[solved] ACPI override

archdub · 2025-08-05 14:34:29

At the bottom of section 3.3.4 in this page
https://wiki.archlinux.org/title/Mkinit … mmon_hooks

there is a mention of acpi_override. I read it and would like to try it to see if it could be a workaround for acpi bios issues in my PC. However the two folders referred to in the wiki, namely /usr/initcpio/acpi_override/ and /etc/initcpio/acpi_override/, don't exist. As a result when I run mkinitcpio, I get an error message saying that the files could not be found. I could not find any package with such files.

Is the wiki wrong, or is it missing information about where to find aml files?

Last edited by archdub (2025-08-06 06:12:36)

seth · 2025-08-05 15:28:42

The idea would be that you create those directories and more importantly place the .aml's relevant to address your problem there.
If this is just about ACPI error noise in the journal: stop.
As long as there's no misbehavior that can be linked to the ACPI errors they mean little more than the board vendor not properly implementing the ACPI spec but poking around until windows started - what is common.

archdub · 2025-08-06 06:12:56

Thanks Seth. I won't spend more time on ACPI override, it looks like too much work for uncertain benefit.

There are plenty of ACPI error messages in the dmesg and journal, but i found a workaround that makes them disappear, a kernel parameter. The wider context is that i am using the lts61 kernel because that is the last lts kernel with which waking up from suspend works reliably. Although i suspect it does not work with later kernels due to an interaction between bios and the kernel, i have no proof.

My question was if there is an error in the wiki and where to find .aml files. The answer is that the wiki is not wrong and instead i failed to read the information on one of the provided links (DSDT) which would have answered my question. Arguably the writing of that paragraph could be improved, but that is a minor thing so i will mark this thread as solved.

seth · 2025-08-06 07:16:47

nb. that noaer will only suppress those messages, not really change anything.
It might also suppress more relevant messages about actual errors, so keep that in mind and remove it if you're facing some problems later on.

archdub · 2025-08-06 09:44:35

With “libata.noacpi=1” on the kernel command line, all the red ACPI error messages disappear from dmesg/journal in my computer (MSI motherboard). However that did not solve the issue that with kernel 6.6 several sleep/suspend cycles work well in a row and then one does not and i have to reboot the PC.

Kernel 6.1 is the last that works reasonably reliably. The status quo is ok, as 6.1 still has about 2.5 years before going EOL. Before then i will have to buy a new computer.

seth · 2025-08-06 12:10:50

issue that with kernel 6.6 several sleep/suspend cycles work well in a row and then one does not

Do you have an open thread on that?

archdub · 2025-08-07 10:53:43

seth wrote:

Do you have an open thread on that?

I have been aware of this issue for years, but never opened a thread on that. That was a conscious decision. The main reasons are that: a) I was and still am afraid it would take a lot of my time with uncertain, even unlikely, positive results (i might need to do kernel bisection, for example, and that is time consuming); b) because this issue does not happen on every sleep/wakeup cycle, it might happen at inconvenient times, like during work hours or when i need my pc to do something, so it would be inconvenient to investigate; c) there is a workaround that is to compile myself linux-lts61.

Most likely this is a kernel issue in that it is not compatible with my bios. Long ago i found some discussion in the kernel mailing list that might be related to my issue: a fix was put in a kernel, and some time later that fix was changed or removed. I remember that is what happened with a version post 6.1 on my pc (i forgot the precise version number): it stopped working, then it worked for a while, and stopped working again. This is circumstancial evidence at its best, but it does point to a kernel issue.

seth · 2025-08-07 13:04:21

Open one if you feel like, a journal covering (some) working sleep cycles and then ultimately a failing one would be useful.
See whether you can avoid pushing the power button for a reboot but attempt https://wiki.archlinux.org/title/Keyboa … el_(SysRq) (nb. that you'll have to explicitly enable that before the suspend/failure)

archdub · 2025-08-08 06:59:47

When this issue happens, there is not a single line in the journal, ssh does not work, and the usb keyboard does not work because it seems it gets no power (no LED on) even though the computer is powered (fans heard, LEDs on). It looks like the kernel hangs at an early stage after it wakes up from sleep.

This old computer has a RS232 port with a DB9 connector, I have RS232 cables with DB9 connectors, so it would be easy to set up a RS232 connection. However debugging a kernel hang requires a level of kernel expertise that I don't have. Even if I could figure out why exactly the kernel hangs, coming up with a fix for a later kernel version would not be trivial, considering all the churn between kernel versions, let alone between versions that are years apart.

seth · 2025-08-08 07:13:21

https://wiki.archlinux.org/title/Kdump but it's actually more likely that the kernel doesn't resume at all.
You'd have to look for indications on working resumes ("despite these errors") or ahead of the failing suspend (notably any acpi events, charger, fwupd, …)

Arch Linux

#1 2025-08-05 14:34:29

[solved] ACPI override

#2 2025-08-05 15:28:42

Re: [solved] ACPI override

#3 2025-08-06 06:12:56

Re: [solved] ACPI override

#4 2025-08-06 07:16:47

Re: [solved] ACPI override

#5 2025-08-06 09:44:35

Re: [solved] ACPI override

#6 2025-08-06 12:10:50

Re: [solved] ACPI override

#7 2025-08-07 10:53:43

Re: [solved] ACPI override

#8 2025-08-07 13:04:21

Re: [solved] ACPI override

#9 2025-08-08 06:59:47

Re: [solved] ACPI override

#10 2025-08-08 07:13:21

Re: [solved] ACPI override

Board footer