I want to filter GTP-U packets by Inner Source IP on BF3 device

Infrastructure & Networking BlueField
1

Hi, I’m configuring an UPF application using DPDK now.
I want to filter GTP-U packets by Inner Source IP (in case of me, its sent from packet generator TRex, varying from 239.0.238.1 to 239.0.238.4) on BF3 NIC, by hardware offload. I set FLEX_PARSER_PROFILE_ENABLE value to 3, and checked again after reboot. Then I made a flow rule, using rte_flow API, which is directing packets to queues connected Worker Core according to the value of last octet from Inner source IP. For example, if the packet’s last octet is 1 then forward it to queue 0, and for 2, forward to queue 1…, and so on. I tried various patterns: ETH | IPV4 | UDP | GTPU | IPV4 | UDP | END, ETH | IPV4 | UDP, ETH | IPV4 | UDP | GTPU. In case of ETH | IPV4 | UDP, the flow creation prints out success, even if application of the rule during packet transmission test failed: it goes to only randomly selected one queue. And for the others,

=== GTP-U TEID Flow Rule Debug Info ===
Port: 1, TEID: 0x080004f1, Queue: 0
Expected GTP-U TEID pattern: 0xf1040008 (mask: 0xffffffff)
Flow validation failed: item not supported (type=13)
=== GTP-U TEID Flow Rule Debug Info ===
Port: 1, TEID: 0x080004f2, Queue: 1
Expected GTP-U TEID pattern: 0xf2040008 (mask: 0xffffffff)
Flow validation failed: item not supported (type=13)
=== GTP-U TEID Flow Rule Debug Info ===
Port: 1, TEID: 0x080004f3, Queue: 2
Expected GTP-U TEID pattern: 0xf3040008 (mask: 0xffffffff)
Flow validation failed: item not supported (type=13)
=== GTP-U TEID Flow Rule Debug Info ===
Port: 1, TEID: 0x080004f4, Queue: 3
Expected GTP-U TEID pattern: 0xf4040008 (mask: 0xffffffff)
Flow validation failed: item not supported (type=13)
NIC Initialization success.

these message comes out, and flow creation is failed.

I tried both on DPDK 22.11 and DPDK 25.03, and

root@localhost:/opt/mndu# mlxconfig -d 0000:03:00.1 query FLEX_PARSER_PROFILE_ENABLE

Device #1:

Device type: BlueField3
Name: 900-9D3B6-00CV-A_Ax
Description: NVIDIA BlueField-3 B3220 P-Series FHHL DPU; 200GbE (default mode) / NDR200 IB; Dual-port QSFP112; PCIe Gen5.0 x16 with x16 PCIe extension option; 16 Arm cores; 32GB on-board DDR; integrated BMC; Crypto Enabled
Device: 0000:03:00.1

Configurations: Next Boot
FLEX_PARSER_PROFILE_ENABLE 3

root@localhost:/opt/mndu# lspci | grep Ethernet
03:00.0 Ethernet controller: Mellanox Technologies MT43244 BlueField-3 integrated ConnectX-7 network controller (rev 01)
03:00.1 Ethernet controller: Mellanox Technologies MT43244 BlueField-3 integrated ConnectX-7 network controller (rev 01)

my device is this.

root@localhost:/opt/mndu# modinfo mlx5_core | grep ^version:
version: 23.07-0.5.0

root@localhost:/opt/mndu# dpkg -l | grep rdma-core
ii python3-pyverbs:arm64 2307mlnx47-1.2307050 arm64 Python bindings for rdma-core
ii rdma-core 2307mlnx47-1.2307050 arm64 RDMA core userspace infrastructure and documentation

and these are kernel module and rdma-core version.

root@localhost:/opt/mndu# mlxfwmanager
Querying Mellanox devices firmware …

Device #1:

Device Type: BlueField3
Part Number: 900-9D3B6-00CV-A_Ax
Description: NVIDIA BlueField-3 B3220 P-Series FHHL DPU; 200GbE (default mode) / NDR200 IB; Dual-port QSFP112; PCIe Gen5.0 x16 with x16 PCIe extension option; 16 Arm cores; 32GB on-board DDR; integrated BMC; Crypto Enabled
PSID: MT_0000000884
PCI Device Name: /dev/mst/mt41692_pciconf0
Base MAC: a088c20b5a16
Versions: Current Available
FW 32.39.4082 N/A
PXE 3.7.0300 N/A
UEFI 14.33.0012 N/A
UEFI Virtio blk 22.4.0014 N/A
UEFI Virtio net 21.4.0013 N/A

Status: No matching image found

Finally, These is current FW version.

  1. My first question is that, should I upgrade kernel module, and rdma-core to newer version to figure out whether tunnel_stateless_gtp is enabled? Is this necessary for my goal to filter packets by Inner source IP?
  2. Do I need to upgrade BF3 FW to enable GTP packets filtering and Inner source IP filtering?
  3. Are there any additional conditions or requirements to perform this? If there are any missings, could you let me know?

Regards,
Sunghyun Jang
SNETICT R&D employee in South Korea

2

Hi Sunghyun,

Thanks for your question.

According to my check, mlx5 supports only RTE_FLOW_ITEM_TYPE_GTP flow item type.

The issue may be in the rule pattern.

i.e.

You configured:

ETH | IPV4 | UDP | GTPU | IPV4 | UDP | END

But it should be:

ETH | IPV4 | UDP | GTP | IPV4 | UDP | END

it will match GTP-U packets as well.

Best Regards,

Anatoly

4

Thanks for your detailed reply.
After I searched on my code, however, there are only RTE_FLOW_ITEM_TYPE_GTPU, and there was no file that used RTE_FLOW_ITEM_TYPE_GTP.
Here, I have another question. To parse GTP packets(including GTP-U and GTP-C) on BlueField-3 DPU, I want to know that if I must use DOCA.
To check if the hardware currently supports GTP field parsing on BF3 SmartNIC, using hardware offload, I used testpmd first.
On that, it was revealed that there is no ptype related to gtp, and it does not support rss related to gtp. That means, without DOCA support, BF3 SmartNIC cannot parse GTP packets, nor forward it according to inner src/dst IP.
Is this correct? Or are there any alternatives to parse GTP packets and forward it according to inner src/dst IP, without using DOCA?