In the case of Open vSwitch some things, such as sFlow, want to do
accounting based on the input port.  The same mechanism is used by OVS
internally for packets that traverse userspace so it is important that
they get counted but it is possible for the controller to do this as
well.

The OpenFlow spec also says that in_port should be used in the case
where the table virtual port is used as an output.

On Thu, Sep 1, 2011 at 5:30 PM, Aaron Rosen <[email protected]> wrote:
> Sorry could you clarify what the the roll of in_port is for a packet out?
>
> Also the default value for inport for sending a packet in nox is
> OFPP_CONTROLLER fwiw.
>
> Thanks
>
>    def send_openflow_packet(self, dp_id, packet, actions,
>                             inport=openflow.OFPP_CONTROLLER):
>        """\brief Send an openflow packet to a datapath.
>
>        @param dp_id datapath to send packet to
>        @param packet data to put in openflow packet
>        @param actions list of actions or dp port to send out of
>        @param inport dp port to mark as source (defaults to Controller port)
>        """
>
>
>
> On Thu, Sep 1, 2011 at 8:02 PM, Jesse Gross <[email protected]> wrote:
>> That's the cause of the problem.  OVS is rejecting the packet out
>> because it is being sent from an out-of-range port.  It's not entirely
>> clear to me whether this is allowed by the OpenFlow spec and should be
>> supported.  In this case, I believe that OFPP_NONE is the correct port
>> to use, although OVS will not like that either.
>>
>> In the meantime, either OFPP_LOCAL or a real port should behave correctly.
>>
>> On Thu, Sep 1, 2011 at 4:21 PM, Aaron Rosen <[email protected]> wrote:
>>> Opps, I mean:
>>>
>>>  inport=openflow.OFPP_CONTROLLER
>>>
>>> I was talking about outport.
>>>
>>> Aaron
>>>
>>>
>>> On Thu, Sep 1, 2011 at 7:19 PM, Aaron Rosen <[email protected]> wrote:
>>>> I  tried both OFPP_FLOOD, in addition to OFPAT_OUTPUT 1 and 65534
>>>>
>>>> Thanks,
>>>>
>>>> Aaron
>>>>
>>>> P.S: attached in the first message is a dump of the OF packets that
>>>> are causing this.
>>>>
>>>>
>>>> On Thu, Sep 1, 2011 at 7:15 PM, Jesse Gross <[email protected]> wrote:
>>>>> What in_port are you supplying in the packet out message?  Is it 
>>>>> OFPP_NONE?
>>>>>
>>>>> On Thu, Sep 1, 2011 at 8:29 AM, Aaron Rosen <[email protected]> wrote:
>>>>>> No problem:
>>>>>>
>>>>>> It says,
>>>>>>
>>>>>> Sep  1 15:28:56 planetlab4 ovs-vswitchd: 00192|dpif|WARN|system@dp0:
>>>>>> execute 0,1 failed (Invalid argument) on packet 66:f3:43:38:f4:a2 >
>>>>>> Broadcast, ethertype ARP (0x0806), length 56: Request who-has
>>>>>> 10.43.100.104 tell 10.43.100.54, length 42
>>>>>> Sep  1 15:28:56 planetlab4 kernel: openvswitch_mod:
>>>>>> flow_metadata_from_nlattrs failed
>>>>>> Sep  1 15:28:57 planetlab4 ovs-vswitchd: 00193|dpif|WARN|Dropped 1 log
>>>>>> messages in last 1 seconds (most recently, 1 seconds ago) due to
>>>>>> excessive rate
>>>>>> Sep  1 15:28:57 planetlab4 kernel: openvswitch_mod:
>>>>>> flow_metadata_from_nlattrs failed
>>>>>> Sep  1 15:28:57 planetlab4 ovs-vswitchd: 00194|dpif|WARN|system@dp0:
>>>>>> execute 1 failed (Invalid argument) on packet 00:00:00:00:01:04 >
>>>>>> 01:23:20:00:00:01, ethertype Unknown (0x27a0), length 36:
>>>>>> Sep  1 15:28:57 planetlab4 ovs-vswitchd: #0110x0000:  0209 0764 7069
>>>>>> 643a 3130 3404 0302 0001  ...dpid:104.....
>>>>>> Sep  1 15:28:57 planetlab4 ovs-vswitchd: #0110x0010:  0602 0078 0000
>>>>>>                        ...x..
>>>>>> Sep  1 15:28:58 planetlab4 kernel: openvswitch_mod:
>>>>>> flow_metadata_from_nlattrs failed
>>>>>>
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Aaron
>>>>>>
>>>>>>
>>>>>> On Thu, Sep 1, 2011 at 3:28 AM, Jesse Gross <[email protected]> wrote:
>>>>>>> It does seem to be the same as the message on the mailing list, which
>>>>>>> I don't think was ever understood.  I tried and failed to reproduce
>>>>>>> this with packet outs and I think that Ben was previously unsuccessful
>>>>>>> with LLDP packets.
>>>>>>>
>>>>>>> Can you try the attached patch?  It just breaks out the various causes
>>>>>>> of failure instead of reporting a generic EINVAL.  It's a kernel
>>>>>>> patch, so you'll have to reload the kernel module and then look for
>>>>>>> relevant log messages in dmesg.
>>>>>>>
>>>>>>> On Mon, Aug 29, 2011 at 10:39 PM, Aaron Rosen <[email protected]> 
>>>>>>> wrote:
>>>>>>>> Also, If it helps debug I could provide the address of my controller 
>>>>>>>> and you
>>>>>>>> could point a switch at it (unless you think this is a configuration 
>>>>>>>> issue
>>>>>>>> on my part).
>>>>>>>> btw: found this on the mailing list but I'm not sure if the cause was 
>>>>>>>> ever
>>>>>>>> determined
>>>>>>>> http://www.mail-archive.com/[email protected]/msg01806.html
>>>>>>>> P.S: Added Ben to CC since he was helping me with this problem on IRC 
>>>>>>>> but
>>>>>>>> for some reason it had seemed to resolve itself at the time.
>>>>>>>> Thanks,
>>>>>>>> Aaron
>>>>>>>>
>>>>>>>> On Tue, Aug 30, 2011 at 1:10 AM, Aaron Rosen <[email protected]> 
>>>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> Yup the same,
>>>>>>>>> [root@planetlab4 ~]# dmesg | grep "Open vSwitch"
>>>>>>>>> Open vSwitch 1.2.90, built Aug 29 2011 15:49:49
>>>>>>>>> [root@planetlab4 utilities]# ./ovs-appctl  version
>>>>>>>>> ovs-vswitchd (Open vSwitch) 1.2.90
>>>>>>>>> Compiled Aug 29 2011 15:50:36
>>>>>>>>> Thanks,
>>>>>>>>> Aaron
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Tue, Aug 30, 2011 at 1:07 AM, Justin Pettit <[email protected]> 
>>>>>>>>> wrote:
>>>>>>>>>>
>>>>>>>>>> Can you confirm there's no mismatch?  If you're running the latest 
>>>>>>>>>> from
>>>>>>>>>> git, try running:
>>>>>>>>>>
>>>>>>>>>>        ./ovs-appctl version
>>>>>>>>>>
>>>>>>>>>> and:
>>>>>>>>>>
>>>>>>>>>>        dmesg | grep "Open vSwitch"
>>>>>>>>>>
>>>>>>>>>> to make sure they match.
>>>>>>>>>>
>>>>>>>>>> --Justin
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On Aug 29, 2011, at 7:22 PM, Aaron Rosen wrote:
>>>>>>>>>>
>>>>>>>>>> > I'm using the kernel version not the userspace ( unless my
>>>>>>>>>> > understanding of this is  unclear) Also, I'm near positive I'm not 
>>>>>>>>>> > mixing
>>>>>>>>>> > versions.
>>>>>>>>>> >
>>>>>>>>>> > Aaron
>>>>>>>>>> >
>>>>>>>>>> > On Mon, Aug 29, 2011 at 10:16 PM, Jesse Gross <[email protected]> 
>>>>>>>>>> > wrote:
>>>>>>>>>> > On Tue, Aug 30, 2011 at 12:30 AM, Aaron Rosen <[email protected]>
>>>>>>>>>> > wrote:
>>>>>>>>>> > > Hello,
>>>>>>>>>> > > I'm sending ARP packets from my openflow controller and 
>>>>>>>>>> > > everything
>>>>>>>>>> > > seems to
>>>>>>>>>> > > be working on hardware switches though on OVS I'm getting the
>>>>>>>>>> > > following
>>>>>>>>>> > > error in /var/log/messages.
>>>>>>>>>> > > ovs-vswitchd: 70867|dpif|WARN|Dropped 1 log messages in last 1
>>>>>>>>>> > > seconds (most
>>>>>>>>>> > > recently, 1 seconds ago) due to excessive rate
>>>>>>>>>> > >
>>>>>>>>>> > > ovs-vswitchd: 70868|dpif|WARN|system@dp0: execute 0,1 failed 
>>>>>>>>>> > > (Invalid
>>>>>>>>>> > > argument) on packet 66:f3:43:38:f4:a2 > Broadcast, ethertype ARP
>>>>>>>>>> > > (0x0806),
>>>>>>>>>> > > length 56: Request who-has 10.43.100.104 tell 10.43.100.105, 
>>>>>>>>>> > > length
>>>>>>>>>> > > 42
>>>>>>>>>> >
>>>>>>>>>> > I would check that userspace and kernel versions are the same.
>>>>>>>>>> >
>>>>>>>>>> >
>>>>>>>>>> >
>>>>>>>>>> > --
>>>>>>>>>> > Aaron O. Rosen
>>>>>>>>>> > Masters Student - Network Communication
>>>>>>>>>> > 306B Fluor Daniel
>>>>>>>>>> >
>>>>>>>>>> >
>>>>>>>>>> > _______________________________________________
>>>>>>>>>> > discuss mailing list
>>>>>>>>>> > [email protected]
>>>>>>>>>> > http://openvswitch.org/mailman/listinfo/discuss
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> Aaron O. Rosen
>>>>>>>>> Masters Student - Network Communication
>>>>>>>>> 306B Fluor Daniel
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> Aaron O. Rosen
>>>>>>>> Masters Student - Network Communication
>>>>>>>> 306B Fluor Daniel
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Aaron O. Rosen
>>>>>> Masters Student - Network Communication
>>>>>> 306B Fluor Daniel
>>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Aaron O. Rosen
>>>> Masters Student - Network Communication
>>>> 306B Fluor Daniel
>>>>
>>>
>>>
>>>
>>> --
>>> Aaron O. Rosen
>>> Masters Student - Network Communication
>>> 306B Fluor Daniel
>>>
>>
>
>
>
> --
> Aaron O. Rosen
> Masters Student - Network Communication
> 306B Fluor Daniel
>
_______________________________________________
discuss mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/discuss

Reply via email to