Add d_off support for multiple filesystems.

The d_off field has been added to the dirent structure recently.
Currently filesystems don't support this feature.  Support has been
added and tested for zfs, ufs, ext2fs, fdescfs, msdosfs and unionfs.
A stub implementation is available for cd9660, nandfs, udf and
pseudofs but hasn't been tested.

Motivation for this feature: our usecase is for a userspace nfs server
(nfs-ganesha) with zfs.  At the moment we cache direntry offsets by
calling lseek once per entry, with this patch we can get the offset
directly from getdirentries(2) calls which provides a significant
speedup.

Submitted by:	Jack Halford <jack@gandi.net>
Reviewed by:	mckusick, pfg, rmacklem (previous versions)
Sponsored by:	Gandi.net
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D17917

8 files changed, 19 insertions, 0 deletions

diff --git a/sys/fs/cd9660/cd9660_vnops.c b/sys/fs/cd9660/cd9660_vnops.c
index 84cee260d066..ca291c7c5a3f 100644
--- a/sys/fs/cd9660/cd9660_vnops.c
+++ b/sys/fs/cd9660/cd9660_vnops.c
@@ -576,6 +576,8 @@ cd9660_readdir(ap)
 				entryoffsetinblock;
 
 		idp->curroff += reclen;
+		/* NOTE: d_off is the offset of *next* entry. */
+		idp->current.d_off = idp->curroff;
 
 		switch (imp->iso_ftype) {
 		case ISO_FTYPE_RRIP:
diff --git a/sys/fs/devfs/devfs_vnops.c b/sys/fs/devfs/devfs_vnops.c
index 5c0bf603ab9b..c046c4e88ad3 100644
--- a/sys/fs/devfs/devfs_vnops.c
+++ b/sys/fs/devfs/devfs_vnops.c
@@ -1381,6 +1381,8 @@ devfs_readdir(struct vop_readdir_args *ap)
 		if (dp->d_reclen > uio->uio_resid)
 			break;
 		dp->d_fileno = de->de_inode;
+		/* NOTE: d_off is the offset for the *next* entry. */
+		dp->d_off = off + dp->d_reclen;
 		if (off >= uio->uio_offset) {
 			error = vfs_read_dirent(ap, dp, off);
 			if (error)
diff --git a/sys/fs/ext2fs/ext2_lookup.c b/sys/fs/ext2fs/ext2_lookup.c
index 5de7b9a7a3b5..d5fc6f93542d 100644
--- a/sys/fs/ext2fs/ext2_lookup.c
+++ b/sys/fs/ext2fs/ext2_lookup.c
@@ -224,6 +224,8 @@ ext2_readdir(struct vop_readdir_args *ap)
 			dstdp.d_reclen = GENERIC_DIRSIZ(&dstdp);
 			bcopy(dp->e2d_name, dstdp.d_name, dstdp.d_namlen);
 			dstdp.d_name[dstdp.d_namlen] = '\0';
+			/* NOTE: d_off is the offset of the *next* entry. */
+			dstdp.d_off = offset + dp->e2d_reclen;
 			if (dstdp.d_reclen > uio->uio_resid) {
 				if (uio->uio_resid == startresid)
 					error = EINVAL;
diff --git a/sys/fs/fdescfs/fdesc_vnops.c b/sys/fs/fdescfs/fdesc_vnops.c
index 7d94ddf17a51..d1731938f062 100644
--- a/sys/fs/fdescfs/fdesc_vnops.c
+++ b/sys/fs/fdescfs/fdesc_vnops.c
@@ -574,6 +574,8 @@ fdesc_readdir(struct vop_readdir_args *ap)
 			dp->d_fileno = i + FD_DESC;
 			break;
 		}
+		/* NOTE: d_off is the offset of the *next* entry. */
+		dp->d_off = UIO_MX * (i + 1);
 		if (dp->d_namlen != 0) {
 			/*
 			 * And ship to userland
diff --git a/sys/fs/msdosfs/msdosfs_vnops.c b/sys/fs/msdosfs/msdosfs_vnops.c
index 0abcda4a99b4..e6172a25f6c0 100644
--- a/sys/fs/msdosfs/msdosfs_vnops.c
+++ b/sys/fs/msdosfs/msdosfs_vnops.c
@@ -1558,6 +1558,8 @@ msdosfs_readdir(struct vop_readdir_args *ap)
 					break;
 				}
 				dirbuf.d_reclen = GENERIC_DIRSIZ(&dirbuf);
+				/* NOTE: d_off is the offset of the *next* entry. */
+				dirbuf.d_off = offset + sizeof(struct direntry);
 				if (uio->uio_resid < dirbuf.d_reclen)
 					goto out;
 				error = uiomove(&dirbuf, dirbuf.d_reclen, uio);
@@ -1681,6 +1683,8 @@ msdosfs_readdir(struct vop_readdir_args *ap)
 				mbnambuf_flush(&nb, &dirbuf);
 			chksum = -1;
 			dirbuf.d_reclen = GENERIC_DIRSIZ(&dirbuf);
+			/* NOTE: d_off is the offset of the *next* entry. */
+			dirbuf.d_off = offset + sizeof(struct direntry);
 			if (uio->uio_resid < dirbuf.d_reclen) {
 				brelse(bp);
 				goto out;
diff --git a/sys/fs/nandfs/nandfs_vnops.c b/sys/fs/nandfs/nandfs_vnops.c
index 80e98c368cd7..2b9f5364f69b 100644
--- a/sys/fs/nandfs/nandfs_vnops.c
+++ b/sys/fs/nandfs/nandfs_vnops.c
@@ -1233,6 +1233,8 @@ nandfs_readdir(struct vop_readdir_args *ap)
 				dirent.d_namlen = name_len;
 				strncpy(dirent.d_name, ndirent->name, name_len);
 				dirent.d_reclen = GENERIC_DIRSIZ(&dirent);
+				/* NOTE: d_off is the offset of the *next* entry. */
+				dirent.d_off = diroffset + ndirent->rec_len;
 				DPRINTF(READDIR, ("copying `%*.*s`\n", name_len,
 				    name_len, dirent.d_name));
 			}
diff --git a/sys/fs/pseudofs/pseudofs_vnops.c b/sys/fs/pseudofs/pseudofs_vnops.c
index 664949492906..1ffcb4c8098e 100644
--- a/sys/fs/pseudofs/pseudofs_vnops.c
+++ b/sys/fs/pseudofs/pseudofs_vnops.c
@@ -830,6 +830,8 @@ pfs_readdir(struct vop_readdir_args *va)
 			pfsent->entry.d_name[i] = pn->pn_name[i];
 		pfsent->entry.d_name[i] = 0;
 		pfsent->entry.d_namlen = i;
+		/* NOTE: d_off is the offset of the *next* entry. */
+		pfsent->entry.d_off = offset + PFS_DELEN;
 		switch (pn->pn_type) {
 		case pfstype_procdir:
 			KASSERT(p != NULL,
diff --git a/sys/fs/udf/udf_vnops.c b/sys/fs/udf/udf_vnops.c
index 6e7706c5c7c5..30558cf86931 100644
--- a/sys/fs/udf/udf_vnops.c
+++ b/sys/fs/udf/udf_vnops.c
@@ -846,6 +846,7 @@ udf_readdir(struct vop_readdir_args *a)
 			dir.d_name[1] = '\0';
 			dir.d_namlen = 1;
 			dir.d_reclen = GENERIC_DIRSIZ(&dir);
+			dir.d_off = 1;
 			uiodir.dirent = &dir;
 			error = udf_uiodir(&uiodir, dir.d_reclen, uio, 1);
 			if (error)
@@ -858,6 +859,7 @@ udf_readdir(struct vop_readdir_args *a)
 			dir.d_name[2] = '\0';
 			dir.d_namlen = 2;
 			dir.d_reclen = GENERIC_DIRSIZ(&dir);
+			dir.d_off = 2;
 			uiodir.dirent = &dir;
 			error = udf_uiodir(&uiodir, dir.d_reclen, uio, 2);
 		} else {
@@ -867,6 +869,7 @@ udf_readdir(struct vop_readdir_args *a)
 			dir.d_type = (fid->file_char & UDF_FILE_CHAR_DIR) ?
 			    DT_DIR : DT_UNKNOWN;
 			dir.d_reclen = GENERIC_DIRSIZ(&dir);
+			dir.d_off = ds->this_off;
 			uiodir.dirent = &dir;
 			error = udf_uiodir(&uiodir, dir.d_reclen, uio,
 			    ds->this_off);