On Mar 26, 2006, at 1:21 PM, Martin Husemann wrote:

> I just ran accross a case where (eroneously) a filesystem exeeded
> the maximum file id size _VFS_MAXFIDSIZ. This truncates the values
> returned by the getfh() syscall at least - maybe has other  
> consequences
> in NFS.
>
> Since _VFS_MAXFIDSIZ is defined to be the maximum possible size,
> should we enforce it a bit, like in this change?

Yes, I think this is a good change.

>
> Index: vfs_syscalls.c
> ===================================================================
> RCS file: /cvsroot/src/sys/kern/vfs_syscalls.c,v
> retrieving revision 1.238
> diff -p -U5 -r1.238 vfs_syscalls.c
> --- vfs_syscalls.c	1 Mar 2006 12:38:21 -0000	1.238
> +++ vfs_syscalls.c	26 Mar 2006 21:16:22 -0000
> @@ -1218,10 +1218,11 @@ sys_getfh(struct lwp *l, void *v, regist
>  	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
>  	error = VFS_VPTOFH(vp, &fh.fh_fid);
>  	vput(vp);
>  	if (error)
>  		return (error);
> +	KASSERT(fh.fh_fid.len <= _VFS_MAXFIDSIZ);
>  	error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
>  	return (error);
>  }
>
>
> Should this be a KDASSERT instead?
> Should I add this, and equivalent changes to the NFS code (guarding
> each VFS_VPTOFH)?
>
> Martin

-- thorpej