OpenGrok

Cross Reference: libzfs_status.c
xref: /onnv/onnv-gate/usr/src/lib/libzfs/common/libzfs_status.c
Home | History | Annotate | Line # | Download | only in common
      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  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
     23  */
     24 
     25 /*
     26  * This file contains the functions which analyze the status of a pool.  This
     27  * include both the status of an active pool, as well as the status exported
     28  * pools.  Returns one of the ZPOOL_STATUS_* defines describing the status of
     29  * the pool.  This status is independent (to a certain degree) from the state of
     30  * the pool.  A pool's state describes only whether or not it is capable of
     31  * providing the necessary fault tolerance for data.  The status describes the
     32  * overall status of devices.  A pool that is online can still have a device
     33  * that is experiencing errors.
     34  *
     35  * Only a subset of the possible faults can be detected using 'zpool status',
     36  * and not all possible errors correspond to a FMA message ID.  The explanation
     37  * is left up to the caller, depending on whether it is a live pool or an
     38  * import.
     39  */
     40 
     41 #include <libzfs.h>
     42 #include <string.h>
     43 #include <unistd.h>
     44 #include "libzfs_impl.h"
     45 
     46 /*
     47  * Message ID table.  This must be kept in sync with the ZPOOL_STATUS_* defines
     48  * in libzfs.h.  Note that there are some status results which go past the end
     49  * of this table, and hence have no associated message ID.
     50  */
     51 static char *zfs_msgid_table[] = {
     52 	"ZFS-8000-14",
     53 	"ZFS-8000-2Q",
     54 	"ZFS-8000-3C",
     55 	"ZFS-8000-4J",
     56 	"ZFS-8000-5E",
     57 	"ZFS-8000-6X",
     58 	"ZFS-8000-72",
     59 	"ZFS-8000-8A",
     60 	"ZFS-8000-9P",
     61 	"ZFS-8000-A5",
     62 	"ZFS-8000-EY",
     63 	"ZFS-8000-HC",
     64 	"ZFS-8000-JQ",
     65 	"ZFS-8000-K4",
     66 };
     67 
     68 #define	NMSGID	(sizeof (zfs_msgid_table) / sizeof (zfs_msgid_table[0]))
     69 
     70 /* ARGSUSED */
     71 static int
     72 vdev_missing(uint64_t state, uint64_t aux, uint64_t errs)
     73 {
     74 	return (state == VDEV_STATE_CANT_OPEN &&
     75 	    aux == VDEV_AUX_OPEN_FAILED);
     76 }
     77 
     78 /* ARGSUSED */
     79 static int
     80 vdev_faulted(uint64_t state, uint64_t aux, uint64_t errs)
     81 {
     82 	return (state == VDEV_STATE_FAULTED);
     83 }
     84 
     85 /* ARGSUSED */
     86 static int
     87 vdev_errors(uint64_t state, uint64_t aux, uint64_t errs)
     88 {
     89 	return (state == VDEV_STATE_DEGRADED || errs != 0);
     90 }
     91 
     92 /* ARGSUSED */
     93 static int
     94 vdev_broken(uint64_t state, uint64_t aux, uint64_t errs)
     95 {
     96 	return (state == VDEV_STATE_CANT_OPEN);
     97 }
     98 
     99 /* ARGSUSED */
    100 static int
    101 vdev_offlined(uint64_t state, uint64_t aux, uint64_t errs)
    102 {
    103 	return (state == VDEV_STATE_OFFLINE);
    104 }
    105 
    106 /* ARGSUSED */
    107 static int
    108 vdev_removed(uint64_t state, uint64_t aux, uint64_t errs)
    109 {
    110 	return (state == VDEV_STATE_REMOVED);
    111 }
    112 
    113 /*
    114  * Detect if any leaf devices that have seen errors or could not be opened.
    115  */
    116 static boolean_t
    117 find_vdev_problem(nvlist_t *vdev, int (*func)(uint64_t, uint64_t, uint64_t))
    118 {
    119 	nvlist_t **child;
    120 	vdev_stat_t *vs;
    121 	uint_t c, children;
    122 	char *type;
    123 
    124 	/*
    125 	 * Ignore problems within a 'replacing' vdev, since we're presumably in
    126 	 * the process of repairing any such errors, and don't want to call them
    127 	 * out again.  We'll pick up the fact that a resilver is happening
    128 	 * later.
    129 	 */
    130 	verify(nvlist_lookup_string(vdev, ZPOOL_CONFIG_TYPE, &type) == 0);
    131 	if (strcmp(type, VDEV_TYPE_REPLACING) == 0)
    132 		return (B_FALSE);
    133 
    134 	if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, &child,
    135 	    &children) == 0) {
    136 		for (c = 0; c < children; c++)
    137 			if (find_vdev_problem(child[c], func))
    138 				return (B_TRUE);
    139 	} else {
    140 		verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
    141 		    (uint64_t **)&vs, &c) == 0);
    142 
    143 		if (func(vs->vs_state, vs->vs_aux,
    144 		    vs->vs_read_errors +
    145 		    vs->vs_write_errors +
    146 		    vs->vs_checksum_errors))
    147 			return (B_TRUE);
    148 	}
    149 
    150 	return (B_FALSE);
    151 }
    152 
    153 /*
    154  * Active pool health status.
    155  *
    156  * To determine the status for a pool, we make several passes over the config,
    157  * picking the most egregious error we find.  In order of importance, we do the
    158  * following:
    159  *
    160  *	- Check for a complete and valid configuration
    161  *	- Look for any faulted or missing devices in a non-replicated config
    162  *	- Check for any data errors
    163  *	- Check for any faulted or missing devices in a replicated config
    164  *	- Look for any devices showing errors
    165  *	- Check for any resilvering devices
    166  *
    167  * There can obviously be multiple errors within a single pool, so this routine
    168  * only picks the most damaging of all the current errors to report.
    169  */
    170 static zpool_status_t
    171 check_status(nvlist_t *config, boolean_t isimport)
    172 {
    173 	nvlist_t *nvroot;
    174 	vdev_stat_t *vs;
    175 	pool_scan_stat_t *ps = NULL;
    176 	uint_t vsc, psc;
    177 	uint64_t nerr;
    178 	uint64_t version;
    179 	uint64_t stateval;
    180 	uint64_t suspended;
    181 	uint64_t hostid = 0;
    182 
    183 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
    184 	    &version) == 0);
    185 	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
    186 	    &nvroot) == 0);
    187 	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
    188 	    (uint64_t **)&vs, &vsc) == 0);
    189 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
    190 	    &stateval) == 0);
    191 
    192 	/*
    193 	 * Currently resilvering a vdev
    194 	 */
    195 	(void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
    196 	    (uint64_t **)&ps, &psc);
    197 	if (ps && ps->pss_func == POOL_SCAN_RESILVER &&
    198 	    ps->pss_state == DSS_SCANNING)
    199 		return (ZPOOL_STATUS_RESILVERING);
    200 
    201 	/*
    202 	 * Pool last accessed by another system.
    203 	 */
    204 	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
    205 	if (hostid != 0 && (unsigned long)hostid != gethostid() &&
    206 	    stateval == POOL_STATE_ACTIVE)
    207 		return (ZPOOL_STATUS_HOSTID_MISMATCH);
    208 
    209 	/*
    210 	 * Newer on-disk version.
    211 	 */
    212 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    213 	    vs->vs_aux == VDEV_AUX_VERSION_NEWER)
    214 		return (ZPOOL_STATUS_VERSION_NEWER);
    215 
    216 	/*
    217 	 * Check that the config is complete.
    218 	 */
    219 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    220 	    vs->vs_aux == VDEV_AUX_BAD_GUID_SUM)
    221 		return (ZPOOL_STATUS_BAD_GUID_SUM);
    222 
    223 	/*
    224 	 * Check whether the pool has suspended due to failed I/O.
    225 	 */
    226 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED,
    227 	    &suspended) == 0) {
    228 		if (suspended == ZIO_FAILURE_MODE_CONTINUE)
    229 			return (ZPOOL_STATUS_IO_FAILURE_CONTINUE);
    230 		return (ZPOOL_STATUS_IO_FAILURE_WAIT);
    231 	}
    232 
    233 	/*
    234 	 * Could not read a log.
    235 	 */
    236 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    237 	    vs->vs_aux == VDEV_AUX_BAD_LOG) {
    238 		return (ZPOOL_STATUS_BAD_LOG);
    239 	}
    240 
    241 	/*
    242 	 * Bad devices in non-replicated config.
    243 	 */
    244 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    245 	    find_vdev_problem(nvroot, vdev_faulted))
    246 		return (ZPOOL_STATUS_FAULTED_DEV_NR);
    247 
    248 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    249 	    find_vdev_problem(nvroot, vdev_missing))
    250 		return (ZPOOL_STATUS_MISSING_DEV_NR);
    251 
    252 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    253 	    find_vdev_problem(nvroot, vdev_broken))
    254 		return (ZPOOL_STATUS_CORRUPT_LABEL_NR);
    255 
    256 	/*
    257 	 * Corrupted pool metadata
    258 	 */
    259 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
    260 	    vs->vs_aux == VDEV_AUX_CORRUPT_DATA)
    261 		return (ZPOOL_STATUS_CORRUPT_POOL);
    262 
    263 	/*
    264 	 * Persistent data errors.
    265 	 */
    266 	if (!isimport) {
    267 		if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
    268 		    &nerr) == 0 && nerr != 0)
    269 			return (ZPOOL_STATUS_CORRUPT_DATA);
    270 	}
    271 
    272 	/*
    273 	 * Missing devices in a replicated config.
    274 	 */
    275 	if (find_vdev_problem(nvroot, vdev_faulted))
    276 		return (ZPOOL_STATUS_FAULTED_DEV_R);
    277 	if (find_vdev_problem(nvroot, vdev_missing))
    278 		return (ZPOOL_STATUS_MISSING_DEV_R);
    279 	if (find_vdev_problem(nvroot, vdev_broken))
    280 		return (ZPOOL_STATUS_CORRUPT_LABEL_R);
    281 
    282 	/*
    283 	 * Devices with errors
    284 	 */
    285 	if (!isimport && find_vdev_problem(nvroot, vdev_errors))
    286 		return (ZPOOL_STATUS_FAILING_DEV);
    287 
    288 	/*
    289 	 * Offlined devices
    290 	 */
    291 	if (find_vdev_problem(nvroot, vdev_offlined))
    292 		return (ZPOOL_STATUS_OFFLINE_DEV);
    293 
    294 	/*
    295 	 * Removed device
    296 	 */
    297 	if (find_vdev_problem(nvroot, vdev_removed))
    298 		return (ZPOOL_STATUS_REMOVED_DEV);
    299 
    300 	/*
    301 	 * Outdated, but usable, version
    302 	 */
    303 	if (version < SPA_VERSION)
    304 		return (ZPOOL_STATUS_VERSION_OLDER);
    305 
    306 	return (ZPOOL_STATUS_OK);
    307 }
    308 
    309 zpool_status_t
    310 zpool_get_status(zpool_handle_t *zhp, char **msgid)
    311 {
    312 	zpool_status_t ret = check_status(zhp->zpool_config, B_FALSE);
    313 
    314 	if (ret >= NMSGID)
    315 		*msgid = NULL;
    316 	else
    317 		*msgid = zfs_msgid_table[ret];
    318 
    319 	return (ret);
    320 }
    321 
    322 zpool_status_t
    323 zpool_import_status(nvlist_t *config, char **msgid)
    324 {
    325 	zpool_status_t ret = check_status(config, B_TRUE);
    326 
    327 	if (ret >= NMSGID)
    328 		*msgid = NULL;
    329 	else
    330 		*msgid = zfs_msgid_table[ret];
    331 
    332 	return (ret);
    333 }
    334 
    335 static void
    336 dump_ddt_stat(const ddt_stat_t *dds, int h)
    337 {
    338 	char refcnt[6];
    339 	char blocks[6], lsize[6], psize[6], dsize[6];
    340 	char ref_blocks[6], ref_lsize[6], ref_psize[6], ref_dsize[6];
    341 
    342 	if (dds == NULL || dds->dds_blocks == 0)
    343 		return;
    344 
    345 	if (h == -1)
    346 		(void) strcpy(refcnt, "Total");
    347 	else
    348 		zfs_nicenum(1ULL << h, refcnt, sizeof (refcnt));
    349 
    350 	zfs_nicenum(dds->dds_blocks, blocks, sizeof (blocks));
    351 	zfs_nicenum(dds->dds_lsize, lsize, sizeof (lsize));
    352 	zfs_nicenum(dds->dds_psize, psize, sizeof (psize));
    353 	zfs_nicenum(dds->dds_dsize, dsize, sizeof (dsize));
    354 	zfs_nicenum(dds->dds_ref_blocks, ref_blocks, sizeof (ref_blocks));
    355 	zfs_nicenum(dds->dds_ref_lsize, ref_lsize, sizeof (ref_lsize));
    356 	zfs_nicenum(dds->dds_ref_psize, ref_psize, sizeof (ref_psize));
    357 	zfs_nicenum(dds->dds_ref_dsize, ref_dsize, sizeof (ref_dsize));
    358 
    359 	(void) printf("%6s   %6s   %5s   %5s   %5s   %6s   %5s   %5s   %5s\n",
    360 	    refcnt,
    361 	    blocks, lsize, psize, dsize,
    362 	    ref_blocks, ref_lsize, ref_psize, ref_dsize);
    363 }
    364 
    365 /*
    366  * Print the DDT histogram and the column totals.
    367  */
    368 void
    369 zpool_dump_ddt(const ddt_stat_t *dds_total, const ddt_histogram_t *ddh)
    370 {
    371 	int h;
    372 
    373 	(void) printf("\n");
    374 
    375 	(void) printf("bucket   "
    376 	    "           allocated             "
    377 	    "          referenced          \n");
    378 	(void) printf("______   "
    379 	    "______________________________   "
    380 	    "______________________________\n");
    381 
    382 	(void) printf("%6s   %6s   %5s   %5s   %5s   %6s   %5s   %5s   %5s\n",
    383 	    "refcnt",
    384 	    "blocks", "LSIZE", "PSIZE", "DSIZE",
    385 	    "blocks", "LSIZE", "PSIZE", "DSIZE");
    386 
    387 	(void) printf("%6s   %6s   %5s   %5s   %5s   %6s   %5s   %5s   %5s\n",
    388 	    "------",
    389 	    "------", "-----", "-----", "-----",
    390 	    "------", "-----", "-----", "-----");
    391 
    392 	for (h = 0; h < 64; h++)
    393 		dump_ddt_stat(&ddh->ddh_stat[h], h);
    394 
    395 	dump_ddt_stat(dds_total, -1);
    396 
    397 	(void) printf("\n");
    398 }
    399