[ovs-discuss] Resubmit vs Recirculate

K.R Kishore krkishore at yahoo.com
Mon Aug 5 19:58:17 UTC 2013


Thanks for the quick response. The answer is not clear/complete to me, let me rephrase the question based on what I understand.

- The ofproto-provider which implements the parser and contains the Nicira extension module (for OF) will interpret OF commands and derive flow entries that it plumbs into the datapath. Do commands that require "Resubmit" go only in the user space datapath? Does kernel datapath ever deal with multiple table lookups?

- Secondly, I see a chain of activity (http://openvswitch.org/pipermail/dev/2013-June/028032.html) regarding submission (not release) of "Recirculation" feature into OVS and I can gather from the context of the thread mentioned above that it is related to supporting MPLS packet handling (at least initially). Does this feature result in entries in the user space datapath table or kernel space table or both? If only the former, what is the packet flow when packets that require "Recirculation"? Does it take the kernel-table miss path? Is there a special flow/path to handle this instead of a miss?

thx,
Kishore



>________________________________
> From: Ben Pfaff <blp at nicira.com>
>To: K.R Kishore <krkishore at yahoo.com> 
>Cc: "discuss at openvswitch.org" <discuss at openvswitch.org> 
>Sent: Monday, August 5, 2013 10:54 AM
>Subject: Re: [ovs-discuss] Resubmit vs Recirculate
> 
>
>On Mon, Aug 05, 2013 at 10:48:25AM -0700, K.R Kishore wrote:
>> I wanted to understand the difference between the Resubmit and
>> Recirculate features in OVS. From what I can gather resubmit exists
>> in an extension module and supports use of multiple OpenFlow tables
>> while Recirculate is part of the standard OVS to support MPLS like
>> operations, correct?
>
>Resubmit is a Open vSwitch extension to OpenFlow.
>OVS doesn't have a recirculate feature.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20130805/bcc2bc76/attachment.html>


More information about the discuss mailing list