[Bug 2083716] Re: s2idle regression with sk hynix BC901 nvme in noble

That did the job, thank you!

** Attachment added: "20241029-20-56-s0ix-output.log"
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2083716/+attachment/5832852/+files/20241029-20-56-s0ix-output.log

--
You received this bug notification because you are subscribed to linux
in Ubuntu.
Matching subscriptions: Bgg, Bmail, Nb
https://bugs.launchpad.net/bugs/2083716

Title:
s2idle regression with sk hynix BC901 nvme in noble

Status in linux package in Ubuntu:
New

Bug description:
With lunar's and mantic's kernels, my laptop (Dell 7630) is able to
properly s2idle, with <0.5% battery consumption per hour. On Noble's
kernel, I'm unable to reach s2idle.

Notably, the mainline kernel also doesn't do s2idle properly for
equivalent versions (6.2.x and 6.5.x). There's something special
about lunar's and mantic's kernels that makes s2idle work properly on
my laptop, and noble has lost this. For now I'm running mantic's
kernel and noble userspace, which seems to work fine.

In mantic, intel's S0ixSelftestTool says, "Congratulations! Your
system achieved the deepest S0ix substate! "

In noble and in mainline kernels, I get the following:


[...]
-[10000:e0]---06.0-[e1]----00.0 SK hynix BC901 NVMe Solid State Drive (DRAM-less)

Pcieport is not in D3cold:
10000:e0:06.0


Full output is in the attached S0ixSelftestTool log file.

I know there've been issues with SK hynix drives and s2idle in the
past. As best I can guess, maybe there's some kind of fix in the
lunar and mantic kernels that didn't make it into noble or mainline.

For now the only workaround I've found is to run 6.5.0-44-generic from
mantic. I haven't been able to find any settings or tweaks that get
s2idle working properly in noble's kernel. Specifically, I tried
`nvme_core.force_apst=1` on the kernel command line but that didn't
help.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2083716/+subscriptions