NULL, CHECKPOINT_OPT},
3285 		{0, 0, 0, 0}
3286 	};
3287 
3288 	/* check options */
3289 	while ((c = getopt_long(argc, argv, ":aCc:d:DEfFlmnNo:R:stT:VX",
3290 	    long_options, NULL)) != -1) {
3291 		switch (c) {
3292 		case 'a':
3293 			do_all = B_TRUE;
3294 			break;
3295 		case 'c':
3296 			cachefile = optarg;
3297 			break;
3298 		case 'd':
3299 			if (searchdirs == NULL) {
3300 				searchdirs = safe_malloc(sizeof (char *));
3301 			} else {
3302 				char **tmp = safe_malloc((nsearch + 1) *
3303 				    sizeof (char *));
3304 				bcopy(searchdirs, tmp, nsearch *
3305 				    sizeof (char *));
3306 				free(searchdirs);
3307 				searchdirs = tmp;
3308 			}
3309 			searchdirs[nsearch++] = optarg;
3310 			break;
3311 		case 'D':
3312 			do_destroyed = B_TRUE;
3313 			break;
3314 		case 'f':
3315 			flags |= ZFS_IMPORT_ANY_HOST;
3316 			break;
3317 		case 'F':
3318 			do_rewind = B_TRUE;
3319 			break;
3320 		case 'l':
3321 			flags |= ZFS_IMPORT_LOAD_KEYS;
3322 			break;
3323 		case 'm':
3324 			flags |= ZFS_IMPORT_MISSING_LOG;
3325 			break;
3326 		case 'n':
3327 			dryrun = B_TRUE;
3328 			break;
3329 		case 'N':
3330 			flags |= ZFS_IMPORT_ONLY;
3331 			break;
3332 		case 'o':
3333 			if ((propval = strchr(optarg, '=')) != NULL) {
3334 				*propval = '\0';
3335 				propval++;
3336 				if (add_prop_list(optarg, propval,
3337 				    &props, B_TRUE))
3338 					goto error;
3339 			} else {
3340 				mntopts = optarg;
3341 			}
3342 			break;
3343 		case 'R':
3344 			if (add_prop_list(zpool_prop_to_name(
3345 			    ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
3346 				goto error;
3347 			if (add_prop_list_default(zpool_prop_to_name(
3348 			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
3349 				goto error;
3350 			break;
3351 		case 's':
3352 			do_scan = B_TRUE;
3353 			break;
3354 		case 't':
3355 			flags |= ZFS_IMPORT_TEMP_NAME;
3356 			if (add_prop_list_default(zpool_prop_to_name(
3357 			    ZPOOL_PROP_CACHEFILE), "none", &props, B_TRUE))
3358 				goto error;
3359 			break;
3360 
3361 		case 'T':
3362 			errno = 0;
3363 			txg = strtoull(optarg, &endptr, 0);
3364 			if (errno != 0 || *endptr != '\0') {
3365 				(void) fprintf(stderr,
3366 				    gettext("invalid txg value\n"));
3367 				usage(B_FALSE);
3368 			}
3369 			rewind_policy = ZPOOL_DO_REWIND | ZPOOL_EXTREME_REWIND;
3370 			break;
3371 		case 'V':
3372 			flags |= ZFS_IMPORT_VERBATIM;
3373 			break;
3374 		case 'X':
3375 			xtreme_rewind = B_TRUE;
3376 			break;
3377 		case CHECKPOINT_OPT:
3378 			flags |= ZFS_IMPORT_CHECKPOINT;
3379 			break;
3380 		case ':':
3381 			(void) fprintf(stderr, gettext("missing argument for "
3382 			    "'%c' option\n"), optopt);
3383 			usage(B_FALSE);
3384 			break;
3385 		case '?':
3386 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
3387 			    optopt);
3388 			usage(B_FALSE);
3389 		}
3390 	}
3391 
3392 	argc -= optind;
3393 	argv += optind;
3394 
3395 	if (cachefile && nsearch != 0) {
3396 		(void) fprintf(stderr, gettext("-c is incompatible with -d\n"));
3397 		usage(B_FALSE);
3398 	}
3399 
3400 	if ((flags & ZFS_IMPORT_LOAD_KEYS) && (flags & ZFS_IMPORT_ONLY)) {
3401 		(void) fprintf(stderr, gettext("-l is incompatible with -N\n"));
3402 		usage(B_FALSE);
3403 	}
3404 
3405 	if ((flags & ZFS_IMPORT_LOAD_KEYS) && !do_all && argc == 0) {
3406 		(void) fprintf(stderr, gettext("-l is only meaningful during "
3407 		    "an import\n"));
3408 		usage(B_FALSE);
3409 	}
3410 
3411 	if ((dryrun || xtreme_rewind) && !do_rewind) {
3412 		(void) fprintf(stderr,
3413 		    gettext("-n or -X only meaningful with -F\n"));
3414 		usage(B_FALSE);
3415 	}
3416 	if (dryrun)
3417 		rewind_policy = ZPOOL_TRY_REWIND;
3418 	else if (do_rewind)
3419 		rewind_policy = ZPOOL_DO_REWIND;
3420 	if (xtreme_rewind)
3421 		rewind_policy |= ZPOOL_EXTREME_REWIND;
3422 
3423 	/* In the future, we can capture further policy and include it here */
3424 	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 ||
3425 	    nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, txg) != 0 ||
3426 	    nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
3427 	    rewind_policy) != 0)
3428 		goto error;
3429 
3430 	/* check argument count */
3431 	if (do_all) {
3432 		if (argc != 0) {
3433 			(void) fprintf(stderr, gettext("too many arguments\n"));
3434 			usage(B_FALSE);
3435 		}
3436 	} else {
3437 		if (argc > 2) {
3438 			(void) fprintf(stderr, gettext("too many arguments\n"));
3439 			usage(B_FALSE);
3440 		}
3441 	}
3442 
3443 	/*
3444 	 * Check for the effective uid.  We do this explicitly here because
3445 	 * otherwise any attempt to discover pools will silently fail.
3446 	 */
3447 	if (argc == 0 && geteuid() != 0) {
3448 		(void) fprintf(stderr, gettext("cannot "
3449 		    "discover pools: permission denied\n"));
3450 		if (searchdirs != NULL)
3451 			free(searchdirs);
3452 
3453 		nvlist_free(props);
3454 		nvlist_free(policy);
3455 		return (1);
3456 	}
3457 
3458 	/*
3459 	 * Depending on the arguments given, we do one of the following:
3460 	 *
3461 	 *	<none>	Iterate through all pools and display information about
3462 	 *		each one.
3463 	 *
3464 	 *	-a	Iterate through all pools and try to import each one.
3465 	 *
3466 	 *	<id>	Find the pool that corresponds to the given GUID/pool
3467 	 *		name and import that one.
3468 	 *
3469 	 *	-D	Above options applies only to destroyed pools.
3470 	 */
3471 	if (argc != 0) {
3472 		char *endptr;
3473 
3474 		errno = 0;
3475 		searchguid = strtoull(argv[0], &endptr, 10);
3476 		if (errno != 0 || *endptr != '\0') {
3477 			searchname = argv[0];
3478 			searchguid = 0;
3479 		}
3480 		found_config = NULL;
3481 
3482 		/*
3483 		 * User specified a name or guid.  Ensure it's unique.
3484 		 */
3485 		target_exists_args_t search = {searchname, searchguid};
3486 		pool_exists = zpool_iter(g_zfs, name_or_guid_exists, &search);
3487 	}
3488 
3489 	/*
3490 	 * Check the environment for the preferred search path.
3491 	 */
3492 	if ((searchdirs == NULL) && (env = getenv("ZPOOL_IMPORT_PATH"))) {
3493 		char *dir;
3494 
3495 		envdup = strdup(env);
3496 
3497 		dir = strtok(envdup, ":");
3498 		while (dir != NULL) {
3499 			if (searchdirs == NULL) {
3500 				searchdirs = safe_malloc(sizeof (char *));
3501 			} else {
3502 				char **tmp = safe_malloc((nsearch + 1) *
3503 				    sizeof (char *));
3504 				bcopy(searchdirs, tmp, nsearch *
3505 				    sizeof (char *));
3506 				free(searchdirs);
3507 				searchdirs = tmp;
3508 			}
3509 			searchdirs[nsearch++] = dir;
3510 			dir = strtok(NULL, ":");
3511 		}
3512 	}
3513 
3514 	idata.path = searchdirs;
3515 	idata.paths = nsearch;
3516 	idata.poolname = searchname;
3517 	idata.guid = searchguid;
3518 	idata.cachefile = cachefile;
3519 	idata.scan = do_scan;
3520 	idata.policy = policy;
3521 
3522 	pools = zpool_search_import(g_zfs, &idata, &libzfs_config_ops);
3523 
3524 	if (pools != NULL && pool_exists &&
3525 	    (argc == 1 || strcmp(argv[0], argv[1]) == 0)) {
3526 		(void) fprintf(stderr, gettext("cannot import '%s': "
3527 		    "a pool with that name already exists\n"),
3528 		    argv[0]);
3529 		(void) fprintf(stderr, gettext("use the form '%s "
3530 		    "<pool | id> <newpool>' to give it a new name\n"),
3531 		    "zpool import");
3532 		err = 1;
3533 	} else if (pools == NULL && pool_exists) {
3534 		(void) fprintf(stderr, gettext("cannot import '%s': "
3535 		    "a pool with that name is already created/imported,\n"),
3536 		    argv[0]);
3537 		(void) fprintf(stderr, gettext("and no additional pools "
3538 		    "with that name were found\n"));
3539 		err = 1;
3540 	} else if (pools == NULL) {
3541 		if (argc != 0) {
3542 			(void) fprintf(stderr, gettext("cannot import '%s': "
3543 			    "no such pool available\n"), argv[0]);
3544 		}
3545 		err = 1;
3546 	}
3547 
3548 	if (err == 1) {
3549 		if (searchdirs != NULL)
3550 			free(searchdirs);
3551 		if (envdup != NULL)
3552 			free(envdup);
3553 		nvlist_free(policy);
3554 		nvlist_free(pools);
3555 		nvlist_free(props);
3556 		return (1);
3557 	}
3558 
3559 	/*
3560 	 * At this point we have a list of import candidate configs. Even if
3561 	 * we were searching by pool name or guid, we still need to
3562 	 * post-process the list to deal with pool state and possible
3563 	 * duplicate names.
3564 	 */
3565 	err = 0;
3566 	elem = NULL;
3567 	first = B_TRUE;
3568 	while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
3569 
3570 		verify(nvpair_value_nvlist(elem, &config) == 0);
3571 
3572 		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
3573 		    &pool_state) == 0);
3574 		if (!do_destroyed && pool_state == POOL_STATE_DESTROYED)
3575 			continue;
3576 		if (do_destroyed && pool_state != POOL_STATE_DESTROYED)
3577 			continue;
3578 
3579 		verify(nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
3580 		    policy) == 0);
3581 
3582 		if (argc == 0) {
3583 			if (first)
3584 				first = B_FALSE;
3585 			else if (!do_all)
3586 				(void) printf("\n");
3587 
3588 			if (do_all) {
3589 				err |= do_import(config, NULL, mntopts,
3590 				    props, flags);
3591 			} else {
3592 				show_import(config);
3593 			}
3594 		} else if (searchname != NULL) {
3595 			char *name;
3596 
3597 			/*
3598 			 * We are searching for a pool based on name.
3599 			 */
3600 			verify(nvlist_lookup_string(config,
3601 			    ZPOOL_CONFIG_POOL_NAME, &name) == 0);
3602 
3603 			if (strcmp(name, searchname) == 0) {
3604 				if (found_config != NULL) {
3605 					(void) fprintf(stderr, gettext(
3606 					    "cannot import '%s': more than "
3607 					    "one matching pool\n"), searchname);
3608 					(void) fprintf(stderr, gettext(
3609 					    "import by numeric ID instead\n"));
3610 					err = B_TRUE;
3611 				}
3612 				found_config = config;
3613 			}
3614 		} else {
3615 			uint64_t guid;
3616 
3617 			/*
3618 			 * Search for a pool by guid.
3619 			 */
3620 			verify(nvlist_lookup_uint64(config,
3621 			    ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
3622 
3623 			if (guid == searchguid)
3624 				found_config = config;
3625 		}
3626 	}
3627 
3628 	/*
3629 	 * If we were searching for a specific pool, verify that we found a
3630 	 * pool, and then do the import.
3631 	 */
3632 	if (argc != 0 && err == 0) {
3633 		if (found_config == NULL) {
3634 			(void) fprintf(stderr, gettext("cannot import '%s': "
3635 			    "no such pool available\n"), argv[0]);
3636 			err = B_TRUE;
3637 		} else {
3638 			err |= do_import(found_config, argc == 1 ? NULL :
3639 			    argv[1], mntopts, props, flags);
3640 		}
3641 	}
3642 
3643 	/*
3644 	 * If we were just looking for pools, report an error if none were
3645 	 * found.
3646 	 */
3647 	if (argc == 0 && first)
3648 		(void) fprintf(stderr,
3649 		    gettext("no pools available to import\n"));
3650 
3651 error:
3652 	nvlist_free(props);
3653 	nvlist_free(pools);
3654 	nvlist_free(policy);
3655 	if (searchdirs != NULL)
3656 		free(searchdirs);
3657 	if (envdup != NULL)
3658 		free(envdup);
3659 
3660 	return (err ? 1 : 0);
3661 }
3662 
3663 /*
3664  * zpool sync [-f] [pool] ...
3665  *
3666  * -f (undocumented) force uberblock (and config including zpool cache file)
3667  *    update.
3668  *
3669  * Sync the specified pool(s).
3670  * Without arguments "zpool sync" will sync all pools.
3671  * This command initiates TXG sync(s) and will return after the TXG(s) commit.
3672  *
3673  */
3674 static int
3675 zpool_do_sync(int argc, char **argv)
3676 {
3677 	int ret;
3678 	boolean_t force = B_FALSE;
3679 
3680 	/* check options */
3681 	while ((ret  = getopt(argc, argv, "f")) != -1) {
3682 		switch (ret) {
3683 		case 'f':
3684 			force = B_TRUE;
3685 			break;
3686 		case '?':
3687 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
3688 			    optopt);
3689 			usage(B_FALSE);
3690 		}
3691 	}
3692 
3693 	argc -= optind;
3694 	argv += optind;
3695 
3696 	/* if argc == 0 we will execute zpool_sync_one on all pools */
3697 	ret = for_each_pool(argc, argv, B_FALSE, NULL, B_FALSE, zpool_sync_one,
3698 	    &force);
3699 
3700 	return (ret);
3701 }
3702 
3703 typedef struct iostat_cbdata {
3704 	uint64_t cb_flags;
3705 	int cb_name_flags;
3706 	int cb_namewidth;
3707 	int cb_iteration;
3708 	char **cb_vdev_names; /* Only show these vdevs */
3709 	unsigned int cb_vdev_names_count;
3710 	boolean_t cb_verbose;
3711 	boolean_t cb_literal;
3712 	boolean_t cb_scripted;
3713 	zpool_list_t *cb_list;
3714 	vdev_cmd_data_list_t *vcdl;
3715 } iostat_cbdata_t;
3716 
3717 /*  iostat labels */
3718 typedef struct name_and_columns {
3719 	const char *name;	/* Column name */
3720 	unsigned int columns;	/* Center name to this number of columns */
3721 } name_and_columns_t;
3722 
3723 #define	IOSTAT_MAX_LABELS	13	/* Max number of labels on one line */
3724 
3725 static const name_and_columns_t iostat_top_labels[][IOSTAT_MAX_LABELS] =
3726 {
3727 	[IOS_DEFAULT] = {{"capacity", 2}, {"operations", 2}, {"bandwidth", 2},
3728 	    {NULL}},
3729 	[IOS_LATENCY] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
3730 	    {"asyncq_wait", 2}, {"scrub", 1}, {"trim", 1}, {NULL}},
3731 	[IOS_QUEUES] = {{"syncq_read", 2}, {"syncq_write", 2},
3732 	    {"asyncq_read", 2}, {"asyncq_write", 2}, {"scrubq_read", 2},
3733 	    {"trimq_write", 2}, {NULL}},
3734 	[IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
3735 	    {"asyncq_wait", 2}, {NULL}},
3736 	[IOS_RQ_HISTO] = {{"sync_read", 2}, {"sync_write", 2},
3737 	    {"async_read", 2}, {"async_write", 2}, {"scrub", 2},
3738 	    {"trim", 2}, {NULL}},
3739 };
3740 
3741 /* Shorthand - if "columns" field not set, default to 1 column */
3742 static const name_and_columns_t iostat_bottom_labels[][IOSTAT_MAX_LABELS] =
3743 {
3744 	[IOS_DEFAULT] = {{"alloc"}, {"free"}, {"read"}, {"write"}, {"read"},
3745 	    {"write"}, {NULL}},
3746 	[IOS_LATENCY] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
3747 	    {"write"}, {"read"}, {"write"}, {"wait"}, {"wait"}, {NULL}},
3748 	[IOS_QUEUES] = {{"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"},
3749 	    {"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"},
3750 	    {"pend"}, {"activ"}, {NULL}},
3751 	[IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
3752 	    {"write"}, {"read"}, {"write"}, {"scrub"}, {"trim"}, {NULL}},
3753 	[IOS_RQ_HISTO] = {{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
3754 	    {"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"}, {NULL}},
3755 };
3756 
3757 static const char *histo_to_title[] = {
3758 	[IOS_L_HISTO] = "latency",
3759 	[IOS_RQ_HISTO] = "req_size",
3760 };
3761 
3762 /*
3763  * Return the number of labels in a null-terminated name_and_columns_t
3764  * array.
3765  *
3766  */
3767 static unsigned int
3768 label_array_len(const name_and_columns_t *labels)
3769 {
3770 	int i = 0;
3771 
3772 	while (labels[i].name)
3773 		i++;
3774 
3775 	return (i);
3776 }
3777 
3778 /*
3779  * Return the number of strings in a null-terminated string array.
3780  * For example:
3781  *
3782  *     const char foo[] = {"bar", "baz", NULL}
3783  *
3784  * returns 2
3785  */
3786 static uint64_t
3787 str_array_len(const char *array[])
3788 {
3789 	uint64_t i = 0;
3790 	while (array[i])
3791 		i++;
3792 
3793 	return (i);
3794 }
3795 
3796 
3797 /*
3798  * Return a default column width for default/latency/queue columns. This does
3799  * not include histograms, which have their columns autosized.
3800  */
3801 static unsigned int
3802 default_column_width(iostat_cbdata_t *cb, enum iostat_type type)
3803 {
3804 	unsigned long column_width = 5; /* Normal niceprint */
3805 	static unsigned long widths[] = {
3806 		/*
3807 		 * Choose some sane default column sizes for printing the
3808 		 * raw numbers.
3809 		 */
3810 		[IOS_DEFAULT] = 15, /* 1PB capacity */
3811 		[IOS_LATENCY] = 10, /* 1B ns = 10sec */
3812 		[IOS_QUEUES] = 6,   /* 1M queue entries */
3813 		[IOS_L_HISTO] = 10, /* 1B ns = 10sec */
3814 		[IOS_RQ_HISTO] = 6, /* 1M queue entries */
3815 	};
3816 
3817 	if (cb->cb_literal)
3818 		column_width = widths[type];
3819 
3820 	return (column_width);
3821 }
3822 
3823 /*
3824  * Print the column labels, i.e:
3825  *
3826  *   capacity     operations     bandwidth
3827  * alloc   free   read  write   read  write  ...
3828  *
3829  * If force_column_width is set, use it for the column width.  If not set, use
3830  * the default column width.
3831  */
3832 static void
3833 print_iostat_labels(iostat_cbdata_t *cb, unsigned int force_column_width,
3834     const name_and_columns_t labels[][IOSTAT_MAX_LABELS])
3835 {
3836 	int i, idx, s;
3837 	int text_start, rw_column_width, spaces_to_end;
3838 	uint64_t flags = cb->cb_flags;
3839 	uint64_t f;
3840 	unsigned int column_width = force_column_width;
3841 
3842 	/* For each bit set in flags */
3843 	for (f = flags; f; f &= ~(1ULL << idx)) {
3844 		idx = lowbit64(f) - 1;
3845 		if (!force_column_width)
3846 			column_width = default_column_width(cb, idx);
3847 		/* Print our top labels centered over "read  write" label. */
3848 		for (i = 0; i < label_array_len(labels[idx]); i++) {
3849 			const char *name = labels[idx][i].name;
3850 			/*
3851 			 * We treat labels[][].columns == 0 as shorthand
3852 			 * for one column.  It makes writing out the label
3853 			 * tables more concise.
3854 			 */
3855 			unsigned int columns = MAX(1, labels[idx][i].columns);
3856 			unsigned int slen = strlen(name);
3857 
3858 			rw_column_width = (column_width * columns) +
3859 			    (2 * (columns - 1));
3860 
3861 			text_start = (int)((rw_column_width) / columns -
3862 			    slen / columns);
3863 			if (text_start < 0)
3864 				text_start = 0;
3865 
3866 			printf("  ");	/* Two spaces between columns */
3867 
3868 			/* Space from beginning of column to label */
3869 			for (s = 0; s < text_start; s++)
3870 				printf(" ");
3871 
3872 			printf("%s", name);
3873 
3874 			/* Print space after label to end of column */
3875 			spaces_to_end = rw_column_width - text_start - slen;
3876 			if (spaces_to_end < 0)
3877 				spaces_to_end = 0;
3878 
3879 			for (s = 0; s < spaces_to_end; s++)
3880 				printf(" ");
3881 		}
3882 	}
3883 }
3884 
3885 
3886 /*
3887  * print_cmd_columns - Print custom column titles from -c
3888  *
3889  * If the user specified the "zpool status|iostat -c" then print their custom
3890  * column titles in the header.  For example, print_cmd_columns() would print
3891  * the "  col1  col2" part of this:
3892  *
3893  * $ zpool iostat -vc 'echo col1=val1; echo col2=val2'
3894  * ...
3895  *	      capacity     operations     bandwidth
3896  * pool        alloc   free   read  write   read  write  col1  col2
3897  * ----------  -----  -----  -----  -----  -----  -----  ----  ----
3898  * mypool       269K  1008M      0      0    107    946
3899  *   mirror     269K  1008M      0      0    107    946
3900  *     sdb         -      -      0      0    102    473  val1  val2
3901  *     sdc         -      -      0      0      5    473  val1  val2
3902  * ----------  -----  -----  -----  -----  -----  -----  ----  ----
3903  */
3904 static void
3905 print_cmd_columns(vdev_cmd_data_list_t *vcdl, int use_dashes)
3906 {
3907 	int i, j;
3908 	vdev_cmd_data_t *data = &vcdl->data[0];
3909 
3910 	if (vcdl->count == 0 || data == NULL)
3911 		return;
3912 
3913 	/*
3914 	 * Each vdev cmd should have the same column names unless the user did
3915 	 * something weird with their cmd.  Just take the column names from the
3916 	 * first vdev and assume it works for all of them.
3917 	 */
3918 	for (i = 0; i < vcdl->uniq_cols_cnt; i++) {
3919 		printf("  ");
3920 		if (use_dashes) {
3921 			for (j = 0; j < vcdl->uniq_cols_width[i]; j++)
3922 				printf("-");
3923 		} else {
3924 			printf_color(ANSI_BOLD, "%*s", vcdl->uniq_cols_width[i],
3925 			    vcdl->uniq_cols[i]);
3926 		}
3927 	}
3928 }
3929 
3930 
3931 /*
3932  * Utility function to print out a line of dashes like:
3933  *
3934  * 	--------------------------------  -----  -----  -----  -----  -----
3935  *
3936  * ...or a dashed named-row line like:
3937  *
3938  * 	logs                                  -      -      -      -      -
3939  *
3940  * @cb:				iostat data
3941  *
3942  * @force_column_width		If non-zero, use the value as the column width.
3943  * 				Otherwise use the default column widths.
3944  *
3945  * @name:			Print a dashed named-row line starting
3946  * 				with @name.  Otherwise, print a regular
3947  * 				dashed line.
3948  */
3949 static void
3950 print_iostat_dashes(iostat_cbdata_t *cb, unsigned int force_column_width,
3951     const char *name)
3952 {
3953 	int i;
3954 	unsigned int namewidth;
3955 	uint64_t flags = cb->cb_flags;
3956 	uint64_t f;
3957 	int idx;
3958 	const name_and_columns_t *labels;
3959 	const char *title;
3960 
3961 
3962 	if (cb->cb_flags & IOS_ANYHISTO_M) {
3963 		title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
3964 	} else if (cb->cb_vdev_names_count) {
3965 		title = "vdev";
3966 	} else  {
3967 		title = "pool";
3968 	}
3969 
3970 	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
3971 	    name ? strlen(name) : 0);
3972 
3973 
3974 	if (name) {
3975 		printf("%-*s", namewidth, name);
3976 	} else {
3977 		for (i = 0; i < namewidth; i++)
3978 			(void) printf("-");
3979 	}
3980 
3981 	/* For each bit in flags */
3982 	for (f = flags; f; f &= ~(1ULL << idx)) {
3983 		unsigned int column_width;
3984 		idx = lowbit64(f) - 1;
3985 		if (force_column_width)
3986 			column_width = force_column_width;
3987 		else
3988 			column_width = default_column_width(cb, idx);
3989 
3990 		labels = iostat_bottom_labels[idx];
3991 		for (i = 0; i < label_array_len(labels); i++) {
3992 			if (name)
3993 				printf("  %*s-", column_width - 1, " ");
3994 			else
3995 				printf("  %.*s", column_width,
3996 				    "--------------------");
3997 		}
3998 	}
3999 }
4000 
4001 
4002 static void
4003 print_iostat_separator_impl(iostat_cbdata_t *cb,
4004     unsigned int force_column_width)
4005 {
4006 	print_iostat_dashes(cb, force_column_width, NULL);
4007 }
4008 
4009 static void
4010 print_iostat_separator(iostat_cbdata_t *cb)
4011 {
4012 	print_iostat_separator_impl(cb, 0);
4013 }
4014 
4015 static void
4016 print_iostat_header_impl(iostat_cbdata_t *cb, unsigned int force_column_width,
4017     const char *histo_vdev_name)
4018 {
4019 	unsigned int namewidth;
4020 	const char *title;
4021 
4022 	if (cb->cb_flags & IOS_ANYHISTO_M) {
4023 		title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
4024 	} else if (cb->cb_vdev_names_count) {
4025 		title = "vdev";
4026 	} else  {
4027 		title = "pool";
4028 	}
4029 
4030 	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
4031 	    histo_vdev_name ? strlen(histo_vdev_name) : 0);
4032 
4033 	if (histo_vdev_name)
4034 		printf("%-*s", namewidth, histo_vdev_name);
4035 	else
4036 		printf("%*s", namewidth, "");
4037 
4038 
4039 	print_iostat_labels(cb, force_column_width, iostat_top_labels);
4040 	printf("\n");
4041 
4042 	printf("%-*s", namewidth, title);
4043 
4044 	print_iostat_labels(cb, force_column_width, iostat_bottom_labels);
4045 	if (cb->vcdl != NULL)
4046 		print_cmd_columns(cb->vcdl, 0);
4047 
4048 	printf("\n");
4049 
4050 	print_iostat_separator_impl(cb, force_column_width);
4051 
4052 	if (cb->vcdl != NULL)
4053 		print_cmd_columns(cb->vcdl, 1);
4054 
4055 	printf("\n");
4056 }
4057 
4058 static void
4059 print_iostat_header(iostat_cbdata_t *cb)
4060 {
4061 	print_iostat_header_impl(cb, 0, NULL);
4062 }
4063 
4064 
4065 /*
4066  * Display a single statistic.
4067  */
4068 static void
4069 print_one_stat(uint64_t value, enum zfs_nicenum_format format,
4070     unsigned int column_size, boolean_t scripted)
4071 {
4072 	char buf[64];
4073 
4074 	zfs_nicenum_format(value, buf, sizeof (buf), format);
4075 
4076 	if (scripted)
4077 		printf("\t%s", buf);
4078 	else
4079 		printf("  %*s", column_size, buf);
4080 }
4081 
4082 /*
4083  * Calculate the default vdev stats
4084  *
4085  * Subtract oldvs from newvs, apply a scaling factor, and save the resulting
4086  * stats into calcvs.
4087  */
4088 static void
4089 calc_default_iostats(vdev_stat_t *oldvs, vdev_stat_t *newvs,
4090     vdev_stat_t *calcvs)
4091 {
4092 	int i;
4093 
4094 	memcpy(calcvs, newvs, sizeof (*calcvs));
4095 	for (i = 0; i < ARRAY_SIZE(calcvs->vs_ops); i++)
4096 		calcvs->vs_ops[i] = (newvs->vs_ops[i] - oldvs->vs_ops[i]);
4097 
4098 	for (i = 0; i < ARRAY_SIZE(calcvs->vs_bytes); i++)
4099 		calcvs->vs_bytes[i] = (newvs->vs_bytes[i] - oldvs->vs_bytes[i]);
4100 }
4101 
4102 /*
4103  * Internal representation of the extended iostats data.
4104  *
4105  * The extended iostat stats are exported in nvlists as either uint64_t arrays
4106  * or single uint64_t's.  We make both look like arrays to make them easier
4107  * to process.  In order to make single uint64_t's look like arrays, we set
4108  * __data to the stat data, and then set *data = &__data with count = 1.  Then,
4109  * we can just use *data and count.
4110  */
4111 struct stat_array {
4112 	uint64_t *data;
4113 	uint_t count;	/* Number of entries in data[] */
4114 	uint64_t __data; /* Only used when data is a single uint64_t */
4115 };
4116 
4117 static uint64_t
4118 stat_histo_max(struct stat_array *nva, unsigned int len)
4119 {
4120 	uint64_t max = 0;
4121 	int i;
4122 	for (i = 0; i < len; i++)
4123 		max = MAX(max, array64_max(nva[i].data, nva[i].count));
4124 
4125 	return (max);
4126 }
4127 
4128 /*
4129  * Helper function to lookup a uint64_t array or uint64_t value and store its
4130  * data as a stat_array.  If the nvpair is a single uint64_t value, then we make
4131  * it look like a one element array to make it easier to process.
4132  */
4133 static int
4134 nvpair64_to_stat_array(nvlist_t *nvl, const char *name,
4135     struct stat_array *nva)
4136 {
4137 	nvpair_t *tmp;
4138 	int ret;
4139 
4140 	verify(nvlist_lookup_nvpair(nvl, name, &tmp) == 0);
4141 	switch (nvpair_type(tmp)) {
4142 	case DATA_TYPE_UINT64_ARRAY:
4143 		ret = nvpair_value_uint64_array(tmp, &nva->data, &nva->count);
4144 		break;
4145 	case DATA_TYPE_UINT64:
4146 		ret = nvpair_value_uint64(tmp, &nva->__data);
4147 		nva->data = &nva->__data;
4148 		nva->count = 1;
4149 		break;
4150 	default:
4151 		/* Not a uint64_t */
4152 		ret = EINVAL;
4153 		break;
4154 	}
4155 
4156 	return (ret);
4157 }
4158 
4159 /*
4160  * Given a list of nvlist names, look up the extended stats in newnv and oldnv,
4161  * subtract them, and return the results in a newly allocated stat_array.
4162  * You must free the returned array after you are done with it with
4163  * free_calc_stats().
4164  *
4165  * Additionally, you can set "oldnv" to NULL if you simply want the newnv
4166  * values.
4167  */
4168 static struct stat_array *
4169 calc_and_alloc_stats_ex(const char **names, unsigned int len, nvlist_t *oldnv,
4170     nvlist_t *newnv)
4171 {
4172 	nvlist_t *oldnvx = NULL, *newnvx;
4173 	struct stat_array *oldnva, *newnva, *calcnva;
4174 	int i, j;
4175 	unsigned int alloc_size = (sizeof (struct stat_array)) * len;
4176 
4177 	/* Extract our extended stats nvlist from the main list */
4178 	verify(nvlist_lookup_nvlist(newnv, ZPOOL_CONFIG_VDEV_STATS_EX,
4179 	    &newnvx) == 0);
4180 	if (oldnv) {
4181 		verify(nvlist_lookup_nvlist(oldnv, ZPOOL_CONFIG_VDEV_STATS_EX,
4182 		    &oldnvx) == 0);
4183 	}
4184 
4185 	newnva = safe_malloc(alloc_size);
4186 	oldnva = safe_malloc(alloc_size);
4187 	calcnva = safe_malloc(alloc_size);
4188 
4189 	for (j = 0; j < len; j++) {
4190 		verify(nvpair64_to_stat_array(newnvx, names[j],
4191 		    &newnva[j]) == 0);
4192 		calcnva[j].count = newnva[j].count;
4193 		alloc_size = calcnva[j].count * sizeof (calcnva[j].data[0]);
4194 		calcnva[j].data = safe_malloc(alloc_size);
4195 		memcpy(calcnva[j].data, newnva[j].data, alloc_size);
4196 
4197 		if (oldnvx) {
4198 			verify(nvpair64_to_stat_array(oldnvx, names[j],
4199 			    &oldnva[j]) == 0);
4200 			for (i = 0; i < oldnva[j].count; i++)
4201 				calcnva[j].data[i] -= oldnva[j].data[i];
4202 		}
4203 	}
4204 	free(newnva);
4205 	free(oldnva);
4206 	return (calcnva);
4207 }
4208 
4209 static void
4210 free_calc_stats(struct stat_array *nva, unsigned int len)
4211 {
4212 	int i;
4213 	for (i = 0; i < len; i++)
4214 		free(nva[i].data);
4215 
4216 	free(nva);
4217 }
4218 
4219 static void
4220 print_iostat_histo(struct stat_array *nva, unsigned int len,
4221     iostat_cbdata_t *cb, unsigned int column_width, unsigned int namewidth,
4222     double scale)
4223 {
4224 	int i, j;
4225 	char buf[6];
4226 	uint64_t val;
4227 	enum zfs_nicenum_format format;
4228 	unsigned int buckets;
4229 	unsigned int start_bucket;
4230 
4231 	if (cb->cb_literal)
4232 		format = ZFS_NICENUM_RAW;
4233 	else
4234 		format = ZFS_NICENUM_1024;
4235 
4236 	/* All these histos are the same size, so just use nva[0].count */
4237 	buckets = nva[0].count;
4238 
4239 	if (cb->cb_flags & IOS_RQ_HISTO_M) {
4240 		/* Start at 512 - req size should never be lower than this */
4241 		start_bucket = 9;
4242 	} else {
4243 		start_bucket = 0;
4244 	}
4245 
4246 	for (j = start_bucket; j < buckets; j++) {
4247 		/* Print histogram bucket label */
4248 		if (cb->cb_flags & IOS_L_HISTO_M) {
4249 			/* Ending range of this bucket */
4250 			val = (1UL << (j + 1)) - 1;
4251 			zfs_nicetime(val, buf, sizeof (buf));
4252 		} else {
4253 			/* Request size (starting range of bucket) */
4254 			val = (1UL << j);
4255 			zfs_nicenum(val, buf, sizeof (buf));
4256 		}
4257 
4258 		if (cb->cb_scripted)
4259 			printf("%llu", (u_longlong_t)val);
4260 		else
4261 			printf("%-*s", namewidth, buf);
4262 
4263 		/* Print the values on the line */
4264 		for (i = 0; i < len; i++) {
4265 			print_one_stat(nva[i].data[j] * scale, format,
4266 			    column_width, cb->cb_scripted);
4267 		}
4268 		printf("\n");
4269 	}
4270 }
4271 
4272 static void
4273 print_solid_separator(unsigned int length)
4274 {
4275 	while (length--)
4276 		printf("-");
4277 	printf("\n");
4278 }
4279 
4280 static void
4281 print_iostat_histos(iostat_cbdata_t *cb, nvlist_t *oldnv,
4282     nvlist_t *newnv, double scale, const char *name)
4283 {
4284 	unsigned int column_width;
4285 	unsigned int namewidth;
4286 	unsigned int entire_width;
4287 	enum iostat_type type;
4288 	struct stat_array *nva;
4289 	const char **names;
4290 	unsigned int names_len;
4291 
4292 	/* What type of histo are we? */
4293 	type = IOS_HISTO_IDX(cb->cb_flags);
4294 
4295 	/* Get NULL-terminated array of nvlist names for our histo */
4296 	names = vsx_type_to_nvlist[type];
4297 	names_len = str_array_len(names); /* num of names */
4298 
4299 	nva = calc_and_alloc_stats_ex(names, names_len, oldnv, newnv);
4300 
4301 	if (cb->cb_literal) {
4302 		column_width = MAX(5,
4303 		    (unsigned int) log10(stat_histo_max(nva, names_len)) + 1);
4304 	} else {
4305 		column_width = 5;
4306 	}
4307 
4308 	namewidth = MAX(cb->cb_namewidth,
4309 	    strlen(histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]));
4310 
4311 	/*
4312 	 * Calculate the entire line width of what we're printing.  The
4313 	 * +2 is for the two spaces between columns:
4314 	 */
4315 	/*	 read  write				*/
4316 	/*	-----  -----				*/
4317 	/*	|___|  <---------- column_width		*/
4318 	/*						*/
4319 	/*	|__________|  <--- entire_width		*/
4320 	/*						*/
4321 	entire_width = namewidth + (column_width + 2) *
4322 	    label_array_len(iostat_bottom_labels[type]);
4323 
4324 	if (cb->cb_scripted)
4325 		printf("%s\n", name);
4326 	else
4327 		print_iostat_header_impl(cb, column_width, name);
4328 
4329 	print_iostat_histo(nva, names_len, cb, column_width,
4330 	    namewidth, scale);
4331 
4332 	free_calc_stats(nva, names_len);
4333 	if (!cb->cb_scripted)
4334 		print_solid_separator(entire_width);
4335 }
4336 
4337 /*
4338  * Calculate the average latency of a power-of-two latency histogram
4339  */
4340 static uint64_t
4341 single_histo_average(uint64_t *histo, unsigned int buckets)
4342 {
4343 	int i;
4344 	uint64_t count = 0, total = 0;
4345 
4346 	for (i = 0; i < buckets; i++) {
4347 		/*
4348 		 * Our buckets are power-of-two latency ranges.  Use the
4349 		 * midpoint latency of each bucket to calculate the average.
4350 		 * For example:
4351 		 *
4352 		 * Bucket          Midpoint
4353 		 * 8ns-15ns:       12ns
4354 		 * 16ns-31ns:      24ns
4355 		 * ...
4356 		 */
4357 		if (histo[i] != 0) {
4358 			total += histo[i] * (((1UL << i) + ((1UL << i)/2)));
4359 			count += histo[i];
4360 		}
4361 	}
4362 
4363 	/* Prevent divide by zero */
4364 	return (count == 0 ? 0 : total / count);
4365 }
4366 
4367 static void
4368 print_iostat_queues(iostat_cbdata_t *cb, nvlist_t *oldnv,
4369     nvlist_t *newnv)
4370 {
4371 	int i;
4372 	uint64_t val;
4373 	const char *names[] = {
4374 		ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE,
4375 		ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
4376 		ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE,
4377 		ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
4378 		ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE,
4379 		ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
4380 		ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE,
4381 		ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
4382 		ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE,
4383 		ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
4384 		ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE,
4385 		ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
4386 	};
4387 
4388 	struct stat_array *nva;
4389 
4390 	unsigned int column_width = default_column_width(cb, IOS_QUEUES);
4391 	enum zfs_nicenum_format format;
4392 
4393 	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), NULL, newnv);
4394 
4395 	if (cb->cb_literal)
4396 		format = ZFS_NICENUM_RAW;
4397 	else
4398 		format = ZFS_NICENUM_1024;
4399 
4400 	for (i = 0; i < ARRAY_SIZE(names); i++) {
4401 		val = nva[i].data[0];
4402 		print_one_stat(val, format, column_width, cb->cb_scripted);
4403 	}
4404 
4405 	free_calc_stats(nva, ARRAY_SIZE(names));
4406 }
4407 
4408 static void
4409 print_iostat_latency(iostat_cbdata_t *cb, nvlist_t *oldnv,
4410     nvlist_t *newnv)
4411 {
4412 	int i;
4413 	uint64_t val;
4414 	const char *names[] = {
4415 		ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
4416 		ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
4417 		ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
4418 		ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
4419 		ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
4420 		ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
4421 		ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
4422 		ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
4423 		ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
4424 		ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
4425 	};
4426 	struct stat_array *nva;
4427 
4428 	unsigned int column_width = default_column_width(cb, IOS_LATENCY);
4429 	enum zfs_nicenum_format format;
4430 
4431 	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), oldnv, newnv);
4432 
4433 	if (cb->cb_literal)
4434 		format = ZFS_NICENUM_RAWTIME;
4435 	else
4436 		format = ZFS_NICENUM_TIME;
4437 
4438 	/* Print our avg latencies on the line */
4439 	for (i = 0; i < ARRAY_SIZE(names); i++) {
4440 		/* Compute average latency for a latency histo */
4441 		val = single_histo_average(nva[i].data, nva[i].count);
4442 		print_one_stat(val, format, column_width, cb->cb_scripted);
4443 	}
4444 	free_calc_stats(nva, ARRAY_SIZE(names));
4445 }
4446 
4447 /*
4448  * Print default statistics (capacity/operations/bandwidth)
4449  */
4450 static void
4451 print_iostat_default(vdev_stat_t *vs, iostat_cbdata_t *cb, double scale)
4452 {
4453 	unsigned int column_width = default_column_width(cb, IOS_DEFAULT);
4454 	enum zfs_nicenum_format format;
4455 	char na;	/* char to print for "not applicable" values */
4456 
4457 	if (cb->cb_literal) {
4458 		format = ZFS_NICENUM_RAW;
4459 		na = '0';
4460 	} else {
4461 		format = ZFS_NICENUM_1024;
4462 		na = '-';
4463 	}
4464 
4465 	/* only toplevel vdevs have capacity stats */
4466 	if (vs->vs_space == 0) {
4467 		if (cb->cb_scripted)
4468 			printf("\t%c\t%c", na, na);
4469 		else
4470 			printf("  %*c  %*c", column_width, na, column_width,
4471 			    na);
4472 	} else {
4473 		print_one_stat(vs->vs_alloc, format, column_width,
4474 		    cb->cb_scripted);
4475 		print_one_stat(vs->vs_space - vs->vs_alloc, format,
4476 		    column_width, cb->cb_scripted);
4477 	}
4478 
4479 	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_READ] * scale),
4480 	    format, column_width, cb->cb_scripted);
4481 	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_WRITE] * scale),
4482 	    format, column_width, cb->cb_scripted);
4483 	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_READ] * scale),
4484 	    format, column_width, cb->cb_scripted);
4485 	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_WRITE] * scale),
4486 	    format, column_width, cb->cb_scripted);
4487 }
4488 
4489 static const char *class_name[] = {
4490 	VDEV_ALLOC_BIAS_DEDUP,
4491 	VDEV_ALLOC_BIAS_SPECIAL,
4492 	VDEV_ALLOC_CLASS_LOGS
4493 };
4494 
4495 /*
4496  * Print out all the statistics for the given vdev.  This can either be the
4497  * toplevel configuration, or called recursively.  If 'name' is NULL, then this
4498  * is a verbose output, and we don't want to display the toplevel pool stats.
4499  *
4500  * Returns the number of stat lines printed.
4501  */
4502 static unsigned int
4503 print_vdev_stats(zpool_handle_t *zhp, const char *name, nvlist_t *oldnv,
4504     nvlist_t *newnv, iostat_cbdata_t *cb, int depth)
4505 {
4506 	nvlist_t **oldchild, **newchild;
4507 	uint_t c, children, oldchildren;
4508 	vdev_stat_t *oldvs, *newvs, *calcvs;
4509 	vdev_stat_t zerovs = { 0 };
4510 	char *vname;
4511 	int i;
4512 	int ret = 0;
4513 	uint64_t tdelta;
4514 	double scale;
4515 
4516 	if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
4517 		return (ret);
4518 
4519 	calcvs = safe_malloc(sizeof (*calcvs));
4520 
4521 	if (oldnv != NULL) {
4522 		verify(nvlist_lookup_uint64_array(oldnv,
4523 		    ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&oldvs, &c) == 0);
4524 	} else {
4525 		oldvs = &zerovs;
4526 	}
4527 
4528 	/* Do we only want to see a specific vdev? */
4529 	for (i = 0; i < cb->cb_vdev_names_count; i++) {
4530 		/* Yes we do.  Is this the vdev? */
4531 		if (strcmp(name, cb->cb_vdev_names[i]) == 0) {
4532 			/*
4533 			 * This is our vdev.  Since it is the only vdev we
4534 			 * will be displaying, make depth = 0 so that it
4535 			 * doesn't get indented.
4536 			 */
4537 			depth = 0;
4538 			break;
4539 		}
4540 	}
4541 
4542 	if (cb->cb_vdev_names_count && (i == cb->cb_vdev_names_count)) {
4543 		/* Couldn't match the name */
4544 		goto children;
4545 	}
4546 
4547 
4548 	verify(nvlist_lookup_uint64_array(newnv, ZPOOL_CONFIG_VDEV_STATS,
4549 	    (uint64_t **)&newvs, &c) == 0);
4550 
4551 	/*
4552 	 * Print the vdev name unless it's is a histogram.  Histograms
4553 	 * display the vdev name in the header itself.
4554 	 */
4555 	if (!(cb->cb_flags & IOS_ANYHISTO_M)) {
4556 		if (cb->cb_scripted) {
4557 			printf("%s", name);
4558 		} else {
4559 			if (strlen(name) + depth > cb->cb_namewidth)
4560 				(void) printf("%*s%s", depth, "", name);
4561 			else
4562 				(void) printf("%*s%s%*s", depth, "", name,
4563 				    (int)(cb->cb_namewidth - strlen(name) -
4564 				    depth), "");
4565 		}
4566 	}
4567 
4568 	/* Calculate our scaling factor */
4569 	tdelta = newvs->vs_timestamp - oldvs->vs_timestamp;
4570 	if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_ANYHISTO_M)) {
4571 		/*
4572 		 * If we specify printing histograms with no time interval, then
4573 		 * print the histogram numbers over the entire lifetime of the
4574 		 * vdev.
4575 		 */
4576 		scale = 1;
4577 	} else {
4578 		if (tdelta == 0)
4579 			scale = 1.0;
4580 		else
4581 			scale = (double)NANOSEC / tdelta;
4582 	}
4583 
4584 	if (cb->cb_flags & IOS_DEFAULT_M) {
4585 		calc_default_iostats(oldvs, newvs, calcvs);
4586 		print_iostat_default(calcvs, cb, scale);
4587 	}
4588 	if (cb->cb_flags & IOS_LATENCY_M)
4589 		print_iostat_latency(cb, oldnv, newnv);
4590 	if (cb->cb_flags & IOS_QUEUES_M)
4591 		print_iostat_queues(cb, oldnv, newnv);
4592 	if (cb->cb_flags & IOS_ANYHISTO_M) {
4593 		printf("\n");
4594 		print_iostat_histos(cb, oldnv, newnv, scale, name);
4595 	}
4596 
4597 	if (cb->vcdl != NULL) {
4598 		char *path;
4599 		if (nvlist_lookup_string(newnv, ZPOOL_CONFIG_PATH,
4600 		    &path) == 0) {
4601 			printf("  ");
4602 			zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
4603 		}
4604 	}
4605 
4606 	if (!(cb->cb_flags & IOS_ANYHISTO_M))
4607 		printf("\n");
4608 
4609 	ret++;
4610 
4611 children:
4612 
4613 	free(calcvs);
4614 
4615 	if (!cb->cb_verbose)
4616 		return (ret);
4617 
4618 	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_CHILDREN,
4619 	    &newchild, &children) != 0)
4620 		return (ret);
4621 
4622 	if (oldnv) {
4623 		if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_CHILDREN,
4624 		    &oldchild, &oldchildren) != 0)
4625 			return (ret);
4626 
4627 		children = MIN(oldchildren, children);
4628 	}
4629 
4630 	/*
4631 	 * print normal top-level devices
4632 	 */
4633 	for (c = 0; c < children; c++) {
4634 		uint64_t ishole = B_FALSE, islog = B_FALSE;
4635 
4636 		(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_HOLE,
4637 		    &ishole);
4638 
4639 		(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_LOG,
4640 		    &islog);
4641 
4642 		if (ishole || islog)
4643 			continue;
4644 
4645 		if (nvlist_exists(newchild[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
4646 			continue;
4647 
4648 		vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
4649 		    cb->cb_name_flags);
4650 		ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL,
4651 		    newchild[c], cb, depth + 2);
4652 		free(vname);
4653 	}
4654 
4655 	/*
4656 	 * print all other top-level devices
4657 	 */
4658 	for (uint_t n = 0; n < 3; n++) {
4659 		boolean_t printed = B_FALSE;
4660 
4661 		for (c = 0; c < children; c++) {
4662 			uint64_t islog = B_FALSE;
4663 			char *bias = NULL;
4664 			char *type = NULL;
4665 
4666 			(void) nvlist_lookup_uint64(newchild[c],
4667 			    ZPOOL_CONFIG_IS_LOG, &islog);
4668 			if (islog) {
4669 				bias = VDEV_ALLOC_CLASS_LOGS;
4670 			} else {
4671 				(void) nvlist_lookup_string(newchild[c],
4672 				    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
4673 				(void) nvlist_lookup_string(newchild[c],
4674 				    ZPOOL_CONFIG_TYPE, &type);
4675 			}
4676 			if (bias == NULL || strcmp(bias, class_name[n]) != 0)
4677 				continue;
4678 			if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
4679 				continue;
4680 
4681 			if (!printed) {
4682 				if ((!(cb->cb_flags & IOS_ANYHISTO_M)) &&
4683 				    !cb->cb_scripted && !cb->cb_vdev_names) {
4684 					print_iostat_dashes(cb, 0,
4685 					    class_name[n]);
4686 				}
4687 				printf("\n");
4688 				printed = B_TRUE;
4689 			}
4690 
4691 			vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
4692 			    cb->cb_name_flags);
4693 			ret += print_vdev_stats(zhp, vname, oldnv ?
4694 			    oldchild[c] : NULL, newchild[c], cb, depth + 2);
4695 			free(vname);
4696 		}
4697 	}
4698 
4699 	/*
4700 	 * Include level 2 ARC devices in iostat output
4701 	 */
4702 	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_L2CACHE,
4703 	    &newchild, &children) != 0)
4704 		return (ret);
4705 
4706 	if (oldnv) {
4707 		if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_L2CACHE,
4708 		    &oldchild, &oldchildren) != 0)
4709 			return (ret);
4710 
4711 		children = MIN(oldchildren, children);
4712 	}
4713 
4714 	if (children > 0) {
4715 		if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted &&
4716 		    !cb->cb_vdev_names) {
4717 			print_iostat_dashes(cb, 0, "cache");
4718 		}
4719 		printf("\n");
4720 
4721 		for (c = 0; c < children; c++) {
4722 			vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
4723 			    cb->cb_name_flags);
4724 			ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c]
4725 			    : NULL, newchild[c], cb, depth + 2);
4726 			free(vname);
4727 		}
4728 	}
4729 
4730 	return (ret);
4731 }
4732 
4733 static int
4734 refresh_iostat(zpool_handle_t *zhp, void *data)
4735 {
4736 	iostat_cbdata_t *cb = data;
4737 	boolean_t missing;
4738 
4739 	/*
4740 	 * If the pool has disappeared, remove it from the list and continue.
4741 	 */
4742 	if (zpool_refresh_stats(zhp, &missing) != 0)
4743 		return (-1);
4744 
4745 	if (missing)
4746 		pool_list_remove(cb->cb_list, zhp);
4747 
4748 	return (0);
4749 }
4750 
4751 /*
4752  * Callback to print out the iostats for the given pool.
4753  */
4754 static int
4755 print_iostat(zpool_handle_t *zhp, void *data)
4756 {
4757 	iostat_cbdata_t *cb = data;
4758 	nvlist_t *oldconfig, *newconfig;
4759 	nvlist_t *oldnvroot, *newnvroot;
4760 	int ret;
4761 
4762 	newconfig = zpool_get_config(zhp, &oldconfig);
4763 
4764 	if (cb->cb_iteration == 1)
4765 		oldconfig = NULL;
4766 
4767 	verify(nvlist_lookup_nvlist(newconfig, ZPOOL_CONFIG_VDEV_TREE,
4768 	    &newnvroot) == 0);
4769 
4770 	if (oldconfig == NULL)
4771 		oldnvroot = NULL;
4772 	else
4773 		verify(nvlist_lookup_nvlist(oldconfig, ZPOOL_CONFIG_VDEV_TREE,
4774 		    &oldnvroot) == 0);
4775 
4776 	ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot,
4777 	    cb, 0);
4778 	if ((ret != 0) && !(cb->cb_flags & IOS_ANYHISTO_M) &&
4779 	    !cb->cb_scripted && cb->cb_verbose && !cb->cb_vdev_names_count) {
4780 		print_iostat_separator(cb);
4781 		if (cb->vcdl != NULL) {
4782 			print_cmd_columns(cb->vcdl, 1);
4783 		}
4784 		printf("\n");
4785 	}
4786 
4787 	return (ret);
4788 }
4789 
4790 static int
4791 get_columns(void)
4792 {
4793 	struct winsize ws;
4794 	int columns = 80;
4795 	int error;
4796 
4797 	if (isatty(STDOUT_FILENO)) {
4798 		error = ioctl(STDOUT_FILENO, TIOCGWINSZ, &ws);
4799 		if (error == 0)
4800 			columns = ws.ws_col;
4801 	} else {
4802 		columns = 999;
4803 	}
4804 
4805 	return (columns);
4806 }
4807 
4808 /*
4809  * Return the required length of the pool/vdev name column.  The minimum
4810  * allowed width and output formatting flags must be provided.
4811  */
4812 static int
4813 get_namewidth(zpool_handle_t *zhp, int min_width, int flags, boolean_t verbose)
4814 {
4815 	nvlist_t *config, *nvroot;
4816 	int width = min_width;
4817 
4818 	if ((config = zpool_get_config(zhp, NULL)) != NULL) {
4819 		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4820 		    &nvroot) == 0);
4821 		unsigned int poolname_len = strlen(zpool_get_name(zhp));
4822 		if (verbose == B_FALSE) {
4823 			width = MAX(poolname_len, min_width);
4824 		} else {
4825 			width = MAX(poolname_len,
4826 			    max_width(zhp, nvroot, 0, min_width, flags));
4827 		}
4828 	}
4829 
4830 	return (width);
4831 }
4832 
4833 /*
4834  * Parse the input string, get the 'interval' and 'count' value if there is one.
4835  */
4836 static void
4837 get_interval_count(int *argcp, char **argv, float *iv,
4838     unsigned long *cnt)
4839 {
4840 	float interval = 0;
4841 	unsigned long count = 0;
4842 	int argc = *argcp;
4843 
4844 	/*
4845 	 * Determine if the last argument is an integer or a pool name
4846 	 */
4847 	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
4848 		char *end;
4849 
4850 		errno = 0;
4851 		interval = strtof(argv[argc - 1], &end);
4852 
4853 		if (*end == '\0' && errno == 0) {
4854 			if (interval == 0) {
4855 				(void) fprintf(stderr, gettext("interval "
4856 				    "cannot be zero\n"));
4857 				usage(B_FALSE);
4858 			}
4859 			/*
4860 			 * Ignore the last parameter
4861 			 */
4862 			argc--;
4863 		} else {
4864 			/*
4865 			 * If this is not a valid number, just plow on.  The
4866 			 * user will get a more informative error message later
4867 			 * on.
4868 			 */
4869 			interval = 0;
4870 		}
4871 	}
4872 
4873 	/*
4874 	 * If the last argument is also an integer, then we have both a count
4875 	 * and an interval.
4876 	 */
4877 	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
4878 		char *end;
4879 
4880 		errno = 0;
4881 		count = interval;
4882 		interval = strtof(argv[argc - 1], &end);
4883 
4884 		if (*end == '\0' && errno == 0) {
4885 			if (interval == 0) {
4886 				(void) fprintf(stderr, gettext("interval "
4887 				    "cannot be zero\n"));
4888 				usage(B_FALSE);
4889 			}
4890 
4891 			/*
4892 			 * Ignore the last parameter
4893 			 */
4894 			argc--;
4895 		} else {
4896 			interval = 0;
4897 		}
4898 	}
4899 
4900 	*iv = interval;
4901 	*cnt = count;
4902 	*argcp = argc;
4903 }
4904 
4905 static void
4906 get_timestamp_arg(char c)
4907 {
4908 	if (c == 'u')
4909 		timestamp_fmt = UDATE;
4910 	else if (c == 'd')
4911 		timestamp_fmt = DDATE;
4912 	else
4913 		usage(B_FALSE);
4914 }
4915 
4916 /*
4917  * Return stat flags that are supported by all pools by both the module and
4918  * zpool iostat.  "*data" should be initialized to all 0xFFs before running.
4919  * It will get ANDed down until only the flags that are supported on all pools
4920  * remain.
4921  */
4922 static int
4923 get_stat_flags_cb(zpool_handle_t *zhp, void *data)
4924 {
4925 	uint64_t *mask = data;
4926 	nvlist_t *config, *nvroot, *nvx;
4927 	uint64_t flags = 0;
4928 	int i, j;
4929 
4930 	config = zpool_get_config(zhp, NULL);
4931 	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4932 	    &nvroot) == 0);
4933 
4934 	/* Default stats are always supported, but for completeness.. */
4935 	if (nvlist_exists(nvroot, ZPOOL_CONFIG_VDEV_STATS))
4936 		flags |= IOS_DEFAULT_M;
4937 
4938 	/* Get our extended stats nvlist from the main list */
4939 	if (nvlist_lookup_nvlist(nvroot, ZPOOL_CONFIG_VDEV_STATS_EX,
4940 	    &nvx) != 0) {
4941 		/*
4942 		 * No extended stats; they're probably running an older
4943 		 * module.  No big deal, we support that too.
4944 		 */
4945 		goto end;
4946 	}
4947 
4948 	/* For each extended stat, make sure all its nvpairs are supported */
4949 	for (j = 0; j < ARRAY_SIZE(vsx_type_to_nvlist); j++) {
4950 		if (!vsx_type_to_nvlist[j][0])
4951 			continue;
4952 
4953 		/* Start off by assuming the flag is supported, then check */
4954 		flags |= (1ULL << j);
4955 		for (i = 0; vsx_type_to_nvlist[j][i]; i++) {
4956 			if (!nvlist_exists(nvx, vsx_type_to_nvlist[j][i])) {
4957 				/* flag isn't supported */
4958 				flags = flags & ~(1ULL  << j);
4959 				break;
4960 			}
4961 		}
4962 	}
4963 end:
4964 	*mask = *mask & flags;
4965 	return (0);
4966 }
4967 
4968 /*
4969  * Return a bitmask of stats that are supported on all pools by both the module
4970  * and zpool iostat.
4971  */
4972 static uint64_t
4973 get_stat_flags(zpool_list_t *list)
4974 {
4975 	uint64_t mask = -1;
4976 
4977 	/*
4978 	 * get_stat_flags_cb() will lop off bits from "mask" until only the
4979 	 * flags that are supported on all pools remain.
4980 	 */
4981 	pool_list_iter(list, B_FALSE, get_stat_flags_cb, &mask);
4982 	return (mask);
4983 }
4984 
4985 /*
4986  * Return 1 if cb_data->cb_vdev_names[0] is this vdev's name, 0 otherwise.
4987  */
4988 static int
4989 is_vdev_cb(zpool_handle_t *zhp, nvlist_t *nv, void *cb_data)
4990 {
4991 	iostat_cbdata_t *cb = cb_data;
4992 	char *name = NULL;
4993 	int ret = 0;
4994 
4995 	name = zpool_vdev_name(g_zfs, zhp, nv, cb->cb_name_flags);
4996 
4997 	if (strcmp(name, cb->cb_vdev_names[0]) == 0)
4998 		ret = 1; /* match */
4999 	free(name);
5000 
5001 	return (ret);
5002 }
5003 
5004 /*
5005  * Returns 1 if cb_data->cb_vdev_names[0] is a vdev name, 0 otherwise.
5006  */
5007 static int
5008 is_vdev(zpool_handle_t *zhp, void *cb_data)
5009 {
5010 	return (for_each_vdev(zhp, is_vdev_cb, cb_data));
5011 }
5012 
5013 /*
5014  * Check if vdevs are in a pool
5015  *
5016  * Return 1 if all argv[] strings are vdev names in pool "pool_name". Otherwise
5017  * return 0.  If pool_name is NULL, then search all pools.
5018  */
5019 static int
5020 are_vdevs_in_pool(int argc, char **argv, char *pool_name,
5021     iostat_cbdata_t *cb)
5022 {
5023 	char **tmp_name;
5024 	int ret = 0;
5025 	int i;
5026 	int pool_count = 0;
5027 
5028 	if ((argc == 0) || !*argv)
5029 		return (0);
5030 
5031 	if (pool_name)
5032 		pool_count = 1;
5033 
5034 	/* Temporarily hijack cb_vdev_names for a second... */
5035 	tmp_name = cb->cb_vdev_names;
5036 
5037 	/* Go though our list of prospective vdev names */
5038 	for (i = 0; i < argc; i++) {
5039 		cb->cb_vdev_names = argv + i;
5040 
5041 		/* Is this name a vdev in our pools? */
5042 		ret = for_each_pool(pool_count, &pool_name, B_TRUE, NULL,
5043 		    B_FALSE, is_vdev, cb);
5044 		if (!ret) {
5045 			/* No match */
5046 			break;
5047 		}
5048 	}
5049 
5050 	cb->cb_vdev_names = tmp_name;
5051 
5052 	return (ret);
5053 }
5054 
5055 static int
5056 is_pool_cb(zpool_handle_t *zhp, void *data)
5057 {
5058 	char *name = data;
5059 	if (strcmp(name, zpool_get_name(zhp)) == 0)
5060 		return (1);
5061 
5062 	return (0);
5063 }
5064 
5065 /*
5066  * Do we have a pool named *name?  If so, return 1, otherwise 0.
5067  */
5068 static int
5069 is_pool(char *name)
5070 {
5071 	return (for_each_pool(0, NULL, B_TRUE, NULL, B_FALSE, is_pool_cb,
5072 	    name));
5073 }
5074 
5075 /* Are all our argv[] strings pool names?  If so return 1, 0 otherwise. */
5076 static int
5077 are_all_pools(int argc, char **argv)
5078 {
5079 	if ((argc == 0) || !*argv)
5080 		return (0);
5081 
5082 	while (--argc >= 0)
5083 		if (!is_pool(argv[argc]))
5084 			return (0);
5085 
5086 	return (1);
5087 }
5088 
5089 /*
5090  * Helper function to print out vdev/pool names we can't resolve.  Used for an
5091  * error message.
5092  */
5093 static void
5094 error_list_unresolved_vdevs(int argc, char **argv, char *pool_name,
5095     iostat_cbdata_t *cb)
5096 {
5097 	int i;
5098 	char *name;
5099 	char *str;
5100 	for (i = 0; i < argc; i++) {
5101 		name = argv[i];
5102 
5103 		if (is_pool(name))
5104 			str = gettext("pool");
5105 		else if (are_vdevs_in_pool(1, &name, pool_name, cb))
5106 			str = gettext("vdev in this pool");
5107 		else if (are_vdevs_in_pool(1, &name, NULL, cb))
5108 			str = gettext("vdev in another pool");
5109 		else
5110 			str = gettext("unknown");
5111 
5112 		fprintf(stderr, "\t%s (%s)\n", name, str);
5113 	}
5114 }
5115 
5116 /*
5117  * Same as get_interval_count(), but with additional checks to not misinterpret
5118  * guids as interval/count values.  Assumes VDEV_NAME_GUID is set in
5119  * cb.cb_name_flags.
5120  */
5121 static void
5122 get_interval_count_filter_guids(int *argc, char **argv, float *interval,
5123     unsigned long *count, iostat_cbdata_t *cb)
5124 {
5125 	char **tmpargv = argv;
5126 	int argc_for_interval = 0;
5127 
5128 	/* Is the last arg an interval value?  Or a guid? */
5129 	if (*argc >= 1 && !are_vdevs_in_pool(1, &argv[*argc - 1], NULL, cb)) {
5130 		/*
5131 		 * The last arg is not a guid, so it's probably an
5132 		 * interval value.
5133 		 */
5134 		argc_for_interval++;
5135 
5136 		if (*argc >= 2 &&
5137 		    !are_vdevs_in_pool(1, &argv[*argc - 2], NULL, cb)) {
5138 			/*
5139 			 * The 2nd to last arg is not a guid, so it's probably
5140 			 * an interval value.
5141 			 */
5142 			argc_for_interval++;
5143 		}
5144 	}
5145 
5146 	/* Point to our list of possible intervals */
5147 	tmpargv = &argv[*argc - argc_for_interval];
5148 
5149 	*argc = *argc - argc_for_interval;
5150 	get_interval_count(&argc_for_interval, tmpargv,
5151 	    interval, count);
5152 }
5153 
5154 /*
5155  * Floating point sleep().  Allows you to pass in a floating point value for
5156  * seconds.
5157  */
5158 static void
5159 fsleep(float sec)
5160 {
5161 	struct timespec req;
5162 	req.tv_sec = floor(sec);
5163 	req.tv_nsec = (sec - (float)req.tv_sec) * NANOSEC;
5164 	nanosleep(&req, NULL);
5165 }
5166 
5167 /*
5168  * Terminal height, in rows. Returns -1 if stdout is not connected to a TTY or
5169  * if we were unable to determine its size.
5170  */
5171 static int
5172 terminal_height(void)
5173 {
5174 	struct winsize win;
5175 
5176 	if (isatty(STDOUT_FILENO) == 0)
5177 		return (-1);
5178 
5179 	if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) != -1 && win.ws_row > 0)
5180 		return (win.ws_row);
5181 
5182 	return (-1);
5183 }
5184 
5185 /*
5186  * Run one of the zpool status/iostat -c scripts with the help (-h) option and
5187  * print the result.
5188  *
5189  * name:	Short name of the script ('iostat').
5190  * path:	Full path to the script ('/usr/local/etc/zfs/zpool.d/iostat');
5191  */
5192 static void
5193 print_zpool_script_help(char *name, char *path)
5194 {
5195 	char *argv[] = {path, "-h", NULL};
5196 	char **lines = NULL;
5197 	int lines_cnt = 0;
5198 	int rc;
5199 
5200 	rc = libzfs_run_process_get_stdout_nopath(path, argv, NULL, &lines,
5201 	    &lines_cnt);
5202 	if (rc != 0 || lines == NULL || lines_cnt <= 0) {
5203 		if (lines != NULL)
5204 			libzfs_free_str_array(lines, lines_cnt);
5205 		return;
5206 	}
5207 
5208 	for (int i = 0; i < lines_cnt; i++)
5209 		if (!is_blank_str(lines[i]))
5210 			printf("  %-14s  %s\n", name, lines[i]);
5211 
5212 	libzfs_free_str_array(lines, lines_cnt);
5213 }
5214 
5215 /*
5216  * Go though the zpool status/iostat -c scripts in the user's path, run their
5217  * help option (-h), and print out the results.
5218  */
5219 static void
5220 print_zpool_dir_scripts(char *dirpath)
5221 {
5222 	DIR *dir;
5223 	struct dirent *ent;
5224 	char fullpath[MAXPATHLEN];
5225 	struct stat dir_stat;
5226 
5227 	if ((dir = opendir(dirpath)) != NULL) {
5228 		/* print all the files and directories within directory */
5229 		while ((ent = readdir(dir)) != NULL) {
5230 			sprintf(fullpath, "%s/%s", dirpath, ent->d_name);
5231 
5232 			/* Print the scripts */
5233 			if (stat(fullpath, &dir_stat) == 0)
5234 				if (dir_stat.st_mode & S_IXUSR &&
5235 				    S_ISREG(dir_stat.st_mode))
5236 					print_zpool_script_help(ent->d_name,
5237 					    fullpath);
5238 		}
5239 		closedir(dir);
5240 	}
5241 }
5242 
5243 /*
5244  * Print out help text for all zpool status/iostat -c scripts.
5245  */
5246 static void
5247 print_zpool_script_list(char *subcommand)
5248 {
5249 	char *dir, *sp;
5250 
5251 	printf(gettext("Available 'zpool %s -c' commands:\n"), subcommand);
5252 
5253 	sp = zpool_get_cmd_search_path();
5254 	if (sp == NULL)
5255 		return;
5256 
5257 	dir = strtok(sp, ":");
5258 	while (dir != NULL) {
5259 		print_zpool_dir_scripts(dir);
5260 		dir = strtok(NULL, ":");
5261 	}
5262 
5263 	free(sp);
5264 }
5265 
5266 /*
5267  * Set the minimum pool/vdev name column width.  The width must be at least 10,
5268  * but may be as large as the column width - 42 so it still fits on one line.
5269  * NOTE: 42 is the width of the default capacity/operations/bandwidth output
5270  */
5271 static int
5272 get_namewidth_iostat(zpool_handle_t *zhp, void *data)
5273 {
5274 	iostat_cbdata_t *cb = data;
5275 	int width, available_width;
5276 
5277 	/*
5278 	 * get_namewidth() returns the maximum width of any name in that column
5279 	 * for any pool/vdev/device line that will be output.
5280 	 */
5281 	width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags,
5282 	    cb->cb_verbose);
5283 
5284 	/*
5285 	 * The width we are calculating is the width of the header and also the
5286 	 * padding width for names that are less than maximum width.  The stats
5287 	 * take up 42 characters, so the width available for names is:
5288 	 */
5289 	available_width = get_columns() - 42;
5290 
5291 	/*
5292 	 * If the maximum width fits on a screen, then great!  Make everything
5293 	 * line up by justifying all lines to the same width.  If that max
5294 	 * width is larger than what's available, the name plus stats won't fit
5295 	 * on one line, and justifying to that width would cause every line to
5296 	 * wrap on the screen.  We only want lines with long names to wrap.
5297 	 * Limit the padding to what won't wrap.
5298 	 */
5299 	if (width > available_width)
5300 		width = available_width;
5301 
5302 	/*
5303 	 * And regardless of whatever the screen width is (get_columns can
5304 	 * return 0 if the width is not known or less than 42 for a narrow
5305 	 * terminal) have the width be a minimum of 10.
5306 	 */
5307 	if (width < 10)
5308 		width = 10;
5309 
5310 	/* Save the calculated width */
5311 	cb->cb_namewidth = width;
5312 
5313 	return (0);
5314 }
5315 
5316 /*
5317  * zpool iostat [[-c [script1,script2,...]] [-lq]|[-rw]] [-ghHLpPvy] [-n name]
5318  *              [-T d|u] [[ pool ...]|[pool vdev ...]|[vdev ...]]
5319  *              [interval [count]]
5320  *
5321  *	-c CMD  For each vdev, run command CMD
5322  *	-g	Display guid for individual vdev name.
5323  *	-L	Follow links when resolving vdev path name.
5324  *	-P	Display full path for vdev name.
5325  *	-v	Display statistics for individual vdevs
5326  *	-h	Display help
5327  *	-p	Display values in parsable (exact) format.
5328  *	-H	Scripted mode.  Don't display headers, and separate properties
5329  *		by a single tab.
5330  *	-l	Display average latency
5331  *	-q	Display queue depths
5332  *	-w	Display latency histograms
5333  *	-r	Display request size histogram
5334  *	-T	Display a timestamp in date(1) or Unix format
5335  *	-n	Only print headers once
5336  *
5337  * This command can be tricky because we want to be able to deal with pool
5338  * creation/destruction as well as vdev configuration changes.  The bulk of this
5339  * processing is handled by the pool_list_* routines in zpool_iter.c.  We rely
5340  * on pool_list_update() to detect the addition of new pools.  Configuration
5341  * changes are all handled within libzfs.
5342  */
5343 int
5344 zpool_do_iostat(int argc, char **argv)
5345 {
5346 	int c;
5347 	int ret;
5348 	int npools;
5349 	float interval = 0;
5350 	unsigned long count = 0;
5351 	int winheight = 24;
5352 	zpool_list_t *list;
5353 	boolean_t verbose = B_FALSE;
5354 	boolean_t latency = B_FALSE, l_histo = B_FALSE, rq_histo = B_FALSE;
5355 	boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE;
5356 	boolean_t omit_since_boot = B_FALSE;
5357 	boolean_t guid = B_FALSE;
5358 	boolean_t follow_links = B_FALSE;
5359 	boolean_t full_name = B_FALSE;
5360 	boolean_t headers_once = B_FALSE;
5361 	iostat_cbdata_t cb = { 0 };
5362 	char *cmd = NULL;
5363 
5364 	/* Used for printing error message */
5365 	const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q',
5366 	    [IOS_L_HISTO] = 'w', [IOS_RQ_HISTO] = 'r'};
5367 
5368 	uint64_t unsupported_flags;
5369 
5370 	/* check options */
5371 	while ((c = getopt(argc, argv, "c:gLPT:vyhplqrwnH")) != -1) {
5372 		switch (c) {
5373 		case 'c':
5374 			if (cmd != NULL) {
5375 				fprintf(stderr,
5376 				    gettext("Can't set -c flag twice\n"));
5377 				exit(1);
5378 			}
5379 
5380 			if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
5381 			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
5382 				fprintf(stderr, gettext(
5383 				    "Can't run -c, disabled by "
5384 				    "ZPOOL_SCRIPTS_ENABLED.\n"));
5385 				exit(1);
5386 			}
5387 
5388 			if ((getuid() <= 0 || geteuid() <= 0) &&
5389 			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
5390 				fprintf(stderr, gettext(
5391 				    "Can't run -c with root privileges "
5392 				    "unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
5393 				exit(1);
5394 			}
5395 			cmd = optarg;
5396 			verbose = B_TRUE;
5397 			break;
5398 		case 'g':
5399 			guid = B_TRUE;
5400 			break;
5401 		case 'L':
5402 			follow_links = B_TRUE;
5403 			break;
5404 		case 'P':
5405 			full_name = B_TRUE;
5406 			break;
5407 		case 'T':
5408 			get_timestamp_arg(*optarg);
5409 			break;
5410 		case 'v':
5411 			verbose = B_TRUE;
5412 			break;
5413 		case 'p':
5414 			parsable = B_TRUE;
5415 			break;
5416 		case 'l':
5417 			latency = B_TRUE;
5418 			break;
5419 		case 'q':
5420 			queues = B_TRUE;
5421 			break;
5422 		case 'H':
5423 			scripted = B_TRUE;
5424 			break;
5425 		case 'w':
5426 			l_histo = B_TRUE;
5427 			break;
5428 		case 'r':
5429 			rq_histo = B_TRUE;
5430 			break;
5431 		case 'y':
5432 			omit_since_boot = B_TRUE;
5433 			break;
5434 		case 'n':
5435 			headers_once = B_TRUE;
5436 			break;
5437 		case 'h':
5438 			usage(B_FALSE);
5439 			break;
5440 		case '?':
5441 			if (optopt == 'c') {
5442 				print_zpool_script_list("iostat");
5443 				exit(0);
5444 			} else {
5445 				fprintf(stderr,
5446 				    gettext("invalid option '%c'\n"), optopt);
5447 			}
5448 			usage(B_FALSE);
5449 		}
5450 	}
5451 
5452 	argc -= optind;
5453 	argv += optind;
5454 
5455 	cb.cb_literal = parsable;
5456 	cb.cb_scripted = scripted;
5457 
5458 	if (guid)
5459 		cb.cb_name_flags |= VDEV_NAME_GUID;
5460 	if (follow_links)
5461 		cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS;
5462 	if (full_name)
5463 		cb.cb_name_flags |= VDEV_NAME_PATH;
5464 	cb.cb_iteration = 0;
5465 	cb.cb_namewidth = 0;
5466 	cb.cb_verbose = verbose;
5467 
5468 	/* Get our interval and count values (if any) */
5469 	if (guid) {
5470 		get_interval_count_filter_guids(&argc, argv, &interval,
5471 		    &count, &cb);
5472 	} else {
5473 		get_interval_count(&argc, argv, &interval, &count);
5474 	}
5475 
5476 	if (argc == 0) {
5477 		/* No args, so just print the defaults. */
5478 	} else if (are_all_pools(argc, argv)) {
5479 		/* All the args are pool names */
5480 	} else if (are_vdevs_in_pool(argc, argv, NULL, &cb)) {
5481 		/* All the args are vdevs */
5482 		cb.cb_vdev_names = argv;
5483 		cb.cb_vdev_names_count = argc;
5484 		argc = 0; /* No pools to process */
5485 	} else if (are_all_pools(1, argv)) {
5486 		/* The first arg is a pool name */
5487 		if (are_vdevs_in_pool(argc - 1, argv + 1, argv[0], &cb)) {
5488 			/* ...and the rest are vdev names */
5489 			cb.cb_vdev_names = argv + 1;
5490 			cb.cb_vdev_names_count = argc - 1;
5491 			argc = 1; /* One pool to process */
5492 		} else {
5493 			fprintf(stderr, gettext("Expected either a list of "));
5494 			fprintf(stderr, gettext("pools, or list of vdevs in"));
5495 			fprintf(stderr, " \"%s\", ", argv[0]);
5496 			fprintf(stderr, gettext("but got:\n"));
5497 			error_list_unresolved_vdevs(argc - 1, argv + 1,
5498 			    argv[0], &cb);
5499 			fprintf(stderr, "\n");
5500 			usage(B_FALSE);
5501 			return (1);
5502 		}
5503 	} else {
5504 		/*
5505 		 * The args don't make sense. The first arg isn't a pool name,
5506 		 * nor are all the args vdevs.
5507 		 */
5508 		fprintf(stderr, gettext("Unable to parse pools/vdevs list.\n"));
5509 		fprintf(stderr, "\n");
5510 		return (1);
5511 	}
5512 
5513 	if (cb.cb_vdev_names_count != 0) {
5514 		/*
5515 		 * If user specified vdevs, it implies verbose.
5516 		 */
5517 		cb.cb_verbose = B_TRUE;
5518 	}
5519 
5520 	/*
5521 	 * Construct the list of all interesting pools.
5522 	 */
5523 	ret = 0;
5524 	if ((list = pool_list_get(argc, argv, NULL, parsable, &ret)) == NULL)
5525 		return (1);
5526 
5527 	if (pool_list_count(list) == 0 && argc != 0) {
5528 		pool_list_free(list);
5529 		return (1);
5530 	}
5531 
5532 	if (pool_list_count(list) == 0 && interval == 0) {
5533 		pool_list_free(list);
5534 		(void) fprintf(stderr, gettext("no pools available\n"));
5535 		return (1);
5536 	}
5537 
5538 	if ((l_histo || rq_histo) && (cmd != NULL || latency || queues)) {
5539 		pool_list_free(list);
5540 		(void) fprintf(stderr,
5541 		    gettext("[-r|-w] isn't allowed with [-c|-l|-q]\n"));
5542 		usage(B_FALSE);
5543 		return (1);
5544 	}
5545 
5546 	if (l_histo && rq_histo) {
5547 		pool_list_free(list);
5548 		(void) fprintf(stderr,
5549 		    gettext("Only one of [-r|-w] can be passed at a time\n"));
5550 		usage(B_FALSE);
5551 		return (1);
5552 	}
5553 
5554 	/*
5555 	 * Enter the main iostat loop.
5556 	 */
5557 	cb.cb_list = list;
5558 
5559 	if (l_histo) {
5560 		/*
5561 		 * Histograms tables look out of place when you try to display
5562 		 * them with the other stats, so make a rule that you can only
5563 		 * print histograms by themselves.
5564 		 */
5565 		cb.cb_flags = IOS_L_HISTO_M;
5566 	} else if (rq_histo) {
5567 		cb.cb_flags = IOS_RQ_HISTO_M;
5568 	} else {
5569 		cb.cb_flags = IOS_DEFAULT_M;
5570 		if (latency)
5571 			cb.cb_flags |= IOS_LATENCY_M;
5572 		if (queues)
5573 			cb.cb_flags |= IOS_QUEUES_M;
5574 	}
5575 
5576 	/*
5577 	 * See if the module supports all the stats we want to display.
5578 	 */
5579 	unsupported_flags = cb.cb_flags & ~get_stat_flags(list);
5580 	if (unsupported_flags) {
5581 		uint64_t f;
5582 		int idx;
5583 		fprintf(stderr,
5584 		    gettext("The loaded zfs module doesn't support:"));
5585 
5586 		/* for each bit set in unsupported_flags */
5587 		for (f = unsupported_flags; f; f &= ~(1ULL << idx)) {
5588 			idx = lowbit64(f) - 1;
5589 			fprintf(stderr, " -%c", flag_to_arg[idx]);
5590 		}
5591 
5592 		fprintf(stderr, ".  Try running a newer module.\n");
5593 		pool_list_free(list);
5594 
5595 		return (1);
5596 	}
5597 
5598 	for (;;) {
5599 		if ((npools = pool_list_count(list)) == 0)
5600 			(void) fprintf(stderr, gettext("no pools available\n"));
5601 		else {
5602 			/*
5603 			 * If this is the first iteration and -y was supplied
5604 			 * we skip any printing.
5605 			 */
5606 			boolean_t skip = (omit_since_boot &&
5607 			    cb.cb_iteration == 0);
5608 
5609 			/*
5610 			 * Refresh all statistics.  This is done as an
5611 			 * explicit step before calculating the maximum name
5612 			 * width, so that any * configuration changes are
5613 			 * properly accounted for.
5614 			 */
5615 			(void) pool_list_iter(list, B_FALSE, refresh_iostat,
5616 			    &cb);
5617 
5618 			/*
5619 			 * Iterate over all pools to determine the maximum width
5620 			 * for the pool / device name column across all pools.
5621 			 */
5622 			cb.cb_namewidth = 0;
5623 			(void) pool_list_iter(list, B_FALSE,
5624 			    get_namewidth_iostat, &cb);
5625 
5626 			if (timestamp_fmt != NODATE)
5627 				print_timestamp(timestamp_fmt);
5628 
5629 			if (cmd != NULL && cb.cb_verbose &&
5630 			    !(cb.cb_flags & IOS_ANYHISTO_M)) {
5631 				cb.vcdl = all_pools_for_each_vdev_run(argc,
5632 				    argv, cmd, g_zfs, cb.cb_vdev_names,
5633 				    cb.cb_vdev_names_count, cb.cb_name_flags);
5634 			} else {
5635 				cb.vcdl = NULL;
5636 			}
5637 
5638 
5639 			/*
5640 			 * Check terminal size so we can print headers
5641 			 * even when terminal window has its height
5642 			 * changed.
5643 			 */
5644 			winheight = terminal_height();
5645 			/*
5646 			 * Are we connected to TTY? If not, headers_once
5647 			 * should be true, to avoid breaking scripts.
5648 			 */
5649 			if (winheight < 0)
5650 				headers_once = B_TRUE;
5651 
5652 			/*
5653 			 * If it's the first time and we're not skipping it,
5654 			 * or either skip or verbose mode, print the header.
5655 			 *
5656 			 * The histogram code explicitly prints its header on
5657 			 * every vdev, so skip this for histograms.
5658 			 */
5659 			if (((++cb.cb_iteration == 1 && !skip) ||
5660 			    (skip != verbose) ||
5661 			    (!headers_once &&
5662 			    (cb.cb_iteration % winheight) == 0)) &&
5663 			    (!(cb.cb_flags & IOS_ANYHISTO_M)) &&
5664 			    !cb.cb_scripted)
5665 				print_iostat_header(&cb);
5666 
5667 			if (skip) {
5668 				(void) fsleep(interval);
5669 				continue;
5670 			}
5671 
5672 			pool_list_iter(list, B_FALSE, print_iostat, &cb);
5673 
5674 			/*
5675 			 * If there's more than one pool, and we're not in
5676 			 * verbose mode (which prints a separator for us),
5677 			 * then print a separator.
5678 			 *
5679 			 * In addition, if we're printing specific vdevs then
5680 			 * we also want an ending separator.
5681 			 */
5682 			if (((npools > 1 && !verbose &&
5683 			    !(cb.cb_flags & IOS_ANYHISTO_M)) ||
5684 			    (!(cb.cb_flags & IOS_ANYHISTO_M) &&
5685 			    cb.cb_vdev_names_count)) &&
5686 			    !cb.cb_scripted) {
5687 				print_iostat_separator(&cb);
5688 				if (cb.vcdl != NULL)
5689 					print_cmd_columns(cb.vcdl, 1);
5690 				printf("\n");
5691 			}
5692 
5693 			if (cb.vcdl != NULL)
5694 				free_vdev_cmd_data_list(cb.vcdl);
5695 
5696 		}
5697 
5698 		/*
5699 		 * Flush the output so that redirection to a file isn't buffered
5700 		 * indefinitely.
5701 		 */
5702 		(void) fflush(stdout);
5703 
5704 		if (interval == 0)
5705 			break;
5706 
5707 		if (count != 0 && --count == 0)
5708 			break;
5709 
5710 		(void) fsleep(interval);
5711 	}
5712 
5713 	pool_list_free(list);
5714 
5715 	return (ret);
5716 }
5717 
5718 typedef struct list_cbdata {
5719 	boolean_t	cb_verbose;
5720 	int		cb_name_flags;
5721 	int		cb_namewidth;
5722 	boolean_t	cb_scripted;
5723 	zprop_list_t	*cb_proplist;
5724 	boolean_t	cb_literal;
5725 } list_cbdata_t;
5726 
5727 
5728 /*
5729  * Given a list of columns to display, output appropriate headers for each one.
5730  */
5731 static void
5732 print_header(list_cbdata_t *cb)
5733 {
5734 	zprop_list_t *pl = cb->cb_proplist;
5735 	char headerbuf[ZPOOL_MAXPROPLEN];
5736 	const char *header;
5737 	boolean_t first = B_TRUE;
5738 	boolean_t right_justify;
5739 	size_t width = 0;
5740 
5741 	for (; pl != NULL; pl = pl->pl_next) {
5742 		width = pl->pl_width;
5743 		if (first && cb->cb_verbose) {
5744 			/*
5745 			 * Reset the width to accommodate the verbose listing
5746 			 * of devices.
5747 			 */
5748 			width = cb->cb_namewidth;
5749 		}
5750 
5751 		if (!first)
5752 			(void) printf("  ");
5753 		else
5754 			first = B_FALSE;
5755 
5756 		right_justify = B_FALSE;
5757 		if (pl->pl_prop != ZPROP_INVAL) {
5758 			header = zpool_prop_column_name(pl->pl_prop);
5759 			right_justify = zpool_prop_align_right(pl->pl_prop);
5760 		} else {
5761 			int i;
5762 
5763 			for (i = 0; pl->pl_user_prop[i] != '\0'; i++)
5764 				headerbuf[i] = toupper(pl->pl_user_prop[i]);
5765 			headerbuf[i] = '\0';
5766 			header = headerbuf;
5767 		}
5768 
5769 		if (pl->pl_next == NULL && !right_justify)
5770 			(void) printf("%s", header);
5771 		else if (right_justify)
5772 			(void) printf("%*s", (int)width, header);
5773 		else
5774 			(void) printf("%-*s", (int)width, header);
5775 	}
5776 
5777 	(void) printf("\n");
5778 }
5779 
5780 /*
5781  * Given a pool and a list of properties, print out all the properties according
5782  * to the described layout. Used by zpool_do_list().
5783  */
5784 static void
5785 print_pool(zpool_handle_t *zhp, list_cbdata_t *cb)
5786 {
5787 	zprop_list_t *pl = cb->cb_proplist;
5788 	boolean_t first = B_TRUE;
5789 	char property[ZPOOL_MAXPROPLEN];
5790 	char *propstr;
5791 	boolean_t right_justify;
5792 	size_t width;
5793 
5794 	for (; pl != NULL; pl = pl->pl_next) {
5795 
5796 		width = pl->pl_width;
5797 		if (first && cb->cb_verbose) {
5798 			/*
5799 			 * Reset the width to accommodate the verbose listing
5800 			 * of devices.
5801 			 */
5802 			width = cb->cb_namewidth;
5803 		}
5804 
5805 		if (!first) {
5806 			if (cb->cb_scripted)
5807 				(void) printf("\t");
5808 			else
5809 				(void) printf("  ");
5810 		} else {
5811 			first = B_FALSE;
5812 		}
5813 
5814 		right_justify = B_FALSE;
5815 		if (pl->pl_prop != ZPROP_INVAL) {
5816 			if (zpool_get_prop(zhp, pl->pl_prop, property,
5817 			    sizeof (property), NULL, cb->cb_literal) != 0)
5818 				propstr = "-";
5819 			else
5820 				propstr = property;
5821 
5822 			right_justify = zpool_prop_align_right(pl->pl_prop);
5823 		} else if ((zpool_prop_feature(pl->pl_user_prop) ||
5824 		    zpool_prop_unsupported(pl->pl_user_prop)) &&
5825 		    zpool_prop_get_feature(zhp, pl->pl_user_prop, property,
5826 		    sizeof (property)) == 0) {
5827 			propstr = property;
5828 		} else {
5829 			propstr = "-";
5830 		}
5831 
5832 
5833 		/*
5834 		 * If this is being called in scripted mode, or if this is the
5835 		 * last column and it is left-justified, don't include a width
5836 		 * format specifier.
5837 		 */
5838 		if (cb->cb_scripted || (pl->pl_next == NULL && !right_justify))
5839 			(void) printf("%s", propstr);
5840 		else if (right_justify)
5841 			(void) printf("%*s", (int)width, propstr);
5842 		else
5843 			(void) printf("%-*s", (int)width, propstr);
5844 	}
5845 
5846 	(void) printf("\n");
5847 }
5848 
5849 static void
5850 print_one_column(zpool_prop_t prop, uint64_t value, const char *str,
5851     boolean_t scripted, boolean_t valid, enum zfs_nicenum_format format)
5852 {
5853 	char propval[64];
5854 	boolean_t fixed;
5855 	size_t width = zprop_width(prop, &fixed, ZFS_TYPE_POOL);
5856 
5857 	switch (prop) {
5858 	case ZPOOL_PROP_EXPANDSZ:
5859 	case ZPOOL_PROP_CHECKPOINT:
5860 	case ZPOOL_PROP_DEDUPRATIO:
5861 		if (value == 0)
5862 			(void) strlcpy(propval, "-", sizeof (propval));
5863 		else
5864 			zfs_nicenum_format(value, propval, sizeof (propval),
5865 			    format);
5866 		break;
5867 	case ZPOOL_PROP_FRAGMENTATION:
5868 		if (value == ZFS_FRAG_INVALID) {
5869 			(void) strlcpy(propval, "-", sizeof (propval));
5870 		} else if (format == ZFS_NICENUM_RAW) {
5871 			(void) snprintf(propval, sizeof (propval), "%llu",
5872 			    (unsigned long long)value);
5873 		} else {
5874 			(void) snprintf(propval, sizeof (propval), "%llu%%",
5875 			    (unsigned long long)value);
5876 		}
5877 		break;
5878 	case ZPOOL_PROP_CAPACITY:
5879 		/* capacity value is in parts-per-10,000 (aka permyriad) */
5880 		if (format == ZFS_NICENUM_RAW)
5881 			(void) snprintf(propval, sizeof (propval), "%llu",
5882 			    (unsigned long long)value / 100);
5883 		else
5884 			(void) snprintf(propval, sizeof (propval),
5885 			    value < 1000 ? "%1.2f%%" : value < 10000 ?
5886 			    "%2.1f%%" : "%3.0f%%", value / 100.0);
5887 		break;
5888 	case ZPOOL_PROP_HEALTH:
5889 		width = 8;
5890 		snprintf(propval, sizeof (propval), "%-*s", (int)width, str);
5891 		break;
5892 	default:
5893 		zfs_nicenum_format(value, propval, sizeof (propval), format);
5894 	}
5895 
5896 	if (!valid)
5897 		(void) strlcpy(propval, "-", sizeof (propval));
5898 
5899 	if (scripted)
5900 		(void) printf("\t%s", propval);
5901 	else
5902 		(void) printf("  %*s", (int)width, propval);
5903 }
5904 
5905 /*
5906  * print static default line per vdev
5907  * not compatible with '-o' <proplist> option
5908  */
5909 static void
5910 print_list_stats(zpool_handle_t *zhp, const char *name, nvlist_t *nv,
5911     list_cbdata_t *cb, int depth, boolean_t isspare)
5912 {
5913 	nvlist_t **child;
5914 	vdev_stat_t *vs;
5915 	uint_t c, children;
5916 	char *vname;
5917 	boolean_t scripted = cb->cb_scripted;
5918 	uint64_t islog = B_FALSE;
5919 	char *dashes = "%-*s      -      -      -        -         "
5920 	    "-      -      -      -  -\n";
5921 
5922 	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
5923 	    (uint64_t **)&vs, &c) == 0);
5924 
5925 	if (name != NULL) {
5926 		boolean_t toplevel = (vs->vs_space != 0);
5927 		uint64_t cap;
5928 		enum zfs_nicenum_format format;
5929 		const char *state;
5930 
5931 		if (cb->cb_literal)
5932 			format = ZFS_NICENUM_RAW;
5933 		else
5934 			format = ZFS_NICENUM_1024;
5935 
5936 		if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
5937 			return;
5938 
5939 		if (scripted)
5940 			(void) printf("\t%s", name);
5941 		else if (strlen(name) + depth > cb->cb_namewidth)
5942 			(void) printf("%*s%s", depth, "", name);
5943 		else
5944 			(void) printf("%*s%s%*s", depth, "", name,
5945 			    (int)(cb->cb_namewidth - strlen(name) - depth), "");
5946 
5947 		/*
5948 		 * Print the properties for the individual vdevs. Some
5949 		 * properties are only applicable to toplevel vdevs. The
5950 		 * 'toplevel' boolean value is passed to the print_one_column()
5951 		 * to indicate that the value is valid.
5952 		 */
5953 		print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL, scripted,
5954 		    toplevel, format);
5955 		print_one_column(ZPOOL_PROP_ALLOCATED, vs->vs_alloc, NULL,
5956 		    scripted, toplevel, format);
5957 		print_one_column(ZPOOL_PROP_FREE, vs->vs_space - vs->vs_alloc,
5958 		    NULL, scripted, toplevel, format);
5959 		print_one_column(ZPOOL_PROP_CHECKPOINT,
5960 		    vs->vs_checkpoint_space, NULL, scripted, toplevel, format);
5961 		print_one_column(ZPOOL_PROP_EXPANDSZ, vs->vs_esize, NULL,
5962 		    scripted, B_TRUE, format);
5963 		print_one_column(ZPOOL_PROP_FRAGMENTATION,
5964 		    vs->vs_fragmentation, NULL, scripted,
5965 		    (vs->vs_fragmentation != ZFS_FRAG_INVALID && toplevel),
5966 		    format);
5967 		cap = (vs->vs_space == 0) ? 0 :
5968 		    (vs->vs_alloc * 10000 / vs->vs_space);
5969 		print_one_column(ZPOOL_PROP_CAPACITY, cap, NULL,
5970 		    scripted, toplevel, format);
5971 		print_one_column(ZPOOL_PROP_DEDUPRATIO, 0, NULL,
5972 		    scripted, toplevel, format);
5973 		state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
5974 		if (isspare) {
5975 			if (vs->vs_aux == VDEV_AUX_SPARED)
5976 				state = "INUSE";
5977 			else if (vs->vs_state == VDEV_STATE_HEALTHY)
5978 				state = "AVAIL";
5979 		}
5980 		print_one_column(ZPOOL_PROP_HEALTH, 0, state, scripted,
5981 		    B_TRUE, format);
5982 		(void) printf("\n");
5983 	}
5984 
5985 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
5986 	    &child, &children) != 0)
5987 		return;
5988 
5989 	/* list the normal vdevs first */
5990 	for (c = 0; c < children; c++) {
5991 		uint64_t ishole = B_FALSE;
5992 
5993 		if (nvlist_lookup_uint64(child[c],
5994 		    ZPOOL_CONFIG_IS_HOLE, &ishole) == 0 && ishole)
5995 			continue;
5996 
5997 		if (nvlist_lookup_uint64(child[c],
5998 		    ZPOOL_CONFIG_IS_LOG, &islog) == 0 && islog)
5999 			continue;
6000 
6001 		if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
6002 			continue;
6003 
6004 		vname = zpool_vdev_name(g_zfs, zhp, child[c],
6005 		    cb->cb_name_flags);
6006 		print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE);
6007 		free(vname);
6008 	}
6009 
6010 	/* list the classes: 'logs', 'dedup', and 'special' */
6011 	for (uint_t n = 0; n < 3; n++) {
6012 		boolean_t printed = B_FALSE;
6013 
6014 		for (c = 0; c < children; c++) {
6015 			char *bias = NULL;
6016 			char *type = NULL;
6017 
6018 			if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
6019 			    &islog) == 0 && islog) {
6020 				bias = VDEV_ALLOC_CLASS_LOGS;
6021 			} else {
6022 				(void) nvlist_lookup_string(child[c],
6023 				    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
6024 				(void) nvlist_lookup_string(child[c],
6025 				    ZPOOL_CONFIG_TYPE, &type);
6026 			}
6027 			if (bias == NULL || strcmp(bias, class_name[n]) != 0)
6028 				continue;
6029 			if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
6030 				continue;
6031 
6032 			if (!printed) {
6033 				/* LINTED E_SEC_PRINTF_VAR_FMT */
6034 				(void) printf(dashes, cb->cb_namewidth,
6035 				    class_name[n]);
6036 				printed = B_TRUE;
6037 			}
6038 			vname = zpool_vdev_name(g_zfs, zhp, child[c],
6039 			    cb->cb_name_flags);
6040 			print_list_stats(zhp, vname, child[c], cb, depth + 2,
6041 			    B_FALSE);
6042 			free(vname);
6043 		}
6044 	}
6045 
6046 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
6047 	    &child, &children) == 0 && children > 0) {
6048 		/* LINTED E_SEC_PRINTF_VAR_FMT */
6049 		(void) printf(dashes, cb->cb_namewidth, "cache");
6050 		for (c = 0; c < children; c++) {
6051 			vname = zpool_vdev_name(g_zfs, zhp, child[c],
6052 			    cb->cb_name_flags);
6053 			print_list_stats(zhp, vname, child[c], cb, depth + 2,
6054 			    B_FALSE);
6055 			free(vname);
6056 		}
6057 	}
6058 
6059 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child,
6060 	    &children) == 0 && children > 0) {
6061 		/* LINTED E_SEC_PRINTF_VAR_FMT */
6062 		(void) printf(dashes, cb->cb_namewidth, "spare");
6063 		for (c = 0; c < children; c++) {
6064 			vname = zpool_vdev_name(g_zfs, zhp, child[c],
6065 			    cb->cb_name_flags);
6066 			print_list_stats(zhp, vname, child[c], cb, depth + 2,
6067 			    B_TRUE);
6068 			free(vname);
6069 		}
6070 	}
6071 }
6072 
6073 /*
6074  * Generic callback function to list a pool.
6075  */
6076 static int
6077 list_callback(zpool_handle_t *zhp, void *data)
6078 {
6079 	list_cbdata_t *cbp = data;
6080 
6081 	print_pool(zhp, cbp);
6082 
6083 	if (cbp->cb_verbose) {
6084 		nvlist_t *config, *nvroot;
6085 
6086 		config = zpool_get_config(zhp, NULL);
6087 		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
6088 		    &nvroot) == 0);
6089 		print_list_stats(zhp, NULL, nvroot, cbp, 0, B_FALSE);
6090 	}
6091 
6092 	return (0);
6093 }
6094 
6095 /*
6096  * Set the minimum pool/vdev name column width.  The width must be at least 9,
6097  * but may be as large as needed.
6098  */
6099 static int
6100 get_namewidth_list(zpool_handle_t *zhp, void *data)
6101 {
6102 	list_cbdata_t *cb = data;
6103 	int width;
6104 
6105 	width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags,
6106 	    cb->cb_verbose);
6107 
6108 	if (width < 9)
6109 		width = 9;
6110 
6111 	cb->cb_namewidth = width;
6112 
6113 	return (0);
6114 }
6115 
6116 /*
6117  * zpool list [-gHLpP] [-o prop[,prop]*] [-T d|u] [pool] ... [interval [count]]
6118  *
6119  *	-g	Display guid for individual vdev name.
6120  *	-H	Scripted mode.  Don't display headers, and separate properties
6121  *		by a single tab.
6122  *	-L	Follow links when resolving vdev path name.
6123  *	-o	List of properties to display.  Defaults to
6124  *		"name,size,allocated,free,expandsize,fragmentation,capacity,"
6125  *		"dedupratio,health,altroot"
6126  *	-p	Display values in parsable (exact) format.
6127  *	-P	Display full path for vdev name.
6128  *	-T	Display a timestamp in date(1) or Unix format
6129  *
6130  * List all pools in the system, whether or not they're healthy.  Output space
6131  * statistics for each one, as well as health status summary.
6132  */
6133 int
6134 zpool_do_list(int argc, char **argv)
6135 {
6136 	int c;
6137 	int ret = 0;
6138 	list_cbdata_t cb = { 0 };
6139 	static char default_props[] =
6140 	    "name,size,allocated,free,checkpoint,expandsize,fragmentation,"
6141 	    "capacity,dedupratio,health,altroot";
6142 	char *props = default_props;
6143 	float interval = 0;
6144 	unsigned long count = 0;
6145 	zpool_list_t *list;
6146 	boolean_t first = B_TRUE;
6147 
6148 	/* check options */
6149 	while ((c = getopt(argc, argv, ":gHLo:pPT:v")) != -1) {
6150 		switch (c) {
6151 		case 'g':
6152 			cb.cb_name_flags |= VDEV_NAME_GUID;
6153 			break;
6154 		case 'H':
6155 			cb.cb_scripted = B_TRUE;
6156 			break;
6157 		case 'L':
6158 			cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS;
6159 			break;
6160 		case 'o':
6161 			props = optarg;
6162 			break;
6163 		case 'P':
6164 			cb.cb_name_flags |= VDEV_NAME_PATH;
6165 			break;
6166 		case 'p':
6167 			cb.cb_literal = B_TRUE;
6168 			break;
6169 		case 'T':
6170 			get_timestamp_arg(*optarg);
6171 			break;
6172 		case 'v':
6173 			cb.cb_verbose = B_TRUE;
6174 			cb.cb_namewidth = 8;	/* 8 until precalc is avail */
6175 			break;
6176 		case ':':
6177 			(void) fprintf(stderr, gettext("missing argument for "
6178 			    "'%c' option\n"), optopt);
6179 			usage(B_FALSE);
6180 			break;
6181 		case '?':
6182 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6183 			    optopt);
6184 			usage(B_FALSE);
6185 		}
6186 	}
6187 
6188 	argc -= optind;
6189 	argv += optind;
6190 
6191 	get_interval_count(&argc, argv, &interval, &count);
6192 
6193 	if (zprop_get_list(g_zfs, props, &cb.cb_proplist, ZFS_TYPE_POOL) != 0)
6194 		usage(B_FALSE);
6195 
6196 	for (;;) {
6197 		if ((list = pool_list_get(argc, argv, &cb.cb_proplist,
6198 		    cb.cb_literal, &ret)) == NULL)
6199 			return (1);
6200 
6201 		if (pool_list_count(list) == 0)
6202 			break;
6203 
6204 		cb.cb_namewidth = 0;
6205 		(void) pool_list_iter(list, B_FALSE, get_namewidth_list, &cb);
6206 
6207 		if (timestamp_fmt != NODATE)
6208 			print_timestamp(timestamp_fmt);
6209 
6210 		if (!cb.cb_scripted && (first || cb.cb_verbose)) {
6211 			print_header(&cb);
6212 			first = B_FALSE;
6213 		}
6214 		ret = pool_list_iter(list, B_TRUE, list_callback, &cb);
6215 
6216 		if (interval == 0)
6217 			break;
6218 
6219 		if (count != 0 && --count == 0)
6220 			break;
6221 
6222 		pool_list_free(list);
6223 		(void) fsleep(interval);
6224 	}
6225 
6226 	if (argc == 0 && !cb.cb_scripted && pool_list_count(list) == 0) {
6227 		(void) printf(gettext("no pools available\n"));
6228 		ret = 0;
6229 	}
6230 
6231 	pool_list_free(list);
6232 	zprop_free_list(cb.cb_proplist);
6233 	return (ret);
6234 }
6235 
6236 static int
6237 zpool_do_attach_or_replace(int argc, char **argv, int replacing)
6238 {
6239 	boolean_t force = B_FALSE;
6240 	boolean_t rebuild = B_FALSE;
6241 	boolean_t wait = B_FALSE;
6242 	int c;
6243 	nvlist_t *nvroot;
6244 	char *poolname, *old_disk, *new_disk;
6245 	zpool_handle_t *zhp;
6246 	nvlist_t *props = NULL;
6247 	char *propval;
6248 	int ret;
6249 
6250 	/* check options */
6251 	while ((c = getopt(argc, argv, "fo:sw")) != -1) {
6252 		switch (c) {
6253 		case 'f':
6254 			force = B_TRUE;
6255 			break;
6256 		case 'o':
6257 			if ((propval = strchr(optarg, '=')) == NULL) {
6258 				(void) fprintf(stderr, gettext("missing "
6259 				    "'=' for -o option\n"));
6260 				usage(B_FALSE);
6261 			}
6262 			*propval = '\0';
6263 			propval++;
6264 
6265 			if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) ||
6266 			    (add_prop_list(optarg, propval, &props, B_TRUE)))
6267 				usage(B_FALSE);
6268 			break;
6269 		case 's':
6270 			rebuild = B_TRUE;
6271 			break;
6272 		case 'w':
6273 			wait = B_TRUE;
6274 			break;
6275 		case '?':
6276 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6277 			    optopt);
6278 			usage(B_FALSE);
6279 		}
6280 	}
6281 
6282 	argc -= optind;
6283 	argv += optind;
6284 
6285 	/* get pool name and check number of arguments */
6286 	if (argc < 1) {
6287 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
6288 		usage(B_FALSE);
6289 	}
6290 
6291 	poolname = argv[0];
6292 
6293 	if (argc < 2) {
6294 		(void) fprintf(stderr,
6295 		    gettext("missing <device> specification\n"));
6296 		usage(B_FALSE);
6297 	}
6298 
6299 	old_disk = argv[1];
6300 
6301 	if (argc < 3) {
6302 		if (!replacing) {
6303 			(void) fprintf(stderr,
6304 			    gettext("missing <new_device> specification\n"));
6305 			usage(B_FALSE);
6306 		}
6307 		new_disk = old_disk;
6308 		argc -= 1;
6309 		argv += 1;
6310 	} else {
6311 		new_disk = argv[2];
6312 		argc -= 2;
6313 		argv += 2;
6314 	}
6315 
6316 	if (argc > 1) {
6317 		(void) fprintf(stderr, gettext("too many arguments\n"));
6318 		usage(B_FALSE);
6319 	}
6320 
6321 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
6322 		nvlist_free(props);
6323 		return (1);
6324 	}
6325 
6326 	if (zpool_get_config(zhp, NULL) == NULL) {
6327 		(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
6328 		    poolname);
6329 		zpool_close(zhp);
6330 		nvlist_free(props);
6331 		return (1);
6332 	}
6333 
6334 	/* unless manually specified use "ashift" pool property (if set) */
6335 	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
6336 		int intval;
6337 		zprop_source_t src;
6338 		char strval[ZPOOL_MAXPROPLEN];
6339 
6340 		intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
6341 		if (src != ZPROP_SRC_DEFAULT) {
6342 			(void) sprintf(strval, "%" PRId32, intval);
6343 			verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
6344 			    &props, B_TRUE) == 0);
6345 		}
6346 	}
6347 
6348 	nvroot = make_root_vdev(zhp, props, force, B_FALSE, replacing, B_FALSE,
6349 	    argc, argv);
6350 	if (nvroot == NULL) {
6351 		zpool_close(zhp);
6352 		nvlist_free(props);
6353 		return (1);
6354 	}
6355 
6356 	ret = zpool_vdev_attach(zhp, old_disk, new_disk, nvroot, replacing,
6357 	    rebuild);
6358 
6359 	if (ret == 0 && wait)
6360 		ret = zpool_wait(zhp,
6361 		    replacing ? ZPOOL_WAIT_REPLACE : ZPOOL_WAIT_RESILVER);
6362 
6363 	nvlist_free(props);
6364 	nvlist_free(nvroot);
6365 	zpool_close(zhp);
6366 
6367 	return (ret);
6368 }
6369 
6370 /*
6371  * zpool replace [-fsw] [-o property=value] <pool> <device> <new_device>
6372  *
6373  *	-f	Force attach, even if <new_device> appears to be in use.
6374  *	-s	Use sequential instead of healing reconstruction for resilver.
6375  *	-o	Set property=value.
6376  *	-w	Wait for replacing to complete before returning
6377  *
6378  * Replace <device> with <new_device>.
6379  */
6380 /* ARGSUSED */
6381 int
6382 zpool_do_replace(int argc, char **argv)
6383 {
6384 	return (zpool_do_attach_or_replace(argc, argv, B_TRUE));
6385 }
6386 
6387 /*
6388  * zpool attach [-fsw] [-o property=value] <pool> <device> <new_device>
6389  *
6390  *	-f	Force attach, even if <new_device> appears to be in use.
6391  *	-s	Use sequential instead of healing reconstruction for resilver.
6392  *	-o	Set property=value.
6393  *	-w	Wait for resilvering to complete before returning
6394  *
6395  * Attach <new_device> to the mirror containing <device>.  If <device> is not
6396  * part of a mirror, then <device> will be transformed into a mirror of
6397  * <device> and <new_device>.  In either case, <new_device> will begin life
6398  * with a DTL of [0, now], and will immediately begin to resilver itself.
6399  */
6400 int
6401 zpool_do_attach(int argc, char **argv)
6402 {
6403 	return (zpool_do_attach_or_replace(argc, argv, B_FALSE));
6404 }
6405 
6406 /*
6407  * zpool detach [-f] <pool> <device>
6408  *
6409  *	-f	Force detach of <device>, even if DTLs argue against it
6410  *		(not supported yet)
6411  *
6412  * Detach a device from a mirror.  The operation will be refused if <device>
6413  * is the last device in the mirror, or if the DTLs indicate that this device
6414  * has the only valid copy of some data.
6415  */
6416 /* ARGSUSED */
6417 int
6418 zpool_do_detach(int argc, char **argv)
6419 {
6420 	int c;
6421 	char *poolname, *path;
6422 	zpool_handle_t *zhp;
6423 	int ret;
6424 
6425 	/* check options */
6426 	while ((c = getopt(argc, argv, "")) != -1) {
6427 		switch (c) {
6428 		case '?':
6429 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6430 			    optopt);
6431 			usage(B_FALSE);
6432 		}
6433 	}
6434 
6435 	argc -= optind;
6436 	argv += optind;
6437 
6438 	/* get pool name and check number of arguments */
6439 	if (argc < 1) {
6440 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
6441 		usage(B_FALSE);
6442 	}
6443 
6444 	if (argc < 2) {
6445 		(void) fprintf(stderr,
6446 		    gettext("missing <device> specification\n"));
6447 		usage(B_FALSE);
6448 	}
6449 
6450 	poolname = argv[0];
6451 	path = argv[1];
6452 
6453 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6454 		return (1);
6455 
6456 	ret = zpool_vdev_detach(zhp, path);
6457 
6458 	zpool_close(zhp);
6459 
6460 	return (ret);
6461 }
6462 
6463 /*
6464  * zpool split [-gLnP] [-o prop=val] ...
6465  *		[-o mntopt] ...
6466  *		[-R altroot] <pool> <newpool> [<device> ...]
6467  *
6468  *	-g      Display guid for individual vdev name.
6469  *	-L	Follow links when resolving vdev path name.
6470  *	-n	Do not split the pool, but display the resulting layout if
6471  *		it were to be split.
6472  *	-o	Set property=value, or set mount options.
6473  *	-P	Display full path for vdev name.
6474  *	-R	Mount the split-off pool under an alternate root.
6475  *	-l	Load encryption keys while importing.
6476  *
6477  * Splits the named pool and gives it the new pool name.  Devices to be split
6478  * off may be listed, provided that no more than one device is specified
6479  * per top-level vdev mirror.  The newly split pool is left in an exported
6480  * state unless -R is specified.
6481  *
6482  * Restrictions: the top-level of the pool pool must only be made up of
6483  * mirrors; all devices in the pool must be healthy; no device may be
6484  * undergoing a resilvering operation.
6485  */
6486 int
6487 zpool_do_split(int argc, char **argv)
6488 {
6489 	char *srcpool, *newpool, *propval;
6490 	char *mntopts = NULL;
6491 	splitflags_t flags;
6492 	int c, ret = 0;
6493 	boolean_t loadkeys = B_FALSE;
6494 	zpool_handle_t *zhp;
6495 	nvlist_t *config, *props = NULL;
6496 
6497 	flags.dryrun = B_FALSE;
6498 	flags.import = B_FALSE;
6499 	flags.name_flags = 0;
6500 
6501 	/* check options */
6502 	while ((c = getopt(argc, argv, ":gLR:lno:P")) != -1) {
6503 		switch (c) {
6504 		case 'g':
6505 			flags.name_flags |= VDEV_NAME_GUID;
6506 			break;
6507 		case 'L':
6508 			flags.name_flags |= VDEV_NAME_FOLLOW_LINKS;
6509 			break;
6510 		case 'R':
6511 			flags.import = B_TRUE;
6512 			if (add_prop_list(
6513 			    zpool_prop_to_name(ZPOOL_PROP_ALTROOT), optarg,
6514 			    &props, B_TRUE) != 0) {
6515 				nvlist_free(props);
6516 				usage(B_FALSE);
6517 			}
6518 			break;
6519 		case 'l':
6520 			loadkeys = B_TRUE;
6521 			break;
6522 		case 'n':
6523 			flags.dryrun = B_TRUE;
6524 			break;
6525 		case 'o':
6526 			if ((propval = strchr(optarg, '=')) != NULL) {
6527 				*propval = '\0';
6528 				propval++;
6529 				if (add_prop_list(optarg, propval,
6530 				    &props, B_TRUE) != 0) {
6531 					nvlist_free(props);
6532 					usage(B_FALSE);
6533 				}
6534 			} else {
6535 				mntopts = optarg;
6536 			}
6537 			break;
6538 		case 'P':
6539 			flags.name_flags |= VDEV_NAME_PATH;
6540 			break;
6541 		case ':':
6542 			(void) fprintf(stderr, gettext("missing argument for "
6543 			    "'%c' option\n"), optopt);
6544 			usage(B_FALSE);
6545 			break;
6546 		case '?':
6547 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6548 			    optopt);
6549 			usage(B_FALSE);
6550 			break;
6551 		}
6552 	}
6553 
6554 	if (!flags.import && mntopts != NULL) {
6555 		(void) fprintf(stderr, gettext("setting mntopts is only "
6556 		    "valid when importing the pool\n"));
6557 		usage(B_FALSE);
6558 	}
6559 
6560 	if (!flags.import && loadkeys) {
6561 		(void) fprintf(stderr, gettext("loading keys is only "
6562 		    "valid when importing the pool\n"));
6563 		usage(B_FALSE);
6564 	}
6565 
6566 	argc -= optind;
6567 	argv += optind;
6568 
6569 	if (argc < 1) {
6570 		(void) fprintf(stderr, gettext("Missing pool name\n"));
6571 		usage(B_FALSE);
6572 	}
6573 	if (argc < 2) {
6574 		(void) fprintf(stderr, gettext("Missing new pool name\n"));
6575 		usage(B_FALSE);
6576 	}
6577 
6578 	srcpool = argv[0];
6579 	newpool = argv[1];
6580 
6581 	argc -= 2;
6582 	argv += 2;
6583 
6584 	if ((zhp = zpool_open(g_zfs, srcpool)) == NULL) {
6585 		nvlist_free(props);
6586 		return (1);
6587 	}
6588 
6589 	config = split_mirror_vdev(zhp, newpool, props, flags, argc, argv);
6590 	if (config == NULL) {
6591 		ret = 1;
6592 	} else {
6593 		if (flags.dryrun) {
6594 			(void) printf(gettext("would create '%s' with the "
6595 			    "following layout:\n\n"), newpool);
6596 			print_vdev_tree(NULL, newpool, config, 0, "",
6597 			    flags.name_flags);
6598 			print_vdev_tree(NULL, "dedup", config, 0,
6599 			    VDEV_ALLOC_BIAS_DEDUP, 0);
6600 			print_vdev_tree(NULL, "special", config, 0,
6601 			    VDEV_ALLOC_BIAS_SPECIAL, 0);
6602 		}
6603 	}
6604 
6605 	zpool_close(zhp);
6606 
6607 	if (ret != 0 || flags.dryrun || !flags.import) {
6608 		nvlist_free(config);
6609 		nvlist_free(props);
6610 		return (ret);
6611 	}
6612 
6613 	/*
6614 	 * The split was successful. Now we need to open the new
6615 	 * pool and import it.
6616 	 */
6617 	if ((zhp = zpool_open_canfail(g_zfs, newpool)) == NULL) {
6618 		nvlist_free(config);
6619 		nvlist_free(props);
6620 		return (1);
6621 	}
6622 
6623 	if (loadkeys) {
6624 		ret = zfs_crypto_attempt_load_keys(g_zfs, newpool);
6625 		if (ret != 0)
6626 			ret = 1;
6627 	}
6628 
6629 	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
6630 	    zpool_enable_datasets(zhp, mntopts, 0) != 0) {
6631 		ret = 1;
6632 		(void) fprintf(stderr, gettext("Split was successful, but "
6633 		    "the datasets could not all be mounted\n"));
6634 		(void) fprintf(stderr, gettext("Try doing '%s' with a "
6635 		    "different altroot\n"), "zpool import");
6636 	}
6637 	zpool_close(zhp);
6638 	nvlist_free(config);
6639 	nvlist_free(props);
6640 
6641 	return (ret);
6642 }
6643 
6644 
6645 
6646 /*
6647  * zpool online <pool> <device> ...
6648  */
6649 int
6650 zpool_do_online(int argc, char **argv)
6651 {
6652 	int c, i;
6653 	char *poolname;
6654 	zpool_handle_t *zhp;
6655 	int ret = 0;
6656 	vdev_state_t newstate;
6657 	int flags = 0;
6658 
6659 	/* check options */
6660 	while ((c = getopt(argc, argv, "e")) != -1) {
6661 		switch (c) {
6662 		case 'e':
6663 			flags |= ZFS_ONLINE_EXPAND;
6664 			break;
6665 		case '?':
6666 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6667 			    optopt);
6668 			usage(B_FALSE);
6669 		}
6670 	}
6671 
6672 	argc -= optind;
6673 	argv += optind;
6674 
6675 	/* get pool name and check number of arguments */
6676 	if (argc < 1) {
6677 		(void) fprintf(stderr, gettext("missing pool name\n"));
6678 		usage(B_FALSE);
6679 	}
6680 	if (argc < 2) {
6681 		(void) fprintf(stderr, gettext("missing device name\n"));
6682 		usage(B_FALSE);
6683 	}
6684 
6685 	poolname = argv[0];
6686 
6687 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6688 		return (1);
6689 
6690 	for (i = 1; i < argc; i++) {
6691 		if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
6692 			if (newstate != VDEV_STATE_HEALTHY) {
6693 				(void) printf(gettext("warning: device '%s' "
6694 				    "onlined, but remains in faulted state\n"),
6695 				    argv[i]);
6696 				if (newstate == VDEV_STATE_FAULTED)
6697 					(void) printf(gettext("use 'zpool "
6698 					    "clear' to restore a faulted "
6699 					    "device\n"));
6700 				else
6701 					(void) printf(gettext("use 'zpool "
6702 					    "replace' to replace devices "
6703 					    "that are no longer present\n"));
6704 			}
6705 		} else {
6706 			ret = 1;
6707 		}
6708 	}
6709 
6710 	zpool_close(zhp);
6711 
6712 	return (ret);
6713 }
6714 
6715 /*
6716  * zpool offline [-ft] <pool> <device> ...
6717  *
6718  *	-f	Force the device into a faulted state.
6719  *
6720  *	-t	Only take the device off-line temporarily.  The offline/faulted
6721  *		state will not be persistent across reboots.
6722  */
6723 /* ARGSUSED */
6724 int
6725 zpool_do_offline(int argc, char **argv)
6726 {
6727 	int c, i;
6728 	char *poolname;
6729 	zpool_handle_t *zhp;
6730 	int ret = 0;
6731 	boolean_t istmp = B_FALSE;
6732 	boolean_t fault = B_FALSE;
6733 
6734 	/* check options */
6735 	while ((c = getopt(argc, argv, "ft")) != -1) {
6736 		switch (c) {
6737 		case 'f':
6738 			fault = B_TRUE;
6739 			break;
6740 		case 't':
6741 			istmp = B_TRUE;
6742 			break;
6743 		case '?':
6744 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6745 			    optopt);
6746 			usage(B_FALSE);
6747 		}
6748 	}
6749 
6750 	argc -= optind;
6751 	argv += optind;
6752 
6753 	/* get pool name and check number of arguments */
6754 	if (argc < 1) {
6755 		(void) fprintf(stderr, gettext("missing pool name\n"));
6756 		usage(B_FALSE);
6757 	}
6758 	if (argc < 2) {
6759 		(void) fprintf(stderr, gettext("missing device name\n"));
6760 		usage(B_FALSE);
6761 	}
6762 
6763 	poolname = argv[0];
6764 
6765 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6766 		return (1);
6767 
6768 	for (i = 1; i < argc; i++) {
6769 		if (fault) {
6770 			uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]);
6771 			vdev_aux_t aux;
6772 			if (istmp == B_FALSE) {
6773 				/* Force the fault to persist across imports */
6774 				aux = VDEV_AUX_EXTERNAL_PERSIST;
6775 			} else {
6776 				aux = VDEV_AUX_EXTERNAL;
6777 			}
6778 
6779 			if (guid == 0 || zpool_vdev_fault(zhp, guid, aux) != 0)
6780 				ret = 1;
6781 		} else {
6782 			if (zpool_vdev_offline(zhp, argv[i], istmp) != 0)
6783 				ret = 1;
6784 		}
6785 	}
6786 
6787 	zpool_close(zhp);
6788 
6789 	return (ret);
6790 }
6791 
6792 /*
6793  * zpool clear <pool> [device]
6794  *
6795  * Clear all errors associated with a pool or a particular device.
6796  */
6797 int
6798 zpool_do_clear(int argc, char **argv)
6799 {
6800 	int c;
6801 	int ret = 0;
6802 	boolean_t dryrun = B_FALSE;
6803 	boolean_t do_rewind = B_FALSE;
6804 	boolean_t xtreme_rewind = B_FALSE;
6805 	uint32_t rewind_policy = ZPOOL_NO_REWIND;
6806 	nvlist_t *policy = NULL;
6807 	zpool_handle_t *zhp;
6808 	char *pool, *device;
6809 
6810 	/* check options */
6811 	while ((c = getopt(argc, argv, "FnX")) != -1) {
6812 		switch (c) {
6813 		case 'F':
6814 			do_rewind = B_TRUE;
6815 			break;
6816 		case 'n':
6817 			dryrun = B_TRUE;
6818 			break;
6819 		case 'X':
6820 			xtreme_rewind = B_TRUE;
6821 			break;
6822 		case '?':
6823 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6824 			    optopt);
6825 			usage(B_FALSE);
6826 		}
6827 	}
6828 
6829 	argc -= optind;
6830 	argv += optind;
6831 
6832 	if (argc < 1) {
6833 		(void) fprintf(stderr, gettext("missing pool name\n"));
6834 		usage(B_FALSE);
6835 	}
6836 
6837 	if (argc > 2) {
6838 		(void) fprintf(stderr, gettext("too many arguments\n"));
6839 		usage(B_FALSE);
6840 	}
6841 
6842 	if ((dryrun || xtreme_rewind) && !do_rewind) {
6843 		(void) fprintf(stderr,
6844 		    gettext("-n or -X only meaningful with -F\n"));
6845 		usage(B_FALSE);
6846 	}
6847 	if (dryrun)
6848 		rewind_policy = ZPOOL_TRY_REWIND;
6849 	else if (do_rewind)
6850 		rewind_policy = ZPOOL_DO_REWIND;
6851 	if (xtreme_rewind)
6852 		rewind_policy |= ZPOOL_EXTREME_REWIND;
6853 
6854 	/* In future, further rewind policy choices can be passed along here */
6855 	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 ||
6856 	    nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
6857 	    rewind_policy) != 0) {
6858 		return (1);
6859 	}
6860 
6861 	pool = argv[0];
6862 	device = argc == 2 ? argv[1] : NULL;
6863 
6864 	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
6865 		nvlist_free(policy);
6866 		return (1);
6867 	}
6868 
6869 	if (zpool_clear(zhp, device, policy) != 0)
6870 		ret = 1;
6871 
6872 	zpool_close(zhp);
6873 
6874 	nvlist_free(policy);
6875 
6876 	return (ret);
6877 }
6878 
6879 /*
6880  * zpool reguid <pool>
6881  */
6882 int
6883 zpool_do_reguid(int argc, char **argv)
6884 {
6885 	int c;
6886 	char *poolname;
6887 	zpool_handle_t *zhp;
6888 	int ret = 0;
6889 
6890 	/* check options */
6891 	while ((c = getopt(argc, argv, "")) != -1) {
6892 		switch (c) {
6893 		case '?':
6894 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6895 			    optopt);
6896 			usage(B_FALSE);
6897 		}
6898 	}
6899 
6900 	argc -= optind;
6901 	argv += optind;
6902 
6903 	/* get pool name and check number of arguments */
6904 	if (argc < 1) {
6905 		(void) fprintf(stderr, gettext("missing pool name\n"));
6906 		usage(B_FALSE);
6907 	}
6908 
6909 	if (argc > 1) {
6910 		(void) fprintf(stderr, gettext("too many arguments\n"));
6911 		usage(B_FALSE);
6912 	}
6913 
6914 	poolname = argv[0];
6915 	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
6916 		return (1);
6917 
6918 	ret = zpool_reguid(zhp);
6919 
6920 	zpool_close(zhp);
6921 	return (ret);
6922 }
6923 
6924 
6925 /*
6926  * zpool reopen <pool>
6927  *
6928  * Reopen the pool so that the kernel can update the sizes of all vdevs.
6929  */
6930 int
6931 zpool_do_reopen(int argc, char **argv)
6932 {
6933 	int c;
6934 	int ret = 0;
6935 	boolean_t scrub_restart = B_TRUE;
6936 
6937 	/* check options */
6938 	while ((c = getopt(argc, argv, "n")) != -1) {
6939 		switch (c) {
6940 		case 'n':
6941 			scrub_restart = B_FALSE;
6942 			break;
6943 		case '?':
6944 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
6945 			    optopt);
6946 			usage(B_FALSE);
6947 		}
6948 	}
6949 
6950 	argc -= optind;
6951 	argv += optind;
6952 
6953 	/* if argc == 0 we will execute zpool_reopen_one on all pools */
6954 	ret = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE, zpool_reopen_one,
6955 	    &scrub_restart);
6956 
6957 	return (ret);
6958 }
6959 
6960 typedef struct scrub_cbdata {
6961 	int	cb_type;
6962 	pool_scrub_cmd_t cb_scrub_cmd;
6963 } scrub_cbdata_t;
6964 
6965 static boolean_t
6966 zpool_has_checkpoint(zpool_handle_t *zhp)
6967 {
6968 	nvlist_t *config, *nvroot;
6969 
6970 	config = zpool_get_config(zhp, NULL);
6971 
6972 	if (config != NULL) {
6973 		pool_checkpoint_stat_t *pcs = NULL;
6974 		uint_t c;
6975 
6976 		nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
6977 		(void) nvlist_lookup_uint64_array(nvroot,
6978 		    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
6979 
6980 		if (pcs == NULL || pcs->pcs_state == CS_NONE)
6981 			return (B_FALSE);
6982 
6983 		assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS ||
6984 		    pcs->pcs_state == CS_CHECKPOINT_DISCARDING);
6985 		return (B_TRUE);
6986 	}
6987 
6988 	return (B_FALSE);
6989 }
6990 
6991 static int
6992 scrub_callback(zpool_handle_t *zhp, void *data)
6993 {
6994 	scrub_cbdata_t *cb = data;
6995 	int err;
6996 
6997 	/*
6998 	 * Ignore faulted pools.
6999 	 */
7000 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
7001 		(void) fprintf(stderr, gettext("cannot scan '%s': pool is "
7002 		    "currently unavailable\n"), zpool_get_name(zhp));
7003 		return (1);
7004 	}
7005 
7006 	err = zpool_scan(zhp, cb->cb_type, cb->cb_scrub_cmd);
7007 
7008 	if (err == 0 && zpool_has_checkpoint(zhp) &&
7009 	    cb->cb_type == POOL_SCAN_SCRUB) {
7010 		(void) printf(gettext("warning: will not scrub state that "
7011 		    "belongs to the checkpoint of pool '%s'\n"),
7012 		    zpool_get_name(zhp));
7013 	}
7014 
7015 	return (err != 0);
7016 }
7017 
7018 static int
7019 wait_callback(zpool_handle_t *zhp, void *data)
7020 {
7021 	zpool_wait_activity_t *act = data;
7022 	return (zpool_wait(zhp, *act));
7023 }
7024 
7025 /*
7026  * zpool scrub [-s | -p] [-w] <pool> ...
7027  *
7028  *	-s	Stop.  Stops any in-progress scrub.
7029  *	-p	Pause. Pause in-progress scrub.
7030  *	-w	Wait.  Blocks until scrub has completed.
7031  */
7032 int
7033 zpool_do_scrub(int argc, char **argv)
7034 {
7035 	int c;
7036 	scrub_cbdata_t cb;
7037 	boolean_t wait = B_FALSE;
7038 	int error;
7039 
7040 	cb.cb_type = POOL_SCAN_SCRUB;
7041 	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;
7042 
7043 	/* check options */
7044 	while ((c = getopt(argc, argv, "spw")) != -1) {
7045 		switch (c) {
7046 		case 's':
7047 			cb.cb_type = POOL_SCAN_NONE;
7048 			break;
7049 		case 'p':
7050 			cb.cb_scrub_cmd = POOL_SCRUB_PAUSE;
7051 			break;
7052 		case 'w':
7053 			wait = B_TRUE;
7054 			break;
7055 		case '?':
7056 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
7057 			    optopt);
7058 			usage(B_FALSE);
7059 		}
7060 	}
7061 
7062 	if (cb.cb_type == POOL_SCAN_NONE &&
7063 	    cb.cb_scrub_cmd == POOL_SCRUB_PAUSE) {
7064 		(void) fprintf(stderr, gettext("invalid option combination: "
7065 		    "-s and -p are mutually exclusive\n"));
7066 		usage(B_FALSE);
7067 	}
7068 
7069 	if (wait && (cb.cb_type == POOL_SCAN_NONE ||
7070 	    cb.cb_scrub_cmd == POOL_SCRUB_PAUSE)) {
7071 		(void) fprintf(stderr, gettext("invalid option combination: "
7072 		    "-w cannot be used with -p or -s\n"));
7073 		usage(B_FALSE);
7074 	}
7075 
7076 	argc -= optind;
7077 	argv += optind;
7078 
7079 	if (argc < 1) {
7080 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
7081 		usage(B_FALSE);
7082 	}
7083 
7084 	error = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE,
7085 	    scrub_callback, &cb);
7086 
7087 	if (wait && !error) {
7088 		zpool_wait_activity_t act = ZPOOL_WAIT_SCRUB;
7089 		error = for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE,
7090 		    wait_callback, &act);
7091 	}
7092 
7093 	return (error);
7094 }
7095 
7096 /*
7097  * zpool resilver <pool> ...
7098  *
7099  *	Restarts any in-progress resilver
7100  */
7101 int
7102 zpool_do_resilver(int argc, char **argv)
7103 {
7104 	int c;
7105 	scrub_cbdata_t cb;
7106 
7107 	cb.cb_type = POOL_SCAN_RESILVER;
7108 	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;
7109 
7110 	/* check options */
7111 	while ((c = getopt(argc, argv, "")) != -1) {
7112 		switch (c) {
7113 		case '?':
7114 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
7115 			    optopt);
7116 			usage(B_FALSE);
7117 		}
7118 	}
7119 
7120 	argc -= optind;
7121 	argv += optind;
7122 
7123 	if (argc < 1) {
7124 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
7125 		usage(B_FALSE);
7126 	}
7127 
7128 	return (for_each_pool(argc, argv, B_TRUE, NULL, B_FALSE,
7129 	    scrub_callback, &cb));
7130 }
7131 
7132 /*
7133  * zpool trim [-d] [-r <rate>] [-c | -s] <pool> [<device> ...]
7134  *
7135  *	-c		Cancel. Ends any in-progress trim.
7136  *	-d		Secure trim.  Requires kernel and device support.
7137  *	-r <rate>	Sets the TRIM rate in bytes (per second). Supports
7138  *			adding a multiplier suffix such as 'k' or 'm'.
7139  *	-s		Suspend. TRIM can then be restarted with no flags.
7140  *	-w		Wait. Blocks until trimming has completed.
7141  */
7142 int
7143 zpool_do_trim(int argc, char **argv)
7144 {
7145 	struct option long_options[] = {
7146 		{"cancel",	no_argument,		NULL,	'c'},
7147 		{"secure",	no_argument,		NULL,	'd'},
7148 		{"rate",	required_argument,	NULL,	'r'},
7149 		{"suspend",	no_argument,		NULL,	's'},
7150 		{"wait",	no_argument,		NULL,	'w'},
7151 		{0, 0, 0, 0}
7152 	};
7153 
7154 	pool_trim_func_t cmd_type = POOL_TRIM_START;
7155 	uint64_t rate = 0;
7156 	boolean_t secure = B_FALSE;
7157 	boolean_t wait = B_FALSE;
7158 
7159 	int c;
7160 	while ((c = getopt_long(argc, argv, "cdr:sw", long_options, NULL))
7161 	    != -1) {
7162 		switch (c) {
7163 		case 'c':
7164 			if (cmd_type != POOL_TRIM_START &&
7165 			    cmd_type != POOL_TRIM_CANCEL) {
7166 				(void) fprintf(stderr, gettext("-c cannot be "
7167 				    "combined with other options\n"));
7168 				usage(B_FALSE);
7169 			}
7170 			cmd_type = POOL_TRIM_CANCEL;
7171 			break;
7172 		case 'd':
7173 			if (cmd_type != POOL_TRIM_START) {
7174 				(void) fprintf(stderr, gettext("-d cannot be "
7175 				    "combined with the -c or -s options\n"));
7176 				usage(B_FALSE);
7177 			}
7178 			secure = B_TRUE;
7179 			break;
7180 		case 'r':
7181 			if (cmd_type != POOL_TRIM_START) {
7182 				(void) fprintf(stderr, gettext("-r cannot be "
7183 				    "combined with the -c or -s options\n"));
7184 				usage(B_FALSE);
7185 			}
7186 			if (zfs_nicestrtonum(NULL, optarg, &rate) == -1) {
7187 				(void) fprintf(stderr,
7188 				    gettext("invalid value for rate\n"));
7189 				usage(B_FALSE);
7190 			}
7191 			break;
7192 		case 's':
7193 			if (cmd_type != POOL_TRIM_START &&
7194 			    cmd_type != POOL_TRIM_SUSPEND) {
7195 				(void) fprintf(stderr, gettext("-s cannot be "
7196 				    "combined with other options\n"));
7197 				usage(B_FALSE);
7198 			}
7199 			cmd_type = POOL_TRIM_SUSPEND;
7200 			break;
7201 		case 'w':
7202 			wait = B_TRUE;
7203 			break;
7204 		case '?':
7205 			if (optopt != 0) {
7206 				(void) fprintf(stderr,
7207 				    gettext("invalid option '%c'\n"), optopt);
7208 			} else {
7209 				(void) fprintf(stderr,
7210 				    gettext("invalid option '%s'\n"),
7211 				    argv[optind - 1]);
7212 			}
7213 			usage(B_FALSE);
7214 		}
7215 	}
7216 
7217 	argc -= optind;
7218 	argv += optind;
7219 
7220 	if (argc < 1) {
7221 		(void) fprintf(stderr, gettext("missing pool name argument\n"));
7222 		usage(B_FALSE);
7223 		return (-1);
7224 	}
7225 
7226 	if (wait && (cmd_type != POOL_TRIM_START)) {
7227 		(void) fprintf(stderr, gettext("-w cannot be used with -c or "
7228 		    "-s\n"));
7229 		usage(B_FALSE);
7230 	}
7231 
7232 	char *poolname = argv[0];
7233 	zpool_handle_t *zhp = zpool_open(g_zfs, poolname);
7234 	if (zhp == NULL)
7235 		return (-1);
7236 
7237 	trimflags_t trim_flags = {
7238 		.secure = secure,
7239 		.rate = rate,
7240 		.wait = wait,
7241 	};
7242 
7243 	nvlist_t *vdevs = fnvlist_alloc();
7244 	if (argc == 1) {
7245 		/* no individual leaf vdevs specified, so add them all */
7246 		nvlist_t *config = zpool_get_config(zhp, NULL);
7247 		nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
7248 		    ZPOOL_CONFIG_VDEV_TREE);
7249 		zpool_collect_leaves(zhp, nvroot, vdevs);
7250 		trim_flags.fullpool = B_TRUE;
7251 	} else {
7252 		trim_flags.fullpool = B_FALSE;
7253 		for (int i = 1; i < argc; i++) {
7254 			fnvlist_add_boolean(vdevs, argv[i]);
7255 		}
7256 	}
7257 
7258 	int error = zpool_trim(zhp, cmd_type, vdevs, &trim_flags);
7259 
7260 	fnvlist_free(vdevs);
7261 	zpool_close(zhp);
7262 
7263 	return (error);
7264 }
7265 
7266 /*
7267  * Converts a total number of seconds to a human readable string broken
7268  * down in to days/hours/minutes/seconds.
7269  */
7270 static void
7271 secs_to_dhms(uint64_t total, char *buf)
7272 {
7273 	uint64_t days = total / 60 / 60 / 24;
7274 	uint64_t hours = (total / 60 / 60) % 24;
7275 	uint64_t mins = (total / 60) % 60;
7276 	uint64_t secs = (total % 60);
7277 
7278 	if (days > 0) {
7279 		(void) sprintf(buf, "%llu days %02llu:%02llu:%02llu",
7280 		    (u_longlong_t)days, (u_longlong_t)hours,
7281 		    (u_longlong_t)mins, (u_longlong_t)secs);
7282 	} else {
7283 		(void) sprintf(buf, "%02llu:%02llu:%02llu",
7284 		    (u_longlong_t)hours, (u_longlong_t)mins,
7285 		    (u_longlong_t)secs);
7286 	}
7287 }
7288 
7289 /*
7290  * Print out detailed scrub status.
7291  */
7292 static void
7293 print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
7294 {
7295 	time_t start, end, pause;
7296 	uint64_t pass_scanned, scanned, pass_issued, issued, total;
7297 	uint64_t elapsed, scan_rate, issue_rate;
7298 	double fraction_done;
7299 	char processed_buf[7], scanned_buf[7], issued_buf[7], total_buf[7];
7300 	char srate_buf[7], irate_buf[7], time_buf[32];
7301 
7302 	printf("  ");
7303 	printf_color(ANSI_BOLD, gettext("scan:"));
7304 	printf(" ");
7305 
7306 	/* If there's never been a scan, there's not much to say. */
7307 	if (ps == NULL || ps->pss_func == POOL_SCAN_NONE ||
7308 	    ps->pss_func >= POOL_SCAN_FUNCS) {
7309 		(void) printf(gettext("none requested\n"));
7310 		return;
7311 	}
7312 
7313 	start = ps->pss_start_time;
7314 	end = ps->pss_end_time;
7315 	pause = ps->pss_pass_scrub_pause;
7316 
7317 	zfs_nicebytes(ps->pss_processed, processed_buf, sizeof (processed_buf));
7318 
7319 	assert(ps->pss_func == POOL_SCAN_SCRUB ||
7320 	    ps->pss_func == POOL_SCAN_RESILVER);
7321 
7322 	/* Scan is finished or canceled. */
7323 	if (ps->pss_state == DSS_FINISHED) {
7324 		secs_to_dhms(end - start, time_buf);
7325 
7326 		if (ps->pss_func == POOL_SCAN_SCRUB) {
7327 			(void) printf(gettext("scrub repaired %s "
7328 			    "in %s with %llu errors on %s"), processed_buf,
7329 			    time_buf, (u_longlong_t)ps->pss_errors,
7330 			    ctime(&end));
7331 		} else if (ps->pss_func == POOL_SCAN_RESILVER) {
7332 			(void) printf(gettext("resilvered %s "
7333 			    "in %s with %llu errors on %s"), processed_buf,
7334 			    time_buf, (u_longlong_t)ps->pss_errors,
7335 			    ctime(&end));
7336 		}
7337 		return;
7338 	} else if (ps->pss_state == DSS_CANCELED) {
7339 		if (ps->pss_func == POOL_SCAN_SCRUB) {
7340 			(void) printf(gettext("scrub canceled on %s"),
7341 			    ctime(&end));
7342 		} else if (ps->pss_func == POOL_SCAN_RESILVER) {
7343 			(void) printf(gettext("resilver canceled on %s"),
7344 			    ctime(&end));
7345 		}
7346 		return;
7347 	}
7348 
7349 	assert(ps->pss_state == DSS_SCANNING);
7350 
7351 	/* Scan is in progress. Resilvers can't be paused. */
7352 	if (ps->pss_func == POOL_SCAN_SCRUB) {
7353 		if (pause == 0) {
7354 			(void) printf(gettext("scrub in progress since %s"),
7355 			    ctime(&start));
7356 		} else {
7357 			(void) printf(gettext("scrub paused since %s"),
7358 			    ctime(&pause));
7359 			(void) printf(gettext("\tscrub started on %s"),
7360 			    ctime(&start));
7361 		}
7362 	} else if (ps->pss_func == POOL_SCAN_RESILVER) {
7363 		(void) printf(gettext("resilver in progress since %s"),
7364 		    ctime(&start));
7365 	}
7366 
7367 	scanned = ps->pss_examined;
7368 	pass_scanned = ps->pss_pass_exam;
7369 	issued = ps->pss_issued;
7370 	pass_issued = ps->pss_pass_issued;
7371 	total = ps->pss_to_examine;
7372 
7373 	/* we are only done with a block once we have issued the IO for it */
7374 	fraction_done = (double)issued / total;
7375 
7376 	/* elapsed time for this pass, rounding up to 1 if it's 0 */
7377 	elapsed = time(NULL) - ps->pss_pass_start;
7378 	elapsed -= ps->pss_pass_scrub_spent_paused;
7379 	elapsed = (elapsed != 0) ? elapsed : 1;
7380 
7381 	scan_rate = pass_scanned / elapsed;
7382 	issue_rate = pass_issued / elapsed;
7383 	uint64_t total_secs_left = (issue_rate != 0 && total >= issued) ?
7384 	    ((total - issued) / issue_rate) : UINT64_MAX;
7385 	secs_to_dhms(total_secs_left, time_buf);
7386 
7387 	/* format all of the numbers we will be reporting */
7388 	zfs_nicebytes(scanned, scanned_buf, sizeof (scanned_buf));
7389 	zfs_nicebytes(issued, issued_buf, sizeof (issued_buf));
7390 	zfs_nicebytes(total, total_buf, sizeof (total_buf));
7391 	zfs_nicebytes(scan_rate, srate_buf, sizeof (srate_buf));
7392 	zfs_nicebytes(issue_rate, irate_buf, sizeof (irate_buf));
7393 
7394 	/* do not print estimated time if we have a paused scrub */
7395 	if (pause == 0) {
7396 		(void) printf(gettext("\t%s scanned at %s/s, "
7397 		    "%s issued at %s/s, %s total\n"),
7398 		    scanned_buf, srate_buf, issued_buf, irate_buf, total_buf);
7399 	} else {
7400 		(void) printf(gettext("\t%s scanned, %s issued, %s total\n"),
7401 		    scanned_buf, issued_buf, total_buf);
7402 	}
7403 
7404 	if (ps->pss_func == POOL_SCAN_RESILVER) {
7405 		(void) printf(gettext("\t%s resilvered, %.2f%% done"),
7406 		    processed_buf, 100 * fraction_done);
7407 	} else if (ps->pss_func == POOL_SCAN_SCRUB) {
7408 		(void) printf(gettext("\t%s repaired, %.2f%% done"),
7409 		    processed_buf, 100 * fraction_done);
7410 	}
7411 
7412 	if (pause == 0) {
7413 		if (total_secs_left != UINT64_MAX &&
7414 		    issue_rate >= 10 * 1024 * 1024) {
7415 			(void) printf(gettext(", %s to go\n"), time_buf);
7416 		} else {
7417 			(void) printf(gettext(", no estimated "
7418 			    "completion time\n"));
7419 		}
7420 	} else {
7421 		(void) printf(gettext("\n"));
7422 	}
7423 }
7424 
7425 static void
7426 print_rebuild_status_impl(vdev_rebuild_stat_t *vrs, char *vdev_name)
7427 {
7428 	if (vrs == NULL || vrs->vrs_state == VDEV_REBUILD_NONE)
7429 		return;
7430 
7431 	printf("  ");
7432 	printf_color(ANSI_BOLD, gettext("scan:"));
7433 	printf(" ");
7434 
7435 	uint64_t bytes_scanned = vrs->vrs_bytes_scanned;
7436 	uint64_t bytes_issued = vrs->vrs_bytes_issued;
7437 	uint64_t bytes_rebuilt = vrs->vrs_bytes_rebuilt;
7438 	uint64_t bytes_est = vrs->vrs_bytes_est;
7439 	uint64_t scan_rate = (vrs->vrs_pass_bytes_scanned /
7440 	    (vrs->vrs_pass_time_ms + 1)) * 1000;
7441 	uint64_t issue_rate = (vrs->vrs_pass_bytes_issued /
7442 	    (vrs->vrs_pass_time_ms + 1)) * 1000;
7443 	double scan_pct = MIN((double)bytes_scanned * 100 /
7444 	    (bytes_est + 1), 100);
7445 
7446 	/* Format all of the numbers we will be reporting */
7447 	char bytes_scanned_buf[7], bytes_issued_buf[7];
7448 	char bytes_rebuilt_buf[7], bytes_est_buf[7];
7449 	char scan_rate_buf[7], issue_rate_buf[7], time_buf[32];
7450 	zfs_nicebytes(bytes_scanned, bytes_scanned_buf,
7451 	    sizeof (bytes_scanned_buf));
7452 	zfs_nicebytes(bytes_issued, bytes_issued_buf,
7453 	    sizeof (bytes_issued_buf));
7454 	zfs_nicebytes(bytes_rebuilt, bytes_rebuilt_buf,
7455 	    sizeof (bytes_rebuilt_buf));
7456 	zfs_nicebytes(bytes_est, bytes_est_buf, sizeof (bytes_est_buf));
7457 	zfs_nicebytes(scan_rate, scan_rate_buf, sizeof (scan_rate_buf));
7458 	zfs_nicebytes(issue_rate, issue_rate_buf, sizeof (issue_rate_buf));
7459 
7460 	time_t start = vrs->vrs_start_time;
7461 	time_t end = vrs->vrs_end_time;
7462 
7463 	/* Rebuild is finished or canceled. */
7464 	if (vrs->vrs_state == VDEV_REBUILD_COMPLETE) {
7465 		secs_to_dhms(vrs->vrs_scan_time_ms / 1000, time_buf);
7466 		(void) printf(gettext("resilvered (%s) %s in %s "
7467 		    "with %llu errors on %s"), vdev_name, bytes_rebuilt_buf,
7468 		    time_buf, (u_longlong_t)vrs->vrs_errors, ctime(&end));
7469 		return;
7470 	} else if (vrs->vrs_state == VDEV_REBUILD_CANCELED) {
7471 		(void) printf(gettext("resilver (%s) canceled on %s"),
7472 		    vdev_name, ctime(&end));
7473 		return;
7474 	} else if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
7475 		(void) printf(gettext("resilver (%s) in progress since %s"),
7476 		    vdev_name, ctime(&start));
7477 	}
7478 
7479 	assert(vrs->vrs_state == VDEV_REBUILD_ACTIVE);
7480 
7481 	secs_to_dhms(MAX((int64_t)bytes_est - (int64_t)bytes_scanned, 0) /
7482 	    MAX(scan_rate, 1), time_buf);
7483 
7484 	(void) printf(gettext("\t%s scanned at %s/s, %s issued %s/s, "
7485 	    "%s total\n"), bytes_scanned_buf, scan_rate_buf,
7486 	    bytes_issued_buf, issue_rate_buf, bytes_est_buf);
7487 	(void) printf(gettext("\t%s resilvered, %.2f%% done"),
7488 	    bytes_rebuilt_buf, scan_pct);
7489 
7490 	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
7491 		if (scan_rate >= 10 * 1024 * 1024) {
7492 			(void) printf(gettext(", %s to go\n"), time_buf);
7493 		} else {
7494 			(void) printf(gettext(", no estimated "
7495 			    "completion time\n"));
7496 		}
7497 	} else {
7498 		(void) printf(gettext("\n"));
7499 	}
7500 }
7501 
7502 /*
7503  * Print rebuild status for top-level vdevs.
7504  */
7505 static void
7506 print_rebuild_status(zpool_handle_t *zhp, nvlist_t *nvroot)
7507 {
7508 	nvlist_t **child;
7509 	uint_t children;
7510 
7511 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
7512 	    &child, &children) != 0)
7513 		children = 0;
7514 
7515 	for (uint_t c = 0; c < children; c++) {
7516 		vdev_rebuild_stat_t *vrs;
7517 		uint_t i;
7518 
7519 		if (nvlist_lookup_uint64_array(child[c],
7520 		    ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {
7521 			char *name = zpool_vdev_name(g_zfs, zhp,
7522 			    child[c], VDEV_NAME_TYPE_ID);
7523 			print_rebuild_status_impl(vrs, name);
7524 			free(name);
7525 		}
7526 	}
7527 }
7528 
7529 /*
7530  * As we don't scrub checkpointed blocks, we want to warn the user that we
7531  * skipped scanning some blocks if a checkpoint exists or existed at any
7532  * time during the scan.  If a sequential instead of healing reconstruction
7533  * was performed then the blocks were reconstructed.  However, their checksums
7534  * have not been verified so we still print the warning.
7535  */
7536 static void
7537 print_checkpoint_scan_warning(pool_scan_stat_t *ps, pool_checkpoint_stat_t *pcs)
7538 {
7539 	if (ps == NULL || pcs == NULL)
7540 		return;
7541 
7542 	if (pcs->pcs_state == CS_NONE ||
7543 	    pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
7544 		return;
7545 
7546 	assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS);
7547 
7548 	if (ps->pss_state == DSS_NONE)
7549 		return;
7550 
7551 	if ((ps->pss_state == DSS_FINISHED || ps->pss_state == DSS_CANCELED) &&
7552 	    ps->pss_end_time < pcs->pcs_start_time)
7553 		return;
7554 
7555 	if (ps->pss_state == DSS_FINISHED || ps->pss_state == DSS_CANCELED) {
7556 		(void) printf(gettext("    scan warning: skipped blocks "
7557 		    "that are only referenced by the checkpoint.\n"));
7558 	} else {
7559 		assert(ps->pss_state == DSS_SCANNING);
7560 		(void) printf(gettext("    scan warning: skipping blocks "
7561 		    "that are only referenced by the checkpoint.\n"));
7562 	}
7563 }
7564 
7565 /*
7566  * Returns B_TRUE if there is an active rebuild in progress.  Otherwise,
7567  * B_FALSE is returned and 'rebuild_end_time' is set to the end time for
7568  * the last completed (or cancelled) rebuild.
7569  */
7570 static boolean_t
7571 check_rebuilding(nvlist_t *nvroot, uint64_t *rebuild_end_time)
7572 {
7573 	nvlist_t **child;
7574 	uint_t children;
7575 	boolean_t rebuilding = B_FALSE;
7576 	uint64_t end_time = 0;
7577 
7578 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
7579 	    &child, &children) != 0)
7580 		children = 0;
7581 
7582 	for (uint_t c = 0; c < children; c++) {
7583 		vdev_rebuild_stat_t *vrs;
7584 		uint_t i;
7585 
7586 		if (nvlist_lookup_uint64_array(child[c],
7587 		    ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {
7588 
7589 			if (vrs->vrs_end_time > end_time)
7590 				end_time = vrs->vrs_end_time;
7591 
7592 			if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
7593 				rebuilding = B_TRUE;
7594 				end_time = 0;
7595 				break;
7596 			}
7597 		}
7598 	}
7599 
7600 	if (rebuild_end_time != NULL)
7601 		*rebuild_end_time = end_time;
7602 
7603 	return (rebuilding);
7604 }
7605 
7606 /*
7607  * Print the scan status.
7608  */
7609 static void
7610 print_scan_status(zpool_handle_t *zhp, nvlist_t *nvroot)
7611 {
7612 	uint64_t rebuild_end_time = 0, resilver_end_time = 0;
7613 	boolean_t have_resilver = B_FALSE, have_scrub = B_FALSE;
7614 	boolean_t active_resilver = B_FALSE;
7615 	pool_checkpoint_stat_t *pcs = NULL;
7616 	pool_scan_stat_t *ps = NULL;
7617 	uint_t c;
7618 
7619 	if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
7620 	    (uint64_t **)&ps, &c) == 0) {
7621 		if (ps->pss_func == POOL_SCAN_RESILVER) {
7622 			resilver_end_time = ps->pss_end_time;
7623 			active_resilver = (ps->pss_state == DSS_SCANNING);
7624 		}
7625 
7626 		have_resilver = (ps->pss_func == POOL_SCAN_RESILVER);
7627 		have_scrub = (ps->pss_func == POOL_SCAN_SCRUB);
7628 	}
7629 
7630 	boolean_t active_rebuild = check_rebuilding(nvroot, &rebuild_end_time);
7631 	boolean_t have_rebuild = (active_rebuild || (rebuild_end_time > 0));
7632 
7633 	/* Always print the scrub status when available. */
7634 	if (have_scrub)
7635 		print_scan_scrub_resilver_status(ps);
7636 
7637 	/*
7638 	 * When there is an active resilver or rebuild print its status.
7639 	 * Otherwise print the status of the last resilver or rebuild.
7640 	 */
7641 	if (active_resilver || (!active_rebuild && have_resilver &&
7642 	    resilver_end_time && resilver_end_time > rebuild_end_time)) {
7643 		print_scan_scrub_resilver_status(ps);
7644 	} else if (active_rebuild || (!active_resilver && have_rebuild &&
7645 	    rebuild_end_time && rebuild_end_time > resilver_end_time)) {
7646 		print_rebuild_status(zhp, nvroot);
7647 	}
7648 
7649 	(void) nvlist_lookup_uint64_array(nvroot,
7650 	    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
7651 	print_checkpoint_scan_warning(ps, pcs);
7652 }
7653 
7654 /*
7655  * Print out detailed removal status.
7656  */
7657 static void
7658 print_removal_status(zpool_handle_t *zhp, pool_removal_stat_t *prs)
7659 {
7660 	char copied_buf[7], examined_buf[7], total_buf[7], rate_buf[7];
7661 	time_t start, end;
7662 	nvlist_t *config, *nvroot;
7663 	nvlist_t **child;
7664 	uint_t children;
7665 	char *vdev_name;
7666 
7667 	if (prs == NULL || prs->prs_state == DSS_NONE)
7668 		return;
7669 
7670 	/*
7671 	 * Determine name of vdev.
7672 	 */
7673 	config = zpool_get_config(zhp, NULL);
7674 	nvroot = fnvlist_lookup_nvlist(config,
7675 	    ZPOOL_CONFIG_VDEV_TREE);
7676 	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
7677 	    &child, &children) == 0);
7678 	assert(prs->prs_removing_vdev < children);
7679 	vdev_name = zpool_vdev_name(g_zfs, zhp,
7680 	    child[prs->prs_removing_vdev], B_TRUE);
7681 
7682 	printf_color(ANSI_BOLD, gettext("remove: "));
7683 
7684 	start = prs->prs_start_time;
7685 	end = prs->prs_end_time;
7686 	zfs_nicenum(prs->prs_copied, copied_buf, sizeof (copied_buf));
7687 
7688 	/*
7689 	 * Removal is finished or canceled.
7690 	 */
7691 	if (prs->prs_state == DSS_FINISHED) {
7692 		uint64_t minutes_taken = (end - start) / 60;
7693 
7694 		(void) printf(gettext("Removal of vdev %llu copied %s "
7695 		    "in %lluh%um, completed on %s"),
7696 		    (longlong_t)prs->prs_removing_vdev,
7697 		    copied_buf,
7698 		    (u_longlong_t)(minutes_taken / 60),
7699 		    (uint_t)(minutes_taken % 60),
7700 		    ctime((time_t *)&end));
7701 	} else if (prs->prs_state == DSS_CANCELED) {
7702 		(void) printf(gettext("Removal of %s canceled on %s"),
7703 		    vdev_name, ctime(&end));
7704 	} else {
7705 		uint64_t copied, total, elapsed, mins_left, hours_left;
7706 		double fraction_done;
7707 		uint_t rate;
7708 
7709 		assert(prs->prs_state == DSS_SCANNING);
7710 
7711 		/*
7712 		 * Removal is in progress.
7713 		 */
7714 		(void) printf(gettext(
7715 		    "Evacuation of %s in progress since %s"),
7716 		    vdev_name, ctime(&start));
7717 
7718 		copied = prs->prs_copied > 0 ? prs->prs_copied : 1;
7719 		total = prs->prs_to_copy;
7720 		fraction_done = (double)copied / total;
7721 
7722 		/* elapsed time for this pass */
7723 		elapsed = time(NULL) - prs->prs_start_time;
7724 		elapsed = elapsed > 0 ? elapsed : 1;
7725 		rate = copied / elapsed;
7726 		rate = rate > 0 ? rate : 1;
7727 		mins_left = ((total - copied) / rate) / 60;
7728 		hours_left = mins_left / 60;
7729 
7730 		zfs_nicenum(copied, examined_buf, sizeof (examined_buf));
7731 		zfs_nicenum(total, total_buf, sizeof (total_buf));
7732 		zfs_nicenum(rate, rate_buf, sizeof (rate_buf));
7733 
7734 		/*
7735 		 * do not print estimated time if hours_left is more than
7736 		 * 30 days
7737 		 */
7738 		(void) printf(gettext("    %s copied out of %s at %s/s, "
7739 		    "%.2f%% done"),
7740 		    examined_buf, total_buf, rate_buf, 100 * fraction_done);
7741 		if (hours_left < (30 * 24)) {
7742 			(void) printf(gettext(", %lluh%um to go\n"),
7743 			    (u_longlong_t)hours_left, (uint_t)(mins_left % 60));
7744 		} else {
7745 			(void) printf(gettext(
7746 			    ", (copy is slow, no estimated time)\n"));
7747 		}
7748 	}
7749 	free(vdev_name);
7750 
7751 	if (prs->prs_mapping_memory > 0) {
7752 		char mem_buf[7];
7753 		zfs_nicenum(prs->prs_mapping_memory, mem_buf, sizeof (mem_buf));
7754 		(void) printf(gettext("    %s memory used for "
7755 		    "removed device mappings\n"),
7756 		    mem_buf);
7757 	}
7758 }
7759 
7760 static void
7761 print_checkpoint_status(pool_checkpoint_stat_t *pcs)
7762 {
7763 	time_t start;
7764 	char space_buf[7];
7765 
7766 	if (pcs == NULL || pcs->pcs_state == CS_NONE)
7767 		return;
7768 
7769 	(void) printf(gettext("checkpoint: "));
7770 
7771 	start = pcs->pcs_start_time;
7772 	zfs_nicenum(pcs->pcs_space, space_buf, sizeof (space_buf));
7773 
7774 	if (pcs->pcs_state == CS_CHECKPOINT_EXISTS) {
7775 		char *date = ctime(&start);
7776 
7777 		/*
7778 		 * ctime() adds a newline at the end of the generated
7779 		 * string, thus the weird format specifier and the
7780 		 * strlen() call used to chop it off from the output.
7781 		 */
7782 		(void) printf(gettext("created %.*s, consumes %s\n"),
7783 		    (int)(strlen(date) - 1), date, space_buf);
7784 		return;
7785 	}
7786 
7787 	assert(pcs->pcs_state == CS_CHECKPOINT_DISCARDING);
7788 
7789 	(void) printf(gettext("discarding, %s remaining.\n"),
7790 	    space_buf);
7791 }
7792 
7793 static void
7794 print_error_log(zpool_handle_t *zhp)
7795 {
7796 	nvlist_t *nverrlist = NULL;
7797 	nvpair_t *elem;
7798 	char *pathname;
7799 	size_t len = MAXPATHLEN * 2;
7800 
7801 	if (zpool_get_errlog(zhp, &nverrlist) != 0)
7802 		return;
7803 
7804 	(void) printf("errors: Permanent errors have been "
7805 	    "detected in the following files:\n\n");
7806 
7807 	pathname = safe_malloc(len);
7808 	elem = NULL;
7809 	while ((elem = nvlist_next_nvpair(nverrlist, elem)) != NULL) {
7810 		nvlist_t *nv;
7811 		uint64_t dsobj, obj;
7812 
7813 		verify(nvpair_value_nvlist(elem, &nv) == 0);
7814 		verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_DATASET,
7815 		    &dsobj) == 0);
7816 		verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_OBJECT,
7817 		    &obj) == 0);
7818 		zpool_obj_to_path(zhp, dsobj, obj, pathname, len);
7819 		(void) printf("%7s %s\n", "", pathname);
7820 	}
7821 	free(pathname);
7822 	nvlist_free(nverrlist);
7823 }
7824 
7825 static void
7826 print_spares(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t **spares,
7827     uint_t nspares)
7828 {
7829 	uint_t i;
7830 	char *name;
7831 
7832 	if (nspares == 0)
7833 		return;
7834 
7835 	(void) printf(gettext("\tspares\n"));
7836 
7837 	for (i = 0; i < nspares; i++) {
7838 		name = zpool_vdev_name(g_zfs, zhp, spares[i],
7839 		    cb->cb_name_flags);
7840 		print_status_config(zhp, cb, name, spares[i], 2, B_TRUE, NULL);
7841 		free(name);
7842 	}
7843 }
7844 
7845 static void
7846 print_l2cache(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t **l2cache,
7847     uint_t nl2cache)
7848 {
7849 	uint_t i;
7850 	char *name;
7851 
7852 	if (nl2cache == 0)
7853 		return;
7854 
7855 	(void) printf(gettext("\tcache\n"));
7856 
7857 	for (i = 0; i < nl2cache; i++) {
7858 		name = zpool_vdev_name(g_zfs, zhp, l2cache[i],
7859 		    cb->cb_name_flags);
7860 		print_status_config(zhp, cb, name, l2cache[i], 2,
7861 		    B_FALSE, NULL);
7862 		free(name);
7863 	}
7864 }
7865 
7866 static void
7867 print_dedup_stats(nvlist_t *config)
7868 {
7869 	ddt_histogram_t *ddh;
7870 	ddt_stat_t *dds;
7871 	ddt_object_t *ddo;
7872 	uint_t c;
7873 	char dspace[6], mspace[6];
7874 
7875 	/*
7876 	 * If the pool was faulted then we may not have been able to
7877 	 * obtain the config. Otherwise, if we have anything in the dedup
7878 	 * table continue processing the stats.
7879 	 */
7880 	if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_OBJ_STATS,
7881 	    (uint64_t **)&ddo, &c) != 0)
7882 		return;
7883 
7884 	(void) printf("\n");
7885 	(void) printf(gettext(" dedup: "));
7886 	if (ddo->ddo_count == 0) {
7887 		(void) printf(gettext("no DDT entries\n"));
7888 		return;
7889 	}
7890 
7891 	zfs_nicebytes(ddo->ddo_dspace, dspace, sizeof (dspace));
7892 	zfs_nicebytes(ddo->ddo_mspace, mspace, sizeof (mspace));
7893 	(void) printf("DDT entries %llu, size %s on disk, %s in core\n",
7894 	    (u_longlong_t)ddo->ddo_count,
7895 	    dspace,
7896 	    mspace);
7897 
7898 	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_STATS,
7899 	    (uint64_t **)&dds, &c) == 0);
7900 	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_HISTOGRAM,
7901 	    (uint64_t **)&ddh, &c) == 0);
7902 	zpool_dump_ddt(dds, ddh);
7903 }
7904 
7905 /*
7906  * Display a summary of pool status.  Displays a summary such as:
7907  *
7908  *        pool: tank
7909  *	status: DEGRADED
7910  *	reason: One or more devices ...
7911  *         see: https://openzfs.github.io/openzfs-docs/msg/ZFS-xxxx-01
7912  *	config:
7913  *		mirror		DEGRADED
7914  *                c1t0d0	OK
7915  *                c2t0d0	UNAVAIL
7916  *
7917  * When given the '-v' option, we print out the complete config.  If the '-e'
7918  * option is specified, then we print out error rate information as well.
7919  */
7920 static int
7921 status_callback(zpool_handle_t *zhp, void *data)
7922 {
7923 	status_cbdata_t *cbp = data;
7924 	nvlist_t *config, *nvroot;
7925 	char *msgid;
7926 	zpool_status_t reason;
7927 	zpool_errata_t errata;
7928 	const char *health;
7929 	uint_t c;
7930 	vdev_stat_t *vs;
7931 
7932 	config = zpool_get_config(zhp, NULL);
7933 	reason = zpool_get_status(zhp, &msgid, &errata);
7934 
7935 	cbp->cb_count++;
7936 
7937 	/*
7938 	 * If we were given 'zpool status -x', only report those pools with
7939 	 * problems.
7940 	 */
7941 	if (cbp->cb_explain &&
7942 	    (reason == ZPOOL_STATUS_OK ||
7943 	    reason == ZPOOL_STATUS_VERSION_OLDER ||
7944 	    reason == ZPOOL_STATUS_FEAT_DISABLED)) {
7945 		if (!cbp->cb_allpools) {
7946 			(void) printf(gettext("pool '%s' is healthy\n"),
7947 			    zpool_get_name(zhp));
7948 			if (cbp->cb_first)
7949 				cbp->cb_first = B_FALSE;
7950 		}
7951 		return (0);
7952 	}
7953 
7954 	if (cbp->cb_first)
7955 		cbp->cb_first = B_FALSE;
7956 	else
7957 		(void) printf("\n");
7958 
7959 	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
7960 	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
7961 	    (uint64_t **)&vs, &c) == 0);
7962 
7963 	health = zpool_get_state_str(zhp);
7964 
7965 	printf("  ");
7966 	printf_color(ANSI_BOLD, gettext("pool:"));
7967 	printf(" %s\n", zpool_get_name(zhp));
7968 	printf(" ");
7969 	printf_color(ANSI_BOLD, gettext("state: "));
7970 
7971 	printf_color(health_str_to_color(health), "%s", health);
7972 
7973 	printf("\n");
7974 
7975 	switch (reason) {
7976 	case ZPOOL_STATUS_MISSING_DEV_R:
7977 		printf_color(ANSI_BOLD, gettext("status: "));
7978 		printf_color(ANSI_YELLOW, gettext("One or more devices could "
7979 		    "not be opened.  Sufficient replicas exist for\n\tthe pool "
7980 		    "to continue functioning in a degraded state.\n"));
7981 		printf_color(ANSI_BOLD, gettext("action: "));
7982 		printf_color(ANSI_YELLOW, gettext("Attach the missing device "
7983 		    "and online it using 'zpool online'.\n"));
7984 		break;
7985 
7986 	case ZPOOL_STATUS_MISSING_DEV_NR:
7987 		printf_color(ANSI_BOLD, gettext("status: "));
7988 		printf_color(ANSI_YELLOW, gettext("One or more devices could "
7989 		    "not be opened.  There are insufficient\n\treplicas for the"
7990 		    " pool to continue functioning.\n"));
7991 		printf_color(ANSI_BOLD, gettext("action: "));
7992 		printf_color(ANSI_YELLOW, gettext("Attach the missing device "
7993 		    "and online it using 'zpool online'.\n"));
7994 		break;
7995 
7996 	case ZPOOL_STATUS_CORRUPT_LABEL_R:
7997 		printf_color(ANSI_BOLD, gettext("status: "));
7998 		printf_color(ANSI_YELLOW, gettext("One or more devices could "
7999 		    "not be used because the label is missing or\n\tinvalid.  "
8000 		    "Sufficient replicas exist for the pool to continue\n\t"
8001 		    "functioning in a degraded state.\n"));
8002 		printf_color(ANSI_BOLD, gettext("action: "));
8003 		printf_color(ANSI_YELLOW, gettext("Replace the device using "
8004 		    "'zpool replace'.\n"));
8005 		break;
8006 
8007 	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
8008 		printf_color(ANSI_BOLD, gettext("status: "));
8009 		printf_color(ANSI_YELLOW, gettext("One or more devices could "
8010 		    "not be used because the label is missing \n\tor invalid.  "
8011 		    "There are insufficient replicas for the pool to "
8012 		    "continue\n\tfunctioning.\n"));
8013 		zpool_explain_recover(zpool_get_handle(zhp),
8014 		    zpool_get_name(zhp), reason, config);
8015 		break;
8016 
8017 	case ZPOOL_STATUS_FAILING_DEV:
8018 		printf_color(ANSI_BOLD, gettext("status: "));
8019 		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8020 		    "experienced an unrecoverable error.  An\n\tattempt was "
8021 		    "made to correct the error.  Applications are "
8022 		    "unaffected.\n"));
8023 		printf_color(ANSI_BOLD, gettext("action: "));
8024 			printf_color(ANSI_YELLOW, gettext("Determine if the "
8025 		    "device needs to be replaced, and clear the errors\n\tusing"
8026 		    " 'zpool clear' or replace the device with 'zpool "
8027 		    "replace'.\n"));
8028 		break;
8029 
8030 	case ZPOOL_STATUS_OFFLINE_DEV:
8031 		printf_color(ANSI_BOLD, gettext("status: "));
8032 		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8033 		    "been taken offline by the administrator.\n\tSufficient "
8034 		    "replicas exist for the pool to continue functioning in "
8035 		    "a\n\tdegraded state.\n"));
8036 		printf_color(ANSI_BOLD, gettext("action: "));
8037 		printf_color(ANSI_YELLOW, gettext("Online the device "
8038 		    "using 'zpool online' or replace the device with\n\t'zpool "
8039 		    "replace'.\n"));
8040 		break;
8041 
8042 	case ZPOOL_STATUS_REMOVED_DEV:
8043 		printf_color(ANSI_BOLD, gettext("status: "));
8044 		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8045 		    "been removed by the administrator.\n\tSufficient "
8046 		    "replicas exist for the pool to continue functioning in "
8047 		    "a\n\tdegraded state.\n"));
8048 		printf_color(ANSI_BOLD, gettext("action: "));
8049 		printf_color(ANSI_YELLOW, gettext("Online the device "
8050 		    "using zpool online' or replace the device with\n\t'zpool "
8051 		    "replace'.\n"));
8052 		break;
8053 
8054 	case ZPOOL_STATUS_RESILVERING:
8055 	case ZPOOL_STATUS_REBUILDING:
8056 		printf_color(ANSI_BOLD, gettext("status: "));
8057 		printf_color(ANSI_YELLOW, gettext("One or more devices is "
8058 		    "currently being resilvered.  The pool will\n\tcontinue "
8059 		    "to function, possibly in a degraded state.\n"));
8060 		printf_color(ANSI_BOLD, gettext("action: "));
8061 		printf_color(ANSI_YELLOW, gettext("Wait for the resilver to "
8062 		    "complete.\n"));
8063 		break;
8064 
8065 	case ZPOOL_STATUS_REBUILD_SCRUB:
8066 		printf_color(ANSI_BOLD, gettext("status: "));
8067 		printf_color(ANSI_YELLOW, gettext("One or more devices have "
8068 		    "been sequentially resilvered, scrubbing\n\tthe pool "
8069 		    "is recommended.\n"));
8070 		printf_color(ANSI_BOLD, gettext("action: "));
8071 		printf_color(ANSI_YELLOW, gettext("Use 'zpool scrub' to "
8072 		    "verify all data checksums.\n"));
8073 		break;
8074 
8075 	case ZPOOL_STATUS_CORRUPT_DATA:
8076 		printf_color(ANSI_BOLD, gettext("status: "));
8077 		printf_color(ANSI_YELLOW, gettext("One or more devices has "
8078 		    "experienced an error resulting in data\n\tcorruption.  "
8079 		    "Applications may be affected.\n"));
8080 		printf_color(ANSI_BOLD, gettext("action: "));
8081 		printf_color(ANSI_YELLOW, gettext("Restore the file in question"
8082 		    " if possible.  Otherwise restore the\n\tentire pool from "
8083 		    "backup.\n"));
8084 		break;
8085 
8086 	case ZPOOL_STATUS_CORRUPT_POOL:
8087 		printf_color(ANSI_BOLD, gettext("status: "));
8088 		printf_color(ANSI_YELLOW, gettext("The pool metadata is "
8089 		    "corrupted and the pool cannot be opened.\n"));
8090 		zpool_explain_recover(zpool_get_handle(zhp),
8091 		    zpool_get_name(zhp), reason, config);
8092 		break;
8093 
8094 	case ZPOOL_STATUS_VERSION_OLDER:
8095 		printf_color(ANSI_BOLD, gettext("status: "));
8096 		printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
8097 		    "a legacy on-disk format.  The pool can\n\tstill be used, "
8098 		    "but some features are unavailable.\n"));
8099 		printf_color(ANSI_BOLD, gettext("action: "));
8100 		printf_color(ANSI_YELLOW, gettext("Upgrade the pool using "
8101 		    "'zpool upgrade'.  Once this is done, the\n\tpool will no "
8102 		    "longer be accessible on software that does not support\n\t"
8103 		    "feature flags.\n"));
8104 		break;
8105 
8106 	case ZPOOL_STATUS_VERSION_NEWER:
8107 		printf_color(ANSI_BOLD, gettext("status: "));
8108 		printf_color(ANSI_YELLOW, gettext("The pool has been upgraded "
8109 		    "to a newer, incompatible on-disk version.\n\tThe pool "
8110 		    "cannot be accessed on this system.\n"));
8111 		printf_color(ANSI_BOLD, gettext("action: "));
8112 		printf_color(ANSI_YELLOW, gettext("Access the pool from a "
8113 		    "system running more recent software, or\n\trestore the "
8114 		    "pool from backup.\n"));
8115 		break;
8116 
8117 	case ZPOOL_STATUS_FEAT_DISABLED:
8118 		printf_color(ANSI_BOLD, gettext("status: "));
8119 		printf_color(ANSI_YELLOW, gettext("Some supported features are "
8120 		    "not enabled on the pool. The pool can\n\tstill be used, "
8121 		    "but some features are unavailable.\n"));
8122 		printf_color(ANSI_BOLD, gettext("action: "));
8123 		printf_color(ANSI_YELLOW, gettext("Enable all features using "
8124 		    "'zpool upgrade'. Once this is done,\n\tthe pool may no "
8125 		    "longer be accessible by software that does not support\n\t"
8126 		    "the features. See zpool-features(5) for details.\n"));
8127 		break;
8128 
8129 	case ZPOOL_STATUS_UNSUP_FEAT_READ:
8130 		printf_color(ANSI_BOLD, gettext("status: "));
8131 		printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
8132 		    "on this system because it uses the\n\tfollowing feature(s)"
8133 		    " not supported on this system:\n"));
8134 		zpool_print_unsup_feat(config);
8135 		(void) printf("\n");
8136 		printf_color(ANSI_BOLD, gettext("action: "));
8137 		printf_color(ANSI_YELLOW, gettext("Access the pool from a "
8138 		    "system that supports the required feature(s),\n\tor "
8139 		    "restore the pool from backup.\n"));
8140 		break;
8141 
8142 	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
8143 		printf_color(ANSI_BOLD, gettext("status: "));
8144 		printf_color(ANSI_YELLOW, gettext("The pool can only be "
8145 		    "accessed in read-only mode on this system. It\n\tcannot be"
8146 		    " accessed in read-write mode because it uses the "
8147 		    "following\n\tfeature(s) not supported on this system:\n"));
8148 		zpool_print_unsup_feat(config);
8149 		(void) printf("\n");
8150 		printf_color(ANSI_BOLD, gettext("action: "));
8151 		printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
8152 		    "in read-write mode. Import the pool with\n"
8153 		    "\t\"-o readonly=on\", access the pool from a system that "
8154 		    "supports the\n\trequired feature(s), or restore the "
8155 		    "pool from backup.\n"));
8156 		break;
8157 
8158 	case ZPOOL_STATUS_FAULTED_DEV_R:
8159 		printf_color(ANSI_BOLD, gettext("status: "));
8160 		printf_color(ANSI_YELLOW, gettext("One or more devices are "
8161 		    "faulted in response to persistent errors.\n\tSufficient "
8162 		    "replicas exist for the pool to continue functioning "
8163 		    "in a\n\tdegraded state.\n"));
8164 		printf_color(ANSI_BOLD, gettext("action: "));
8165 		printf_color(ANSI_YELLOW, gettext("Replace the faulted device, "
8166 		    "or use 'zpool clear' to mark the device\n\trepaired.\n"));
8167 		break;
8168 
8169 	case ZPOOL_STATUS_FAULTED_DEV_NR:
8170 		printf_color(ANSI_BOLD, gettext("status: "));
8171 		printf_color(ANSI_YELLOW, gettext("One or more devices are "
8172 		    "faulted in response to persistent errors.  There are "
8173 		    "insufficient replicas for the pool to\n\tcontinue "
8174 		    "functioning.\n"));
8175 		printf_color(ANSI_BOLD, gettext("action: "));
8176 		printf_color(ANSI_YELLOW, gettext("Destroy and re-create the "
8177 		    "pool from a backup source.  Manually marking the device\n"
8178 		    "\trepaired using 'zpool clear' may allow some data "
8179 		    "to be recovered.\n"));
8180 		break;
8181 
8182 	case ZPOOL_STATUS_IO_FAILURE_MMP:
8183 		printf_color(ANSI_BOLD, gettext("status: "));
8184 		printf_color(ANSI_YELLOW, gettext("The pool is suspended "
8185 		    "because multihost writes failed or were delayed;\n\t"
8186 		    "another system could import the pool undetected.\n"));
8187 		printf_color(ANSI_BOLD, gettext("action: "));
8188 		printf_color(ANSI_YELLOW, gettext("Make sure the pool's devices"
8189 		    " are connected, then reboot your system and\n\timport the "
8190 		    "pool.\n"));
8191 		break;
8192 
8193 	case ZPOOL_STATUS_IO_FAILURE_WAIT:
8194 	case ZPOOL_STATUS_IO_FAILURE_CONTINUE:
8195 		printf_color(ANSI_BOLD, gettext("status: "));
8196 		printf_color(ANSI_YELLOW, gettext("One or more devices are "
8197 		    "faulted in response to IO failures.\n"));
8198 		printf_color(ANSI_BOLD, gettext("action: "));
8199 		printf_color(ANSI_YELLOW, gettext("Make sure the affected "
8200 		    "devices are connected, then run 'zpool clear'.\n"));
8201 		break;
8202 
8203 	case ZPOOL_STATUS_BAD_LOG:
8204 		printf_color(ANSI_BOLD, gettext("status: "));
8205 		printf_color(ANSI_YELLOW, gettext("An intent log record "
8206 		    "could not be read.\n"
8207 		    "\tWaiting for administrator intervention to fix the "
8208 		    "faulted pool.\n"));
8209 		printf_color(ANSI_BOLD, gettext("action: "));
8210 		printf_color(ANSI_YELLOW, gettext("Either restore the affected "
8211 		    "device(s) and run 'zpool online',\n"
8212 		    "\tor ignore the intent log records by running "
8213 		    "'zpool clear'.\n"));
8214 		break;
8215 
8216 	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
8217 		(void) printf(gettext("status: One or more devices are "
8218 		    "configured to use a non-native block size.\n"
8219 		    "\tExpect reduced performance.\n"));
8220 		(void) printf(gettext("action: Replace affected devices with "
8221 		    "devices that support the\n\tconfigured block size, or "
8222 		    "migrate data to a properly configured\n\tpool.\n"));
8223 		break;
8224 
8225 	case ZPOOL_STATUS_HOSTID_MISMATCH:
8226 		printf_color(ANSI_BOLD, gettext("status: "));
8227 		printf_color(ANSI_YELLOW, gettext("Mismatch between pool hostid"
8228 		    " and system hostid on imported pool.\n\tThis pool was "
8229 		    "previously imported into a system with a different "
8230 		    "hostid,\n\tand then was verbatim imported into this "
8231 		    "system.\n"));
8232 		printf_color(ANSI_BOLD, gettext("action: "));
8233 		printf_color(ANSI_YELLOW, gettext("Export this pool on all "
8234 		    "systems on which it is imported.\n"
8235 		    "\tThen import it to correct the mismatch.\n"));
8236 		break;
8237 
8238 	case ZPOOL_STATUS_ERRATA:
8239 		printf_color(ANSI_BOLD, gettext("status: "));
8240 		printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
8241 		    errata);
8242 
8243 		switch (errata) {
8244 		case ZPOOL_ERRATA_NONE:
8245 			break;
8246 
8247 		case ZPOOL_ERRATA_ZOL_2094_SCRUB:
8248 			printf_color(ANSI_BOLD, gettext("action: "));
8249 			printf_color(ANSI_YELLOW, gettext("To correct the issue"
8250 			    " run 'zpool scrub'.\n"));
8251 			break;
8252 
8253 		case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
8254 			(void) printf(gettext("\tExisting encrypted datasets "
8255 			    "contain an on-disk incompatibility\n\twhich "
8256 			    "needs to be corrected.\n"));
8257 			printf_color(ANSI_BOLD, gettext("action: "));
8258 			printf_color(ANSI_YELLOW, gettext("To correct the issue"
8259 			    " backup existing encrypted datasets to new\n\t"
8260 			    "encrypted datasets and destroy the old ones. "
8261 			    "'zfs mount -o ro' can\n\tbe used to temporarily "
8262 			    "mount existing encrypted datasets readonly.\n"));
8263 			break;
8264 
8265 		case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
8266 			(void) printf(gettext("\tExisting encrypted snapshots "
8267 			    "and bookmarks contain an on-disk\n\tincompat"
8268 			    "ibility. This may cause on-disk corruption if "
8269 			    "they are used\n\twith 'zfs recv'.\n"));
8270 			printf_color(ANSI_BOLD, gettext("action: "));
8271 			printf_color(ANSI_YELLOW, gettext("To correct the"
8272 			    "issue, enable the bookmark_v2 feature. No "
8273 			    "additional\n\taction is needed if there are no "
8274 			    "encrypted snapshots or bookmarks.\n\tIf preserving"
8275 			    "the encrypted snapshots and bookmarks is required,"
8276 			    " use\n\ta non-raw send to backup and restore them."
8277 			    " Alternately, they may be\n\tremoved to resolve "
8278 			    "the incompatibility.\n"));
8279 			break;
8280 
8281 		default:
8282 			/*
8283 			 * All errata which allow the pool to be imported
8284 			 * must contain an action message.
8285 			 */
8286 			assert(0);
8287 		}
8288 		break;
8289 
8290 	default:
8291 		/*
8292 		 * The remaining errors can't actually be generated, yet.
8293 		 */
8294 		assert(reason == ZPOOL_STATUS_OK);
8295 	}
8296 
8297 	if (msgid != NULL) {
8298 		printf("   ");
8299 		printf_color(ANSI_BOLD, gettext("see:"));
8300 		printf(gettext(
8301 		    " https://openzfs.github.io/openzfs-docs/msg/%s\n"),
8302 		    msgid);
8303 	}
8304 
8305 	if (config != NULL) {
8306 		uint64_t nerr;
8307 		nvlist_t **spares, **l2cache;
8308 		uint_t nspares, nl2cache;
8309 		pool_checkpoint_stat_t *pcs = NULL;
8310 		pool_removal_stat_t *prs = NULL;
8311 
8312 		print_scan_status(zhp, nvroot);
8313 
8314 		(void) nvlist_lookup_uint64_array(nvroot,
8315 		    ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
8316 		print_removal_status(zhp, prs);
8317 
8318 		(void) nvlist_lookup_uint64_array(nvroot,
8319 		    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
8320 		print_checkpoint_status(pcs);
8321 
8322 		cbp->cb_namewidth = max_width(zhp, nvroot, 0, 0,
8323 		    cbp->cb_name_flags | VDEV_NAME_TYPE_ID);
8324 		if (cbp->cb_namewidth < 10)
8325 			cbp->cb_namewidth = 10;
8326 
8327 		color_start(ANSI_BOLD);
8328 		(void) printf(gettext("config:\n\n"));
8329 		(void) printf(gettext("\t%-*s  %-8s %5s %5s %5s"),
8330 		    cbp->cb_namewidth, "NAME", "STATE", "READ", "WRITE",
8331 		    "CKSUM");
8332 		color_end();
8333 
8334 		if (cbp->cb_print_slow_ios) {
8335 			printf_color(ANSI_BOLD, " %5s", gettext("SLOW"));
8336 		}
8337 
8338 		if (cbp->vcdl != NULL)
8339 			print_cmd_columns(cbp->vcdl, 0);
8340 
8341 		printf("\n");
8342 
8343 		print_status_config(zhp, cbp, zpool_get_name(zhp), nvroot, 0,
8344 		    B_FALSE, NULL);
8345 
8346 		print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_DEDUP);
8347 		print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
8348 		print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_CLASS_LOGS);
8349 
8350 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
8351 		    &l2cache, &nl2cache) == 0)
8352 			print_l2cache(zhp, cbp, l2cache, nl2cache);
8353 
8354 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
8355 		    &spares, &nspares) == 0)
8356 			print_spares(zhp, cbp, spares, nspares);
8357 
8358 		if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
8359 		    &nerr) == 0) {
8360 			nvlist_t *nverrlist = NULL;
8361 
8362 			/*
8363 			 * If the approximate error count is small, get a
8364 			 * precise count by fetching the entire log and
8365 			 * uniquifying the results.
8366 			 */
8367 			if (nerr > 0 && nerr < 100 && !cbp->cb_verbose &&
8368 			    zpool_get_errlog(zhp, &nverrlist) == 0) {
8369 				nvpair_t *elem;
8370 
8371 				elem = NULL;
8372 				nerr = 0;
8373 				while ((elem = nvlist_next_nvpair(nverrlist,
8374 				    elem)) != NULL) {
8375 					nerr++;
8376 				}
8377 			}
8378 			nvlist_free(nverrlist);
8379 
8380 			(void) printf("\n");
8381 
8382 			if (nerr == 0)
8383 				(void) printf(gettext("errors: No known data "
8384 				    "errors\n"));
8385 			else if (!cbp->cb_verbose)
8386 				(void) printf(gettext("errors: %llu data "
8387 				    "errors, use '-v' for a list\n"),
8388 				    (u_longlong_t)nerr);
8389 			else
8390 				print_error_log(zhp);
8391 		}
8392 
8393 		if (cbp->cb_dedup_stats)
8394 			print_dedup_stats(config);
8395 	} else {
8396 		(void) printf(gettext("config: The configuration cannot be "
8397 		    "determined.\n"));
8398 	}
8399 
8400 	return (0);
8401 }
8402 
8403 /*
8404  * zpool status [-c [script1,script2,...]] [-igLpPstvx] [-T d|u] [pool] ...
8405  *              [interval [count]]
8406  *
8407  *	-c CMD	For each vdev, run command CMD
8408  *	-i	Display vdev initialization status.
8409  *	-g	Display guid for individual vdev name.
8410  *	-L	Follow links when resolving vdev path name.
8411  *	-p	Display values in parsable (exact) format.
8412  *	-P	Display full path for vdev name.
8413  *	-s	Display slow IOs column.
8414  *	-v	Display complete error logs
8415  *	-x	Display only pools with potential problems
8416  *	-D	Display dedup status (undocumented)
8417  *	-t	Display vdev TRIM status.
8418  *	-T	Display a timestamp in date(1) or Unix format
8419  *
8420  * Describes the health status of all pools or some subset.
8421  */
8422 int
8423 zpool_do_status(int argc, char **argv)
8424 {
8425 	int c;
8426 	int ret;
8427 	float interval = 0;
8428 	unsigned long count = 0;
8429 	status_cbdata_t cb = { 0 };
8430 	char *cmd = NULL;
8431 
8432 	/* check options */
8433 	while ((c = getopt(argc, argv, "c:igLpPsvxDtT:")) != -1) {
8434 		switch (c) {
8435 		case 'c':
8436 			if (cmd != NULL) {
8437 				fprintf(stderr,
8438 				    gettext("Can't set -c flag twice\n"));
8439 				exit(1);
8440 			}
8441 
8442 			if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
8443 			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
8444 				fprintf(stderr, gettext(
8445 				    "Can't run -c, disabled by "
8446 				    "ZPOOL_SCRIPTS_ENABLED.\n"));
8447 				exit(1);
8448 			}
8449 
8450 			if ((getuid() <= 0 || geteuid() <= 0) &&
8451 			    !libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
8452 				fprintf(stderr, gettext(
8453 				    "Can't run -c with root privileges "
8454 				    "unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
8455 				exit(1);
8456 			}
8457 			cmd = optarg;
8458 			break;
8459 		case 'i':
8460 			cb.cb_print_vdev_init = B_TRUE;
8461 			break;
8462 		case 'g':
8463 			cb.cb_name_flags |= VDEV_NAME_GUID;
8464 			break;
8465 		case 'L':
8466 			cb.cb_name_flags |= VDEV_NAME_FOLLOW_LINKS;
8467 			break;
8468 		case 'p':
8469 			cb.cb_literal = B_TRUE;
8470 			break;
8471 		case 'P':
8472 			cb.cb_name_flags |= VDEV_NAME_PATH;
8473 			break;
8474 		case 's':
8475 			cb.cb_print_slow_ios = B_TRUE;
8476 			break;
8477 		case 'v':
8478 			cb.cb_verbose = B_TRUE;
8479 			break;
8480 		case 'x':
8481 			cb.cb_explain = B_TRUE;
8482 			break;
8483 		case 'D':
8484 			cb.cb_dedup_stats = B_TRUE;
8485 			break;
8486 		case 't':
8487 			cb.cb_print_vdev_trim = B_TRUE;
8488 			break;
8489 		case 'T':
8490 			get_timestamp_arg(*optarg);
8491 			break;
8492 		case '?':
8493 			if (optopt == 'c') {
8494 				print_zpool_script_list("status");
8495 				exit(0);
8496 			} else {
8497 				fprintf(stderr,
8498 				    gettext("invalid option '%c'\n"), optopt);
8499 			}
8500 			usage(B_FALSE);
8501 		}
8502 	}
8503 
8504 	argc -= optind;
8505 	argv += optind;
8506 
8507 	get_interval_count(&argc, argv, &interval, &count);
8508 
8509 	if (argc == 0)
8510 		cb.cb_allpools = B_TRUE;
8511 
8512 	cb.cb_first = B_TRUE;
8513 	cb.cb_print_status = B_TRUE;
8514 
8515 	for (;;) {
8516 		if (timestamp_fmt != NODATE)
8517 			print_timestamp(timestamp_fmt);
8518 
8519 		if (cmd != NULL)
8520 			cb.vcdl = all_pools_for_each_vdev_run(argc, argv, cmd,
8521 			    NULL, NULL, 0, 0);
8522 
8523 		ret = for_each_pool(argc, argv, B_TRUE, NULL, cb.cb_literal,
8524 		    status_callback, &cb);
8525 
8526 		if (cb.vcdl != NULL)
8527 			free_vdev_cmd_data_list(cb.vcdl);
8528 
8529 		if (argc == 0 && cb.cb_count == 0)
8530 			(void) fprintf(stderr, gettext("no pools available\n"));
8531 		else if (cb.cb_explain && cb.cb_first && cb.cb_allpools)
8532 			(void) printf(gettext("all pools are healthy\n"));
8533 
8534 		if (ret != 0)
8535 			return (ret);
8536 
8537 		if (interval == 0)
8538 			break;
8539 
8540 		if (count != 0 && --count == 0)
8541 			break;
8542 
8543 		(void) fsleep(interval);
8544 	}
8545 
8546 	return (0);
8547 }
8548 
8549 typedef struct upgrade_cbdata {
8550 	int	cb_first;
8551 	int	cb_argc;
8552 	uint64_t cb_version;
8553 	char	**cb_argv;
8554 } upgrade_cbdata_t;
8555 
8556 static int
8557 check_unsupp_fs(zfs_handle_t *zhp, void *unsupp_fs)
8558 {
8559 	int zfs_version = (int)zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
8560 	int *count = (int *)unsupp_fs;
8561 
8562 	if (zfs_version > ZPL_VERSION) {
8563 		(void) printf(gettext("%s (v%d) is not supported by this "
8564 		    "implementation of ZFS.\n"),
8565 		    zfs_get_name(zhp), zfs_version);
8566 		(*count)++;
8567 	}
8568 
8569 	zfs_iter_filesystems(zhp, check_unsupp_fs, unsupp_fs);
8570 
8571 	zfs_close(zhp);
8572 
8573 	return (0);
8574 }
8575 
8576 static int
8577 upgrade_version(zpool_handle_t *zhp, uint64_t version)
8578 {
8579 	int ret;
8580 	nvlist_t *config;
8581 	uint64_t oldversion;
8582 	int unsupp_fs = 0;
8583 
8584 	config = zpool_get_config(zhp, NULL);
8585 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8586 	    &oldversion) == 0);
8587 
8588 	assert(SPA_VERSION_IS_SUPPORTED(oldversion));
8589 	assert(oldversion < version);
8590 
8591 	ret = zfs_iter_root(zpool_get_handle(zhp), check_unsupp_fs, &unsupp_fs);
8592 	if (ret != 0)
8593 		return (ret);
8594 
8595 	if (unsupp_fs) {
8596 		(void) fprintf(stderr, gettext("Upgrade not performed due "
8597 		    "to %d unsupported filesystems (max v%d).\n"),
8598 		    unsupp_fs, (int)ZPL_VERSION);
8599 		return (1);
8600 	}
8601 
8602 	ret = zpool_upgrade(zhp, version);
8603 	if (ret != 0)
8604 		return (ret);
8605 
8606 	if (version >= SPA_VERSION_FEATURES) {
8607 		(void) printf(gettext("Successfully upgraded "
8608 		    "'%s' from version %llu to feature flags.\n"),
8609 		    zpool_get_name(zhp), (u_longlong_t)oldversion);
8610 	} else {
8611 		(void) printf(gettext("Successfully upgraded "
8612 		    "'%s' from version %llu to version %llu.\n"),
8613 		    zpool_get_name(zhp), (u_longlong_t)oldversion,
8614 		    (u_longlong_t)version);
8615 	}
8616 
8617 	return (0);
8618 }
8619 
8620 static int
8621 upgrade_enable_all(zpool_handle_t *zhp, int *countp)
8622 {
8623 	int i, ret, count;
8624 	boolean_t firstff = B_TRUE;
8625 	nvlist_t *enabled = zpool_get_features(zhp);
8626 
8627 	count = 0;
8628 	for (i = 0; i < SPA_FEATURES; i++) {
8629 		const char *fname = spa_feature_table[i].fi_uname;
8630 		const char *fguid = spa_feature_table[i].fi_guid;
8631 		if (!nvlist_exists(enabled, fguid)) {
8632 			char *propname;
8633 			verify(-1 != asprintf(&propname, "feature@%s", fname));
8634 			ret = zpool_set_prop(zhp, propname,
8635 			    ZFS_FEATURE_ENABLED);
8636 			if (ret != 0) {
8637 				free(propname);
8638 				return (ret);
8639 			}
8640 			count++;
8641 
8642 			if (firstff) {
8643 				(void) printf(gettext("Enabled the "
8644 				    "following features on '%s':\n"),
8645 				    zpool_get_name(zhp));
8646 				firstff = B_FALSE;
8647 			}
8648 			(void) printf(gettext("  %s\n"), fname);
8649 			free(propname);
8650 		}
8651 	}
8652 
8653 	if (countp != NULL)
8654 		*countp = count;
8655 	return (0);
8656 }
8657 
8658 static int
8659 upgrade_cb(zpool_handle_t *zhp, void *arg)
8660 {
8661 	upgrade_cbdata_t *cbp = arg;
8662 	nvlist_t *config;
8663 	uint64_t version;
8664 	boolean_t printnl = B_FALSE;
8665 	int ret;
8666 
8667 	config = zpool_get_config(zhp, NULL);
8668 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8669 	    &version) == 0);
8670 
8671 	assert(SPA_VERSION_IS_SUPPORTED(version));
8672 
8673 	if (version < cbp->cb_version) {
8674 		cbp->cb_first = B_FALSE;
8675 		ret = upgrade_version(zhp, cbp->cb_version);
8676 		if (ret != 0)
8677 			return (ret);
8678 		printnl = B_TRUE;
8679 
8680 		/*
8681 		 * If they did "zpool upgrade -a", then we could
8682 		 * be doing ioctls to different pools.  We need
8683 		 * to log this history once to each pool, and bypass
8684 		 * the normal history logging that happens in main().
8685 		 */
8686 		(void) zpool_log_history(g_zfs, history_str);
8687 		log_history = B_FALSE;
8688 	}
8689 
8690 	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
8691 		int count;
8692 		ret = upgrade_enable_all(zhp, &count);
8693 		if (ret != 0)
8694 			return (ret);
8695 
8696 		if (count > 0) {
8697 			cbp->cb_first = B_FALSE;
8698 			printnl = B_TRUE;
8699 		}
8700 	}
8701 
8702 	if (printnl) {
8703 		(void) printf(gettext("\n"));
8704 	}
8705 
8706 	return (0);
8707 }
8708 
8709 static int
8710 upgrade_list_older_cb(zpool_handle_t *zhp, void *arg)
8711 {
8712 	upgrade_cbdata_t *cbp = arg;
8713 	nvlist_t *config;
8714 	uint64_t version;
8715 
8716 	config = zpool_get_config(zhp, NULL);
8717 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8718 	    &version) == 0);
8719 
8720 	assert(SPA_VERSION_IS_SUPPORTED(version));
8721 
8722 	if (version < SPA_VERSION_FEATURES) {
8723 		if (cbp->cb_first) {
8724 			(void) printf(gettext("The following pools are "
8725 			    "formatted with legacy version numbers and can\n"
8726 			    "be upgraded to use feature flags.  After "
8727 			    "being upgraded, these pools\nwill no "
8728 			    "longer be accessible by software that does not "
8729 			    "support feature\nflags.\n\n"));
8730 			(void) printf(gettext("VER  POOL\n"));
8731 			(void) printf(gettext("---  ------------\n"));
8732 			cbp->cb_first = B_FALSE;
8733 		}
8734 
8735 		(void) printf("%2llu   %s\n", (u_longlong_t)version,
8736 		    zpool_get_name(zhp));
8737 	}
8738 
8739 	return (0);
8740 }
8741 
8742 static int
8743 upgrade_list_disabled_cb(zpool_handle_t *zhp, void *arg)
8744 {
8745 	upgrade_cbdata_t *cbp = arg;
8746 	nvlist_t *config;
8747 	uint64_t version;
8748 
8749 	config = zpool_get_config(zhp, NULL);
8750 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
8751 	    &version) == 0);
8752 
8753 	if (version >= SPA_VERSION_FEATURES) {
8754 		int i;
8755 		boolean_t poolfirst = B_TRUE;
8756 		nvlist_t *enabled = zpool_get_features(zhp);
8757 
8758 		for (i = 0; i < SPA_FEATURES; i++) {
8759 			const char *fguid = spa_feature_table[i].fi_guid;
8760 			const char *fname = spa_feature_table[i].fi_uname;
8761 			if (!nvlist_exists(enabled, fguid)) {
8762 				if (cbp->cb_first) {
8763 					(void) printf(gettext("\nSome "
8764 					    "supported features are not "
8765 					    "enabled on the following pools. "
8766 					    "Once a\nfeature is enabled the "
8767 					    "pool may become incompatible with "
8768 					    "software\nthat does not support "
8769 					    "the feature. See "
8770 					    "zpool-features(5) for "
8771 					    "details.\n\n"));
8772 					(void) printf(gettext("POOL  "
8773 					    "FEATURE\n"));
8774 					(void) printf(gettext("------"
8775 					    "---------\n"));
8776 					cbp->cb_first = B_FALSE;
8777 				}
8778 
8779 				if (poolfirst) {
8780 					(void) printf(gettext("%s\n"),
8781 					    zpool_get_name(zhp));
8782 					poolfirst = B_FALSE;
8783 				}
8784 
8785 				(void) printf(gettext("      %s\n"), fname);
8786 			}
8787 			/*
8788 			 * If they did "zpool upgrade -a", then we could
8789 			 * be doing ioctls to different pools.  We need
8790 			 * to log this history once to each pool, and bypass
8791 			 * the normal history logging that happens in main().
8792 			 */
8793 			(void) zpool_log_history(g_zfs, history_str);
8794 			log_history = B_FALSE;
8795 		}
8796 	}
8797 
8798 	return (0);
8799 }
8800 
8801 /* ARGSUSED */
8802 static int
8803 upgrade_one(zpool_handle_t *zhp, void *data)
8804 {
8805 	boolean_t printnl = B_FALSE;
8806 	upgrade_cbdata_t *cbp = data;
8807 	uint64_t cur_version;
8808 	int ret;
8809 
8810 	if (strcmp("log", zpool_get_name(zhp)) == 0) {
8811 		(void) fprintf(stderr, gettext("'log' is now a reserved word\n"
8812 		    "Pool 'log' must be renamed using export and import"
8813 		    " to upgrade.\n"));
8814 		return (1);
8815 	}
8816 
8817 	cur_version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
8818 	if (cur_version > cbp->cb_version) {
8819 		(void) printf(gettext("Pool '%s' is already formatted "
8820 		    "using more current version '%llu'.\n\n"),
8821 		    zpool_get_name(zhp), (u_longlong_t)cur_version);
8822 		return (0);
8823 	}
8824 
8825 	if (cbp->cb_version != SPA_VERSION && cur_version == cbp->cb_version) {
8826 		(void) printf(gettext("Pool '%s' is already formatted "
8827 		    "using version %llu.\n\n"), zpool_get_name(zhp),
8828 		    (u_longlong_t)cbp->cb_version);
8829 		return (0);
8830 	}
8831 
8832 	if (cur_version != cbp->cb_version) {
8833 		printnl = B_TRUE;
8834 		ret = upgrade_version(zhp, cbp->cb_version);
8835 		if (ret != 0)
8836 			return (ret);
8837 	}
8838 
8839 	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
8840 		int count = 0;
8841 		ret = upgrade_enable_all(zhp, &count);
8842 		if (ret != 0)
8843 			return (ret);
8844 
8845 		if (count != 0) {
8846 			printnl = B_TRUE;
8847 		} else if (cur_version == SPA_VERSION) {
8848 			(void) printf(gettext("Pool '%s' already has all "
8849 			    "supported features enabled.\n"),
8850 			    zpool_get_name(zhp));
8851 		}
8852 	}
8853 
8854 	if (printnl) {
8855 		(void) printf(gettext("\n"));
8856 	}
8857 
8858 	return (0);
8859 }
8860 
8861 /*
8862  * zpool upgrade
8863  * zpool upgrade -v
8864  * zpool upgrade [-V version] <-a | pool ...>
8865  *
8866  * With no arguments, display downrev'd ZFS pool available for upgrade.
8867  * Individual pools can be upgraded by specifying the pool, and '-a' will
8868  * upgrade all pools.
8869  */
8870 int
8871 zpool_do_upgrade(int argc, char **argv)
8872 {
8873 	int c;
8874 	upgrade_cbdata_t cb = { 0 };
8875 	int ret = 0;
8876 	boolean_t showversions = B_FALSE;
8877 	boolean_t upgradeall = B_FALSE;
8878 	char *end;
8879 
8880 
8881 	/* check options */
8882 	while ((c = getopt(argc, argv, ":avV:")) != -1) {
8883 		switch (c) {
8884 		case 'a':
8885 			upgradeall = B_TRUE;
8886 			break;
8887 		case 'v':
8888 			showversions = B_TRUE;
8889 			break;
8890 		case 'V':
8891 			cb.cb_version = strtoll(optarg, &end, 10);
8892 			if (*end != '\0' ||
8893 			    !SPA_VERSION_IS_SUPPORTED(cb.cb_version)) {
8894 				(void) fprintf(stderr,
8895 				    gettext("invalid version '%s'\n"), optarg);
8896 				usage(B_FALSE);
8897 			}
8898 			break;
8899 		case ':':
8900 			(void) fprintf(stderr, gettext("missing argument for "
8901 			    "'%c' option\n"), optopt);
8902 			usage(B_FALSE);
8903 			break;
8904 		case '?':
8905 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
8906 			    optopt);
8907 			usage(B_FALSE);
8908 		}
8909 	}
8910 
8911 	cb.cb_argc = argc;
8912 	cb.cb_argv = argv;
8913 	argc -= optind;
8914 	argv += optind;
8915 
8916 	if (cb.cb_version == 0) {
8917 		cb.cb_version = SPA_VERSION;
8918 	} else if (!upgradeall && argc == 0) {
8919 		(void) fprintf(stderr, gettext("-V option is "
8920 		    "incompatible with other arguments\n"));
8921 		usage(B_FALSE);
8922 	}
8923 
8924 	if (showversions) {
8925 		if (upgradeall || argc != 0) {
8926 			(void) fprintf(stderr, gettext("-v option is "
8927 			    "incompatible with other arguments\n"));
8928 			usage(B_FALSE);
8929 		}
8930 	} else if (upgradeall) {
8931 		if (argc != 0) {
8932 			(void) fprintf(stderr, gettext("-a option should not "
8933 			    "be used along with a pool name\n"));
8934 			usage(B_FALSE);
8935 		}
8936 	}
8937 
8938 	(void) printf(gettext("This system supports ZFS pool feature "
8939 	    "flags.\n\n"));
8940 	if (showversions) {
8941 		int i;
8942 
8943 		(void) printf(gettext("The following features are "
8944 		    "supported:\n\n"));
8945 		(void) printf(gettext("FEAT DESCRIPTION\n"));
8946 		(void) printf("----------------------------------------------"
8947 		    "---------------\n");
8948 		for (i = 0; i < SPA_FEATURES; i++) {
8949 			zfeature_info_t *fi = &spa_feature_table[i];
8950 			const char *ro =
8951 			    (fi->fi_flags & ZFEATURE_FLAG_READONLY_COMPAT) ?
8952 			    " (read-only compatible)" : "";
8953 
8954 			(void) printf("%-37s%s\n", fi->fi_uname, ro);
8955 			(void) printf("     %s\n", fi->fi_desc);
8956 		}
8957 		(void) printf("\n");
8958 
8959 		(void) printf(gettext("The following legacy versions are also "
8960 		    "supported:\n\n"));
8961 		(void) printf(gettext("VER  DESCRIPTION\n"));
8962 		(void) printf("---  -----------------------------------------"
8963 		    "---------------\n");
8964 		(void) printf(gettext(" 1   Initial ZFS version\n"));
8965 		(void) printf(gettext(" 2   Ditto blocks "
8966 		    "(replicated metadata)\n"));
8967 		(void) printf(gettext(" 3   Hot spares and double parity "
8968 		    "RAID-Z\n"));
8969 		(void) printf(gettext(" 4   zpool history\n"));
8970 		(void) printf(gettext(" 5   Compression using the gzip "
8971 		    "algorithm\n"));
8972 		(void) printf(gettext(" 6   bootfs pool property\n"));
8973 		(void) printf(gettext(" 7   Separate intent log devices\n"));
8974 		(void) printf(gettext(" 8   Delegated administration\n"));
8975 		(void) printf(gettext(" 9   refquota and refreservation "
8976 		    "properties\n"));
8977 		(void) printf(gettext(" 10  Cache devices\n"));
8978 		(void) printf(gettext(" 11  Improved scrub performance\n"));
8979 		(void) printf(gettext(" 12  Snapshot properties\n"));
8980 		(void) printf(gettext(" 13  snapused property\n"));
8981 		(void) printf(gettext(" 14  passthrough-x aclinherit\n"));
8982 		(void) printf(gettext(" 15  user/group space accounting\n"));
8983 		(void) printf(gettext(" 16  stmf property support\n"));
8984 		(void) printf(gettext(" 17  Triple-parity RAID-Z\n"));
8985 		(void) printf(gettext(" 18  Snapshot user holds\n"));
8986 		(void) printf(gettext(" 19  Log device removal\n"));
8987 		(void) printf(gettext(" 20  Compression using zle "
8988 		    "(zero-length encoding)\n"));
8989 		(void) printf(gettext(" 21  Deduplication\n"));
8990 		(void) printf(gettext(" 22  Received properties\n"));
8991 		(void) printf(gettext(" 23  Slim ZIL\n"));
8992 		(void) printf(gettext(" 24  System attributes\n"));
8993 		(void) printf(gettext(" 25  Improved scrub stats\n"));
8994 		(void) printf(gettext(" 26  Improved snapshot deletion "
8995 		    "performance\n"));
8996 		(void) printf(gettext(" 27  Improved snapshot creation "
8997 		    "performance\n"));
8998 		(void) printf(gettext(" 28  Multiple vdev replacements\n"));
8999 		(void) printf(gettext("\nFor more information on a particular "
9000 		    "version, including supported releases,\n"));
9001 		(void) printf(gettext("see the ZFS Administration Guide.\n\n"));
9002 	} else if (argc == 0 && upgradeall) {
9003 		cb.cb_first = B_TRUE;
9004 		ret = zpool_iter(g_zfs, upgrade_cb, &cb);
9005 		if (ret == 0 && cb.cb_first) {
9006 			if (cb.cb_version == SPA_VERSION) {
9007 				(void) printf(gettext("All pools are already "
9008 				    "formatted using feature flags.\n\n"));
9009 				(void) printf(gettext("Every feature flags "
9010 				    "pool already has all supported features "
9011 				    "enabled.\n"));
9012 			} else {
9013 				(void) printf(gettext("All pools are already "
9014 				    "formatted with version %llu or higher.\n"),
9015 				    (u_longlong_t)cb.cb_version);
9016 			}
9017 		}
9018 	} else if (argc == 0) {
9019 		cb.cb_first = B_TRUE;
9020 		ret = zpool_iter(g_zfs, upgrade_list_older_cb, &cb);
9021 		assert(ret == 0);
9022 
9023 		if (cb.cb_first) {
9024 			(void) printf(gettext("All pools are formatted "
9025 			    "using feature flags.\n\n"));
9026 		} else {
9027 			(void) printf(gettext("\nUse 'zpool upgrade -v' "
9028 			    "for a list of available legacy versions.\n"));
9029 		}
9030 
9031 		cb.cb_first = B_TRUE;
9032 		ret = zpool_iter(g_zfs, upgrade_list_disabled_cb, &cb);
9033 		assert(ret == 0);
9034 
9035 		if (cb.cb_first) {
9036 			(void) printf(gettext("Every feature flags pool has "
9037 			    "all supported features enabled.\n"));
9038 		} else {
9039 			(void) printf(gettext("\n"));
9040 		}
9041 	} else {
9042 		ret = for_each_pool(argc, argv, B_FALSE, NULL, B_FALSE,
9043 		    upgrade_one, &cb);
9044 	}
9045 
9046 	return (ret);
9047 }
9048 
9049 typedef struct hist_cbdata {
9050 	boolean_t first;
9051 	boolean_t longfmt;
9052 	boolean_t internal;
9053 } hist_cbdata_t;
9054 
9055 static void
9056 print_history_records(nvlist_t *nvhis, hist_cbdata_t *cb)
9057 {
9058 	nvlist_t **records;
9059 	uint_t numrecords;
9060 	int i;
9061 
9062 	verify(nvlist_lookup_nvlist_array(nvhis, ZPOOL_HIST_RECORD,
9063 	    &records, &numrecords) == 0);
9064 	for (i = 0; i < numrecords; i++) {
9065 		nvlist_t *rec = records[i];
9066 		char tbuf[30] = "";
9067 
9068 		if (nvlist_exists(rec, ZPOOL_HIST_TIME)) {
9069 			time_t tsec;
9070 			struct tm t;
9071 
9072 			tsec = fnvlist_lookup_uint64(records[i],
9073 			    ZPOOL_HIST_TIME);
9074 			(void) localtime_r(&tsec, &t);
9075 			(void) strftime(tbuf, sizeof (tbuf), "%F.%T", &t);
9076 		}
9077 
9078 		if (nvlist_exists(rec, ZPOOL_HIST_CMD)) {
9079 			(void) printf("%s %s", tbuf,
9080 			    fnvlist_lookup_string(rec, ZPOOL_HIST_CMD));
9081 		} else if (nvlist_exists(rec, ZPOOL_HIST_INT_EVENT)) {
9082 			int ievent =
9083 			    fnvlist_lookup_uint64(rec, ZPOOL_HIST_INT_EVENT);
9084 			if (!cb->internal)
9085 				continue;
9086 			if (ievent >= ZFS_NUM_LEGACY_HISTORY_EVENTS) {
9087 				(void) printf("%s unrecognized record:\n",
9088 				    tbuf);
9089 				dump_nvlist(rec, 4);
9090 				continue;
9091 			}
9092 			(void) printf("%s [internal %s txg:%lld] %s", tbuf,
9093 			    zfs_history_event_names[ievent],
9094 			    (longlong_t)fnvlist_lookup_uint64(
9095 			    rec, ZPOOL_HIST_TXG),
9096 			    fnvlist_lookup_string(rec, ZPOOL_HIST_INT_STR));
9097 		} else if (nvlist_exists(rec, ZPOOL_HIST_INT_NAME)) {
9098 			if (!cb->internal)
9099 				continue;
9100 			(void) printf("%s [txg:%lld] %s", tbuf,
9101 			    (longlong_t)fnvlist_lookup_uint64(
9102 			    rec, ZPOOL_HIST_TXG),
9103 			    fnvlist_lookup_string(rec, ZPOOL_HIST_INT_NAME));
9104 			if (nvlist_exists(rec, ZPOOL_HIST_DSNAME)) {
9105 				(void) printf(" %s (%llu)",
9106 				    fnvlist_lookup_string(rec,
9107 				    ZPOOL_HIST_DSNAME),
9108 				    (u_longlong_t)fnvlist_lookup_uint64(rec,
9109 				    ZPOOL_HIST_DSID));
9110 			}
9111 			(void) printf(" %s", fnvlist_lookup_string(rec,
9112 			    ZPOOL_HIST_INT_STR));
9113 		} else if (nvlist_exists(rec, ZPOOL_HIST_IOCTL)) {
9114 			if (!cb->internal)
9115 				continue;
9116 			(void) printf("%s ioctl %s\n", tbuf,
9117 			    fnvlist_lookup_string(rec, ZPOOL_HIST_IOCTL));
9118 			if (nvlist_exists(rec, ZPOOL_HIST_INPUT_NVL)) {
9119 				(void) printf("    input:\n");
9120 				dump_nvlist(fnvlist_lookup_nvlist(rec,
9121 				    ZPOOL_HIST_INPUT_NVL), 8);
9122 			}
9123 			if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_NVL)) {
9124 				(void) printf("    output:\n");
9125 				dump_nvlist(fnvlist_lookup_nvlist(rec,
9126 				    ZPOOL_HIST_OUTPUT_NVL), 8);
9127 			}
9128 			if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_SIZE)) {
9129 				(void) printf("    output nvlist omitted; "
9130 				    "original size: %lldKB\n",
9131 				    (longlong_t)fnvlist_lookup_int64(rec,
9132 				    ZPOOL_HIST_OUTPUT_SIZE) / 1024);
9133 			}
9134 			if (nvlist_exists(rec, ZPOOL_HIST_ERRNO)) {
9135 				(void) printf("    errno: %lld\n",
9136 				    (longlong_t)fnvlist_lookup_int64(rec,
9137 				    ZPOOL_HIST_ERRNO));
9138 			}
9139 		} else {
9140 			if (!cb->internal)
9141 				continue;
9142 			(void) printf("%s unrecognized record:\n", tbuf);
9143 			dump_nvlist(rec, 4);
9144 		}
9145 
9146 		if (!cb->longfmt) {
9147 			(void) printf("\n");
9148 			continue;
9149 		}
9150 		(void) printf(" [");
9151 		if (nvlist_exists(rec, ZPOOL_HIST_WHO)) {
9152 			uid_t who = fnvlist_lookup_uint64(rec, ZPOOL_HIST_WHO);
9153 			struct passwd *pwd = getpwuid(who);
9154 			(void) printf("user %d ", (int)who);
9155 			if (pwd != NULL)
9156 				(void) printf("(%s) ", pwd->pw_name);
9157 		}
9158 		if (nvlist_exists(rec, ZPOOL_HIST_HOST)) {
9159 			(void) printf("on %s",
9160 			    fnvlist_lookup_string(rec, ZPOOL_HIST_HOST));
9161 		}
9162 		if (nvlist_exists(rec, ZPOOL_HIST_ZONE)) {
9163 			(void) printf(":%s",
9164 			    fnvlist_lookup_string(rec, ZPOOL_HIST_ZONE));
9165 		}
9166 
9167 		(void) printf("]");
9168 		(void) printf("\n");
9169 	}
9170 }
9171 
9172 /*
9173  * Print out the command history for a specific pool.
9174  */
9175 static int
9176 get_history_one(zpool_handle_t *zhp, void *data)
9177 {
9178 	nvlist_t *nvhis;
9179 	int ret;
9180 	hist_cbdata_t *cb = (hist_cbdata_t *)data;
9181 	uint64_t off = 0;
9182 	boolean_t eof = B_FALSE;
9183 
9184 	cb->first = B_FALSE;
9185 
9186 	(void) printf(gettext("History for '%s':\n"), zpool_get_name(zhp));
9187 
9188 	while (!eof) {
9189 		if ((ret = zpool_get_history(zhp, &nvhis, &off, &eof)) != 0)
9190 			return (ret);
9191 
9192 		print_history_records(nvhis, cb);
9193 		nvlist_free(nvhis);
9194 	}
9195 	(void) printf("\n");
9196 
9197 	return (ret);
9198 }
9199 
9200 /*
9201  * zpool history <pool>
9202  *
9203  * Displays the history of commands that modified pools.
9204  */
9205 int
9206 zpool_do_history(int argc, char **argv)
9207 {
9208 	hist_cbdata_t cbdata = { 0 };
9209 	int ret;
9210 	int c;
9211 
9212 	cbdata.first = B_TRUE;
9213 	/* check options */
9214 	while ((c = getopt(argc, argv, "li")) != -1) {
9215 		switch (c) {
9216 		case 'l':
9217 			cbdata.longfmt = B_TRUE;
9218 			break;
9219 		case 'i':
9220 			cbdata.internal = B_TRUE;
9221 			break;
9222 		case '?':
9223 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9224 			    optopt);
9225 			usage(B_FALSE);
9226 		}
9227 	}
9228 	argc -= optind;
9229 	argv += optind;
9230 
9231 	ret = for_each_pool(argc, argv, B_FALSE, NULL, B_FALSE, get_history_one,
9232 	    &cbdata);
9233 
9234 	if (argc == 0 && cbdata.first == B_TRUE) {
9235 		(void) fprintf(stderr, gettext("no pools available\n"));
9236 		return (0);
9237 	}
9238 
9239 	return (ret);
9240 }
9241 
9242 typedef struct ev_opts {
9243 	int verbose;
9244 	int scripted;
9245 	int follow;
9246 	int clear;
9247 	char poolname[ZFS_MAX_DATASET_NAME_LEN];
9248 } ev_opts_t;
9249 
9250 static void
9251 zpool_do_events_short(nvlist_t *nvl, ev_opts_t *opts)
9252 {
9253 	char ctime_str[26], str[32], *ptr;
9254 	int64_t *tv;
9255 	uint_t n;
9256 
9257 	verify(nvlist_lookup_int64_array(nvl, FM_EREPORT_TIME, &tv, &n) == 0);
9258 	memset(str, ' ', 32);
9259 	(void) ctime_r((const time_t *)&tv[0], ctime_str);
9260 	(void) memcpy(str, ctime_str+4,  6);		/* 'Jun 30' */
9261 	(void) memcpy(str+7, ctime_str+20, 4);		/* '1993' */
9262 	(void) memcpy(str+12, ctime_str+11, 8);		/* '21:49:08' */
9263 	(void) sprintf(str+20, ".%09lld", (longlong_t)tv[1]); /* '.123456789' */
9264 	if (opts->scripted)
9265 		(void) printf(gettext("%s\t"), str);
9266 	else
9267 		(void) printf(gettext("%s "), str);
9268 
9269 	verify(nvlist_lookup_string(nvl, FM_CLASS, &ptr) == 0);
9270 	(void) printf(gettext("%s\n"), ptr);
9271 }
9272 
9273 static void
9274 zpool_do_events_nvprint(nvlist_t *nvl, int depth)
9275 {
9276 	nvpair_t *nvp;
9277 
9278 	for (nvp = nvlist_next_nvpair(nvl, NULL);
9279 	    nvp != NULL; nvp = nvlist_next_nvpair(nvl, nvp)) {
9280 
9281 		data_type_t type = nvpair_type(nvp);
9282 		const char *name = nvpair_name(nvp);
9283 
9284 		boolean_t b;
9285 		uint8_t i8;
9286 		uint16_t i16;
9287 		uint32_t i32;
9288 		uint64_t i64;
9289 		char *str;
9290 		nvlist_t *cnv;
9291 
9292 		printf(gettext("%*s%s = "), depth, "", name);
9293 
9294 		switch (type) {
9295 		case DATA_TYPE_BOOLEAN:
9296 			printf(gettext("%s"), "1");
9297 			break;
9298 
9299 		case DATA_TYPE_BOOLEAN_VALUE:
9300 			(void) nvpair_value_boolean_value(nvp, &b);
9301 			printf(gettext("%s"), b ? "1" : "0");
9302 			break;
9303 
9304 		case DATA_TYPE_BYTE:
9305 			(void) nvpair_value_byte(nvp, &i8);
9306 			printf(gettext("0x%x"), i8);
9307 			break;
9308 
9309 		case DATA_TYPE_INT8:
9310 			(void) nvpair_value_int8(nvp, (void *)&i8);
9311 			printf(gettext("0x%x"), i8);
9312 			break;
9313 
9314 		case DATA_TYPE_UINT8:
9315 			(void) nvpair_value_uint8(nvp, &i8);
9316 			printf(gettext("0x%x"), i8);
9317 			break;
9318 
9319 		case DATA_TYPE_INT16:
9320 			(void) nvpair_value_int16(nvp, (void *)&i16);
9321 			printf(gettext("0x%x"), i16);
9322 			break;
9323 
9324 		case DATA_TYPE_UINT16:
9325 			(void) nvpair_value_uint16(nvp, &i16);
9326 			printf(gettext("0x%x"), i16);
9327 			break;
9328 
9329 		case DATA_TYPE_INT32:
9330 			(void) nvpair_value_int32(nvp, (void *)&i32);
9331 			printf(gettext("0x%x"), i32);
9332 			break;
9333 
9334 		case DATA_TYPE_UINT32:
9335 			(void) nvpair_value_uint32(nvp, &i32);
9336 			printf(gettext("0x%x"), i32);
9337 			break;
9338 
9339 		case DATA_TYPE_INT64:
9340 			(void) nvpair_value_int64(nvp, (void *)&i64);
9341 			printf(gettext("0x%llx"), (u_longlong_t)i64);
9342 			break;
9343 
9344 		case DATA_TYPE_UINT64:
9345 			(void) nvpair_value_uint64(nvp, &i64);
9346 			/*
9347 			 * translate vdev state values to readable
9348 			 * strings to aide zpool events consumers
9349 			 */
9350 			if (strcmp(name,
9351 			    FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE) == 0 ||
9352 			    strcmp(name,
9353 			    FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE) == 0) {
9354 				printf(gettext("\"%s\" (0x%llx)"),
9355 				    zpool_state_to_name(i64, VDEV_AUX_NONE),
9356 				    (u_longlong_t)i64);
9357 			} else {
9358 				printf(gettext("0x%llx"), (u_longlong_t)i64);
9359 			}
9360 			break;
9361 
9362 		case DATA_TYPE_HRTIME:
9363 			(void) nvpair_value_hrtime(nvp, (void *)&i64);
9364 			printf(gettext("0x%llx"), (u_longlong_t)i64);
9365 			break;
9366 
9367 		case DATA_TYPE_STRING:
9368 			(void) nvpair_value_string(nvp, &str);
9369 			printf(gettext("\"%s\""), str ? str : "<NULL>");
9370 			break;
9371 
9372 		case DATA_TYPE_NVLIST:
9373 			printf(gettext("(embedded nvlist)\n"));
9374 			(void) nvpair_value_nvlist(nvp, &cnv);
9375 			zpool_do_events_nvprint(cnv, depth + 8);
9376 			printf(gettext("%*s(end %s)"), depth, "", name);
9377 			break;
9378 
9379 		case DATA_TYPE_NVLIST_ARRAY: {
9380 			nvlist_t **val;
9381 			uint_t i, nelem;
9382 
9383 			(void) nvpair_value_nvlist_array(nvp, &val, &nelem);
9384 			printf(gettext("(%d embedded nvlists)\n"), nelem);
9385 			for (i = 0; i < nelem; i++) {
9386 				printf(gettext("%*s%s[%d] = %s\n"),
9387 				    depth, "", name, i, "(embedded nvlist)");
9388 				zpool_do_events_nvprint(val[i], depth + 8);
9389 				printf(gettext("%*s(end %s[%i])\n"),
9390 				    depth, "", name, i);
9391 			}
9392 			printf(gettext("%*s(end %s)\n"), depth, "", name);
9393 			}
9394 			break;
9395 
9396 		case DATA_TYPE_INT8_ARRAY: {
9397 			int8_t *val;
9398 			uint_t i, nelem;
9399 
9400 			(void) nvpair_value_int8_array(nvp, &val, &nelem);
9401 			for (i = 0; i < nelem; i++)
9402 				printf(gettext("0x%x "), val[i]);
9403 
9404 			break;
9405 			}
9406 
9407 		case DATA_TYPE_UINT8_ARRAY: {
9408 			uint8_t *val;
9409 			uint_t i, nelem;
9410 
9411 			(void) nvpair_value_uint8_array(nvp, &val, &nelem);
9412 			for (i = 0; i < nelem; i++)
9413 				printf(gettext("0x%x "), val[i]);
9414 
9415 			break;
9416 			}
9417 
9418 		case DATA_TYPE_INT16_ARRAY: {
9419 			int16_t *val;
9420 			uint_t i, nelem;
9421 
9422 			(void) nvpair_value_int16_array(nvp, &val, &nelem);
9423 			for (i = 0; i < nelem; i++)
9424 				printf(gettext("0x%x "), val[i]);
9425 
9426 			break;
9427 			}
9428 
9429 		case DATA_TYPE_UINT16_ARRAY: {
9430 			uint16_t *val;
9431 			uint_t i, nelem;
9432 
9433 			(void) nvpair_value_uint16_array(nvp, &val, &nelem);
9434 			for (i = 0; i < nelem; i++)
9435 				printf(gettext("0x%x "), val[i]);
9436 
9437 			break;
9438 			}
9439 
9440 		case DATA_TYPE_INT32_ARRAY: {
9441 			int32_t *val;
9442 			uint_t i, nelem;
9443 
9444 			(void) nvpair_value_int32_array(nvp, &val, &nelem);
9445 			for (i = 0; i < nelem; i++)
9446 				printf(gettext("0x%x "), val[i]);
9447 
9448 			break;
9449 			}
9450 
9451 		case DATA_TYPE_UINT32_ARRAY: {
9452 			uint32_t *val;
9453 			uint_t i, nelem;
9454 
9455 			(void) nvpair_value_uint32_array(nvp, &val, &nelem);
9456 			for (i = 0; i < nelem; i++)
9457 				printf(gettext("0x%x "), val[i]);
9458 
9459 			break;
9460 			}
9461 
9462 		case DATA_TYPE_INT64_ARRAY: {
9463 			int64_t *val;
9464 			uint_t i, nelem;
9465 
9466 			(void) nvpair_value_int64_array(nvp, &val, &nelem);
9467 			for (i = 0; i < nelem; i++)
9468 				printf(gettext("0x%llx "),
9469 				    (u_longlong_t)val[i]);
9470 
9471 			break;
9472 			}
9473 
9474 		case DATA_TYPE_UINT64_ARRAY: {
9475 			uint64_t *val;
9476 			uint_t i, nelem;
9477 
9478 			(void) nvpair_value_uint64_array(nvp, &val, &nelem);
9479 			for (i = 0; i < nelem; i++)
9480 				printf(gettext("0x%llx "),
9481 				    (u_longlong_t)val[i]);
9482 
9483 			break;
9484 			}
9485 
9486 		case DATA_TYPE_STRING_ARRAY: {
9487 			char **str;
9488 			uint_t i, nelem;
9489 
9490 			(void) nvpair_value_string_array(nvp, &str, &nelem);
9491 			for (i = 0; i < nelem; i++)
9492 				printf(gettext("\"%s\" "),
9493 				    str[i] ? str[i] : "<NULL>");
9494 
9495 			break;
9496 			}
9497 
9498 		case DATA_TYPE_BOOLEAN_ARRAY:
9499 		case DATA_TYPE_BYTE_ARRAY:
9500 		case DATA_TYPE_DOUBLE:
9501 		case DATA_TYPE_DONTCARE:
9502 		case DATA_TYPE_UNKNOWN:
9503 			printf(gettext("<unknown>"));
9504 			break;
9505 		}
9506 
9507 		printf(gettext("\n"));
9508 	}
9509 }
9510 
9511 static int
9512 zpool_do_events_next(ev_opts_t *opts)
9513 {
9514 	nvlist_t *nvl;
9515 	int zevent_fd, ret, dropped;
9516 	char *pool;
9517 
9518 	zevent_fd = open(ZFS_DEV, O_RDWR);
9519 	VERIFY(zevent_fd >= 0);
9520 
9521 	if (!opts->scripted)
9522 		(void) printf(gettext("%-30s %s\n"), "TIME", "CLASS");
9523 
9524 	while (1) {
9525 		ret = zpool_events_next(g_zfs, &nvl, &dropped,
9526 		    (opts->follow ? ZEVENT_NONE : ZEVENT_NONBLOCK), zevent_fd);
9527 		if (ret || nvl == NULL)
9528 			break;
9529 
9530 		if (dropped > 0)
9531 			(void) printf(gettext("dropped %d events\n"), dropped);
9532 
9533 		if (strlen(opts->poolname) > 0 &&
9534 		    nvlist_lookup_string(nvl, FM_FMRI_ZFS_POOL, &pool) == 0 &&
9535 		    strcmp(opts->poolname, pool) != 0)
9536 			continue;
9537 
9538 		zpool_do_events_short(nvl, opts);
9539 
9540 		if (opts->verbose) {
9541 			zpool_do_events_nvprint(nvl, 8);
9542 			printf(gettext("\n"));
9543 		}
9544 		(void) fflush(stdout);
9545 
9546 		nvlist_free(nvl);
9547 	}
9548 
9549 	VERIFY(0 == close(zevent_fd));
9550 
9551 	return (ret);
9552 }
9553 
9554 static int
9555 zpool_do_events_clear(ev_opts_t *opts)
9556 {
9557 	int count, ret;
9558 
9559 	ret = zpool_events_clear(g_zfs, &count);
9560 	if (!ret)
9561 		(void) printf(gettext("cleared %d events\n"), count);
9562 
9563 	return (ret);
9564 }
9565 
9566 /*
9567  * zpool events [-vHf [pool] | -c]
9568  *
9569  * Displays events logs by ZFS.
9570  */
9571 int
9572 zpool_do_events(int argc, char **argv)
9573 {
9574 	ev_opts_t opts = { 0 };
9575 	int ret;
9576 	int c;
9577 
9578 	/* check options */
9579 	while ((c = getopt(argc, argv, "vHfc")) != -1) {
9580 		switch (c) {
9581 		case 'v':
9582 			opts.verbose = 1;
9583 			break;
9584 		case 'H':
9585 			opts.scripted = 1;
9586 			break;
9587 		case 'f':
9588 			opts.follow = 1;
9589 			break;
9590 		case 'c':
9591 			opts.clear = 1;
9592 			break;
9593 		case '?':
9594 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9595 			    optopt);
9596 			usage(B_FALSE);
9597 		}
9598 	}
9599 	argc -= optind;
9600 	argv += optind;
9601 
9602 	if (argc > 1) {
9603 		(void) fprintf(stderr, gettext("too many arguments\n"));
9604 		usage(B_FALSE);
9605 	} else if (argc == 1) {
9606 		(void) strlcpy(opts.poolname, argv[0], sizeof (opts.poolname));
9607 		if (!zfs_name_valid(opts.poolname, ZFS_TYPE_POOL)) {
9608 			(void) fprintf(stderr,
9609 			    gettext("invalid pool name '%s'\n"), opts.poolname);
9610 			usage(B_FALSE);
9611 		}
9612 	}
9613 
9614 	if ((argc == 1 || opts.verbose || opts.scripted || opts.follow) &&
9615 	    opts.clear) {
9616 		(void) fprintf(stderr,
9617 		    gettext("invalid options combined with -c\n"));
9618 		usage(B_FALSE);
9619 	}
9620 
9621 	if (opts.clear)
9622 		ret = zpool_do_events_clear(&opts);
9623 	else
9624 		ret = zpool_do_events_next(&opts);
9625 
9626 	return (ret);
9627 }
9628 
9629 static int
9630 get_callback(zpool_handle_t *zhp, void *data)
9631 {
9632 	zprop_get_cbdata_t *cbp = (zprop_get_cbdata_t *)data;
9633 	char value[MAXNAMELEN];
9634 	zprop_source_t srctype;
9635 	zprop_list_t *pl;
9636 
9637 	for (pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) {
9638 
9639 		/*
9640 		 * Skip the special fake placeholder. This will also skip
9641 		 * over the name property when 'all' is specified.
9642 		 */
9643 		if (pl->pl_prop == ZPOOL_PROP_NAME &&
9644 		    pl == cbp->cb_proplist)
9645 			continue;
9646 
9647 		if (pl->pl_prop == ZPROP_INVAL &&
9648 		    (zpool_prop_feature(pl->pl_user_prop) ||
9649 		    zpool_prop_unsupported(pl->pl_user_prop))) {
9650 			srctype = ZPROP_SRC_LOCAL;
9651 
9652 			if (zpool_prop_get_feature(zhp, pl->pl_user_prop,
9653 			    value, sizeof (value)) == 0) {
9654 				zprop_print_one_property(zpool_get_name(zhp),
9655 				    cbp, pl->pl_user_prop, value, srctype,
9656 				    NULL, NULL);
9657 			}
9658 		} else {
9659 			if (zpool_get_prop(zhp, pl->pl_prop, value,
9660 			    sizeof (value), &srctype, cbp->cb_literal) != 0)
9661 				continue;
9662 
9663 			zprop_print_one_property(zpool_get_name(zhp), cbp,
9664 			    zpool_prop_to_name(pl->pl_prop), value, srctype,
9665 			    NULL, NULL);
9666 		}
9667 	}
9668 	return (0);
9669 }
9670 
9671 /*
9672  * zpool get [-Hp] [-o "all" | field[,...]] <"all" | property[,...]> <pool> ...
9673  *
9674  *	-H	Scripted mode.  Don't display headers, and separate properties
9675  *		by a single tab.
9676  *	-o	List of columns to display.  Defaults to
9677  *		"name,property,value,source".
9678  * 	-p	Display values in parsable (exact) format.
9679  *
9680  * Get properties of pools in the system. Output space statistics
9681  * for each one as well as other attributes.
9682  */
9683 int
9684 zpool_do_get(int argc, char **argv)
9685 {
9686 	zprop_get_cbdata_t cb = { 0 };
9687 	zprop_list_t fake_name = { 0 };
9688 	int ret;
9689 	int c, i;
9690 	char *value;
9691 
9692 	cb.cb_first = B_TRUE;
9693 
9694 	/*
9695 	 * Set up default columns and sources.
9696 	 */
9697 	cb.cb_sources = ZPROP_SRC_ALL;
9698 	cb.cb_columns[0] = GET_COL_NAME;
9699 	cb.cb_columns[1] = GET_COL_PROPERTY;
9700 	cb.cb_columns[2] = GET_COL_VALUE;
9701 	cb.cb_columns[3] = GET_COL_SOURCE;
9702 	cb.cb_type = ZFS_TYPE_POOL;
9703 
9704 	/* check options */
9705 	while ((c = getopt(argc, argv, ":Hpo:")) != -1) {
9706 		switch (c) {
9707 		case 'p':
9708 			cb.cb_literal = B_TRUE;
9709 			break;
9710 		case 'H':
9711 			cb.cb_scripted = B_TRUE;
9712 			break;
9713 		case 'o':
9714 			bzero(&cb.cb_columns, sizeof (cb.cb_columns));
9715 			i = 0;
9716 			while (*optarg != '\0') {
9717 				static char *col_subopts[] =
9718 				{ "name", "property", "value", "source",
9719 				"all", NULL };
9720 
9721 				if (i == ZFS_GET_NCOLS) {
9722 					(void) fprintf(stderr, gettext("too "
9723 					"many fields given to -o "
9724 					"option\n"));
9725 					usage(B_FALSE);
9726 				}
9727 
9728 				switch (getsubopt(&optarg, col_subopts,
9729 				    &value)) {
9730 				case 0:
9731 					cb.cb_columns[i++] = GET_COL_NAME;
9732 					break;
9733 				case 1:
9734 					cb.cb_columns[i++] = GET_COL_PROPERTY;
9735 					break;
9736 				case 2:
9737 					cb.cb_columns[i++] = GET_COL_VALUE;
9738 					break;
9739 				case 3:
9740 					cb.cb_columns[i++] = GET_COL_SOURCE;
9741 					break;
9742 				case 4:
9743 					if (i > 0) {
9744 						(void) fprintf(stderr,
9745 						    gettext("\"all\" conflicts "
9746 						    "with specific fields "
9747 						    "given to -o option\n"));
9748 						usage(B_FALSE);
9749 					}
9750 					cb.cb_columns[0] = GET_COL_NAME;
9751 					cb.cb_columns[1] = GET_COL_PROPERTY;
9752 					cb.cb_columns[2] = GET_COL_VALUE;
9753 					cb.cb_columns[3] = GET_COL_SOURCE;
9754 					i = ZFS_GET_NCOLS;
9755 					break;
9756 				default:
9757 					(void) fprintf(stderr,
9758 					    gettext("invalid column name "
9759 					    "'%s'\n"), value);
9760 					usage(B_FALSE);
9761 				}
9762 			}
9763 			break;
9764 		case '?':
9765 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9766 			    optopt);
9767 			usage(B_FALSE);
9768 		}
9769 	}
9770 
9771 	argc -= optind;
9772 	argv += optind;
9773 
9774 	if (argc < 1) {
9775 		(void) fprintf(stderr, gettext("missing property "
9776 		    "argument\n"));
9777 		usage(B_FALSE);
9778 	}
9779 
9780 	if (zprop_get_list(g_zfs, argv[0], &cb.cb_proplist,
9781 	    ZFS_TYPE_POOL) != 0)
9782 		usage(B_FALSE);
9783 
9784 	argc--;
9785 	argv++;
9786 
9787 	if (cb.cb_proplist != NULL) {
9788 		fake_name.pl_prop = ZPOOL_PROP_NAME;
9789 		fake_name.pl_width = strlen(gettext("NAME"));
9790 		fake_name.pl_next = cb.cb_proplist;
9791 		cb.cb_proplist = &fake_name;
9792 	}
9793 
9794 	ret = for_each_pool(argc, argv, B_TRUE, &cb.cb_proplist, cb.cb_literal,
9795 	    get_callback, &cb);
9796 
9797 	if (cb.cb_proplist == &fake_name)
9798 		zprop_free_list(fake_name.pl_next);
9799 	else
9800 		zprop_free_list(cb.cb_proplist);
9801 
9802 	return (ret);
9803 }
9804 
9805 typedef struct set_cbdata {
9806 	char *cb_propname;
9807 	char *cb_value;
9808 	boolean_t cb_any_successful;
9809 } set_cbdata_t;
9810 
9811 static int
9812 set_callback(zpool_handle_t *zhp, void *data)
9813 {
9814 	int error;
9815 	set_cbdata_t *cb = (set_cbdata_t *)data;
9816 
9817 	error = zpool_set_prop(zhp, cb->cb_propname, cb->cb_value);
9818 
9819 	if (!error)
9820 		cb->cb_any_successful = B_TRUE;
9821 
9822 	return (error);
9823 }
9824 
9825 int
9826 zpool_do_set(int argc, char **argv)
9827 {
9828 	set_cbdata_t cb = { 0 };
9829 	int error;
9830 
9831 	if (argc > 1 && argv[1][0] == '-') {
9832 		(void) fprintf(stderr, gettext("invalid option '%c'\n"),
9833 		    argv[1][1]);
9834 		usage(B_FALSE);
9835 	}
9836 
9837 	if (argc < 2) {
9838 		(void) fprintf(stderr, gettext("missing property=value "
9839 		    "argument\n"));
9840 		usage(B_FALSE);
9841 	}
9842 
9843 	if (argc < 3) {
9844 		(void) fprintf(stderr, gettext("missing pool name\n"));
9845 		usage(B_FALSE);
9846 	}
9847 
9848 	if (argc > 3) {
9849 		(void) fprintf(stderr, gettext("too many pool names\n"));
9850 		usage(B_FALSE);
9851 	}
9852 
9853 	cb.cb_propname = argv[1];
9854 	cb.cb_value = strchr(cb.cb_propname, '=');
9855 	if (cb.cb_value == NULL) {
9856 		(void) fprintf(stderr, gettext("missing value in "
9857 		    "property=value argument\n"));
9858 		usage(B_FALSE);
9859 	}
9860 
9861 	*(cb.cb_value) = '\0';
9862 	cb.cb_value++;
9863 
9864 	error = for_each_pool(argc - 2, argv + 2, B_TRUE, NULL, B_FALSE,
9865 	    set_callback, &cb);
9866 
9867 	return (error);
9868 }
9869 
9870 /* Add up the total number of bytes left to initialize/trim across all vdevs */
9871 static uint64_t
9872 vdev_activity_remaining(nvlist_t *nv, zpool_wait_activity_t activity)
9873 {
9874 	uint64_t bytes_remaining;
9875 	nvlist_t **child;
9876 	uint_t c, children;
9877 	vdev_stat_t *vs;
9878 
9879 	assert(activity == ZPOOL_WAIT_INITIALIZE ||
9880 	    activity == ZPOOL_WAIT_TRIM);
9881 
9882 	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
9883 	    (uint64_t **)&vs, &c) == 0);
9884 
9885 	if (activity == ZPOOL_WAIT_INITIALIZE &&
9886 	    vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE)
9887 		bytes_remaining = vs->vs_initialize_bytes_est -
9888 		    vs->vs_initialize_bytes_done;
9889 	else if (activity == ZPOOL_WAIT_TRIM &&
9890 	    vs->vs_trim_state == VDEV_TRIM_ACTIVE)
9891 		bytes_remaining = vs->vs_trim_bytes_est -
9892 		    vs->vs_trim_bytes_done;
9893 	else
9894 		bytes_remaining = 0;
9895 
9896 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
9897 	    &child, &children) != 0)
9898 		children = 0;
9899 
9900 	for (c = 0; c < children; c++)
9901 		bytes_remaining += vdev_activity_remaining(child[c], activity);
9902 
9903 	return (bytes_remaining);
9904 }
9905 
9906 /* Add up the total number of bytes left to rebuild across top-level vdevs */
9907 static uint64_t
9908 vdev_activity_top_remaining(nvlist_t *nv)
9909 {
9910 	uint64_t bytes_remaining = 0;
9911 	nvlist_t **child;
9912 	uint_t children;
9913 	int error;
9914 
9915 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
9916 	    &child, &children) != 0)
9917 		children = 0;
9918 
9919 	for (uint_t c = 0; c < children; c++) {
9920 		vdev_rebuild_stat_t *vrs;
9921 		uint_t i;
9922 
9923 		error = nvlist_lookup_uint64_array(child[c],
9924 		    ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i);
9925 		if (error == 0) {
9926 			if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
9927 				bytes_remaining += (vrs->vrs_bytes_est -
9928 				    vrs->vrs_bytes_rebuilt);
9929 			}
9930 		}
9931 	}
9932 
9933 	return (bytes_remaining);
9934 }
9935 
9936 /* Whether any vdevs are 'spare' or 'replacing' vdevs */
9937 static boolean_t
9938 vdev_any_spare_replacing(nvlist_t *nv)
9939 {
9940 	nvlist_t **child;
9941 	uint_t c, children;
9942 	char *vdev_type;
9943 
9944 	(void) nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &vdev_type);
9945 
9946 	if (strcmp(vdev_type, VDEV_TYPE_REPLACING) == 0 ||
9947 	    strcmp(vdev_type, VDEV_TYPE_SPARE) == 0 ||
9948 	    strcmp(vdev_type, VDEV_TYPE_DRAID_SPARE) == 0) {
9949 		return (B_TRUE);
9950 	}
9951 
9952 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
9953 	    &child, &children) != 0)
9954 		children = 0;
9955 
9956 	for (c = 0; c < children; c++) {
9957 		if (vdev_any_spare_replacing(child[c]))
9958 			return (B_TRUE);
9959 	}
9960 
9961 	return (B_FALSE);
9962 }
9963 
9964 typedef struct wait_data {
9965 	char *wd_poolname;
9966 	boolean_t wd_scripted;
9967 	boolean_t wd_exact;
9968 	boolean_t wd_headers_once;
9969 	boolean_t wd_should_exit;
9970 	/* Which activities to wait for */
9971 	boolean_t wd_enabled[ZPOOL_WAIT_NUM_ACTIVITIES];
9972 	float wd_interval;
9973 	pthread_cond_t wd_cv;
9974 	pthread_mutex_t wd_mutex;
9975 } wait_data_t;
9976 
9977 /*
9978  * Print to stdout a single line, containing one column for each activity that
9979  * we are waiting for specifying how many bytes of work are left for that
9980  * activity.
9981  */
9982 static void
9983 print_wait_status_row(wait_data_t *wd, zpool_handle_t *zhp, int row)
9984 {
9985 	nvlist_t *config, *nvroot;
9986 	uint_t c;
9987 	int i;
9988 	pool_checkpoint_stat_t *pcs = NULL;
9989 	pool_scan_stat_t *pss = NULL;
9990 	pool_removal_stat_t *prs = NULL;
9991 	char *headers[] = {"DISCARD", "FREE", "INITIALIZE", "REPLACE",
9992 	    "REMOVE", "RESILVER", "SCRUB", "TRIM"};
9993 	int col_widths[ZPOOL_WAIT_NUM_ACTIVITIES];
9994 
9995 	/* Calculate the width of each column */
9996 	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
9997 		/*
9998 		 * Make sure we have enough space in the col for pretty-printed
9999 		 * numbers and for the column header, and then leave a couple
10000 		 * spaces between cols for readability.
10001 		 */
10002 		col_widths[i] = MAX(strlen(headers[i]), 6) + 2;
10003 	}
10004 
10005 	/* Print header if appropriate */
10006 	int term_height = terminal_height();
10007 	boolean_t reprint_header = (!wd->wd_headers_once && term_height > 0 &&
10008 	    row % (term_height-1) == 0);
10009 	if (!wd->wd_scripted && (row == 0 || reprint_header)) {
10010 		for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10011 			if (wd->wd_enabled[i])
10012 				(void) printf("%*s", col_widths[i], headers[i]);
10013 		}
10014 		(void) printf("\n");
10015 	}
10016 
10017 	/* Bytes of work remaining in each activity */
10018 	int64_t bytes_rem[ZPOOL_WAIT_NUM_ACTIVITIES] = {0};
10019 
10020 	bytes_rem[ZPOOL_WAIT_FREE] =
10021 	    zpool_get_prop_int(zhp, ZPOOL_PROP_FREEING, NULL);
10022 
10023 	config = zpool_get_config(zhp, NULL);
10024 	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
10025 
10026 	(void) nvlist_lookup_uint64_array(nvroot,
10027 	    ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
10028 	if (pcs != NULL && pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
10029 		bytes_rem[ZPOOL_WAIT_CKPT_DISCARD] = pcs->pcs_space;
10030 
10031 	(void) nvlist_lookup_uint64_array(nvroot,
10032 	    ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
10033 	if (prs != NULL && prs->prs_state == DSS_SCANNING)
10034 		bytes_rem[ZPOOL_WAIT_REMOVE] = prs->prs_to_copy -
10035 		    prs->prs_copied;
10036 
10037 	(void) nvlist_lookup_uint64_array(nvroot,
10038 	    ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&pss, &c);
10039 	if (pss != NULL && pss->pss_state == DSS_SCANNING &&
10040 	    pss->pss_pass_scrub_pause == 0) {
10041 		int64_t rem = pss->pss_to_examine - pss->pss_issued;
10042 		if (pss->pss_func == POOL_SCAN_SCRUB)
10043 			bytes_rem[ZPOOL_WAIT_SCRUB] = rem;
10044 		else
10045 			bytes_rem[ZPOOL_WAIT_RESILVER] = rem;
10046 	} else if (check_rebuilding(nvroot, NULL)) {
10047 		bytes_rem[ZPOOL_WAIT_RESILVER] =
10048 		    vdev_activity_top_remaining(nvroot);
10049 	}
10050 
10051 	bytes_rem[ZPOOL_WAIT_INITIALIZE] =
10052 	    vdev_activity_remaining(nvroot, ZPOOL_WAIT_INITIALIZE);
10053 	bytes_rem[ZPOOL_WAIT_TRIM] =
10054 	    vdev_activity_remaining(nvroot, ZPOOL_WAIT_TRIM);
10055 
10056 	/*
10057 	 * A replace finishes after resilvering finishes, so the amount of work
10058 	 * left for a replace is the same as for resilvering.
10059 	 *
10060 	 * It isn't quite correct to say that if we have any 'spare' or
10061 	 * 'replacing' vdevs and a resilver is happening, then a replace is in
10062 	 * progress, like we do here. When a hot spare is used, the faulted vdev
10063 	 * is not removed after the hot spare is resilvered, so parent 'spare'
10064 	 * vdev is not removed either. So we could have a 'spare' vdev, but be
10065 	 * resilvering for a different reason. However, we use it as a heuristic
10066 	 * because we don't have access to the DTLs, which could tell us whether
10067 	 * or not we have really finished resilvering a hot spare.
10068 	 */
10069 	if (vdev_any_spare_replacing(nvroot))
10070 		bytes_rem[ZPOOL_WAIT_REPLACE] =  bytes_rem[ZPOOL_WAIT_RESILVER];
10071 
10072 	if (timestamp_fmt != NODATE)
10073 		print_timestamp(timestamp_fmt);
10074 
10075 	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10076 		char buf[64];
10077 		if (!wd->wd_enabled[i])
10078 			continue;
10079 
10080 		if (wd->wd_exact)
10081 			(void) snprintf(buf, sizeof (buf), "%" PRIi64,
10082 			    bytes_rem[i]);
10083 		else
10084 			zfs_nicenum(bytes_rem[i], buf, sizeof (buf));
10085 
10086 		if (wd->wd_scripted)
10087 			(void) printf(i == 0 ? "%s" : "\t%s", buf);
10088 		else
10089 			(void) printf(" %*s", col_widths[i] - 1, buf);
10090 	}
10091 	(void) printf("\n");
10092 	(void) fflush(stdout);
10093 }
10094 
10095 static void *
10096 wait_status_thread(void *arg)
10097 {
10098 	wait_data_t *wd = (wait_data_t *)arg;
10099 	zpool_handle_t *zhp;
10100 
10101 	if ((zhp = zpool_open(g_zfs, wd->wd_poolname)) == NULL)
10102 		return (void *)(1);
10103 
10104 	for (int row = 0; ; row++) {
10105 		boolean_t missing;
10106 		struct timespec timeout;
10107 		int ret = 0;
10108 		(void) clock_gettime(CLOCK_REALTIME, &timeout);
10109 
10110 		if (zpool_refresh_stats(zhp, &missing) != 0 || missing ||
10111 		    zpool_props_refresh(zhp) != 0) {
10112 			zpool_close(zhp);
10113 			return (void *)(uintptr_t)(missing ? 0 : 1);
10114 		}
10115 
10116 		print_wait_status_row(wd, zhp, row);
10117 
10118 		timeout.tv_sec += floor(wd->wd_interval);
10119 		long nanos = timeout.tv_nsec +
10120 		    (wd->wd_interval - floor(wd->wd_interval)) * NANOSEC;
10121 		if (nanos >= NANOSEC) {
10122 			timeout.tv_sec++;
10123 			timeout.tv_nsec = nanos - NANOSEC;
10124 		} else {
10125 			timeout.tv_nsec = nanos;
10126 		}
10127 		pthread_mutex_lock(&wd->wd_mutex);
10128 		if (!wd->wd_should_exit)
10129 			ret = pthread_cond_timedwait(&wd->wd_cv, &wd->wd_mutex,
10130 			    &timeout);
10131 		pthread_mutex_unlock(&wd->wd_mutex);
10132 		if (ret == 0) {
10133 			break; /* signaled by main thread */
10134 		} else if (ret != ETIMEDOUT) {
10135 			(void) fprintf(stderr, gettext("pthread_cond_timedwait "
10136 			    "failed: %s\n"), strerror(ret));
10137 			zpool_close(zhp);
10138 			return (void *)(uintptr_t)(1);
10139 		}
10140 	}
10141 
10142 	zpool_close(zhp);
10143 	return (void *)(0);
10144 }
10145 
10146 int
10147 zpool_do_wait(int argc, char **argv)
10148 {
10149 	boolean_t verbose = B_FALSE;
10150 	int c;
10151 	char *value;
10152 	int i;
10153 	unsigned long count;
10154 	pthread_t status_thr;
10155 	int error = 0;
10156 	zpool_handle_t *zhp;
10157 
10158 	wait_data_t wd;
10159 	wd.wd_scripted = B_FALSE;
10160 	wd.wd_exact = B_FALSE;
10161 	wd.wd_headers_once = B_FALSE;
10162 	wd.wd_should_exit = B_FALSE;
10163 
10164 	pthread_mutex_init(&wd.wd_mutex, NULL);
10165 	pthread_cond_init(&wd.wd_cv, NULL);
10166 
10167 	/* By default, wait for all types of activity. */
10168 	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++)
10169 		wd.wd_enabled[i] = B_TRUE;
10170 
10171 	while ((c = getopt(argc, argv, "HpT:t:")) != -1) {
10172 		switch (c) {
10173 		case 'H':
10174 			wd.wd_scripted = B_TRUE;
10175 			break;
10176 		case 'n':
10177 			wd.wd_headers_once = B_TRUE;
10178 			break;
10179 		case 'p':
10180 			wd.wd_exact = B_TRUE;
10181 			break;
10182 		case 'T':
10183 			get_timestamp_arg(*optarg);
10184 			break;
10185 		case 't':
10186 		{
10187 			static char *col_subopts[] = { "discard", "free",
10188 			    "initialize", "replace", "remove", "resilver",
10189 			    "scrub", "trim", NULL };
10190 
10191 			/* Reset activities array */
10192 			bzero(&wd.wd_enabled, sizeof (wd.wd_enabled));
10193 			while (*optarg != '\0') {
10194 				int activity = getsubopt(&optarg, col_subopts,
10195 				    &value);
10196 
10197 				if (activity < 0) {
10198 					(void) fprintf(stderr,
10199 					    gettext("invalid activity '%s'\n"),
10200 					    value);
10201 					usage(B_FALSE);
10202 				}
10203 
10204 				wd.wd_enabled[activity] = B_TRUE;
10205 			}
10206 			break;
10207 		}
10208 		case '?':
10209 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
10210 			    optopt);
10211 			usage(B_FALSE);
10212 		}
10213 	}
10214 
10215 	argc -= optind;
10216 	argv += optind;
10217 
10218 	get_interval_count(&argc, argv, &wd.wd_interval, &count);
10219 	if (count != 0) {
10220 		/* This subcmd only accepts an interval, not a count */
10221 		(void) fprintf(stderr, gettext("too many arguments\n"));
10222 		usage(B_FALSE);
10223 	}
10224 
10225 	if (wd.wd_interval != 0)
10226 		verbose = B_TRUE;
10227 
10228 	if (argc < 1) {
10229 		(void) fprintf(stderr, gettext("missing 'pool' argument\n"));
10230 		usage(B_FALSE);
10231 	}
10232 	if (argc > 1) {
10233 		(void) fprintf(stderr, gettext("too many arguments\n"));
10234 		usage(B_FALSE);
10235 	}
10236 
10237 	wd.wd_poolname = argv[0];
10238 
10239 	if ((zhp = zpool_open(g_zfs, wd.wd_poolname)) == NULL)
10240 		return (1);
10241 
10242 	if (verbose) {
10243 		/*
10244 		 * We use a separate thread for printing status updates because
10245 		 * the main thread will call lzc_wait(), which blocks as long
10246 		 * as an activity is in progress, which can be a long time.
10247 		 */
10248 		if (pthread_create(&status_thr, NULL, wait_status_thread, &wd)
10249 		    != 0) {
10250 			(void) fprintf(stderr, gettext("failed to create status"
10251 			    "thread: %s\n"), strerror(errno));
10252 			zpool_close(zhp);
10253 			return (1);
10254 		}
10255 	}
10256 
10257 	/*
10258 	 * Loop over all activities that we are supposed to wait for until none
10259 	 * of them are in progress. Note that this means we can end up waiting
10260 	 * for more activities to complete than just those that were in progress
10261 	 * when we began waiting; if an activity we are interested in begins
10262 	 * while we are waiting for another activity, we will wait for both to
10263 	 * complete before exiting.
10264 	 */
10265 	for (;;) {
10266 		boolean_t missing = B_FALSE;
10267 		boolean_t any_waited = B_FALSE;
10268 
10269 		for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
10270 			boolean_t waited;
10271 
10272 			if (!wd.wd_enabled[i])
10273 				continue;
10274 
10275 			error = zpool_wait_status(zhp, i, &missing, &waited);
10276 			if (error != 0 || missing)
10277 				break;
10278 
10279 			any_waited = (any_waited || waited);
10280 		}
10281 
10282 		if (error != 0 || missing || !any_waited)
10283 			break;
10284 	}
10285 
10286 	zpool_close(zhp);
10287 
10288 	if (verbose) {
10289 		uintptr_t status;
10290 		pthread_mutex_lock(&wd.wd_mutex);
10291 		wd.wd_should_exit = B_TRUE;
10292 		pthread_cond_signal(&wd.wd_cv);
10293 		pthread_mutex_unlock(&wd.wd_mutex);
10294 		(void) pthread_join(status_thr, (void *)&status);
10295 		if (status != 0)
10296 			error = status;
10297 	}
10298 
10299 	pthread_mutex_destroy(&wd.wd_mutex);
10300 	pthread_cond_destroy(&wd.wd_cv);
10301 	return (error);
10302 }
10303 
10304 static int
10305 find_command_idx(char *command, int *idx)
10306 {
10307 	int i;
10308 
10309 	for (i = 0; i < NCOMMAND; i++) {
10310 		if (command_table[i].name == NULL)
10311 			continue;
10312 
10313 		if (strcmp(command, command_table[i].name) == 0) {
10314 			*idx = i;
10315 			return (0);
10316 		}
10317 	}
10318 	return (1);
10319 }
10320 
10321 /*
10322  * Display version message
10323  */
10324 static int
10325 zpool_do_version(int argc, char **argv)
10326 {
10327 	if (zfs_version_print() == -1)
10328 		return (1);
10329 
10330 	return (0);
10331 }
10332 
10333 int
10334 main(int argc, char **argv)
10335 {
10336 	int ret = 0;
10337 	int i = 0;
10338 	char *cmdname;
10339 	char **newargv;
10340 
10341 	(void) setlocale(LC_ALL, "");
10342 	(void) setlocale(LC_NUMERIC, "C");
10343 	(void) textdomain(TEXT_DOMAIN);
10344 	srand(time(NULL));
10345 
10346 	opterr = 0;
10347 
10348 	/*
10349 	 * Make sure the user has specified some command.
10350 	 */
10351 	if (argc < 2) {
10352 		(void) fprintf(stderr, gettext("missing command\n"));
10353 		usage(B_FALSE);
10354 	}
10355 
10356 	cmdname = argv[1];
10357 
10358 	/*
10359 	 * Special case '-?'
10360 	 */
10361 	if ((strcmp(cmdname, "-?") == 0) || strcmp(cmdname, "--help") == 0)
10362 		usage(B_TRUE);
10363 
10364 	/*
10365 	 * Special case '-V|--version'
10366 	 */
10367 	if ((strcmp(cmdname, "-V") == 0) || (strcmp(cmdname, "--version") == 0))
10368 		return (zpool_do_version(argc, argv));
10369 
10370 	if ((g_zfs = libzfs_init()) == NULL) {
10371 		(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
10372 		return (1);
10373 	}
10374 
10375 	libzfs_print_on_error(g_zfs, B_TRUE);
10376 
10377 	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));
10378 
10379 	/*
10380 	 * Many commands modify input strings for string parsing reasons.
10381 	 * We create a copy to protect the original argv.
10382 	 */
10383 	newargv = malloc((argc + 1) * sizeof (newargv[0]));
10384 	for (i = 0; i < argc; i++)
10385 		newargv[i] = strdup(argv[i]);
10386 	newargv[argc] = NULL;
10387 
10388 	/*
10389 	 * Run the appropriate command.
10390 	 */
10391 	if (find_command_idx(cmdname, &i) == 0) {
10392 		current_command = &command_table[i];
10393 		ret = command_table[i].func(argc - 1, newargv + 1);
10394 	} else if (strchr(cmdname, '=')) {
10395 		verify(find_command_idx("set", &i) == 0);
10396 		current_command = &command_table[i];
10397 		ret = command_table[i].func(argc, newargv);
10398 	} else if (strcmp(cmdname, "freeze") == 0 && argc == 3) {
10399 		/*
10400 		 * 'freeze' is a vile debugging abomination, so we treat
10401 		 * it as such.
10402 		 */
10403 		zfs_cmd_t zc = {"\0"};
10404 
10405 		(void) strlcpy(zc.zc_name, argv[2], sizeof (zc.zc_name));
10406 		ret = zfs_ioctl(g_zfs, ZFS_IOC_POOL_FREEZE, &zc);
10407 		if (ret != 0) {
10408 			(void) fprintf(stderr,
10409 			gettext("failed to freeze pool: %d\n"), errno);
10410 			ret = 1;
10411 		}
10412 
10413 		log_history = 0;
10414 	} else {
10415 		(void) fprintf(stderr, gettext("unrecognized "
10416 		    "command '%s'\n"), cmdname);
10417 		usage(B_FALSE);
10418 		ret = 1;
10419 	}
10420 
10421 	for (i = 0; i < argc; i++)
10422 		free(newargv[i]);
10423 	free(newargv);
10424 
10425 	if (ret == 0 && log_history)
10426 		(void) zpool_log_history(g_zfs, history_str);
10427 
10428 	libzfs_fini(g_zfs);
10429 
10430 	/*
10431 	 * The 'ZFS_ABORT' environment variable causes us to dump core on exit
10432 	 * for the purposes of running ::findleaks.
10433 	 */
10434 	if (getenv("ZFS_ABORT") != NULL) {
10435 		(void) printf("dumping core by request\n");
10436 		abort();
10437 	}
10438 
10439 	return (ret);
10440 }
10441