FreeBSD: Implement xattr=sa

FreeBSD historically has not cared about the xattr property; it was
always treated as xattr=on.  With xattr=on, xattrs are stored as files
in a hidden xattr directory.  With xattr=sa, xattrs are stored as
system attributes and get cached in nvlists during xattr operations.
This makes SA xattrs simpler and more efficient to manipulate.  FreeBSD
needs to implement the SA xattr operations for feature parity with
Linux and to ensure that SA xattrs are accessible when migrated or
replicated from Linux.

Following the example set by Linux, refactor our existing extattr vnops
to split off the parts handling dir style xattrs, and add the
corresponding SA handling parts.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Ryan Moeller <ryan@iXsystems.com>
Closes #11997

man/man8/zfsprops.8

@@ -38,7 +38,7 @@
 .\" Copyright 2019 Joyent, Inc.
 .\" Copyright (c) 2019, Kjeld Schouten-Lebbing
 .\"
-.Dd September 1, 2020
+.Dd May 5, 2021
 .Dt ZFSPROPS 8
 .Os
 .Sh NAME
@@ -1778,7 +1778,7 @@ The default value is
 This property is not used on Linux.
 .It Sy xattr Ns = Ns Sy on Ns | Ns Sy off Ns | Ns Sy sa
 Controls whether extended attributes are enabled for this file system. Two
-styles of extended attributes are supported either directory based or system
+styles of extended attributes are supported: either directory based or system
 attribute based.
 .Pp
 The default value of
@@ -1790,7 +1790,7 @@ can be set on a file. Although under Linux the
 and
 .Xr setxattr 2
 system calls limit the maximum size to 64K. This is the most compatible
-style of extended attribute and is supported by all OpenZFS implementations.
+style of extended attribute and is supported by all ZFS implementations.
 .Pp
 System attribute based xattrs can be enabled by setting the value to
 .Sy sa .
@@ -1803,6 +1803,9 @@ based xattr. System attribute based extended attributes are not accessible
 on platforms which do not support the
 .Sy xattr=sa
 feature.
+OpenZFS supports
+.Sy xattr=sa
+on both FreeBSD and Linux.
 .Pp
 The use of system attribute based xattrs is strongly encouraged for users of
 SELinux or POSIX ACLs. Both of these features heavily rely on extended

module/os/freebsd/zfs/zfs_vfsops.c

@@ -846,6 +846,10 @@ zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os)
 		    &sa_obj);
 		if (error != 0)
 			return (error);
+
+		error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val);
+		if (error == 0 && val == ZFS_XATTR_SA)
+			zfsvfs->z_xattr_sa = B_TRUE;
 	}
 
 	error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,

module/os/freebsd/zfs/zfs_vnops_os.c

@@ -5227,6 +5227,29 @@ zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
 	return (0);
 }
 
+static int
+zfs_ensure_xattr_cached(znode_t *zp)
+{
+	int error = 0;
+
+	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
+
+	if (zp->z_xattr_cached != NULL)
+		return (0);
+
+	if (rw_write_held(&zp->z_xattr_lock))
+		return (zfs_sa_get_xattr(zp));
+
+	if (!rw_tryupgrade(&zp->z_xattr_lock)) {
+		rw_exit(&zp->z_xattr_lock);
+		rw_enter(&zp->z_xattr_lock, RW_WRITER);
+	}
+	if (zp->z_xattr_cached == NULL)
+		error = zfs_sa_get_xattr(zp);
+	rw_downgrade(&zp->z_xattr_lock);
+	return (error);
+}
+
 #ifndef _SYS_SYSPROTO_H_
 struct vop_getextattr {
 	IN struct vnode *a_vp;
@@ -5239,26 +5262,85 @@ struct vop_getextattr {
 };
 #endif
 
-/*
- * Vnode operating to retrieve a named extended attribute.
- */
 static int
-zfs_getextattr(struct vop_getextattr_args *ap)
+zfs_getextattr_dir(struct vop_getextattr_args *ap, const char *attrname)
 {
-	zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
 	struct thread *td = ap->a_td;
 	struct nameidata nd;
-	char attrname[255];
 	struct vattr va;
 	vnode_t *xvp = NULL, *vp;
 	int error, flags;
 
+	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
+	    LOOKUP_XATTR, B_FALSE);
+	if (error != 0)
+		return (error);
+
+	flags = FREAD;
+	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
+	    xvp, td);
+	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_INVFS, ap->a_cred, NULL);
+	vp = nd.ni_vp;
+	NDFREE(&nd, NDF_ONLY_PNBUF);
+	if (error != 0)
+		return (error);
+
+	if (ap->a_size != NULL) {
+		error = VOP_GETATTR(vp, &va, ap->a_cred);
+		if (error == 0)
+			*ap->a_size = (size_t)va.va_size;
+	} else if (ap->a_uio != NULL)
+		error = VOP_READ(vp, ap->a_uio, IO_UNIT, ap->a_cred);
+
+	VOP_UNLOCK1(vp);
+	vn_close(vp, flags, ap->a_cred, td);
+	return (error);
+}
+
+static int
+zfs_getextattr_sa(struct vop_getextattr_args *ap, const char *attrname)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	uchar_t *nv_value;
+	uint_t nv_size;
+	int error;
+
+	error = zfs_ensure_xattr_cached(zp);
+	if (error != 0)
+		return (error);
+
+	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
+	ASSERT3P(zp->z_xattr_cached, !=, NULL);
+
+	error = nvlist_lookup_byte_array(zp->z_xattr_cached, attrname,
+	    &nv_value, &nv_size);
+	if (error)
+		return (error);
+
+	if (ap->a_size != NULL)
+		*ap->a_size = nv_size;
+	else if (ap->a_uio != NULL)
+		error = uiomove(nv_value, nv_size, ap->a_uio);
+
+	return (error);
+}
+
+/*
+ * Vnode operation to retrieve a named extended attribute.
+ */
+static int
+zfs_getextattr(struct vop_getextattr_args *ap)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	zfsvfs_t *zfsvfs = ZTOZSB(zp);
+	char attrname[EXTATTR_MAXNAMELEN+1];
+	int error;
+
 	/*
 	 * If the xattr property is off, refuse the request.
 	 */
-	if (!(zfsvfs->z_flags & ZSB_XATTR)) {
+	if (!(zfsvfs->z_flags & ZSB_XATTR))
 		return (SET_ERROR(EOPNOTSUPP));
-	}
 
 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
 	    ap->a_cred, ap->a_td, VREAD);
@@ -5270,38 +5352,18 @@ zfs_getextattr(struct vop_getextattr_args *ap)
 	if (error != 0)
 		return (error);
 
+	error = ENOENT;
 	ZFS_ENTER(zfsvfs);
-
+	ZFS_VERIFY_ZP(zp)
-	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
+	rw_enter(&zp->z_xattr_lock, RW_READER);
-	    LOOKUP_XATTR, B_FALSE);
+	if (zfsvfs->z_use_sa && zp->z_is_sa)
-	if (error != 0) {
+		error = zfs_getextattr_sa(ap, attrname);
-		ZFS_EXIT(zfsvfs);
+	if (error == ENOENT)
-		return (error);
+		error = zfs_getextattr_dir(ap, attrname);
-	}
+	rw_exit(&zp->z_xattr_lock);
-
-	flags = FREAD;
-	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
-	    xvp, td);
-	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_INVFS, ap->a_cred, NULL);
-	vp = nd.ni_vp;
-	NDFREE(&nd, NDF_ONLY_PNBUF);
-	if (error != 0) {
-		ZFS_EXIT(zfsvfs);
-		if (error == ENOENT)
-			error = ENOATTR;
-		return (error);
-	}
-
-	if (ap->a_size != NULL) {
-		error = VOP_GETATTR(vp, &va, ap->a_cred);
-		if (error == 0)
-			*ap->a_size = (size_t)va.va_size;
-	} else if (ap->a_uio != NULL)
-		error = VOP_READ(vp, ap->a_uio, IO_UNIT, ap->a_cred);
-
-	VOP_UNLOCK1(vp);
-	vn_close(vp, flags, ap->a_cred, td);
 	ZFS_EXIT(zfsvfs);
+	if (error == ENOENT)
+		error = SET_ERROR(ENOATTR);
 	return (error);
 }
 
@@ -5315,54 +5377,25 @@ struct vop_deleteextattr {
 };
 #endif
 
-/*
- * Vnode operation to remove a named attribute.
- */
 static int
-zfs_deleteextattr(struct vop_deleteextattr_args *ap)
+zfs_deleteextattr_dir(struct vop_deleteextattr_args *ap, const char *attrname)
 {
-	zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
 	struct thread *td = ap->a_td;
 	struct nameidata nd;
-	char attrname[255];
 	vnode_t *xvp = NULL, *vp;
 	int error;
 
-	/*
-	 * If the xattr property is off, refuse the request.
-	 */
-	if (!(zfsvfs->z_flags & ZSB_XATTR)) {
-		return (SET_ERROR(EOPNOTSUPP));
-	}
-
-	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
-	    ap->a_cred, ap->a_td, VWRITE);
-	if (error != 0)
-		return (error);
-
-	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
-	    sizeof (attrname));
-	if (error != 0)
-		return (error);
-
-	ZFS_ENTER(zfsvfs);
-
 	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
 	    LOOKUP_XATTR, B_FALSE);
-	if (error != 0) {
+	if (error != 0)
-		ZFS_EXIT(zfsvfs);
 		return (error);
-	}
 
 	NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF,
 	    UIO_SYSSPACE, attrname, xvp, td);
 	error = namei(&nd);
 	vp = nd.ni_vp;
 	if (error != 0) {
-		ZFS_EXIT(zfsvfs);
 		NDFREE(&nd, NDF_ONLY_PNBUF);
-		if (error == ENOENT)
-			error = ENOATTR;
 		return (error);
 	}
 
@@ -5374,11 +5407,90 @@ zfs_deleteextattr(struct vop_deleteextattr_args *ap)
 		vrele(vp);
 	else
 		vput(vp);
-	ZFS_EXIT(zfsvfs);
 
 	return (error);
 }
 
+static int
+zfs_deleteextattr_sa(struct vop_deleteextattr_args *ap, const char *attrname)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	nvlist_t *nvl;
+	int error;
+
+	error = zfs_ensure_xattr_cached(zp);
+	if (error != 0)
+		return (error);
+
+	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
+	ASSERT3P(zp->z_xattr_cached, !=, NULL);
+
+	nvl = zp->z_xattr_cached;
+	error = nvlist_remove(nvl, attrname, DATA_TYPE_BYTE_ARRAY);
+	if (error == 0)
+		error = zfs_sa_set_xattr(zp);
+	if (error != 0) {
+		zp->z_xattr_cached = NULL;
+		nvlist_free(nvl);
+	}
+	return (error);
+}
+
+/*
+ * Vnode operation to remove a named attribute.
+ */
+static int
+zfs_deleteextattr(struct vop_deleteextattr_args *ap)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	zfsvfs_t *zfsvfs = ZTOZSB(zp);
+	char attrname[EXTATTR_MAXNAMELEN+1];
+	int error;
+
+	/*
+	 * If the xattr property is off, refuse the request.
+	 */
+	if (!(zfsvfs->z_flags & ZSB_XATTR))
+		return (SET_ERROR(EOPNOTSUPP));
+
+	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
+	    ap->a_cred, ap->a_td, VWRITE);
+	if (error != 0)
+		return (error);
+
+	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
+	    sizeof (attrname));
+	if (error != 0)
+		return (error);
+
+	size_t size = 0;
+	struct vop_getextattr_args vga = {
+		.a_vp = ap->a_vp,
+		.a_size = &size,
+		.a_cred = ap->a_cred,
+		.a_td = ap->a_td,
+	};
+	error = ENOENT;
+	ZFS_ENTER(zfsvfs);
+	ZFS_VERIFY_ZP(zp);
+	rw_enter(&zp->z_xattr_lock, RW_WRITER);
+	if (zfsvfs->z_use_sa && zp->z_is_sa) {
+		error = zfs_getextattr_sa(&vga, attrname);
+		if (error == 0)
+			error = zfs_deleteextattr_sa(ap, attrname);
+	}
+	if (error == ENOENT) {
+		error = zfs_getextattr_dir(&vga, attrname);
+		if (error == 0)
+			error = zfs_deleteextattr_dir(ap, attrname);
+	}
+	rw_exit(&zp->z_xattr_lock);
+	ZFS_EXIT(zfsvfs);
+	if (error == ENOENT)
+		error = SET_ERROR(ENOATTR);
+	return (error);
+}
+
 #ifndef _SYS_SYSPROTO_H_
 struct vop_setextattr {
 	IN struct vnode *a_vp;
@@ -5390,56 +5502,28 @@ struct vop_setextattr {
 };
 #endif
 
-/*
- * Vnode operation to set a named attribute.
- */
 static int
-zfs_setextattr(struct vop_setextattr_args *ap)
+zfs_setextattr_dir(struct vop_setextattr_args *ap, const char *attrname)
 {
-	zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
 	struct thread *td = ap->a_td;
 	struct nameidata nd;
-	char attrname[255];
 	struct vattr va;
 	vnode_t *xvp = NULL, *vp;
 	int error, flags;
 
-	/*
-	 * If the xattr property is off, refuse the request.
-	 */
-	if (!(zfsvfs->z_flags & ZSB_XATTR)) {
-		return (SET_ERROR(EOPNOTSUPP));
-	}
-
-	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
-	    ap->a_cred, ap->a_td, VWRITE);
-	if (error != 0)
-		return (error);
-	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
-	    sizeof (attrname));
-	if (error != 0)
-		return (error);
-
-	ZFS_ENTER(zfsvfs);
-
 	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
 	    LOOKUP_XATTR | CREATE_XATTR_DIR, B_FALSE);
-	if (error != 0) {
+	if (error != 0)
-		ZFS_EXIT(zfsvfs);
 		return (error);
-	}
 
 	flags = FFLAGS(O_WRONLY | O_CREAT);
-	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
+	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname, xvp, td);
-	    xvp, td);
 	error = vn_open_cred(&nd, &flags, 0600, VN_OPEN_INVFS, ap->a_cred,
 	    NULL);
 	vp = nd.ni_vp;
 	NDFREE(&nd, NDF_ONLY_PNBUF);
-	if (error != 0) {
+	if (error != 0)
-		ZFS_EXIT(zfsvfs);
 		return (error);
-	}
 
 	VATTR_NULL(&va);
 	va.va_size = 0;
@@ -5449,6 +5533,102 @@ zfs_setextattr(struct vop_setextattr_args *ap)
 
 	VOP_UNLOCK1(vp);
 	vn_close(vp, flags, ap->a_cred, td);
+	return (error);
+}
+
+static int
+zfs_setextattr_sa(struct vop_setextattr_args *ap, const char *attrname)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	nvlist_t *nvl;
+	size_t sa_size;
+	int error;
+
+	error = zfs_ensure_xattr_cached(zp);
+	if (error != 0)
+		return (error);
+
+	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
+	ASSERT3P(zp->z_xattr_cached, !=, NULL);
+
+	nvl = zp->z_xattr_cached;
+	size_t entry_size = ap->a_uio->uio_resid;
+	if (entry_size > DXATTR_MAX_ENTRY_SIZE)
+		return (SET_ERROR(EFBIG));
+	error = nvlist_size(nvl, &sa_size, NV_ENCODE_XDR);
+	if (error != 0)
+		return (error);
+	if (sa_size > DXATTR_MAX_SA_SIZE)
+		return (SET_ERROR(EFBIG));
+	uchar_t *buf = kmem_alloc(entry_size, KM_SLEEP);
+	error = uiomove(buf, entry_size, ap->a_uio);
+	if (error == 0)
+		error = nvlist_add_byte_array(nvl, attrname, buf, entry_size);
+	kmem_free(buf, entry_size);
+	if (error == 0)
+		error = zfs_sa_set_xattr(zp);
+	if (error != 0) {
+		zp->z_xattr_cached = NULL;
+		nvlist_free(nvl);
+	}
+	return (error);
+}
+
+/*
+ * Vnode operation to set a named attribute.
+ */
+static int
+zfs_setextattr(struct vop_setextattr_args *ap)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	zfsvfs_t *zfsvfs = ZTOZSB(zp);
+	char attrname[EXTATTR_MAXNAMELEN+1];
+	int error;
+
+	/*
+	 * If the xattr property is off, refuse the request.
+	 */
+	if (!(zfsvfs->z_flags & ZSB_XATTR))
+		return (SET_ERROR(EOPNOTSUPP));
+
+	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
+	    ap->a_cred, ap->a_td, VWRITE);
+	if (error != 0)
+		return (error);
+
+	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
+	    sizeof (attrname));
+	if (error != 0)
+		return (error);
+
+	struct vop_deleteextattr_args vda = {
+		.a_vp = ap->a_vp,
+		.a_cred = ap->a_cred,
+		.a_td = ap->a_td,
+	};
+	error = ENOENT;
+	ZFS_ENTER(zfsvfs);
+	ZFS_VERIFY_ZP(zp);
+	rw_enter(&zp->z_xattr_lock, RW_WRITER);
+	if (zfsvfs->z_use_sa && zp->z_is_sa && zfsvfs->z_xattr_sa) {
+		error = zfs_setextattr_sa(ap, attrname);
+		if (error == 0)
+			/*
+			 * Successfully put into SA, we need to clear the one
+			 * in dir if present.
+			 */
+			zfs_deleteextattr_dir(&vda, attrname);
+	}
+	if (error) {
+		error = zfs_setextattr_dir(ap, attrname);
+		if (error == 0)
+			/*
+			 * Successfully put into dir, we need to clear the one
+			 * in SA if present.
+			 */
+			zfs_deleteextattr_sa(&vda, attrname);
+	}
+	rw_exit(&zp->z_xattr_lock);
 	ZFS_EXIT(zfsvfs);
 	return (error);
 }
@@ -5464,55 +5644,20 @@ struct vop_listextattr {
 };
 #endif
 
-/*
- * Vnode operation to retrieve extended attributes on a vnode.
- */
 static int
-zfs_listextattr(struct vop_listextattr_args *ap)
+zfs_listextattr_dir(struct vop_listextattr_args *ap, const char *attrprefix)
 {
-	zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
 	struct thread *td = ap->a_td;
 	struct nameidata nd;
-	char attrprefix[16];
 	uint8_t dirbuf[sizeof (struct dirent)];
-	struct dirent *dp;
 	struct iovec aiov;
 	struct uio auio;
-	size_t *sizep = ap->a_size;
-	size_t plen;
 	vnode_t *xvp = NULL, *vp;
-	int done, error, eof, pos;
+	int error, eof;
-	zfs_uio_t uio;
-
-	zfs_uio_init(&uio, ap->a_uio);
-
-	/*
-	 * If the xattr property is off, refuse the request.
-	 */
-	if (!(zfsvfs->z_flags & ZSB_XATTR)) {
-		return (SET_ERROR(EOPNOTSUPP));
-	}
-
-	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
-	    ap->a_cred, ap->a_td, VREAD);
-	if (error != 0)
-		return (error);
-
-	error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
-	    sizeof (attrprefix));
-	if (error != 0)
-		return (error);
-	plen = strlen(attrprefix);
-
-	ZFS_ENTER(zfsvfs);
-
-	if (sizep != NULL)
-		*sizep = 0;
 
 	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
 	    LOOKUP_XATTR, B_FALSE);
 	if (error != 0) {
-		ZFS_EXIT(zfsvfs);
 		/*
 		 * ENOATTR means that the EA directory does not yet exist,
 		 * i.e. there are no extended attributes there.
@@ -5527,10 +5672,8 @@ zfs_listextattr(struct vop_listextattr_args *ap)
 	error = namei(&nd);
 	vp = nd.ni_vp;
 	NDFREE(&nd, NDF_ONLY_PNBUF);
-	if (error != 0) {
+	if (error != 0)
-		ZFS_EXIT(zfsvfs);
 		return (error);
-	}
 
 	auio.uio_iov = &aiov;
 	auio.uio_iovcnt = 1;
@@ -5539,18 +5682,18 @@ zfs_listextattr(struct vop_listextattr_args *ap)
 	auio.uio_rw = UIO_READ;
 	auio.uio_offset = 0;
 
-	do {
+	size_t plen = strlen(attrprefix);
-		uint8_t nlen;
 
+	do {
 		aiov.iov_base = (void *)dirbuf;
 		aiov.iov_len = sizeof (dirbuf);
 		auio.uio_resid = sizeof (dirbuf);
 		error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
-		done = sizeof (dirbuf) - auio.uio_resid;
 		if (error != 0)
 			break;
-		for (pos = 0; pos < done; ) {
+		int done = sizeof (dirbuf) - auio.uio_resid;
-			dp = (struct dirent *)(dirbuf + pos);
+		for (int pos = 0; pos < done; ) {
+			struct dirent *dp = (struct dirent *)(dirbuf + pos);
 			pos += dp->d_reclen;
 			/*
 			 * XXX: Temporarily we also accept DT_UNKNOWN, as this
@@ -5558,24 +5701,23 @@ zfs_listextattr(struct vop_listextattr_args *ap)
 			 */
 			if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
 				continue;
-			if (plen == 0 &&
+			else if (plen == 0 &&
 			    strncmp(dp->d_name, "freebsd:", 8) == 0)
 				continue;
 			else if (strncmp(dp->d_name, attrprefix, plen) != 0)
 				continue;
-			nlen = dp->d_namlen - plen;
+			uint8_t nlen = dp->d_namlen - plen;
-			if (sizep != NULL)
+			if (ap->a_size != NULL) {
-				*sizep += 1 + nlen;
+				*ap->a_size += 1 + nlen;
-			else if (GET_UIO_STRUCT(&uio) != NULL) {
+			} else if (ap->a_uio != NULL) {
 				/*
 				 * Format of extattr name entry is one byte for
 				 * length and the rest for name.
 				 */
-				error = zfs_uiomove(&nlen, 1, zfs_uio_rw(&uio),
+				error = uiomove(&nlen, 1, ap->a_uio);
-				    &uio);
 				if (error == 0) {
-					error = zfs_uiomove(dp->d_name + plen,
+					char *namep = dp->d_name + plen;
-					    nlen, zfs_uio_rw(&uio), &uio);
+					error = uiomove(namep, nlen, ap->a_uio);
 				}
 				if (error != 0)
 					break;
@@ -5584,8 +5726,92 @@ zfs_listextattr(struct vop_listextattr_args *ap)
 	} while (!eof && error == 0);
 
 	vput(vp);
-	ZFS_EXIT(zfsvfs);
+	return (error);
+}
 
+static int
+zfs_listextattr_sa(struct vop_listextattr_args *ap, const char *attrprefix)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	int error;
+
+	error = zfs_ensure_xattr_cached(zp);
+	if (error != 0)
+		return (error);
+
+	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
+	ASSERT3P(zp->z_xattr_cached, !=, NULL);
+
+	size_t plen = strlen(attrprefix);
+	nvpair_t *nvp = NULL;
+	while ((nvp = nvlist_next_nvpair(zp->z_xattr_cached, nvp)) != NULL) {
+		ASSERT3U(nvpair_type(nvp), ==, DATA_TYPE_BYTE_ARRAY);
+
+		const char *name = nvpair_name(nvp);
+		if (plen == 0 && strncmp(name, "freebsd:", 8) == 0)
+			continue;
+		else if (strncmp(name, attrprefix, plen) != 0)
+			continue;
+		uint8_t nlen = strlen(name) - plen;
+		if (ap->a_size != NULL) {
+			*ap->a_size += 1 + nlen;
+		} else if (ap->a_uio != NULL) {
+			/*
+			 * Format of extattr name entry is one byte for
+			 * length and the rest for name.
+			 */
+			error = uiomove(&nlen, 1, ap->a_uio);
+			if (error == 0) {
+				char *namep = __DECONST(char *, name) + plen;
+				error = uiomove(namep, nlen, ap->a_uio);
+			}
+			if (error != 0)
+				break;
+		}
+	}
+
+	return (error);
+}
+
+/*
+ * Vnode operation to retrieve extended attributes on a vnode.
+ */
+static int
+zfs_listextattr(struct vop_listextattr_args *ap)
+{
+	znode_t *zp = VTOZ(ap->a_vp);
+	zfsvfs_t *zfsvfs = ZTOZSB(zp);
+	char attrprefix[16];
+	int error;
+
+	if (ap->a_size != NULL)
+		*ap->a_size = 0;
+
+	/*
+	 * If the xattr property is off, refuse the request.
+	 */
+	if (!(zfsvfs->z_flags & ZSB_XATTR))
+		return (SET_ERROR(EOPNOTSUPP));
+
+	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
+	    ap->a_cred, ap->a_td, VREAD);
+	if (error != 0)
+		return (error);
+
+	error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
+	    sizeof (attrprefix));
+	if (error != 0)
+		return (error);
+
+	ZFS_ENTER(zfsvfs);
+	ZFS_VERIFY_ZP(zp);
+	rw_enter(&zp->z_xattr_lock, RW_READER);
+	if (zfsvfs->z_use_sa && zp->z_is_sa)
+		error = zfs_listextattr_sa(ap, attrprefix);
+	if (error == 0)
+		error = zfs_listextattr_dir(ap, attrprefix);
+	rw_exit(&zp->z_xattr_lock);
+	ZFS_EXIT(zfsvfs);
 	return (error);
 }
 

module/os/freebsd/zfs/zfs_znode.c

@@ -145,10 +145,13 @@ zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
 
 	mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
+	rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);
 
 	zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp);
 
 	zp->z_acl_cached = NULL;
+	zp->z_xattr_cached = NULL;
+	zp->z_xattr_parent = 0;
 	zp->z_vnode = NULL;
 	return (0);
 }
@@ -164,9 +167,11 @@ zfs_znode_cache_destructor(void *buf, void *arg)
 	ASSERT(!list_link_active(&zp->z_link_node));
 	mutex_destroy(&zp->z_lock);
 	mutex_destroy(&zp->z_acl_lock);
+	rw_destroy(&zp->z_xattr_lock);
 	zfs_rangelock_fini(&zp->z_rangelock);
 
-	ASSERT(zp->z_acl_cached == NULL);
+	ASSERT3P(zp->z_acl_cached, ==, NULL);
+	ASSERT3P(zp->z_xattr_cached, ==, NULL);
 }
 
 
@@ -211,7 +216,10 @@ zfs_znode_alloc_kmem(int flags)
 static void
 zfs_znode_free_kmem(znode_t *zp)
 {
-
+	if (zp->z_xattr_cached) {
+		nvlist_free(zp->z_xattr_cached);
+		zp->z_xattr_cached = NULL;
+	}
 	uma_zfree_smr(znode_uma_zone, zp);
 }
 #else
@@ -237,7 +245,10 @@ zfs_znode_alloc_kmem(int flags)
 static void
 zfs_znode_free_kmem(znode_t *zp)
 {
-
+	if (zp->z_xattr_cached) {
+		nvlist_free(zp->z_xattr_cached);
+		zp->z_xattr_cached = NULL;
+	}
 	kmem_cache_free(znode_cache, zp);
 }
 #endif
@@ -1078,9 +1089,16 @@ zfs_rezget(znode_t *zp)
 		zfs_acl_free(zp->z_acl_cached);
 		zp->z_acl_cached = NULL;
 	}
-
 	mutex_exit(&zp->z_acl_lock);
-	ASSERT(zp->z_sa_hdl == NULL);
+
+	rw_enter(&zp->z_xattr_lock, RW_WRITER);
+	if (zp->z_xattr_cached) {
+		nvlist_free(zp->z_xattr_cached);
+		zp->z_xattr_cached = NULL;
+	}
+	rw_exit(&zp->z_xattr_lock);
+
+	ASSERT3P(zp->z_sa_hdl, ==, NULL);
 	err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
 	if (err) {
 		ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);