Illumos 5175 - implement dmu_read_uio_dbuf() to improve cached read performance

5175 implement dmu_read_uio_dbuf() to improve cached read performance
Reviewed by: Adam Leventhal <[email protected]>
Reviewed by: Alex Reece <[email protected]>
Reviewed by: George Wilson <[email protected]>
Reviewed by: Richard Elling <[email protected]>
Approved by: Robert Mustacchi <[email protected]>

References:
  https://www.illumos.org/issues/5175
  https://github.com/illumos/illumos-gate/commit/f8554bb

Porting notes:

This patch doesn't include the changes for the COMSTAR (Common
Multiprotocol SCSI Target) - since it's not available for ZoL.

http://thegreyblog.blogspot.co.at/2010/02/setting-up-solaris-comstar-and.html

Ported by: kernelOfTruth <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #3392

2 files changed, 77 insertions, 11 deletions

diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
index 81f434380..7e8328e77 100644
--- a/module/zfs/dmu.c
+++ b/module/zfs/dmu.c
@@ -1199,8 +1199,8 @@ dmu_write_req(objset_t *os, uint64_t object, struct request *req, dmu_tx_t *tx)
 	return (err);
 }
 
-int
-dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
+static int
+dmu_read_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size)
 {
 	dmu_buf_t **dbp;
 	int numbufs, i, err;
@@ -1210,8 +1210,8 @@ dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
 	 * NB: we could do this block-at-a-time, but it's nice
 	 * to be reading in parallel.
 	 */
-	err = dmu_buf_hold_array(os, object, uio->uio_loffset, size, TRUE, FTAG,
-	    &numbufs, &dbp);
+	err = dmu_buf_hold_array_by_dnode(dn, uio->uio_loffset, size,
+	    TRUE, FTAG, &numbufs, &dbp, 0);
 	if (err)
 		return (err);
 
@@ -1253,6 +1253,58 @@ dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
 	return (err);
 }
 
+/*
+ * Read 'size' bytes into the uio buffer.
+ * From object zdb->db_object.
+ * Starting at offset uio->uio_loffset.
+ *
+ * If the caller already has a dbuf in the target object
+ * (e.g. its bonus buffer), this routine is faster than dmu_read_uio(),
+ * because we don't have to find the dnode_t for the object.
+ */
+int
+dmu_read_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size)
+{
+	dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb;
+	dnode_t *dn;
+	int err;
+
+	if (size == 0)
+		return (0);
+
+	DB_DNODE_ENTER(db);
+	dn = DB_DNODE(db);
+	err = dmu_read_uio_dnode(dn, uio, size);
+	DB_DNODE_EXIT(db);
+
+	return (err);
+}
+
+/*
+ * Read 'size' bytes into the uio buffer.
+ * From the specified object
+ * Starting at offset uio->uio_loffset.
+ */
+int
+dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
+{
+	dnode_t *dn;
+	int err;
+
+	if (size == 0)
+		return (0);
+
+	err = dnode_hold(os, object, FTAG, &dn);
+	if (err)
+		return (err);
+
+	err = dmu_read_uio_dnode(dn, uio, size);
+
+	dnode_rele(dn, FTAG);
+
+	return (err);
+}
+
 static int
 dmu_write_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size, dmu_tx_t *tx)
 {
@@ -1305,6 +1357,15 @@ dmu_write_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size, dmu_tx_t *tx)
 	return (err);
 }
 
+/*
+ * Write 'size' bytes from the uio buffer.
+ * To object zdb->db_object.
+ * Starting at offset uio->uio_loffset.
+ *
+ * If the caller already has a dbuf in the target object
+ * (e.g. its bonus buffer), this routine is faster than dmu_write_uio(),
+ * because we don't have to find the dnode_t for the object.
+ */
 int
 dmu_write_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size,
     dmu_tx_t *tx)
@@ -1324,6 +1385,11 @@ dmu_write_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size,
 	return (err);
 }
 
+/*
+ * Write 'size' bytes from the uio buffer.
+ * To the specified object.
+ * Starting at offset uio->uio_loffset.
+ */
 int
 dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size,
     dmu_tx_t *tx)
diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
index 19a4132e4..7780fe902 100644
--- a/module/zfs/zfs_vnops.c
+++ b/module/zfs/zfs_vnops.c
@@ -376,7 +376,6 @@ mappedread(struct inode *ip, int nbytes, uio_t *uio)
 	struct address_space *mp = ip->i_mapping;
 	struct page *pp;
 	znode_t *zp = ITOZ(ip);
-	objset_t *os = ITOZSB(ip)->z_os;
 	int64_t	start, off;
 	uint64_t bytes;
 	int len = nbytes;
@@ -403,7 +402,8 @@ mappedread(struct inode *ip, int nbytes, uio_t *uio)
 			unlock_page(pp);
 			page_cache_release(pp);
 		} else {
-			error = dmu_read_uio(os, zp->z_id, uio, bytes);
+			error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
+			    uio, bytes);
 		}
 
 		len -= bytes;
@@ -440,7 +440,6 @@ zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 {
 	znode_t		*zp = ITOZ(ip);
 	zfs_sb_t	*zsb = ITOZSB(ip);
-	objset_t	*os;
 	ssize_t		n, nbytes;
 	int		error = 0;
 	rl_t		*rl;
@@ -450,7 +449,6 @@ zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 
 	ZFS_ENTER(zsb);
 	ZFS_VERIFY_ZP(zp);
-	os = zsb->z_os;
 
 	if (zp->z_pflags & ZFS_AV_QUARANTINED) {
 		ZFS_EXIT(zsb);
@@ -531,10 +529,12 @@ zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 		nbytes = MIN(n, zfs_read_chunk_size -
 		    P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
 
-		if (zp->z_is_mapped && !(ioflag & O_DIRECT))
+		if (zp->z_is_mapped && !(ioflag & O_DIRECT)) {
 			error = mappedread(ip, nbytes, uio);
-		else
-			error = dmu_read_uio(os, zp->z_id, uio, nbytes);
+		} else {
+			error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
+			    uio, nbytes);
+		}
 
 		if (error) {
 			/* convert checksum errors into IO errors */