Ilya Maximets <[email protected]> writes:

> On 8/31/23 09:15, Frode Nordahl wrote:
>> On Wed, Aug 30, 2023 at 9:30 PM Paolo Valerio <[email protected]> wrote:
>>>
>>> From: hepeng <[email protected]>
>>>
>>> The patch avoids the extra allocation for nat_conn.
>>> Currently, when doing NAT, the userspace conntrack will use an extra
>>> conn for the two directions in a flow. However, each conn has actually
>>> the two keys for both orig and rev directions. This patch introduces a
>>> key_node[CT_DIRS] member as per Aaron's suggestion in the conn which
>>> consists of a key, direction, and a cmap_node for hash lookup so
>>> addressing the feedback received by the original patch [0].
>>>
>>> [0]
>>> https://patchwork.ozlabs.org/project/openvswitch/patch/[email protected]/
>>>
>>> Signed-off-by: Peng He <[email protected]>
>>> Co-authored-by: Paolo Valerio <[email protected]>
>>> Signed-off-by: Paolo Valerio <[email protected]>
>> 
>> Thanks alot for working on this, should we perhaps reference the
>> original bug report, i.e:
>> Reported-by: Michael Plato <[email protected]>
>> Reported-at: 
>> https://mail.openvswitch.org/pipermail/ovs-discuss/2022-September/052065.html
>
> Can be added while applying, I think.  It also may be worth adding
> a sentence about fixing the assertion to the commit message.

Done.

>> 
>> We have a reproducer for the issue and we no longer see it occurring
>> with this patch.
>> Tested-by: Frode Nordahl <[email protected]>
>
> Thanks!

Thanks everyone!  I've applied and backported down to branch-3.0, and
will work on the backport to branch-2.17.

>> 
>> Is there a plan for a backport, we have users on OVS 2.17 that would
>> be interested in having this fixed, and AFAICT this patch does not
>> cleanly backport.
>
> Yes, the plan was to get this one reviewed, accepted and backported
> down to 3.0.  3.1 and 3.0 seems to require only minor rebase (no support
> for SCTP and parent key dumps).  Then post a backport patch for 2.17 to
> get it reviewed, since it will have some extra modifications.
>
> Aaron, I suppose your Ack from v2 still mostly relevant, but could
> you please take another look at v3?
>
> From my side the code looks fine:
>
> Acked-by: Ilya Maximets <[email protected]>
>
> Best regards, Ilya Maximets.

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to