* under sponsorship from the FreeBSD Foundation.
* Copyright (c) 2021 Allan Jude
* Copyright (c) 2021 Toomas Soome <tsoome@me.com>
+ * Copyright (c) 2023, Klara Inc.
+ * Copyright (c) 2023, Rob Norris <robn@despairlabs.com>
*/
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>
#include <getopt.h>
+#include <openssl/evp.h>
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/dsl_crypt.h>
#include <sys/dsl_scan.h>
#include <sys/btree.h>
+#include <sys/brt.h>
+#include <sys/brt_impl.h>
#include <zfs_comutil.h>
#include <sys/zstd/zstd.h>
#include <libnvpair.h>
#include <libzutil.h>
+#include <libzdb.h>
+
#include "zdb.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_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) : \
- (idx) == DMU_OTN_ZAP_DATA || (idx) == DMU_OTN_ZAP_METADATA ? \
- DMU_OT_ZAP_OTHER : \
- (idx) == DMU_OTN_UINT64_DATA || (idx) == DMU_OTN_UINT64_METADATA ? \
- DMU_OT_UINT64_OTHER : DMU_OT_NUMTYPES)
-
-/* Some platforms require part of inode IDs to be remapped */
-#ifdef __APPLE__
-#define ZDB_MAP_OBJECT_ID(obj) INO_XNUTOZFS(obj, 2)
-#else
-#define ZDB_MAP_OBJECT_ID(obj) (obj)
-#endif
-
-static const char *
-zdb_ot_name(dmu_object_type_t type)
-{
- if (type < DMU_OT_NUMTYPES)
- return (dmu_ot[type].ot_name);
- else if ((type & DMU_OT_NEWTYPE) &&
- ((type & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS))
- return (dmu_ot_byteswap[type & DMU_OT_BYTESWAP_MASK].ob_name);
- else
- return ("UNKNOWN");
-}
extern int reference_tracking_enable;
extern int zfs_recover;
-extern unsigned long zfs_arc_meta_min, zfs_arc_meta_limit;
-extern int zfs_vdev_async_read_max_active;
+extern uint_t zfs_vdev_async_read_max_active;
extern boolean_t spa_load_verify_dryrun;
extern boolean_t spa_mode_readable_spacemaps;
-extern int zfs_reconstruct_indirect_combinations_max;
+extern uint_t zfs_reconstruct_indirect_combinations_max;
extern uint_t zfs_btree_verify_intensity;
static const char cmdname[] = "zdb";
typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size);
-uint64_t *zopt_metaslab = NULL;
+static uint64_t *zopt_metaslab = NULL;
static unsigned zopt_metaslab_args = 0;
-typedef struct zopt_object_range {
- uint64_t zor_obj_start;
- uint64_t zor_obj_end;
- uint64_t zor_flags;
-} zopt_object_range_t;
-zopt_object_range_t *zopt_object_ranges = NULL;
+
+static zopt_object_range_t *zopt_object_ranges = NULL;
static unsigned zopt_object_args = 0;
static int flagbits[256];
-#define ZOR_FLAG_PLAIN_FILE 0x0001
-#define ZOR_FLAG_DIRECTORY 0x0002
-#define ZOR_FLAG_SPACE_MAP 0x0004
-#define ZOR_FLAG_ZAP 0x0008
-#define ZOR_FLAG_ALL_TYPES -1
-#define ZOR_SUPPORTED_FLAGS (ZOR_FLAG_PLAIN_FILE | \
- ZOR_FLAG_DIRECTORY | \
- ZOR_FLAG_SPACE_MAP | \
- ZOR_FLAG_ZAP)
-
-#define ZDB_FLAG_CHECKSUM 0x0001
-#define ZDB_FLAG_DECOMPRESS 0x0002
-#define ZDB_FLAG_BSWAP 0x0004
-#define ZDB_FLAG_GBH 0x0008
-#define ZDB_FLAG_INDIRECT 0x0010
-#define ZDB_FLAG_RAW 0x0020
-#define ZDB_FLAG_PRINT_BLKPTR 0x0040
-#define ZDB_FLAG_VERBOSE 0x0080
-uint64_t max_inflight_bytes = 256 * 1024 * 1024; /* 256MB */
+static uint64_t max_inflight_bytes = 256 * 1024 * 1024; /* 256MB */
static int leaked_objects = 0;
static range_tree_t *mos_refd_objs;
static int dump_bpobj_cb(void *arg, const blkptr_t *bp, boolean_t free,
dmu_tx_t *tx);
-typedef struct sublivelist_verify {
- /* FREE's that haven't yet matched to an ALLOC, in one sub-livelist */
- zfs_btree_t sv_pair;
-
- /* ALLOC's without a matching FREE, accumulates across sub-livelists */
- zfs_btree_t sv_leftover;
-} sublivelist_verify_t;
-
-static int
-livelist_compare(const void *larg, const void *rarg)
-{
- const blkptr_t *l = larg;
- const blkptr_t *r = rarg;
-
- /* Sort them according to dva[0] */
- uint64_t l_dva0_vdev, r_dva0_vdev;
- l_dva0_vdev = DVA_GET_VDEV(&l->blk_dva[0]);
- r_dva0_vdev = DVA_GET_VDEV(&r->blk_dva[0]);
- if (l_dva0_vdev < r_dva0_vdev)
- return (-1);
- else if (l_dva0_vdev > r_dva0_vdev)
- return (+1);
-
- /* if vdevs are equal, sort by offsets. */
- uint64_t l_dva0_offset;
- uint64_t r_dva0_offset;
- l_dva0_offset = DVA_GET_OFFSET(&l->blk_dva[0]);
- r_dva0_offset = DVA_GET_OFFSET(&r->blk_dva[0]);
- if (l_dva0_offset < r_dva0_offset) {
- return (-1);
- } else if (l_dva0_offset > r_dva0_offset) {
- return (+1);
- }
-
- /*
- * Since we're storing blkptrs without cancelling FREE/ALLOC pairs,
- * it's possible the offsets are equal. In that case, sort by txg
- */
- if (l->blk_birth < r->blk_birth) {
- return (-1);
- } else if (l->blk_birth > r->blk_birth) {
- return (+1);
- }
- return (0);
-}
-
-typedef struct sublivelist_verify_block {
- dva_t svb_dva;
- /*
- * We need this to check if the block marked as allocated
- * in the livelist was freed (and potentially reallocated)
- * in the metaslab spacemaps at a later TXG.
- */
- uint64_t svb_allocated_txg;
-} sublivelist_verify_block_t;
static void zdb_print_blkptr(const blkptr_t *bp, int flags);
int err;
struct sublivelist_verify *sv = args;
- zfs_btree_create(&sv->sv_pair, sublivelist_block_refcnt_compare,
+ zfs_btree_create(&sv->sv_pair, sublivelist_block_refcnt_compare, NULL,
sizeof (sublivelist_verify_block_refcnt_t));
err = bpobj_iterate_nofree(&dle->dle_bpobj, sublivelist_verify_blkptr,
{
(void) args;
sublivelist_verify_t sv;
- zfs_btree_create(&sv.sv_leftover, livelist_block_compare,
+ zfs_btree_create(&sv.sv_leftover, livelist_block_compare, NULL,
sizeof (sublivelist_verify_block_t));
int err = sublivelist_verify_func(&sv, dle);
zfs_btree_clear(&sv.sv_leftover);
verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg,
uint64_t offset, uint64_t size)
{
- sublivelist_verify_block_t svb;
+ sublivelist_verify_block_t svb = {{{0}}};
DVA_SET_VDEV(&svb.svb_dva, mv->mv_vdid);
DVA_SET_OFFSET(&svb.svb_dva, offset);
DVA_SET_ASIZE(&svb.svb_dva, size);
(void) printf("Verifying deleted livelist entries\n");
sublivelist_verify_t sv;
- zfs_btree_create(&sv.sv_leftover, livelist_block_compare,
+ zfs_btree_create(&sv.sv_leftover, livelist_block_compare, NULL,
sizeof (sublivelist_verify_block_t));
iterate_deleted_livelists(spa, livelist_verify, &sv);
mv.mv_start = m->ms_start;
mv.mv_end = m->ms_start + m->ms_size;
zfs_btree_create(&mv.mv_livelist_allocs,
- livelist_block_compare,
+ livelist_block_compare, NULL,
sizeof (sublivelist_verify_block_t));
mv_populate_livelist_allocs(&mv, &sv);
"Usage:\t%s [-AbcdDFGhikLMPsvXy] [-e [-V] [-p <path> ...]] "
"[-I <inflight I/Os>]\n"
"\t\t[-o <var>=<value>]... [-t <txg>] [-U <cache>] [-x <dumpdir>]\n"
+ "\t\t[-K <key>]\n"
"\t\t[<poolname>[/<dataset | objset id>] [<object | range> ...]]\n"
- "\t%s [-AdiPv] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
+ "\t%s [-AdiPv] [-e [-V] [-p <path> ...]] [-U <cache>] [-K <key>]\n"
"\t\t[<poolname>[/<dataset | objset id>] [<object | range> ...]\n"
+ "\t%s -B [-e [-V] [-p <path> ...]] [-I <inflight I/Os>]\n"
+ "\t\t[-o <var>=<value>]... [-t <txg>] [-U <cache>] [-x <dumpdir>]\n"
+ "\t\t[-K <key>] <poolname>/<objset id> [<backupflags>]\n"
"\t%s [-v] <bookmark>\n"
- "\t%s -C [-A] [-U <cache>]\n"
+ "\t%s -C [-A] [-U <cache>] [<poolname>]\n"
"\t%s -l [-Aqu] <device>\n"
"\t%s -m [-AFLPX] [-e [-V] [-p <path> ...]] [-t <txg>] "
"[-U <cache>]\n\t\t<poolname> [<vdev> [<metaslab> ...]]\n"
- "\t%s -O <dataset> <path>\n"
- "\t%s -r <dataset> <path> <destination>\n"
+ "\t%s -O
[-K <key>] <dataset> <path>\n"
+ "\t%s -r
[-K <key>] <dataset> <path> <destination>\n"
"\t%s -R [-A] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
"\t\t<poolname> <vdev>:<offset>:<size>[:<flags>]\n"
"\t%s -E [-A] word0:word1:...:word15\n"
"\t%s -S [-AP] [-e [-V] [-p <path> ...]] [-U <cache>] "
"<poolname>\n\n",
cmdname, cmdname, cmdname, cmdname, cmdname, cmdname, cmdname,
- cmdname, cmdname, cmdname, cmdname);
+ cmdname, cmdname, cmdname, cmdname, cmdname);
(void) fprintf(stderr, " Dataset name must include at least one "
"separator character '/' or '@'\n");
(void) fprintf(stderr, " Options to control amount of output:\n");
(void) fprintf(stderr, " -b --block-stats "
"block statistics\n");
+ (void) fprintf(stderr, " -B --backup "
+ "backup stream\n");
(void) fprintf(stderr, " -c --checksum "
"checksum all metadata (twice for all data) blocks\n");
(void) fprintf(stderr, " -C --config "
(void) fprintf(stderr, " -I --inflight=INTEGER "
"specify the maximum number of checksumming I/Os "
"[default is 200]\n");
+ (void) fprintf(stderr, " -K --key=KEY "
+ "decryption key for encrypted dataset\n");
(void) fprintf(stderr, " -o --option=\"OPTION=INTEGER\" "
"set global variable to an unsigned 32-bit integer\n");
(void) fprintf(stderr, " -p --path==PATH "
"don't print label contents\n");
(void) fprintf(stderr, " -t --txg=INTEGER "
"highest txg to use when searching for uberblocks\n");
+ (void) fprintf(stderr, " -T --brt-stats "
+ "BRT statistics\n");
(void) fprintf(stderr, " -u --uberblock "
"uberblock\n");
(void) fprintf(stderr, " -U --cachefile=PATH "
if (dump_opt['P'])
(void) snprintf(buf, buflen, "%llu", (longlong_t)num);
else
- nicenum(num, buf, sizeof (buf));
+ nicenum(num, buf, buflen);
+}
+
+static void
+zdb_nicebytes(uint64_t bytes, char *buf, size_t buflen)
+{
+ if (dump_opt['P'])
+ (void) snprintf(buf, buflen, "%llu", (longlong_t)bytes);
+ else
+ zfs_nicebytes(bytes, buf, buflen);
}
static const char histo_stars[] = "****************************************";
uint64_t max = 0;
for (i = 0; i < size; i++) {
+ if (histo[i] == 0)
+ continue;
if (histo[i] > max)
max = histo[i];
- if (histo[i] > 0 && i > maxidx)
+ if (i > maxidx)
maxidx = i;
- if (histo[i] > 0 && i < minidx)
+ if (i < minidx)
minidx = i;
}
}
if (size == 0) {
+ if (data == NULL)
+ kmem_free(arr, oursize);
(void) printf("\t\t[]\n");
return;
}
dump_dedup_ratio(&dds_total);
}
+static void
+dump_brt(spa_t *spa)
+{
+ if (!spa_feature_is_enabled(spa, SPA_FEATURE_BLOCK_CLONING)) {
+ printf("BRT: unsupported on this pool\n");
+ return;
+ }
+
+ if (!spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING)) {
+ printf("BRT: empty\n");
+ return;
+ }
+
+ brt_t *brt = spa->spa_brt;
+ VERIFY(brt);
+
+ char count[32], used[32], saved[32];
+ zdb_nicebytes(brt_get_used(spa), used, sizeof (used));
+ zdb_nicebytes(brt_get_saved(spa), saved, sizeof (saved));
+ uint64_t ratio = brt_get_ratio(spa);
+ printf("BRT: used %s; saved %s; ratio %llu.%02llux\n", used, saved,
+ (u_longlong_t)(ratio / 100), (u_longlong_t)(ratio % 100));
+
+ if (dump_opt['T'] < 2)
+ return;
+
+ for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
+ brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
+ if (brtvd == NULL)
+ continue;
+
+ if (!brtvd->bv_initiated) {
+ printf("BRT: vdev %" PRIu64 ": empty\n", vdevid);
+ continue;
+ }
+
+ zdb_nicenum(brtvd->bv_totalcount, count, sizeof (count));
+ zdb_nicebytes(brtvd->bv_usedspace, used, sizeof (used));
+ zdb_nicebytes(brtvd->bv_savedspace, saved, sizeof (saved));
+ printf("BRT: vdev %" PRIu64 ": refcnt %s; used %s; saved %s\n",
+ vdevid, count, used, saved);
+ }
+
+ if (dump_opt['T'] < 3)
+ return;
+
+ char dva[64];
+ printf("\n%-16s %-10s\n", "DVA", "REFCNT");
+
+ for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
+ brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
+ if (brtvd == NULL || !brtvd->bv_initiated)
+ continue;
+
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ for (zap_cursor_init(&zc, brt->brt_mos, brtvd->bv_mos_entries);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ uint64_t offset = *(uint64_t *)za.za_name;
+ uint64_t refcnt = za.za_first_integer;
+
+ snprintf(dva, sizeof (dva), "%" PRIu64 ":%llx", vdevid,
+ (u_longlong_t)offset);
+ printf("%-16s %-10llu\n", dva, (u_longlong_t)refcnt);
+ }
+ zap_cursor_fini(&zc);
+ }
+}
+
static void
dump_dtl_seg(void *arg, uint64_t start, uint64_t size)
{
snprintf_zstd_header(spa_t *spa, char *blkbuf, size_t buflen,
const blkptr_t *bp)
{
- abd_t *pabd;
+ static abd_t *pabd = NULL;
void *buf;
zio_t *zio;
zfs_zstdhdr_t zstd_hdr;
return;
}
- pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
+ if (!pabd)
+ pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
zio = zio_root(spa, NULL, NULL, 0);
/* Decrypt but don't decompress so we can read the compression header */
(void) snprintf(blkbuf + strlen(blkbuf),
buflen - strlen(blkbuf), " %s", "FREE");
(void) snprintf(blkbuf + strlen(blkbuf),
- buflen - strlen(blkbuf), " cksum=%llx:%llx:%llx:%llx",
+ buflen - strlen(blkbuf),
+ " cksum=%016llx:%016llx:%016llx:%016llx",
(u_longlong_t)bp->blk_cksum.zc_word[0],
(u_longlong_t)bp->blk_cksum.zc_word[1],
(u_longlong_t)bp->blk_cksum.zc_word[2],
zap_cursor_advance(&zc)) {
char osname[ZFS_MAX_DATASET_NAME_LEN];
char buf[ZFS_MAX_DATASET_NAME_LEN];
+ int len;
dmu_objset_name(os, osname);
- VERIFY3S(0, <=, snprintf(buf, sizeof (buf), "%s#%s", osname,
- attr.za_name));
+ len = snprintf(buf, sizeof (buf), "%s#%s", osname,
+ attr.za_name);
+ VERIFY3S(len, <, ZFS_MAX_DATASET_NAME_LEN);
(void) dump_bookmark(dp, buf, verbosity >= 5, verbosity >= 6);
}
zap_cursor_fini(&zc);
return (0);
}
+static char *key_material = NULL;
+
+static boolean_t
+zdb_derive_key(dsl_dir_t *dd, uint8_t *key_out)
+{
+ uint64_t keyformat, salt, iters;
+ int i;
+ unsigned char c;
+
+ VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset, dd->dd_crypto_obj,
+ zfs_prop_to_name(ZFS_PROP_KEYFORMAT), sizeof (uint64_t),
+ 1, &keyformat));
+
+ switch (keyformat) {
+ case ZFS_KEYFORMAT_HEX:
+ for (i = 0; i < WRAPPING_KEY_LEN * 2; i += 2) {
+ if (!isxdigit(key_material[i]) ||
+ !isxdigit(key_material[i+1]))
+ return (B_FALSE);
+ if (sscanf(&key_material[i], "%02hhx", &c) != 1)
+ return (B_FALSE);
+ key_out[i / 2] = c;
+ }
+ break;
+
+ case ZFS_KEYFORMAT_PASSPHRASE:
+ VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset,
+ dd->dd_crypto_obj, zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT),
+ sizeof (uint64_t), 1, &salt));
+ VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset,
+ dd->dd_crypto_obj, zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS),
+ sizeof (uint64_t), 1, &iters));
+
+ if (PKCS5_PBKDF2_HMAC_SHA1(key_material, strlen(key_material),
+ ((uint8_t *)&salt), sizeof (uint64_t), iters,
+ WRAPPING_KEY_LEN, key_out) != 1)
+ return (B_FALSE);
+
+ break;
+
+ default:
+ fatal("no support for key format %u\n",
+ (unsigned int) keyformat);
+ }
+
+ return (B_TRUE);
+}
+
+static char encroot[ZFS_MAX_DATASET_NAME_LEN];
+static boolean_t key_loaded = B_FALSE;
+
+static void
+zdb_load_key(objset_t *os)
+{
+ dsl_pool_t *dp;
+ dsl_dir_t *dd, *rdd;
+ uint8_t key[WRAPPING_KEY_LEN];
+ uint64_t rddobj;
+ int err;
+
+ dp = spa_get_dsl(os->os_spa);
+ dd = os->os_dsl_dataset->ds_dir;
+
+ dsl_pool_config_enter(dp, FTAG);
+ VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset, dd->dd_crypto_obj,
+ DSL_CRYPTO_KEY_ROOT_DDOBJ, sizeof (uint64_t), 1, &rddobj));
+ VERIFY0(dsl_dir_hold_obj(dd->dd_pool, rddobj, NULL, FTAG, &rdd));
+ dsl_dir_name(rdd, encroot);
+ dsl_dir_rele(rdd, FTAG);
+
+ if (!zdb_derive_key(dd, key))
+ fatal("couldn't derive encryption key");
+
+ dsl_pool_config_exit(dp, FTAG);
+
+ ASSERT3U(dsl_dataset_get_keystatus(dd), ==, ZFS_KEYSTATUS_UNAVAILABLE);
+
+ dsl_crypto_params_t *dcp;
+ nvlist_t *crypto_args;
+
+ crypto_args = fnvlist_alloc();
+ fnvlist_add_uint8_array(crypto_args, "wkeydata",
+ (uint8_t *)key, WRAPPING_KEY_LEN);
+ VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
+ NULL, crypto_args, &dcp));
+ err = spa_keystore_load_wkey(encroot, dcp, B_FALSE);
+
+ dsl_crypto_params_free(dcp, (err != 0));
+ fnvlist_free(crypto_args);
+
+ if (err != 0)
+ fatal(
+ "couldn't load encryption key for %s: %s",
+ encroot, err == ZFS_ERR_CRYPTO_NOTSUP ?
+ "crypto params not supported" : strerror(err));
+
+ ASSERT3U(dsl_dataset_get_keystatus(dd), ==, ZFS_KEYSTATUS_AVAILABLE);
+
+ printf("Unlocked encryption root: %s\n", encroot);
+ key_loaded = B_TRUE;
+}
+
+static void
+zdb_unload_key(void)
+{
+ if (!key_loaded)
+ return;
+
+ VERIFY0(spa_keystore_unload_wkey(encroot));
+ key_loaded = B_FALSE;
+}
+
static avl_tree_t idx_tree;
static avl_tree_t domain_tree;
static boolean_t fuid_table_loaded;
uint64_t version = 0;
VERIFY3P(sa_os, ==, NULL);
+
/*
* We can't own an objset if it's redacted. Therefore, we do this
* dance: hold the objset, then acquire a long hold on its dataset, then
* release the pool (which is held as part of holding the objset).
*/
- err = dmu_objset_hold(path, tag, osp);
+
+ if (dump_opt['K']) {
+ /* decryption requested, try to load keys */
+ err = dmu_objset_hold(path, tag, osp);
+ if (err != 0) {
+ (void) fprintf(stderr, "failed to hold dataset "
+ "'%s': %s\n",
+ path, strerror(err));
+ return (err);
+ }
+ dsl_dataset_long_hold(dmu_objset_ds(*osp), tag);
+ dsl_pool_rele(dmu_objset_pool(*osp), tag);
+
+ /* succeeds or dies */
+ zdb_load_key(*osp);
+
+ /* release it all */
+ dsl_dataset_long_rele(dmu_objset_ds(*osp), tag);
+ dsl_dataset_rele(dmu_objset_ds(*osp), tag);
+ }
+
+ int ds_hold_flags = key_loaded ? DS_HOLD_FLAG_DECRYPT : 0;
+
+ err = dmu_objset_hold_flags(path, ds_hold_flags, tag, osp);
if (err != 0) {
(void) fprintf(stderr, "failed to hold dataset '%s': %s\n",
path, strerror(err));
dsl_dataset_long_hold(dmu_objset_ds(*osp), tag);
dsl_pool_rele(dmu_objset_pool(*osp), tag);
- if (dmu_objset_type(*osp) == DMU_OST_ZFS && !(*osp)->os_encrypted) {
+ if (dmu_objset_type(*osp) == DMU_OST_ZFS &&
+ (key_loaded || !(*osp)->os_encrypted)) {
(void) zap_lookup(*osp, MASTER_NODE_OBJ, ZPL_VERSION_STR,
8, 1, &version);
if (version >= ZPL_VERSION_SA) {
(void) fprintf(stderr, "sa_setup failed: %s\n",
strerror(err));
dsl_dataset_long_rele(dmu_objset_ds(*osp), tag);
- dsl_dataset_rele(dmu_objset_ds(*osp), tag);
+ dsl_dataset_rele_flags(dmu_objset_ds(*osp),
+ ds_hold_flags, tag);
*osp = NULL;
}
}
sa_os = *osp;
- return (0);
+ return (err);
}
static void
if (os->os_sa != NULL)
sa_tear_down(os);
dsl_dataset_long_rele(dmu_objset_ds(os), tag);
- dsl_dataset_rele(dmu_objset_ds(os), tag);
+ dsl_dataset_rele_flags(dmu_objset_ds(os),
+ key_loaded ? DS_HOLD_FLAG_DECRYPT : 0, tag);
sa_attr_table = NULL;
sa_os = NULL;
+
+ zdb_unload_key();
}
static void
(void) printf("\tSA xattrs: %d bytes, %d entries\n\n",
sa_xattr_size, sa_xattr_entries);
while ((elem = nvlist_next_nvpair(sa_xattr, elem)) != NULL) {
+ boolean_t can_print = !dump_opt['P'];
uchar_t *value;
uint_t cnt, idx;
(void) printf("\t\t%s = ", nvpair_name(elem));
nvpair_value_byte_array(elem, &value, &cnt);
+
for (idx = 0; idx < cnt; ++idx) {
- if (isprint(value[idx]))
+ if (!isprint(value[idx])) {
+ can_print = B_FALSE;
+ break;
+ }
+ }
+
+ for (idx = 0; idx < cnt; ++idx) {
+ if (can_print)
(void) putchar(value[idx]);
else
(void) printf("\\%3.3o", value[idx]);
if (error)
fatal("dmu_object_info() failed, errno %u", error);
- if (os->os_encrypted &&
+ if (!key_loaded && os->os_encrypted &&
DMU_OT_IS_ENCRYPTED(doi.doi_bonus_type)) {
error = dnode_hold(os, object, FTAG, &dn);
if (error)
zdb_nicenum(doi.doi_physical_blocks_512 << 9, asize, sizeof (asize));
zdb_nicenum(doi.doi_bonus_size, bonus_size, sizeof (bonus_size));
zdb_nicenum(doi.doi_dnodesize, dnsize, sizeof (dnsize));
- (void) sprintf(fill, "%6.2f", 100.0 * doi.doi_fill_count *
- doi.doi_data_block_size / (object == 0 ? DNODES_PER_BLOCK : 1) /
- doi.doi_max_offset);
+ (void) snprintf(fill, sizeof (fill), "%6.2f", 100.0 *
+ doi.doi_fill_count * doi.doi_data_block_size / (object == 0 ?
+ DNODES_PER_BLOCK : 1) / doi.doi_max_offset);
aux[0] = '\0';
(void) printf("\t\t(bonus encrypted)\n");
}
- if (!os->os_encrypted || !DMU_OT_IS_ENCRYPTED(doi.doi_type)) {
+ if (key_loaded ||
+ (!os->os_encrypted || !DMU_OT_IS_ENCRYPTED(doi.doi_type))) {
object_viewer[ZDB_OT_TYPE(doi.doi_type)](os, object,
NULL, 0);
} else {
}
(void) printf("\tcheckpoint_txg = %llu\n",
(u_longlong_t)ub->ub_checkpoint_txg);
+
+ (void) printf("\traidz_reflow state=%u off=%llu\n",
+ (int)RRSS_GET_STATE(ub),
+ (u_longlong_t)RRSS_GET_OFFSET(ub));
+
(void) printf("%s", footer ? footer : "");
}
{
nvlist_t *list, **array;
nvpair_t *nvp = NULL;
- char *name;
+ const char *name;
uint_t i, items;
stats->zns_list_count++;
}
static void
-dump_l2arc_log_blkptr(l2arc_log_blkptr_t lbps)
+dump_l2arc_log_blkptr(const l2arc_log_blkptr_t *lbps)
{
- (void) printf("|\t\tdaddr: %llu\n", (u_longlong_t)lbps.lbp_daddr);
+ (void) printf("|\t\tdaddr: %llu\n", (u_longlong_t)lbps->lbp_daddr);
(void) printf("|\t\tpayload_asize: %llu\n",
- (u_longlong_t)lbps.lbp_payload_asize);
+ (u_longlong_t)lbps->lbp_payload_asize);
(void) printf("|\t\tpayload_start: %llu\n",
- (u_longlong_t)lbps.lbp_payload_start);
+ (u_longlong_t)lbps->lbp_payload_start);
(void) printf("|\t\tlsize: %llu\n",
- (u_longlong_t)L2BLK_GET_LSIZE((&lbps)->lbp_prop));
+ (u_longlong_t)L2BLK_GET_LSIZE(lbps->lbp_prop));
(void) printf("|\t\tasize: %llu\n",
- (u_longlong_t)L2BLK_GET_PSIZE((&lbps)->lbp_prop));
+ (u_longlong_t)L2BLK_GET_PSIZE(lbps->lbp_prop));
(void) printf("|\t\tcompralgo: %llu\n",
- (u_longlong_t)L2BLK_GET_COMPRESS((&lbps)->lbp_prop));
+ (u_longlong_t)L2BLK_GET_COMPRESS(lbps->lbp_prop));
(void) printf("|\t\tcksumalgo: %llu\n",
- (u_longlong_t)L2BLK_GET_CHECKSUM((&lbps)->lbp_prop));
+ (u_longlong_t)L2BLK_GET_CHECKSUM(lbps->lbp_prop));
(void) printf("|\n\n");
}
static void
-dump_l2arc_log_blocks(int fd, l2arc_dev_hdr_phys_t l2dhdr,
+dump_l2arc_log_blocks(int fd, const l2arc_dev_hdr_phys_t *l2dhdr,
l2arc_dev_hdr_phys_t *rebuild)
{
l2arc_log_blk_phys_t this_lb;
if (!dump_opt['q'])
print_l2arc_log_blocks();
- memcpy(lbps, l2dhdr.dh_start_lbps, sizeof (lbps));
+ memcpy(lbps, l2dhdr->dh_start_lbps, sizeof (lbps));
- dev.l2ad_evict = l2dhdr.dh_evict;
- dev.l2ad_start = l2dhdr.dh_start;
- dev.l2ad_end = l2dhdr.dh_end;
+ dev.l2ad_evict = l2dhdr->dh_evict;
+ dev.l2ad_start = l2dhdr->dh_start;
+ dev.l2ad_end = l2dhdr->dh_end;
- if (l2dhdr.dh_start_lbps[0].lbp_daddr == 0) {
+ if (l2dhdr->dh_start_lbps[0].lbp_daddr == 0) {
/* no log blocks to read */
if (!dump_opt['q']) {
(void) printf("No log blocks to read\n");
L2BLK_GET_PSIZE((&lbps[0])->lbp_prop);
}
- dev.l2ad_first = !!(l2dhdr.dh_flags & L2ARC_DEV_HDR_EVICT_FIRST);
+ dev.l2ad_first = !!(l2dhdr->dh_flags & L2ARC_DEV_HDR_EVICT_FIRST);
for (;;) {
if (!l2arc_log_blkptr_valid(&dev, &lbps[0]))
failed++;
if (!dump_opt['q']) {
(void) printf("Invalid cksum\n");
- dump_l2arc_log_blkptr(lbps[0]);
+ dump_l2arc_log_blkptr(&lbps[0]);
}
break;
}
(void) printf("lb[%4llu]\tmagic: %llu\n",
(u_longlong_t)rebuild->dh_lb_count,
(u_longlong_t)this_lb.lb_magic);
- dump_l2arc_log_blkptr(lbps[0]);
+ dump_l2arc_log_blkptr(&lbps[0]);
}
if (dump_opt['l'] > 2 && !dump_opt['q'])
- dump_l2arc_log_entries(l2dhdr.dh_log_entries,
+ dump_l2arc_log_entries(l2dhdr->dh_log_entries,
this_lb.lb_entries,
rebuild->dh_lb_count);
(u_longlong_t)l2dhdr.dh_trim_state);
}
- dump_l2arc_log_blocks(fd, l2dhdr, &rebuild);
+ dump_l2arc_log_blocks(fd, &l2dhdr, &rebuild);
/*
* The total aligned size of log blocks and the number of log blocks
* reported in the header of the device may be less than what zdb
return (err);
}
+static int
+dump_backup_bytes(objset_t *os, void *buf, int len, void *arg)
+{
+ const char *p = (const char *)buf;
+ ssize_t nwritten;
+
+ (void) os;
+ (void) arg;
+
+ /* Write the data out, handling short writes and signals. */
+ while ((nwritten = write(STDOUT_FILENO, p, len)) < len) {
+ if (nwritten < 0) {
+ if (errno == EINTR)
+ continue;
+ return (errno);
+ }
+ p += nwritten;
+ len -= nwritten;
+ }
+
+ return (0);
+}
+
+static void
+dump_backup(const char *pool, uint64_t objset_id, const char *flagstr)
+{
+ boolean_t embed = B_FALSE;
+ boolean_t large_block = B_FALSE;
+ boolean_t compress = B_FALSE;
+ boolean_t raw = B_FALSE;
+
+ const char *c;
+ for (c = flagstr; c != NULL && *c != '\0'; c++) {
+ switch (*c) {
+ case 'e':
+ embed = B_TRUE;
+ break;
+ case 'L':
+ large_block = B_TRUE;
+ break;
+ case 'c':
+ compress = B_TRUE;
+ break;
+ case 'w':
+ raw = B_TRUE;
+ break;
+ default:
+ fprintf(stderr, "dump_backup: invalid flag "
+ "'%c'\n", *c);
+ return;
+ }
+ }
+
+ if (isatty(STDOUT_FILENO)) {
+ fprintf(stderr, "dump_backup: stream cannot be written "
+ "to a terminal\n");
+ return;
+ }
+
+ offset_t off = 0;
+ dmu_send_outparams_t out = {
+ .dso_outfunc = dump_backup_bytes,
+ .dso_dryrun = B_FALSE,
+ };
+
+ int err = dmu_send_obj(pool, objset_id, /* fromsnap */0, embed,
+ large_block, compress, raw, /* saved */ B_FALSE, STDOUT_FILENO,
+ &off, &out);
+ if (err != 0) {
+ fprintf(stderr, "dump_backup: dmu_send_obj: %s\n",
+ strerror(err));
+ return;
+ }
+}
+
static int
zdb_copy_object(objset_t *os, uint64_t srcobj, char *destfile)
{
if (nvlist_size(config, &size, NV_ENCODE_XDR) != 0)
size = buflen;
- /* If the device is a cache device clear the header. */
+ /* If the device is a cache device read the header. */
if (!read_l2arc_header) {
if (nvlist_lookup_uint64(config,
ZPOOL_CONFIG_POOL_STATE, &l2cache) == 0 &&
avl_first(&dmu_objset_ds(os)->ds_bookmarks); dbn != NULL;
dbn = AVL_NEXT(&dmu_objset_ds(os)->ds_bookmarks, dbn)) {
mos_obj_refd(dbn->dbn_phys.zbm_redaction_obj);
- if (dbn->dbn_phys.zbm_redaction_obj != 0)
- global_feature_count[SPA_FEATURE_REDACTION_BOOKMARKS]++;
+ if (dbn->dbn_phys.zbm_redaction_obj != 0) {
+ global_feature_count[
+ SPA_FEATURE_REDACTION_BOOKMARKS]++;
+ objset_t *mos = os->os_spa->spa_meta_objset;
+ dnode_t *rl;
+ VERIFY0(dnode_hold(mos,
+ dbn->dbn_phys.zbm_redaction_obj, FTAG, &rl));
+ if (rl->dn_have_spill) {
+ global_feature_count[
+ SPA_FEATURE_REDACTION_LIST_SPILL]++;
+ }
+ }
if (dbn->dbn_phys.zbm_flags & ZBM_FLAG_HAS_FBN)
global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN]++;
}
#define ZB_TOTAL DN_MAX_LEVELS
#define SPA_MAX_FOR_16M (SPA_MAXBLOCKSHIFT+1)
+typedef struct zdb_brt_entry {
+ dva_t zbre_dva;
+ uint64_t zbre_refcount;
+ avl_node_t zbre_node;
+} zdb_brt_entry_t;
+
typedef struct zdb_cb {
zdb_blkstats_t zcb_type[ZB_TOTAL + 1][ZDB_OT_TOTAL + 1];
uint64_t zcb_removing_size;
uint64_t zcb_checkpoint_size;
uint64_t zcb_dedup_asize;
uint64_t zcb_dedup_blocks;
+ uint64_t zcb_clone_asize;
+ uint64_t zcb_clone_blocks;
uint64_t zcb_psize_count[SPA_MAX_FOR_16M];
uint64_t zcb_lsize_count[SPA_MAX_FOR_16M];
uint64_t zcb_asize_count[SPA_MAX_FOR_16M];
int zcb_haderrors;
spa_t *zcb_spa;
uint32_t **zcb_vd_obsolete_counts;
+ avl_tree_t zcb_brt;
+ boolean_t zcb_brt_is_active;
} zdb_cb_t;
/* test if two DVA offsets from same vdev are within the same metaslab */
zcb->zcb_asize_len[bin] += BP_GET_ASIZE(bp);
zcb->zcb_asize_total += BP_GET_ASIZE(bp);
+ if (zcb->zcb_brt_is_active && brt_maybe_exists(zcb->zcb_spa, bp)) {
+ /*
+ * Cloned blocks are special. We need to count them, so we can
+ * later uncount them when reporting leaked space, and we must
+ * only claim them them once.
+ *
+ * To do this, we keep our own in-memory BRT. For each block
+ * we haven't seen before, we look it up in the real BRT and
+ * if its there, we note it and its refcount then proceed as
+ * normal. If we see the block again, we count it as a clone
+ * and then give it no further consideration.
+ */
+ zdb_brt_entry_t zbre_search, *zbre;
+ avl_index_t where;
+
+ zbre_search.zbre_dva = bp->blk_dva[0];
+ zbre = avl_find(&zcb->zcb_brt, &zbre_search, &where);
+ if (zbre != NULL) {
+ zcb->zcb_clone_asize += BP_GET_ASIZE(bp);
+ zcb->zcb_clone_blocks++;
+
+ zbre->zbre_refcount--;
+ if (zbre->zbre_refcount == 0) {
+ avl_remove(&zcb->zcb_brt, zbre);
+ umem_free(zbre, sizeof (zdb_brt_entry_t));
+ }
+ return;
+ }
+
+ uint64_t crefcnt = brt_entry_get_refcount(zcb->zcb_spa, bp);
+ if (crefcnt > 0) {
+ zbre = umem_zalloc(sizeof (zdb_brt_entry_t),
+ UMEM_NOFAIL);
+ zbre->zbre_dva = bp->blk_dva[0];
+ zbre->zbre_refcount = crefcnt;
+ avl_insert(&zcb->zcb_brt, zbre, where);
+ }
+ }
+
if (dump_opt['L'])
return;
iterate_deleted_livelists(spa, dump_livelist_cb, NULL);
}
+static int
+zdb_brt_entry_compare(const void *zcn1, const void *zcn2)
+{
+ const dva_t *dva1 = &((const zdb_brt_entry_t *)zcn1)->zbre_dva;
+ const dva_t *dva2 = &((const zdb_brt_entry_t *)zcn2)->zbre_dva;
+ int cmp;
+
+ cmp = TREE_CMP(DVA_GET_VDEV(dva1), DVA_GET_VDEV(dva2));
+ if (cmp == 0)
+ cmp = TREE_CMP(DVA_GET_OFFSET(dva1), DVA_GET_OFFSET(dva2));
+
+ return (cmp);
+}
+
static int
dump_block_stats(spa_t *spa)
{
zcb = umem_zalloc(sizeof (zdb_cb_t), UMEM_NOFAIL);
+ if (spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING)) {
+ avl_create(&zcb->zcb_brt, zdb_brt_entry_compare,
+ sizeof (zdb_brt_entry_t),
+ offsetof(zdb_brt_entry_t, zbre_node));
+ zcb->zcb_brt_is_active = B_TRUE;
+ }
+
(void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
(dump_opt['c'] || !dump_opt['L']) ? "to verify " : "",
(dump_opt['c'] == 1) ? "metadata " : "",
metaslab_class_get_alloc(spa_special_class(spa)) +
metaslab_class_get_alloc(spa_dedup_class(spa)) +
get_unflushed_alloc_space(spa);
- total_found = tzb->zb_asize - zcb->zcb_dedup_asize +
+ total_found =
+ tzb->zb_asize - zcb->zcb_dedup_asize - zcb->zcb_clone_asize +
zcb->zcb_removing_size + zcb->zcb_checkpoint_size;
if (total_found == total_alloc && !dump_opt['L']) {
"bp deduped:", (u_longlong_t)zcb->zcb_dedup_asize,
(u_longlong_t)zcb->zcb_dedup_blocks,
(double)zcb->zcb_dedup_asize / tzb->zb_asize + 1.0);
+ (void) printf("\t%-16s %14llu count: %6llu\n",
+ "bp cloned:", (u_longlong_t)zcb->zcb_clone_asize,
+ (u_longlong_t)zcb->zcb_clone_blocks);
(void) printf("\t%-16s %14llu used: %5.2f%%\n", "Normal class:",
(u_longlong_t)norm_alloc, 100.0 * norm_alloc / norm_space);
if (dump_opt['b'] >= 2) {
int l, t, level;
+ char csize[32], lsize[32], psize[32], asize[32];
+ char avg[32], gang[32];
(void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
"\t avg\t comp\t%%Total\tType\n");
+ zfs_blkstat_t *mdstats = umem_zalloc(sizeof (zfs_blkstat_t),
+ UMEM_NOFAIL);
+
for (t = 0; t <= ZDB_OT_TOTAL; t++) {
- char csize[32], lsize[32], psize[32], asize[32];
- char avg[32], gang[32];
const char *typename;
/* make sure nicenum has enough space */
if (zb->zb_asize == 0)
continue;
+ if (level != ZB_TOTAL && t < DMU_OT_NUMTYPES &&
+ (level > 0 || DMU_OT_IS_METADATA(t))) {
+ mdstats->zb_count += zb->zb_count;
+ mdstats->zb_lsize += zb->zb_lsize;
+ mdstats->zb_psize += zb->zb_psize;
+ mdstats->zb_asize += zb->zb_asize;
+ mdstats->zb_gangs += zb->zb_gangs;
+ }
+
if (dump_opt['b'] < 3 && level != ZB_TOTAL)
continue;
}
}
}
+ zdb_nicenum(mdstats->zb_count, csize,
+ sizeof (csize));
+ zdb_nicenum(mdstats->zb_lsize, lsize,
+ sizeof (lsize));
+ zdb_nicenum(mdstats->zb_psize, psize,
+ sizeof (psize));
+ zdb_nicenum(mdstats->zb_asize, asize,
+ sizeof (asize));
+ zdb_nicenum(mdstats->zb_asize / mdstats->zb_count, avg,
+ sizeof (avg));
+ zdb_nicenum(mdstats->zb_gangs, gang, sizeof (gang));
+
+ (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
+ "\t%5.2f\t%6.2f\t",
+ csize, lsize, psize, asize, avg,
+ (double)mdstats->zb_lsize / mdstats->zb_psize,
+ 100.0 * mdstats->zb_asize / tzb->zb_asize);
+ (void) printf("%s\n", "Metadata Total");
/* Output a table summarizing block sizes in the pool */
if (dump_opt['b'] >= 2) {
dump_size_histograms(zcb);
}
+
+ umem_free(mdstats, sizeof (zfs_blkstat_t));
}
(void) printf("\n");
}
typedef struct zdb_ddt_entry {
+ /* key must be first for ddt_key_compare */
ddt_key_t zdde_key;
uint64_t zdde_ref_blocks;
uint64_t zdde_ref_lsize;
ddt_histogram_t ddh_total = {{{0}}};
ddt_stat_t dds_total = {0};
- avl_create(&t, ddt_entry_compare,
+ avl_create(&t, ddt_key_compare,
sizeof (zdb_ddt_entry_t), offsetof(zdb_ddt_entry_t, zdde_node));
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
freecfg = B_TRUE;
}
- if (asprintf(&bogus_name, "%s%s", poolname, BOGUS_SUFFIX) == -1)
+ if (asprintf(&bogus_name, "%s%s", poolname, BOGUS_SUFFIX) == -1) {
+ if (target != poolname)
+ free(poolname);
return (NULL);
+ }
fnvlist_add_string(cfg, ZPOOL_CONFIG_POOL_NAME, bogus_name);
error = spa_import(bogus_name, cfg, NULL,
if (new_path != NULL && path_start != NULL) {
if (asprintf(new_path, "%s%s", bogus_name, path_start) == -1) {
+ free(bogus_name);
if (path_start != NULL)
free(poolname);
return (NULL);
mos_obj_refd(space_map_object(ms->ms_sm));
}
+ if (vd->vdev_root_zap != 0)
+ mos_obj_refd(vd->vdev_root_zap);
+
if (vd->vdev_top_zap != 0) {
mos_obj_refd(vd->vdev_top_zap);
mos_leak_vdev_top_zap(vd);
mos_obj_refd(sls->sls_sm_obj);
}
+static void
+errorlog_count_refd(objset_t *mos, uint64_t errlog)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ for (zap_cursor_init(&zc, mos, errlog);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ mos_obj_refd(za.za_first_integer);
+ }
+ zap_cursor_fini(&zc);
+}
+
static int
dump_mos_leaks(spa_t *spa)
{
mos_obj_refd(spa->spa_history);
mos_obj_refd(spa->spa_errlog_last);
mos_obj_refd(spa->spa_errlog_scrub);
+
+ if (spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
+ errorlog_count_refd(mos, spa->spa_errlog_last);
+ errorlog_count_refd(mos, spa->spa_errlog_scrub);
+ }
+
mos_obj_refd(spa->spa_all_vdev_zaps);
mos_obj_refd(spa->spa_dsl_pool->dp_bptree_obj);
mos_obj_refd(spa->spa_dsl_pool->dp_tmp_userrefs_obj);
}
}
+ if (spa->spa_brt != NULL) {
+ brt_t *brt = spa->spa_brt;
+ for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
+ brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
+ if (brtvd != NULL && brtvd->bv_initiated) {
+ mos_obj_refd(brtvd->bv_mos_brtvdev);
+ mos_obj_refd(brtvd->bv_mos_entries);
+ }
+ }
+ }
+
/*
* Visit all allocated objects and make sure they are referenced.
*/
if (dump_opt['D'])
dump_all_ddts(spa);
+ if (dump_opt['T'])
+ dump_brt(spa);
+
if (dump_opt['d'] > 2 || dump_opt['m'])
dump_metaslabs(spa);
if (dump_opt['M'])
for (spa_feature_t f = 0; f < SPA_FEATURES; f++)
global_feature_count[f] = UINT64_MAX;
global_feature_count[SPA_FEATURE_REDACTION_BOOKMARKS] = 0;
+ global_feature_count[SPA_FEATURE_REDACTION_LIST_SPILL] = 0;
global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN] = 0;
global_feature_count[SPA_FEATURE_LIVELIST] = 0;
#define ZIO_COMPRESS_MASK(alg) (1ULL << (ZIO_COMPRESS_##alg))
static boolean_t
+try_decompress_block(abd_t *pabd, uint64_t lsize, uint64_t psize,
+ int flags, int cfunc, void *lbuf, void *lbuf2)
+{
+ if (flags & ZDB_FLAG_VERBOSE) {
+ (void) fprintf(stderr,
+ "Trying %05llx -> %05llx (%s)\n",
+ (u_longlong_t)psize,
+ (u_longlong_t)lsize,
+ zio_compress_table[cfunc].ci_name);
+ }
+
+ /*
+ * We set lbuf to all zeros and lbuf2 to all
+ * ones, then decompress to both buffers and
+ * compare their contents. This way we can
+ * know if decompression filled exactly to
+ * lsize or if it left some bytes unwritten.
+ */
+
+ memset(lbuf, 0x00, lsize);
+ memset(lbuf2, 0xff, lsize);
+
+ if (zio_decompress_data(cfunc, pabd,
+ lbuf, psize, lsize, NULL) == 0 &&
+ zio_decompress_data(cfunc, pabd,
+ lbuf2, psize, lsize, NULL) == 0 &&
+ memcmp(lbuf, lbuf2, lsize) == 0)
+ return (B_TRUE);
+ return (B_FALSE);
+}
+
+static uint64_t
zdb_decompress_block(abd_t *pabd, void *buf, void *lbuf, uint64_t lsize,
uint64_t psize, int flags)
{
(void) buf;
- boolean_t exceeded = B_FALSE;
+ uint64_t orig_lsize = lsize;
+ boolean_t tryzle = ((getenv("ZDB_NO_ZLE") == NULL));
+ boolean_t found = B_FALSE;
/*
* We don't know how the data was compressed, so just try
* every decompress function at every inflated blocksize.
uint64_t maxlsize = SPA_MAXBLOCKSIZE;
uint64_t mask = ZIO_COMPRESS_MASK(ON) | ZIO_COMPRESS_MASK(OFF) |
ZIO_COMPRESS_MASK(INHERIT) | ZIO_COMPRESS_MASK(EMPTY) |
- (getenv("ZDB_NO_ZLE") ? ZIO_COMPRESS_MASK(ZLE) : 0);
+ ZIO_COMPRESS_MASK(ZLE);
*cfuncp++ = ZIO_COMPRESS_LZ4;
*cfuncp++ = ZIO_COMPRESS_LZJB;
mask |= ZIO_COMPRESS_MASK(LZ4) | ZIO_COMPRESS_MASK(LZJB);
+ /*
+ * Every gzip level has the same decompressor, no need to
+ * run it 9 times per bruteforce attempt.
+ */
+ mask |= ZIO_COMPRESS_MASK(GZIP_2) | ZIO_COMPRESS_MASK(GZIP_3);
+ mask |= ZIO_COMPRESS_MASK(GZIP_4) | ZIO_COMPRESS_MASK(GZIP_5);
+ mask |= ZIO_COMPRESS_MASK(GZIP_6) | ZIO_COMPRESS_MASK(GZIP_7);
+ mask |= ZIO_COMPRESS_MASK(GZIP_8) | ZIO_COMPRESS_MASK(GZIP_9);
for (int c = 0; c < ZIO_COMPRESS_FUNCTIONS; c++)
if (((1ULL << c) & mask) == 0)
*cfuncp++ = c;
lsize += SPA_MINBLOCKSIZE;
else
maxlsize = lsize;
+
for (; lsize <= maxlsize; lsize += SPA_MINBLOCKSIZE) {
for (cfuncp = cfuncs; *cfuncp; cfuncp++) {
- if (flags & ZDB_FLAG_VERBOSE) {
- (void) fprintf(stderr,
- "Trying %05llx -> %05llx (%s)\n",
- (u_longlong_t)psize,
- (u_longlong_t)lsize,
- zio_compress_table[*cfuncp].\
- ci_name);
- }
-
- /*
- * We randomize lbuf2, and decompress to both
- * lbuf and lbuf2. This way, we will know if
- * decompression fill exactly to lsize.
- */
- VERIFY0(random_get_pseudo_bytes(lbuf2, lsize));
-
- if (zio_decompress_data(*cfuncp, pabd,
- lbuf, psize, lsize, NULL) == 0 &&
- zio_decompress_data(*cfuncp, pabd,
- lbuf2, psize, lsize, NULL) == 0 &&
- memcmp(lbuf, lbuf2, lsize) == 0)
+ if (try_decompress_block(pabd, lsize, psize, flags,
+ *cfuncp, lbuf, lbuf2)) {
+ found = B_TRUE;
break;
+ }
}
if (*cfuncp != 0)
break;
}
+ if (!found && tryzle) {
+ for (lsize = orig_lsize; lsize <= maxlsize;
+ lsize += SPA_MINBLOCKSIZE) {
+ if (try_decompress_block(pabd, lsize, psize, flags,
+ ZIO_COMPRESS_ZLE, lbuf, lbuf2)) {
+ *cfuncp = ZIO_COMPRESS_ZLE;
+ found = B_TRUE;
+ break;
+ }
+ }
+ }
umem_free(lbuf2, SPA_MAXBLOCKSIZE);
- if (lsize > maxlsize) {
- exceeded = B_TRUE;
- }
if (*cfuncp == ZIO_COMPRESS_ZLE) {
printf("\nZLE decompression was selected. If you "
"suspect the results are wrong,\ntry avoiding ZLE "
"by setting and exporting ZDB_NO_ZLE=\"true\"\n");
}
- return (exceeded);
+ return (lsize > maxlsize ? -1 : lsize);
}
/*
vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev);
if (vd == NULL) {
(void) printf("***Invalid vdev: %s\n", vdev);
- free(dup);
- return;
+ goto done;
} else {
if (vd->vdev_path)
(void) fprintf(stderr, "Found vdev: %s\n",
*/
zio_nowait(zio_vdev_child_io(zio, bp, vd, offset, pabd,
psize, ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ,
- ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_PROPAGATE |
- ZIO_FLAG_DONT_RETRY | ZIO_FLAG_CANFAIL | ZIO_FLAG_RAW |
- ZIO_FLAG_OPTIONAL, NULL, NULL));
+ ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY |
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_RAW | ZIO_FLAG_OPTIONAL,
+ NULL, NULL));
}
error = zio_wait(zio);
uint64_t orig_lsize = lsize;
buf = lbuf;
if (flags & ZDB_FLAG_DECOMPRESS) {
- boolean_t failed = zdb_decompress_block(pabd, buf, lbuf,
+ lsize = zdb_decompress_block(pabd, buf, lbuf,
lsize, psize, flags);
- if (failed) {
+ if (lsize == -1) {
(void) printf("Decompress of %s failed\n", thing);
goto out;
}
!(flags & ZDB_FLAG_DECOMPRESS)) {
const blkptr_t *b = (const blkptr_t *)(void *)
((uintptr_t)buf + (uintptr_t)blkptr_offset);
- if (zfs_blkptr_verify(spa, b, B_FALSE, BLK_VERIFY_ONLY) ==
- B_FALSE) {
+ if (zfs_blkptr_verify(spa, b,
+ BLK_CONFIG_NEEDED, BLK_VERIFY_ONLY) == B_FALSE) {
abd_return_buf_copy(pabd, buf, lsize);
borrowed = B_FALSE;
buf = lbuf;
- boolean_t failed = zdb_decompress_block(pabd, buf,
+ lsize = zdb_decompress_block(pabd, buf,
lbuf, lsize, psize, flags);
b = (const blkptr_t *)(void *)
((uintptr_t)buf + (uintptr_t)blkptr_offset);
- if (failed || zfs_blkptr_verify(spa, b, B_FALSE,
- BLK_VERIFY_LOG) == B_FALSE) {
+ if (lsize == -1 || zfs_blkptr_verify(spa, b,
+ BLK_CONFIG_NEEDED, BLK_VERIFY_LOG) == B_FALSE) {
printf("invalid block pointer at this DVA\n");
goto out;
}
BP_SET_CHECKSUM(bp, ck);
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
czio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
- czio->io_bp = bp;
-
if (vd == vd->vdev_top) {
zio_nowait(zio_read(czio, spa, bp, pabd, psize,
NULL, NULL,
zio_nowait(zio_vdev_child_io(czio, bp, vd,
offset, pabd, psize, ZIO_TYPE_READ,
ZIO_PRIORITY_SYNC_READ,
- ZIO_FLAG_DONT_CACHE |
ZIO_FLAG_DONT_PROPAGATE |
ZIO_FLAG_DONT_RETRY |
ZIO_FLAG_CANFAIL | ZIO_FLAG_RAW |
}
error = zio_wait(czio);
if (error == 0 || error == ECKSUM) {
- zio_t *ck_zio = zio_root(spa, NULL, NULL, 0);
+ zio_t *ck_zio = zio_null(NULL, spa, NULL,
+ NULL, NULL, 0);
ck_zio->io_offset =
DVA_GET_OFFSET(&bp->blk_dva[0]);
ck_zio->io_bp = bp;
zio_checksum_compute(ck_zio, ck, pabd, lsize);
- printf("%12s\tcksum=%llx:%llx:%llx:%llx\n",
+ printf(
+ "%12s\t"
+ "cksum=%016llx:%016llx:%016llx:%016llx\n",
zio_checksum_table[ck].ci_name,
(u_longlong_t)bp->blk_cksum.zc_word[0],
(u_longlong_t)bp->blk_cksum.zc_word[1],
struct option long_options[] = {
{"ignore-assertions", no_argument, NULL, 'A'},
{"block-stats", no_argument, NULL, 'b'},
+ {"backup", no_argument, NULL, 'B'},
{"checksum", no_argument, NULL, 'c'},
{"config", no_argument, NULL, 'C'},
{"datasets", no_argument, NULL, 'd'},
{"intent-logs", no_argument, NULL, 'i'},
{"inflight", required_argument, NULL, 'I'},
{"checkpointed-state", no_argument, NULL, 'k'},
+ {"key", required_argument, NULL, 'K'},
{"label", no_argument, NULL, 'l'},
{"disable-leak-tracking", no_argument, NULL, 'L'},
{"metaslabs", no_argument, NULL, 'm'},
{"io-stats", no_argument, NULL, 's'},
{"simulate-dedup", no_argument, NULL, 'S'},
{"txg", required_argument, NULL, 't'},
+ {"brt-stats", no_argument, NULL, 'T'},
{"uberblock", no_argument, NULL, 'u'},
{"cachefile", required_argument, NULL, 'U'},
{"verbose", no_argument, NULL, 'v'},
};
while ((c = getopt_long(argc, argv,
- "AbcCdDeEFGhiI:klLmMNo:Op:PqrRsSt:uU:vVx:XYyZ",
+ "AbBcCdDeEFGhiI:kK:lLmMNo:Op:PqrRsSt:TuU:vVx:XYyZ",
long_options, NULL)) != -1) {
switch (c) {
case 'b':
+ case 'B':
case 'c':
case 'C':
case 'd':
case 'R':
case 's':
case 'S':
+ case 'T':
case 'u':
case 'y':
case 'Z':
usage();
}
break;
+ case 'K':
+ dump_opt[c]++;
+ key_material = strdup(optarg);
+ /* redact key material in process table */
+ while (*optarg != '\0') { *optarg++ = '*'; }
+ break;
case 'o':
error = set_global_var(optarg);
if (error != 0)
* ZDB does not typically re-read blocks; therefore limit the ARC
* to 256 MB, which can be used entirely for metadata.
*/
- zfs_arc_min = zfs_arc_meta_min = 2ULL << SPA_MAXBLOCKSHIFT;
- zfs_arc_max = zfs_arc_meta_limit = 256 * 1024 * 1024;
+ zfs_arc_min = 2ULL << SPA_MAXBLOCKSHIFT;
+ zfs_arc_max = 256 * 1024 * 1024;
#endif
/*
verbose = MAX(verbose, 1);
for (c = 0; c < 256; c++) {
- if (dump_all && strchr("AeEFklLNOPrRSXy", c) == NULL)
+ if (dump_all && strchr("ABeEFkKlLNOPrRSXy", c) == NULL)
dump_opt[c] = 1;
if (dump_opt[c])
dump_opt[c] += verbose;
if (dump_opt['l'])
return (dump_label(argv[0]));
- if (dump_opt['O']) {
- if (argc != 2)
- usage();
- dump_opt['v'] = verbose + 3;
- return (dump_path(argv[0], argv[1], NULL));
- }
- if (dump_opt['r']) {
- target_is_spa = B_FALSE;
- if (argc != 3)
- usage();
- dump_opt['v'] = verbose;
- error = dump_path(argv[0], argv[1], &object);
- if (error != 0)
- fatal("internal error: %s", strerror(error));
- }
-
if (dump_opt['X'] || dump_opt['F'])
rewind = ZPOOL_DO_REWIND |
(dump_opt['X'] ? ZPOOL_EXTREME_REWIND : 0);
args.path = searchdirs;
args.can_be_active = B_TRUE;
- error = zpool_find_config(NULL, target_pool, &cfg, &args,
- &libzpool_config_ops);
+ libpc_handle_t lpch = {
+ .lpc_lib_handle = NULL,
+ .lpc_ops = &libzpool_config_ops,
+ .lpc_printerr = B_TRUE
+ };
+ error = zpool_find_config(&lpch, target_pool, &cfg, &args);
if (error == 0) {
searchdirs = NULL;
}
+ /*
+ * We need to make sure to process -O option or call
+ * dump_path after the -e option has been processed,
+ * which imports the pool to the namespace if it's
+ * not in the cachefile.
+ */
+ if (dump_opt['O']) {
+ if (argc != 2)
+ usage();
+ dump_opt['v'] = verbose + 3;
+ return (dump_path(argv[0], argv[1], NULL));
+ }
+
+ if (dump_opt['r']) {
+ target_is_spa = B_FALSE;
+ if (argc != 3)
+ usage();
+ dump_opt['v'] = verbose;
+ error = dump_path(argv[0], argv[1], &object);
+ if (error != 0)
+ fatal("internal error: %s", strerror(error));
+ }
+
/*
* import_checkpointed_state makes the assumption that the
* target pool that we pass it is already part of the spa
checkpoint_pool, error);
}
- } else if (target_is_spa || dump_opt['R'] || objset_id == 0) {
+ } else if (target_is_spa || dump_opt['R'] || dump_opt['B'] ||
+ objset_id == 0) {
zdb_set_skip_mmp(target);
error = spa_open_rewind(target, &spa, FTAG, policy,
NULL);
strerror(errno));
}
}
- if (os != NULL) {
+ if (dump_opt['B']) {
+ dump_backup(target, objset_id,
+ argc > 0 ? argv[0] : NULL);
+ } else if (os != NULL) {
dump_objset(os);
} else if (zopt_object_args > 0 && !dump_opt['m']) {
dump_objset(spa->spa_meta_objset);
projects / mirror_zfs.git / blobdiff