nixos/config/sysfs: init module

[illdefined]

This module introduces a config option boot.kernel.sysfs, which permits setting of sysfs parameters.

The configuration option accepts a nested attribute set of sysfs path components with arbitrary values, for example:

{
  kernel.mm.transparent_hugepage = {
    enabled = "always";
    defrag = "defer";
    shmem_enabled = "within_size";
  };
}

The options will be applied through systemd path units watching the individual files or glob patterns and writing the configured value through a small service using systemd-tmpfiles once they exist.

The supplied NixOS test checks if configured parameters are actually applied.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 25.05 Release Notes (or backporting 24.11 and 25.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[illdefined]

The NixOS manual build failure appears to be unrelated to this change.

I am able to build the manual locally without any issues:
dpc1yk271wl343axkdy7q3h86jx815m4-nixos-manual-html.tar.gz

[r-vdp]

One issue here, is that a lot of sysfs paths depend on kernel modules being loaded, which may happen at a later time.
On my systems, I use systemd path units to trigger a service whenever a certain path appears (because a module got loaded, e.g. hotplugging a device), and then that service writes a value to the file in sysfs.

I wonder if we'd want to support something like that? Currently the module in this PR will only work for sysfs paths that are present at the time systemd-tmpfiles runs.

[SuperSandro2000]

types have descriptions? That's new to me.

[illdefined]

The description is usually generated automatically, but due to the nesting this would lead to an infinite recursion. I therefore override the description attribute and describe the type manually.

This description is included in the generated option documentation.

[SuperSandro2000]

One issue here, is that a lot of sysfs paths depend on kernel modules being loaded, which may happen at a later time.

sysctl has the same problem. I usually work around it by loading the kernel module already in the initrd.

[illdefined]

One issue here, is that a lot of sysfs paths depend on kernel modules being loaded, which may happen at a later time.

sysctl has the same problem. I usually work around it by loading the kernel module already in the initrd.

Instead of relying on systemd-tmpfiles, I could create individual path units for every entry.

[flokli]

Continued in #392127, closing this one.

[illdefined]

Continued in #392127, closing this one.

This PR is about sysfs options, not sysctls. I could revert / reset to the original proposal which used a single tmpfiles configuration to set options on startup.

[SuperSandro2000]

I sadly cannot add much for review here but closing the PR looked like a mistake especially based on the comment above.

[pigsinablanket]

I tested enabling zram zswap with boot.kernel.sysfs.module.zswap.parameters.enabled = 1; and it worked and was persistent through a reboot.

[ShamrockLee]

I tested enabling zram with boot.kernel.sysfs.module.zswap.parameters.enabled = 1; and it worked and was persistent through a reboot.

Did you mean "zswap"? That's a great news!

[ShamrockLee]

It also works well configuring zswap on my system.

[jopejoe1]

This looks interesting, would be nice to get this merged.

[illdefined]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 391329 --checkout commit --package nixosTests.sysfs
Commit: 1305aee196b53922abfc0959d03ac90f34c8c4f9

---

x86_64-linux

✅ 1 test built:

• nixosTests.sysfs

---

aarch64-linux

✅ 1 test built:

• nixosTests.sysfs