Currently, you need to override `stdenv.hostPlatform` to request a
compressed kernel on AArch64, and the kernel configuration is split
between the central structured configuration and string snippets in
platform definitions. This has consequently made the latter bitrot
terribly. Since the platform‐specific logic is now very limited after
cleaning up the detritus, we can move it into the kernel derivation
and expose the relevant configuration there for anyone who wants to
customize it further or needs to read it out.
Co-authored-by: zowoq <59103226+zowoq@users.noreply.github.com>
Locally this just prefixes activation commands with `run0 --`, so the
user's normal polkit agent (graphical or pkttyagent) handles auth.
For --target-host, run0 would need a controlling terminal, which we
deliberately do not allocate over SSH. Instead the equivalent
`systemd-run --uid=0 --pipe --wait ...` form is used. Without
--ask-elevate-password this relies on a polkit rule on the target
granting the deploying user `org.freedesktop.systemd1.manage-units`.
With --ask-elevate-password the command is wrapped in
polkit-stdin-agent, which registers a per-process polkit agent for the
child and answers the PAM conversation from stdin.
A transient unit does not inherit the SSH login environment. On NixOS
the unit's PATH is just the systemd store path, so neither `env` nor
`nix-env` would be found inside it. Elevator.wrap_remote() therefore
now takes the env mapping and the command, and returns the full remote
argv: sudo keeps the existing `/bin/sh -c 'exec /usr/bin/env -i ...'`
wrapper inside itself, while run0 puts the shell wrapper *around*
systemd-run and forwards the resolved values into the unit via
--setenv. The Arg/Args/EnvValue/PRESERVE_ENV types and the env-shell
helper move from process.py into elevate.py to avoid a circular import,
and callers are updated to import them from there.
polkit-stdin-agent is resolved on the machine doing the elevation
rather than baked in as a host-arch store path, which would be wrong
for --no-reexec, cross-arch deploys (re-exec hits ENOEXEC and falls
back), --rollback/--store-path, and Darwin deployers. Locally that is
PATH. Remotely, Elevator.for_target_config() binds the elevator to the
toplevel just placed on the target, and a small /bin/sh picker tries
<toplevel>/sw/bin/polkit-stdin-agent (target-arch, in the copied
closure) then PATH, exiting with an actionable error pointing at
system.tools.nixos-rebuild.enableRun0Elevation when neither resolves.
That option (added here) puts the agent in environment.systemPackages
and asserts security.polkit.enable.
A nixos-rebuild-target-host subtest exercises the full remote run0
path (polkit-stdin-agent + systemd-run + activation) end-to-end.
Drop the now-unused stdenv argument from package.nix while here.
Closes#507054.
For systems where the package is marked as broken, use `false` instead.
The Perl script checks for a non-zero exit status, so it will handle it
as if the command failed.
NixOS has traditionally enabled the `ext` family of file systems by
default. Originally, when switching to systemd initrd, we wanted to
transition to making this explicit so that initrds could be made
without `ext`. The problem is that anyone with `fsType = "auto";` for
an `ext` file system in initrd will fail to boot, which is not really
an acceptable regression as we switch to systemd initrd by default.
By removing `default = "auto"` from `fsType`, we rule out the vast
majority of these regressions as eval errors, since most users of
`fsType = "auto"` for ext file systems are using it because of the
default value.
In hindsight, this is probably what #225352 was really about.
Allow running shell commands after the root filesystem image has been
prepared but before the final SD image is assembled. This is useful for
adjusting the rootfs, such as resizing it to a fixed size or injecting
additional files, without having to override the entire build
derivation.
Signed-off-by: Tanel Dettenborn <tanel.dettenborn@tii.ae>
The option was missing a type declaration. Use types.lines so multiple
modules can contribute commands that get merged with newlines.
While at it, document the $img shell variable so users do not have to
read the module source to discover it.
Convert all remaining boot.postBootCommands that call
nix-store --load-db into register-nix-paths systemd services for
consistency with qemu-vm.nix and to make the ordering dependency
with nix-channel-init explicit.
All services use the same unit name (register-nix-paths) and the
same sysinit.target ordering, so nix-channel-init's
after=register-nix-paths.service dependency works uniformly across
all platforms.
Platform-specific logic (netboot password setup, sd-image partition
expansion) is kept in the service script or in boot.postBootCommands
as appropriate.
Replace boot.postBootCommands with a nix-channel-init systemd
service. This is needed because the nix store DB registration in
QEMU VMs now happens in a systemd service (register-nix-paths)
rather than in postBootCommands, so the channel init must also be
a systemd service to express the ordering dependency.
The service runs in sysinit.target, after local-fs.target and
(when present) register-nix-paths.service, but before
nix-daemon.socket. Since nix-env is invoked as root before the
daemon socket exists, it accesses the store directly.
since commit 68afbf9d6385("installer: add deprecation warning about
sd-card file move"), it has been 5 years passed, and this should be
sufficient before the final deletion.
Signed-off-by: Cryolitia PukNgae <Cryolitia@gmail.com>
Sergei Zimmerman