Nucleo F429ZI with LWIP6 ethernet fails on eth.connect() with error -3010

[teetak01]

• Type: Bug
• Priority: Blocker

Nucleo board reliable fails to get IPv6 address with LWIP6 enabled and using easy-connect.

10:45:36.473 | D1 <<< DutThread: [EasyConnect] IPv6 mode
10:45:36.474 | D1 <<< DutThread: [EasyConnect] Using Ethernet
10:45:53.207 | D1 <<< DutThread: [EasyConnect] Connection to Network Failed -3010!
10:45:53.208 | TC     MainThread: Test Case fails because of: Failed to initialize connection

In the same network and with same application, K64F is able to connect with IPv6.

[teetak01]

@MarceloSalazar @bridadan @JanneKiiskila

[0xc0170]

cc @bcostm @adustm @LMESTM @jeromecoutant

[JanneKiiskila]

Hei,

some extra info. Easy-connect, using tip of master, mbed OS 5.4.6 release.
You can use mbed-os-example-client for reproducing the issue.

https://github.com/ARMmbed/mbed-os-example-client

That already has the same things in place.

For the mbed_app.json, use this; https://github.com/ARMmbed/mbed-os-example-client/blob/master/configs/eth_v6.json

[bridadan]

@teetak01 We have seen similar issues. Do you have multiple Nucleo F429ZI devices running network tests at once on the same network? All Nucleos 429ZI devices have the same MAC address, so it can fail to get an address this way.

EDIT: I was totally wrong here, please ignore...

[teetak01]

@bridadan Yes we do, multiple devices. Should something like this be implemented for every board if the Hardware does not provide a unique one? #550

[JanneKiiskila]

Yes, this is a serious problem and really, we should allow any board that doesn't support unique MAC address to be a reference for anything, really. It's not that a difficult thing to implement and one can use for example the unique chip id as the root for that.

[jeromecoutant]

Hi
Note that ST doesn't provide any MAC address for any boards.
You have to code it.

[JanneKiiskila]

Why? Code how, you should provide instructions.

[SeppoTakalo]

I'm not sure that I'm looking at the correct location, but for me it looks like the functionality is already there:

https://github.com/ARMmbed/mbed-os/blob/master/features/FEATURE_LWIP/lwip-interface/lwip-eth/arch/TARGET_STM/stm32xx_emac.c#L480
https://github.com/ARMmbed/mbed-os/blob/master/features/FEATURE_LWIP/lwip-interface/lwip-eth/arch/TARGET_STM/stm32xx_emac.c#L498

Similar to NXP boards (K66F and K64F):
https://github.com/ARMmbed/mbed-os/blob/master/targets/TARGET_Freescale/TARGET_MCUXpresso_MCUS/TARGET_MCU_K64F/TARGET_FRDM/mbed_overrides.c#L45

Both seems to be using chip's unique ID reqisters to generate MAC addresses for testing purposes. Why this doesn't work, is beyond me...

[jeromecoutant]

Yes, MAC coding is already there.

@bridadan

All Nucleos 429ZI devices have the same MAC address

This is not normal.
Is it possible to display this MAC address in one test?
And ideally add this kind of code:

uint32_t ST_UID[3] = {0};
HAL_GetUID(ST_UID);
printf("Unique ID %08x %08x %08x\n", ST_UID[0], ST_UID[1], ST_UID[2]);

[teetak01]

It most likely is not MAC issue. I checked our logs from couple of last days:

Board	MAC
079602210B2668133462FEB8	00:80:e1:2b:00:20
079602210B2668133462FEB8	00:80:e1:2b:00:20
079602210B2068073464FEAC	00:80:e1:2d:00:34
079602210B3368093477FEA2	00:80:e1:3e:00:3a
079602210B3368093477FEA2	00:80:e1:3e:00:3a

So a specific board always has its own unique ID.

[bridadan]

I ran some tests locally, turns out I was mistaken, thanks for not taking my word for it @teetak01!

Did this bug start appearing on 5.4.6? Was it ok with 5.4.5?