1 unstable release

0.1.1 Mar 25, 2023

#1042 in Hardware support

Unlicense

11KB
86 lines

post-clock

Once a system has booted, use the display for POST diagnostic codes as a clock.

Background on POST diagnostic codes

During the boot process of x86 and x86-64 systems, the firmware typically outputs diagnostic information to I/O port 0x80. These are called power-on self test (POST) codes, and some motherboards are able to display them. There are also add-on cards that display POST codes.

After the firmware hands off execution to the OS kernel, no more POST codes are output. At that point some motherboards can be configured to use the display for other purposes, like on-board temperature sensor readouts.

However, it's also possible for the kernel and userspace to write arbitrary bytes to I/O port 0x80 and, if the POST code display hasn't been switched to a different input, have it display them.

One small complication is that some programs, including the Linux kernel, may by default use writes to port 0x80 as a short delay timer.^1 But this behavior can be disabled: for example, Linux can be configured to use port 0xED instead, which is also generally safe.

System requirements and preparation

post-clock requires a recent Linux kernel and a x86 or x86-64 system.

The kernel should be configured to use a port different than 0x80 for I/O delays, or no port at all. This can be done at build time with CONFIG_IO_DELAY_0XED and similar options or, more commonly, with the io_delay boot parameter:

io_delay=       [X86] I/O delay method
        0x80
                Standard port 0x80 based delay
        0xed
                Alternate port 0xed based delay (needed on some systems)
        udelay
                Simple two microseconds delay
        none
                No delay

Other programs using port 0x80 for short delays should also be configured to use some other port or delay mechanism.

The motherboard should be configured to not output other information to that display once the system has booted.

And, finally, the CAP_SYS_RAWIO capability is required for the ioperm(2) system call (post-clock will drop it and any other capabilities before entering its main loop). Alternatively, post-clock can be executed as root.

Linux packages

Distribution Package name
ArchLinux post-clockAUR

Manual installation

Because post-clock requires the CAP_SYS_RAWIO capability to run, it generally shouldn't be installed in a user-writable location. This unfortunately also means that cargo install post-clock isn't a good fit.[^3]

Instead, the recommended way to manually install post-clock is to clone the repository at the latest release tag, build with cargo, and copy the resulting executable to the desired location:

git clone https://github.com/jonasmalacofilho/post-clock --branch v0.1.1
cd post-clock
cargo build --release
sudo install -Dm0755 -t /usr/local/bin/ target/release/post-clock

Running as a service

A systemd system service file is provided in post-clock.service. After installing it to a suitable location,[^4] reload all unit files:

sudo install -Dm0644 -t /etc/systemd/system/ post-clock.service
sudo systemctl daemon-reload

Then, enable and start the service:

sudo systemctl enable --now post-clock.service

[^3]: While it's possible to pass --root /usr/local/bin, that would require running cargo as root, and it would build post-clock as root, which isn't advised.

[^4]: See systemd.unit(5).

Dependencies

~1MB
~22K SLoC