phygate-tauri-l-imx8mm eth0 and eth1 are switched

[duncan-dommerdich-beebryte]

Hi,

In our phyGATE Tauri-L a ticket recently arrived with the following remarks :

Test case:

- configure both eth 0 and eth1 with DHCP

- connect only eth0 to a DHCP server

- connect only eth1 to a DHCP server

Result:

- eth0 on port LAN2 with MAC address 1

- eth1 on port LAN1 with MAC address 2

I've checked the following:

• no udev rule set the interface name
• ethernet0 alias is set in arch/arm64/boot/dts/freescale/imx8mm.dtsi, I've tried to change to ethernet1 nothing changed

If you need any further information, don't hesitate to ask :smiley:

[tremmet]

@duncan-dommerdich-beebryte Which BSP/kernel/yocto version are you using?

[duncan-dommerdich-beebryte]

thanks for the rapid answer, here are my checkout:

• poky: kirkstone-4.0.16 commit: 54af8c5e80ebf63707ef4e51cc9d374f716da603
• meta-oe: commit: 79a6f60dabad9e5b0e041efa91379447ef030482
• kernel : 5.15
• meta-freescale: commit: c82d4634e7aba8bc0de73ce1dfc997b630051571
• meta-phytec: tag: BSP-Yocto-NXP-i.MX8MM-PD23.1.0
• meta-arm: yocto-4.0.3
• meta-imx-bsp: rel_imx_5.15.71_2.2.2

[tremmet]

@duncan-dommerdich-beebryte Thank you. We will check internally.

[gportay]

@tremmet, @duncan-dommerdich-beebryte, In otherwords, I do think there is printing error on the LAN numbers; LAN1 is actually LAN2 and vice versa.

@tremmet, If I am not mistaken, the MAC-ID-1 is stored in the OTP/FUSE. How is stored the MAC-ID-2? in the Intel IGB internal memory?

[tremmet]

@gportay Yes, this is how it looks for me, too. I am waiting for feedback of the team that created the housing if there is anything behind this. As we mostly work without housing, we did not notice our self's.

And yes the MAC of the fec interface inside of the i.MX8MM is stored in FUSEs and the Intel one as an internal EEPROM.

[gportay]

@gportay Yes, this is how it looks for me, too. I am waiting for feedback of the team that created the housing if there is anything behind this. As we mostly work without housing, we did not notice our self's.

And yes the MAC of the fec interface inside of the i.MX8MM is stored in FUSEs and the Intel one as an internal EEPROM.

Thanks.

[tremmet]

@duncan-dommerdich-beebryte The LAN[1|2] numbering of the housing was independently done from schematics and BSP. It was created based on user expectations from top to bottom. See: https://www.phytec.de/cdocuments/?doc=woHCFw#HardwareandBSPReferenceManualphyGATETauriLi-MX8MMiniKitL1028e-A1-ElectricalConnection

What are your requirements here?

[gportay]

@duncan-dommerdich-beebryte The LAN[1|2] numbering of the housing was independently done from schematics and BSP. It was created based on user expectations from top to bottom.

Does the LAN1 expect to use the MAC-ID-1? and LAN2/MAC-ID-2 or the numbering is not correlated?

[tremmet]

@duncan-dommerdich-beebryte The LAN[1|2] numbering of the housing was independently done from schematics and BSP. It was created based on user expectations from top to bottom.

Does the LAN1 expect to use the MAC-ID-1? and LAN2/MAC-ID-2 or the numbering is not correlated?

It is directly the inverse.

MAC-ID-1 (i.MX8MM FEC) -> ETH0_CON -> X30 -> LAN2 MAC-ID-2 (Intel IGB) -> ETH1_CON -> X29 -> LAN1

[gportay]

MAC-ID-1 (i.MX8MM FEC) -> ETH0_CON -> X30 -> LAN2 MAC-ID-2 (Intel IGB) -> ETH1_CON -> X29 -> LAN1

Thank you very much for the clarification, this is pretty much confusing the numbers DOES NOT match, isn't it?

It took us a while to figure out that issue comes from the sticker/printing and not from the BSP itself.

It would have been nice to have this correspondence somewhere, directly printed in the sticker or in the documentation.

Just saying.

Again, thank you for the support.

[tremmet]

@gportay I agree. Thank you for your feedback. I asked the colleagues to improve the documentation about the ethernet interface naming.