On 15/09/2026 23:32, Viktor Donner wrote:
|Hi, I reproduced the issue again after a clean Proxmox VE installation.
Updated environment: - Proxmox VE: 9.2.20 - Kernel: 7.0.14-17-pve - NIC:
Intel I219-LM - Driver: e1000e - Firmware: 0.4-4 During the failure: -
nic0 remained UP and LOWER_UP - link remained detected at 1000 Mb/s/full
duplex - the ARP/neighbour entry for the gateway 192.168.1.10 changed to
FAILED - ping returned "Destination Host Unreachable" from the host -
rx_missed_errors reached 14867 - rx_errors, tx_errors, CRC, frame,
carrier and timeout counters remained zero - the kernel log contained no
e1000e Tx hang, watchdog, reset or register dump Bringing nic0 down and
up, running ifreload, and flushing the neighbour table restored
connectivity without rebooting. I then disabled the following offloads:
ethtool -K nic0 tso off gso off gro off After that, several speed tests
completed successfully and rx_missed_errors did not increase. This
appears reproducible with offloads enabled and is mitigated by disabling
them. Please let me know whether testing TSO, GSO and GRO individually,
or collecting any specific e1000e trace/debug data, would be useful.
Regards, Viktor|
Hi Viktor,
Sorry for only replying now. This mail arrived while I was away on
vacation, and almost got lost in my mailbox.
Would it be possible for you to enable netif_msg_hw so that
e1000e_dump() prints detailed info when a hang is detected?
I think you can do it via ethtool or module_param.
Additionally, the global system log level may need to be raised to cover
KERN_INFO.
--Dima
