Re: Hurd-related research help requested

[Moritz Schulte]

nisse@lysator.liu.se (Niels Möller) writes:

>  * Particular translators, like shadow-fs. As I understand it, it's
>    interactino with the rest of the system is not entirely trivial,
>    and it might be a good example for really understanding translators
>    and the dir_lookup mechanism.

Btw: there was a consensus that it should be better called unionfs;
that's also the name which identifies the package contained in the
hurd-extras repository of Savannah.

Well, actually it is not that complicated, how unionfs does the
lookups on the underlying filesystems; it's just the fs API, which
makes it AFAICS impossible to implement without any race conditions.

The tricky part is that unionfs should return a looked up port
directly to the user, in case it is the root of a new translator (or a
symlink).  In that case unionfs' job is over.  But this is not
possible _without_ a _new_ lookup of a given port.  That new lookup is
necessary, since _first_ unionfs has to do O_NOTRANS lookups to be
able to _know_ what kind of node it has found.

                moritz
-- 
moritz@duesseldorf.ccc.de - http://duesseldorf.ccc.de/~moritz/
GPG fingerprint = 3A14 3923 15BE FD57 FC06  B501 0841 2D7B 6F98 4199