[ovs-discuss] _APPLY_ vs _WRITE_ actions
Ben Pfaff
blp at nicira.com
Thu Aug 20 17:51:24 UTC 2015
[adding the list back]
This may be a bug. Can you provide a small example that allows me to
reproduce the problem?
On Thu, Aug 20, 2015 at 08:56:01AM -0500, David Evans wrote:
> I have a header modifier (custom action) rule in a table, Apply actions,
> go to next table. - This works and output pcaps show the packet to be
> correctly modified.
> Then in the next table I have a filter rule. (the contents of the packet
> should' be visible to the filter rule now because of the header modifier
> in previous table)
> But do not get the expected match
>
> My confusion is OFPTTFPT13_APPLY_ACTIONS in the spec says it ought to
> apply immediately,
> should mean that the following table match¹ should be testing against the
> modified packet.
> But the mini flow is unchanged and following table matches don¹t happen.
>
> Anyhow, I ended up doing exactly the same as MPLS header removal does,
> to engage recirculation in compose action using was_mpls etc
> and CHECK_MPLS_RECIRCULATION() in the switch for do_xlate_actions()
> - which works.
>
>
> I know you¹re crazy busy. Thanks for your time helping out.
> Cheers,
>
> Dave.
>
> On 8/19/15, 11:13 PM, "Ben Pfaff" <blp at nicira.com> wrote:
>
> >On Mon, Aug 10, 2015 at 02:42:18PM -0500, David Evans wrote:
> >> I would like to OFPTFPT13_APPLY_ACTIONS and have have subsequent table
> >> matches on the modified packet.
> >> The OpenFlow spec indicates that apply actions are applied immediately,
> >>and
> >> the matches defined in subsequent tables are tests against the modified
> >> packet. but this doesn't seem to be so for me.
> >
> >I don't understand the sentence below:
> >
> >> Packet tests against an inner header are do not discover the inner
> >>traffic
> >> unless there has been a recirculation. (forcing an update of the packet
> >> metadata/flow data)
> >
>
>
More information about the discuss
mailing list