[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

passive translators

From: Neal H. Walfield
Subject: passive translators
Date: Wed, 20 Jun 2007 19:00:50 +0200
User-agent: Wanderlust/2.14.0 (Africa) SEMI/1.14.6 (Maruoka) FLIM/1.14.6 (Marutamachi) APEL/10.6 Emacs/21.4 (i386-pc-linux-gnu) MULE/5.0 (SAKAKI)

There is a bit of confusion here, I'm going to try to be more
specific, please indulge me.

At Thu, 21 Jun 2007 00:25:33 +0800,
Wei Shen wrote:
> The fs server or the translator?

What is an fs server?  In the Hurd, translators are file systems: some
are one node file systems; and, some expose complex hierarchies.
Translators are linked together using the file_set_translator method.

>                                   I think at least the fs server should
> know. Am I right?

You are right: when a client invokes an operation on a capability/file
handle, the server knows that the file handle is chroot'ed: it has a
capability that it returns when resolving `..' via the file handle.

You talk about associating a symbolic name with this.  The question is
what is the right symbolic name for it?  This depends on the context,
and that is exactly the problem: naming contexts as represented by
capabilities are not persistent.

> The translator needs not to know whether a process is chrooted or not, it
> just always adds (for any process) the virtual root argument saved when it
> is associated with the file node.

I assume that you mean the file handle that is passed to the chroot'ed
process that it uses as its root, and not the on-disk node.

First, the file handle is not persistent: it is destoryed when the
system is rebooted.  How do you recover this information?

Second, the chroot'ed process can create a passive translator and then
stat it.  This causes the translator to start an active translator.
Currently, when a passive translator is instantiated, the resulting
translator's name space is set to the parent translator's.  This is
because the passive translator is a string without a naming context.
This allows the chroot'ed process to escape from its jail.  The
problem here is how to preserve the naming context.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]