Intel e1000e driver bug on stable (6.9.x)

Hi,

I noticed that the NIC started to fail on a couple of notebooks [0]
[1] after upgrading to 6.9.1.

I tracked down the problem to commit 861e8086029e ("e1000e: move force SMBUS from enable ulp function to avoid PHY loss issue", 2024-03-03), included in all 6.9.x releases.

The fix is in commit bfd546a552e1 ("e1000e: move force SMBUS near the end of enable_ulp function", 2024-05-28) from mainline.

The NIC fails right after boot on both systems I tried; I mention because the description is a bit unclear about that on the fix, maybe other systems are affected differently.

Best regards.

[0] HP ZBook 17 Gen 1 (D5D93AV) [8086:153a (rev 04)]
[1] Lenovo Thinkpad P15 Gen 1 [8086:0d4c]

Quoted 8 lines · Ismael Luceno

I noticed that the NIC started to fail on a couple of notebooks [0]
[1] after upgrading to 6.9.1.

I tracked down the problem to commit 861e8086029e ("e1000e: move force SMBUS from enable ulp function to avoid PHY loss issue", 2024-03-03), included in all 6.9.x releases.

Quoted 6 lines

The fix is in commit bfd546a552e1 ("e1000e: move force SMBUS near the end of enable_ulp function", 2024-05-28) from mainline.

The NIC fails right after boot on both systems I tried; I mention because the description is a bit unclear about that on the fix, maybe other systems are affected differently.

Now queued up, thanks.

Quoted 9 lines · Greg KH

Quoted 8 lines · Ismael Luceno

I noticed that the NIC started to fail on a couple of notebooks [0]
[1] after upgrading to 6.9.1.

I tracked down the problem to commit 861e8086029e ("e1000e: move force SMBUS from enable ulp function to avoid PHY loss issue", 2024-03-03), included in all 6.9.x releases.

Quoted 8 lines

Quoted 6 lines

The fix is in commit bfd546a552e1 ("e1000e: move force SMBUS near the end of enable_ulp function", 2024-05-28) from mainline.

The NIC fails right after boot on both systems I tried; I mention because the description is a bit unclear about that on the fix, maybe other systems are affected differently.

Now queued up, thanks.

I see that they are in the latest 6.6.y and 6.9.y stable-rcs. Thing is:

bfd546a552e1 causes other regressions, which is why Hui Wang submitted a revert for that one:

https://lore.kernel.org/all/[email protected]/

Vitaly Lifshits meanwhile submitted a change that afaics is meant to fix that regression:

https://lore.kernel.org/all/[email protected]/

CCed both so they can comment.

Not sure what's the best way forward here, maybe it is "not picking up bfd546a552e1 for now and waiting a few more days till the dust settles".