Attention is currently required from: Michael Niewöhner. Benjamin Doron has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/49140 )
Change subject: soc/intel/skylake/acpi: Add PEP table ......................................................................
Patch Set 3:
(1 comment)
Commit Message:
https://review.coreboot.org/c/coreboot/+/49140/comment/ca3ab6a3_43c3cc7a PS3, Line 11:
Awesome, then at least S0ix states work just fine.
That error message comes from the kernel SlpS0 self-test. So, let's test something. I assume `cat /sys/power/mem_sleep` has `s2idle` selected? (With s0ix_enable=1 it should. If not, we have another problem.)
Yes. (I think that's based on the FADT)
Right, there is some bit in FADT for that.
Does your system stay in sleep after executing `echo freeze >/sys/power/state`?
Is the alternative that it wakes immediately? Then yes, it stays in sleep. However, when it does wake, it has always been unlocked. I don't know if that's relevant.
Unlocked vs. locked is user-space and depends on the desktop environment. XFCE has an option for locking before going to sleep but that only works when sleep was initiated from the DE.
That makes sense; I thought it might have been a possibility.
If it wakes immediately with that error, the self-test failed. I'm surprised to see the self-test fail *without* immediate wake, though.... Maybe we have to exclude the SlpS0 state from the table LPIT, when s0ix is not supported on a board :S Could you test this, please? Insert a "return" right before the comment "System (Slp_S0) residency counter" in `src/soc/intel/common/block/acpi/lpit.c` and see if the error disappears.
Based on https://01.org/blogs/qwang59/2020/linux-s0ix-troubleshooting, PC10 without S0ix residency is a possible case. If I understand correctly, that situation would be what I see with this board: No S0ix/failed self-test, but PC10 residency/no immediate wake. I will attempt to troubleshoot that sometime by disabling devices, I suppose, following that link.
I'm also not certain what you mean. If S0ix is disabled, would LPIT factor? Besides, if S0ix does relate to hardware state, but not board-design, shouldn't LPIT always advertise both PC10 residency and system residency to warn about broken S0ix support for a board?
But regardless: `return` or `return current`? The PC10 residency counter portion does increment "current"
-
Benjamin Doron (Code Review)