Illumos 4757, 4913

4757 ZFS embedded-data block pointers ("zero block compression")
4913 zfs release should not be subject to space checks

Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Max Grossman <max.grossman@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Dan McDonald <danmcd@omniti.com>
Approved by: Dan McDonald <danmcd@omniti.com>

References:
  https://www.illumos.org/issues/4757
  https://www.illumos.org/issues/4913
  illumos/illumos-gate@5d7b4d4

Porting notes:

For compatibility with the fastpath code the zio_done() function
needed to be updated.  Because embedded-data block pointers do
not require DVAs to be allocated the associated vdevs will not
be marked and therefore should not be unmarked.

Ported by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #2544

Author: ahrens

cmd/zdb/zdb.c

@@ -1047,6 +1047,16 @@ snprintf_blkptr_compact(char *blkbuf, size_t buflen, const blkptr_t *bp)
 		return;
 	}
 
+	if (BP_IS_EMBEDDED(bp)) {
+		(void) sprintf(blkbuf,
+		    "EMBEDDED et=%u %llxL/%llxP B=%llu",
+		    (int)BPE_GET_ETYPE(bp),
+		    (u_longlong_t)BPE_GET_LSIZE(bp),
+		    (u_longlong_t)BPE_GET_PSIZE(bp),
+		    (u_longlong_t)bp->blk_birth);
+		return;
+	}
+
 	blkbuf[0] = '\0';
 
 	for (i = 0; i < ndvas; i++)
@@ -1066,7 +1076,7 @@ snprintf_blkptr_compact(char *blkbuf, size_t buflen, const blkptr_t *bp)
 		    "%llxL/%llxP F=%llu B=%llu/%llu",
 		    (u_longlong_t)BP_GET_LSIZE(bp),
 		    (u_longlong_t)BP_GET_PSIZE(bp),
-		    (u_longlong_t)bp->blk_fill,
+		    (u_longlong_t)BP_GET_FILL(bp),
 		    (u_longlong_t)bp->blk_birth,
 		    (u_longlong_t)BP_PHYSICAL_BIRTH(bp));
 	}
@@ -1079,8 +1089,10 @@ print_indirect(blkptr_t *bp, const zbookmark_t *zb,
 	char blkbuf[BP_SPRINTF_LEN];
 	int l;
 
-	ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
-	ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
+	if (!BP_IS_EMBEDDED(bp)) {
+		ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
+		ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
+	}
 
 	(void) printf("%16llx ", (u_longlong_t)blkid2offset(dnp, bp, zb));
 
@@ -1134,10 +1146,10 @@ visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
 			err = visit_indirect(spa, dnp, cbp, &czb);
 			if (err)
 				break;
-			fill += cbp->blk_fill;
+			fill += BP_GET_FILL(cbp);
 		}
 		if (!err)
-			ASSERT3U(fill, ==, bp->blk_fill);
+			ASSERT3U(fill, ==, BP_GET_FILL(bp));
 		(void) arc_buf_remove_ref(buf, &buf);
 	}
 
@@ -1861,14 +1873,14 @@ dump_dir(objset_t *os)
 
 	if (dds.dds_type == DMU_OST_META) {
 		dds.dds_creation_txg = TXG_INITIAL;
-		usedobjs = os->os_rootbp->blk_fill;
+		usedobjs = BP_GET_FILL(os->os_rootbp);
 		refdbytes = os->os_spa->spa_dsl_pool->
 		    dp_mos_dir->dd_phys->dd_used_bytes;
 	} else {
 		dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch);
 	}
 
-	ASSERT3U(usedobjs, ==, os->os_rootbp->blk_fill);
+	ASSERT3U(usedobjs, ==, BP_GET_FILL(os->os_rootbp));
 
 	zdb_nicenum(refdbytes, numbuf);
 
@@ -2171,6 +2183,9 @@ typedef struct zdb_cb {
 	zdb_blkstats_t	zcb_type[ZB_TOTAL + 1][ZDB_OT_TOTAL + 1];
 	uint64_t	zcb_dedup_asize;
 	uint64_t	zcb_dedup_blocks;
+	uint64_t	zcb_embedded_blocks[NUM_BP_EMBEDDED_TYPES];
+	uint64_t	zcb_embedded_histogram[NUM_BP_EMBEDDED_TYPES]
+	    [BPE_PAYLOAD_SIZE];
 	uint64_t	zcb_start;
 	uint64_t	zcb_lastprint;
 	uint64_t	zcb_totalasize;
@@ -2204,6 +2219,13 @@ zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
 		zb->zb_psize_histogram[BP_GET_PSIZE(bp) >> SPA_MINBLOCKSHIFT]++;
 	}
 
+	if (BP_IS_EMBEDDED(bp)) {
+		zcb->zcb_embedded_blocks[BPE_GET_ETYPE(bp)]++;
+		zcb->zcb_embedded_histogram[BPE_GET_ETYPE(bp)]
+		    [BPE_GET_PSIZE(bp)]++;
+		return;
+	}
+
 	if (dump_opt['L'])
 		return;
 
@@ -2301,7 +2323,8 @@ zdb_blkptr_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 
 	is_metadata = (BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type));
 
-	if (dump_opt['c'] > 1 || (dump_opt['c'] && is_metadata)) {
+	if (!BP_IS_EMBEDDED(bp) &&
+	    (dump_opt['c'] > 1 || (dump_opt['c'] && is_metadata))) {
 		size_t size = BP_GET_PSIZE(bp);
 		void *data = zio_data_buf_alloc(size);
 		int flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB | ZIO_FLAG_RAW;
@@ -2497,8 +2520,9 @@ dump_block_stats(spa_t *spa)
 	zdb_blkstats_t *zb, *tzb;
 	uint64_t norm_alloc, norm_space, total_alloc, total_found;
 	int flags = TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA | TRAVERSE_HARD;
-	int leaks = 0;
+	boolean_t leaks = B_FALSE;
 	int e;
+	bp_embedded_type_t i;
 
 	(void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
 	    (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "",
@@ -2587,7 +2611,7 @@ dump_block_stats(spa_t *spa)
 		    (u_longlong_t)total_alloc,
 		    (dump_opt['L']) ? "unreachable" : "leaked",
 		    (longlong_t)(total_alloc - total_found));
-		leaks = 1;
+		leaks = B_TRUE;
 	}
 
 	if (tzb->zb_count == 0)
@@ -2617,6 +2641,23 @@ dump_block_stats(spa_t *spa)
 	(void) printf("\tSPA allocated: %10llu     used: %5.2f%%\n",
 	    (u_longlong_t)norm_alloc, 100.0 * norm_alloc / norm_space);
 
+	for (i = 0; i < NUM_BP_EMBEDDED_TYPES; i++) {
+		if (zcb.zcb_embedded_blocks[i] == 0)
+			continue;
+		(void) printf("\n");
+		(void) printf("\tadditional, non-pointer bps of type %u: "
+		    "%10llu\n",
+		    i, (u_longlong_t)zcb.zcb_embedded_blocks[i]);
+
+		if (dump_opt['b'] >= 3) {
+			(void) printf("\t number of (compressed) bytes:  "
+			    "number of bps\n");
+			dump_histogram(zcb.zcb_embedded_histogram[i],
+			    sizeof (zcb.zcb_embedded_histogram[i]) /
+			    sizeof (zcb.zcb_embedded_histogram[i][0]), 0);
+		}
+	}
+
 	if (dump_opt['b'] >= 2) {
 		int l, t, level;
 		(void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
@@ -2718,14 +2759,14 @@ 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_IS_HOLE(bp))
+	if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp))
 		return (0);
 
 	if (dump_opt['S'] > 1 && zb->zb_level == ZB_ROOT_LEVEL) {
 		(void) printf("traversing objset %llu, %llu objects, "
 		    "%lu blocks so far\n",
 		    (u_longlong_t)zb->zb_objset,
-		    (u_longlong_t)bp->blk_fill,
+		    (u_longlong_t)BP_GET_FILL(bp),
 		    avl_numnodes(t));
 	}
 

cmd/zfs/zfs_main.c

@@ -258,9 +258,9 @@ get_usage(zfs_help_t idx)
 	case HELP_ROLLBACK:
 		return (gettext("\trollback [-rRf] <snapshot>\n"));
 	case HELP_SEND:
-		return (gettext("\tsend [-DnPpRrv] [-[iI] snapshot] "
+		return (gettext("\tsend [-DnPpRrve] [-[iI] snapshot] "
 		    "<snapshot>\n"
-		    "\tsend [-i snapshot|bookmark] "
+		    "\tsend [-e] [-i snapshot|bookmark] "
 		    "<filesystem|volume|snapshot>\n"));
 	case HELP_SET:
 		return (gettext("\tset <property=value> "
@@ -3338,6 +3338,8 @@ rollback_check_dependent(zfs_handle_t *zhp, void *data)
 	zfs_close(zhp);
 	return (0);
 }
+
+
 /*
  * Report any snapshots more recent than the one specified.  Used when '-r' is
  * not specified.  We reuse this same callback for the snapshot dependents - if
@@ -3677,7 +3679,7 @@ zfs_do_send(int argc, char **argv)
 	boolean_t extraverbose = B_FALSE;
 
 	/* check options */
-	while ((c = getopt(argc, argv, ":i:I:RDpvnP")) != -1) {
+	while ((c = getopt(argc, argv, ":i:I:RDpvnPe")) != -1) {
 		switch (c) {
 		case 'i':
 			if (fromname)
@@ -3712,6 +3714,9 @@ zfs_do_send(int argc, char **argv)
 		case 'n':
 			flags.dryrun = B_TRUE;
 			break;
+		case 'e':
+			flags.embed_data = B_TRUE;
+			break;
 		case ':':
 			(void) fprintf(stderr, gettext("missing argument for "
 			    "'%c' option\n"), optopt);
@@ -3750,6 +3755,7 @@ zfs_do_send(int argc, char **argv)
 	if (strchr(argv[0], '@') == NULL ||
 	    (fromname && strchr(fromname, '#') != NULL)) {
 		char frombuf[ZFS_MAXNAMELEN];
+		enum lzc_send_flags lzc_flags = 0;
 
 		if (flags.replicate || flags.doall || flags.props ||
 		    flags.dedup || flags.dryrun || flags.verbose ||
@@ -3764,6 +3770,9 @@ zfs_do_send(int argc, char **argv)
 		if (zhp == NULL)
 			return (1);
 
+		if (flags.embed_data)
+			lzc_flags |= LZC_SEND_FLAG_EMBED_DATA;
+
 		if (fromname != NULL &&
 		    (fromname[0] == '#' || fromname[0] == '@')) {
 			/*
@@ -3777,7 +3786,7 @@ zfs_do_send(int argc, char **argv)
 			(void) strlcat(frombuf, fromname, sizeof (frombuf));
 			fromname = frombuf;
 		}
-		err = zfs_send_one(zhp, fromname, STDOUT_FILENO);
+		err = zfs_send_one(zhp, fromname, STDOUT_FILENO, lzc_flags);
 		zfs_close(zhp);
 		return (err != 0);
 	}

cmd/zstreamdump/zstreamdump.c

@@ -36,7 +36,6 @@
 #include <sys/zfs_ioctl.h>
 #include <zfs_fletcher.h>
 
-uint64_t drr_record_count[DRR_NUMTYPES];
 uint64_t total_write_size = 0;
 uint64_t total_stream_len = 0;
 FILE *send_stream = 0;
@@ -81,6 +80,8 @@ int
 main(int argc, char *argv[])
 {
 	char *buf = malloc(INITIAL_BUFLEN);
+	uint64_t drr_record_count[DRR_NUMTYPES] = { 0 };
+	uint64_t total_records = 0;
 	dmu_replay_record_t thedrr;
 	dmu_replay_record_t *drr = &thedrr;
 	struct drr_begin *drrb = &thedrr.drr_u.drr_begin;
@@ -91,6 +92,7 @@ main(int argc, char *argv[])
 	struct drr_write_byref *drrwbr = &thedrr.drr_u.drr_write_byref;
 	struct drr_free *drrf = &thedrr.drr_u.drr_free;
 	struct drr_spill *drrs = &thedrr.drr_u.drr_spill;
+	struct drr_write_embedded *drrwe = &thedrr.drr_u.drr_write_embedded;
 	char c;
 	boolean_t verbose = B_FALSE;
 	boolean_t first = B_TRUE;
@@ -170,6 +172,7 @@ main(int argc, char *argv[])
 		}
 
 		drr_record_count[drr->drr_type]++;
+		total_records++;
 
 		switch (drr->drr_type) {
 		case DRR_BEGIN:
@@ -286,8 +289,8 @@ main(int argc, char *argv[])
 				    drro->drr_bonuslen);
 			}
 			if (drro->drr_bonuslen > 0) {
-				(void) ssread(buf, P2ROUNDUP(drro->drr_bonuslen,
-				    8), &zc);
+				(void) ssread(buf,
+				    P2ROUNDUP(drro->drr_bonuslen, 8), &zc);
 			}
 			break;
 
@@ -397,6 +400,38 @@ main(int argc, char *argv[])
 			}
 			(void) ssread(buf, drrs->drr_length, &zc);
 			break;
+		case DRR_WRITE_EMBEDDED:
+			if (do_byteswap) {
+				drrwe->drr_object =
+				    BSWAP_64(drrwe->drr_object);
+				drrwe->drr_offset =
+				    BSWAP_64(drrwe->drr_offset);
+				drrwe->drr_length =
+				    BSWAP_64(drrwe->drr_length);
+				drrwe->drr_toguid =
+				    BSWAP_64(drrwe->drr_toguid);
+				drrwe->drr_lsize =
+				    BSWAP_32(drrwe->drr_lsize);
+				drrwe->drr_psize =
+				    BSWAP_32(drrwe->drr_psize);
+			}
+			if (verbose) {
+				(void) printf("WRITE_EMBEDDED object = %llu "
+				    "offset = %llu length = %llu\n"
+				    "toguid = %llx comp = %u etype = %u "
+				    "lsize = %u psize = %u\n",
+				    (u_longlong_t)drrwe->drr_object,
+				    (u_longlong_t)drrwe->drr_offset,
+				    (u_longlong_t)drrwe->drr_length,
+				    (u_longlong_t)drrwe->drr_toguid,
+				    drrwe->drr_compression,
+				    drrwe->drr_etype,
+				    drrwe->drr_lsize,
+				    drrwe->drr_psize);
+			}
+			(void) ssread(buf,
+			    P2ROUNDUP(drrwe->drr_psize, 8), &zc);
+			break;
 		case DRR_NUMTYPES:
 			/* should never be reached */
 			exit(1);
@@ -418,18 +453,16 @@ main(int argc, char *argv[])
 	    (u_longlong_t)drr_record_count[DRR_FREEOBJECTS]);
 	(void) printf("\tTotal DRR_WRITE records = %lld\n",
 	    (u_longlong_t)drr_record_count[DRR_WRITE]);
+	(void) printf("\tTotal DRR_WRITE_BYREF records = %lld\n",
+	    (u_longlong_t)drr_record_count[DRR_WRITE_BYREF]);
+	(void) printf("\tTotal DRR_WRITE_EMBEDDED records = %lld\n",
+	    (u_longlong_t)drr_record_count[DRR_WRITE_EMBEDDED]);
 	(void) printf("\tTotal DRR_FREE records = %lld\n",
 	    (u_longlong_t)drr_record_count[DRR_FREE]);
 	(void) printf("\tTotal DRR_SPILL records = %lld\n",
 	    (u_longlong_t)drr_record_count[DRR_SPILL]);
 	(void) printf("\tTotal records = %lld\n",
-	    (u_longlong_t)(drr_record_count[DRR_BEGIN] +
-	    drr_record_count[DRR_OBJECT] +
-	    drr_record_count[DRR_FREEOBJECTS] +
-	    drr_record_count[DRR_WRITE] +
-	    drr_record_count[DRR_FREE] +
-	    drr_record_count[DRR_SPILL] +
-	    drr_record_count[DRR_END]));
+	    (u_longlong_t)total_records);
 	(void) printf("\tTotal write size = %lld (0x%llx)\n",
 	    (u_longlong_t)total_write_size, (u_longlong_t)total_write_size);
 	(void) printf("\tTotal stream length = %lld (0x%llx)\n",