wt/finit
1
0
Fork 0
mirror of https://github.com/troglobit/finit.git synced 2026-10-09 16:50:59 +07:00
Code Issues Packages Projects Releases Wiki Activity
Files
9875e764197319ee632466344a7c2f93e088ccef
finit/doc/dbus.md
T
Joachim Wiberg 63aabaa6df libink/dbus: identify the caller behind a broker 
On the local bus SO_PEERCRED says who is calling and the kernel is the
one saying it.  Behind a broker one connection carries every caller,
so that credential describes dbus-daemon and nothing else, and every
privileged method was refused there, root included.

Ask the bus driver instead.  libink parks the call and hands us the
sender; we ask GetConnectionUnixUser and answer when the reply lands,
through the same event loop as everything else.  Nothing blocks:
blocking in PID 1 is why libuEv exists.  That needs calls libink can
make on a connection it already has, so it gained those too.

Answers are cached, since a bus never reuses a unique name while it
runs.  Not across a restart though: a new dbus-daemon numbers from
scratch and :1.7 becomes somebody else, so the cache goes when the
broker does.  A sender name too long to key on is refused rather than
truncated, two callers sharing a truncated key would share an
identity.

Privilege is no longer uid 0 alone.  The socket is already owned by
the --with-group group, so refusing its members every method that
changes anything left a wheel user able to open the bus and unable to
reboot.  Both gates now say the same thing.

Group membership needs NSS, which the C library loads with dlopen(),
so the lookup is compiled out where Finit is built to link statically.
That leaves such a build root-only, which is worth saying out loud
rather than leaving to be discovered.

Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
2026-08-13 10:15:06 +02:00

339 lines
17 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
D-Bus Integration
=================
Finit ships with a built-in, brokerless [D-Bus][] implementation, **libink**
(`-link`), that exposes the running init system as a peer on its own private
bus, and optionally on the system bus when `dbus-daemon` is available.
Everything `initctl` does is also reachable from any generic D-Bus tooling —
`dbus-send`, `dbus-monitor`, `gdbus`, language bindings, dashboards,
monitoring agents, etc.
> [!NOTE]
> D-Bus support is enabled by default, opt out at build time with
> `--disable-dbus`, see [Building](build.md) for details. When disabled,
> `initctl` keeps using the legacy `INIT_SOCKET` transport and Finit
> exposes no bus.
Bus address
-----------
| Bus | Address |
|------------------------|---------------------------------------------|
| Local (always) | `unix:path=/run/finit/bus` |
| System (opportunistic) | `unix:path=/var/run/dbus/system_bus_socket` |
The **local** bus is brokerless: clients connect straight to Finit over a
Unix-domain socket using the standard D-Bus SASL EXTERNAL handshake. No
`dbus-daemon` is required, which makes it suitable for embedded systems that
don't ship one.
The socket is `0660`, owned by `root` and the group given to
`--with-group` at build time, the same gate as `/run/finit/socket` that
`initctl` falls back on. The bus reaches every operation `initctl`
does, so restricting one and not the other would leave the door open.
Members of that group may use it, see [Authorization](#authorization).
The **system** bus is best-effort: Finit probes for a running `dbus-daemon`
and, when reachable, claims the well-known name `org.finit` so that standard
tooling sees Finit just like any other system service:
```sh
dbus-send --system --print-reply --dest=org.finit /org/finit/manager \
org.finit.Manager1.ListServices
dbus-monitor --system "sender='org.finit'"
```
If no system bus is present (the common case on embedded targets), this step
is silently skipped.
Object tree
-----------
```
/
├── org/
│ └── finit/
│ ├── manager Manager1
│ ├── cond Cond1
│ └── service/
│ ├── keventd Service1 (one per service)
│ ├── sshd
│ └── …
└── org/freedesktop/DBus Standard well-known interfaces
```
Every node implements the usual stock interfaces:
| Interface | Purpose |
|---------------------------------------|-------------------------------------------|
| `org.freedesktop.DBus` | `Hello`, `AddMatch`, `RemoveMatch` |
| `org.freedesktop.DBus.Peer` | `Ping`, `GetMachineId` |
| `org.freedesktop.DBus.Introspectable` | `Introspect()` — XML description |
| `org.freedesktop.DBus.Properties` | `Get`, `GetAll` (Set not yet implemented) |
Note: `Hello`, `AddMatch`, and `RemoveMatch` are answered on the canonical
`/org/freedesktop/DBus` object only, as per the D-Bus specification.
`org.finit.Manager1`
--------------------
Lives at **`/org/finit/manager`**. Owns the global init operations
and the service registry.
### Methods
| Method | In sig | Out sig | Priv. | Notes |
|------------------------------|--------|---------|-------|-----------------------------------------------------------|
| `ListServices` | — | `as` | no | Returns identities (`name`, `name:id`) of loaded services |
| `GetService` | `s` | `o` | no | Resolves an identity to its `Service1` object path |
| `Start` | `s` | — | yes | Start the service(s) matching the identity |
| `Stop` | `s` | — | yes | Stop the service(s) matching the identity |
| `Restart` | `s` | — | yes | Restart (stop + start) the service(s) |
| `Reload` | — | — | yes | Re-read all `*.conf` and apply changes |
| `SetRunlevel` | `u` | — | yes | Transition to runlevel `u` (0–6) |
| `SetDebug` | — | — | yes | Toggle Finit's runtime debug flag |
| `Signal` | `su` | — | yes | Send signal `u` (1–31) to services matching identity `s` |
| `Suspend` | — | — | yes | `sync()` + suspend-to-RAM |
| `Reboot`, `Halt`, `Poweroff` | — | — | yes | Trigger the corresponding shutdown sequence |
### Properties
All read-only strings; observable via `Properties.Get` and
`Properties.GetAll`.
| Property | Type | Returns |
|----------------|------|--------------------------------------------------------|
| `Runlevel` | `s` | Current runlevel as a digit (`"2"`, `"3"`, …) or `"S"` |
| `PrevRunlevel` | `s` | Previous runlevel, same encoding, `"N"` when none |
| `Version` | `s` | Finit's version string (`PACKAGE_VERSION`) |
### Signals
| Signal | Body | Fires when |
|-----------------------|----------------------------------------|------------------------|
| `ServiceStateChanged` | `sss` — identity, old state, new state | Service transitions |
| `RunlevelChanged` | `ss` — old level, new level | System runlevel change |
State names emitted by `ServiceStateChanged` are stable wire strings:
`halted`, `done`, `dead`, `cleanup`, `teardown`, `stopping`, `setup`,
`paused`, `waiting`, `starting`, `running`. `RunlevelChanged` levels
use the same encoding as the `Runlevel` property: digits, `"S"`, `"N"`.
`org.finit.Service1` (per-service objects)
------------------------------------------
Lives at **`/org/finit/service/<encoded>`**, one object per loaded service.
`<encoded>` is the service identity (name, or `name:id` for templated
services) put through systemd-style `_HH` hex escaping — ASCII alphanumerics
and `_` pass through, anything else becomes `_HH` where `HH` is the hex byte.
Use `Manager1.GetService(identity)` to look up the exact path rather than
constructing it by hand.
| Method | In sig | Out sig | Priv. | Notes |
|-----------|--------|---------|-------|-------------------------------------------------------------|
| `Start` | — | — | yes | Equivalent to `Manager1.Start(<identity>)` for this service |
| `Stop` | — | — | yes | … |
| `Restart` | — | — | yes | … |
| `Reload` | — | — | yes | Reload (SIGHUP if supported, else restart) |
### Properties
All read-only; observable via `Properties.Get` and `Properties.GetAll`.
| Property | Type | Returns |
|-----------------|------|----------------------------------------------------------------|
| `Identity` | `s` | Service identity, `name` or `name:id` |
| `Name` | `s` | Program name (basename of the command) |
| `State` | `s` | Current status, same vocabulary as `initctl status` |
| `Pid` | `u` | Current PID, 0 when not running |
| `RestartCount` | `u` | Restarts since the last stable run |
| `Runlevels` | `u` | Allowed runlevels as a bitmask, bit N = runlevel N, bit 10 = S |
| `Description` | `s` | The service's `description` string |
| `Command` | `s` | Full command line, arguments included |
| `Conditions` | `s` | Declared conditions, raw `.conf` form |
| `Type` | `s` | Unit type: `service`, `task`, `run`, `sysv`, `tty`, `free` |
| `Origin` | `s` | Source `.conf` file, empty for built-ins |
| `Environment` | `s` | The service's `env` setting, raw |
| `PidFile` | `s` | Declared PID file, raw (`!` prefix included) |
| `User` | `s` | User the service runs as |
| `Group` | `s` | Group the service runs as |
| `Uptime` | `u` | Seconds since start, 0 when not running |
| `ExitStatus` | `u` | Raw `waitpid(2)` status from the last exit |
| `RestartsTotal` | `u` | Restarts over the service's lifetime |
| `RestartMax` | `u` | Restart limit before the service is blocked |
| `Starts` | `u` | Times started, for `manual-start` units |
| `ManualStart` | `b` | `manual-start` set in the `.conf` |
| `Forking` | `b` | Daemon forks to background |
| `Started` | `b` | Run/task completed successfully |
On every state transition the object also emits the standard
`org.freedesktop.DBus.Properties.PropertiesChanged` signal: `State`
in the changed dictionary, `Pid` and `RestartCount` invalidated (call
`Get` for fresh values).
The per-service surface lets generic tooling supply an object handle
once and then invoke methods on it, instead of repeatedly passing the
identity string.
`org.finit.Cond1`
-----------------
Lives at **`/org/finit/cond`**. Exposes Finit's
[condition system](conditions.md) to bus clients.
### Methods
| Method | In sig | Out sig | Priv. | Notes |
|---------|--------|---------|-------|--------------------------------------------------------------|
| `Get` | `s` | `s` | no | Returns `"on"`, `"off"`, or `"flux"` for the named condition |
| `Set` | `s` | — | yes | Assert a `usr/<name>` condition |
| `Clear` | `s` | — | yes | Deassert a `usr/<name>` condition |
| `List` | — | `as` | no | Names of all known conditions |
| `Dump` | — | `a(ss)` | no | `(name, state)` pairs for everything `List` returns |
Note: non-`usr/*` paths are rejected with `InvalidArgs` -- system
conditions belong to Finit's state machine.
### Signals
| Signal | Body | Fires when |
|--------------------|------------------------|---------------------------------------|
| `ConditionChanged` | `ss` — name, new state | A condition is asserted or deasserted |
Authorization
-------------
Privileged methods accept `root`, and any caller belonging to the group given
to `--with-group` at build time. That is the same set the socket mode already
admits, so the two gates agree instead of the socket letting the group in and
every method turning it away.
A build configured with `--enable-static` accepts only `root`. Group
membership comes from NSS, which the C library loads with `dlopen()`, and a
build meant to link statically cannot count on that, so the lookup is compiled
out rather than left to fail open. The socket mode is unchanged, so the group
can still connect, it just cannot invoke a privileged method.
On the **local** bus the kernel decides this at `connect()`, supplementary
groups included, and `SO_PEERCRED` tells Finit exactly who is calling, so
privilege escalation through the bus is impossible.
On the **system** bus one connection carries every caller, so `SO_PEERCRED`
describes `dbus-daemon` rather than whoever asked. For a privileged method
Finit asks the bus driver `GetConnectionUnixUser` about the message sender and
holds the call until the answer arrives. Nothing blocks: the reply comes back
through the same event loop as everything else, and the held call is then
dispatched or refused on its merits.
Answers are cached per sender. A bus never reuses a unique name while it
runs, so an answer holds for as long as that bus does; Finit empties the cache
when the broker goes away, since a new one numbers its clients from scratch.
A caller Finit cannot identify is refused, so the failure mode is a denial
rather than an escalation.
When a privileged method is rejected the error name is exactly
`org.freedesktop.DBus.Error.AccessDenied`, and the body carries a short reason
string.
`initctl` integration
---------------------
`initctl` transparently routes through D-Bus when the bus socket is present,
and falls back to the legacy `INIT_SOCKET` transport otherwise. Concretely,
the following subcommands use the bus first:
| Subcommand | Method |
|----------------------------|--------------------------------------------------|
| `initctl start` | `Manager1.Start(svc)` |
| `initctl stop` | `Manager1.Stop(svc)` |
| `initctl restart` | `Manager1.Restart(svc)` |
| `initctl reload` | `Manager1.Reload()` |
| `initctl reload S` | `Service1.Reload()` (per-svc) |
| `initctl reboot` | `Manager1.Reboot()` |
| `initctl halt` | `Manager1.Halt()` |
| `initctl poweroff` | `Manager1.Poweroff()` |
| `initctl suspend` | `Manager1.Suspend()` |
| `initctl debug` | `Manager1.SetDebug()` |
| `initctl signal` | `Manager1.Signal(svc, signo)` |
| `initctl runlevel` | `Properties.Get(Manager1.Runlevel/PrevRunlevel)` |
| `initctl cond set/get/clear` | `Cond1.{Set,Get,Clear}`, `clr` is an alias |
Two `initctl` subcommands are pure D-Bus features without legacy equivalents:
* `initctl monitor` — subscribes to every signal on the local bus and prints
one line per delivery (with timestamp, interface and member). Same idea as
`dbus-monitor`, but scoped to Finit and with no need to pass `--address`.
* `initctl cond` emits the standard `Cond1.ConditionChanged` signal as a side
effect, so subscribers observe user-driven state changes the same way they
observe service-driven ones.
Examples
--------
The examples below use `dbus-send` and `dbus-monitor`, which ship as part of
the [dbus][] reference implementation; they're widely packaged and don't pull
in any extra runtime. Any tool that speaks D-Bus over an AF_UNIX socket works
equally well — `gdbus`, Python's `jeepney`/`dasbus`, etc. — substitute their
syntax for setting the bus address. The wire protocol is the compatibility
surface: Finit's own **libink** is an internal implementation detail, external
clients should use any standard D-Bus library.
When `org.finit` is registered on the system bus you can replace
`--address=unix:path=/run/finit/bus` with `--system` in any example below.
List the running services:
```sh
dbus-send --address=unix:path=/run/finit/bus \
--type=method_call --print-reply --dest=org.finit \
/org/finit/manager \
org.finit.Manager1.ListServices
```
Read the current runlevel via the Properties interface:
```sh
dbus-send --address=unix:path=/run/finit/bus \
--type=method_call --print-reply --dest=org.finit \
/org/finit/manager \
org.freedesktop.DBus.Properties.Get \
string:org.finit.Manager1 string:Runlevel
```
Subscribe to every state change on the manager object:
```sh
dbus-monitor --address=unix:path=/run/finit/bus \
"type='signal',interface='org.finit.Manager1'"
```
Or use `initctl monitor`, which does the same without any address
plumbing.
Restart a service by its object path:
```sh
dbus-send --address=unix:path=/run/finit/bus \
--type=method_call --dest=org.finit \
/org/finit/service/sshd \
org.finit.Service1.Restart
```
Trigger a `usr/`-condition assertion that wakes any dependent service:
```sh
dbus-send --address=unix:path=/run/finit/bus \
--type=method_call --dest=org.finit \
/org/finit/cond \
org.finit.Cond1.Set string:"data-ready"
```
The `--dest=org.finit` argument is informational on the local
brokerless bus — Finit accepts any destination because there's no
broker to route by name — but `dbus-send` requires it syntactically.
[dbus]: https://gitlab.freedesktop.org/dbus/dbus
[D-Bus]: https://dbus.freedesktop.org/doc/dbus-specification.html
Reference in New Issue View Git Blame Copy Permalink