Ethernet does not work without active UART session

I could not upload .txt files so here are pastebin links:
No working ethernet: https://pastebin.com/1EDpUs8S
Working ethernet: https://pastebin.com/RRdNT6n2

To provide more details on the problem,

Interface is brought up but it assigns itself a “random” ip address based on the “/etc/networks” file:

I’ve tried to manually assign an ip address, changed back to dhcp, enable and disable the interface using connmanctl but with no luck. Ive manually enabled and disabled various services from “/etc/systemd/system/” folder and that still doesn’t help.

Am I missing something super simple that its hiding in plain sight?

Regards,

Kartik

• This reply was modified 5 years, 1 month ago by slisgrinder. Reason: incorrect console output

An update of my findings from yesterday:

I double checked the pull-up/down states of UART0 and UART2 pins and they are correct however, if I connect +3v3 to any of the UART0 or UART2 Rx lines, the system boots up fine and ethernet interface sets up and authenticates correctly to the network.

Is there a system service that is watching these lines?
Is there some timing requirement that I am missing?

Regards,
Kartik

Good afternoon Eshtaartha,

I have ensured that the internal pull up is enabled on UART0 Rx as well as an external pull-up on UART2 Rx. I am getting intermittent boots with networking configured correctly. When the in the “failed” state, I checked PHY operation using ethtool and mii-tool, the PHY is operational with an appropriate link configuration. Simply, networking service has configured the eth0 interface to a random address as part of the “link-local” address (169.254.x.x) from /etc/networks file.

– Manually assigning a static address doesn’t yield any results. I just see a second address using ifconfig but nothing on wireshark.
– Using dhclient to issue a dhcp request doesn’t yield any results
– A reset cannot be performed using ethtool “Operation not supported”
– A reset using mii-tool does perform a soft reset but yields no results

What did work was manually hard resetting the PHY (I used tweezers to pull the reset line of the PHY low) and rebooting – without external pull ups.

May be I am interpreting this incorrectly, I wonder if it has something to with the following advisory from the AM335x ERRATA “Advisory 1.0.23 – Ethernet Media Access Controller (EMAC) and Switch Subsystem Reset Isolation Feature is not Supported”. Could it possibly be “looking” at a different interface as part of the dual Gigabit interface?

As a note, I always have wireshark running to sniff the packets and “doesn’t yield any results” is I don’t see any packets on the line.

I appreciate you help,
Kartik

Good afternoon Neeraj,

It seems it is some weird timing issue between the PHY coming out of reset on power-up and the Octavo. I connected the reset line from the PHY to one of the GPIO on the Octavo, configured it as a GPIO with pull-up and assigned it to the MDIO interface definition in the device tree as a reset gpio (see: https://github.com/torvalds/linux/blob/4d88e3d24905eafa98cef0fc29365649ad8977b5/Documentation/devicetree/bindings/net/mdio.txt).

I am now curious to see if there is a software workaround to this hardware jumper mod so that we don’t have to do a design rev nor any modification to our existing inventory.

Thank you for your help!
Regards,
Kartik

Good afternoon Neeraj,

Thank you for your help thus far, its really appreciated. We will make the necessary changes to our design and rev it.

Will there be a change/note in the reference design? I only ask this in case someone has a similar problem down the road.

Regards,
Kartik