>Small clarification: I'm not using "venus". I'm working with a
>venus-replacement called "podfuk" to hook-up my own filesystems (like
>ftpfs, tarfs and so on...).

Thanks for the clarification ... that makes better sense now.

>And to answer your question: The file dtpad tries to mmap() is within
>the CODA filesystem (better: the filesystem managed by the CODA kernel
>module).
>
>Any ideas what's going wrong ?

Let me quote from the code:

  if (off + len > cn->attr.va_size) {
    cmn_err (CE_NOTE, "coda_map: off+len (0x%x) > va_size (0x%x)!",
             (unsigned long)off+len, cn->attr.va_size);
    error = EINVAL;

So what it looks like is that the offset (start of mapping in the
file) plus the length of the mapping ends up being out side of
the bits in the file (attr.va_size).

In fact, if you see, your message appears to say that your "container"
file has no length.  (va_size of 0x0)

There are two possibilities:

  a) The underlying mmap of "container" files wouldn't work and thus
	the arla (and thus coda) kernel module didn't allow this,

or
  b) coda (and arla) kernel module is being too picky and isn't allowing
	it should.

I suspect it is the second and that this case should be handled by
extending the file to the proper place.

-- 
Phil Nelson                       NetBSD: http://www.netbsd.org
e-mail: phil@cs.wwu.edu           Coda: http://www.coda.cs.cmu.edu
http://cs.wwu.edu/faculty/nelson