Testing on various upcoming OEM systems shows commit 7b167c4cb4 ("ACPI:
PM: Only mark EC GPE for wakeup on Intel systems") was short
sighted and the symptoms were indicative of other problems. Some OEMs
do have the dedicated GPIOs for the power button but also rely upon
an interrupt to the EC SCI to let the lid work.
The original commit showed spurious activity on Lenovo systems:
* On both Lenovo T14 and P14s the keyboard wakeup doesn't work, and
sometimes the power button event doesn't work.
This was confirmed on my end at that time.
However further development in the kernel showed that the issue was
actually the IRQ for the GPIO controller was also shared with the EC SCI.
This was actually fixed by commit 2d54067fcd ("pinctrl: amd: Fix
wakeups when IRQ is shared with SCI").
The original commit also showed problems with AC adapter:
* On HP 635 G7 detaching or attaching AC during suspend will cause
the system not to wakeup
* On Asus vivobook to prevent detaching AC causing resume problems
* On Lenovo 14ARE05 to prevent detaching AC causing resume problems
* On HP ENVY x360 to prevent detaching AC causing resume problems
Detaching AC adapter causing problems appears to have been a problem
because the EC SCI went off to notify the OS of the power adapter change
but the SCI was ignored and there was no other way to wake up this system
since GPIO controller wasn't properly enabled. The wakeups were fixed by
enabling the GPIO controller in commit acd47b9f28 ("pinctrl: amd: Handle
wake-up interrupt").
I've confirmed on a variety of OEM notebooks with the following test
1) echo 1 | sudo tee /sys/power/pm_debug_messages
2) sudo systemctl suspend
3) unplug AC adapter, make sure system is still asleep
4) wake system from lid (which is provided by ACPI SCI on some of them)
5) dmesg
a) see the EC GPE dispatched, timekeeping for X seconds (matching ~time
until AC adapter plug out)
b) see timekeeping for Y seconds until woke (matching ~time from AC
adapter until lid event)
6) Look at /sys/kernel/debug/amd_pmc/s0ix_stats
"Time (in us) in S0i3" = X + Y - firmware processing time
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
d6ebb17ccc