Flashing the Cisco CX623106A NIC to OEM firmware

[Jaket]

In this just for some updates.

 

[Thomas42]

Were you successful

 

[bitbckt]

Were you successful

Yes.

https://forums.servethehome.com/index.php?threads/flashing-the-cisco-cx623106a-nic-to-oem-firmware.40390/post-380647

 

[jnsdf]

I got only 9 and need 15 more in case someone is willing to sell them to me. Thanks!

 

[NablaSquaredG]

I got only 9 and need 15 more in case someone is willing to sell them to me. Thanks!

Find out who bought 700 pcs and ask them

 

[dreamkass]

Sorry is the jumper required after all? never done this, still waiting on my cards

 

[bitbckt]

Sorry is the jumper required after all? never done this, still waiting on my cards

Nope.

 

[NablaSquaredG]

Nope.

hu?

Well, you need to bring the card into livefish mode to flash...

You don't need a jumpter for that. A pair of sufficiently thin tweezers will also work

 

[bitbckt]

You don't need a jumpter for that. A pair of sufficiently thin tweezers will also work

Exactly!

 

[Slothstronaut]

Just for some confirmation, this does work. Was able to use a small jumper, boot the card and do the PSID change, power off then remove jumper, then power on and flash normally. Was able to convert all 20 cards from the Cisco flash to the fully-featured (with Crypto) Nvidia model MCX623106AC-CDA. Great guide! I had one freak out and start acting very strange in ESXi but the others have been great. Need to do some more troubleshooting whats going on.

The card acting up was the first card I tried to flash, and I ended up flashing it first to the non-crypto firmware, then changed it to the crypto and reflashed it. Not sure if that's what caused the weird behavior or what. Going to spend some time on it this weekend to see whats up. Basically the card is passing traffic from one port out the other when connected as redundant uplinks on a vDS in ESXi. This causes a massive network loop and brings the entire network to a halt eventually. Swapped out the card and it stops. I'm guessing something about messing with the firmware so much broke something with the hardware offload and its not doing what its supposed to.

Anyways, great guide, awesome that you found such a handy workaround!

 

[JBB]

What am I doing wrong? I have a ConnectX-6 DX from eBay that spits out the following:

MST devices:
------------

mt530_pciconf0

Soldered the jumper on and the card also burns constant orange and gets the following when updating:

flint -d mt530_pciconf0 --psid MT_0000000437 -i fw-ConnectX6Dx-rel-22_38_1900-MCX623106AS-CDA_Ax-UEFI-14.31.20-FlexBoot-3.7.201.signed.bin -ocr --nofs --allow_psid_change --yes burn

-W- Firmware flash cache access is enabled. Running in this mode may cause the firmware to hang.
-E- Cannot open Device: mt530_pciconf0. MFE_NO_FLASH_DETECTED

 

[twinkletoes]

What am I doing wrong? I have a ConnectX-6 DX from eBay that spits out the following:

-W- Firmware flash cache access is enabled. Running in this mode may cause the firmware to hang.
-E- Cannot open Device: mt530_pciconf0. MFE_NO_FLASH_DETECTED

I got MFE_NO_FLASH_DETECTED using older version of flint, but not all the time. Most of the time!

Using the latest mft from NVIDIA, it worked every time.

 

[twinkletoes]

Well, this turned out to be easier than I anticipated.

I issued the following:

# mst start
# mst status // /dev/mst/mt4125_pciconf0 is the CX-6.
# flint -d /dev/mst/mt530_pciconf0 --psid MT_0000000437 -i fw-ConnectX6Dx-rel-22_35_2302-MCX623106AS-CDA_Ax-UEFI-14.28.16-FlexBoot-3.6.805.signed.bin -ocr --nofs --allow_psid_change --yes burn
# shutdown -r now

The 'mst start' isn't necessary. That appears to run some utility that doesn't do anything in this application.

Also I don't think it's necessary to blacklist the mlx5_core or the i2c.

At least, from what I can tell, when you jumper JP7, it changes the PCI device ID to the Mellanox flash thingie and nothing wants to attach anyways.

Just flint away, using the latest MFT...

 

[bitbckt]

The 'mst start' isn't necessary.

I wrote exactly what was required for a successful flash. You‘re welcome to a refund for what you paid for the information. Good luck.

 

[twinkletoes]

Hey, just sayin'. Your directions were excellent, but I don't think mst start is required. On newer NVIDIA Mellanox Firmware Tools, the 'mst start' command is now called 'mst server start' and those Mellanox services are not used in the flint process. Flint talks directly to the card and does the flash either way.

 

[bitbckt]

The appropriate way to say that is along the lines of: “as of MFT version XYZ, mst start is no longer necessary.”

I provided the environment details to avoid exactly that sort of future visitor sharpshooting BS.

 

[twinkletoes]

Using NVIDIA MFT 4.26.0 I got several flashed without using mst start.

 

[dreamkass]

Using NVIDIA MFT 4.26.0 I got several flashed without using mst start.

Could you post the procedure?

 

[Michal_MTTJ]

Nope.

You don't need a jumpter for that. A pair of sufficiently thin tweezers will also work

Just for some confirmation, this does work. Was able to use a small jumper, boot the card and do the PSID change

Hi, I might have a strange question about J7 pins (maybe JP2 ???). I just received the "Dell Mellanox CX6" card. I converted it from Dell IB to Dell IB/Ethernet 100GbE ("flint" command under windows10 - like Cx-3/4/5). but I can't flash to OEM Mellanox 200GbE because of "Secure FW" and PSID from Dell not OEM Mellanox (flashing MT_0000000223 into DEL0000000013).

My question to you who found the pins - which pins are J7? Because there are many around "J7" or could it be JP2?
can you please look into the photo and tell which pins should I connect?



or maybe talking about JP2 ?

 

[NablaSquaredG]

or maybe talking about JP2 ?

Yes JP2