Linux Mint - Intel 82579LM NIC and Hardware Unit Hang

Question

A few months ago, I installed Linux Mint 19.1 x64 on a system with a Supermicro X9SCM system board. The X9SCM has two integrated NICs:

Intel 82574L Gigabit Network Connection
Intel 82579LM Gigabit Network Connection (Lewisville)

The Intel 82574L Gigabit Network Connection has been issue free.

The Intel 82579LM Gigabit Network Connection (Lewisville) is using a Linux kernel bridge with NetworkManager disabled.

Approximately every six minutes, the Intel 82579LM Gigabit Network Connection (Lewisville) reported the following:

e1000e 0000:00:19.0 eno1: Detected Hardware Unit Hang:
    TDH                  <d2>
    TDT                  <43>
    next_to_use          <43>
    next_to_clean        <d1>
    buffer_info[next_to_clean]:
    time_stamp           <103d3f01c>
    next_to_watch        <d2>
    jiffies              <103d3f8e0>
    next_to_watch.status <0>
    MAC Status             <40080083>
    PHY Status             <796d>
    PHY 1000BASE-T Status  <3c00>
    PHY Extended Status    <3000>
    PCI Status             <10>
e1000e 0000:00:19.0 eno1: Reset adapter unexpectedly
br0: port 1(eno1) entered disabled state
e1000e: eno1 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx
br0: port 1(eno1) entered blocking state
br0: port 1(eno1) entered forwarding state

Various ethernet cables and switch ports were tried, but it changed nothing.

Answer 1

The Intel 82579LM Gigabit Network Connection (Lewisville) defaults to offloading enabled, and disabling offloading works around the issue.

Determine NIC device:

ip addr show
2: eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel master br0 state UP group default qlen 1000

Determine whether offloading is enabled (on):

sudo ethtool -k eno1
tcp-segmentation-offload: on
generic-segmentation-offload: on

To test, temporarily (reset after every reboot) disable offloading:

sudo ethtool -K eno1 tso off
sudo ethtool -K eno1 gso off

The log entries for "Detected Hardware Unit Hang" and "Reset adapter unexpectedly" should cease.

To add a permanent configuration change with NetworkManager disabled:

sudo nano /etc/network/interfaces

Add the following:

post-up /sbin/ethtool -K eno1 tso off gso off

-> Ctrl + o -> Enter -> Ctrl + x

Restart the computer.

Determine whether offloading is disabled (off):

sudo ethtool -k eno1
tcp-segmentation-offload: off
generic-segmentation-offload: off

I spent several days troubleshooting this, and I hope it helps others.

---------- NetworkManager ----------

I only needed to disable offloading in this specific scenario with NetworkManager disabled and using a Linux kernel bridge for QEMU/KVM. Before I configured the bridge, I did not notice whether there was an issue with hangs and resets. If there is, you may be able to add the appropriate commands to a startup script to accomplish the same result while continuing to use NetworkManager. The following is untested:

sudo crontab -e

Press button "1" for nano.

On the last line enter the following:

@reboot sleep 10 && ethtool -K eno1 tso off
@reboot sleep 10 && ethtool -K eno1 gso off

sudo shutdown -r now

---------- lshw output ----------

*-network
     description: Ethernet interface
     product: 82579LM Gigabit Network Connection (Lewisville)
     vendor: Intel Corporation
     physical id: 19
     bus info: pci@0000:00:19.0
     logical name: eno1
     version: 05
     serial: <REMOVED>
     size: 1Gbit/s
     capacity: 1Gbit/s
     width: 32 bits
     clock: 33MHz
     capabilities: pm msi bus_master cap_list ethernet physical tp 10bt 10bt-fd 100bt 100bt-fd 1000bt-fd autonegotiation
     configuration: autonegotiation=on broadcast=yes driver=e1000e driverversion=3.2.6-k duplex=full firmware=0.13-4 latency=0 link=yes multicast=yes port=twisted pair speed=1Gbit/s
     resources: irq:30 memory:df900000-df91ffff memory:df925000-df925fff ioport:f020(size=32)