On Wed, Feb 15, 2017 at 02:46:15PM +0000, Dr. David Alan Gilbert wrote: > * Daniel P. Berrange (berra...@redhat.com) wrote: > > On Mon, Jan 23, 2017 at 10:32:13PM +0100, Juan Quintela wrote: > > > We create new channels for each new thread created. We only send through > > > them a character to be sure that we are creating the channels in the > > > right order. > > > > > > Note: Reference count/freeing of channels is not done > > > > > > Signed-off-by: Juan Quintela <quint...@redhat.com> > > > --- > > > include/migration/migration.h | 6 +++++ > > > migration/ram.c | 45 +++++++++++++++++++++++++++++++++- > > > migration/socket.c | 56 > > > +++++++++++++++++++++++++++++++++++++++++-- > > > > BTW, right now libvirt never uses QEMU's tcp: protocol - it does everything > > with the fd: protocol. So either we need multi-fd support for fd: protocol, > > or libvirt needs to switch to use tcp: > > I thought using fd was safer than tcp: because of the race when something else > could listen on the proposed port on the incoming side between the point of > libvirt > picking the port number and qemu starting.
Hmm, good point. > > In fact, having said that, we're going to have to switch to use the tcp: > > protocol anyway in order to support TLS, so this is just another good > > reason for the switch. > > I thought you had a way of allowing fd to work for TLS? Oh yes, I forgot that I made that work :-) 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/ :|