Illumos 5162 - zfs recv should use loaned arc buffer to avoid copy

5162 zfs recv should use loaned arc buffer to avoid copy Reviewed by: Christopher Siden <[email protected]> Reviewed by: George Wilson <[email protected]> Reviewed by: Bayard Bell <[email protected]> Reviewed by: Richard Elling <[email protected]> Approved by: Garrett D'Amore <[email protected]> References: https://www.illumos.org/issues/5162 illumos/illumos-gate@8a90470 Porting notes: Fix spelling error 's/arena/area/' in dmu.c. In restore_write() declare bonus and abuf at the top of the function. Ported by: Turbo Fredriksson <[email protected]> Signed-off-by: Brian Behlendorf <[email protected]> Closes #2696
openzfs · Oct 21, 2014 · 88904bb · 88904bb
1 parent 4b20a6f
commit 88904bb
Show file tree

Hide file tree

Showing 2 changed files with 38 additions and 18 deletions.
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
@@ -1387,7 +1387,14 @@ dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf,
 	rw_exit(&dn->dn_struct_rwlock);
 	DB_DNODE_EXIT(dbuf);
 
-	if (offset == db->db.db_offset && blksz == db->db.db_size) {
+	/*
+	 * We can only assign if the offset is aligned, the arc buf is the
+	 * same size as the dbuf, and the dbuf is not metadata.  It
+	 * can't be metadata because the loaned arc buf comes from the
+	 * user-data kmem area.
+	 */
+	if (offset == db->db.db_offset && blksz == db->db.db_size &&
+	    DBUF_GET_BUFC_TYPE(db) == ARC_BUFC_DATA) {
 		dbuf_assign_arcbuf(db, buf, tx);
 		dbuf_rele(db, FTAG);
 	} else {

diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c
@@ -1206,19 +1206,21 @@ free_guid_map_onexit(void *arg)
 }
 
 static void *
-restore_read(struct restorearg *ra, int len)
+restore_read(struct restorearg *ra, int len, char *buf)
 {
-	void *rv;
 	int done = 0;
 
+	if (buf == NULL)
+		buf = ra->buf;
+
 	/* some things will require 8-byte alignment, so everything must */
 	ASSERT0(len % 8);
 
 	while (done < len) {
 		ssize_t resid;
 
 		ra->err = vn_rdwr(UIO_READ, ra->vp,
-		    (caddr_t)ra->buf + done, len - done,
+		    buf + done, len - done,
 		    ra->voff, UIO_SYSSPACE, FAPPEND,
 		    RLIM64_INFINITY, CRED(), &resid);
 
@@ -1231,12 +1233,11 @@ restore_read(struct restorearg *ra, int len)
 	}
 
 	ASSERT3U(done, ==, len);
-	rv = ra->buf;
 	if (ra->byteswap)
-		fletcher_4_incremental_byteswap(rv, len, &ra->cksum);
+		fletcher_4_incremental_byteswap(buf, len, &ra->cksum);
 	else
-		fletcher_4_incremental_native(rv, len, &ra->cksum);
-	return (rv);
+		fletcher_4_incremental_native(buf, len, &ra->cksum);
+	return (buf);
 }
 
 noinline static void
@@ -1367,7 +1368,7 @@ restore_object(struct restorearg *ra, objset_t *os, struct drr_object *drro)
 	object = err == 0 ? drro->drr_object : DMU_NEW_OBJECT;
 
 	if (drro->drr_bonuslen) {
-		data = restore_read(ra, P2ROUNDUP(drro->drr_bonuslen, 8));
+		data = restore_read(ra, P2ROUNDUP(drro->drr_bonuslen, 8), NULL);
 		if (ra->err != 0)
 			return (ra->err);
 	}
@@ -1473,26 +1474,38 @@ restore_write(struct restorearg *ra, objset_t *os,
     struct drr_write *drrw)
 {
 	dmu_tx_t *tx;
+	dmu_buf_t *bonus;
+	arc_buf_t *abuf;
 	void *data;
 	int err;
 
 	if (drrw->drr_offset + drrw->drr_length < drrw->drr_offset ||
 	    !DMU_OT_IS_VALID(drrw->drr_type))
 		return (SET_ERROR(EINVAL));
 
-	data = restore_read(ra, drrw->drr_length);
-	if (data == NULL)
-		return (ra->err);
-
 	if (dmu_object_info(os, drrw->drr_object, NULL) != 0)
 		return (SET_ERROR(EINVAL));
 
+	if (dmu_bonus_hold(os, drrw->drr_object, FTAG, &bonus) != 0)
+		return (SET_ERROR(EINVAL));
+
+	abuf = dmu_request_arcbuf(bonus, drrw->drr_length);
+
+	data = restore_read(ra, drrw->drr_length, abuf->b_data);
+	if (data == NULL) {
+		dmu_return_arcbuf(abuf);
+		dmu_buf_rele(bonus, FTAG);
+		return (ra->err);
+	}
+
 	tx = dmu_tx_create(os);
 
 	dmu_tx_hold_write(tx, drrw->drr_object,
 	    drrw->drr_offset, drrw->drr_length);
 	err = dmu_tx_assign(tx, TXG_WAIT);
 	if (err != 0) {
+		dmu_return_arcbuf(abuf);
+		dmu_buf_rele(bonus, FTAG);
 		dmu_tx_abort(tx);
 		return (err);
 	}
@@ -1501,9 +1514,9 @@ restore_write(struct restorearg *ra, objset_t *os,
 		    DMU_OT_BYTESWAP(drrw->drr_type);
 		dmu_ot_byteswap[byteswap].ob_func(data, drrw->drr_length);
 	}
-	dmu_write(os, drrw->drr_object,
-	    drrw->drr_offset, drrw->drr_length, data, tx);
+	dmu_assign_arcbuf(bonus, drrw->drr_offset, abuf, tx);
 	dmu_tx_commit(tx);
+	dmu_buf_rele(bonus, FTAG);
 	return (0);
 }
 
@@ -1585,7 +1598,7 @@ restore_write_embedded(struct restorearg *ra, objset_t *os,
 	if (drrwnp->drr_compression >= ZIO_COMPRESS_FUNCTIONS)
 		return (EINVAL);
 
-	data = restore_read(ra, P2ROUNDUP(drrwnp->drr_psize, 8));
+	data = restore_read(ra, P2ROUNDUP(drrwnp->drr_psize, 8), NULL);
 	if (data == NULL)
 		return (ra->err);
 
@@ -1620,7 +1633,7 @@ restore_spill(struct restorearg *ra, objset_t *os, struct drr_spill *drrs)
 	    drrs->drr_length > SPA_MAXBLOCKSIZE)
 		return (SET_ERROR(EINVAL));
 
-	data = restore_read(ra, drrs->drr_length);
+	data = restore_read(ra, drrs->drr_length, NULL);
 	if (data == NULL)
 		return (ra->err);
 
@@ -1761,7 +1774,7 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp,
 	 */
 	pcksum = ra.cksum;
 	while (ra.err == 0 &&
-	    NULL != (drr = restore_read(&ra, sizeof (*drr)))) {
+	    NULL != (drr = restore_read(&ra, sizeof (*drr), NULL))) {
 		if (issig(JUSTLOOKING) && issig(FORREAL)) {
 			ra.err = SET_ERROR(EINTR);
 			goto out;