Hi Lollita,

>       Thank you for update.  I have do quick browse on the source code of  
> "punt socket", but what is your consider about supporting VRF?

The current packet descriptor is:

 * Packet descriptor header. Version 1
 * If this header changes, the version must also change to notify clients.
typedef struct __attribute__ ((packed))
  u32 sw_if_index;              /* RX or TX interface */
  enum punt_action_e action;
} punt_packetdesc_t;

It should be just add fib_index to that structure and modify 
punt.c:punt_socket_rx_fd() to set the fib_index in packet.
Same in the other direction.


> BR/Lollita Liu
> -----Original Message-----
> From: Ole Troan [mailto:otr...@employees.org]
> Sent: Thursday, March 01, 2018 6:35 PM
> To: Lollita Liu <lollita....@ericsson.com>
> Cc: vpp-dev@lists.fd.io
> Subject: Re: [vpp-dev] FW: method for forward pacekt between vpp and kernel
> Lollita,
>> We are plan to implement application based on VPP with following
>> architecture. The application is running on container. All traffic will be 
>> received by physical NIC take over by VPP via DPDK. Data plane traffic will 
>> be handled by nodes developed by ourselves, and control plane traffic should 
>> be sent to or from Linux socket.
>> <image001.png>
>> Looks there are plenty of method on VPP can forward traffic with Kernel:
>>      • Using tuntap with: enable tuntap in startup.conf and configure “set 
>> punt udp”
>> It works, but only packet received from interface belong to IP table 0 will 
>> be send to Linux. VRF is not supported by tuntap. What is t
>>      • Using stn plugin. UDP packet can not handle by VPP but match STN rule 
>> will be sent to Linux via tap interface, but the ethernet header is 
>> hardcoded. The packet is able be captured by tcpdump, but will not arrive on 
>> socket layer. Looks it is been droped because of wrong MAC address. Then 
>> what is the use case of stn?
>>      • Using ip punt redirect via tap device. We have do simple test on it 
>> and looks vpp is plan to send arp to kernel via node ip4-arp. I just want 
>> check what is the standard use case of “ip punt redirect”? Is there any 
>> standard sample based on punt redirect?
>> In general, we have another question, is there any other method can fulfill 
>> our requirement with VRF support in vpp?
> The short answer is of course "it depends".
> If you want an unchanged Linux application you get into the issues of 
> mirroring each VPP interface into Linux and having the same IP address / MAC 
> address be used both in VPP and in Linux.
> And you can make hacks around those issues NAT or MAC NAT. I wouldn't go down 
> this path.
> On the other hand if you can change the application, then you could have a 
> single punt socket. Currently we have base implementation in:
> https://gerrit.fd.io/r/#/c/6721/
> It supports meta-data, which allows the control plane application to set 
> outgoing interface, and next-node (ip4-lookup, ip6-lookup or 
> interface-output).
> It's currently limited to UDP applications, but I have been thinking of 
> extending it to any type of control traffic. No limitation in what you can 
> send.
> We have also been thinking of extending it to using the memif interface for 
> higher performance than Unix domain sockets. Patches are welcome!
> Cheers,
> Ole


You receive all messages sent to this group.

View/Reply Online (#8386): https://lists.fd.io/g/vpp-dev/message/8386
View All Messages In Topic (1): https://lists.fd.io/g/vpp-dev/topic/13285373
Mute This Topic: https://lists.fd.io/mt/13285373/21656
New Topic: https://lists.fd.io/g/vpp-dev/post

Change Your Subscription: https://lists.fd.io/g/vpp-dev/editsub/21656
Group Home: https://lists.fd.io/g/vpp-dev
Contact Group Owner: vpp-dev+ow...@lists.fd.io
Terms of Service: https://lists.fd.io/static/tos
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub

Attachment: signature.asc
Description: Message signed with OpenPGP

Reply via email to