intel i219-LM on Dell 5420 with bad NVM checksum doesn't work with e1000e driver

[Michał Andrzejczak]

[SRU Justification]

[Impact]

Kernel module e1000e on load performs NVM checksum validation. This may
fails when the EEPROM got corrupted, but if bypassed the NIC still works
as it supposed to. This was once resolved for Presice in LP #1070182,
and carried all the way up until dropped in Trusty for unknown reason.

This patch adds the eeprom_bad_csum_allow module parameter for the
e1000 and e1000e drivers. This allows users of affected hardware
to bypass EEPROM/NVM checksum validation.

This patch was first sent to kernel mailing list as
https://<email address hidden>/, but was
rejected for the favor of fixing the actual root cause in the EEPROM.
Then this was once accepted for Presice in LP #1070182, and had been
carried all the way up until dropped in Trusty for unknown reason.

[Fix]

Cherry pick commit fc650e376e72 ("UBUNTU: SAUCE: add
eeprom_bad_csum_allow module parameter") from linux/precise.

[Test Case]

1. Boot the patched kernel. The dmesg should have following error
message on devices with corrupted NIC EEPROM:

  e1000e 0000:00:19.0: >The NVM Checksum Is Not Valid
  e1000e: probe of 0000:00:19.0 failed with error -5

2. Unload the driver and reload with eeprom_bad_csum_allow=1:

  $ sudo modprobe -r e1000e
  $ sudo modprobe e1000e eeprom_bad_csum_allow=1

3. Check if netdev is up:

  $ ip link

[Where problems could occur]

From the discussion thread, skipping checksum validation works only as a
quick work-around, and the configurations stored in EEPROM still matter
in runtime. The adapter may still fail to work as expected.

[Other Info]

To pick up this patch for all living kernels. Nominate for Trusty,
Xenial, Bionic, Focal, Jammy, Noble, Oracular, Plucky and Unstable.

============ original bug report ============

I work with the company that has obtained 300 used Dell Latitude 5420 laptops. They ware running Windows previously without problems. However under Ubuntu (22.04; 24.10) there is a problem with Intel i219-LM (13) NIC. Kernel module e1000e on load performs NVM checksum validation and fails. Tested sample suggests about 70% of laptops are affected.

I modified e1000e module to bypass checksum and after that NIC works as it supposed to. I tried to correct NVM using ethtool and intel provided tool but it seems Dell locked writing to NVM on these laptops.

I searched internet and this is not the first time this problem has occurred. There even used to be a patch adding module parameter to force checksum bypass (https://patchwork<email address hidden>/#470600). I wonder why ubuntu kernel is no longer shipped with this patch? Can you add this patch back or similar workaround?

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[You-Sheng Yang (vicamo)]

I'm checking why this patch were dropped during all these years.

[Michał Andrzejczak (michal-andrzejczak)]

Thank you. I hope we can find some permanent workaround for this problem.

[Jürg Häfliger (juergh)]

It seems to have been dropped in Trusty 3.12 but it's unclear why that happened.

[You-Sheng Yang (vicamo)]

SRU: https://lists.ubuntu.com/archives/kernel-team/2025-March/158335.html (trusty, xenial, bionic, focal, jammy, noble, oracular, plucky, unstable)