MFC r286705: 5960 zfs recv should prefetch indirect blocks

5925 zfs receive -o origin=

Reviewed by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Author: Paul Dagnelie <pcd@delphix.com>

While running 'zfs recv' we noticed that every 128th 8K block required a
read. We were seeing that restore_write() was calling dmu_tx_hold_write()
and the indirect block was not cached. We should prefetch upcoming indirect
blocks to avoid having to go to disk and blocking the restore_write().

Allow an incremental send stream to be received as a clone, even if the
stream does not mark it as a clone.

1 files changed, 9 insertions, 9 deletions

diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dnode.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dnode.c
index 4af50f5..697a45d 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dnode.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dnode.c
@@ -1116,7 +1116,7 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag,
 		drop_struct_lock = TRUE;
 	}
 
-	blk = dbuf_whichblock(mdn, object * sizeof (dnode_phys_t));
+	blk = dbuf_whichblock(mdn, 0, object * sizeof (dnode_phys_t));
 
 	db = dbuf_hold(mdn, blk, FTAG);
 	if (drop_struct_lock)
@@ -1413,7 +1413,7 @@ dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx)
 		goto fail;
 
 	/* resize the old block */
-	err = dbuf_hold_impl(dn, 0, 0, TRUE, FTAG, &db);
+	err = dbuf_hold_impl(dn, 0, 0, TRUE, FALSE, FTAG, &db);
 	if (err == 0)
 		dbuf_new_size(db, size, tx);
 	else if (err != ENOENT)
@@ -1586,8 +1586,8 @@ dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx)
 		ASSERT3U(blkoff + head, ==, blksz);
 		if (len < head)
 			head = len;
-		if (dbuf_hold_impl(dn, 0, dbuf_whichblock(dn, off), TRUE,
-		    FTAG, &db) == 0) {
+		if (dbuf_hold_impl(dn, 0, dbuf_whichblock(dn, 0, off),
+		    TRUE, FALSE, FTAG, &db) == 0) {
 			caddr_t data;
 
 			/* don't dirty if it isn't on disk and isn't dirty */
@@ -1624,8 +1624,8 @@ dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx)
 	if (tail) {
 		if (len < tail)
 			tail = len;
-		if (dbuf_hold_impl(dn, 0, dbuf_whichblock(dn, off+len),
-		    TRUE, FTAG, &db) == 0) {
+		if (dbuf_hold_impl(dn, 0, dbuf_whichblock(dn, 0, off+len),
+		    TRUE, FALSE, FTAG, &db) == 0) {
 			/* don't dirty if not on disk and not dirty */
 			if (db->db_last_dirty ||
 			    (db->db_blkptr && !BP_IS_HOLE(db->db_blkptr))) {
@@ -1854,7 +1854,7 @@ dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx)
  */
 static int
 dnode_next_offset_level(dnode_t *dn, int flags, uint64_t *offset,
-	int lvl, uint64_t blkfill, uint64_t txg)
+    int lvl, uint64_t blkfill, uint64_t txg)
 {
 	dmu_buf_impl_t *db = NULL;
 	void *data = NULL;
@@ -1876,8 +1876,8 @@ dnode_next_offset_level(dnode_t *dn, int flags, uint64_t *offset,
 		epb = dn->dn_phys->dn_nblkptr;
 		data = dn->dn_phys->dn_blkptr;
 	} else {
-		uint64_t blkid = dbuf_whichblock(dn, *offset) >> (epbs * lvl);
-		error = dbuf_hold_impl(dn, lvl, blkid, TRUE, FTAG, &db);
+		uint64_t blkid = dbuf_whichblock(dn, lvl, *offset);
+		error = dbuf_hold_impl(dn, lvl, blkid, TRUE, FALSE, FTAG, &db);
 		if (error) {
 			if (error != ENOENT)
 				return (error);