The ollama module in its default configuration relies on systemd's
`DynamicUser=` feature for user allocation. In #305076 that allocation
was made conditional and tied to the `sandboxing` option, that was
intended to fix access to model directories outside the allocated state
directory.
However, by disabling sandboxing ollama would inadvertently run as root,
given that `User=` and `Group=` are not required to be set.
The correct way to grant access to other paths is to allocate static
user and group, and grant permissions to the destination path to that
allocation.
We therefore replace the sandboxing option user and group options, that
default to `null`, which means they default to `DynamicUser=`, but can
be replaced with a statically allocated user/group, and thereby a stable
uid/gid.
Fixes: 552eb759 ("nixos/ollama: add options to bypass sandboxing")
Quoting from upstream's documentation [1]:
> Basically everyone recommends not doing this. Please use [a webserver]
> to handle media file serving.
Given that this commit broke the module for unrelated reasons, I've
decided to just revert it and let downstream users make the choice of
easy vs. secure.
[1]: https://docs.tandoor.dev/system/configuration/#gunicorn-media
This reverts commit e8c56de827.
Regardless of mutable or immutable users, systemd-sysupdate never
updates existing user records and thus will for example never change
passwords for you.
It only support initial passwords and now actively asserts agains other
paswords.
On Linux we cannot feasbibly generate users statically because we need
to take care to not change or re-use UIDs over the lifetime of a machine
(i.e. over multiple generations). This means we need the context of the
running machine.
Thus, stop creating users statically and instead generate them at
runtime irrespective of mutableUsers.
When /etc is immutable, the password files (e.g. /etc/passwd etc.) are
created in a separate directory (/var/lib/nixos/etc). /etc will be
pre-populated with symlinks to this separate directory.
Immutable users are now implemented by bind-mounting the password files
read-only onto themselves and only briefly re-mounting them writable to
re-execute sysusers. The biggest limitation of this design is that you
now need to manually unmount this bind mount to change passwords because
sysusers cannot change passwords for you. This shouldn't be too much of
an issue because system users should only rarely need to change their
passwords.
systemd-sysusers cannot create normal users (i.e. with a UID > 1000).
Thus we stop trying an explitily only use systemd-sysusers when there
are no normal users on the system (e.g. appliances).
We want to get rid of specialFileSystems / earlyMountScript eventually and
there is no need to run this before systemd anymore now that
the wrappers themselves are set up in a systemd unit since https://github.com/NixOS/nixpkgs/pull/263203
Also this is needed to make soft-reboot work. We want to make sure
that we remount /run/wrappers with the nosuid bit removed on soft-reboot
but because @earlyMountScript@ happens in initrd, this wouldn't happen
