xref: /illumos-gate/usr/src/lib/libzfs/common/libzfs_status.c (revision 34bdffbf3e3c188027e767e631f717b10159316d)
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 (c) 2012 by Delphix. All rights reserved.
25  */
26 
27 /*
28  * This file contains the functions which analyze the status of a pool.  This
29  * include both the status of an active pool, as well as the status exported
30  * pools.  Returns one of the ZPOOL_STATUS_* defines describing the status of
31  * the pool.  This status is independent (to a certain degree) from the state of
32  * the pool.  A pool's state describes only whether or not it is capable of
33  * providing the necessary fault tolerance for data.  The status describes the
34  * overall status of devices.  A pool that is online can still have a device
35  * that is experiencing errors.
36  *
37  * Only a subset of the possible faults can be detected using 'zpool status',
38  * and not all possible errors correspond to a FMA message ID.  The explanation
39  * is left up to the caller, depending on whether it is a live pool or an
40  * import.
41  */
42 
43 #include <libzfs.h>
44 #include <string.h>
45 #include <unistd.h>
46 #include "libzfs_impl.h"
47 #include "zfeature_common.h"
48 
49 /*
50  * Message ID table.  This must be kept in sync with the ZPOOL_STATUS_* defines
51  * in libzfs.h.  Note that there are some status results which go past the end
52  * of this table, and hence have no associated message ID.
53  */
54 static char *zfs_msgid_table[] = {
55 	"ZFS-8000-14",
56 	"ZFS-8000-2Q",
57 	"ZFS-8000-3C",
58 	"ZFS-8000-4J",
59 	"ZFS-8000-5E",
60 	"ZFS-8000-6X",
61 	"ZFS-8000-72",
62 	"ZFS-8000-8A",
63 	"ZFS-8000-9P",
64 	"ZFS-8000-A5",
65 	"ZFS-8000-EY",
66 	"ZFS-8000-HC",
67 	"ZFS-8000-JQ",
68 	"ZFS-8000-K4",
69 };
70 
71 #define	NMSGID	(sizeof (zfs_msgid_table) / sizeof (zfs_msgid_table[0]))
72 
73 /* ARGSUSED */
74 static int
75 vdev_missing(uint64_t state, uint64_t aux, uint64_t errs)
76 {
77 	return (state == VDEV_STATE_CANT_OPEN &&
78 	    aux == VDEV_AUX_OPEN_FAILED);
79 }
80 
81 /* ARGSUSED */
82 static int
83 vdev_faulted(uint64_t state, uint64_t aux, uint64_t errs)
84 {
85 	return (state == VDEV_STATE_FAULTED);
86 }
87 
88 /* ARGSUSED */
89 static int
90 vdev_errors(uint64_t state, uint64_t aux, uint64_t errs)
91 {
92 	return (state == VDEV_STATE_DEGRADED || errs != 0);
93 }
94 
95 /* ARGSUSED */
96 static int
97 vdev_broken(uint64_t state, uint64_t aux, uint64_t errs)
98 {
99 	return (state == VDEV_STATE_CANT_OPEN);
100 }
101 
102 /* ARGSUSED */
103 static int
104 vdev_offlined(uint64_t state, uint64_t aux, uint64_t errs)
105 {
106 	return (state == VDEV_STATE_OFFLINE);
107 }
108 
109 /* ARGSUSED */
110 static int
111 vdev_removed(uint64_t state, uint64_t aux, uint64_t errs)
112 {
113 	return (state == VDEV_STATE_REMOVED);
114 }
115 
116 /*
117  * Detect if any leaf devices that have seen errors or could not be opened.
118  */
119 static boolean_t
120 find_vdev_problem(nvlist_t *vdev, int (*func)(uint64_t, uint64_t, uint64_t))
121 {
122 	nvlist_t **child;
123 	vdev_stat_t *vs;
124 	uint_t c, children;
125 	char *type;
126 
127 	/*
128 	 * Ignore problems within a 'replacing' vdev, since we're presumably in
129 	 * the process of repairing any such errors, and don't want to call them
130 	 * out again.  We'll pick up the fact that a resilver is happening
131 	 * later.
132 	 */
133 	verify(nvlist_lookup_string(vdev, ZPOOL_CONFIG_TYPE, &type) == 0);
134 	if (strcmp(type, VDEV_TYPE_REPLACING) == 0)
135 		return (B_FALSE);
136 
137 	if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, &child,
138 	    &children) == 0) {
139 		for (c = 0; c < children; c++)
140 			if (find_vdev_problem(child[c], func))
141 				return (B_TRUE);
142 	} else {
143 		verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
144 		    (uint64_t **)&vs, &c) == 0);
145 
146 		if (func(vs->vs_state, vs->vs_aux,
147 		    vs->vs_read_errors +
148 		    vs->vs_write_errors +
149 		    vs->vs_checksum_errors))
150 			return (B_TRUE);
151 	}
152 
153 	return (B_FALSE);
154 }
155 
156 /*
157  * Active pool health status.
158  *
159  * To determine the status for a pool, we make several passes over the config,
160  * picking the most egregious error we find.  In order of importance, we do the
161  * following:
162  *
163  *	- Check for a complete and valid configuration
164  *	- Look for any faulted or missing devices in a non-replicated config
165  *	- Check for any data errors
166  *	- Check for any faulted or missing devices in a replicated config
167  *	- Look for any devices showing errors
168  *	- Check for any resilvering devices
169  *
170  * There can obviously be multiple errors within a single pool, so this routine
171  * only picks the most damaging of all the current errors to report.
172  */
173 static zpool_status_t
174 check_status(nvlist_t *config, boolean_t isimport)
175 {
176 	nvlist_t *nvroot;
177 	vdev_stat_t *vs;
178 	pool_scan_stat_t *ps = NULL;
179 	uint_t vsc, psc;
180 	uint64_t nerr;
181 	uint64_t version;
182 	uint64_t stateval;
183 	uint64_t suspended;
184 	uint64_t hostid = 0;
185 
186 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
187 	    &version) == 0);
188 	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
189 	    &nvroot) == 0);
190 	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
191 	    (uint64_t **)&vs, &vsc) == 0);
192 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
193 	    &stateval) == 0);
194 
195 	/*
196 	 * Currently resilvering a vdev
197 	 */
198 	(void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
199 	    (uint64_t **)&ps, &psc);
200 	if (ps && ps->pss_func == POOL_SCAN_RESILVER &&
201 	    ps->pss_state == DSS_SCANNING)
202 		return (ZPOOL_STATUS_RESILVERING);
203 
204 	/*
205 	 * Pool last accessed by another system.
206 	 */
207 	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
208 	if (hostid != 0 && (unsigned long)hostid != gethostid() &&
209 	    stateval == POOL_STATE_ACTIVE)
210 		return (ZPOOL_STATUS_HOSTID_MISMATCH);
211 
212 	/*
213 	 * Newer on-disk version.
214 	 */
215 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
216 	    vs->vs_aux == VDEV_AUX_VERSION_NEWER)
217 		return (ZPOOL_STATUS_VERSION_NEWER);
218 
219 	/*
220 	 * Unsupported feature(s).
221 	 */
222 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
223 	    vs->vs_aux == VDEV_AUX_UNSUP_FEAT) {
224 		nvlist_t *nvinfo;
225 
226 		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
227 		    &nvinfo) == 0);
228 		if (nvlist_exists(nvinfo, ZPOOL_CONFIG_CAN_RDONLY))
229 			return (ZPOOL_STATUS_UNSUP_FEAT_WRITE);
230 		return (ZPOOL_STATUS_UNSUP_FEAT_READ);
231 	}
232 
233 	/*
234 	 * Check that the config is complete.
235 	 */
236 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
237 	    vs->vs_aux == VDEV_AUX_BAD_GUID_SUM)
238 		return (ZPOOL_STATUS_BAD_GUID_SUM);
239 
240 	/*
241 	 * Check whether the pool has suspended due to failed I/O.
242 	 */
243 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED,
244 	    &suspended) == 0) {
245 		if (suspended == ZIO_FAILURE_MODE_CONTINUE)
246 			return (ZPOOL_STATUS_IO_FAILURE_CONTINUE);
247 		return (ZPOOL_STATUS_IO_FAILURE_WAIT);
248 	}
249 
250 	/*
251 	 * Could not read a log.
252 	 */
253 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
254 	    vs->vs_aux == VDEV_AUX_BAD_LOG) {
255 		return (ZPOOL_STATUS_BAD_LOG);
256 	}
257 
258 	/*
259 	 * Bad devices in non-replicated config.
260 	 */
261 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
262 	    find_vdev_problem(nvroot, vdev_faulted))
263 		return (ZPOOL_STATUS_FAULTED_DEV_NR);
264 
265 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
266 	    find_vdev_problem(nvroot, vdev_missing))
267 		return (ZPOOL_STATUS_MISSING_DEV_NR);
268 
269 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
270 	    find_vdev_problem(nvroot, vdev_broken))
271 		return (ZPOOL_STATUS_CORRUPT_LABEL_NR);
272 
273 	/*
274 	 * Corrupted pool metadata
275 	 */
276 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
277 	    vs->vs_aux == VDEV_AUX_CORRUPT_DATA)
278 		return (ZPOOL_STATUS_CORRUPT_POOL);
279 
280 	/*
281 	 * Persistent data errors.
282 	 */
283 	if (!isimport) {
284 		if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
285 		    &nerr) == 0 && nerr != 0)
286 			return (ZPOOL_STATUS_CORRUPT_DATA);
287 	}
288 
289 	/*
290 	 * Missing devices in a replicated config.
291 	 */
292 	if (find_vdev_problem(nvroot, vdev_faulted))
293 		return (ZPOOL_STATUS_FAULTED_DEV_R);
294 	if (find_vdev_problem(nvroot, vdev_missing))
295 		return (ZPOOL_STATUS_MISSING_DEV_R);
296 	if (find_vdev_problem(nvroot, vdev_broken))
297 		return (ZPOOL_STATUS_CORRUPT_LABEL_R);
298 
299 	/*
300 	 * Devices with errors
301 	 */
302 	if (!isimport && find_vdev_problem(nvroot, vdev_errors))
303 		return (ZPOOL_STATUS_FAILING_DEV);
304 
305 	/*
306 	 * Offlined devices
307 	 */
308 	if (find_vdev_problem(nvroot, vdev_offlined))
309 		return (ZPOOL_STATUS_OFFLINE_DEV);
310 
311 	/*
312 	 * Removed device
313 	 */
314 	if (find_vdev_problem(nvroot, vdev_removed))
315 		return (ZPOOL_STATUS_REMOVED_DEV);
316 
317 	/*
318 	 * Outdated, but usable, version
319 	 */
320 	if (SPA_VERSION_IS_SUPPORTED(version) && version != SPA_VERSION)
321 		return (ZPOOL_STATUS_VERSION_OLDER);
322 
323 	/*
324 	 * Usable pool with disabled features
325 	 */
326 	if (version >= SPA_VERSION_FEATURES) {
327 		int i;
328 		nvlist_t *feat;
329 
330 		if (isimport) {
331 			feat = fnvlist_lookup_nvlist(config,
332 			    ZPOOL_CONFIG_LOAD_INFO);
333 			feat = fnvlist_lookup_nvlist(feat,
334 			    ZPOOL_CONFIG_ENABLED_FEAT);
335 		} else {
336 			feat = fnvlist_lookup_nvlist(config,
337 			    ZPOOL_CONFIG_FEATURE_STATS);
338 		}
339 
340 		for (i = 0; i < SPA_FEATURES; i++) {
341 			zfeature_info_t *fi = &spa_feature_table[i];
342 			if (!nvlist_exists(feat, fi->fi_guid))
343 				return (ZPOOL_STATUS_FEAT_DISABLED);
344 		}
345 	}
346 
347 	return (ZPOOL_STATUS_OK);
348 }
349 
350 zpool_status_t
351 zpool_get_status(zpool_handle_t *zhp, char **msgid)
352 {
353 	zpool_status_t ret = check_status(zhp->zpool_config, B_FALSE);
354 
355 	if (ret >= NMSGID)
356 		*msgid = NULL;
357 	else
358 		*msgid = zfs_msgid_table[ret];
359 
360 	return (ret);
361 }
362 
363 zpool_status_t
364 zpool_import_status(nvlist_t *config, char **msgid)
365 {
366 	zpool_status_t ret = check_status(config, B_TRUE);
367 
368 	if (ret >= NMSGID)
369 		*msgid = NULL;
370 	else
371 		*msgid = zfs_msgid_table[ret];
372 
373 	return (ret);
374 }
375 
376 static void
377 dump_ddt_stat(const ddt_stat_t *dds, int h)
378 {
379 	char refcnt[6];
380 	char blocks[6], lsize[6], psize[6], dsize[6];
381 	char ref_blocks[6], ref_lsize[6], ref_psize[6], ref_dsize[6];
382 
383 	if (dds == NULL || dds->dds_blocks == 0)
384 		return;
385 
386 	if (h == -1)
387 		(void) strcpy(refcnt, "Total");
388 	else
389 		zfs_nicenum(1ULL << h, refcnt, sizeof (refcnt));
390 
391 	zfs_nicenum(dds->dds_blocks, blocks, sizeof (blocks));
392 	zfs_nicenum(dds->dds_lsize, lsize, sizeof (lsize));
393 	zfs_nicenum(dds->dds_psize, psize, sizeof (psize));
394 	zfs_nicenum(dds->dds_dsize, dsize, sizeof (dsize));
395 	zfs_nicenum(dds->dds_ref_blocks, ref_blocks, sizeof (ref_blocks));
396 	zfs_nicenum(dds->dds_ref_lsize, ref_lsize, sizeof (ref_lsize));
397 	zfs_nicenum(dds->dds_ref_psize, ref_psize, sizeof (ref_psize));
398 	zfs_nicenum(dds->dds_ref_dsize, ref_dsize, sizeof (ref_dsize));
399 
400 	(void) printf("%6s   %6s   %5s   %5s   %5s   %6s   %5s   %5s   %5s\n",
401 	    refcnt,
402 	    blocks, lsize, psize, dsize,
403 	    ref_blocks, ref_lsize, ref_psize, ref_dsize);
404 }
405 
406 /*
407  * Print the DDT histogram and the column totals.
408  */
409 void
410 zpool_dump_ddt(const ddt_stat_t *dds_total, const ddt_histogram_t *ddh)
411 {
412 	int h;
413 
414 	(void) printf("\n");
415 
416 	(void) printf("bucket   "
417 	    "           allocated             "
418 	    "          referenced          \n");
419 	(void) printf("______   "
420 	    "______________________________   "
421 	    "______________________________\n");
422 
423 	(void) printf("%6s   %6s   %5s   %5s   %5s   %6s   %5s   %5s   %5s\n",
424 	    "refcnt",
425 	    "blocks", "LSIZE", "PSIZE", "DSIZE",
426 	    "blocks", "LSIZE", "PSIZE", "DSIZE");
427 
428 	(void) printf("%6s   %6s   %5s   %5s   %5s   %6s   %5s   %5s   %5s\n",
429 	    "------",
430 	    "------", "-----", "-----", "-----",
431 	    "------", "-----", "-----", "-----");
432 
433 	for (h = 0; h < 64; h++)
434 		dump_ddt_stat(&ddh->ddh_stat[h], h);
435 
436 	dump_ddt_stat(dds_total, -1);
437 
438 	(void) printf("\n");
439 }
440