RDMA write failing with RAE (AETH syndrome 96)

First column is source and second column is destination.

12.12.12.10 is Mellanox and 12.12.12.12 is our XRNIC target. Last but column is

AETH syndrome.

If you see for every write request going out there are 2

incoming responses, out of them one with with RAE (AETH syndrome ‘d96) and next

one is just positive ACK and there is not much timing difference between these

responses only 15 us.

This is protocol violation, Please suggest what would be wrong for this

Our card

root@xhd-ipsspdk1 mailto:root@xhd-ipsspdk1 ~]# ibv_devinfo

hca_id: mlx4_0

transport: InfiniBand (0)

fw_ver: 2.40.7000

node_guid: 7cfe:9003:009c:2400

sys_image_guid: 7cfe:9003:009c:2400

vendor_id: 0x02c9

vendor_part_id: 4103

hw_ver: 0x0

board_id: MT_1090111023

phys_port_cnt: 2

Device ports:

sniffer.pcap.zip (3.78 MB)