Re: [PATCH v3 2/3] QIOChannelSocket: Implement io_async_write & io_async_flush

[Daniel P . Berrangé]

On Wed, Sep 29, 2021 at 04:32:12PM -0300, Leonardo Bras Soares Passos wrote:
> Hello Daniel,
> 
> On Fri, Sep 24, 2021 at 2:38 PM Daniel P. Berrangé <berrange@redhat.com> 
> wrote:
> [...]
> > > @@ -154,6 +171,19 @@ int qio_channel_socket_connect_sync(QIOChannelSocket 
> > > *ioc,
> > >          return -1;
> > >      }
> > >
> > > +#ifdef CONFIG_LINUX
> > > +    if (addr->type != SOCKET_ADDRESS_TYPE_INET) {
> > > +        return 0;
> > > +    }
> > > +
> > > +    ret = qemu_setsockopt(fd, SOL_SOCKET, SO_ZEROCOPY, &v, sizeof(v));
> > > +    if (ret >= 0) {
> > > +        QIOChannelClass *klass = QIO_CHANNEL_GET_CLASS(ioc);
> > > +        klass->io_async_writev = qio_channel_socket_async_writev;
> > > +        klass->io_async_flush = qio_channel_socket_async_flush;
> > > +    }
> > > +#endif
> >
> > This is not write - the async APIs should not be tied 1:1 to ZEROCOPY
> > usage - we should have them take a flag to request ZEROCOPY behaviour.
> 
> I agree, but I am not aware of how to do asynchronous send in a socket
> without MSG_ZEROCOPY.
> 
> I mean, I know of the non-blocking send, but I am not sure how it
> checks if everything was sent (i.e. the flush part).
> Would it also be using the ERRQUEUE for that?
> 
> What would you suggest?

Yeah, there isn't any really. I guess I'm anticipating a future that
probably won't exist.  Lets just call the callbacks 'io_write_zerocopy'
and 'io_flush_zerocopy' and ignore the flag.


Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|