WIP: Detect and prevent mixed raw and non-raw sends

Currently, there is an issue in the raw receive code where
raw receives are allowed to happen on top of previously
non-raw received datasets. This is a problem because the
source-side dataset doesn't know about how the blocks on
the destination were encrypted. As a result, any MAC in
the objset's checksum-of-MACs tree that is a parent of both
blocks encrypted on the source and blocks encrypted by the
destination will be incorrect. This will result in
authentication errors when we decrypt the dataset.

This patch fixes this issue by adding a new check to the
raw receive code. The code now maintains an "IVset guid",
which acts as an identifier for the set of IVs used to
encrypt a given snapshot. When a snapshot is raw received,
the destination snapshot will take this value from the
DRR_BEGIN payload. Non-raw receives and normal "zfs snap"
operations will cause ZFS to generate a new IVset guid.
When a raw incremental stream is received, ZFS will check
that the "from" IVset guid in the stream matches that of
the "from" destination snapshot. If they do not match, the
code will error out the receive, preventing the problem.

This patch requires an on-disk format change to add the
IVset guids to snapshots and bookmarks. As a result, this
patch has errata handling and a tunable to help affected
users resolve the issue with as little interruption as
possible.

Signed-off-by: Tom Caputi <tcaputi@datto.com>

19 files changed, 296 insertions, 38 deletions

diff --git a/usr/src/cmd/zpool/zpool_main.c b/usr/src/cmd/zpool/zpool_main.c
index 03caf7caf6..54d86cd041 100644
--- a/usr/src/cmd/zpool/zpool_main.c
+++ b/usr/src/cmd/zpool/zpool_main.c
@@ -1744,7 +1744,19 @@ print_import_config(const char *name, nvlist_t *nv, int namewidth, int depth)
 			(void) printf(gettext("all children offline"));
 			break;
 
-		default:
+		case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
+			(void) printf(gettext(" action: Existing "
+			    "encrypted datasets contain an on-disk "
+			    "incompatibility which\n\tmay cause "
+			    "on-disk corruption with 'zfs recv' and "
+			    "which needs to be\n\tcorrected. Enable "
+			    "the bookmark_v2 feature and backup "
+			    "these datasets to new encrypted "
+			    "datasets\n\tand destroy the "
+			    "old ones.\n"));
+			break;
+
+	   default:
 			(void) printf(gettext("corrupted data"));
 			break;
 		}
diff --git a/usr/src/common/zfs/zfeature_common.c b/usr/src/common/zfs/zfeature_common.c
index 49566ced2f..6f2bf7b114 100644
--- a/usr/src/common/zfs/zfeature_common.c
+++ b/usr/src/common/zfs/zfeature_common.c
@@ -327,6 +327,7 @@ zpool_feature_init(void)
 
 	static const spa_feature_t encryption_deps[] = {
 		SPA_FEATURE_EXTENSIBLE_DATASET,
+		SPA_FEATURE_BOOKMARK_V2,
 		SPA_FEATURE_NONE
 	};
 	zfeature_register(SPA_FEATURE_ENCRYPTION,
diff --git a/usr/src/common/zfs/zfs_prop.c b/usr/src/common/zfs/zfs_prop.c
index 291b5b3d02..27c730a5ac 100644
--- a/usr/src/common/zfs/zfs_prop.c
+++ b/usr/src/common/zfs/zfs_prop.c
@@ -507,6 +507,9 @@ zfs_prop_init(void)
 	    PROP_READONLY, ZFS_TYPE_DATASET, "OBJSETID");
 	zprop_register_hidden(ZFS_PROP_INCONSISTENT, "inconsistent",
 	    PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, "INCONSISTENT");
+	zprop_register_hidden(ZFS_PROP_IVSET_GUID, "ivsetguid",
+	    PROP_TYPE_NUMBER, PROP_READONLY,
+	    ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "IVSETGUID");
 	zprop_register_hidden(ZFS_PROP_PREV_SNAP, "prevsnap", PROP_TYPE_STRING,
 	    PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "PREVSNAP");
 	zprop_register_hidden(ZFS_PROP_PBKDF2_SALT, "pbkdf2salt",
diff --git a/usr/src/lib/libzfs/common/libzfs_iter.c b/usr/src/lib/libzfs/common/libzfs_iter.c
index 59eb50fef0..679926f228 100644
--- a/usr/src/lib/libzfs/common/libzfs_iter.c
+++ b/usr/src/lib/libzfs/common/libzfs_iter.c
@@ -202,6 +202,7 @@ zfs_iter_bookmarks(zfs_handle_t *zhp, zfs_iter_f func, void *data)
 	fnvlist_add_boolean(props, zfs_prop_to_name(ZFS_PROP_GUID));
 	fnvlist_add_boolean(props, zfs_prop_to_name(ZFS_PROP_CREATETXG));
 	fnvlist_add_boolean(props, zfs_prop_to_name(ZFS_PROP_CREATION));
+	fnvlist_add_boolean(props, zfs_prop_to_name(ZFS_PROP_IVSET_GUID));
 
 	if ((err = lzc_get_bookmarks(zhp->zfs_name, props, &bmarks)) != 0)
 		goto out;
diff --git a/usr/src/lib/libzfs/common/libzfs_sendrecv.c b/usr/src/lib/libzfs/common/libzfs_sendrecv.c
index 33d3922358..98088168c5 100644
--- a/usr/src/lib/libzfs/common/libzfs_sendrecv.c
+++ b/usr/src/lib/libzfs/common/libzfs_sendrecv.c
@@ -3980,6 +3980,20 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			    "destination %s space quota exceeded"), zc.zc_name);
 			(void) zfs_error(hdl, EZFS_NOSPC, errbuf);
 			break;
+		case ZFS_ERR_FROM_IVSET_GUID_MISSING:
+			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+			    "IV set guid missing. See errata %u at"
+			    "http://zfsonlinux.org/msg/ZFS-8000-ER"),
+			    ZPOOL_ERRATA_ZOL_8308_ENCRYPTION);
+			(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
+			break;
+		case ZFS_ERR_FROM_IVSET_GUID_MISMATCH:
+			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+			    "IV set guid mismatch. See the 'zfs receive' "
+			    "man page section\n discussing the limitations "
+			    "of raw encrypted send streams."));
+			(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
+			break;
 		default:
 			(void) zfs_standard_error(hdl, ioctl_errno, errbuf);
 		}
diff --git a/usr/src/lib/libzfs_core/common/libzfs_core.c b/usr/src/lib/libzfs_core/common/libzfs_core.c
index 9c46af2434..e8ecbe6caf 100644
--- a/usr/src/lib/libzfs_core/common/libzfs_core.c
+++ b/usr/src/lib/libzfs_core/common/libzfs_core.c
@@ -889,6 +889,7 @@ lzc_bookmark(nvlist_t *bookmarks, nvlist_t **errlist)
  * "guid" - globally unique identifier of the snapshot it refers to
  * "createtxg" - txg when the snapshot it refers to was created
  * "creation" - timestamp when the snapshot it refers to was created
+ * "ivsetguid" - IVset guid for identifying encrypted snapshots
  *
  * The format of the returned nvlist as follows:
  * <short name of bookmark> -> {
diff --git a/usr/src/man/man1m/zfs.1m b/usr/src/man/man1m/zfs.1m
index c64e2f5dea..02ebca0ba2 100644
--- a/usr/src/man/man1m/zfs.1m
+++ b/usr/src/man/man1m/zfs.1m
@@ -3287,6 +3287,31 @@ inheritance or by specifying encryption parameters with the
 .Fl o
 options.
 .Pp
+The added security provided by raw sends adds some restrictions to the send
+and receive process. ZFS will not allow a mix of raw receives and non-raw
+receives. Specifically, any raw incremental receives that are attempted after
+a non-raw receive will fail. Non-raw receives do not have this restriction and,
+therefore, are always possible. Because of this, it is best practice to always
+use either raw sends for their security benefits or non-raw sends for their
+flexibility when working with encrypted datasets, but not a combination.
+.Pp
+The reason for this restriction stems from the inherent restrictions of the
+AEAD ciphers that ZFS uses to encrypt data. When using ZFS native encryption,
+each block of data is encrypted against a randomly generated number known as
+the "initialization vector" (IV), which is stored in the filesystem metadata.
+This number is required by the encryption algorithms whenever the data is to
+be decrypted. Together, all of the IVs provided for all of the blocks in a
+given snapshot are collectively called an "IV set". When ZFS performs a raw
+send, the IV set is transferred from the source to the destination in the send
+stream. When ZFS performs a non-raw send, the data is decrypted by the source
+system and re-encrypted by the destination system, creating a snapshot with
+effectively the same data, but a different IV set. In order for decryption to
+work after a raw send, ZFS must ensure that the IV set used on both the source
+and destination side match. When an incremental raw receive is performed on
+top of an existing snapshot, ZFS will check to confirm that the "from"
+snapshot on both the source and destination were using the same IV set,
+ensuring the new IV set is consistent.
+.Pp
 The name of the snapshot
 .Pq and file system, if a full stream is received
 that this subcommand creates depends on the argument type and the use of the
diff --git a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg
index 3336a0f8c4..daa714c1fa 100644
--- a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg
+++ b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg
@@ -77,4 +77,5 @@ typeset -a properties=(
     "feature@zpool_checkpoint"
     "feature@spacemap_v2"
     "feature@encryption"
+    "feature@bookmark_v2"
 )
diff --git a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_errata3.ksh b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_errata3.ksh
index fa431a166c..6fd172dd8e 100755
--- a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_errata3.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_errata3.ksh
@@ -95,5 +95,5 @@ log_must zfs destroy -r $POOL_NAME/testvol
 
 log_must zpool export $POOL_NAME
 log_must zpool import -d /$TESTPOOL/ $POOL_NAME
-log_mustnot eval "zpool status | grep -q Errata"
+log_mustnot eval "zpool status | grep -q 'Errata #3'"
 log_pass "Errata 3 is properly handled"
diff --git a/usr/src/uts/common/fs/zfs/dmu_send.c b/usr/src/uts/common/fs/zfs/dmu_send.c
index ef3c43b87d..db5e2e01ef 100644
--- a/usr/src/uts/common/fs/zfs/dmu_send.c
+++ b/usr/src/uts/common/fs/zfs/dmu_send.c
@@ -1073,9 +1073,13 @@ dmu_send_impl(void *tag, dsl_pool_t *dp, dsl_dataset_t *to_ds,
 		}
 
 		if (featureflags & DMU_BACKUP_FEATURE_RAW) {
+			uint64_t ivset_guid = (ancestor_zb != NULL) ?
+			    ancestor_zb->zbm_ivset_guid : 0;
+
 			ASSERT(os->os_encrypted);
 
-			err = dsl_crypto_populate_key_nvlist(to_ds, &keynvl);
+			err = dsl_crypto_populate_key_nvlist(to_ds,
+			    ivset_guid, &keynvl);
 			if (err != 0) {
 				fnvlist_free(nvl);
 				goto out;
@@ -1189,7 +1193,7 @@ dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
 	}
 
 	if (fromsnap != 0) {
-		zfs_bookmark_phys_t zb;
+		zfs_bookmark_phys_t zb = { 0 };
 		boolean_t is_clone;
 
 		err = dsl_dataset_hold_obj(dp, fromsnap, FTAG, &fromds);
@@ -1198,12 +1202,25 @@ dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
 			dsl_pool_rele(dp, FTAG);
 			return (err);
 		}
-		if (!dsl_dataset_is_before(ds, fromds, 0))
+		if (!dsl_dataset_is_before(ds, fromds, 0)) {
 			err = SET_ERROR(EXDEV);
+			dsl_dataset_rele(fromds, FTAG);
+			dsl_dataset_rele_flags(ds, dsflags, FTAG);
+			dsl_pool_rele(dp, FTAG);
+			return (err);
+		}
+
 		zb.zbm_creation_time =
 		    dsl_dataset_phys(fromds)->ds_creation_time;
 		zb.zbm_creation_txg = dsl_dataset_phys(fromds)->ds_creation_txg;
 		zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
+
+		if (dsl_dataset_is_zapified(fromds)) {
+			(void) zap_lookup(dp->dp_meta_objset,
+			    fromds->ds_object, DS_FIELD_IVSET_GUID, 8, 1,
+			    &zb.zbm_ivset_guid);
+		}
+
 		is_clone = (fromds->ds_dir != ds->ds_dir);
 		dsl_dataset_rele(fromds, FTAG);
 		err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone,
@@ -1253,7 +1270,7 @@ dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
 	}
 
 	if (fromsnap != NULL) {
-		zfs_bookmark_phys_t zb;
+		zfs_bookmark_phys_t zb = { 0 };
 		boolean_t is_clone = B_FALSE;
 		int fsnamelen = strchr(tosnap, '@') - tosnap;
 
@@ -1279,6 +1296,13 @@ dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
 				    dsl_dataset_phys(fromds)->ds_creation_txg;
 				zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
 				is_clone = (ds->ds_dir != fromds->ds_dir);
+
+				if (dsl_dataset_is_zapified(fromds)) {
+					(void) zap_lookup(dp->dp_meta_objset,
+					    fromds->ds_object,
+					    DS_FIELD_IVSET_GUID, 8, 1,
+					    &zb.zbm_ivset_guid);
+				}
 				dsl_dataset_rele(fromds, FTAG);
 			}
 		} else {
@@ -1476,7 +1500,6 @@ typedef struct dmu_recv_begin_arg {
 	dmu_recv_cookie_t *drba_cookie;
 	cred_t *drba_cred;
 	dsl_crypto_params_t *drba_dcp;
-	uint64_t drba_snapobj;
 } dmu_recv_begin_arg_t;
 
 static int
@@ -1524,7 +1547,7 @@ recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds,
 		dsl_dataset_t *snap;
 		uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
 
-		/* Can't perform a raw receive on top of a non-raw receive */
+		/* Can't raw receive on top of an unencrypted dataset */
 		if (!encrypted && raw)
 			return (SET_ERROR(EINVAL));
 
@@ -1551,7 +1574,7 @@ recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds,
 			return (SET_ERROR(ENODEV));
 
 		if (drba->drba_cookie->drc_force) {
-			drba->drba_snapobj = obj;
+			drba->drba_cookie->drc_fromsnapobj = obj;
 		} else {
 			/*
 			 * If we are not forcing, there must be no
@@ -1561,7 +1584,8 @@ recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds,
 				dsl_dataset_rele(snap, FTAG);
 				return (SET_ERROR(ETXTBSY));
 			}
-			drba->drba_snapobj = ds->ds_prev->ds_object;
+			drba->drba_cookie->drc_fromsnapobj =
+			    ds->ds_prev->ds_object;
 		}
 
 		dsl_dataset_rele(snap, FTAG);
@@ -1596,7 +1620,7 @@ recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds,
 				return (SET_ERROR(EINVAL));
 		}
 
-		drba->drba_snapobj = 0;
+		drba->drba_cookie->drc_fromsnapobj = 0;
 	}
 
 	return (0);
@@ -1820,7 +1844,7 @@ dmu_recv_begin_sync(void *arg, dmu_tx_t *tx)
 	 * the raw cmd set. Raw incremental recvs do not use a dcp
 	 * since the encryption parameters are already set in stone.
 	 */
-	if (dcp == NULL && drba->drba_snapobj == 0 &&
+	if (dcp == NULL && drba->drba_cookie->drc_fromsnapobj == 0 &&
 	    drba->drba_origin == NULL) {
 		ASSERT3P(dcp, ==, NULL);
 		dcp = &dummy_dcp;
@@ -1834,15 +1858,15 @@ dmu_recv_begin_sync(void *arg, dmu_tx_t *tx)
 		/* create temporary clone */
 		dsl_dataset_t *snap = NULL;
 
-		if (drba->drba_snapobj != 0) {
+		if (drba->drba_cookie->drc_fromsnapobj != 0) {
 			VERIFY0(dsl_dataset_hold_obj(dp,
-			    drba->drba_snapobj, FTAG, &snap));
+			    drba->drba_cookie->drc_fromsnapobj, FTAG, &snap));
 			ASSERT3P(dcp, ==, NULL);
 		}
 
 		dsobj = dsl_dataset_create_sync(ds->ds_dir, recv_clone_name,
 		    snap, crflags, drba->drba_cred, dcp, tx);
-		if (drba->drba_snapobj != 0)
+		if (drba->drba_cookie->drc_fromsnapobj != 0)
 			dsl_dataset_rele(snap, FTAG);
 		dsl_dataset_rele_flags(ds, dsflags, FTAG);
 	} else {
@@ -3734,11 +3758,15 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp,
 		 * the keynvl away until then.
 		 */
 		err = dsl_crypto_recv_raw(spa_name(ra.os->os_spa),
-		    drc->drc_ds->ds_object, drc->drc_drrb->drr_type,
-		    keynvl, drc->drc_newfs);
+		    drc->drc_ds->ds_object, drc->drc_fromsnapobj,
+		    drc->drc_drrb->drr_type, keynvl, drc->drc_newfs);
 		if (err != 0)
 			goto out;
 
+		/* see comment in dmu_recv_end_sync() */
+		(void) nvlist_lookup_uint64(keynvl, "to_ivset_guid",
+		    &drc->drc_ivset_guid);
+
 		if (!drc->drc_newfs)
 			drc->drc_keynvl = fnvlist_dup(keynvl);
 	}
@@ -4054,6 +4082,23 @@ dmu_recv_end_sync(void *arg, dmu_tx_t *tx)
 	}
 
 	/*
+	 * If this is a raw receive, the crypt_keydata nvlist will include
+	 * a to_ivset_guid for us to set on the new snapshot. This value
+	 * will override the value generated by the snapshot code. Older
+	 * implementations of the raw send code did not include this
+	 * value. By setting the zfs_disable_ivset_guid_check tunable, the
+	 * user can still receive these streams, using the generated value
+	 * if one hasn't been provided.
+	 */
+	if (drc->drc_raw && drc->drc_ivset_guid != 0) {
+		dmu_object_zapify(dp->dp_meta_objset, drc->drc_newsnapobj,
+		    DMU_OT_DSL_DATASET, tx);
+		VERIFY0(zap_update(dp->dp_meta_objset, drc->drc_newsnapobj,
+		    DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1,
+		    &drc->drc_ivset_guid, tx));
+	}
+
+	/*
 	 * Release the hold from dmu_recv_begin.  This must be done before
 	 * we return to open context, so that when we free the dataset's dnode
 	 * we can evict its bonus buffer. Since the dataset may be destroyed
diff --git a/usr/src/uts/common/fs/zfs/dsl_bookmark.c b/usr/src/uts/common/fs/zfs/dsl_bookmark.c
index 08e8355414..a32198402f 100644
--- a/usr/src/uts/common/fs/zfs/dsl_bookmark.c
+++ b/usr/src/uts/common/fs/zfs/dsl_bookmark.c
@@ -221,6 +221,26 @@ dsl_bookmark_create_sync(void *arg, dmu_tx_t *tx)
 		bmark_phys.zbm_creation_time =
 		    dsl_dataset_phys(snapds)->ds_creation_time;
 
+		/*
+		 * If the dataset is encrypted create a larger bookmark to
+		 * accommodate the IVset guid. The IVset guid was added
+		 * after the encryption feature to prevent a problem with
+		 * raw sends. If we encounter an encrypted dataset without
+		 * an IVset guid we fall back to a normal bookmark.
+		 */
+		if (snapds->ds_dir->dd_crypto_obj != 0 &&
+		    spa_feature_is_enabled(dp->dp_spa,
+		    SPA_FEATURE_BOOKMARK_V2)) {
+			int err = zap_lookup(mos, snapds->ds_object,
+			    DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1,
+			    &bmark_phys.zbm_ivset_guid);
+			if (err == 0) {
+				bmark_len = BOOKMARK_PHYS_SIZE_V2;
+				spa_feature_incr(dp->dp_spa,
+				    SPA_FEATURE_BOOKMARK_V2, tx);
+			}
+		}
+
 		VERIFY0(zap_add(mos, bmark_fs->ds_bookmarks,
 		    shortname, sizeof (uint64_t),
 		    bmark_len / sizeof (uint64_t), &bmark_phys, tx));
@@ -273,7 +293,7 @@ dsl_get_bookmarks_impl(dsl_dataset_t *ds, nvlist_t *props, nvlist_t *outnvl)
 	    zap_cursor_retrieve(&zc, &attr) == 0;
 	    zap_cursor_advance(&zc)) {
 		char *bmark_name = attr.za_name;
-		zfs_bookmark_phys_t bmark_phys;
+		zfs_bookmark_phys_t bmark_phys = { 0 };
 
 		err = dsl_dataset_bmark_lookup(ds, bmark_name, &bmark_phys);
 		ASSERT3U(err, !=, ENOENT);
@@ -296,6 +316,11 @@ dsl_get_bookmarks_impl(dsl_dataset_t *ds, nvlist_t *props, nvlist_t *outnvl)
 			dsl_prop_nvlist_add_uint64(out_props,
 			    ZFS_PROP_CREATION, bmark_phys.zbm_creation_time);
 		}
+		if (nvlist_exists(props,
+		    zfs_prop_to_name(ZFS_PROP_IVSET_GUID))) {
+			dsl_prop_nvlist_add_uint64(out_props,
+			    ZFS_PROP_IVSET_GUID, bmark_phys.zbm_ivset_guid);
+		}
 
 		fnvlist_add_nvlist(outnvl, bmark_name, out_props);
 		fnvlist_free(out_props);
@@ -343,13 +368,26 @@ typedef struct dsl_bookmark_destroy_arg {
 static int
 dsl_dataset_bookmark_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx)
 {
+	int err;
 	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
 	uint64_t bmark_zapobj = ds->ds_bookmarks;
 	matchtype_t mt = 0;
+	uint64_t int_size, num_ints;
 
 	if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET)
 		mt = MT_NORMALIZE;
 
+	err = zap_length(mos, bmark_zapobj, name, &int_size, &num_ints);
+	if (err != 0)
+		return (err);
+
+	ASSERT3U(int_size, ==, sizeof (uint64_t));
+
+	if (num_ints * int_size > BOOKMARK_PHYS_SIZE_V1) {
+		spa_feature_decr(dmu_objset_spa(mos),
+		    SPA_FEATURE_BOOKMARK_V2, tx);
+	}
+
 	return (zap_remove_norm(mos, bmark_zapobj, name, mt, tx));
 }
 
diff --git a/usr/src/uts/common/fs/zfs/dsl_crypt.c b/usr/src/uts/common/fs/zfs/dsl_crypt.c
index 6ef478e956..a0544a7ad2 100644
--- a/usr/src/uts/common/fs/zfs/dsl_crypt.c
+++ b/usr/src/uts/common/fs/zfs/dsl_crypt.c
@@ -71,6 +71,13 @@
  * object is also refcounted.
  */
 
+/*
+ * This tunable allows datasets to be raw received even if the stream does
+ * not include IVset guids or if the guids don't match. This is used as part
+ * of the resolution for ZPOOL_ERRATA_ZOL_8308_ENCRYPTION.
+ */
+int zfs_disable_ivset_guid_check = 0;
+
 static void
 dsl_wrapping_key_hold(dsl_wrapping_key_t *wkey, void *tag)
 {
@@ -1941,21 +1948,23 @@ dsl_dataset_create_crypt_sync(uint64_t dsobj, dsl_dir_t *dd,
 
 typedef struct dsl_crypto_recv_key_arg {
 	uint64_t dcrka_dsobj;
+	uint64_t dcrka_fromobj;
 	dmu_objset_type_t dcrka_ostype;
 	nvlist_t *dcrka_nvl;
 	boolean_t dcrka_do_key;
 } dsl_crypto_recv_key_arg_t;
 
 static int
-dsl_crypto_recv_raw_objset_check(dsl_dataset_t *ds, dmu_objset_type_t ostype,
-    nvlist_t *nvl, dmu_tx_t *tx)
+dsl_crypto_recv_raw_objset_check(dsl_dataset_t *ds, dsl_dataset_t *fromds,
+    dmu_objset_type_t ostype, nvlist_t *nvl, dmu_tx_t *tx)
 {
 	int ret;
 	objset_t *os;
 	dnode_t *mdn;
 	uint8_t *buf = NULL;
 	uint_t len;
-	uint64_t intval, nlevels, blksz, ibs, nblkptr, maxblkid;
+	uint64_t intval, nlevels, blksz, ibs;
+	uint64_t nblkptr, maxblkid, from_ivset_guid = 0;
 
 	if (ostype != DMU_OST_ZFS && ostype != DMU_OST_ZVOL)
 		return (SET_ERROR(EINVAL));
@@ -2021,6 +2030,29 @@ dsl_crypto_recv_raw_objset_check(dsl_dataset_t *ds, dmu_objset_type_t ostype,
 	}
 	rrw_exit(&ds->ds_bp_rwlock, FTAG);
 
+	/*
+	 * Check that the ivset guid of the fromds matches the one from the
+	 * send stream. Older versions of the encryption code did not have
+	 * an ivset guid on the from dataset and did not send one in the
+	 * stream. For these streams we provide the
+	 * zfs_disable_ivset_guid_check tunable to allow these datasets to
+	 * be received with a generated ivset guid.
+	 */
+	if (fromds != NULL) {
+		intval = 0;
+		(void) nvlist_lookup_uint64(nvl, "from_ivset_guid", &intval);
+		(void) zap_lookup(tx->tx_pool->dp_meta_objset,
+		    fromds->ds_object, DS_FIELD_IVSET_GUID,
+		    sizeof (from_ivset_guid), 1, &from_ivset_guid);
+
+		if (!zfs_disable_ivset_guid_check &&
+		    (intval == 0 || from_ivset_guid == 0))
+			return (SET_ERROR(ZFS_ERR_FROM_IVSET_GUID_MISSING));
+
+		if (!zfs_disable_ivset_guid_check && intval != from_ivset_guid)
+			return (SET_ERROR(ZFS_ERR_FROM_IVSET_GUID_MISMATCH));
+	}
+
 	return (0);
 }
 
@@ -2040,7 +2072,11 @@ dsl_crypto_recv_raw_objset_sync(dsl_dataset_t *ds, dmu_objset_type_t ostype,
 	VERIFY0(dmu_objset_from_ds(ds, &os));
 	mdn = DMU_META_DNODE(os);
 
-	/* fetch the values we need from the nvlist */
+	/*
+	 * Fetch the values we need from the nvlist. "to_ivset_guid" must
+	 * be set on the snapshot, which doesn't exist yet. The receive
+	 * code will take care of this for us later.
+	 */
 	compress = fnvlist_lookup_uint64(nvl, "mdn_compress");
 	checksum = fnvlist_lookup_uint64(nvl, "mdn_checksum");
 	nlevels = fnvlist_lookup_uint64(nvl, "mdn_nlevels");
@@ -2104,7 +2140,7 @@ dsl_crypto_recv_raw_key_check(dsl_dataset_t *ds, nvlist_t *nvl, dmu_tx_t *tx)
 	objset_t *mos = tx->tx_pool->dp_meta_objset;
 	uint8_t *buf = NULL;
 	uint_t len;
-	uint64_t intval, guid, version;
+	uint64_t intval, key_guid, version;
 	boolean_t is_passphrase = B_FALSE;
 
 	ASSERT(dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT);
@@ -2129,10 +2165,10 @@ dsl_crypto_recv_raw_key_check(dsl_dataset_t *ds, nvlist_t *nvl, dmu_tx_t *tx)
 	 */
 	if (ds->ds_dir->dd_crypto_obj != 0) {
 		ret = zap_lookup(mos, ds->ds_dir->dd_crypto_obj,
-		    DSL_CRYPTO_KEY_GUID, 8, 1, &guid);
+		    DSL_CRYPTO_KEY_GUID, 8, 1, &key_guid);
 		if (ret != 0)
 			return (ret);
-		if (intval != guid)
+		if (intval != key_guid)
 			return (SET_ERROR(EACCES));
 	}
 
@@ -2198,13 +2234,13 @@ dsl_crypto_recv_raw_key_sync(dsl_dataset_t *ds, nvlist_t *nvl, dmu_tx_t *tx)
 	uint_t len;
 	uint64_t rddobj, one = 1;
 	uint8_t *keydata, *hmac_keydata, *iv, *mac;
-	uint64_t crypt, guid, keyformat, iters, salt;
+	uint64_t crypt, key_guid, keyformat, iters, salt;
 	uint64_t version = ZIO_CRYPT_KEY_CURRENT_VERSION;
 	char *keylocation = "prompt";
 
 	/* lookup the values we need to create the DSL Crypto Key */
 	crypt = fnvlist_lookup_uint64(nvl, DSL_CRYPTO_KEY_CRYPTO_SUITE);
-	guid = fnvlist_lookup_uint64(nvl, DSL_CRYPTO_KEY_GUID);
+	key_guid = fnvlist_lookup_uint64(nvl, DSL_CRYPTO_KEY_GUID);
 	keyformat = fnvlist_lookup_uint64(nvl,
 	    zfs_prop_to_name(ZFS_PROP_KEYFORMAT));
 	iters = fnvlist_lookup_uint64(nvl,
@@ -2259,7 +2295,7 @@ dsl_crypto_recv_raw_key_sync(dsl_dataset_t *ds, nvlist_t *nvl, dmu_tx_t *tx)
 
 	/* sync the key data to the ZAP object on disk */
 	dsl_crypto_key_sync_impl(mos, dd->dd_crypto_obj, crypt,
-	    rddobj, guid, iv, mac, keydata, hmac_keydata, keyformat, salt,
+	    rddobj, key_guid, iv, mac, keydata, hmac_keydata, keyformat, salt,
 	    iters, tx);
 }
 
@@ -2268,14 +2304,21 @@ dsl_crypto_recv_key_check(void *arg, dmu_tx_t *tx)
 {
 	int ret;
 	dsl_crypto_recv_key_arg_t *dcrka = arg;
-	dsl_dataset_t *ds = NULL;
+	dsl_dataset_t *ds = NULL, *fromds = NULL;
 
 	ret = dsl_dataset_hold_obj(tx->tx_pool, dcrka->dcrka_dsobj,
 	    FTAG, &ds);
 	if (ret != 0)
 		goto error;
 
-	ret = dsl_crypto_recv_raw_objset_check(ds,
+	if (dcrka->dcrka_fromobj != 0) {
+		ret = dsl_dataset_hold_obj(tx->tx_pool, dcrka->dcrka_fromobj,
+		    FTAG, &fromds);
+		if (ret != 0)
+			goto error;
+	}
+
+	ret = dsl_crypto_recv_raw_objset_check(ds, fromds,
 	    dcrka->dcrka_ostype, dcrka->dcrka_nvl, tx);
 	if (ret != 0)
 		goto error;
@@ -2290,11 +2333,16 @@ dsl_crypto_recv_key_check(void *arg, dmu_tx_t *tx)
 		goto error;
 
 	dsl_dataset_rele(ds, FTAG);
+	if (fromds != NULL)
+		dsl_dataset_rele(fromds, FTAG);
+
 	return (0);
 
 error:
 	if (ds != NULL)
 		dsl_dataset_rele(ds, FTAG);
+	if (fromds != NULL)
+		dsl_dataset_rele(fromds, FTAG);
 	return (ret);
 }
 
@@ -2319,12 +2367,13 @@ dsl_crypto_recv_key_sync(void *arg, dmu_tx_t *tx)
  * without wrapping it.
  */
 int
-dsl_crypto_recv_raw(const char *poolname, uint64_t dsobj,
+dsl_crypto_recv_raw(const char *poolname, uint64_t dsobj, uint64_t fromobj,
     dmu_objset_type_t ostype, nvlist_t *nvl, boolean_t do_key)
 {
 	dsl_crypto_recv_key_arg_t dcrka;
 
 	dcrka.dcrka_dsobj = dsobj;
+	dcrka.dcrka_fromobj = fromobj;
 	dcrka.dcrka_ostype = ostype;
 	dcrka.dcrka_nvl = nvl;
 	dcrka.dcrka_do_key = do_key;
@@ -2334,7 +2383,8 @@ dsl_crypto_recv_raw(const char *poolname, uint64_t dsobj,
 }
 
 int
-dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, nvlist_t **nvl_out)
+dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, uint64_t from_ivset_guid,
+    nvlist_t **nvl_out)
 {
 	int ret;
 	objset_t *os;
@@ -2345,8 +2395,9 @@ dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, nvlist_t **nvl_out)
 	dsl_dir_t *rdd = NULL;
 	dsl_pool_t *dp = ds->ds_dir->dd_pool;
 	objset_t *mos = dp->dp_meta_objset;
-	uint64_t crypt = 0, guid = 0, format = 0;
+	uint64_t crypt = 0, key_guid = 0, format = 0;
 	uint64_t iters = 0, salt = 0, version = 0;
+	uint64_t to_ivset_guid = 0;
 	uint8_t raw_keydata[MASTER_KEY_MAX_LEN];
 	uint8_t raw_hmac_keydata[SHA512_HMAC_KEYLEN];
 	uint8_t iv[WRAPPING_IV_LEN];
@@ -2367,7 +2418,7 @@ dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, nvlist_t **nvl_out)
 	if (ret != 0)
 		goto error;
 
-	ret = zap_lookup(mos, dckobj, DSL_CRYPTO_KEY_GUID, 8, 1, &guid);
+	ret = zap_lookup(mos, dckobj, DSL_CRYPTO_KEY_GUID, 8, 1, &key_guid);
 	if (ret != 0)
 		goto error;
 
@@ -2391,6 +2442,12 @@ dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, nvlist_t **nvl_out)
 	if (ret != 0)
 		goto error;
 
+	/* see zfs_disable_ivset_guid_check tunable for errata info */
+	ret = zap_lookup(mos, ds->ds_object, DS_FIELD_IVSET_GUID, 8, 1,
+	    &to_ivset_guid);
+	if (ret != 0)
+		ASSERT3U(dp->dp_spa->spa_errata, !=, 0);
+
 	/*
 	 * We don't support raw sends of legacy on-disk formats. See the
 	 * comment in dsl_crypto_recv_key_check() for details.
@@ -2440,7 +2497,7 @@ dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, nvlist_t **nvl_out)
 	dsl_pool_config_exit(dp, FTAG);
 
 	fnvlist_add_uint64(nvl, DSL_CRYPTO_KEY_CRYPTO_SUITE, crypt);
-	fnvlist_add_uint64(nvl, DSL_CRYPTO_KEY_GUID, guid);
+	fnvlist_add_uint64(nvl, DSL_CRYPTO_KEY_GUID, key_guid);
 	fnvlist_add_uint64(nvl, DSL_CRYPTO_KEY_VERSION, version);
 	VERIFY0(nvlist_add_uint8_array(nvl, DSL_CRYPTO_KEY_MASTER_KEY,
 	    raw_keydata, MASTER_KEY_MAX_LEN));
@@ -2462,6 +2519,8 @@ dsl_crypto_populate_key_nvlist(dsl_dataset_t *ds, nvlist_t **nvl_out)
 	fnvlist_add_uint64(nvl, "mdn_indblkshift", mdn->dn_indblkshift);
 	fnvlist_add_uint64(nvl, "mdn_nblkptr", mdn->dn_nblkptr);
 	fnvlist_add_uint64(nvl, "mdn_maxblkid", mdn->dn_maxblkid);
+	fnvlist_add_uint64(nvl, "to_ivset_guid", to_ivset_guid);
+	fnvlist_add_uint64(nvl, "from_ivset_guid", from_ivset_guid);
 
 	*nvl_out = nvl;
 	return (0);
@@ -2572,6 +2631,10 @@ dsl_dataset_crypt_stats(dsl_dataset_t *ds, nvlist_t *nv)
 	    zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 8, 1, &intval) == 0) {
 		dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_PBKDF2_ITERS, intval);
 	}
+	if (zap_lookup(dd->dd_pool->dp_meta_objset, ds->ds_object,
+	    DS_FIELD_IVSET_GUID, 8, 1, &intval) == 0) {
+		dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_IVSET_GUID, intval);
+	}
 
 	if (dsl_dir_get_encryption_root_ddobj(dd, &intval) == 0) {
 		VERIFY0(dsl_dir_hold_obj(dd->dd_pool, intval, NULL, FTAG,
@@ -2807,3 +2870,9 @@ error:
 
 	return (ret);
 }
+
+#if defined(_KERNEL)
+module_param(zfs_disable_ivset_guid_check, int, 0644);
+MODULE_PARM_DESC(zfs_disable_ivset_guid_check,
+	"Set to allow raw receives without IVset guids");
+#endif
diff --git a/usr/src/uts/common/fs/zfs/dsl_dataset.c b/usr/src/uts/common/fs/zfs/dsl_dataset.c
index e9fdaef927..8786a849e4 100644
--- a/usr/src/uts/common/fs/zfs/dsl_dataset.c
+++ b/usr/src/uts/common/fs/zfs/dsl_dataset.c
@@ -540,6 +540,13 @@ dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag,
 			ds->ds_reserved = ds->ds_quota = 0;
 		}
 
+		if (err == 0 && ds->ds_dir->dd_crypto_obj != 0 &&
+		    ds->ds_is_snapshot &&
+		    zap_contains(mos, dsobj, DS_FIELD_IVSET_GUID) != 0) {
+			dp->dp_spa->spa_errata =
+			    ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
+		}
+
 		dsl_deadlist_open(&ds->ds_deadlist,
 		    mos, dsl_dataset_phys(ds)->ds_deadlist_obj);
 		uint64_t remap_deadlist_obj =
@@ -1548,6 +1555,20 @@ dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname,
 		    sizeof (remap_deadlist_obj), 1, &remap_deadlist_obj, tx));
 	}
 
+	/*
+	 * Create a ivset guid for this snapshot if the dataset is
+	 * encrypted. This may be overridden by a raw receive. We
+	 * only do this if the bookmark_v2 feature is enabled.
+	 */
+	if (ds->ds_dir->dd_crypto_obj != 0 &&
+	    spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_BOOKMARK_V2)) {
+		uint64_t ivset_guid = unique_create();
+
+		dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
+		VERIFY0(zap_add(mos, dsobj, DS_FIELD_IVSET_GUID,
+		    sizeof (ivset_guid), 1, &ivset_guid, tx));
+	}
+
 	ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, <, tx->tx_txg);
 	dsl_dataset_phys(ds)->ds_prev_snap_obj = dsobj;
 	dsl_dataset_phys(ds)->ds_prev_snap_txg = crtxg;
diff --git a/usr/src/uts/common/fs/zfs/spa.c b/usr/src/uts/common/fs/zfs/spa.c
index 48a4b2167e..89afe2511d 100644
--- a/usr/src/uts/common/fs/zfs/spa.c
+++ b/usr/src/uts/common/fs/zfs/spa.c
@@ -3037,6 +3037,16 @@ spa_ld_check_features(spa_t *spa, boolean_t *missing_feat_writep)
 			return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
 	}
 
+	/*
+	 * Encryption was added before bookmark_v2, even though bookmark_v2
+	 * is now a dependency. If this pool has encryption enabled without
+	 * bookmark_v2, trigger an errata message.
+	 */
+	if (spa_feature_is_enabled(spa, SPA_FEATURE_ENCRYPTION) &&
+	    !spa_feature_is_enabled(spa, SPA_FEATURE_BOOKMARK_V2)) {
+		spa->spa_errata = ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
+	}
+
 	return (0);
 }
 
diff --git a/usr/src/uts/common/fs/zfs/sys/dmu_send.h b/usr/src/uts/common/fs/zfs/sys/dmu_send.h
index 391d27f2f6..7d850fe4b1 100644
--- a/usr/src/uts/common/fs/zfs/sys/dmu_send.h
+++ b/usr/src/uts/common/fs/zfs/sys/dmu_send.h
@@ -67,7 +67,9 @@ typedef struct dmu_recv_cookie {
 	struct avl_tree *drc_guid_to_ds_map;
 	nvlist_t *drc_keynvl;
 	zio_cksum_t drc_cksum;
+	uint64_t drc_fromsnapobj;
 	uint64_t drc_newsnapobj;
+	uint64_t drc_ivset_guid;
 	void *drc_owner;
 	cred_t *drc_cred;
 } dmu_recv_cookie_t;
diff --git a/usr/src/uts/common/fs/zfs/sys/dsl_crypt.h b/usr/src/uts/common/fs/zfs/sys/dsl_crypt.h
index ca448ead48..8f9e163582 100644
--- a/usr/src/uts/common/fs/zfs/sys/dsl_crypt.h
+++ b/usr/src/uts/common/fs/zfs/sys/dsl_crypt.h
@@ -189,12 +189,13 @@ void key_mapping_rele(spa_t *spa, dsl_key_mapping_t *km, void *tag);
 int spa_keystore_lookup_key(spa_t *spa, uint64_t dsobj, void *tag,
     dsl_crypto_key_t **dck_out);
 
-int dsl_crypto_populate_key_nvlist(struct dsl_dataset *ds, nvlist_t **nvl_out);
+int dsl_crypto_populate_key_nvlist(struct dsl_dataset *ds,
+    uint64_t from_ivset_guid, nvlist_t **nvl_out);
 int dsl_crypto_recv_raw_key_check(struct dsl_dataset *ds,
     nvlist_t *nvl, dmu_tx_t *tx);
 void dsl_crypto_recv_raw_key_sync(struct dsl_dataset *ds,
     nvlist_t *nvl, dmu_tx_t *tx);
-int dsl_crypto_recv_raw(const char *poolname, uint64_t dsobj,
+int dsl_crypto_recv_raw(const char *poolname, uint64_t dsobj, uint64_t fromobj,
     dmu_objset_type_t ostype, nvlist_t *nvl, boolean_t do_key);
 
 int spa_keystore_change_key(const char *dsname, dsl_crypto_params_t *dcp);
diff --git a/usr/src/uts/common/fs/zfs/sys/dsl_dataset.h b/usr/src/uts/common/fs/zfs/sys/dsl_dataset.h
index 384d42c298..189376eefc 100644
--- a/usr/src/uts/common/fs/zfs/sys/dsl_dataset.h
+++ b/usr/src/uts/common/fs/zfs/sys/dsl_dataset.h
@@ -117,6 +117,12 @@ struct dsl_key_mapping;
 #define	DS_FIELD_REMAP_DEADLIST	"com.delphix:remap_deadlist"
 
 /*
+ * This field is set to the ivset guid for encrypted snapshots. This is used
+ * for validating raw receives.
+ */
+#define	DS_FIELD_IVSET_GUID	"com.datto:ivset_guid"
+
+/*
  * DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
  * name lookups should be performed case-insensitively.
  */
diff --git a/usr/src/uts/common/fs/zfs/zcp_get.c b/usr/src/uts/common/fs/zfs/zcp_get.c
index 85e00a0f93..19d330257b 100644
--- a/usr/src/uts/common/fs/zfs/zcp_get.c
+++ b/usr/src/uts/common/fs/zfs/zcp_get.c
@@ -422,6 +422,11 @@ get_special_prop(lua_State *state, dsl_dataset_t *ds, const char *dsname,
 	case ZFS_PROP_INCONSISTENT:
 		numval = dsl_get_inconsistent(ds);
 		break;
+	case ZFS_PROP_IVSET_GUID:
+		error = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset,
+		    ds->ds_object, DS_FIELD_IVSET_GUID, sizeof (numval),
+		    1, &numval);
+		break;
 	case ZFS_PROP_RECEIVE_RESUME_TOKEN: {
 		char *token = get_receive_resume_stats_impl(ds);
 		VERIFY3U(strlcpy(strval, token, ZAP_MAXVALUELEN), <,
diff --git a/usr/src/uts/common/sys/fs/zfs.h b/usr/src/uts/common/sys/fs/zfs.h
index e48e107cf7..4bea1ef40c 100644
--- a/usr/src/uts/common/sys/fs/zfs.h
+++ b/usr/src/uts/common/sys/fs/zfs.h
@@ -171,6 +171,7 @@ typedef enum {
 	ZFS_PROP_ENCRYPTION_ROOT,
 	ZFS_PROP_KEY_GUID,
 	ZFS_PROP_KEYSTATUS,
+	ZFS_PROP_IVSET_GUID,            /* not exposed to the user */
 	ZFS_NUM_PROPS
 } zfs_prop_t;
 
@@ -1041,7 +1042,9 @@ typedef enum {
 	ZFS_ERR_DISCARDING_CHECKPOINT,
 	ZFS_ERR_NO_CHECKPOINT,
 	ZFS_ERR_DEVRM_IN_PROGRESS,
-	ZFS_ERR_VDEV_TOO_BIG
+	ZFS_ERR_VDEV_TOO_BIG,
+	ZFS_ERR_FROM_IVSET_GUID_MISSING,
+	ZFS_ERR_FROM_IVSET_GUID_MISMATCH,
 } zfs_errno_t;
 
 /*