On Mon, Oct 12, 2015 at 03:35:41PM +0200, Paolo Bonzini wrote: > The QemuOpts-based code treats "option not set" and "option set > to false" the same way for the ipv4 and ipv6 options, because it > is meant to handle only the ",ipv4" and ",ipv6" substrings in > hand-crafted option parsers. > > When converting InetSocketAddress to QemuOpts, however, it is > necessary to handle all three cases (not set, set to true, set > to false). Currently we are not handling all cases correctly. > The rules are: > > * if none or both options are absent, leave things as is > > * if the single present option is Y, the other should be N. > This can be implemented by leaving things as is, or by setting > the other option to N as done in this patch. > > * if the single present option is N, the other should be Y. > This is handled by the "else if" branch of this patch. > > This ensures that the ipv4 option has an effect on Windows, > where creating the socket with PF_UNSPEC makes an ipv6 > socket. With this patch, ",ipv4" will result in a PF_INET > socket instead. > > Reported-by: Sair, Umair <umair_s...@mentor.com> > Tested-by: Sair, Umair <umair_s...@mentor.com> > Signed-off-by: Paolo Bonzini <pbonz...@redhat.com> > --- > util/qemu-sockets.c | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-)
Reviewed-by: Daniel P. Berrange <berra...@redhat.com> I look forward to formally submitting the remainder of my patches to kill off QemuOpts from this qemu-sockets.c code :-) 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 :|
