Moving this back on-list.
On Fri, Sep 9, 2011 at 3:13 PM, Pravin Shelar <pshe...@nicira.com> wrote:
>>> diff --git a/datapath/vport-internal_dev.c b/datapath/vport-internal_dev.c
>>> index f777637..492bef7 100644
>>> --- a/datapath/vport-internal_dev.c
>>> +++ b/datapath/vport-internal_dev.c
>>> -static int internal_dev_change_mtu(struct net_device *netdev, int new_mtu)
>>> -{
>>> - struct vport *vport = internal_dev_get_vport(netdev);
>>> -
>>> - if (new_mtu < 68)
>>> - return -EINVAL;
>>> -
>>> - if (new_mtu > dp_min_mtu(vport->dp))
>>> - return -EINVAL;
>>> -
>>> - netdev->mtu = new_mtu;
>>> - return 0;
>>> -}
>>
>> You can't remove this function because this is what the network stack
>> calls to change the MTU of the device. Removing it means that nothing
>> can ever change the MTU.
>>
>
> I am not sure what do u mean?
> dev_set_mtu() is checking for ->change_mtu() if that is null it sets new MTU.
> do you want extra check on new MTU set for internal_dev in kernel?
Sorry, I forgot that dev_set_mtu() will go ahead and set it if there
is no function. I do think that we need to enforce the minimum frame
size though since we know that we are dealing with Ethernet devices
and the generic code does not.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
