I've managed to setup VXLAN forwarding using the BVI approach, and have a
question regarding the overlay VRF.
when I create a vxlan tunnel using:
create vxlan tunnel src 11.11.11.3 dst 100.0.0.2 vni 50 encap-vrf-id 1
decap-next l2
encap-vrf-id 1 is the underlay VRF. From what I can see, after p
Hi,
I'm not interested in bridging so I tried creating the tunnels in L3 mode
and perform routing between VNI 50 and VNI 70 using 2 linux namespaces :
>
> create loopback interface
> set interface mac address loop0 02:10:18:34:9f:cd
> set interface ip address loop0 11.11.11.3/24
> set int l2 br
Hello,
All of the VXLAN use cases I've found make VPP act as a VXLAN endpoint, example:
https://wiki.fd.io/view/VPP/Using_VPP_as_a_VXLAN_Tunnel_Terminator
The packet is either going through vxlan encap or decap. What I want is to have
a packet
flow that:
1. performs vxlan decap -> routing on th
Hi Neale,
I have a custom plugin in which I want to extract this information and I much
prefer this approach.
Do you have any references on how to walk the FIB graph or existing features
that perform similar tasks?
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View
posting again because not sure message went through:
Thanks for the quick reply Stanislav,
I'm inserting hooks into the VPP control plane in order to offload certain
information to an external hardware dataplane.
In the case of route resolutions, I want to be able to say : "Ok, I got the arp
re
Hello,
>From reading the routing data model:
https://fdio-vpp.readthedocs.io/en/latest/gettingstarted/developers/fib20/routes.html#id2
I understand that when I add a static route (using the "ip route add" command)
to an attached next hop, VPP will resolve that
route, which is triggered (in my cas
using DPDK with Mellanox MT27800 [ConnectX-5] (100G)
show buffers indeed shows a leak:
Pool Name Index NUMA Size Data Size Total Avail Cached Used
default-numa-0 0 0 2496 2048 430185 430185 0 0
default-numa-1 1 1 2496 2048 430185
When testing Damjan's version under traffic (using trex with a few thousand pps)
after about 5 minutes into the test vlib_buffer_clone fails to create 2 copies
and vpp crashes due to double-free error.
looks like the buffer pool leaks.
1. Is it the node function responsibility to vlib_buffer_free
Hemant,
in my case, VPP acts as a bump in the wire, not as an interface to an
application, so my metadata about the original packet is limited.
Damjan,
the hack with reacquiring the pointer after the clone is working as expected.
I found an example using the L2 flag approach in src/vnet/interfa
Damjan,
After looking at the vlib_buffer_clone_256 function I realize that it modifies
the original buffer pointer, like you said.
my packets are coming in down a custom node path (originating from an asic data
plane), so they will always have the l2 header.
The node that performs the cloning is
[Edited Message Follows]
moving the buffer backwards by 14 looks correct for small packets, but for 1500
byte packets I get `truncated-ip - 14 bytes missing` error from tcpdump. After
the clone I'm restoring the original offset:
vlib_buffer_advance (b0, -14);
u16 n_cloned = vlib_buffer_clone (vm
moving the buffer backwards by 14 looks correct for small packets, but for 1500
byte packets I get `truncated-ip - 14 bytes missing` error from tcpdump
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#19176): https://lists.fd.io/g/vpp-dev/message/191
Hemant,
I think you meant a negative 14, since the buffer points to the start of the IP
header, and we need to go backwards the size of the Ethernet header. Is there a
programmatic way to set the buffer to point to the start of the data, without
knowing the packet structure? offset calculations
Damjan's version works :) However, according to tcpdump the packet going to the
host interface contains only IP information,
even though I'm cloning bi0 buffer which also contains the Ethernet header.
How do I make it send the entire raw packet to the host interface?
-=-=-=-=-=-=-=-=-=-=-=-
Links
Damjan,
I'll test your version later today and report
Hemant,
Correct me if I'm wrong, but span mirroring mirrors all the traffic between
physical interfaces. My use case
is mirroring some of the traffic coming to a node (according to some business
logic) to a host interface.
-=-=-=-=-=-=-=-=
> why do you need to open host_if_frame for each packet
If you refer to the fact that I can just create one frame outside the loop, you
are correct :).
I tried to get the code to the simplest working example that will correctly
mirror the packets so I started to "dumb down" things :) I'm aware
t
;
> Executing `to_frame = vlib_get_frame_to_node (vm, node_index);`
> n_left_from times and `vlib_put_frame_to_node (vm, node_index, to_frame);`
> only once doesn’t sound sane. You are leaking frames…
>
> —
> Damjan
>
>
> > On 08.04.2021., at 14:12, David Gohberg wrote:
> >
&
On Thu, Apr 8, 2021 at 2:35 PM Damjan Marion wrote:
>
> Setting frame->n_vectors may help :)
>
> —
> Damjan
>
> On 08.04.2021., at 13:11, David Gohberg wrote:
>
> Thanks Damjan,
>
> This clarifies things regarding the cloning mechanism. However using
> n_clon
bytes of payload from
> original buffer
>- set b->next_buffer_index to point to the original buffer
> - advance b->current_data for VLIB_BUFFER_CLONE_HEAD_SIZE
> - set b->ref_cnt of the original buffer (and all his tails) to be n_clones
>
>
> Hope this explains…
>
&g
Hi,
In my node processing function, I want some buffers to be cloned and sent
to a specific host interface while also be sent to the "normal" interface
resolved in the node.
At the end, the node should be able to send the buffers to two different
interfaces simultaneously.
packet -> node -> inter
20 matches
Mail list logo
