Move the VA_MARKATIME flag for VOP_SETATTR() out into its own VOP:

VOP_MARKATIME() since unlike the rest of VOP_SETATTR(), VA_MARKATIME
can be performed while holding a shared vnode lock (the same functionality
is done internally by VOP_READ which can run with a shared vnode lock).
Add missing locking of the vnode interlock to the ufs implementation and
remove a special note and test from the NFS client about not supporting the
feature.

Inspired by:	ups
Tested by:	pho

1 files changed, 2 insertions, 5 deletions

diff --git a/sys/kern/vfs_subr.c b/sys/kern/vfs_subr.c
index 1e4bf90..753f95f 100644
--- a/sys/kern/vfs_subr.c
+++ b/sys/kern/vfs_subr.c
@@ -4206,18 +4206,15 @@ vfs_read_dirent(struct vop_readdir_args *ap, struct dirent *dp, off_t off)
 
 /*
  * Mark for update the access time of the file if the filesystem
- * supports VA_MARK_ATIME.  This functionality is used by execve
+ * supports VOP_MARKATIME.  This functionality is used by execve
  * and mmap, so we want to avoid the synchronous I/O implied by
  * directly setting va_atime for the sake of efficiency.
  */
 void
 vfs_mark_atime(struct vnode *vp, struct ucred *cred)
 {
-	struct vattr atimeattr;
 
 	if ((vp->v_mount->mnt_flag & (MNT_NOATIME | MNT_RDONLY)) == 0) {
-		VATTR_NULL(&atimeattr);
-		atimeattr.va_vaflags |= VA_MARK_ATIME;
-		(void)VOP_SETATTR(vp, &atimeattr, cred);
+		(void)VOP_MARKATIME(vp);
 	}
 }