On Thu, Mar 8, 2012 at 3:26 PM, Ben Pfaff <[email protected]> wrote:
> On Thu, Mar 08, 2012 at 07:26:36AM -0800, Pravin B Shelar wrote:
>> Fixed according to comments from Ben.
>> v1-v2:
>>      - Changed name from netdev_policy_error to netdev_policing_error.
>>
>> --8<--------------------------cut here-------------------------->8--
>>
>> Signed-off-by: Pravin B Shelar <[email protected]>
>
> The subject line still says "policy".
>
> I'm a little reluctant to try to cache policing settings too
> aggressively.  The kernel API is kind of difficult to deal with and
> I'm not 100% certain that we do it exactly right.  I think it might be
> too much to, after we get one kernel error back, refuse to ever try
> setting any other values.
>
> So, how about this: we always save the desired settings into our data
> structure, regardless of whether the previous call was successful or
> not.  Thereafter, we cache the error status only for future calls that
> supply the same desired settings, but call into the kernel for another
> try for any future call that supplies different settings.
>
> Does that make sense?

yes, its good idea. I will post updated patch.

Thanks,
Pravin.
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev

Reply via email to