Illumos 5027 - zfs large block support

5027 zfs large block support
Reviewed by: Alek Pinchuk <[email protected]>
Reviewed by: George Wilson <[email protected]>
Reviewed by: Josef 'Jeff' Sipek <[email protected]>
Reviewed by: Richard Elling <[email protected]>
Reviewed by: Saso Kiselkov <[email protected]>
Reviewed by: Brian Behlendorf <[email protected]>
Approved by: Dan McDonald <[email protected]>

References:
  https://www.illumos.org/issues/5027
  https://github.com/illumos/illumos-gate/commit/b515258

Porting Notes:

* Included in this patch is a tiny ISP2() cleanup in zio_init() from
Illumos 5255.

* Unlike the upstream Illumos commit this patch does not impose an
arbitrary 128K block size limit on volumes.  Volumes, like filesystems,
are limited by the zfs_max_recordsize=1M module option.

* By default the maximum record size is limited to 1M by the module
option zfs_max_recordsize.  This value may be safely increased up to
16M which is the largest block size supported by the on-disk format.
At the moment, 1M blocks clearly offer a significant performance
improvement but the benefits of going beyond this for the majority
of workloads are less clear.

* The illumos version of this patch increased DMU_MAX_ACCESS to 32M.
This was determined not to be large enough when using 16M blocks
because the zfs_make_xattrdir() function will fail (EFBIG) when
assigning a TX.  This was immediately observed under Linux because
all newly created files must have a security xattr created and
that was failing.  Therefore, we've set DMU_MAX_ACCESS to 64M.

* On 32-bit platforms a hard limit of 1M is set for blocks due
to the limited virtual address space.  We should be able to relax
this one the ABD patches are merged.

Ported-by: Brian Behlendorf <[email protected]>
Closes #354

4 files changed, 61 insertions, 14 deletions

diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c
index 25f44212a..ab6155054 100644
--- a/cmd/zdb/zdb.c
+++ b/cmd/zdb/zdb.c
@@ -2185,6 +2185,8 @@ dump_label(const char *dev)
 	(void) close(fd);
 }
 
+static uint64_t num_large_blocks;
+
 /*ARGSUSED*/
 static int
 dump_one_dir(const char *dsname, void *arg)
@@ -2197,6 +2199,8 @@ dump_one_dir(const char *dsname, void *arg)
 		(void) printf("Could not open %s, error %d\n", dsname, error);
 		return (0);
 	}
+	if (dmu_objset_ds(os)->ds_large_blocks)
+		num_large_blocks++;
 	dump_dir(os);
 	dmu_objset_disown(os, FTAG);
 	fuid_table_destroy();
@@ -2207,7 +2211,7 @@ dump_one_dir(const char *dsname, void *arg)
 /*
  * Block statistics.
  */
-#define	PSIZE_HISTO_SIZE (SPA_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 1)
+#define	PSIZE_HISTO_SIZE (SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 2)
 typedef struct zdb_blkstats {
 	uint64_t zb_asize;
 	uint64_t zb_lsize;
@@ -2273,7 +2277,15 @@ zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
 		zb->zb_lsize += BP_GET_LSIZE(bp);
 		zb->zb_psize += BP_GET_PSIZE(bp);
 		zb->zb_count++;
-		zb->zb_psize_histogram[BP_GET_PSIZE(bp) >> SPA_MINBLOCKSHIFT]++;
+
+		/*
+		 * The histogram is only big enough to record blocks up to
+		 * SPA_OLD_MAXBLOCKSIZE; larger blocks go into the last,
+		 * "other", bucket.
+		 */
+		int idx = BP_GET_PSIZE(bp) >> SPA_MINBLOCKSHIFT;
+		idx = MIN(idx, SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 1);
+		zb->zb_psize_histogram[idx]++;
 
 		zb->zb_gangs += BP_COUNT_GANG(bp);
 
@@ -2979,6 +2991,7 @@ dump_zpool(spa_t *spa)
 		dump_metaslab_groups(spa);
 
 	if (dump_opt['d'] || dump_opt['i']) {
+		uint64_t refcount;
 		dump_dir(dp->dp_meta_objset);
 		if (dump_opt['d'] >= 3) {
 			dump_bpobj(&spa->spa_deferred_bpobj,
@@ -2998,8 +3011,21 @@ dump_zpool(spa_t *spa)
 		}
 		(void) dmu_objset_find(spa_name(spa), dump_one_dir,
 		    NULL, DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
+
+		(void) feature_get_refcount(spa,
+		    &spa_feature_table[SPA_FEATURE_LARGE_BLOCKS], &refcount);
+		if (num_large_blocks != refcount) {
+			(void) printf("large_blocks feature refcount mismatch: "
+			    "expected %lld != actual %lld\n",
+			    (longlong_t)num_large_blocks,
+			    (longlong_t)refcount);
+			rc = 2;
+		} else {
+			(void) printf("Verified large_blocks feature refcount "
+			    "is correct (%llu)\n", (longlong_t)refcount);
+		}
 	}
-	if (dump_opt['b'] || dump_opt['c'])
+	if (rc == 0 && (dump_opt['b'] || dump_opt['c']))
 		rc = dump_block_stats(spa);
 
 	if (rc == 0)
diff --git a/cmd/zfs/zfs_main.c b/cmd/zfs/zfs_main.c
index 50ac59fba..32b9239b7 100644
--- a/cmd/zfs/zfs_main.c
+++ b/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 [-DnPpRrve] [-[iI] snapshot] "
+		return (gettext("\tsend [-DnPpRvLe] [-[iI] snapshot] "
 		    "<snapshot>\n"
-		    "\tsend [-e] [-i snapshot|bookmark] "
+		    "\tsend [-Le] [-i snapshot|bookmark] "
 		    "<filesystem|volume|snapshot>\n"));
 	case HELP_SET:
 		return (gettext("\tset <property=value> "
@@ -3683,7 +3683,7 @@ zfs_do_send(int argc, char **argv)
 	boolean_t extraverbose = B_FALSE;
 
 	/* check options */
-	while ((c = getopt(argc, argv, ":i:I:RDpvnPe")) != -1) {
+	while ((c = getopt(argc, argv, ":i:I:RDpvnPLe")) != -1) {
 		switch (c) {
 		case 'i':
 			if (fromname)
@@ -3718,6 +3718,9 @@ zfs_do_send(int argc, char **argv)
 		case 'n':
 			flags.dryrun = B_TRUE;
 			break;
+		case 'L':
+			flags.largeblock = B_TRUE;
+			break;
 		case 'e':
 			flags.embed_data = B_TRUE;
 			break;
@@ -3774,6 +3777,8 @@ zfs_do_send(int argc, char **argv)
 		if (zhp == NULL)
 			return (1);
 
+		if (flags.largeblock)
+			lzc_flags |= LZC_SEND_FLAG_LARGE_BLOCK;
 		if (flags.embed_data)
 			lzc_flags |= LZC_SEND_FLAG_EMBED_DATA;
 
diff --git a/cmd/zstreamdump/zstreamdump.c b/cmd/zstreamdump/zstreamdump.c
index c51a9c806..176dd66b2 100644
--- a/cmd/zstreamdump/zstreamdump.c
+++ b/cmd/zstreamdump/zstreamdump.c
@@ -56,7 +56,6 @@ uint64_t total_stream_len = 0;
 FILE *send_stream = 0;
 boolean_t do_byteswap = B_FALSE;
 boolean_t do_cksum = B_TRUE;
-#define	INITIAL_BUFLEN (1<<20)
 
 static void
 usage(void)
@@ -69,6 +68,18 @@ usage(void)
 	exit(1);
 }
 
+static void *
+safe_malloc(size_t size)
+{
+	void *rv = malloc(size);
+	if (rv == NULL) {
+		(void) fprintf(stderr, "ERROR; failed to allocate %u bytes\n",
+		    (unsigned)size);
+		abort();
+	}
+	return (rv);
+}
+
 /*
  * ssread - send stream read.
  *
@@ -160,7 +171,7 @@ print_block(char *buf, int length)
 int
 main(int argc, char *argv[])
 {
-	char *buf = malloc(INITIAL_BUFLEN);
+	char *buf = safe_malloc(SPA_MAXBLOCKSIZE);
 	uint64_t drr_record_count[DRR_NUMTYPES] = { 0 };
 	uint64_t total_records = 0;
 	dmu_replay_record_t thedrr;
@@ -308,9 +319,9 @@ main(int argc, char *argv[])
 				nvlist_t *nv;
 				int sz = drr->drr_payloadlen;
 
-				if (sz > INITIAL_BUFLEN) {
+				if (sz > SPA_MAXBLOCKSIZE) {
 					free(buf);
-					buf = malloc(sz);
+					buf = safe_malloc(sz);
 				}
 				(void) ssread(buf, sz, &zc);
 				if (ferror(send_stream))
diff --git a/cmd/ztest/ztest.c b/cmd/ztest/ztest.c
index 0602a7ec5..6b939bdb6 100644
--- a/cmd/ztest/ztest.c
+++ b/cmd/ztest/ztest.c
@@ -1040,9 +1040,14 @@ ztest_spa_get_ashift(void) {
 static int
 ztest_random_blocksize(void)
 {
-	// Choose a block size >= the ashift.
-	uint64_t block_shift =
-	    ztest_random(SPA_MAXBLOCKSHIFT - ztest_spa_get_ashift() + 1);
+	/*
+	 * Choose a block size >= the ashift.
+	 * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks.
+	 */
+	int maxbs = SPA_OLD_MAXBLOCKSHIFT;
+	if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE)
+		maxbs = 20;
+	uint64_t block_shift = ztest_random(maxbs - ztest_spa_get_ashift() + 1);
 	return (1 << (SPA_MINBLOCKSHIFT + block_shift));
 }
 
@@ -4972,7 +4977,7 @@ ztest_fault_inject(ztest_ds_t *zd, uint64_t id)
 	char *path0;
 	char *pathrand;
 	size_t fsize;
-	int bshift = SPA_MAXBLOCKSHIFT + 2;	/* don't scrog all labels */
+	int bshift = SPA_OLD_MAXBLOCKSHIFT + 2;	/* don't scrog all labels */
 	int iters = 1000;
 	int maxfaults;
 	int mirror_save;