[Lguest] [PATCH 0/5] Boot protocol changes

[Chris Malley]

On Thu, 2007-10-04 at 20:47 +1000, Rusty Russell wrote:
> On Thu, 2007-10-04 at 10:47 +0100, Chris Malley wrote:
> > Hi guys
> > 
> > I gave these patches a try (on top of 2.6.23-rc9 plus the previously
> > submitted 2.6.24 patch set).
> > 
> > The last two seem to cause Badness on my system, whereby if I start a
> > guest (using the same bzImage as the host, as before) it seems to boot
> > OK, and the host system still superficially looks stable (my X session
> > is OK and I can interact with existing processes) but if I attempt to
> > launch any new process on the host (even typing "ls" into a terminal
> > window) I get a Segmentation fault.
> 
> I think this might actually be caused by a nasty bug in one of the
> previous patches (patch 27).  I found it because every second time I
> launched a guest it failed.  Strange, huh?
> 
> These lines in drivers/lguest/page_tables.c:
> 
> +               pte_t ro_gpte = gpte;
> +               pte_wrprotect(ro_gpte);
> 
> Should be:
> 
> +               pte_t ro_gpte = pte_wrprotct(gpte);
just being pedantic I assume you mean:
  +               pte_t ro_gpte = pte_wrprotect(gpte);

> That's the only "break the host" issue I know of...
> 
> Thanks!
> Rusty.

Yep, that fixes it, thanks Rusty!

--
Chris