Linux – Where is the log for ACPI events

This is often also a BIOS setting. You can check it with:

grep ^PWRB /proc/acpi/wakeup  

and enable it with

echo "PWRB" > /proc/acpi/wakeup  

By default, Linux kernel reserves lowest 64K of system memory for BIOS and repeatedly scans that part of the memory for unexpected changes. If background scanning process notices that the memory has been unexpectedly changed (corrupted) it spams the kernel log with something along the lines

Corrupted low memory at <virtual address> (<address> phys) = <memory contents after corruption>
...
Memory corruption detected in low memory
...
Hardware name: <BIOS info>

where virtual address is the virtual address that points to the corrupted memory. This can be practically anything and unless you know that some other software is using said address you can safely ignore this. The physical address is the interesting part. As long as that address (always printed in hexadecimal) is lower than 64K (2^16 bytes) you will be fine and the message is really trying to say

The motherboard manufacturer is incompetent and the system BIOS is writing to low memory area (0-64K of physical memory) without reserving that area using ACPI memory tables. Don't worry, Linux kernel expected this bug and did not store anything important there. If your BIOS ever gets fixed, you could use the last 64K of your real RAM, too. However, you will probably need to re-compile your kernel to disable the expectation that the BIOS has this bug.

The kernel developers were trying to collect a list of known bad BIOS versions and only disable the usage of lowest 64K of RAM if BIOS was known buggy. In practice, it turns out that BIOS manufacturers are incompetent enough that it would make more sense to create a list of known good BIOS versions. That list would be much shorter. In long run, it will be much easier to just declare lowest 64K of memory as always bad and ignore the whole mess. That would cause lost 64K of RAM for all i386/amd64 compatible linux computers but would allow developers to spend their time for something more productive.

If you know that your BIOS is bad (above warning is emitted to kernel log) and just want to skip the scanning for bad BIOS and keep reserving the memory you can reserve that memory area and disable scanning. Just add following kernel flags:

memmap=64K$0 memory_corruption_check=0

The flag memmap reserves 64K of RAM at physical address 0 (start of the physical RAM) and flag memory_corruption_check disables background scanning for corruption.

If your system has unexpected RAM corruption and you suspect the BIOS, adding memmap=1024K$0 will reserve 1MB at the start of physical RAM from Linux kernel. Windows 10 is rumoured to do this by default so I would not be surprised if future BIOS versions corrupted all over the first 1MB if that corruption did not cause any issues using Windows 10.