On Sat, Jun 25, 2016 at 01:07:10AM +0300, Michael S. Tsirkin wrote: > On Fri, Jun 24, 2016 at 02:08:52PM +0200, Marc-André Lureau wrote: > > On Thu, Jun 23, 2016 at 7:01 PM, Michael S. Tsirkin <m...@redhat.com> wrote: > > >> > Maybe what you want is a need_unlink feature. > > >> > Set it for unix sockets only, that would make some sense. > > >> > > >> Oh perhaps what you mean is that if the fd was passed, we should cleanup > > >> the unix socket? Yes, I think we should do that then. I'll update the > > >> series. > > > > Actually it's not possible to pass a listening fd to a socket chardev > > today (the path argument doesn't understand /dev/fdset), so only path > > created by qemu will be cleaned up. > > > > > > > > I'd like it better contained - that's all. So let's set a flag that says > > > "must unlink" as opposed to "it's listening". > > > > You suggest to rename QIO_CHANNEL_FEATURE_LISTEN to > > QIO_CHANNEL_FEATURE_LISTEN_MUST_UNLINK ? > > QIO_CHANNEL_FEATURE_UNLINK_ON_CLOSE > > or something like this. > > Or maybe QIO_CHANNEL_FEATURE_BOUND > > > Or to add another feature > > flag? I don't think that brings anything useful here. > > The point is that in the future we might be listening on sockets where > we did not bind it. I would think that in that case, we do not want to > unlink it. So name should reflect this somehow.
There is no possibility of listening on sockets without binding to them with the current design of the code. If & when we want such support we can adapt as needed, but I don't see any real point in trying to support something that is impossible right now. 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 :|
