[src/trunk]: src/sys/kern Do not follow symlinks in sys_unmount()

[manu <manu%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/538d20973d28
branches:  trunk
changeset: 334014:538d20973d28
user:      manu <manu%NetBSD.org@localhost>
date:      Wed Nov 26 10:50:36 2014 +0000

description:
Do not follow symlinks in sys_unmount()

There are situations where the underlying filesystem is unreachable
(e.g: NFS) causing symlink resolution to hang. Such a situation
should be avoided by using umount -f -R (force and raw), but while -R
causes the symlink resolution to be skipped in umount(8), the kernel was
still doing it in sys_unmount(). This changes fixes that.

When the -R flag is not given, umount(8) does symlinks resolution through
realpath(3) before calling unmount(2), hence not doing it in the kernel
would not change behavior.

diffstat:

 sys/kern/vfs_syscalls.c |  6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diffs (27 lines):

diff -r 77deee49217c -r 538d20973d28 sys/kern/vfs_syscalls.c
--- a/sys/kern/vfs_syscalls.c   Wed Nov 26 10:18:37 2014 +0000
+++ b/sys/kern/vfs_syscalls.c   Wed Nov 26 10:50:36 2014 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: vfs_syscalls.c,v 1.491 2014/09/05 09:20:59 matt Exp $  */
+/*     $NetBSD: vfs_syscalls.c,v 1.492 2014/11/26 10:50:36 manu Exp $  */
 
 /*-
  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
@@ -70,7 +70,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.491 2014/09/05 09:20:59 matt Exp $");
+__KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.492 2014/11/26 10:50:36 manu Exp $");
 
 #ifdef _KERNEL_OPT
 #include "opt_fileassoc.h"
@@ -574,7 +574,7 @@
                return error;
        }
 
-       NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
+       NDINIT(&nd, LOOKUP, LOCKLEAF | TRYEMULROOT, pb);
        if ((error = namei(&nd)) != 0) {
                pathbuf_destroy(pb);
                return error;