BTW, I also randomly discovered that in sys_mount(), failure to allocate
syncvnode can be sometimes ignored - around line 351 of vfs_syscalls.c,
it is like:

               VOP_UNLOCK(vp, 0);
                if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
                        error = vfs_allocate_syncvnode(mp);
                vfs_unbusy(mp);
                (void) VFS_STATFS(mp, &mp->mnt_stat, p);
                if ((error = VFS_START(mp, 0, p)))
                        vrele(vp);

The sync vnode could probably be allocated before the mountpoint
is added to the mountlist and error should be handled.
Also handling of error if VFS_START() fails seems to be incomplete -
shouldn't it remove the mountpoint from mountlist as well ?
dounmount() ignores return value from vfs_allocate_syncvnode()
altogether -
it that ok ?

Jaromir