[Lguest] mercurial repo

[Rusty Russell]

On Sun, 2007-07-29 at 18:13 -0700, ron minnich wrote:
> On 7/29/07, Rusty Russell <rusty at rustcorp.com.au> wrote:
> > Indeed, I've told Ron we'll be breaking his port soon.  Not
> > gratuitously, but no doubt there'll be churn when we add our second
> > arch, use kvm-style ioctls and change I/O models to virtio.  Should be
> > simple to adapt, tho.
> 
> That's fine. I assume the virtio will mean it's also simpler to run under kvm.
> 
> One question/suggestion. As you retire old HCALLS and add new ones,
> don't reuse the numbers. Rather, if someone uses a deprectaed HCALL,
> kill them instantly with an error. This will simplify my life as
> people try to run older plan 9 guests under newer linux kernels.

Yes, this helps everyone.  I've done one already (removed the
GET_WALLCLOCK in favour of a field in lguest_data).

> Will the churn be in the context of 2.6.23?

No, 2.6.23 is nailed at this point.  It'll be 2.6.24, so we have a
while.

> One last question: I pulled linus' tree and the Linux guest won't
> start. A quick look shows that the return EIP after the second mmap
> call in the lguest program is completely wrong. Anyone seen this? I
> can send more info if not.

How strange!  I have an issue with bzImages crashing (despite bss fix):
seems to be seeing odd memory.  I'm tracking it now (vmlinux seems to
work fine).

Rusty.