*
* CDDL HEADER END
*/
+
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012 by Delphix. All rights reserved.
*/
#include <stdio.h>
#include <sys/zfs_fuid.h>
#include <sys/arc.h>
#include <sys/ddt.h>
+#include <sys/zfeature.h>
#undef ZFS_MAXNAMELEN
#include <libzfs.h>
-#define ZDB_COMPRESS_NAME(idx) ((idx) < ZIO_COMPRESS_FUNCTIONS ? \
- zio_compress_table[(idx)].ci_name : "UNKNOWN")
-#define ZDB_CHECKSUM_NAME(idx) ((idx) < ZIO_CHECKSUM_FUNCTIONS ? \
- zio_checksum_table[(idx)].ci_name : "UNKNOWN")
-#define ZDB_OT_NAME(idx) ((idx) < DMU_OT_NUMTYPES ? \
- dmu_ot[(idx)].ot_name : "UNKNOWN")
-#define ZDB_OT_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) : DMU_OT_NUMTYPES)
+#define ZDB_COMPRESS_NAME(idx) ((idx) < ZIO_COMPRESS_FUNCTIONS ? \
+ zio_compress_table[(idx)].ci_name : "UNKNOWN")
+#define ZDB_CHECKSUM_NAME(idx) ((idx) < ZIO_CHECKSUM_FUNCTIONS ? \
+ zio_checksum_table[(idx)].ci_name : "UNKNOWN")
+#define ZDB_OT_NAME(idx) ((idx) < DMU_OT_NUMTYPES ? \
+ dmu_ot[(idx)].ot_name : DMU_OT_IS_VALID(idx) ? \
+ dmu_ot_byteswap[DMU_OT_BYTESWAP(idx)].ob_name : "UNKNOWN")
+#define ZDB_OT_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) : \
+ (((idx) == DMU_OTN_ZAP_DATA || (idx) == DMU_OTN_ZAP_METADATA) ? \
+ DMU_OT_ZAP_OTHER : DMU_OT_NUMTYPES))
#ifndef lint
extern int zfs_recover;
uint64_t *zopt_object = NULL;
int zopt_objects = 0;
libzfs_handle_t *g_zfs;
+uint64_t max_inflight = 200;
/*
* These libumem hooks provide a reasonable set of defaults for the allocator's
usage(void)
{
(void) fprintf(stderr,
- "Usage: %s [-CumdibcsDvhL] poolname [object...]\n"
- " %s [-div] dataset [object...]\n"
- " %s -m [-L] poolname [vdev [metaslab...]]\n"
- " %s -R poolname vdev:offset:size[:flags]\n"
- " %s -S poolname\n"
- " %s -l [-u] device\n"
- " %s -C\n\n",
+ "Usage: %s [-CumdibcsDvhLXFPA] [-t txg] [-e [-p path...]] "
+ "[-U config] [-M inflight I/Os] poolname [object...]\n"
+ " %s [-divPA] [-e -p path...] [-U config] dataset "
+ "[object...]\n"
+ " %s -m [-LXFPA] [-t txg] [-e [-p path...]] [-U config] "
+ "poolname [vdev [metaslab...]]\n"
+ " %s -R [-A] [-e [-p path...]] poolname "
+ "vdev:offset:size[:flags]\n"
+ " %s -S [-PA] [-e [-p path...]] [-U config] poolname\n"
+ " %s -l [-uA] device\n"
+ " %s -C [-A] [-U config]\n\n",
cmdname, cmdname, cmdname, cmdname, cmdname, cmdname, cmdname);
(void) fprintf(stderr, " Dataset name must include at least one "
"has altroot/not in a cachefile\n");
(void) fprintf(stderr, " -p <path> -- use one or more with "
"-e to specify path to vdev dir\n");
- (void) fprintf(stderr, " -P print numbers parsable\n");
+ (void) fprintf(stderr, " -P print numbers in parseable form\n");
(void) fprintf(stderr, " -t <txg> -- highest txg to use when "
"searching for uberblocks\n");
+ (void) fprintf(stderr, " -M <number of inflight I/Os> -- "
+ "specify the maximum number of checksumming I/Os [default is 200]\n");
(void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
"to make only that option verbose\n");
(void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
dump_metaslab_stats(metaslab_t *msp)
{
char maxbuf[32];
- space_map_t *sm = &msp->ms_map;
+ space_map_t *sm = msp->ms_map;
avl_tree_t *t = sm->sm_pp_root;
int free_pct = sm->sm_space * 100 / sm->sm_size;
{
vdev_t *vd = msp->ms_group->mg_vd;
spa_t *spa = vd->vdev_spa;
- space_map_t *sm = &msp->ms_map;
+ space_map_t *sm = msp->ms_map;
space_map_obj_t *smo = &msp->ms_smo;
char freebuf[32];
return;
ASSERT(error == 0);
- if ((count = ddt_object_count(ddt, type, class)) == 0)
+ error = ddt_object_count(ddt, type, class, &count);
+ ASSERT(error == 0);
+ if (count == 0)
return;
dspace = doi.doi_physical_blocks_512 << 9;
arc_buf_t *buf;
uint64_t fill = 0;
- err = arc_read_nolock(NULL, spa, bp, arc_getbuf_func, &buf,
+ err = arc_read(NULL, spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
if (err)
return (err);
ASSERT(size == sizeof (*ds));
crtime = ds->ds_creation_time;
- zdb_nicenum(ds->ds_used_bytes, used);
+ zdb_nicenum(ds->ds_referenced_bytes, used);
zdb_nicenum(ds->ds_compressed_bytes, compressed);
zdb_nicenum(ds->ds_uncompressed_bytes, uncompressed);
zdb_nicenum(ds->ds_unique_bytes, unique);
/* ARGSUSED */
static int
+dump_bptree_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
+{
+ char blkbuf[BP_SPRINTF_LEN];
+
+ if (bp->blk_birth != 0) {
+ sprintf_blkptr(blkbuf, bp);
+ (void) printf("\t%s\n", blkbuf);
+ }
+ return (0);
+}
+
+static void
+dump_bptree(objset_t *os, uint64_t obj, char *name)
+{
+ char bytes[32];
+ bptree_phys_t *bt;
+ dmu_buf_t *db;
+
+ if (dump_opt['d'] < 3)
+ return;
+
+ VERIFY3U(0, ==, dmu_bonus_hold(os, obj, FTAG, &db));
+ bt = db->db_data;
+ zdb_nicenum(bt->bt_bytes, bytes);
+ (void) printf("\n %s: %llu datasets, %s\n",
+ name, (unsigned long long)(bt->bt_end - bt->bt_begin), bytes);
+ dmu_buf_rele(db, FTAG);
+
+ if (dump_opt['d'] < 5)
+ return;
+
+ (void) printf("\n");
+
+ (void) bptree_iterate(os, obj, B_FALSE, dump_bptree_cb, NULL, NULL);
+}
+
+/* ARGSUSED */
+static int
dump_bpobj_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
char blkbuf[BP_SPRINTF_LEN];
exit(1);
}
- if (fstat64(fd, &statbuf) != 0) {
+ if (fstat64_blk(fd, &statbuf) != 0) {
(void) printf("failed to stat '%s': %s\n", path,
strerror(errno));
free(path);
exit(1);
}
- if (S_ISBLK(statbuf.st_mode)) {
- (void) printf("cannot use '%s': character device required\n",
- path);
- free(path);
- (void) close(fd);
- exit(1);
- }
-
psize = statbuf.st_size;
psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
*/
#define ZDB_OT_DEFERRED (DMU_OT_NUMTYPES + 0)
#define ZDB_OT_DITTO (DMU_OT_NUMTYPES + 1)
-#define ZDB_OT_TOTAL (DMU_OT_NUMTYPES + 2)
+#define ZDB_OT_OTHER (DMU_OT_NUMTYPES + 2)
+#define ZDB_OT_TOTAL (DMU_OT_NUMTYPES + 3)
static char *zdb_ot_extname[] = {
"deferred free",
"dedup ditto",
+ "other",
"Total",
};
bp, NULL, NULL, ZIO_FLAG_CANFAIL)), ==, 0);
}
-/* ARGSUSED */
+static void
+zdb_blkptr_done(zio_t *zio)
+{
+ spa_t *spa = zio->io_spa;
+ blkptr_t *bp = zio->io_bp;
+ int ioerr = zio->io_error;
+ zdb_cb_t *zcb = zio->io_private;
+ zbookmark_t *zb = &zio->io_bookmark;
+
+ zio_data_buf_free(zio->io_data, zio->io_size);
+
+ mutex_enter(&spa->spa_scrub_lock);
+ spa->spa_scrub_inflight--;
+ cv_broadcast(&spa->spa_scrub_io_cv);
+
+ if (ioerr && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
+ char blkbuf[BP_SPRINTF_LEN];
+
+ zcb->zcb_haderrors = 1;
+ zcb->zcb_errors[ioerr]++;
+
+ if (dump_opt['b'] >= 2)
+ sprintf_blkptr(blkbuf, bp);
+ else
+ blkbuf[0] = '\0';
+
+ (void) printf("zdb_blkptr_cb: "
+ "Got error %d reading "
+ "<%llu, %llu, %lld, %llx> %s -- skipping\n",
+ ioerr,
+ (u_longlong_t)zb->zb_objset,
+ (u_longlong_t)zb->zb_object,
+ (u_longlong_t)zb->zb_level,
+ (u_longlong_t)zb->zb_blkid,
+ blkbuf);
+ }
+ mutex_exit(&spa->spa_scrub_lock);
+}
+
static int
-zdb_blkptr_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, arc_buf_t *pbuf,
+zdb_blkptr_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
{
zdb_cb_t *zcb = arg;
type = BP_GET_TYPE(bp);
- zdb_count_block(zcb, zilog, bp, type);
+ zdb_count_block(zcb, zilog, bp,
+ (type & DMU_OT_NEWTYPE) ? ZDB_OT_OTHER : type);
- is_metadata = (BP_GET_LEVEL(bp) != 0 || dmu_ot[type].ot_metadata);
+ is_metadata = (BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type));
if (dump_opt['c'] > 1 || (dump_opt['c'] && is_metadata)) {
- int ioerr;
size_t size = BP_GET_PSIZE(bp);
- void *data = malloc(size);
+ void *data = zio_data_buf_alloc(size);
int flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB | ZIO_FLAG_RAW;
/* If it's an intent log block, failure is expected. */
if (zb->zb_level == ZB_ZIL_LEVEL)
flags |= ZIO_FLAG_SPECULATIVE;
- ioerr = zio_wait(zio_read(NULL, spa, bp, data, size,
- NULL, NULL, ZIO_PRIORITY_ASYNC_READ, flags, zb));
+ mutex_enter(&spa->spa_scrub_lock);
+ while (spa->spa_scrub_inflight > max_inflight)
+ cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
+ spa->spa_scrub_inflight++;
+ mutex_exit(&spa->spa_scrub_lock);
- free(data);
+ zio_nowait(zio_read(NULL, spa, bp, data, size,
+ zdb_blkptr_done, zcb, ZIO_PRIORITY_ASYNC_READ, flags, zb));
- if (ioerr && !(flags & ZIO_FLAG_SPECULATIVE)) {
- zcb->zcb_haderrors = 1;
- zcb->zcb_errors[ioerr]++;
-
- if (dump_opt['b'] >= 2)
- sprintf_blkptr(blkbuf, bp);
- else
- blkbuf[0] = '\0';
-
- (void) printf("zdb_blkptr_cb: "
- "Got error %d reading "
- "<%llu, %llu, %lld, %llx> %s -- skipping\n",
- ioerr,
- (u_longlong_t)zb->zb_objset,
- (u_longlong_t)zb->zb_object,
- (u_longlong_t)zb->zb_level,
- (u_longlong_t)zb->zb_blkid,
- blkbuf);
- }
}
zcb->zcb_readfails = 0;
for (m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *msp = vd->vdev_ms[m];
mutex_enter(&msp->ms_lock);
- space_map_unload(&msp->ms_map);
- VERIFY(space_map_load(&msp->ms_map,
+ space_map_unload(msp->ms_map);
+ VERIFY(space_map_load(msp->ms_map,
&zdb_space_map_ops, SM_ALLOC, &msp->ms_smo,
spa->spa_meta_objset) == 0);
- msp->ms_map.sm_ppd = vd;
+ msp->ms_map->sm_ppd = vd;
mutex_exit(&msp->ms_lock);
}
}
for (m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *msp = vd->vdev_ms[m];
mutex_enter(&msp->ms_lock);
- space_map_unload(&msp->ms_map);
+ space_map_unload(msp->ms_map);
mutex_exit(&msp->ms_lock);
}
}
count_block_cb, &zcb, NULL);
(void) bpobj_iterate_nofree(&spa->spa_dsl_pool->dp_free_bpobj,
count_block_cb, &zcb, NULL);
+ if (spa_feature_is_active(spa,
+ &spa_feature_table[SPA_FEATURE_ASYNC_DESTROY])) {
+ VERIFY3U(0, ==, bptree_iterate(spa->spa_meta_objset,
+ spa->spa_dsl_pool->dp_bptree_obj, B_FALSE, count_block_cb,
+ &zcb, NULL));
+ }
if (dump_opt['c'] > 1)
flags |= TRAVERSE_PREFETCH_DATA;
zcb.zcb_haderrors |= traverse_pool(spa, 0, flags, zdb_blkptr_cb, &zcb);
+ /*
+ * If we've traversed the data blocks then we need to wait for those
+ * I/Os to complete. We leverage "The Godfather" zio to wait on
+ * all async I/Os to complete.
+ */
+ if (dump_opt['c']) {
+ (void) zio_wait(spa->spa_async_zio_root);
+ spa->spa_async_zio_root = zio_root(spa, NULL, NULL,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
+ ZIO_FLAG_GODFATHER);
+ }
+
if (zcb.zcb_haderrors) {
(void) printf("\nError counts:\n\n");
(void) printf("\t%5s %s\n", "errno", "count");
/* ARGSUSED */
static int
zdb_ddt_add_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
- arc_buf_t *pbuf, const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
+ const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
{
avl_tree_t *t = arg;
avl_index_t where;
}
if (BP_IS_HOLE(bp) || BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_OFF ||
- BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata)
+ BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
return (0);
ddt_key_fill(&zdde_search.zdde_key, bp);
dump_bpobj(&spa->spa_deferred_bpobj, "Deferred frees");
if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
dump_bpobj(&spa->spa_dsl_pool->dp_free_bpobj,
- "Pool frees");
+ "Pool snapshot frees");
+ }
+
+ if (spa_feature_is_active(spa,
+ &spa_feature_table[SPA_FEATURE_ASYNC_DESTROY])) {
+ dump_bptree(spa->spa_meta_objset,
+ spa->spa_dsl_pool->dp_bptree_obj,
+ "Pool dataset frees");
}
dump_dtl(spa->spa_root_vdev, 0);
}
psize = size;
lsize = size;
- pbuf = umem_alloc(SPA_MAXBLOCKSIZE, UMEM_NOFAIL);
+ pbuf = umem_alloc_aligned(SPA_MAXBLOCKSIZE, 512, UMEM_NOFAIL);
lbuf = umem_alloc(SPA_MAXBLOCKSIZE, UMEM_NOFAIL);
BP_ZERO(bp);
nvlist_t *policy = NULL;
uint64_t max_txg = UINT64_MAX;
int rewind = ZPOOL_NEVER_REWIND;
+ char *spa_config_path_env;
(void) setrlimit(RLIMIT_NOFILE, &rl);
(void) enable_extended_FILE_stdio(-1, -1);
dprintf_setup(&argc, argv);
- while ((c = getopt(argc, argv, "bcdhilmsuCDRSAFLXevp:t:U:P")) != -1) {
+ /*
+ * If there is an environment variable SPA_CONFIG_PATH it overrides
+ * default spa_config_path setting. If -U flag is specified it will
+ * override this environment variable settings once again.
+ */
+ spa_config_path_env = getenv("SPA_CONFIG_PATH");
+ if (spa_config_path_env != NULL)
+ spa_config_path = spa_config_path_env;
+
+ while ((c = getopt(argc, argv, "bcdhilmM:suCDRSAFLXevp:t:U:P")) != -1) {
switch (c) {
case 'b':
case 'c':
case 'v':
verbose++;
break;
+ case 'M':
+ max_inflight = strtoull(optarg, NULL, 0);
+ if (max_inflight == 0) {
+ (void) fprintf(stderr, "maximum number "
+ "of inflight I/Os must be greater "
+ "than 0\n");
+ usage();
+ }
+ break;
case 'p':
if (searchdirs == NULL) {
searchdirs = umem_alloc(sizeof (char *),
}
kernel_init(FREAD);
- g_zfs = libzfs_init();
- ASSERT(g_zfs != NULL);
+ if ((g_zfs = libzfs_init()) == NULL)
+ return (1);
if (dump_all)
verbose = MAX(verbose, 1);
argv[i], strerror(errno));
}
}
- (os != NULL) ? dump_dir(os) : dump_zpool(spa);
+ if (os != NULL) {
+ dump_dir(os);
+ } else if (zopt_objects > 0 && !dump_opt['m']) {
+ dump_dir(spa->spa_meta_objset);
+ } else {
+ dump_zpool(spa);
+ }
} else {
flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR;
flagbits['c'] = ZDB_FLAG_CHECKSUM;