Unepected power off during boot; incorrect power readings; unexpected undocking behavior

[Lewis Rosenthal]

Hardware:

IBM ThinkPad T43
2.26GHz Pentium M 780 (single core)
2GB PC-4200 RAM
Intel 82801FBM ICH6-M southbridge
ATI Radeon X300

Software:

eCS 2.1 GA
14.106_SMP retail kernel
NO APM.SYS
NO VAPM.SYS
NO APMDAEMN.EXE
NO GSVDAEMN.EXE

Symptom:

Upon installing ACPI.PSD, while the system is still booted without any PSD loaded and while still running APM.SYS (10.163), system shuts down/powers off or goes down for reboot normally. The subsequent cold or warm boot loads the ACPI PSD, and the system boots normally. Upon then shutting down/powering off or rebooting, the subsequent boot fails, either before base device drivers begin loading (the kernel build string is still visible on-screen) or immediately after ACPIDAEMON starts. The failure is an abrupt power off. Subsequent reboots result in similar failures. Rebooting with no PSD and no APM.SYS are successful; however, until APM.SYS has been added back to CONFIG.SYS and the system successfully booted, ACPI will continue to display this behavior. Once successfully booted with APM.SYS active (ny editing CONFIG.SYS), the next reboot with ACPI and ACPIDAEMON will succeed, although rebooting or shutting down from that point will result in the original (unexpected power off) behavior.

ACPID.CFG:

Initially, there was no ACPID.CFG on this system. Creating one based on the sample file made no difference, and neither did specifically stating PowerButton = None (the power off event can happen long before ACPID.CFG is even read - before the IFS even loads to read any files in %ETC%).

Possibly-related symptoms:

When fully booted with ACPIDAEMON running, neither the system power object nor the Beer Battery widget are able to determine battery charge or correctly determine source. In the case of the widget, battery charge is consistently reported as 101% and on AC power (even when the system is running on battery), and the power object reports "unknown" both for power source and power state.

Finally, under normal circumstances (APM.SYS, APMDAEMN.EXE, and GSVDAEMN.EXE in CONFIG.SYS), when the system is docked (​ThinkPad Dock II), the dock is locked. It is not possible to undock the machine while hot (under OS/2). Pressing the dock's release button produces a dull click, but does not release the button to its "up" position to allow it to be pressed to eject the machine - until power off, when the button will immediately snap up. Under ACPI, the button operates freely; the system does not behave as though docked. This might have some bearing on the battery state being unknown, as the system should detect that it is getting power from the docking connector, and if power is coming through the docking connector, then the system should properly "lock" to the dock.

Debug logs attached.

[Lewis Rosenthal]

ACPI debug logs

[lewisr]

With ACPI 3.23.13 and ArcaOS 5.0.4, all of this behaves as expected (no unexpected shutdown events, laptop is properly locked on the dock, battery widget produces expected results, etc.)