On 5/7/20, Paul Durrant <[email protected]> wrote: >> -----Original Message----- >> From: Denis Kirjanov <[email protected]> >> Sent: 06 May 2020 18:45 >> To: [email protected] >> Cc: [email protected]; [email protected]; [email protected]; >> [email protected] >> Subject: Re: [PATCH net-next v7 2/2] xen networking: add XDP offset >> adjustment to xen-netback >> >> On 5/5/20, Paul Durrant <[email protected]> wrote: >> >> -----Original Message----- >> >> >> @@ -417,6 +431,11 @@ static void frontend_changed(struct >> >> >> xenbus_device >> >> >> *dev, >> >> >> set_backend_state(be, XenbusStateConnected); >> >> >> break; >> >> >> >> >> >> + case XenbusStateReconfiguring: >> >> >> + read_xenbus_frontend_xdp(be, dev); >> >> > >> >> > Is the frontend always expected to trigger a re-configure, or could >> >> > feature-xdp already be enabled prior to connection? >> >> >> >> Yes, feature-xdp is set by the frontend when xdp code is loaded. >> >> >> > >> > That's still ambiguous... what I'm getting at is whether you also need >> > to >> > read the xdp state when transitioning into Connected as well as >> > Reconfiguring? >> >> I have to read the state only during the Reconfiguring state since >> that's where an XDP program is loaded / unloaded and then we transition >> from Reconfigred to Connected >> > > Ok, but what about netback re-connection? It is possible that netback can be > disconnected, unloaded, reloaded and re-attached to a running frontend. In > this case XDP would be active so I still think read_xenbus_frontend_xdp() > needs to form part of ring connection (if only in this case).
I made a change to xdp-netfront to keep the state of XDP and then just pass the saved state in talk_to_netback() > > Paul > >
