On Wed, May 04, 2011 at 09:39:02AM +0100, n...@bytemark.co.uk wrote: > Hi, > > Currently migration-tcp.c uses the IPv4-only socket functions, making > migrations over IPv6 impossible. Following is a tentative patch that switches > it to use inet_connect() and inet_listen(). > > However, the patch loses the non-blocking connect() behaviour seen with the > previous code. I'm not sure how much of an issue this is - if connect() blocks > here, does it block execution of the VM? > > If so, I guess we need a non-blocking form of inet_connect(), or some way of > replicating the behaviour - it would potentially be needed for my NBD > reconnection patches too? I can see that a blocking connect() might not be an > issue while the KVM process is starting up, but could cause problems if we > try to reconnect while emulation is ongoing. > > Thoughts?
FWIW, Juan Quintela also posted a set of patches to add IPv6 support for migration a few weeks back, but unfortunately they don't appear to have been merged yet: http://www.mail-archive.com/qemu-devel@nongnu.org/msg58954.html IIUC, Juan's patches don't have the blocking connect() problem you mention. Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
