[Intel-wired-lan] [PATCH iwl-net v1 1/1] e1000e: Fix S0ix residency on Meteor Lake corporate systems

[Mark Pearson]

Hi Paul

On Wed, May 29, 2024, at 7:23 AM, Paul Menzel wrote:
> Dear Vitaly,
>
>
> Thank you for your reply.
>
> Am 29.05.24 um 13:13 schrieb Lifshits, Vitaly:
>
>> On 5/28/2024 1:43 PM, Paul Menzel wrote:
>
>>> Am 28.05.24 um 12:33 schrieb Vitaly Lifshits:
>>>> From: Dima Ruinskiy <dima.ruinskiy at intel.com>
>>>>
>>>> On vPro systems,the  configuration of the I219-LM to achieve power
>>>
>>> s/,the  /, the /
>> Thank you for noticing it.
>> I will fix it in a v2.
>>
>>>> gating and S0ix residency is split between the driver and the CSME FW.
>>>> It was discovered that in some scenarios, where the network cable is
>>>> connected and then disconnected, S0ix residency is not always reached.
>>>
>>> Disconnected at any point, or just during suspend?
>> Yes, at any point.
>>
>>> Any URL to the reports?
>> Yes, https://bugs.launchpad.net/sutton/+bug/2054657
>
> This page does not exist.
>
>      $ curl -I https://bugs.launchpad.net/sutton/+bug/2054657
>      HTTP/1.1 404 Not Found
>      Date: Wed, 29 May 2024 11:20:52 GMT
>      Server: gunicorn
>      […]
>
> Please reference the (working) URL in a Link: tag in the footer.
>

Unfortunately the URL is a private bug between Lenovo and Canonical - we hit the issue as we've been working on the Linux enablement for our 2024 platforms.
I can't make the bug publicly visible I'm afraid - too much internal code name/process in there.
I don't think there will be a public report of the issue yet - because the platforms have only just been released.
If it's useful I guess I can create a kernel.org bug to track against?

>>>> This was root-caused to a subset of I219-LM register writes that are not
>>>> performed by the CSME FW. Therefore, the driver should perform these
>>>> register writes on corporate setups, regardless of the CSME FW state.
>>>
>>> Is that documented somewhere?
>> Only in an internal documentation.
>
> If you can, it’d be great if you summarized why this is triggered by 
> unplugging the network cable.
>
>>> Please add more information about the affected systems, and the test 
>>> environment (firmware versions, …).
>> It is mentioned at the beginning of the commit, Meteorlake vPro systems.
>
> Please be more specific. If there is a Meteor Lake vPro system, where 
> this can *not* be reproduced, it’d be great to know the exact system you 
> tested this with. Meteor Lake has been released a long time ago, and not 
> remembering a lot of bug reports, I assume, it’s not always reproducible.
>
> (You missed to comment on my previous comments at the end.)

I can't comment on the technical details, but I can note that it's reproduced on multiple Lenovo Thinkpad Meteorlake platforms that have ethernet
E14 G6, E16 G2, L14 G5, L16 G1, P14s G5, P16s G3 & P16v G2
I think it's common to any Meteorlake platform - but I defer to the Intel folk on the specifics.

Mark (tech lead for Lenovo PC Linux team)