Re: [Xen-devel] [PATCH] [Linux] Transfer TPM locality info in the ringstructure

[Stefan Berger <stefanb@xxxxxxxxxx>]

Keir Fraser <Keir.Fraser@xxxxxxxxxxxx> wrote
on 01/07/2008 07:58:25 AM:

> Personally I do not much care about compatibility of new backend 
> with old frontend, or vice versa, since I doubt that the vtpm 
> drivers have many users. But others with an interest in the vtpm 
> drivers might care, as the errors that mismatched drivers will cause
> are likely to be moderately hard to debug. If compatibility is 
> required, you’ll have to negotiate the new request format via 
> xenstore. It’s not that hard: see the request-foo/feature-foo 
> handshake that goes on e.g., between netfront and netback.

Actually, intializing the ring to zeroes once the
mapping is done does solve the problem for the frontend that currently
does not touch the uninitialized variable. Though, to correct this, I am
also initializing the 'unused' member on the frontend in a patch that I'l
post soon.

   Stefan

> 
>  -- Keir
> 
> On 7/1/08 11:53, "Stefan Berger" <stefanb@xxxxxxxxxx>
wrote:

> Keir Fraser <Keir.Fraser@xxxxxxxxxxxx>
wrote on 01/06/2008 06:12:06 PM:
> 
> > How does that fix the problem if an old tpm frontend runs against
a 
> > new tpm backend? The unused field will not be initialised by
the old
> > frontend, hence can contain garbage to confuse the new backend?
> > 
> 
> I don't know of any better solution than initializing this 'now'.
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel