We have noticed undesirable behavior when rejecting packets on an NT40A11_4x1G/10G card (FPGA ID: 200-9565-57-06-00).
The discarded packets are actually transmitted and truncated to 60 bytes. This behavior can be reproduced using the example provided (inline_example.c). This example behaves as expected, i.e., it transmits 1 packet out of 2 and therefore rejects 1 packet out of 2. However, when we capture (tcpdump) what is transmitted to the next device, we see that all (including discarded) packets are transmitted, but the discarded packets are truncated to 60 bytes, which is an issue for the downstream equipment.

Below an extract of 4 packets received by the downstream device after Napatech inline_example.c running (source/destination IP were anonymized):

13:15:19.240786 14:a9:d0:1f:32:14 > a8:2b:cd:b3:68:8e, ethertype IPv4 (0x0800), length 152: 211.152.62.11.53 > 43.85.193.164.24623: 52519 0/1/0 (110)

13:15:19.340791 14:a9:d0:1f:32:14 > a8:2b:cd:b3:68:8e, ethertype IPv4 (0x0800), length 60: truncated-ip - 384 bytes missing! 211.152.62.11.53 > 156.127.61.108.60956: 51284[|domain]

13:15:19.440787 14:a9:d0:1f:32:14 > a8:2b:cd:b3:68:8e, ethertype IPv4 (0x0800), length 212: 211.152.62.11.53 > 156.127.230.237.43168: 18269 8/0/0 A 138.199.26.53, A 185.246.211.22, A 185.246.211.5, A 185.246.211.19, A 138.199.26.52, A 138.199.26.50, A 138.199.26.47, A 138.199.26.43 (170)

13:15:19.540784 14:a9:d0:1f:32:14 > a8:2b:cd:b3:68:8e, ethertype IPv4 (0x0800), length 60: truncated-ip - 228 bytes missing! 211.152.62.3.53 > 43.84.172.19.60955: 5004[|domain]

This behavior was not observed on SmartNIC NT40E3 with same source code.. It has been also reproduced on different NT40A011 SmartNIC instances.

Is it the result of code/FPGA optimization for the packet discarding task or do we miss something?

Thanks you in advance for providing us with clarification.

Support 6cure.