[Top][All Lists]

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

Re: Virtual file systems

From: brunoacf
Subject: Re: Virtual file systems
Date: Thu, 4 Sep 2003 09:01:14 -0300

Thanks Marcus.

So, this way is the L4 really the next
(best?) choice to a microkernel for
the Hurd?


> On Wed, Sep 03, 2003 at 09:31:47AM -0300, brunoacf wrote:
> > There is some effort to implement
> > virtual filesystems like procfs and
> > devfs into the Hurd? Sometime ago I
> > saw a discussion about "Hurd do not
> > need procfs" in the mailing lists but
> > I still believe it would be a good
> > improvement.
> If you say procfs, you are probably thinking about the Linux /proc
> filesystem.  There is some information accessible via the trusted proc
> server, and for other information in the Hurd you can access tasks directly to
> get it (see the msgport program, for example - in the L4 port, this includes
> even basic stuff like inspecting memory).
> This can theoretically also be done by procfs.  But only theoretically.  In
> reality, this would mean that the procfs filesystem contacts the process on
> your behalf.  But this means that if the process is malicious and does block
> indefinitely on the request, the procfs thread would be stuck.  It could use
> a timeout, but even a small timeout makes a server vulnerable to denial of
> service attacks.
> So, there are security issues in the Hurd which make a proc filesystem a
> suboptimal solution.  Only some of the features of a Linux proc filesystem
> could be provided by such a system wide translator.
> So, such a translator could be written, but it would usefully be implemented
> as a translator that should be run by the users themselves, and thus
> consumes their resources, rather than system wide.
> Device fs is different.  Given the pityful state of device drivers in GNU
> Mach, this doesn't make much sense today.  With the port to L4, we will have
> a much more dynamic framework, and it will make sense and probably be
> implemented.  Device fs is different because the device drivers are trusted
> and centralized (as opposed to procfs, where the information is distributed
> and provided by untrusted sources).
> Thanks,
> Marcus
> --
> `Rhubarb is no Egyptian god.' GNU
> Marcus Brinkmann              The Hurd

Acabe com aquelas janelinhas que pulam na sua tela.
AntiPop-up UOL - É grátis!

reply via email to

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