Closes#216989
First of all, a bit of context: in PostgreSQL, newly created users don't
have the CREATE privilege on the public schema of a database even with
`ALL PRIVILEGES` granted via `ensurePermissions` which is how most of
the DB users are currently set up "declaratively"[1]. This means e.g. a
freshly deployed Nextcloud service will break early because Nextcloud
itself cannot CREATE any tables in the public schema anymore.
The other issue here is that `ensurePermissions` is a mere hack. It's
effectively a mixture of SQL code (e.g. `DATABASE foo` is relying on how
a value is substituted in a query. You'd have to parse a subset of SQL
to actually know which object are permissions granted to for a user).
After analyzing the existing modules I realized that in every case with
a single exception[2] the UNIX system user is equal to the db user is
equal to the db name and I don't see a compelling reason why people
would change that in 99% of the cases. In fact, some modules would even
break if you'd change that because the declarations of the system user &
the db user are mixed up[3].
So I decided to go with something new which restricts the ways to use
`ensure*` options rather than expanding those[4]. Effectively this means
that
* The DB user _must_ be equal to the DB name.
* Permissions are granted via `ensureDBOwnerhip` for an attribute-set in
`ensureUsers`. That way, the user is actually the owner and can
perform `CREATE`.
* For such a postgres user, a database must be declared in
`ensureDatabases`.
For anything else, a custom state management should be implemented. This
can either be `initialScript`, doing it manual, outside of the module or
by implementing proper state management for postgresql[5], but the
current state of `ensure*` isn't even declarative, but a convergent tool
which is what Nix actually claims to _not_ do.
Regarding existing setups: there are effectively two options:
* Leave everything as-is (assuming that system user == db user == db
name): then the DB user will automatically become the DB owner and
everything else stays the same.
* Drop the `createDatabase = true;` declarations: nothing will change
because a removal of `ensure*` statements is ignored, so it doesn't
matter at all whether this option is kept after the first deploy (and
later on you'd usually restore from backups anyways).
The DB user isn't the owner of the DB then, but for an existing setup
this is irrelevant because CREATE on the public schema isn't revoked
from existing users (only not granted for new users).
[1] not really declarative though because removals of these statements
are simply ignored for instance: https://github.com/NixOS/nixpkgs/issues/206467
[2] `services.invidious`: I removed the `ensure*` part temporarily
because it IMHO falls into the category "manage the state on your
own" (see the commit message). See also
https://github.com/NixOS/nixpkgs/pull/265857
[3] e.g. roundcube had `"DATABASE ${cfg.database.username}" = "ALL PRIVILEGES";`
[4] As opposed to other changes that are considered a potential fix, but
also add more things like collation for DBs or passwords that are
_never_ touched again when changing those.
[5] As suggested in e.g. https://github.com/NixOS/nixpkgs/issues/206467
Related to #262907 (Django3 removal from nixpkgs).
This package already required an unreasonable amount of maintenance
regularly for a such small leaf-package. It has a few highly outdated
dependencies (e.g. flask 1, jinja2 2.11, sqlalchemy 1.3).
After at least each Python package-set update one had to fix up a lot of
dependencies to fix the package itself, so it was only useful on stable
branches. And having so much outdated software in a security-sensitive
piece of software seems questionable.
Finally, globin and I won't be available for maintaining this now that
Mayflower is migrating to another solution (and we'll do that as well)
and I'd expect this to bitrot extremely quick if we both bail out.
Add jitterentropy-rngd, a tool similar to rng-tools.
While not necessarily needed, it is useful for those
who want to strengthen their kernel entropy input pool
by periodic insertion of an independent source.
The entropy source is a NIST SP800-90B compliant
non-physical true RNG source on most systems.
See the jitterentropy documentation for details
(http://chronox.de/jent/doc/CPU-Jitter-NPTRNG.pdf).
Signed-off-by: Markus Theil <theil.markus@gmail.com>
PR#256295 reintroduced ruleFile option, but set the default as a path
literal, which was a "string path" previously. This breaks evaluation
for being impure:
error: access to absolute path '/var/lib/usbguard/rules.conf' is forbidden in pure eval mode (use '--impure' to override)
https://github.com/kanidm/kanidm/releases/tag/v1.1.0-beta.13
The kanidmd process now creates a unix socket, over which admin tasks
can be done, without having to shut kanidm down first.
The kanidm_unixd process now wants access to /etc/shadow and /etc/group,
so it can rule out collisions with the host system.
The usbguard package includes the dbus daemon, but the NixOS config
option does not provide a service file or its necessary polkit rules.
Enabling the dbus daemon allows use of Gnome's USBGuard support.
Bind mount the base dirs of the tls key and chain into the service.
Make sure to bind every directory just once. The test failed on ofborg
when /nix/store and the certificate path in /nix/store/<some path> were
bound.
Issue: after installing, running `yubikey-agent -setup` produces the
following error:
```
nixOS: Failed to connect to the YubiKey: connecting to pscs: the Smart
card resource manager is not running
```
More on this issue: https://github.com/FiloSottile/yubikey-agent/issues/137
The previous changes for the 3.8 update are ready, but staging got
merged into master, so there are a few more challenges to tackle:
* Use python 3.10 now since it's actually supported and less effort to
build (3.9 isn't recursed into anymore).
* sphinx doesn't build with these overrides, so patch it out entirely
(i.e. drop `sphinxHook` where it's causing problems).
* backport a few jinja2 fixes for python 3.10 that were fixed in later
versions, but break because this env is stuck to 2.11.
- Adds an update script to fetch the compatible web vault version
- Removes `vaultwarden-vault` from top-level to prevent independent
updates through e.g. r-ryantm. Istead the vault is now accessible
at `vaultwarden.webvault`.
- The name webvault was chosen because it is the title of the projects
README and it makes it clearer, that this is the web UI.
Changes sgx-psw to append `aesm` to `LD_LIBRARY_PATH`:
- Append instead of prepend to allow for overriding in service config
- As we already add a wrapper to add `aesm` to `LD_LIBRARY_PATH` it is
not necessary to also set in `LD_LIBRARY_PATH` of the systemd service.
Co-authored-by: Vincent Haupert <mail@vincent-haupert.de>
Add an option for physlock's -m flag, which mutes kernel messages on the
console. This ensures that the password prompt is the only thing on the
screen and isn't lost in a flood of kernel messages.
`privacyidea-token-janitor`[1] is a tool which helps to automate
maintenance of tokens. This is helpful to identify e.g. orphaned tokens,
i.e. tokens of users that were removed or tokens that were unused for a
longer period of time and apply actions to them (e.g. `disable` or
`delete`).
This patch adds two new things:
* A wrapper for `privacyidea-token-janitor` to make sure it's executable
from CLI. To achieve this, it does a `sudo(8)` into the
`privacyidea`-user and sets up the environment to make sure the
configuration file can be found. With that, administrators can
directly invoke it from the CLI without additional steps.
* An optional service is added which performs automatic cleanups of
orphaned and/or unassigned tokens. Yes, the tool can do way more
stuff, but I figured it's reasonable to have an automatic way to clean
up tokens of users who were removed from the PI instance. Additional
automation steps should probably be implemented in additional
services (and are perhaps too custom to add them to this module).
[1] https://privacyidea.readthedocs.io/en/v3.7/workflows_and_tools/tools/index.html
In order to be able to use the unixd service with the `verify_ca` and
`verify_hostnames` set to `true` it needs to be able to read the
certificate store. This change bind mounts the cacert paths for the
unixd service.
most of these are hidden because they're either part of a submodule that
doesn't have its type rendered (eg because the submodule type is used in
an either type) or because they are explicitly hidden. some of them are
merely hidden from nix-doc-munge by how their option is put together.
conversions were done using https://github.com/pennae/nix-doc-munge
using (probably) rev f34e145 running
nix-doc-munge nixos/**/*.nix
nix-doc-munge --import nixos/**/*.nix
the tool ensures that only changes that could affect the generated
manual *but don't* are committed, other changes require manual review
and are discarded.
there are sufficiently few variable list around, and they are
sufficiently simple, that it doesn't seem helpful to add another
markdown extension for them. rendering differences are small, except in
the tor module: admonitions inside other blocks cannot be made to work
well with mistune (and likely most other markdown processors), so those
had to be shuffled a bit. we also lose paragraph breaks in the list
items due to how we have to render from markdown to docbook, but once we
remove docbook from the pipeline those paragraph breaks will be restored.
the way these are written they introduce lots of whitespace in each
line, which will cause those lines to render as code when converted to
markdown. override the whole description instead.
no change in rendered output. the html manual could render <screen>
blocks differently, but so far it hasn't (and if we need to make a
distinction we can use a special info string).
using regular strings works well for docbook because docbook is not as
whitespace-sensitive as markdown. markdown would render all of these as
code blocks when given the chance.
this renders the same in the manpage and a little more clearly in the
html manual. in the manpage there continues to be no distinction from
regular text, the html manual gets code-type markup (which was probably
the intention for most of these uses anyway).
now nix-doc-munge will not introduce whitespace changes when it replaces
manpage references with the MD equivalent.
no change to the manpage, changes to the HTML manual are whitespace only.
make (almost) all links appear on only a single line, with no
unnecessary whitespace, using double quotes for attributes. this lets us
automatically convert them to markdown easily.
the few remaining links are extremely long link in a gnome module, we'll
come back to those at a later date.
the conversion procedure is simple:
- find all things that look like options, ie calls to either `mkOption`
or `lib.mkOption` that take an attrset. remember the attrset as the
option
- for all options, find a `description` attribute who's value is not a
call to `mdDoc` or `lib.mdDoc`
- textually convert the entire value of the attribute to MD with a few
simple regexes (the set from mdize-module.sh)
- if the change produced a change in the manual output, discard
- if the change kept the manual unchanged, add some text to the
description to make sure we've actually found an option. if the
manual changes this time, keep the converted description
this procedure converts 80% of nixos options to markdown. around 2000
options remain to be inspected, but most of those fail the "does not
change the manual output check": currently the MD conversion process
does not faithfully convert docbook tags like <code> and <package>, so
any option using such tags will not be converted at all.
Instead of hard-coding a single `configFile` for
`privacyidea-ldap-proxy.service` which is pretty unmergable with other
declarations it now uses a RFC42-like approach. Also to make sure that
secrets can be handled properly without ending up in the Nix store, it's
possible to inject secrets via envsubst
{
services.privacyidea.ldap-proxy = {
enable = true;
environmentFile = "/run/secrets/ldap-pw";
settings = {
privacyidea.instance = "privacyidea.example.org";
service-account = {
dn = "uid=readonly,ou=serviceaccounts,dc=example,dc=org";
password = "$LDAP_PW";
};
};
};
}
and the following secret file (at `/run/secrets`):
LDAP_PW=<super-secret ldap pw>
For backwards-compat the old `configFile`-option is kept, but it throws
a deprecation warning and is mutually exclusive with the
`settings`-attrset. Also, it doesn't support secrets injection with
`envsubst` & `environmentFile`.
* Update to the latest upstream version of pass-secret-service that includes
systemd service files.
* Add patch to fix use of a function that has been removed from the Python
Cryptography library in NixOS 22.05
* Install systemd service files in the Nix package.
* Add NixOS test to ensure the D-Bus API activates the service unit.
* Add myself as a maintainer to the package and NixOS test.
* Use checkTarget instead of equivalent custom checkPhase.
* nixos/vault: add option to start in dev mode.
This is not only useful for nixos tests i.e. when testing vault agent
setups but also when playing around with vault in local setups. In our
tests we can now make use of this option to test more vault features.
i.e. adding this feature has uncovered the need for a `StateDirectory`.
* Update nixos/modules/services/security/vault.nix
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
Co-authored-by: Jonas Chevalier <zimbatm@zimbatm.com>
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
If `services.tor.client.enable` is set to false (the default), the `SOCKSPort` option is not added to the torrc file but since Tor defaults to listening on port 9050 when the option is not specified, the tor client is not actually disabled. To fix this, simply set `SOCKSPort` to 0, which disables the client.
Use `mkForce` to prevent potentially two different `SOCKSPort` options in the torrc file, with one of them being 0 as this would cause Tor to fail to start. When `services.tor.client.enable` is set to false, this should always be disabled.
When `services.resolved.enable` is set to true, the file /etc/resolv.conf becomes a symlink to /etc/static/resolv.conf, which is a symlink to /run/systemd/resolve/stub-resolv.conf. Without this commit, tor does not have access to this file thanks to systemd confinement. This results in the following warning when tor starts:
```
[warn] Unable to stat resolver configuration in '/etc/resolv.conf': No such file or directory
[warn] Could not read your DNS config from '/etc/resolv.conf' - please investigate your DNS configuration. This is possibly a problem. Meanwhile, falling back to local DNS at 127.0.0.1.
```
To fix this, simply allow read-only access to the file when resolved is in use.
According to https://github.com/NixOS/nixpkgs/pull/161818#discussion_r824820462, the symlink may also point to /run/systemd/resolve/resolv.conf, so allow that as well.
Tor waits ShutdownWaitLength seconds (30s by default) before actually shutting down. Since the systemd timeout is also set to 30 seconds, this results in a race condition that ends up killing Tor most of the time on my machine during shutdown.
To fix this, add the ShutdownWaitLength setting and tell systemd to wait 30 seconds more than that.
Arch Linux also has `TimeoutSec` set to 60 seconds: 6df716fe19/trunk/tor.service.
The Intel SGX DCAP driver makes the SGX application enclave device and
the SGX provisioning enclave available below the path `/dev/sgx/`. Since
Linux 5.11, a derivation of the DCAP driver is part of the kernel and
available through the X86_SGX config option; NixOS enables this option
by default.
In contrast to the out-of-tree DCAP driver, the in-tree SGX driver uses
a flat hierarchy for the SGX devices resulting in the paths
`/dev/sgx_enclave` for the application enclave device and
`/dev/sgx_provison` for the provisioning enclave device.
As of this commit, even the latest version of the Intel SGX PSW
libraries still tries to open the (legacy) DCAP paths only. This means
that SGX software currently cannot find the required SGX devices even if
the system actually supports SGX through the in-tree driver. Intel wants
to change this behavior in an upcoming release of intel/linux-sgx.
Having said that, SGX software assuming the SGX devices below
`/dev/sgx/` will prevail. Therefore, this commit introduces the NixOS
configuration option `hardware.cpu.intel.sgx.enableDcapCompat` which
creates the necessary symlinks to support existing SGX software. The
option defaults to true as it is currently the only way to support SGX
software. Also, enabling the SGX AESM service enables the option.
The permissions of the devices `/dev/sgx_enclave` and
`/dev/sgx_provison` remain the same, i.e., are not affected regardless
of having the new option enabled or not.
most modules can be evaluated for their documentation in a very
restricted environment that doesn't include all of nixpkgs. this
evaluation can then be cached and reused for subsequent builds, merging
only documentation that has changed into the cached set. since nixos
ships with a large number of modules of which only a few are used in any
given config this can save evaluation a huge percentage of nixos
options available in any given config.
in tests of this caching, despite having to copy most of nixos/, saves
about 80% of the time needed to build the system manual, or about two
second on the machine used for testing. build time for a full system
config shrank from 9.4s to 7.4s, while turning documentation off
entirely shortened the build to 7.1s.
When accessing the Audit log, I get an HTTP 502 when the frontend
requests `/audit` and I get the following error in my `nginx`-log:
Dec 20 22:12:48 ldap nginx[336]: 2021/12/20 22:12:48 [error] 336#336: *8421 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 10.237.0.1, server: _, request: "GET /audit/?action=**&action_detail=**&administrator=**&client=**&date=**&duration=**&info=**&page=1&page_size=10&policies=**&privacyidea_server=**&realm=**&resolver=**&serial=**&sortorder=desc&startdate=**&success=**&tokentype=**&user=** HTTP/1.1", upstream: "uwsgi://unix:/run/privacyidea/socket:", host: "ldap.ist.nicht-so.sexy", referrer: "https://ldap.ist.nicht-so.sexy/"
This is because of an "invalid request block size"-error according to
`journalctl -u privacyidea.service`:
Dec 20 22:12:48 ldap uwsgi[10721]: invalid request block size: 4245 (max 4096)...skip
Increasing the buffer to 8192 fixes the problem for me.
adds defaultText for all options that use `cfg.*` values in their
defaults, but only for interpolations with no extra processing (other
than toString where necessary)
