• State Resolved
  • Replies 3 replies
  • Subscribers 101 subscribers
  • Views 61 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

AM62P: Continuous pinging of the network port will get phy down and then up.

YuYuan Wang
Prodigy 140 points
Part Number: AM62P
Other Parts Discussed in Thread: DP83867IR

Tool/software:

Hi TI experts,

Our device uses AM62P52 SOC and DP83867IR phy. 

Currently, during the stress test of the network port (using another device to continuously ping our device), the following message occasionally appears on the debug console:

[ 8342.335374] am65-cpsw-nuss 8000000.ethernet eth0: Link is Down
[ 8342.360058] audit: type=1334 audit(1709344099.236:25): prog-id=23 op=LOAD
[ 8342.366936] audit: type=1334 audit(1709344099.248:26): prog-id=24 op=LOAD
[ 8342.373771] audit: type=1334 audit(1709344099.252:27): prog-id=25 op=LOAD
ping: sendto: Network is unreachable
root@am62pxx-ghmi:~# [ 8343.359342] am65-cpsw-nuss 8000000.ethernet eth0: Link is Up - 100Mbps/Full - flow control off
[ 8345.407383] am65-cpsw-nuss 8000000.ethernet eth0: Link is Down
[ 8346.431203] am65-cpsw-nuss 8000000.ethernet eth0: Link is Up - 100Mbps/Full - flow control off
[ 8374.078953] am65-cpsw-nuss 8000000.ethernet eth0: Link is Down
[ 8375.103622] am65-cpsw-nuss 8000000.ethernet eth0: Link is Up - 100Mbps/Full - flow control off
[ 8405.208808] audit: type=1334 audit(1709344162.088:28): prog-id=25 op=UNLOAD
[ 8405.215825] audit: type=1334 audit(1709344162.088:29): prog-id=24 op=UNLOAD
[ 8405.222810] audit: type=1334 audit(1709344162.088:30): prog-id=23 op=UNLOAD

We are using TI SDK10, and the hardware design is based on the EVM. However, this issue does not occur on the EVM.

I have no idea how to debug. 

If anyone knows, I would greatly appreciate it if you could guide us on how to debug this issue. 

Thanks!

  • +1
    TI__Genius 9025 points

    Hello YuYuan, 

    I noticed from your log that the link speed is 100Mbps. Is that speed intentional or simply because the link partner device you are connecting to your custom design (device under test - DUT) is limited to 100Mbps? If you can test 1Gbps, do you also see the problem occur?

    Can you describe your test setup? What device do you have connected to your DUT for this stress test? Is it a direct DUT-to-link partner connection or is there a network switch in between?

    The AM62Px should have fixed the MDIO corruption errata that affected AM62x and AM64x processors, so this unstable link shouldn't be due to an MDIO corruption. There could be a chance there is an issue from the Ethernet PHY side. To determine if that is the case, can you try following the steps specified in "3.4 Unstable Link Up Debug in 1Gbps communication" in this DP83867 Debug App Note? https://www.ti.com/lit/an/snla246c/snla246c.pdf 

    Although the steps are specifying 1Gbps, some steps such as the schematic and layout review info maybe helpful especially considering this issue is only observed on your custom design and not on the EVM.

    -Daolin

  • 0 in reply to Daolin Qiu
    Prodigy 140 points

    Thanks Daolin,

    Yes, we limited our ETH speed to 100Mbps by the following command

    /sbin/ethtool -s eth0 autoneg off speed 100 duplex full
/sbin/ethtool -s eth1 autoneg off speed 100 duplex full

    We set up a Windows laptop and used the ping command to continuously ping the device (the issue can be reproduced whether or not a switch is involved).

    Thanks for your reply, I will check it.

    Thanks!

  • 0 in reply to YuYuan Wang
    TI__Genius 9025 points

    Hi YuYuan,

    YuYuan Wang said:
    Yes, we limited our ETH speed to 100Mbps by the following command

    Do you observe the same issue when the speed is 1Gbps?

    YuYuan Wang said:
    Thanks for your reply, I will check it.

    Let me know if you have additional questions/issues after reviewing the steps in the application note.

    -Daolin