Hi, first time reporting a kernel bug so let me know if I need to attach anything else. Recently my LG Gram (16T90R-K.ADB9U1, i7-1360p) started having problems with Management Engine, possibly because Windows or LG Update did something while I was logged into Windows 11. The first symptom was SOF firmware not loading, but I "fixed" that by resetting BIOS to default settings and am unable to reproduce it now. Windows 11 also reacted by logging me out of my Microsoft Account. Now I'm noticing that Linux can't enter s0ix/s2idle with this problem, and it's now draining ~4%/hour or ~3W while sleeping. I confirmed it with S0ixSelftestTool on many different kernels ranging from 6.7-11 to the current LTS 6.6 and the current mainline 6.12rc4, with a few firmware and microcode versions, and also a few live distros. The first `s0ix-selftest-tool.sh -s` run terminates on "LTR ignore 6", and the second run complains about CSME_PG_STS. I also ran powercfg /SleepStudy on Windows 11. It's also complaining about NoHwDrips.C8.LTR_ME (which is LTR 6), but it seems to be still correctly entering sleep and consuming a more reasonable 660mW. Unfortunately my LG Gram is cursed with an AMI AptIO BIOS so there are no options for me to fiddle around with. For now I'm able to work around this by going into S3 deep sleep instead (which has its own quirks with iwlwifi), but I figured I should file a bug, especially since Windows seems to know how to work around it. The closest bug I've found is https://bugs.launchpad.net/intel/+bug/1744638, but it doesn't quite fit the bill. Attachments below: * s0ix-selftest-tool.sh -s # two runs * journalctl -b | grep -Pi 'kernel|intel|acpi|bios|(lg.laptop)|PM:|csme|mei|s0ix|s2idle' # with intel_pmc_core.warn_on_s0ix_failures=1 * dmesg # with intel_pmc_core.warn_on_s0ix_failures=1 * acpidump before s0ix tests gzipped * acpidump after s0ix tests gzipped * lspci -vvvv * lshw * intel_csme_version_detection_tool * powercfg.exe /SleepStudy
Created attachment 307043 [details] First `s0ix-selftest-tool.sh -s` run First `s0ix-selftest-tool.sh -s` run Note: I've noticed that https://github.com/intel/S0ixSelftestTool/blob/main/s0ix-selftest-tool.sh#L1086 should be assigning to pc10, which I fixed on my local copy.
Created attachment 307044 [details] Second `s0ix-selftest-tool.sh -s` run
Created attachment 307045 [details] journalctl -b | grep -Pi 'kernel|intel|acpi|bios|(lg.laptop)|PM:|csme|mei|s0ix|s2idle' # with intel_pmc_core.warn_on_s0ix_failures=1
Created attachment 307046 [details] dmesg # with intel_pmc_core.warn_on_s0ix_failures=1
Created attachment 307047 [details] acpidump # before s0ix-selftest-tool.sh -s runs
Created attachment 307048 [details] acpidump # after s0ix-selftest-tool.sh -s runs
Created attachment 307049 [details] lspci -vvvv
Created attachment 307050 [details] lshw
Created attachment 307051 [details] intel_csme_version_detection_tool
Created attachment 307052 [details] powercfg.exe /SleepStudy
Sorry, I filed this under ACPI/Power-Sleep-Wake but now I'm seeing that other s0ix bugs are under Power Management/Hibernate-Suspend so I've moved it.
Oddly enough, the demons were somehow exorcised from my firmware and my laptop no longer experiences this CSME/MEI issue and can get a good night's sleep under s0ix now.