Re: [Qemu-devel] [PATCH v4 4/5] slirp: VMStatify socket level

[Dr. David Alan Gilbert]

* Daniel P. Berrange (address@hidden) wrote:
> On Thu, Feb 23, 2017 at 10:51:56AM +0000, Dr. David Alan Gilbert wrote:
> > Samual, Jan: Can you just take 1-3 of this series;  4 has a problem;
> > 
> > * Juan Quintela (address@hidden) wrote:
> > > "Dr. David Alan Gilbert (git)" <address@hidden> wrote:
> > > > From: "Dr. David Alan Gilbert" <address@hidden>
> > > >
> > > > Working up the stack, this replaces the slirp_socket_load/save
> > > > with VMState definitions.
> > > >
> > > > A place holder for IPv6 support is added as a comment; it needs
> > > > testing once the rest of the IPv6 code is there.
> > > >
> > > > Signed-off-by: Dr. David Alan Gilbert <address@hidden>
> > > 
> > > Reviewed-by: Juan Quintela <address@hidden>
> > > 
> > > 
> > > > +/* Win has a signed family number */
> > > > +#define VMSTATE_SS_FAMILY(f, s) VMSTATE_INT16(f, s)
> > > 
> > > Great!
> > > Just hope that there is no 32000 families soon :-)
> > 
> > Actually; that's a problem;  it turns out FreeBSD has a char for it's
> > family type rahter than the uint16 that linux and windows have.
> > I need to think hth to abstract that.
> 
> What, if any, promises have we made about the migration data format
> being ABI stable across platforms ?

It should be in theory; it should just reflect the state of the guest, and
the state of the guest should be independent of the platform it's running on.
Now slirp is a bit of an odd case since it involves more host state.

> It strikes me that even if FreeBSD & Linux had the same sized type
> for ss_family, the constants used to popuate it might be different
> on each platform. eg AF_INET6 is 10 on Linux but 23 on Mingw and
> probably something else again on FreeBSD.

Ewww I'd thought that was at least standardised.
It looks like AF_INET is 2 on Linux, mingw and FreeBSD;
but AF_INET6 is more dependent on what snook in before hand.

> IOW if we transmit this data on the wire, we've effectively said that
> our migration data format is *not* portable across different host OS
> platforms. At that point, sending different sized types on BSD vs
> Linux is no big deal since we're already incompatible semantically.

This is a relatively recent error; it comes from eae303ff which added
ss_family to allow IPv6.

I suspect the right fix here is to populate a temporary for ss_family
on the wire; if we make it conveniently match Linux's AF_INET6 values
it might work.

Dave

> Regards,
> Daniel
> -- 
> |: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
> |: http://libvirt.org              -o-             http://virt-manager.org :|
> |: http://entangle-photo.org       -o-    http://search.cpan.org/~danberr/ :|
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK