Re: "loopback mount" via vnd feature request

[christos%astron.com@localhost (Christos Zoulas)]

In article 
<1a69a9d80903021231n5a83bf05kbd2aad2bbd58cf6c%mail.gmail.com@localhost>,
Arnaud Lacombe  <lacombar%gmail.com@localhost> wrote:
>On Mon, Mar 2, 2009 at 8:03 AM, Toru Nishimura
><locore64%alkyltechnology.com@localhost> wrote:
>> NetBSD can do mount regular files just like L*, however,
>> the way to practice it is awkward.
>>
>as is using a "loop" device to do that.
>
>> The issue here is that our vnconfig (vnode-config) is
>> unable to show off which file is bound with vnd device alive.
>>
>I made a patch I kept in my private tree long ago to fix this. The
>trouble is that, AFAIR, namei() is not bijective, so you can't
>translate reliably an inode number into a filename. You can save the
>filename the filename at configure time, but that might be easily
>defeated (even if it will fit in most of the use case).
>

Well we need to fix this at some point. There is the namei cache hack
vnode_to_path for now, that will give you a path to the vnode if one
exists, but we need to add a more reliable reverse lookup.

christos