1

I bought a lot of four HP branded ConnectX card on Ebay. The cards are :

HP InfiniBand 4X DDR ConnectX PCI-E Gen2 Dual-Port HCA 483514-B2

My problem is that two out of four cards identify themselves as ConnectX EN cards.

This one seems ok:

04:00.0 Network controller: Mellanox Technologies MT26418 [ConnectX VPI PCIe 2.0 5GT/s - IB DDR / 10GigE] (rev a0)
    Subsystem: Mellanox Technologies MT26418 [ConnectX VPI PCIe 2.0 5GT/s - IB DDR / 10GigE]
    Capabilities: [48] Vital Product Data
        Product Name: Eagle DDR
        Read-only fields:
            [PN] Part number: 483514-B21          
            [EC] Engineering changes: A2
            [SN] Serial number: IL20944001             
            [V0] Vendor specific: PCIe Gen2 x8   
            [RV] Reserved: checksum good, 0 byte(s) reserved

Image type:      ConnectX
FW Version:      2.8.0
Device ID:       26418
Chip Revision:   A0
Description:     Node             Port1            Port2            Sys image
GUIDs:           0002c9030006828e 0002c9030006828f 0002c90300068290 0002c90300068291
MACs:                             000000000000     000000000001    
Board ID:         (HP_09D0000008)
VSD:            
PSID:            HP_09D0000008

This one seems to suffer from an identity disorder:

04:00.0 Ethernet controller: Mellanox Technologies MT26448 [ConnectX EN 10GigE, PCIe 2.0 5GT/s] (rev a0)
        Subsystem: Mellanox Technologies Device 0006
        Capabilities: [48] Vital Product Data
                Product Name: Eagle DDR
                Read-only fields:
                        [PN] Part number: 483514-B21          
                        [EC] Engineering changes: A2
                        [SN] Serial number: IL21007133             
                        [V0] Vendor specific: PCIe Gen2 x8   

Image type:      ConnectX
FW Version:      2.7.0
Device ID:       26448
Chip Revision:   A0
Description:     Port1            Port2
MACs:            0002c908d3da     0002c908d3db    
Board ID:         (MT_0A70120005)
VSD:            
PSID:            MT_0A70120005

Please note that all cards report the same part number (483514-B21). My uneducated guess is that two of them have been flashed with the wrong firmware. Is that possible ? If that's the case, how do I reflash them with the correct firmware ?

user228840
  • 11
  • 3
  • Hi Martin, what OS/Server will you be using? – hookenz Jul 01 '14 at 23:09
  • As I understand it, ConnectX cards can operate in IB or EN mode. They are multi-protocol. I think you can even run one port in IB and another as EN. At least the Linux drivers support that. – hookenz Jul 01 '14 at 23:11
  • Hi Matt, I'm using Debian wheezy. If I'm not mistaken, ConnectX cards support both protocols but ConnectX EN cards only support 10 GigE. – user228840 Jul 01 '14 at 23:18
  • You might be right. The VPI ones are the multi-protocol versions. – hookenz Jul 02 '14 at 00:01
  • Could you run `flint q dump` on each of the cards and send the results? Do the device GUIDs match the labels on the cards? – haggai_e Jul 02 '14 at 08:04
  • @haggai_e Here's the output of flint dc for all cards: http://pastebin.com/ZkSuLtYD http://pastebin.com/9uZ8Kr8C http://pastebin.com/f0G8MHHv http://pastebin.com/DNy0SVDH – user228840 Jul 02 '14 at 14:35
  • I will check the labels this afternoon – user228840 Jul 02 '14 at 14:39
  • @haggai_e I checked the GUIDs on the labels and they match. The part number is also the same. – user228840 Jul 02 '14 at 17:54
  • I'd probably try burning the non EN firmware to it. I used a process similar to this: http://www.panticz.de/mellanox & http://serverfault.com/questions/542031/is-it-possible-to-update-supermicro-connectx-3-firmware-with-newer-firmware-from. On Ubuntu. It should work for Debian too. – hookenz Jul 02 '14 at 19:58
  • It seems like you have the wrong firmware image on the second card. I believe the firmware for the part number you've posted is available at http://www.mellanox.com/oem/hp/resources_firmware_updates.php – haggai_e Jul 03 '14 at 07:49
  • I reflashed with 'flint -d /dev/mst/mt26448_pci_cr0 --allow_psid_change -i fw-25408-2_9_1000-MHGH29-XSC_A1.bin burn' and it seems to work. Thanks for your help – user228840 Jul 03 '14 at 13:29

0 Answers0