On Mon, Jun 23, 2014 at 6:10 PM, Joe Stringer
wrote:
> On 24 June 2014 05:26, Alex Wang wrote:
>
>> (snip)
>>
>> Totally agree,
>>
>> So, I'm thinking this solution,
>>
>> - in revalidate_ukey(), if need_revalidate is set, we always:
>> 1. save the current xcache to old_xcache pointers.
>> 2. co
On 24 June 2014 05:26, Alex Wang wrote:
> (snip)
>
> Totally agree,
>
> So, I'm thinking this solution,
>
> - in revalidate_ukey(), if need_revalidate is set, we always:
> 1. save the current xcache to old_xcache pointers.
> 2. conduct full revalidation (but never push stats during revalidation)
Hey Joe,
Sorry for the delayed reply,
Please see my comments inline,
On Sun, Jun 22, 2014 at 11:29 PM, Joe Stringer
wrote:
> Alex, could you review this series?
>
> I think that the correct solution is to perform "retroactive
> side-effects", that is, perform side effects as though they had ha
On second thought, I noticed that this doesn't fix odd behaviour in the
netflow expiration test.
I sent out a patch to revert the original patch:
http://openvswitch.org/pipermail/dev/2014-June/041868.html
On 23 June 2014 18:29, Joe Stringer wrote:
> Alex, could you review this series?
>
> I th
Alex, could you review this series?
I think that the correct solution is to perform "retroactive side-effects",
that is, perform side effects as though they had happened at the time that
the flow was hit. Anything that may hide these effects (like mac table
flush) would need to be taken into accou
Commit a48c85b2d6 (revalidator: Use xcache when revalidation is
required.) introduced a bug where xlate_actions effects from the old
cache would be performed. This could cause stale mac-learning entries to
re-appear, and in some cases, learnt flows to be re-learnt even after
the parent flow is dele
