Comment 20 for bug 2007718

> Just to give another data point, I've been running 5.19 from hwe-edge (instead of hwe) for some time but I realized this suspend issue after updating hwe-edge from 5.19.0.28.29~22.04.6 to 5.19.0.32.33~22.04.9.

I had a try with the latest 5.19 HWE kernel on a Lenovo Z13 I have on hand (with a similar APU) but can't seem to reproduce the issue reported.

Since I can't seem to reproduce if this is a correct datapoint, I think ideally this should be bisected between those two git tags to figure out why this is happening.
Ubuntu-hwe-5.19-5.19.0-28.29_22.04.1
Ubuntu-hwe-5.19-5.19.0-32.33_22.04.1

You can clone from https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/jammy to get those two tags to bisect with.

> 6.1 from linux-oem-22.04c indeed fixes this suspend/ resume issue for me.

AFAICT 5.19 was EOL already upstream at that time, so Canonical was tracking patches from 6.0.y to bring into their 5.19 kernel. I /suspect/ what happened is by doing this some dependency commits from 6.0 are missing. If you can identify which commit caused the problem it will help a lot to figure out what else is missing.

Assuming it's an amdgpu regression (which we don't know right now) the one that looks most relevant is:

89bedb408c48 drm/amdgpu: disallow gfxoff until GC IP blocks complete s2idle resume

So another option is for one of the Canonical kernel guys to try to build a 5.19 test kernel with that reverted. If that helps I'll look for the other related commits to this.