Implement zfs_ioc_recv_new() for OpenZFS 2605

Adds ZFS_IOC_RECV_NEW for resumable streams and preserves the legacy
ZFS_IOC_RECV user/kernel interface.  The new interface supports all
stream options but is currently only used for resumable streams.
This way updated user space utilities will interoperate with older
kernel modules.

ZFS_IOC_RECV_NEW is modeled after the existing ZFS_IOC_SEND_NEW
handler.  Non-Linux OpenZFS platforms have opted to change the
legacy interface in an incompatible fashion instead of adding a
new ioctl.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

include/libzfs_core.h

@@ -69,6 +69,9 @@ int lzc_receive_resumable(const char *, nvlist_t *, const char *,
     boolean_t, int);
 int lzc_receive_with_header(const char *, nvlist_t *, const char *, boolean_t,
     boolean_t, int, const struct dmu_replay_record *);
+int lzc_receive_one(const char *, nvlist_t *, const char *, boolean_t,
+    boolean_t, int, const struct dmu_replay_record *, int, uint64_t *,
+    uint64_t *, uint64_t *, nvlist_t **);
 
 boolean_t lzc_exists(const char *);
 

include/sys/fs/zfs.h

@@ -232,6 +232,7 @@ typedef enum {
 
 #define	ZPROP_SOURCE_VAL_RECVD	"$recvd"
 #define	ZPROP_N_MORE_ERRORS	"N_MORE_ERRORS"
+
 /*
  * Dataset flag implemented as a special entry in the props zap object
  * indicating that the dataset has received properties on or after
@@ -923,7 +924,7 @@ typedef struct ddt_histogram {
  */
 typedef enum zfs_ioc {
 	/*
-	 * Illumos - 70/128 numbers reserved.
+	 * Illumos - 71/128 numbers reserved.
 	 */
 	ZFS_IOC_FIRST =	('Z' << 8),
 	ZFS_IOC = ZFS_IOC_FIRST,
@@ -997,6 +998,7 @@ typedef enum zfs_ioc {
 	ZFS_IOC_BOOKMARK,
 	ZFS_IOC_GET_BOOKMARKS,
 	ZFS_IOC_DESTROY_BOOKMARKS,
+	ZFS_IOC_RECV_NEW,
 
 	/*
 	 * Linux - 3/64 numbers reserved.

include/sys/zfs_ioctl.h

@@ -389,15 +389,14 @@ typedef struct zfs_cmd {
 	uint64_t	zc_iflags;		/* internal to zfs(7fs) */
 	zfs_share_t	zc_share;
 	dmu_objset_stats_t zc_objset_stats;
-	dmu_replay_record_t zc_begin_record;
+	struct drr_begin zc_begin_record;
 	zinject_record_t zc_inject_record;
 	uint32_t	zc_defer_destroy;
 	uint32_t	zc_flags;
 	uint64_t	zc_action_handle;
 	int		zc_cleanup_fd;
 	uint8_t		zc_simple;
-	boolean_t	zc_resumable;
+	uint8_t		zc_pad[3];		/* alignment */
-	uint8_t		zc_pad[2];		/* alignment */
 	uint64_t	zc_sendobj;
 	uint64_t	zc_fromobj;
 	uint64_t	zc_createtxg;

lib/libzfs/libzfs_sendrecv.c

@@ -2960,24 +2960,31 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
     avl_tree_t *stream_avl, char **top_zfs, int cleanup_fd,
     uint64_t *action_handlep, const char *finalsnap)
 {
-	zfs_cmd_t zc = {"\0"};
 	time_t begin_time;
 	int ioctl_err, ioctl_errno, err;
 	char *cp;
 	struct drr_begin *drrb = &drr->drr_u.drr_begin;
 	char errbuf[1024];
-	char prop_errbuf[1024];
 	const char *chopprefix;
 	boolean_t newfs = B_FALSE;
 	boolean_t stream_wantsnewfs;
+	boolean_t newprops = B_FALSE;
+	uint64_t read_bytes = 0;
+	uint64_t errflags = 0;
 	uint64_t parent_snapguid = 0;
 	prop_changelist_t *clp = NULL;
 	nvlist_t *snapprops_nvlist = NULL;
 	zprop_errflags_t prop_errflags;
+	nvlist_t *prop_errors = NULL;
 	boolean_t recursive;
 	char *snapname = NULL;
+	char destsnap[MAXPATHLEN * 2];
+	char origin[MAXNAMELEN];
+	char name[MAXPATHLEN];
+	nvlist_t *props = NULL;
 
 	begin_time = time(NULL);
+	bzero(origin, MAXNAMELEN);
 
 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
 	    "cannot receive"));
@@ -2988,24 +2995,19 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 	if (stream_avl != NULL) {
 		nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid,
 		    &snapname);
-		nvlist_t *props;
-		int ret;
 
 		(void) nvlist_lookup_uint64(fs, "parentfromsnap",
 		    &parent_snapguid);
 		err = nvlist_lookup_nvlist(fs, "props", &props);
-		if (err)
+		if (err) {
 			VERIFY(0 == nvlist_alloc(&props, NV_UNIQUE_NAME, 0));
+			newprops = B_TRUE;
+		}
 
 		if (flags->canmountoff) {
 			VERIFY(0 == nvlist_add_uint64(props,
 			    zfs_prop_to_name(ZFS_PROP_CANMOUNT), 0));
 		}
-		ret = zcmd_write_src_nvlist(hdl, &zc, props);
-		if (err)
-			nvlist_free(props);
-		if (ret != 0)
-			return (-1);
 	}
 
 	cp = NULL;
@@ -3026,7 +3028,8 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		if (strchr(tosnap, '@')) {
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
 			    "argument - snapshot not allowed with -e"));
-			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
+			err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
+			goto out;
 		}
 
 		chopprefix = strrchr(sendfs, '/');
@@ -3053,7 +3056,8 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		if (strchr(tosnap, '@')) {
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
 			    "argument - snapshot not allowed with -d"));
-			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
+			err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
+			goto out;
 		}
 
 		chopprefix = strchr(drrb->drr_toname, '/');
@@ -3071,7 +3075,8 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "cannot specify snapshot name for multi-snapshot "
 			    "stream"));
-			return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
+			err = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
+			goto out;
 		}
 		chopprefix = drrb->drr_toname + strlen(drrb->drr_toname);
 	}
@@ -3083,35 +3088,35 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 	    chopprefix[0] == '\0');
 
 	/*
-	 * Determine name of destination snapshot, store in zc_value.
+	 * Determine name of destination snapshot.
 	 */
-	(void) strcpy(zc.zc_value, tosnap);
+	(void) strcpy(destsnap, tosnap);
-	(void) strlcat(zc.zc_value, chopprefix, sizeof (zc.zc_value));
+	(void) strlcat(destsnap, chopprefix, sizeof (destsnap));
 	free(cp);
-	if (!zfs_name_valid(zc.zc_value, ZFS_TYPE_SNAPSHOT)) {
+	if (!zfs_name_valid(destsnap, ZFS_TYPE_SNAPSHOT)) {
-		zcmd_free_nvlists(&zc);
+		err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
-		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
+		goto out;
 	}
 
 	/*
-	 * Determine the name of the origin snapshot, store in zc_string.
+	 * Determine the name of the origin snapshot.
 	 */
 	if (drrb->drr_flags & DRR_FLAG_CLONE) {
-		if (guid_to_name(hdl, zc.zc_value,
+		if (guid_to_name(hdl, destsnap,
-		    drrb->drr_fromguid, B_FALSE, zc.zc_string) != 0) {
+		    drrb->drr_fromguid, B_FALSE, origin) != 0) {
-			zcmd_free_nvlists(&zc);
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "local origin for clone %s does not exist"),
-			    zc.zc_value);
+			    destsnap);
-			return (zfs_error(hdl, EZFS_NOENT, errbuf));
+			err = zfs_error(hdl, EZFS_NOENT, errbuf);
+			goto out;
 		}
 		if (flags->verbose)
-			(void) printf("found clone origin %s\n", zc.zc_string);
+			(void) printf("found clone origin %s\n", origin);
 	} else if (originsnap) {
-		(void) strncpy(zc.zc_string, originsnap, ZFS_MAXNAMELEN);
+		(void) strncpy(origin, originsnap, ZFS_MAXNAMELEN);
 		if (flags->verbose)
 			(void) printf("using provided clone origin %s\n",
-			    zc.zc_string);
+			    origin);
 	}
 
 	boolean_t resuming = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
@@ -3127,18 +3132,18 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
 		    "cannot receive new filesystem stream"));
 
-		(void) strcpy(zc.zc_name, zc.zc_value);
+		(void) strcpy(name, destsnap);
-		cp = strrchr(zc.zc_name, '/');
+		cp = strrchr(name, '/');
 		if (cp)
 			*cp = '\0';
 		if (cp &&
-		    !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) {
+		    !zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
 			char suffix[ZFS_MAXNAMELEN];
-			(void) strcpy(suffix, strrchr(zc.zc_value, '/'));
+			(void) strcpy(suffix, strrchr(destsnap, '/'));
-			if (guid_to_name(hdl, zc.zc_name, parent_snapguid,
+			if (guid_to_name(hdl, name, parent_snapguid,
-			    B_FALSE, zc.zc_value) == 0) {
+			    B_FALSE, destsnap) == 0) {
-				*strchr(zc.zc_value, '@') = '\0';
+				*strchr(destsnap, '@') = '\0';
-				(void) strcat(zc.zc_value, suffix);
+				(void) strcat(destsnap, suffix);
 			}
 		}
 	} else {
@@ -3149,8 +3154,8 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
 		    "cannot receive incremental stream"));
 
-		(void) strcpy(zc.zc_name, zc.zc_value);
+		(void) strcpy(name, destsnap);
-		*strchr(zc.zc_name, '@') = '\0';
+		*strchr(name, '@') = '\0';
 
 		/*
 		 * If the exact receive path was specified and this is the
@@ -3159,23 +3164,26 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		 */
 		if ((flags->isprefix || (*(chopprefix = drrb->drr_toname +
 		    strlen(sendfs)) != '\0' && *chopprefix != '@')) &&
-		    !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) {
+		    !zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
 			char snap[ZFS_MAXNAMELEN];
-			(void) strcpy(snap, strchr(zc.zc_value, '@'));
+			(void) strcpy(snap, strchr(destsnap, '@'));
-			if (guid_to_name(hdl, zc.zc_name, drrb->drr_fromguid,
+			if (guid_to_name(hdl, name, drrb->drr_fromguid,
-			    B_FALSE, zc.zc_value) == 0) {
+			    B_FALSE, destsnap) == 0) {
-				*strchr(zc.zc_value, '@') = '\0';
+				*strchr(destsnap, '@') = '\0';
-				(void) strcat(zc.zc_value, snap);
+				(void) strcat(destsnap, snap);
 			}
 		}
 	}
 
-	(void) strcpy(zc.zc_name, zc.zc_value);
+	(void) strcpy(name, destsnap);
-	*strchr(zc.zc_name, '@') = '\0';
+	*strchr(name, '@') = '\0';
 
-	if (zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) {
+	if (zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
+		zfs_cmd_t zc = {"\0"};
 		zfs_handle_t *zhp;
 
+		(void) strcpy(zc.zc_name, name);
+
 		/*
 		 * Destination fs exists.  It must be one of these cases:
 		 *  - an incremental send stream
@@ -3186,39 +3194,37 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		 */
 		if (stream_wantsnewfs) {
 			if (!flags->force) {
-				zcmd_free_nvlists(&zc);
 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 				    "destination '%s' exists\n"
-				    "must specify -F to overwrite it"),
+				    "must specify -F to overwrite it"), name);
-				    zc.zc_name);
+				err = zfs_error(hdl, EZFS_EXISTS, errbuf);
-				return (zfs_error(hdl, EZFS_EXISTS, errbuf));
+				goto out;
 			}
 			if (ioctl(hdl->libzfs_fd, ZFS_IOC_SNAPSHOT_LIST_NEXT,
 			    &zc) == 0) {
-				zcmd_free_nvlists(&zc);
 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 				    "destination has snapshots (eg. %s)\n"
 				    "must destroy them to overwrite it"),
-				    zc.zc_name);
+				    name);
-				return (zfs_error(hdl, EZFS_EXISTS, errbuf));
+				err = zfs_error(hdl, EZFS_EXISTS, errbuf);
+				goto out;
 			}
 		}
 
-		if ((zhp = zfs_open(hdl, zc.zc_name,
+		if ((zhp = zfs_open(hdl, name,
 		    ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) == NULL) {
-			zcmd_free_nvlists(&zc);
+			err = -1;
-			return (-1);
+			goto out;
 		}
 
 		if (stream_wantsnewfs &&
 		    zhp->zfs_dmustats.dds_origin[0]) {
-			zcmd_free_nvlists(&zc);
 			zfs_close(zhp);
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "destination '%s' is a clone\n"
-			    "must destroy it to overwrite it"),
+			    "must destroy it to overwrite it"), name);
-			    zc.zc_name);
+			err = zfs_error(hdl, EZFS_EXISTS, errbuf);
-			return (zfs_error(hdl, EZFS_EXISTS, errbuf));
+			goto out;
 		}
 
 		if (!flags->dryrun && zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
@@ -3227,14 +3233,14 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, 0);
 			if (clp == NULL) {
 				zfs_close(zhp);
-				zcmd_free_nvlists(&zc);
+				err = -1;
-				return (-1);
+				goto out;
 			}
 			if (changelist_prefix(clp) != 0) {
 				changelist_free(clp);
 				zfs_close(zhp);
-				zcmd_free_nvlists(&zc);
+				err = -1;
-				return (-1);
+				goto out;
 			}
 		}
 
@@ -3259,11 +3265,11 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		 * contained no slash character).
 		 */
 		if (!stream_wantsnewfs ||
-		    (cp = strrchr(zc.zc_name, '/')) == NULL) {
+		    (cp = strrchr(name, '/')) == NULL) {
-			zcmd_free_nvlists(&zc);
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
-			    "destination '%s' does not exist"), zc.zc_name);
+			    "destination '%s' does not exist"), name);
-			return (zfs_error(hdl, EZFS_NOENT, errbuf));
+			err = zfs_error(hdl, EZFS_NOENT, errbuf);
+			goto out;
 		}
 
 		/*
@@ -3273,45 +3279,34 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		*cp = '\0';
 
 		if (flags->isprefix && !flags->istail && !flags->dryrun &&
-		    create_parents(hdl, zc.zc_value, strlen(tosnap)) != 0) {
+		    create_parents(hdl, destsnap, strlen(tosnap)) != 0) {
-			zcmd_free_nvlists(&zc);
+			err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
-			return (zfs_error(hdl, EZFS_BADRESTORE, errbuf));
+			goto out;
 		}
 
 		newfs = B_TRUE;
 	}
 
-	zc.zc_begin_record = *drr_noswap;
-	zc.zc_cookie = infd;
-	zc.zc_guid = flags->force;
-	zc.zc_resumable = flags->resumable;
 	if (flags->verbose) {
 		(void) printf("%s %s stream of %s into %s\n",
 		    flags->dryrun ? "would receive" : "receiving",
 		    drrb->drr_fromguid ? "incremental" : "full",
-		    drrb->drr_toname, zc.zc_value);
+		    drrb->drr_toname, destsnap);
 		(void) fflush(stdout);
 	}
 
 	if (flags->dryrun) {
-		zcmd_free_nvlists(&zc);
+		err = recv_skip(hdl, infd, flags->byteswap);
-		return (recv_skip(hdl, infd, flags->byteswap));
+		goto out;
 	}
 
-	zc.zc_nvlist_dst = (uint64_t)(uintptr_t)prop_errbuf;
+	err = ioctl_err = lzc_receive_one(destsnap, props, origin,
-	zc.zc_nvlist_dst_size = sizeof (prop_errbuf);
+	    flags->force, flags->resumable, infd, drr_noswap, cleanup_fd,
-	zc.zc_cleanup_fd = cleanup_fd;
+	    &read_bytes, &errflags, action_handlep, &prop_errors);
-	zc.zc_action_handle = *action_handlep;
+	ioctl_errno = ioctl_err;
-
+	prop_errflags = errflags;
-	err = ioctl_err = zfs_ioctl(hdl, ZFS_IOC_RECV, &zc);
-	ioctl_errno = errno;
-	prop_errflags = (zprop_errflags_t)zc.zc_obj;
 
 	if (err == 0) {
-		nvlist_t *prop_errors;
-		VERIFY(0 == nvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst,
-		    zc.zc_nvlist_dst_size, &prop_errors, 0));
-
 		nvpair_t *prop_err = NULL;
 
 		while ((prop_err = nvlist_next_nvpair(prop_errors,
@@ -3344,25 +3339,20 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 				(void) snprintf(tbuf, sizeof (tbuf),
 				    dgettext(TEXT_DOMAIN,
 				    "cannot receive %s property on %s"),
-				    nvpair_name(prop_err), zc.zc_name);
+				    nvpair_name(prop_err), name);
 				zfs_setprop_error(hdl, prop, intval, tbuf);
 			}
 		}
-		nvlist_free(prop_errors);
 	}
 
-	zc.zc_nvlist_dst = 0;
-	zc.zc_nvlist_dst_size = 0;
-	zcmd_free_nvlists(&zc);
-
 	if (err == 0 && snapprops_nvlist) {
-		zfs_cmd_t zc2 = {"\0"};
+		zfs_cmd_t zc = {"\0"};
 
-		(void) strcpy(zc2.zc_name, zc.zc_value);
+		(void) strcpy(zc.zc_name, destsnap);
-		zc2.zc_cookie = B_TRUE; /* received */
+		zc.zc_cookie = B_TRUE; /* received */
-		if (zcmd_write_src_nvlist(hdl, &zc2, snapprops_nvlist) == 0) {
+		if (zcmd_write_src_nvlist(hdl, &zc, snapprops_nvlist) == 0) {
-			(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc2);
+			(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
-			zcmd_free_nvlists(&zc2);
+			zcmd_free_nvlists(&zc);
 		}
 	}
 
@@ -3374,7 +3364,7 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		 */
 		avl_tree_t *local_avl;
 		nvlist_t *local_nv, *fs;
-		cp = strchr(zc.zc_value, '@');
+		cp = strchr(destsnap, '@');
 
 		/*
 		 * XXX Do this faster by just iterating over snaps in
@@ -3382,7 +3372,7 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		 * get a strange "does not exist" error message.
 		 */
 		*cp = '\0';
-		if (gather_nvlist(hdl, zc.zc_value, NULL, NULL, B_FALSE,
+		if (gather_nvlist(hdl, destsnap, NULL, NULL, B_FALSE,
 		    &local_nv, &local_avl) == 0) {
 			*cp = '@';
 			fs = fsavl_find(local_avl, drrb->drr_toguid, NULL);
@@ -3392,7 +3382,7 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			if (fs != NULL) {
 				if (flags->verbose) {
 					(void) printf("snap %s already exists; "
-					    "ignoring\n", zc.zc_value);
+					    "ignoring\n", destsnap);
 				}
 				err = ioctl_err = recv_skip(hdl, infd,
 				    flags->byteswap);
@@ -3404,22 +3394,22 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 	if (ioctl_err != 0) {
 		switch (ioctl_errno) {
 		case ENODEV:
-			cp = strchr(zc.zc_value, '@');
+			cp = strchr(destsnap, '@');
 			*cp = '\0';
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "most recent snapshot of %s does not\n"
-			    "match incremental source"), zc.zc_value);
+			    "match incremental source"), destsnap);
 			(void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
 			*cp = '@';
 			break;
 		case ETXTBSY:
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "destination %s has been modified\n"
-			    "since most recent snapshot"), zc.zc_name);
+			    "since most recent snapshot"), name);
 			(void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
 			break;
 		case EEXIST:
-			cp = strchr(zc.zc_value, '@');
+			cp = strchr(destsnap, '@');
 			if (newfs) {
 				/* it's the containing fs that exists */
 				*cp = '\0';
@@ -3428,14 +3418,18 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			    "destination already exists"));
 			(void) zfs_error_fmt(hdl, EZFS_EXISTS,
 			    dgettext(TEXT_DOMAIN, "cannot restore to %s"),
-			    zc.zc_value);
+			    destsnap);
 			*cp = '@';
 			break;
 		case EINVAL:
+			if (flags->resumable)
+				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+				    "kernel modules must be upgraded to "
+				    "receive this stream."));
 			(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
 			break;
 		case ECKSUM:
-			recv_ecksum_set_aux(hdl, zc.zc_value, flags->resumable);
+			recv_ecksum_set_aux(hdl, destsnap, flags->resumable);
 			(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
 			break;
 		case ENOTSUP:
@@ -3445,7 +3439,7 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			break;
 		case EDQUOT:
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
-			    "destination %s space quota exceeded"), zc.zc_name);
+			    "destination %s space quota exceeded"), name);
 			(void) zfs_error(hdl, EZFS_NOSPC, errbuf);
 			break;
 		default:
@@ -3458,12 +3452,12 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 	 * children of the target filesystem if we did a replication
 	 * receive (indicated by stream_avl being non-NULL).
 	 */
-	cp = strchr(zc.zc_value, '@');
+	cp = strchr(destsnap, '@');
 	if (cp && (ioctl_err == 0 || !newfs)) {
 		zfs_handle_t *h;
 
 		*cp = '\0';
-		h = zfs_open(hdl, zc.zc_value,
+		h = zfs_open(hdl, destsnap,
 		    ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME);
 		if (h != NULL) {
 			if (h->zfs_type == ZFS_TYPE_VOLUME) {
@@ -3474,7 +3468,7 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 				 * for mounting and sharing later.
 				 */
 				if (top_zfs && *top_zfs == NULL)
-					*top_zfs = zfs_strdup(hdl, zc.zc_value);
+					*top_zfs = zfs_strdup(hdl, destsnap);
 			}
 			zfs_close(h);
 		}
@@ -3488,26 +3482,24 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 
 	if (prop_errflags & ZPROP_ERR_NOCLEAR) {
 		(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
-		    "failed to clear unreceived properties on %s"),
+		    "failed to clear unreceived properties on %s"), name);
-		    zc.zc_name);
 		(void) fprintf(stderr, "\n");
 	}
 	if (prop_errflags & ZPROP_ERR_NORESTORE) {
 		(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
-		    "failed to restore original properties on %s"),
+		    "failed to restore original properties on %s"), name);
-		    zc.zc_name);
 		(void) fprintf(stderr, "\n");
 	}
 
-	if (err || ioctl_err)
+	if (err || ioctl_err) {
-		return (-1);
+		err = -1;
-
+		goto out;
-	*action_handlep = zc.zc_action_handle;
+	}
 
 	if (flags->verbose) {
 		char buf1[64];
 		char buf2[64];
-		uint64_t bytes = zc.zc_cookie;
+		uint64_t bytes = read_bytes;
 		time_t delta = time(NULL) - begin_time;
 		if (delta == 0)
 			delta = 1;
@@ -3518,7 +3510,15 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 		    buf1, delta, buf2);
 	}
 
-	return (0);
+	err = 0;
+out:
+	if (prop_errors != NULL)
+		nvlist_free(prop_errors);
+
+	if (newprops)
+		nvlist_free(props);
+
+	return (err);
 }
 
 static int

lib/libzfs_core/libzfs_core.c

@@ -547,82 +547,168 @@ recv_read(int fd, void *buf, int ilen)
 	return (0);
 }
 
+/*
+ * Linux adds ZFS_IOC_RECV_NEW for resumable streams and preserves the legacy
+ * ZFS_IOC_RECV user/kernel interface.  The new interface supports all stream
+ * options but is currently only used for resumable streams.  This way updated
+ * user space utilities will interoperate with older kernel modules.
+ *
+ * Non-Linux OpenZFS platforms have opted to modify the legacy interface.
+ */
 static int
 recv_impl(const char *snapname, nvlist_t *props, const char *origin,
-    boolean_t force, boolean_t resumable, int fd,
+    boolean_t force, boolean_t resumable, int input_fd,
-    const dmu_replay_record_t *begin_record)
+    const dmu_replay_record_t *begin_record, int cleanup_fd,
+    uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle,
+    nvlist_t **errors)
 {
-	/*
+	dmu_replay_record_t drr;
-	 * The receive ioctl is still legacy, so we need to construct our own
+	char fsname[MAXPATHLEN];
-	 * zfs_cmd_t rather than using zfsc_ioctl().
-	 */
-	zfs_cmd_t zc = {"\0"};
 	char *atp;
-	char *packed = NULL;
-	size_t size;
 	int error;
 
-	ASSERT3S(g_refcount, >, 0);
+	/* Set 'fsname' to the name of containing filesystem */
-
+	(void) strlcpy(fsname, snapname, sizeof (fsname));
-	/* zc_name is name of containing filesystem */
+	atp = strchr(fsname, '@');
-	(void) strlcpy(zc.zc_name, snapname, sizeof (zc.zc_name));
-	atp = strchr(zc.zc_name, '@');
 	if (atp == NULL)
 		return (EINVAL);
 	*atp = '\0';
 
-	/* if the fs does not exist, try its parent. */
+	/* If the fs does not exist, try its parent. */
-	if (!lzc_exists(zc.zc_name)) {
+	if (!lzc_exists(fsname)) {
-		char *slashp = strrchr(zc.zc_name, '/');
+		char *slashp = strrchr(fsname, '/');
 		if (slashp == NULL)
 			return (ENOENT);
 		*slashp = '\0';
-
 	}
 
-	/* zc_value is full name of the snapshot to create */
+	/*
+	 * The begin_record is normally a non-byteswapped BEGIN record.
+	 * For resumable streams it may be set to any non-byteswapped
+	 * dmu_replay_record_t.
+	 */
+	if (begin_record == NULL) {
+		error = recv_read(input_fd, &drr, sizeof (drr));
+		if (error != 0)
+			return (error);
+	} else {
+		drr = *begin_record;
+	}
+
+	if (resumable) {
+		nvlist_t *outnvl = NULL;
+		nvlist_t *innvl = fnvlist_alloc();
+
+		fnvlist_add_string(innvl, "snapname", snapname);
+
+		if (props != NULL)
+			fnvlist_add_nvlist(innvl, "props", props);
+
+		if (origin != NULL && strlen(origin))
+			fnvlist_add_string(innvl, "origin", origin);
+
+		fnvlist_add_byte_array(innvl, "begin_record",
+		    (uchar_t *) &drr, sizeof (drr));
+
+		fnvlist_add_int32(innvl, "input_fd", input_fd);
+
+		if (force)
+			fnvlist_add_boolean(innvl, "force");
+
+		if (resumable)
+			fnvlist_add_boolean(innvl, "resumable");
+
+		if (cleanup_fd >= 0)
+			fnvlist_add_int32(innvl, "cleanup_fd", cleanup_fd);
+
+		if (action_handle != NULL)
+			fnvlist_add_uint64(innvl, "action_handle",
+			    *action_handle);
+
+		error = lzc_ioctl(ZFS_IOC_RECV_NEW, fsname, innvl, &outnvl);
+
+		if (error == 0 && read_bytes != NULL)
+			error = nvlist_lookup_uint64(outnvl, "read_bytes",
+			    read_bytes);
+
+		if (error == 0 && errflags != NULL)
+			error = nvlist_lookup_uint64(outnvl, "error_flags",
+			    errflags);
+
+		if (error == 0 && action_handle != NULL)
+			error = nvlist_lookup_uint64(outnvl, "action_handle",
+			    action_handle);
+
+		if (error == 0 && errors != NULL) {
+			nvlist_t *nvl;
+			error = nvlist_lookup_nvlist(outnvl, "errors", &nvl);
+			if (error == 0)
+				*errors = fnvlist_dup(nvl);
+		}
+
+		fnvlist_free(innvl);
+		fnvlist_free(outnvl);
+	} else {
+		zfs_cmd_t zc = {"\0"};
+		char *packed = NULL;
+		size_t size;
+
+		ASSERT3S(g_refcount, >, 0);
+
+		(void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_value));
 		(void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value));
 
 		if (props != NULL) {
-		/* zc_nvlist_src is props to set */
 			packed = fnvlist_pack(props, &size);
 			zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed;
 			zc.zc_nvlist_src_size = size;
 		}
 
-	/* zc_string is name of clone origin (if DRR_FLAG_CLONE) */
 		if (origin != NULL)
-		(void) strlcpy(zc.zc_string, origin, sizeof (zc.zc_string));
+			(void) strlcpy(zc.zc_string, origin,
+			    sizeof (zc.zc_string));
 
-	/* zc_begin_record is non-byteswapped BEGIN record */
+		ASSERT3S(drr.drr_type, ==, DRR_BEGIN);
-	if (begin_record == NULL) {
+		zc.zc_begin_record = drr.drr_u.drr_begin;
-		error = recv_read(fd, &zc.zc_begin_record,
-		    sizeof (zc.zc_begin_record));
-		if (error != 0)
-			goto out;
-	} else {
-		zc.zc_begin_record = *begin_record;
-	}
-
-	/* zc_cookie is fd to read from */
-	zc.zc_cookie = fd;
-
-	/* zc guid is force flag */
 		zc.zc_guid = force;
-
+		zc.zc_cookie = input_fd;
-	zc.zc_resumable = resumable;
-
-	/* zc_cleanup_fd is unused */
 		zc.zc_cleanup_fd = -1;
+		zc.zc_action_handle = 0;
+
+		if (cleanup_fd >= 0)
+			zc.zc_cleanup_fd = cleanup_fd;
+
+		if (action_handle != NULL)
+			zc.zc_action_handle = *action_handle;
+
+		zc.zc_nvlist_dst_size = 128 * 1024;
+		zc.zc_nvlist_dst = (uint64_t)(uintptr_t)
+		    malloc(zc.zc_nvlist_dst_size);
 
 		error = ioctl(g_fd, ZFS_IOC_RECV, &zc);
-	if (error != 0)
+		if (error != 0) {
 			error = errno;
+		} else {
+			if (read_bytes != NULL)
+				*read_bytes = zc.zc_cookie;
+
+			if (errflags != NULL)
+				*errflags = zc.zc_obj;
+
+			if (action_handle != NULL)
+				*action_handle = zc.zc_action_handle;
+
+			if (errors != NULL)
+				VERIFY0(nvlist_unpack(
+				    (void *)(uintptr_t)zc.zc_nvlist_dst,
+				    zc.zc_nvlist_dst_size, errors, KM_SLEEP));
+		}
 
-out:
 		if (packed != NULL)
 			fnvlist_pack_free(packed, size);
 		free((void *)(uintptr_t)zc.zc_nvlist_dst);
+	}
+
 	return (error);
 }
 
@@ -643,7 +729,8 @@ int
 lzc_receive(const char *snapname, nvlist_t *props, const char *origin,
     boolean_t force, int fd)
 {
-	return (recv_impl(snapname, props, origin, force, B_FALSE, fd, NULL));
+	return (recv_impl(snapname, props, origin, force, B_FALSE, fd,
+	    NULL, -1, NULL, NULL, NULL, NULL));
 }
 
 /*
@@ -656,7 +743,8 @@ int
 lzc_receive_resumable(const char *snapname, nvlist_t *props, const char *origin,
     boolean_t force, int fd)
 {
-	return (recv_impl(snapname, props, origin, force, B_TRUE, fd, NULL));
+	return (recv_impl(snapname, props, origin, force, B_TRUE, fd,
+	    NULL, -1, NULL, NULL, NULL, NULL));
 }
 
 /*
@@ -678,7 +766,38 @@ lzc_receive_with_header(const char *snapname, nvlist_t *props,
 	if (begin_record == NULL)
 		return (EINVAL);
 	return (recv_impl(snapname, props, origin, force, resumable, fd,
-	    begin_record));
+	    begin_record, -1, NULL, NULL, NULL, NULL));
+}
+
+/*
+ * Like lzc_receive, but allows the caller to pass all supported arguments
+ * and retrieve all values returned.  The only additional input parameter
+ * is 'cleanup_fd' which is used to set a cleanup-on-exit file descriptor.
+ *
+ * The following parameters all provide return values.  Several may be set
+ * in the failure case and will contain additional information.
+ *
+ * The 'read_bytes' value will be set to the total number of bytes read.
+ *
+ * The 'errflags' value will contain zprop_errflags_t flags which are
+ * used to describe any failures.
+ *
+ * The 'action_handle' is used to pass the handle for this guid/ds mapping.
+ * It should be set to zero on first call and will contain an updated handle
+ * on success, it should be passed in subsequent calls.
+ *
+ * The 'errors' nvlist contains an entry for each unapplied received
+ * property.  Callers are responsible for freeing this nvlist.
+ */
+int lzc_receive_one(const char *snapname, nvlist_t *props,
+    const char *origin, boolean_t force, boolean_t resumable, int input_fd,
+    const dmu_replay_record_t *begin_record, int cleanup_fd,
+    uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle,
+    nvlist_t **errors)
+{
+	return (recv_impl(snapname, props, origin, force, resumable,
+	    input_fd, begin_record, cleanup_fd, read_bytes, errflags,
+	    action_handle, errors));
 }
 
 /*

module/zfs/zfs_ioctl.c

@@ -945,6 +945,13 @@ zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 	    ZFS_DELEG_PERM_CREATE, cr));
 }
 
+/* ARGSUSED */
+static int
+zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
+{
+	return (zfs_secpolicy_recv(zc, innvl, cr));
+}
+
 int
 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 {
@@ -4046,74 +4053,38 @@ static boolean_t zfs_ioc_recv_inject_err;
 #endif
 
 /*
- * inputs:
+ * On failure the 'errors' nvlist may be allocated and will contain a
- * zc_name		name of containing filesystem
+ * descriptions of the failures.  It's the callers responsibilty to free.
- * zc_nvlist_src{_size}	nvlist of properties to apply
- * zc_value		name of snapshot to create
- * zc_string		name of clone origin (if DRR_FLAG_CLONE)
- * zc_cookie		file descriptor to recv from
- * zc_begin_record	the BEGIN record of the stream (not byteswapped)
- * zc_guid		force flag
- * zc_cleanup_fd	cleanup-on-exit file descriptor
- * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
- * zc_resumable		if data is incomplete assume sender will resume
- *
- * outputs:
- * zc_cookie		number of bytes read
- * zc_nvlist_dst{_size} error for each unapplied received property
- * zc_obj		zprop_errflags_t
- * zc_action_handle	handle for this guid/ds mapping
  */
 static int
-zfs_ioc_recv(zfs_cmd_t *zc)
+zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin,
+    nvlist_t *props, boolean_t force, boolean_t resumable, int input_fd,
+    dmu_replay_record_t *begin_record, int cleanup_fd, uint64_t *read_bytes,
+    uint64_t *errflags, uint64_t *action_handle, nvlist_t **errors)
 {
-	file_t *fp;
 	dmu_recv_cookie_t drc;
-	boolean_t force = (boolean_t)zc->zc_guid;
-	int fd;
 	int error = 0;
 	int props_error = 0;
-	nvlist_t *errors;
 	offset_t off;
-	nvlist_t *props = NULL; /* sent properties */
-	nvlist_t *origprops = NULL; /* existing properties */
 	nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
-	char *origin = NULL;
+	nvlist_t *origprops = NULL; /* existing properties */
-	char *tosnap;
-	char tofs[ZFS_MAXNAMELEN];
 	boolean_t first_recvd_props = B_FALSE;
+	file_t *input_fp;
 
-	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
+	*errors = NULL;
-	    strchr(zc->zc_value, '@') == NULL ||
+	input_fp = getf(input_fd);
-	    strchr(zc->zc_value, '%'))
+	if (input_fp == NULL)
-		return (SET_ERROR(EINVAL));
-
-	(void) strcpy(tofs, zc->zc_value);
-	tosnap = strchr(tofs, '@');
-	*tosnap++ = '\0';
-
-	if (zc->zc_nvlist_src != 0 &&
-	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
-	    zc->zc_iflags, &props)) != 0)
-		return (error);
-
-	fd = zc->zc_cookie;
-	fp = getf(fd);
-	if (fp == NULL) {
-		nvlist_free(props);
 		return (SET_ERROR(EBADF));
-	}
-
-	errors = fnvlist_alloc();
-
-	if (zc->zc_string[0])
-		origin = zc->zc_string;
 
 	error = dmu_recv_begin(tofs, tosnap,
-	    &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
+	    begin_record, force, resumable, origin, &drc);
 	if (error != 0)
 		goto out;
 
+	*read_bytes = 0;
+	*errflags = 0;
+	*errors = fnvlist_alloc();
+
 	/*
 	 * Set properties before we receive the stream so that they are applied
 	 * to the new data. Note that we must call dmu_recv_stream() if
@@ -4143,14 +4114,14 @@ zfs_ioc_recv(zfs_cmd_t *zc)
 			if (!first_recvd_props)
 				props_reduce(props, origprops);
 			if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
-				(void) nvlist_merge(errors, errlist, 0);
+				(void) nvlist_merge(*errors, errlist, 0);
 			nvlist_free(errlist);
 
 			if (clear_received_props(tofs, origprops,
 			    first_recvd_props ? NULL : props) != 0)
-				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
+				*errflags |= ZPROP_ERR_NOCLEAR;
 		} else {
-			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
+			*errflags |= ZPROP_ERR_NOCLEAR;
 		}
 	}
 
@@ -4160,13 +4131,13 @@ zfs_ioc_recv(zfs_cmd_t *zc)
 		if (props_error == 0) {
 			delayprops = extract_delay_props(props);
 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
-			    props, errors);
+			    props, *errors);
 		}
 	}
 
-	off = fp->f_offset;
+	off = input_fp->f_offset;
-	error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
+	error = dmu_recv_stream(&drc, input_fp->f_vnode, &off, cleanup_fd,
-	    &zc->zc_action_handle);
+	    action_handle);
 
 	if (error == 0) {
 		zfs_sb_t *zsb = NULL;
@@ -4192,7 +4163,7 @@ zfs_ioc_recv(zfs_cmd_t *zc)
 		/* Set delayed properties now, after we're done receiving. */
 		if (delayprops != NULL && error == 0) {
 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
-			    delayprops, errors);
+			    delayprops, *errors);
 		}
 	}
 
@@ -4210,23 +4181,10 @@ zfs_ioc_recv(zfs_cmd_t *zc)
 		nvlist_free(delayprops);
 	}
 
-	/*
-	 * Now that all props, initial and delayed, are set, report the prop
-	 * errors to the caller.
-	 */
-	if (zc->zc_nvlist_dst_size != 0 &&
-	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
-	    put_nvlist(zc, errors) != 0)) {
-		/*
-		 * Caller made zc->zc_nvlist_dst less than the minimum expected
-		 * size or supplied an invalid address.
-		 */
-		props_error = SET_ERROR(EINVAL);
-	}
 
-	zc->zc_cookie = off - fp->f_offset;
+	*read_bytes = off - input_fp->f_offset;
-	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
+	if (VOP_SEEK(input_fp->f_vnode, input_fp->f_offset, &off, NULL) == 0)
-		fp->f_offset = off;
+	    input_fp->f_offset = off;
 
 #ifdef	DEBUG
 	if (zfs_ioc_recv_inject_err) {
@@ -4245,14 +4203,14 @@ zfs_ioc_recv(zfs_cmd_t *zc)
 			 * Since we may have left a $recvd value on the
 			 * system, we can't clear the $hasrecvd flag.
 			 */
-			zc->zc_obj |= ZPROP_ERR_NORESTORE;
+			*errflags |= ZPROP_ERR_NORESTORE;
 		} else if (first_recvd_props) {
 			dsl_prop_unset_hasrecvd(tofs);
 		}
 
 		if (origprops == NULL && !drc.drc_newfs) {
 			/* We failed to stash the original properties. */
-			zc->zc_obj |= ZPROP_ERR_NORESTORE;
+			*errflags |= ZPROP_ERR_NORESTORE;
 		}
 
 		/*
@@ -4269,14 +4227,12 @@ zfs_ioc_recv(zfs_cmd_t *zc)
 			 * We stashed the original properties but failed to
 			 * restore them.
 			 */
-			zc->zc_obj |= ZPROP_ERR_NORESTORE;
+			*errflags |= ZPROP_ERR_NORESTORE;
 		}
 	}
 out:
-	nvlist_free(props);
+	releasef(input_fd);
 	nvlist_free(origprops);
-	nvlist_free(errors);
-	releasef(fd);
 
 	if (error == 0)
 		error = props_error;
@@ -4284,6 +4240,176 @@ out:
 	return (error);
 }
 
+/*
+ * inputs:
+ * zc_name		name of containing filesystem (unused)
+ * zc_nvlist_src{_size}	nvlist of properties to apply
+ * zc_value		name of snapshot to create
+ * zc_string		name of clone origin (if DRR_FLAG_CLONE)
+ * zc_cookie		file descriptor to recv from
+ * zc_begin_record	the BEGIN record of the stream (not byteswapped)
+ * zc_guid		force flag
+ * zc_cleanup_fd	cleanup-on-exit file descriptor
+ * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
+ *
+ * outputs:
+ * zc_cookie		number of bytes read
+ * zc_obj		zprop_errflags_t
+ * zc_action_handle	handle for this guid/ds mapping
+ * zc_nvlist_dst{_size} error for each unapplied received property
+ */
+static int
+zfs_ioc_recv(zfs_cmd_t *zc)
+{
+	dmu_replay_record_t begin_record;
+	nvlist_t *errors = NULL;
+	nvlist_t *props = NULL;
+	char *origin = NULL;
+	char *tosnap;
+	char tofs[ZFS_MAXNAMELEN];
+	int error = 0;
+
+	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
+	    strchr(zc->zc_value, '@') == NULL ||
+	    strchr(zc->zc_value, '%'))
+		return (SET_ERROR(EINVAL));
+
+	(void) strcpy(tofs, zc->zc_value);
+	tosnap = strchr(tofs, '@');
+	*tosnap++ = '\0';
+
+	if (zc->zc_nvlist_src != 0 &&
+	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+	    zc->zc_iflags, &props)) != 0)
+		return (error);
+
+	if (zc->zc_string[0])
+		origin = zc->zc_string;
+
+	begin_record.drr_type = DRR_BEGIN;
+	begin_record.drr_payloadlen = 0;
+	begin_record.drr_u.drr_begin = zc->zc_begin_record;
+
+	error = zfs_ioc_recv_impl(tofs, tosnap, origin, props, zc->zc_guid,
+	    B_FALSE, zc->zc_cookie, &begin_record, zc->zc_cleanup_fd,
+	    &zc->zc_cookie, &zc->zc_obj, &zc->zc_action_handle, &errors);
+	nvlist_free(props);
+
+	/*
+	 * Now that all props, initial and delayed, are set, report the prop
+	 * errors to the caller.
+	 */
+	if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
+	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
+	    put_nvlist(zc, errors) != 0)) {
+		/*
+		 * Caller made zc->zc_nvlist_dst less than the minimum expected
+		 * size or supplied an invalid address.
+		 */
+		error = SET_ERROR(EINVAL);
+	}
+
+	nvlist_free(errors);
+
+	return (error);
+}
+
+/*
+ * innvl: {
+ *     "snapname" -> full name of the snapshot to create
+ *     (optional) "props" -> properties to set (nvlist)
+ *     (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
+ *     "begin_record" -> non-byteswapped dmu_replay_record_t
+ *     "input_fd" -> file descriptor to read stream from (int32)
+ *     (optional) "force" -> force flag (value ignored)
+ *     (optional) "resumable" -> resumable flag (value ignored)
+ *     (optional) "cleanup_fd" -> cleanup-on-exit file descriptor
+ *     (optional) "action_handle" -> handle for this guid/ds mapping
+ * }
+ *
+ * outnvl: {
+ *     "read_bytes" -> number of bytes read
+ *     "error_flags" -> zprop_errflags_t
+ *     "action_handle" -> handle for this guid/ds mapping
+ *     "errors" -> error for each unapplied received property (nvlist)
+ * }
+ */
+static int
+zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
+{
+	dmu_replay_record_t *begin_record;
+	uint_t begin_record_size;
+	nvlist_t *errors = NULL;
+	nvlist_t *props = NULL;
+	char *snapname = NULL;
+	char *origin = NULL;
+	char *tosnap;
+	char tofs[ZFS_MAXNAMELEN];
+	boolean_t force;
+	boolean_t resumable;
+	uint64_t action_handle = 0;
+	uint64_t read_bytes = 0;
+	uint64_t errflags = 0;
+	int input_fd = -1;
+	int cleanup_fd = -1;
+	int error;
+
+	error = nvlist_lookup_string(innvl, "snapname", &snapname);
+	if (error != 0)
+		return (SET_ERROR(EINVAL));
+
+	if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
+	    strchr(snapname, '@') == NULL ||
+	    strchr(snapname, '%'))
+		return (SET_ERROR(EINVAL));
+
+	(void) strcpy(tofs, snapname);
+	tosnap = strchr(tofs, '@');
+	*tosnap++ = '\0';
+
+	error = nvlist_lookup_string(innvl, "origin", &origin);
+	if (error && error != ENOENT)
+		return (error);
+
+	error = nvlist_lookup_byte_array(innvl, "begin_record",
+	    (uchar_t **) &begin_record, &begin_record_size);
+	if (error != 0 || begin_record_size != sizeof (*begin_record))
+		return (SET_ERROR(EINVAL));
+
+	error = nvlist_lookup_int32(innvl, "input_fd", &input_fd);
+	if (error != 0)
+		return (SET_ERROR(EINVAL));
+
+	force = nvlist_exists(innvl, "force");
+	resumable = nvlist_exists(innvl, "resumable");
+
+	error = nvlist_lookup_int32(innvl, "cleanup_fd", &cleanup_fd);
+	if (error && error != ENOENT)
+		return (error);
+
+	error = nvlist_lookup_uint64(innvl, "action_handle", &action_handle);
+	if (error && error != ENOENT)
+		return (error);
+
+	error = nvlist_lookup_nvlist(innvl, "props", &props);
+	if (error && error != ENOENT)
+		return (error);
+
+	error = zfs_ioc_recv_impl(tofs, tosnap, origin, props, force,
+	    resumable, input_fd, begin_record, cleanup_fd, &read_bytes,
+	    &errflags, &action_handle, &errors);
+
+	fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
+	fnvlist_add_uint64(outnvl, "error_flags", errflags);
+	fnvlist_add_uint64(outnvl, "action_handle", action_handle);
+	fnvlist_add_nvlist(outnvl, "errors", errors);
+
+	nvlist_free(errors);
+	nvlist_free(props);
+
+	return (error);
+}
+
 /*
  * inputs:
  * zc_name	name of snapshot to send
@@ -5555,6 +5681,10 @@ zfs_ioctl_init(void)
 	    POOL_NAME,
 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
 
+	zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
+	    zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
+	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
+
 	/* IOCTLS that use the legacy function signature */
 
 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,