Re: [PATCH 4 of 5] libxl: Add support for passing in the machine's E820 for PCI passthrough

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Fri, Apr 08, 2011 at 03:33:47PM +0100, Ian Campbell wrote:
> On Fri, 2011-04-08 at 15:25 +0100, Tim Deegan wrote:
> > At 15:17 +0100 on 08 Apr (1302275848), Ian Campbell wrote:
> > > > That's a _horrible_ config syntax.
> > > 
> > > fair enough. I thought it nicely represented the "has an empty PCI bus"
> > > myself...
> > 
> > Sure, but it's really not obvious that that means "has a memory layout
> > that matches the host on which it was last booted".
> 
> Hmm, yes, this proposal doesn't really cover migration does it...
> 
> You obviously can't migrate with a device passed through but it should
> be possible to unplug, migrate and replug. Maybe having a sensible hole

Yes.
> on the first host is sufficient... Especially if the annoying case is
> GFX passthrough -- you are much less like to want/be able to unplug your
> gfx over a migration...

So one use case is SR-IOV and depending on what machine you are on, the
BARs can be different.

If you are migrating between the same type of machines, where the BARs
and E820 are the same machine it "ought" to work (hadn't tested it as
I only have one machine with SR-IOV ..but I should be able to
test the save/restore on that machine).

What I did test was save/restore on a single machine (where I unload
the NIC module's before hand and load it after a restore) and it worked
(with some Linux kernel patches that I need to post soon).

But the right way is to augment the Linux kernel code to deal with
a new E820 _after_ a resume. This is something I plan on doing 
during the summer, but not right now.