[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH 2/3] bridge: trigger a bridge calculation upon port changes



On Wed, Apr 30, 2014 at 04:04:34PM -0400, Vlad Yasevich wrote:
> On 04/22/2014 03:43 PM, Luis R. Rodriguez wrote:
> > diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
> > index 54d207d..dcd9378 100644
> > --- a/net/bridge/br_if.c
> > +++ b/net/bridge/br_if.c
> > @@ -315,6 +315,8 @@ netdev_features_t br_features_recompute(struct 
> > net_bridge *br,
> >     features &= ~NETIF_F_ONE_FOR_ALL;
> >  
> >     list_for_each_entry(p, &br->port_list, list) {
> > +           if (p->flags & BR_ROOT_BLOCK)
> > +                   continue;
> >             features = netdev_increment_features(features,
> >                                                  p->dev->features, mask);
> >     }
> >
> Hi Luis
> 
> The hunk above isn't right.  Just because you set ROOT_BLOCK on the port
> doesn't mean that you should ignore it's device features.  If the device
> you just added happens to disable or enable some device offload feature,
> you should take that into account.

OK thanks, how about this part:

On 04/22/2014 03:43 PM, Luis R. Rodriguez wrote:
> On Tue, Mar 18, 2014 at 02:22:43PM -0700, Luis R. Rodriguez wrote:
> On Tue, Mar 18, 2014 at 01:46:49PM -0700, Cong Wang wrote:
>> On Fri, Mar 14, 2014 at 6:39 PM, Luis R. Rodriguez <mcgrof@xxxxxxxx> wrote:
>>> On Thu, Mar 13, 2014 at 11:26:25AM -0700, Cong Wang wrote:
>>>> On Wed, Mar 12, 2014 at 8:15 PM, Luis R. Rodriguez
>>>> <mcgrof@xxxxxxxxxxxxxxxx> wrote:
>>>>>                         spin_unlock_bh(&p->br->lock);
>>>>> +                       if (changed)
>>>>> +                               
>>>>> call_netdevice_notifiers(NETDEV_CHANGEADDR,
>>>>> +                                                        p->br->dev);
>>>>> +                       netdev_update_features(p->br->dev);
>>>>
>>>> I think this is supposed to be in netdev event handler of br->dev
>>>> instead of here.
>>>
>>> Do you mean netdev_update_features() ? I mimic'd what was being done on
>>> br_del_if() given that root blocking is doing something similar. If
>>> we need to change something for the above then I suppose it means we need
>>> to change br_del_if() too. Let me know if you see any reason for something
>>> else.
>>>
>>
>> Yeah, for me it looks like it's better to call netdev_update_features()
>> in the event handler of br->dev, rather than where calling
>> call_netdevice_notifiers(..., br->dev);.
>
> I still don't see why, in fact trying to verify this I am wondering now
> if instead we should actually fix br_features_recompute() to take into
> consideration BR_ROOT_BLOCK as below. Notice how netdev_update_features()
> is called above even if the MAC address did not change, just as is done
> on br_del_if(). There is an NETDEV_FEAT_CHANGE event so would it be more
> appropriate we just call
>
> call_netdevice_notifiers(NETDEV_FEAT_CHANGE, p->br->dev)
>
> for both the above then and also br_del_if()?

  Luis

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.