using readFile instead of fileContents (or using indented strings) can
leave a trailing newline that causes build errors in systemd units and
has previously caused runtime errors in wireguard scripts. use
singleLineStr to strip a trailing newline if it exists, and to fail if
more than one is present.
when a new peer is added, it does not modify any active units, because
the interface unit remains the same. therefore the new peer is not added
until next reboot or manual action.
Make the dynamic-dns refresh systemd service (controlled via the
preexisting option dynamicEndpointRefreshSecond) robust to e.g. dns
failures that happen on intermittent network connections.
Background:
When dns resolution fails with a 'permanent' error ("Name or service not
known" instead of "Temporary failure in name resolution"), wireguard
won't retry despite WG_ENDPOINT_RESOLUTION_RETRIES=infinity.
-> This change should improve reliability/connectivity.
somewhat related thread: https://github.com/NixOS/nixpkgs/issues/63869
Some networks can only transfer packets with a lower than normal maximum
transfer unit size. In these cases, it is necessary to set a MTU that
works for the given upstream network.
Wireguard can tag its packets with a firewall mark. This can be used for
firewalls or policy routing. This is very useful in some setups where
all traffic should go through a wireguard interface. The wireguard
packets cannot go through the wireguard interface and must be routed
differently, which can be done via the Firewall Mark.
The nixos option `config.networking.wireguard.interface.<name>.fwMark`
is of type `types.str` and not `types.int` to allow for specifying the
mark as a hexadecimal value.
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.
As per `man systemd.path`:
> When a service unit triggered by a path unit terminates
> (regardless whether it exited successfully or failed),
> monitored paths are checked immediately again,
> **and the service accordingly restarted instantly**.
Thus the existence of the path unit made it impossible to stop the
wireguard service using e.g.
systemctl stop wireguard-wg0.service
Systemd path units are not intended for program inputs such
as private key files.
This commit simply removes this usage; the private key is still
generated by the `generateKeyServiceUnit`.
Standard best-practice shell quoting, which can prevent the most
horrible production accidents.
Note that we cannot use `+ optionalString someBool '' someString''`
because Nix's multi-line ''double-quoted'' strings remove leading
whitespace.
Until now, the `touch + chmod 600 + write` approach made it possible for
an unprivileged local user read the private key file, by opening
the file after the touch, before the read permissions are restricted.
This was only the case if `generatePrivateKeyFile = true` and the parent
directory of `privateKeyFile` already existed and was readable.
This commit fixes it by using `umask`, which ensures kernel-side that
the `touch` creates the file with the correct permissions atomically.
This commit also:
* Removes `mkdir --mode 0644 -p "${dirOf values.privateKeyFile}"`
because setting permissions `drw-r--r--` ("nobody can enter that dir")
is awkward. `drwx------` would perhaps make sense, like for `.ssh`.
However, setting the permissions on the private key file is enough,
and likely better, because `privateKeyFile` is about that file
specifically and no docs suggest that there's something special
about its parent dir.
* Removes the `chmod 0400 "${values.privateKeyFile}"`
because there isn't really a point in removing write access from
the owner of the private key.
The two new options make it possible to create the interface in one namespace
and move it to a different one, as explained at https://www.wireguard.com/netns/.
Previously each oneshot peer service only ran once and was not
restarted together with the interface unit. Because of this,
defined peers were missing after restarting their corresponding
interface unit.
Co-Authored-By: Franz Pletz <fpletz@fnordicwalking.de>
This can lead to unnecessary failures if the kernel module is already
loaded:
Jun 06 12:38:50 chef bglisn9bz0y5403vdw9hny0ij43r41jk-unit-script-wireguard-wg0-start[13261]: modprobe: FATAL: Module wireguard not found in directory /run/booted-system/kernel-modules/lib/modules/4.19.36
Before, changing any peers caused the entire WireGuard interface to
be torn down and rebuilt. By configuring each peer in a separate
service we're able to only restart the affected peers.
Adding each peer individually also means individual peer
configurations can fail, but the overall interface and all other peers
will still be added.
A WireGuard peer's internal identifier is its public key. This means
it is the only reliable identifier to use for the systemd service.
As a oneshot service, if the startup failed it would never be attempted again.
This is problematic when peer's addresses require DNS. DNS may not be reliably available at
the time wireguard starts. Converting this to a simple service with Restart
and RestartAfter directives allows the service to be reattempted, but at
the cost of losing the oneshot semantics.
Signed-off-by: Maximilian Bosch <maximilian@mbosch.me>
Ideally, private keys never leave the host they're generated on - like
SSH. Setting generatePrivateKeyFile to true causes the PK to be
generate automatically.
Wireguard is now split into two pretty much independent packages:
`wireguard` (Linux-specific kernel module) and `wireguard-tools`,
which is cross-platform.
This is more in line with what other services do; also looks cleaner.
It changes configuration entries for pre-and post-hooks type to lines from
lists of strings which are more logical for them; coersion is provided for
backwards compatibility.
Finally, add several steps to improve robustness:
1. Load kernel module on start if not loaded;
2. Don't remove wireguard interface on start; it is removed on service stop. If
it's not something is wrong.
Sometimes (especially in the default route case) it is required to NOT
add routes for all allowed IP ranges. One might run it's own custom
routing on-top of wireguard and only use the wireguard addresses to
exchange prefixes with the remote host.
Do the right thing, and use multiple interfaces for policy routing. For example, WireGuard interfaces do not allow multiple routes for the same CIDR range.
This adds a convenient per-peer option to set the routing table that associated routes are added to. This functionality is very useful for isolating interfaces from the kernel's global routing and forcing all traffic of a virtual interface (or a group of processes, via e.g. "ip rule add uidrange 10000-10009 lookup 42") through Wireguard.
