OpenBSD on the Desktop (Part II)

GUI Tweaks et cetera

Patrick Bucher

2020-09-12

A week ago, I've installed OpenBSD on my Thinkpad. I've been using it now and then, and already have changed a couple of things in respect to the original setup described in the article. I also installed OpenBSD on the Dell Optiplex on which I previously installed FreeBSD a month before. This means that I'm no longer using FreeBSD on the desktop, at least not for the moment. However, FreeBSD is running on a disk station I built earlier this summer. Maybe I'll describe that particular setup (using ZFS) in a later article.

Except for that storage server, I'd like to use OpenBSD for most of my private computing. In this article, I describe some GUI tweaks and additional setup tasks I perfmormed in order to feel more at home on my OpenBSD machines. Some of the tasks performed are not specific to OpenBSD, but could also be applied to a Linux setup.

doas

sudo originally came from the OpenBSD community. It is almost as widely used in the Unix world as SSH, which is the most prominent OpenBSD project. However, sudo became bigger and harder to configure. Therefore, Ted Unangst came up with a simpler alternative called doas, which stands for Dedicated OpenBSD Application Subexecutor. doas is less powerful than sudo, but much smaller, easier to configure, and, thus, more secure. The full rationale can be read in Ted Unangst's Blog.

A basic doas setup requires to login as root for one last time. The configuration shall be kept extremely simple. I'd like to permit all users from the wheel group (which is just me on my computers) to use doas without entering the password every time but only once when executing a command that requires root permissions. This is only a single line in /etc/doas.conf:

permit persist :wheel

Let's check this setup by logging in as a user of the wheel group and trying to update the packages:

$ doas pkg_add -u

This works, so bye bye root account.

Fonts for dwm, dmenu, and st

By default, dwm, dmenu, and st use a monospace font of size 10, or pixelsize 12, respectively, which is hard to read on a screen with a high resolution. On Linux, I use the the TrueType font DejaVu Sans Mono. For OpenBSD, I'd rather use something more minimalistic: the Terminus bitmap font.

As pkg_info -Q terminus shows, this font comes in different versions. I prefer the version with the centered tilde, which I install:

$ doas pkg_add terminus-font-4.47p0-centered_tilde

Let's reconfigure st first, for testing changes doesn't require a restart of the window manager. I stored my suckless sources in ~/suckless, so the font for st can be configured in ~/suckless/config.h. I replace the existing font configuration

static char *font = "Liberation Mono:pixelsize=12:antialias=true:autohint=true";

with

static char *font = "Terminus:pixelsize=24";

The options antialias and autohinting are not needed for a bitmap font, so I left them away. 24 pixels is rather big, but my screen is big enough to show two text editors with more than 80 characters per line next to each other, so let's keep it this way. I rebuild and reinstall st, then switch to dwm:

$ doas make install
$ cd ../dwm

The font configuration in the config.h file looks a bit different here:

static const char *fonts =      { "monospace:size=10" };
static const char dmenufont =   "monospace:size=10";

Let's just use the same font as for st here:

static const char *fonts =      { "Terminus:pixelsize=24" };
static const char dmenufont =   "Terminus:pixelsize=24";

Note that I'm using pixelsize instead of size here. (24pt would be much bigger than 24px.) Then I rebuild and reinstall dwm.

# make install

This configuration appllies also to dmenu and slstatus, so we're done with the fonts.

X Background

By default, the desktop background is a pattern of black and grey dots, which is a strain to the eye. Even though I rarely look at an empty desktop for long, I'd rather change this to a solid color. This can be done by adding a command to ~/.xinitrc:

xsetroot -solid black

Right before dwm is executed.

USB Flash Drive

Even though SSH is almost ubiquitous nowadays, a USB flash drive is still useful when it comes to exchanging data between computers, especially if Windows is involved, or if the network does not allow SSH.

Block storage devices are accessible through the device nodes /dev/sd*, whereas * stands for the number of the disk. The disks can be listed as follows:

$ sysctl hw.disknames
hw.disknames=sd0:ef0268c97ae7a246

Only sd0 is active, even though I already plugged in my USB dongle. However, the system already figured out that there is a second disk:

$ sysctl hw.diskcount
hw.diskcount=2

The next free disk would have the name sd1. The device nodes can be created by running the MAKEDV script in /dev:

$ cd /dev
$ doas sh MAKEDEV sd1

Let's initialize a new MBR partition schema on sd1:

$ doas fdisk -iy sd1

The new disk layout can be checked using disklabel:

$ doas disklabel sd1
# /dev/rsd1c
...

The first line of the output tells us that there's a partition under /dev/rsd1c. (The r refers to «raw», as opposed to «block».) The partition can be formatted using newfs by referring to that partition name:

$ doas newfs sd1c

This creates a default FFS (Fast File System) partition, which is useful to exchange data between BSD operating systems. The formatted partition is then ready to be mounted:

$ doas mount /dev/sd1c /mnt

Other Partition Types

Other partition types are available under other utilities.

FAT32

The following command creates a FAT32 partition:

$ doas newfs_msdos -F 32 sd1c

The -F 32 parameter specifies FAT32 (as opposed to FAT16 or FAT8). To mount the partition, use the according mount command:

$ doas mount_msdos /dev/sd1c /mnt

EXT2

In order to create an ext2fs file system, the partition type needs to be specified accordingly. First, you might consider a GPT partition schema instead of MBR (additional -g parameter):

$ doas fdisk -igy sd1

Then use disklabel interactively to define a new partition:

$ doas disklabel -E sd1

First, delete all the partitions with z. Then, create a new partition with a, and make sure to specify the type as ext2fs instead of the default 4.2BSD. Notice that the new partition has a different letter (say, a), so you need to use sd1a instead of sd1c for the next steps. Write the changes by typing w, then exit with q. Now you can format and mount the partition:

$ doas newfs_ext2fs sd1a
$ doas mount_ext2fs /dev/sd1a /mnt

SSH Key (GitHub)

In order to access my GitHub repositories, I first create a new SSH key:

$ ssh-keygen -t rsa -b 4096

Since I manage my passwords with pass (of which more later), I don't know most of them by heart. So I can't just login to GitHub and add my public key. Therefore, I copy my public key to my laptop, on which I'm already logged in to GitHub.

This can be either done using scp, for which sshd has to be running on my laptop (which currently has the IP 192.168.178.53):

$ scp ~/.ssh/id_rsa.pub 192.168.178.53:/home/patrick

Or using the USB flash drive formatted with ext2 from before:

$ doas newfs_ext2fs -I sd1a
$ doas mount_ext2fs /dev/sd1a /mnt
$ doas cp ~/.ssh/id_rsa.pub /mnt/

Then id_rsa.pub can be copied into the according GitHub Settings Page, after which cloning GitHub repositories should work on the OpenBSD machine:

$ git clone git@github.com:patrickbucher/conf

GPG Key

My passwords are encrypted using GPG. To encrypt them, I need to copy my private key from my other machine. First, I list my private keys:

$ gpg --list-keys --keyid-format SHORT
pub   rsa2048/73CE6620 2016-11-11 [SC]
      22F91EE20D641CBCF5B8678E82B7FE3A73CE6620
uid         [ultimate] Patrick Bucher <patrick.bucher@mailbox.org>
sub   rsa2048/AF6246E3 2016-11-11 [E]

Then I export both public and private key to an according file using the armored key format:

$ gpg --export --armor 73CE6620 > public.key
$ gpg --export-secret-key --armor 73CE6620 > private.key

The two key files can be copied via SSH or the USB flash disk again, which I won't show here.

Back on my OpenBSD machine, I need to install GnuPG first, because OpenBSD only has signify installed by default:

$ doas pkg_add gnupg

I pick the 2.2 version. Now I can import my keys:

$ gpg2 --import private.key
$ gpg2 --import public.key

The key is not trusted so far, so I need to change that:

$ gpg2 --edit-key 73CE6620
> trust
> 5
> y
> quit

5 stands for ultimate trust, which seems appropriate.

Password Manager

I use pass as a password manager, which can be installed as the password_store package in OpenBSD:

$ doas pkg_add password-store

Now that I have both my GPG private key and a working SSH key for GitHub, I can clone my passwords stored on a private GitHub repository:

$ git clone git@github.com:patrickbucher/pass .password-store

Now I can copy my GitHub password to the clipboard as follows:

$ pass -c github

Aliases

I use a lot of aliases, such as gcl as a shortcut for git clone, and gad for git add, etc. Since OpenBSD uses a Public Domain Korn Shell by default, the .bashrc configuration from my Linux machines won't work here, unless I switch to bash, which is not exactly the point of using OpenBSD.

I define my aliases in ~/.kshrc (excerpt):

alias gcl='git clone'
alias gad='git add'

In order to load those settings, an according ENV parameter needs to be defined in ~/.profile (see man 1 ksh for details):

export ENV=$HOME/.kshrc

After the next login, ~/.profile is reloaded, and the aliases are ready to be used.

Conclusion

Not only is my enhanced setup now ready to do some serious work, but I also increased my understanding of some OpenBSD subjects. There are still things to be improved and to be understood, but my setup is now good enough so that I no longer need a Linux machine running next to it. I'm looking forward to use and learn about OpenBSD in the time to come. I'll write additional articles on the subject as soon as I have enough subject material ready.