Implement Redacted Send/Receive

Redacted send/receive allows users to send subsets of their data to 
a target system. One possible use case for this feature is to not 
transmit sensitive information to a data warehousing, test/dev, or 
analytics environment. Another is to save space by not replicating 
unimportant data within a given dataset, for example in backup tools 
like zrepl.

Redacted send/receive is a three-stage process. First, a clone (or 
clones) is made of the snapshot to be sent to the target. In this 
clone (or clones), all unnecessary or unwanted data is removed or
modified. This clone is then snapshotted to create the "redaction 
snapshot" (or snapshots). Second, the new zfs redact command is used 
to create a redaction bookmark. The redaction bookmark stores the 
list of blocks in a snapshot that were modified by the redaction 
snapshot(s). Finally, the redaction bookmark is passed as a parameter 
to zfs send. When sending to the snapshot that was redacted, the
redaction bookmark is used to filter out blocks that contain sensitive 
or unwanted information, and those blocks are not included in the send 
stream.  When sending from the redaction bookmark, the blocks it 
contains are considered as candidate blocks in addition to those 
blocks in the destination snapshot that were modified since the 
creation_txg of the redaction bookmark.  This step is necessary to 
allow the target to rehydrate data in the case where some blocks are 
accidentally or unnecessarily modified in the redaction snapshot.

The changes to bookmarks to enable fast space estimation involve 
adding deadlists to bookmarks. There is also logic to manage the 
life cycles of these deadlists.

The new size estimation process operates in cases where previously 
an accurate estimate could not be provided. In those cases, a send 
is performed where no data blocks are read, reducing the runtime 
significantly and providing a byte-accurate size estimate.

Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: Prashanth Sreenivasa <pks@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: Chris Williamson <chris.williamson@delphix.com>
Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com>
Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Paul Dagnelie <pcd@delphix.com>
Closes #7958

cmd/zdb/zdb.c

@@ -50,11 +50,13 @@
 #include <sys/dsl_dir.h>
 #include <sys/dsl_dataset.h>
 #include <sys/dsl_pool.h>
+#include <sys/dsl_bookmark.h>
 #include <sys/dbuf.h>
 #include <sys/zil.h>
 #include <sys/zil_impl.h>
 #include <sys/stat.h>
 #include <sys/resource.h>
+#include <sys/dmu_send.h>
 #include <sys/dmu_traverse.h>
 #include <sys/zio_checksum.h>
 #include <sys/zio_compress.h>
@@ -143,6 +145,7 @@ usage(void)
 	    "\t\t[<poolname> [<object> ...]]\n"
 	    "\t%s [-AdiPv] [-e [-V] [-p <path> ...]] [-U <cache>] <dataset>\n"
 	    "\t\t[<object> ...]\n"
+	    "\t%s [-v] <bookmark>\n"
 	    "\t%s -C [-A] [-U <cache>]\n"
 	    "\t%s -l [-Aqu] <device>\n"
 	    "\t%s -m [-AFLPX] [-e [-V] [-p <path> ...]] [-t <txg>] "
@@ -154,7 +157,7 @@ usage(void)
 	    "\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);
 
 	(void) fprintf(stderr, "    Dataset name must include at least one "
 	    "separator character '/' or '@'\n");
@@ -420,6 +423,43 @@ dump_uint8(objset_t *os, uint64_t object, void *data, size_t size)
 static void
 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size)
 {
+	uint64_t *arr;
+
+	if (dump_opt['d'] < 6)
+		return;
+	if (data == NULL) {
+		dmu_object_info_t doi;
+
+		VERIFY0(dmu_object_info(os, object, &doi));
+		size = doi.doi_max_offset;
+		arr = kmem_alloc(size, KM_SLEEP);
+
+		int err = dmu_read(os, object, 0, size, arr, 0);
+		if (err != 0) {
+			(void) printf("got error %u from dmu_read\n", err);
+			kmem_free(arr, size);
+			return;
+		}
+	} else {
+		arr = data;
+	}
+
+	if (size == 0) {
+		(void) printf("\t\t[]\n");
+		return;
+	}
+
+	(void) printf("\t\t[%0llx", (u_longlong_t)arr[0]);
+	for (size_t i = 1; i * sizeof (uint64_t) < size; i++) {
+		if (i % 4 != 0)
+			(void) printf(", %0llx", (u_longlong_t)arr[i]);
+		else
+			(void) printf(",\n\t\t%0llx", (u_longlong_t)arr[i]);
+	}
+	(void) printf("]\n");
+
+	if (data == NULL)
+		kmem_free(arr, size);
 }
 
 /*ARGSUSED*/
@@ -1498,6 +1538,7 @@ visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
 		int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
 		arc_buf_t *buf;
 		uint64_t fill = 0;
+		ASSERT(!BP_IS_REDACTED(bp));
 
 		err = arc_read(NULL, spa, bp, arc_getbuf_func, &buf,
 		    ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
@@ -1784,6 +1825,128 @@ dump_full_bpobj(bpobj_t *bpo, const char *name, int indent)
 	}
 }
 
+static int
+dump_bookmark(dsl_pool_t *dp, char *name, boolean_t print_redact,
+    boolean_t print_list)
+{
+	int err = 0;
+	zfs_bookmark_phys_t prop;
+	objset_t *mos = dp->dp_spa->spa_meta_objset;
+	err = dsl_bookmark_lookup(dp, name, NULL, &prop);
+
+	if (err != 0) {
+		return (err);
+	}
+
+	(void) printf("\t#%s: ", strchr(name, '#') + 1);
+	(void) printf("{guid: %llx creation_txg: %llu creation_time: "
+	    "%llu redaction_obj: %llu}\n", (u_longlong_t)prop.zbm_guid,
+	    (u_longlong_t)prop.zbm_creation_txg,
+	    (u_longlong_t)prop.zbm_creation_time,
+	    (u_longlong_t)prop.zbm_redaction_obj);
+
+	IMPLY(print_list, print_redact);
+	if (!print_redact || prop.zbm_redaction_obj == 0)
+		return (0);
+
+	redaction_list_t *rl;
+	VERIFY0(dsl_redaction_list_hold_obj(dp,
+	    prop.zbm_redaction_obj, FTAG, &rl));
+
+	redaction_list_phys_t *rlp = rl->rl_phys;
+	(void) printf("\tRedacted:\n\t\tProgress: ");
+	if (rlp->rlp_last_object != UINT64_MAX ||
+	    rlp->rlp_last_blkid != UINT64_MAX) {
+		(void) printf("%llu %llu (incomplete)\n",
+		    (u_longlong_t)rlp->rlp_last_object,
+		    (u_longlong_t)rlp->rlp_last_blkid);
+	} else {
+		(void) printf("complete\n");
+	}
+	(void) printf("\t\tSnapshots: [");
+	for (unsigned int i = 0; i < rlp->rlp_num_snaps; i++) {
+		if (i > 0)
+			(void) printf(", ");
+		(void) printf("%0llu",
+		    (u_longlong_t)rlp->rlp_snaps[i]);
+	}
+	(void) printf("]\n\t\tLength: %llu\n",
+	    (u_longlong_t)rlp->rlp_num_entries);
+
+	if (!print_list) {
+		dsl_redaction_list_rele(rl, FTAG);
+		return (0);
+	}
+
+	if (rlp->rlp_num_entries == 0) {
+		dsl_redaction_list_rele(rl, FTAG);
+		(void) printf("\t\tRedaction List: []\n\n");
+		return (0);
+	}
+
+	redact_block_phys_t *rbp_buf;
+	uint64_t size;
+	dmu_object_info_t doi;
+
+	VERIFY0(dmu_object_info(mos, prop.zbm_redaction_obj, &doi));
+	size = doi.doi_max_offset;
+	rbp_buf = kmem_alloc(size, KM_SLEEP);
+
+	err = dmu_read(mos, prop.zbm_redaction_obj, 0, size,
+	    rbp_buf, 0);
+	if (err != 0) {
+		dsl_redaction_list_rele(rl, FTAG);
+		kmem_free(rbp_buf, size);
+		return (err);
+	}
+
+	(void) printf("\t\tRedaction List: [{object: %llx, offset: "
+	    "%llx, blksz: %x, count: %llx}",
+	    (u_longlong_t)rbp_buf[0].rbp_object,
+	    (u_longlong_t)rbp_buf[0].rbp_blkid,
+	    (uint_t)(redact_block_get_size(&rbp_buf[0])),
+	    (u_longlong_t)redact_block_get_count(&rbp_buf[0]));
+
+	for (size_t i = 1; i < rlp->rlp_num_entries; i++) {
+		(void) printf(",\n\t\t{object: %llx, offset: %llx, "
+		    "blksz: %x, count: %llx}",
+		    (u_longlong_t)rbp_buf[i].rbp_object,
+		    (u_longlong_t)rbp_buf[i].rbp_blkid,
+		    (uint_t)(redact_block_get_size(&rbp_buf[i])),
+		    (u_longlong_t)redact_block_get_count(&rbp_buf[i]));
+	}
+	dsl_redaction_list_rele(rl, FTAG);
+	kmem_free(rbp_buf, size);
+	(void) printf("]\n\n");
+	return (0);
+}
+
+static void
+dump_bookmarks(objset_t *os, int verbosity)
+{
+	zap_cursor_t zc;
+	zap_attribute_t attr;
+	dsl_dataset_t *ds = dmu_objset_ds(os);
+	dsl_pool_t *dp = spa_get_dsl(os->os_spa);
+	objset_t *mos = os->os_spa->spa_meta_objset;
+	if (verbosity < 4)
+		return;
+	dsl_pool_config_enter(dp, FTAG);
+
+	for (zap_cursor_init(&zc, mos, ds->ds_bookmarks_obj);
+	    zap_cursor_retrieve(&zc, &attr) == 0;
+	    zap_cursor_advance(&zc)) {
+		char osname[ZFS_MAX_DATASET_NAME_LEN];
+		char buf[ZFS_MAX_DATASET_NAME_LEN];
+		dmu_objset_name(os, osname);
+		VERIFY0(snprintf(buf, sizeof (buf), "%s#%s", osname,
+		    attr.za_name));
+		(void) dump_bookmark(dp, buf, verbosity >= 5, verbosity >= 6);
+	}
+	zap_cursor_fini(&zc);
+	dsl_pool_config_exit(dp, FTAG);
+}
+
 static void
 bpobj_count_refd(bpobj_t *bpo)
 {
@@ -1886,19 +2049,26 @@ static objset_t *sa_os = NULL;
 static sa_attr_type_t *sa_attr_table = NULL;
 
 static int
-open_objset(const char *path, dmu_objset_type_t type, void *tag, objset_t **osp)
+open_objset(const char *path, void *tag, objset_t **osp)
 {
 	int err;
 	uint64_t sa_attrs = 0;
 	uint64_t version = 0;
 
 	VERIFY3P(sa_os, ==, NULL);
-	err = dmu_objset_own(path, type, B_TRUE, B_FALSE, tag, osp);
+	/*
+	 * 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 (err != 0) {
-		(void) fprintf(stderr, "failed to own dataset '%s': %s\n", path,
-		    strerror(err));
+		(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);
 
 	if (dmu_objset_type(*osp) == DMU_OST_ZFS && !(*osp)->os_encrypted) {
 		(void) zap_lookup(*osp, MASTER_NODE_OBJ, ZPL_VERSION_STR,
@@ -1912,7 +2082,8 @@ open_objset(const char *path, dmu_objset_type_t type, void *tag, objset_t **osp)
 		if (err != 0) {
 			(void) fprintf(stderr, "sa_setup failed: %s\n",
 			    strerror(err));
-			dmu_objset_disown(*osp, B_FALSE, tag);
+			dsl_dataset_long_rele(dmu_objset_ds(*osp), tag);
+			dsl_dataset_rele(dmu_objset_ds(*osp), tag);
 			*osp = NULL;
 		}
 	}
@@ -1927,7 +2098,8 @@ close_objset(objset_t *os, void *tag)
 	VERIFY3P(os, ==, sa_os);
 	if (os->os_sa != NULL)
 		sa_tear_down(os);
-	dmu_objset_disown(os, B_FALSE, tag);
+	dsl_dataset_long_rele(dmu_objset_ds(os), tag);
+	dsl_dataset_rele(dmu_objset_ds(os), tag);
 	sa_attr_table = NULL;
 	sa_os = NULL;
 }
@@ -2205,8 +2377,8 @@ static object_viewer_t *object_viewer[DMU_OT_NUMTYPES + 1] = {
 };
 
 static void
-dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header,
-    uint64_t *dnode_slots_used)
+dump_object(objset_t *os, uint64_t object, int verbosity,
+    boolean_t *print_header, uint64_t *dnode_slots_used)
 {
 	dmu_buf_t *db = NULL;
 	dmu_object_info_t doi;
@@ -2325,7 +2497,7 @@ dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header,
 			(void) printf("\t\t(object encrypted)\n");
 		}
 
-		*print_header = 1;
+		*print_header = B_TRUE;
 	}
 
 	if (verbosity >= 5)
@@ -2396,6 +2568,7 @@ count_ds_mos_objects(dsl_dataset_t *ds)
 	mos_obj_refd(dsl_dataset_phys(ds)->ds_props_obj);
 	mos_obj_refd(dsl_dataset_phys(ds)->ds_userrefs_obj);
 	mos_obj_refd(dsl_dataset_phys(ds)->ds_snapnames_zapobj);
+	mos_obj_refd(ds->ds_bookmarks_obj);
 
 	if (!dsl_dataset_is_snapshot(ds)) {
 		count_dir_mos_objects(ds->ds_dir);
@@ -2416,7 +2589,7 @@ dump_dir(objset_t *os)
 	char osname[ZFS_MAX_DATASET_NAME_LEN];
 	const char *type = "UNKNOWN";
 	int verbosity = dump_opt['d'];
-	int print_header = 1;
+	boolean_t print_header;
 	unsigned i;
 	int error;
 	uint64_t total_slots_used = 0;
@@ -2430,6 +2603,8 @@ dump_dir(objset_t *os)
 	dmu_objset_fast_stat(os, &dds);
 	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
 
+	print_header = B_TRUE;
+
 	if (dds.dds_type < DMU_OST_NUMTYPES)
 		type = objset_types[dds.dds_type];
 
@@ -2464,9 +2639,10 @@ dump_dir(objset_t *os)
 	    (dds.dds_inconsistent) ? " (inconsistent)" : "");
 
 	if (zopt_objects != 0) {
-		for (i = 0; i < zopt_objects; i++)
+		for (i = 0; i < zopt_objects; i++) {
 			dump_object(os, zopt_object[i], verbosity,
 			    &print_header, NULL);
+		}
 		(void) printf("\n");
 		return;
 	}
@@ -2485,6 +2661,9 @@ dump_dir(objset_t *os)
 		count_ds_mos_objects(ds);
 	}
 
+	if (dmu_objset_ds(os) != NULL)
+		dump_bookmarks(os, verbosity);
+
 	if (verbosity < 2)
 		return;
 
@@ -2962,7 +3141,7 @@ static int
 dump_path_impl(objset_t *os, uint64_t obj, char *name)
 {
 	int err;
-	int header = 1;
+	boolean_t header = B_TRUE;
 	uint64_t child_obj;
 	char *s;
 	dmu_buf_t *db;
@@ -3033,7 +3212,7 @@ dump_path(char *ds, char *path)
 	objset_t *os;
 	uint64_t root_obj;
 
-	err = open_objset(ds, DMU_OST_ZFS, FTAG, &os);
+	err = open_objset(ds, FTAG, &os);
 	if (err != 0)
 		return (err);
 
@@ -3041,7 +3220,7 @@ dump_path(char *ds, char *path)
 	if (err != 0) {
 		(void) fprintf(stderr, "can't lookup root znode: %s\n",
 		    strerror(err));
-		dmu_objset_disown(os, B_FALSE, FTAG);
+		close_objset(os, FTAG);
 		return (EINVAL);
 	}
 
@@ -3222,6 +3401,7 @@ dump_label(const char *dev)
 }
 
 static uint64_t dataset_feature_count[SPA_FEATURES];
+static uint64_t global_feature_count[SPA_FEATURES];
 static uint64_t remap_deadlist_count = 0;
 
 /*ARGSUSED*/
@@ -3232,7 +3412,7 @@ dump_one_dir(const char *dsname, void *arg)
 	objset_t *os;
 	spa_feature_t f;
 
-	error = open_objset(dsname, DMU_OST_ANY, FTAG, &os);
+	error = open_objset(dsname, FTAG, &os);
 	if (error != 0)
 		return (0);
 
@@ -3248,6 +3428,16 @@ dump_one_dir(const char *dsname, void *arg)
 		remap_deadlist_count++;
 	}
 
+	for (dsl_bookmark_node_t *dbn =
+	    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_flags & ZBM_FLAG_HAS_FBN)
+			global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN]++;
+	}
+
 	dump_dir(os);
 	close_objset(os, FTAG);
 	fuid_table_destroy();
@@ -3484,7 +3674,7 @@ zdb_blkptr_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 	dmu_object_type_t type;
 	boolean_t is_metadata;
 
-	if (bp == NULL)
+	if (zb->zb_level == ZB_DNODE_LEVEL)
 		return (0);
 
 	if (dump_opt['b'] >= 5 && bp->blk_birth > 0) {
@@ -3499,7 +3689,7 @@ zdb_blkptr_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 		    blkbuf);
 	}
 
-	if (BP_IS_HOLE(bp))
+	if (BP_IS_HOLE(bp) || BP_IS_REDACTED(bp))
 		return (0);
 
 	type = BP_GET_TYPE(bp);
@@ -4545,7 +4735,8 @@ zdb_ddt_add_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 	avl_index_t where;
 	zdb_ddt_entry_t *zdde, zdde_search;
 
-	if (bp == NULL || BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp))
+	if (zb->zb_level == ZB_DNODE_LEVEL || BP_IS_HOLE(bp) ||
+	    BP_IS_EMBEDDED(bp))
 		return (0);
 
 	if (dump_opt['S'] > 1 && zb->zb_level == ZB_ROOT_LEVEL) {
@@ -5381,6 +5572,12 @@ dump_zpool(spa_t *spa)
 			}
 			dump_dtl(spa->spa_root_vdev, 0);
 		}
+
+		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_BOOKMARK_WRITTEN] = 0;
+
 		(void) dmu_objset_find(spa_name(spa), dump_one_dir,
 		    NULL, DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
 
@@ -5390,21 +5587,31 @@ dump_zpool(spa_t *spa)
 		for (f = 0; f < SPA_FEATURES; f++) {
 			uint64_t refcount;
 
+			uint64_t *arr;
 			if (!(spa_feature_table[f].fi_flags &
-			    ZFEATURE_FLAG_PER_DATASET) ||
-			    !spa_feature_is_enabled(spa, f)) {
-				ASSERT0(dataset_feature_count[f]);
-				continue;
+			    ZFEATURE_FLAG_PER_DATASET)) {
+				if (global_feature_count[f] == UINT64_MAX)
+					continue;
+				if (!spa_feature_is_enabled(spa, f)) {
+					ASSERT0(global_feature_count[f]);
+					continue;
+				}
+				arr = global_feature_count;
+			} else {
+				if (!spa_feature_is_enabled(spa, f)) {
+					ASSERT0(dataset_feature_count[f]);
+					continue;
+				}
+				arr = dataset_feature_count;
 			}
 			if (feature_get_refcount(spa, &spa_feature_table[f],
 			    &refcount) == ENOTSUP)
 				continue;
-			if (dataset_feature_count[f] != refcount) {
+			if (arr[f] != refcount) {
 				(void) printf("%s feature refcount mismatch: "
-				    "%lld datasets != %lld refcount\n",
+				    "%lld consumers != %lld refcount\n",
 				    spa_feature_table[f].fi_uname,
-				    (longlong_t)dataset_feature_count[f],
-				    (longlong_t)refcount);
+				    (longlong_t)arr[f], (longlong_t)refcount);
 				rc = 2;
 			} else {
 				(void) printf("Verified %s feature refcount "
@@ -6184,9 +6391,23 @@ main(int argc, char **argv)
 					    FTAG, policy, NULL);
 				}
 			}
+		} else if (strpbrk(target, "#") != NULL) {
+			dsl_pool_t *dp;
+			error = dsl_pool_hold(target, FTAG, &dp);
+			if (error != 0) {
+				fatal("can't dump '%s': %s", target,
+				    strerror(error));
+			}
+			error = dump_bookmark(dp, target, B_TRUE, verbose > 1);
+			dsl_pool_rele(dp, FTAG);
+			if (error != 0) {
+				fatal("can't dump '%s': %s", target,
+				    strerror(error));
+			}
+			return (error);
 		} else {
 			zdb_set_skip_mmp(target);
-			error = open_objset(target, DMU_OST_ANY, FTAG, &os);
+			error = open_objset(target, FTAG, &os);
 			if (error == 0)
 				spa = dmu_objset_spa(os);
 		}
@@ -6245,10 +6466,11 @@ main(int argc, char **argv)
 			free(checkpoint_target);
 	}
 
-	if (os != NULL)
+	if (os != NULL) {
 		close_objset(os, FTAG);
-	else
+	} else {
 		spa_close(spa, FTAG);
+	}
 
 	fuid_table_destroy();