Compare commits

..

2 Commits

Author SHA1 Message Date
waldek 93270e3901 merges the local changes 2021-09-05 23:49:43 +02:00
waldek 54f6bcbeeb adds dbus introduction 2021-09-05 23:42:44 +02:00
6 changed files with 279 additions and 19 deletions

Binary file not shown.

After

Width:  |  Height:  |  Size: 28 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 126 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 137 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 158 KiB

View File

@ -1,11 +1,271 @@
# DBus
# dbus
## What is dbus
From our system administrator's point of view, dbus is a bit like systemd.
Most of the time we're unaware of it's existence but it's an essential part of most modern Linux distributions.
But what *is* it and what does it do?
## Command line interfacing with dbus
The plain truth is that dbus is an inter process communication bus or [IPC](https://en.wikipedia.org/wiki/Inter-process_communication).
Each program of server we launch is an independent *process* that is managed by our operating system.
It will claim the necessary memory it needs to run, to store data and variables, but this chunk of memory is exclusive to the process.
No other processes *should* be able to access this memory and data.
But how can programs talk to each other if they need to?
This is where dbus comes on the scene.
## Graphical interfacing
![dbus](./assets/dbus_01.png)
Dbus allows for programs to expose variables and method, functions the program can execute, to the other program connected to dbus.
The program exposing has full control over what it will **do** when methods are called or variables accessed.
Dbus is in charge of connecting and delivering what it's client demand of each other.
While this sounds simple, it does this with grate care and precision.
The dbus [specification](https://dbus.freedesktop.org/doc/dbus-specification.html) states that:
> D-Bus is a system for low-overhead, easy to use interprocess communication (IPC). In more detail:
> * D-Bus is low-overhead because it uses a binary protocol, and does not have to convert to and from a text format such as XML. Because D-Bus is intended for potentially high-resolution same-machine IPC, not primarily for Internet IPC, this is an interesting optimization. D-Bus is also designed to avoid round trips and allow asynchronous operation, much like the X protocol.
> * D-Bus is easy to use because it works in terms of messages rather than byte streams, and automatically handles a lot of the hard IPC issues. Also, the D-Bus library is designed to be wrapped in a way that lets developers use their framework's existing object/type system, rather than learning a new one specifically for IPC.
## Two different busses
On most machines you'll encounter two different, and independent, busses.
* a system bus used by the system to communicate
* a session bus (per logged in user) for programs run by the user
This separation is needed from both practical and security stand point.
Practically speaking I want to control *my* VLC player and not some other one.
From a security point of view unprivileged users should not be allowed to circumvent permissions by hopping on the system bus and triggering all sorts of things they are not allowed to!
Let's have a look at what's running on my home computer.
We clearly see two independent busses running.
```bash
➜ ~ git:(master) ✗ ps aux | grep dbus
message+ 548 0.0 0.0 8864 5192 ? Ss Aug28 0:06 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile --systemd-activation --syslog-only
waldek 1780 0.0 0.0 8628 5136 ? Ss Aug28 0:01 /usr/bin/dbus-daemon --session --address=systemd: --nofork --nopidfile --systemd-activation --syslog-only
waldek 799748 0.0 0.0 6316 656 pts/2 S+ 20:42 0:00 grep --color=auto --exclude-dir=.bzr --exclude-dir=CVS --exclude-dir=.git --exclude-dir=.hg --exclude-dir=.svn --exclude-dir=.idea --exclude-dir=.tox dbus
➜ ~ git:(master) ✗
```
Now on a server.
Where we only see **one** bus which is the system bus.
This server is headless, so no graphical login is possible.
```bash
➜ ~ ps aux | grep dbus
message+ 596 0.0 0.1 8956 3744 ? Ss Mar22 29:09 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile --systemd-activation --syslog-only
waldek 2982 0.0 0.0 6144 824 pts/0 S+ 18:44 0:00 grep --color=auto --exclude-dir=.bzr --exclude-dir=CVS --exclude-dir=.git --exclude-dir=.hg --exclude-dir=.svn --exclude-dir=.idea --exclude-dir=.tox dbus
➜ ~
```
Both `PID`'s are in the same *ballpark* number wise and on the lower end of the spectrum.
This probably means the dbus system bus process is started at boot.
Let's find out a bit more about it.
As it's a program running at boot, it's probably started by `systemd`.
```bash
➜ ~ sudo systemctl list-dependencies dbus.service --reverse
dbus.service
● └─multi-user.target
● └─graphical.target
➜ ~
```
Here we can see that dbus is started when the computer reaches the `multi-user.target` runlevel.
If we would reboot in single user mode, or `rescue.target`, dbus would not be launched!
What about the user bus?
We don't see any specific service files for that one?
Let's have a look at the `systemctl --user` output.
```bash
➜ ~ git:(master) ✗ systemctl --user --type=service | grep dbus
dbus.service loaded active running D-Bus User Message Bus
➜ ~ git:(master) ✗
```
```bash
➜ ~ git:(master) ✗ systemctl --user list-dependencies dbus.service
dbus.service
● ├─app.slice
● ├─dbus.socket
● └─basic.target
● ├─paths.target
● ├─sockets.target
● │ ├─dbus.socket
● │ ├─dirmngr.socket
● │ ├─gpg-agent-browser.socket
● │ ├─gpg-agent-extra.socket
● │ ├─gpg-agent-ssh.socket
● │ ├─gpg-agent.socket
● │ ├─pipewire.socket
● │ ├─pk-debconf-helper.socket
● │ └─pulseaudio.socket
● └─timers.target
➜ ~ git:(master) ✗
```
```bash
➜ ~ git:(master) ✗ systemctl --user list-dependencies dbus.service --reverse
dbus.service
➜ ~ git:(master) ✗
```
## Inspecting the bus
### Graphical
When you're in a graphical environment the *easiest* tool to understand what is exposed on the bus is `d-feet`.
To see what is going on on the bus you're better off with `bustle`.
Let's install both and open up d-feet first.
We can do all of this in the command line as well, but first we'll have a look at the graphical applications.
For an easy demonstration I would like you to install `vlc` as well.
```bash
➜ ~ git:(master) ✗ sudo apt install d-feet bustle vlc
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
bustle is already the newest version (0.8.0-1).
d-feet is already the newest version (0.3.15-3).
vlc is already the newest version (3.0.16-1).
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
➜ ~ git:(master) ✗
```
Open up `d-feet` and `vlc` and search for vlc on the session bus.
You should see something very similar to the screenshot below.
Open a video or audio file in vlc and try to pause it from d-feet.
Have a look at some properties and some methods.
What would the *signals* be?
![d-feet](./assets/dbus_02.png)
Now keep both vlc and d-feet open and we'll add bustle to the mix.
Bustle can record events on either the system or your session bus.
As we're using vlc for the demonstration go ahead and record the session bus.
Try to pause and play your video and try to spit the flow of operations.
![bustle](./assets/dbus_03.png)
### From the command line
Dbus comes with it's own command line tools to interact with programs exposed on the bus.
Let's have a look at them.
First up is `dbus-monitor` and let's see it's arguments.
```bash
➜ ~ git:(master) ✗ dbus-monitor --help
Usage: dbus-monitor [--system | --session | --address ADDRESS] [--monitor | --profile | --pcap | --binary ] [watch expressions]
➜ ~ git:(master) ✗
```
The most general way to invoke it is with either `--system` or `--session`.
We've already looked at the session bus so let's now look at the system bus.
Let's one op one terminal, run `sudo dbus-monitor --system` and keep it running.
In a second terminal stop one of your running services, such as ssh.
You'll see a bunch of messages appearing in the monitor terminal.
Start the service back up.
Do you understand what is happening here?
Try and visualize both terminals at the same time as in the screenshot below.
Try some tabcomplete and see what is happening.
![tab complete](./assets/dbus_04.png)
## Sending messages
We can send things over dbus ourselves with the `dbus-send` program but first we'll use a special program that's very handy to send messages to the notification program of our desktop environment.
It's called `notify-send` and you can install it by installing the `libnotify-bin` package ins Debian.
Once it's installed try it out as follows.
```
dbus-send --session --print-reply --dest=org.freedesktop.Notifications --type=method_call --reply-timeout=10000 /org/freedesktop/Notifications org.freedesktop.Notifications.Notify string:'Test Application' uint32:0 string: string:'NOTIFICATION TEST' string:'This is a test of the notification system via DBus.' array:string: dict:string:string: int32:10000
➜ ~ git:(master) ✗ notify-send "hello"
➜ ~ git:(master) ✗
```
It immediately return but a notification should have popped up on your screen.
You can use this from within bash scripts to signal the user about updates or errors.
But let's dig a little deeper to see what's happening on the inside.
For this we need to inspect the session bus and post a message again.
Look for your message and observe the method that was called.
It should look similar to the one below.
```bash
method call time=1630874537.889652 sender=:1.544 -> destination=:1.52 serial=7 path=/org/freedesktop/Notifications; interface=org.freedesktop.Notifications; member=Notify
string "notify-send"
uint32 0
string "dialog-information"
string "Hello world!"
string "I am a notification..."
array [
]
array [
dict entry(
string "urgency"
variant byte 1
)
]
int32 -1
```
With this information we can construct a raw dbus-send command ourselves (but we'll run into a problem!).
Let's try and map the usage to what we observed above.
Luckily `dbus-send` does autocomplete!
But we can take d-feet on the side to better understand the structure.
```bash
➜ ~ git:(master) ✗ dbus-send --help
Usage: dbus-send [--help] [--system | --session | --bus=ADDRESS | --peer=ADDRESS] [--dest=NAME] [--type=TYPE] [--print-reply[=literal]] [--reply-timeout=MSEC] <destination object path> <message name> [contents ...]
➜ ~ git:(master) ✗ dbus-send --session --print-reply --dest=org.freedesktop.Notifications /org/freedesktop/Notifications org.freedesktop.Notifications.Notify
```
We need to send a message to:
* `--session` bus
* we want to see the reply, or error, if any `--print-reply`
* our destination is `--dest` the notifications
* at this destination we only have one object `/org/freedesktop/Notifications`
* on this object we can call the `org.freedesktop.Notifications.Notify` method
But we run in to an error! (that's why we want the `--print-reply` argument)
```bash
Error org.freedesktop.DBus.Error.InvalidArgs: Type of message, “()”, does not match expected type “(susssasa{sv}i)”
```
The `Notify` method needs arguments and quite a lot of them!
In `d-feet` or our `dbus-monitor` output above we can visualize what these arguments are.
The *type* in the error message looks abstract but when we put it side by side with our dbus-monitor output it starts to make sense.
```bash
string "notify-send"
uint32 0
string "dialog-information"
string "Hello world!"
string "I am a notification..."
array [
]
array [
dict entry(
string "urgency"
variant byte 1
)
]
int32 -1
```
Dbus can send only specific types over the wire and they are described in the specification online.
We could try to get get to work on the command line but sadly it won't!
`dbus-send` and not send the *variant* type in an array so we're out of luck (I might be wrong here!).
An alternative program that can complete this challenge is `gdbus`.
Try to get that one to display a notification!
Let's try to interact with vlc via dbus-send instead.
Below are some examples that you'll have to tweak to suit your needs.
```bash
dbus-send --session --print-reply=literal --dest=org.mpris.MediaPlayer2.vlc /org/mpris/MediaPlayer2 org.mpris.MediaPlayer2.Player.Pause
dbus-send --session --print-reply=literal --dest=org.mpris.MediaPlayer2.vlc /org/mpris/MediaPlayer2 org.mpris.MediaPlayer2.Player.Play
dbus-send --session --print-reply=literal --dest=org.mpris.MediaPlayer2.vlc /org/mpris/MediaPlayer2 org.mpris.MediaPlayer2.Player.SetPosition objpath:/org/videolan/vlc/playlist/3 int64:100000000
dbus-send --session --print-reply=literal --dest=org.mpris.MediaPlayer2.vlc /org/mpris/MediaPlayer2 org.mpris.MediaPlayer2.Player.OpenUri string:"file:///home/waldek/Coralis_MASTER+VO-V5.mp4"
```

View File

@ -22,7 +22,7 @@ We start by writing a small script that serves as a placeholder to prove our wor
It will just echo a line to `journalctl`.
No need to manage our own logging systemd as we can take full advantage of systemd's built in logging.
```
```bash
➜ ~ git:(master) ✗ cat backup.sh
#!/bin/bash
@ -34,7 +34,7 @@ exit 0
We can now take this mini script and turn it into service by creating a `backup.service` file in `/etc/systemd/system`.
Let's do this and populate as such.
```
```ini
➜ ~ git:(master) ✗ cat /etc/systemd/system/backup.service
[Unit]
Description=Our own backup script
@ -49,7 +49,7 @@ WantedBy=default.target
Next we need to reload systemd and test the service.
```
```bash
➜ ~ git:(master) ✗ sudo systemctl daemon-reload
➜ ~ git:(master) ✗ sudo systemctl start backup.service
➜ ~ git:(master) ✗ sudo systemctl status backup.service
@ -70,7 +70,7 @@ It seems to have worked nicely!
Now we will write the udev rule to start our service when a USB stick is plugged in.
We'll start with a very *generic* rule that triggers for each block device that is plugged in.
```
```ini
➜ ~ git:(master) ✗ cat /etc/udev/rules.d/99-backup_to_usb_stick.rules
SUBSYSTEM=="block", ACTION=="add", ENV{SYSTEMD_WANTS}="backup.service"
➜ ~ git:(master) ✗
@ -81,7 +81,7 @@ We can monitor the logs *live* with the `-f` argument if we want.
You'll see the ourput of the placeholder script appear when you plug in any USB stick!
It will also double or even triple trigger, depending on the number of partitions you have on your stick but we know how to fix this.
```
```bash
➜ ~ git:(master) ✗ sudo udevadm control --reload
➜ ~ git:(master) ✗ sudo journalctl -f --unit backup.service
-- Journal begins at Wed 2021-07-14 22:35:36 CEST, ends at Mon 2021-08-23 21:59:26 CEST. --
@ -99,7 +99,7 @@ I only want my backup do be done to a very specific USB stick I trust.
When I plug it in I can inspect it's attributes and environment variables via `udevadm`.
I'll use the filesystem's UUID to identify the disk.
```
```bash
➜ ~ git:(master) ✗ sudo udevadm info --name=/dev/sdd1 | grep UUID
E: ID_PART_TABLE_UUID=d9f8a99f-673a-334c-af4f-18697ed888c9
E: ID_FS_UUID=2469ffe5-e066-476d-805a-cde85a58ea3b
@ -111,7 +111,7 @@ E: ID_PART_ENTRY_UUID=73a25ff8-0e0f-1147-8aaa-7976bf921ced
I modify the rule, reload udev and inspect my logs after plugging in the device.
This works like a charm!
```
```bash
➜ ~ git:(master) ✗ cat /etc/udev/rules.d/99-backup_to_usb_stick.rules
SUBSYSTEM=="block", ENV{ID_FS_UUID}=="2469ffe5-e066-476d-805a-cde85a58ea3b", ACTION=="add", ENV{SYSTEMD_WANTS}="backup.service"
➜ ~ git:(master) ✗ sudo udevadm control --reload
@ -136,7 +136,7 @@ The full solution requires multiple steps and we'll go over them one at time.
First we'll have to modify our rule once more by adding the following.
This passes the drive name to the service.
```
```bash
➜ ~ git:(master) ✗ cat /etc/udev/rules.d/99-backup_to_usb_stick.rules
SUBSYSTEM=="block", ENV{ID_FS_UUID}=="2469ffe5-e066-476d-805a-cde85a58ea3b", ACTION=="add", ENV{SYSTEMD_WANTS}="backup@$name.service"
➜ ~ git:(master) ✗ sudo udevadm control --reload
@ -151,14 +151,14 @@ When the partition we want to mount is `/dev/sde1` the service that will get tri
These are called **template** services and in order to create one, you *just* have to have the `@` symbol in the service name.
We'll move our old backup.service and reload systemd as follows.
```
```bash
➜ ~ git:(master) ✗ sudo mv /etc/systemd/system/backup.service /etc/systemd/system/backup@.service
➜ ~ git:(master) ✗ sudo systemctl daemon-reload
```
Inspecting the logs while plugging in our drive now gives us the following output.
```
```bash
➜ ~ git:(master) ✗ sudo journalctl -f
-- Journal begins at Wed 2021-07-14 22:35:36 CEST. --
Aug 23 22:40:38 deathstar sudo[964916]: waldek : TTY=pts/6 ; PWD=/home/waldek ; USER=root ; COMMAND=/usr/bin/mv /etc/systemd/system/backup.service /etc/systemd/system/backup@.service
@ -203,7 +203,7 @@ We can use the variables passed through via the service name by modifying the `b
The argument used is the `%I` when calling our script.
Let's look at the final service file and reload systemd.
```
```ini
➜ ~ git:(master) ✗ cat /etc/systemd/system/backup@.service
[Unit]
Description=Our own backup script
@ -222,7 +222,7 @@ WantedBy=default.target
The script should now be receiving the partition location as a first argument.
We can echo it out by modifying the script and inspecting the logs via `journalctl`.
```
```bash
➜ ~ git:(master) ✗ cat backup.sh
#!/bin/bash
@ -249,7 +249,7 @@ Our placeholder script is a good proof of concept to debug the flow of operation
Let's write a simple backup script though.
I like using `rsync` so you'll have to install it if you want to play around with this script.
```
```bash
➜ ~ git:(master) ✗ cat backup.sh
#!/bin/bash
@ -271,7 +271,7 @@ You can see all the files that get backed up the first time around.
The second time around no files are copied because I made no changes to any files.
This is the behaviour of the `-a` argument to `rsync`.
```
```bash
➜ ~ git:(master) ✗ sudo journalctl -f
-- Journal begins at Wed 2021-07-14 22:35:36 CEST. --
Aug 23 23:04:40 deathstar sudo[969830]: waldek : TTY=pts/6 ; PWD=/home/waldek ; USER=root ; COMMAND=/usr/bin/journalctl -f