Re: [Qemu-devel] [PATCH 2/3] net: complete all queued packets on VM stop

[Michael S. Tsirkin]

On Thu, Sep 04, 2014 at 06:15:12PM +0800, Jason Wang wrote:
> On 09/04/2014 04:39 PM, Michael S. Tsirkin wrote:
> > This completes all packets, ensuring that callbacks
> > will not run when VM is stopped.
> >
> > Cc: address@hidden
> > Cc: Jason Wang <address@hidden>
> > Signed-off-by: Michael S. Tsirkin <address@hidden>
> > ---
> >  net/net.c | 33 ++++++++++++++++++++++++++++++++-
> >  1 file changed, 32 insertions(+), 1 deletion(-)
> >
> > diff --git a/net/net.c b/net/net.c
> > index 6d930ea..25fdb07 100644
> > --- a/net/net.c
> > +++ b/net/net.c
> > @@ -47,6 +47,7 @@
> >  # define CONFIG_NET_BRIDGE
> >  #endif
> >  
> > +static VMChangeStateEntry *net_change_state_entry;
> >  static QTAILQ_HEAD(, NetClientState) net_clients;
> >  
> >  const char *host_net_devices[] = {
> > @@ -504,7 +505,8 @@ void qemu_purge_queued_packets(NetClientState *nc)
> >      qemu_net_queue_purge(nc->peer->incoming_queue, nc);
> >  }
> >  
> > -void qemu_flush_queued_packets(NetClientState *nc)
> > +static
> > +void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
> >  {
> >      nc->receive_disabled = 0;
> >  
> > @@ -518,9 +520,17 @@ void qemu_flush_queued_packets(NetClientState *nc)
> >           * the file descriptor (for tap, for example).
> >           */
> >          qemu_notify_event();
> > +    } else if (purge) {
> > +        /* Unable to empty the queue, purge remaining packets */
> > +        qemu_net_queue_purge(nc->incoming_queue, nc);
> >      }
> >  }
> >  
> > +void qemu_flush_queued_packets(NetClientState *nc)
> > +{
> > +    qemu_flush_or_purge_queued_packets(nc, false);
> > +}
> > +
> >  static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
> >                                                   unsigned flags,
> >                                                   const uint8_t *buf, int 
> > size,
> > @@ -1168,6 +1178,22 @@ void qmp_set_link(const char *name, bool up, Error 
> > **errp)
> >      }
> >  }
> >  
> > +static void net_vm_change_state_handler(void *opaque, int running,
> > +                                        RunState state)
> > +{
> > +    /* Complete all queued packets, to guarantee we don't modify
> > +     * state later when VM is not running.
> > +     */
> > +    if (!running) {
> > +        NetClientState *nc;
> > +        NetClientState *tmp;
> > +
> > +        QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
> > +            qemu_flush_or_purge_queued_packets(nc, true);
> > +        }
> > +    }
> > +}
> > +
> >  void net_cleanup(void)
> >  {
> >      NetClientState *nc;
> > @@ -1183,6 +1209,8 @@ void net_cleanup(void)
> >              qemu_del_net_client(nc);
> >          }
> >      }
> > +
> > +    qemu_del_vm_change_state_handler(net_change_state_entry);
> >  }
> >  
> >  void net_check_clients(void)
> > @@ -1268,6 +1296,9 @@ int net_init_clients(void)
> >  #endif
> >      }
> >  
> > +    net_change_state_entry =
> > +        qemu_add_vm_change_state_handler(net_vm_change_state_handler, 
> > NULL);
> > +
> >      QTAILQ_INIT(&net_clients);
> >  
> >      if (qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL, 
> > 1) == -1)
> 
> A problem is the dependency between state change handlers (e.g. virtio).
> Current virtio vmstate change handler will be called before this
> handler. Which means vdev->vm_running was false when we purge the queue,
> this will trigger the assert of vdev->vm_running in virtio_net_flush_tx().

True but that's a virtio bug: it changes vm_running too early.
I will send a patch to fix that now.
Long term with the core changes, we mught be able to get rid of
vm_running field completely.