[ovs-dev] [PATCH v2] linux: Signal datapath that unaligned Netlink message can be received

[Jesse Gross]

On Tue, Dec 17, 2013 at 10:08 AM, Thomas Graf <tgraf at redhat.com> wrote:
> On 12/17/2013 06:49 PM, Jesse Gross wrote:
>>
>> I think there's also a potential downgrade issue if we add a new
>> feature to the list of capabilities - it won't automatically reset
>> since userspace is now using v2 of the netlink protocol. Obviously,
>> this isn't an issue yet but it we should make sure that it is
>> addressed before there is a release.
>
>
>>> Defining the NLM_F_REPLACE semantics is non trivial if we want to do
>>> more than just update the settings. I will propose this in a follow up
>>> patch.
>>
>>
>> Couldn't userspace just issue an OVS_DP_CMD_SET on start?
>
>
> Right, that works as well but introduces a small race compared to
> NLM_F_REPLACE which would be atomic. I think we can live with that.
>
> I will send a v3 of this patch with dpif-linux changed to issue
> OVS_DP_CMD_SET first and fall back to OVS_DP_CMD_NEW if no DP exists.

That sounds good to me. I have a hard time imagining a case where the
race condition would matter at all since we are still in the process
of starting up and therefore shouldn't be processing packets yet.