This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AM4376: Netdev watchdog does not reset

Part Number: AM4376


Reporting a customer issue on Linux:

The problem is triggered by a NETDEV WATCHDOG -event coming from the Linux Kernel. The watchdog is triggered when certain ethernet timeouts occur, which may sometimes happen in our System if e.g. the network subsystem is chocked by a higher priority process.

While this is not ideal to begin with, we have to live with this situation sometimes.

The problem yet is that the CPSW in the AM4 is not recovering from the Watchdog event it seems. Normal operation on other tested ethernet hardware is  the subsystem is reset and then resumes operation. On CPSW, this is not the case. The network stays dead, we have to explicitly bring the ethernet interface down and up again (via ifdown/ifup from Linux console) for it to come online again.

Is this a known (maybe even intentional) behavior of the CPSW?

Kernel Version is 4.9.178-rt131

  • Hi,

    Based on the kernel version provided and conferring with the developer here are 2 commits that should fix the problem. Please also note that the patches provided may not apply cleanly with referenced kernel version. 

    A brief summary from the commit message is that when using an RT kernel this first patch fixes a race condition between the start xmit function and the tx handler that caused the tx queue to be disabled forever. 

    commit 62f94c2101f35cd45775df00ba09bde77580e26a

    Author: Grygorii Strashko <grygorii.strashko@ti.com>

    Date:   Tue Feb 6 19:17:06 2018 -0600

         net: ethernet: ti: cpsw: fix net watchdog timeout

    or this one:

    A brief summary of this commit is that the patch fixes a condition when tx timeout occurs the tx queues will need to be emptied and re-initialized.

    commit 75514b6654859e0130b512396dc964d2a9e84967

    Author: Grygorii Strashko <grygorii.strashko@ti.com>

    Date:   Fri Mar 31 18:41:23 2017 -0500

         net: ethernet: ti: cpsw: wake tx queues on ndo_tx_timeout

    Best Regards,

    Schuyler