So, I added a little bit of magic to my /etc/rc.local:

## wifi
if [ "$1" = "" -o "$1" = "wifi" ]; then
  ifconfig ral0 up

  ## Reset ral0
  ifconfig ral0 -bssid -chan media autoselect nwid "" -nwkey -wpa -wpapsk

  ## If at home or at Dad's
  if [ `ifconfig ral0 scan | grep -c "00:0f:66:18:62:e0"` -gt 0 ]; then
    /sbin/ifconfig ral0 nwid MySsid wpa wpapsk `/sbin/wpa-psk MySsid MyKey` > /dev/null 2>&1
    /sbin/dhclient ral0 > /dev/null 2>&1
    echo -n ' wifi'
  else
    ifconfig ral0 down
  fi
fi

So, I took the time this afternoon to make foo2zjs work through the OpenBSD stock lpd and foomatic-rip.

In short, I copied and gunzipped the PPD file from /usr/local/share/foo2zjs/db/source/PPD into /etc/foomatic as the name of my printer. I then updated my /etc/printcap and started lpd.

My /etc/printcap:

lp|2600n|HP Color LaserJet 2600n:\
        :lp=9100@ganymede.home.erdelynet.com:\
        :af=/etc/foomatic/2600n.ppd:\
        :if=/usr/local/bin/foomatic-rip:\
        :sd=/var/spool/output:\
        :lf=/var/log/lpd-errs:\
        :sh:

I added lpd_flags="-s" to my /etc/rc.conf.local.

I ran the following commands to copy the PPD file:

$ cd /etc/foomatic
$ sudo cp /usr/local/share/foo2zjs/db/source/PPD/HP-Color_LaserJet_2600n.ppd.gz 2600n.ppd.gz
$ sudo gunzip 2600n.ppd.gz

Thanks to Darrin Chandler’s Printing on OpenBSD II page for some guidance.

So that I could build and test ports on my “production” server without installing millions of dependencies and polluting my environment, I set up a SSH-based chroot environment to build the ports.

My server has a lot of disk space for my use (1.5 TB). I have it configured to use ~1.2 TB of the available space, leaving ~300 GB available to be partitioned. I created a new 100 GB /home/chroot partition. I have the partition mounted with “rw,softdep” so that I can store and use devices and suid binaries.

I created a “/home/chroot/ports-amd64″ directory and unpacked all of the amd64 snapshot install sets (matching what is installed on the system) to that directory. Then, I changed to /home/chroot/ports-amd64/dev and ran “sudo ./MAKEDEV all” to create devices for the chroot.

Then I created a new user for the chroot. I went with too long of a username, portsamd64, but it’s clear what the purpose is. Then, I copied the following files from /etc to /home/chroot/ports-amd64/etc:

cd /etc
cp group hostname.* master.passwd mk.conf \
 motd mygate myname passwd profile profile.local \
 pwd.db resolv.conf sh_aliases spwd.db sudoers \
 /home/chroot/ports-amd64/etc/

I then ran “sudo ln -sf /usr/share/zoneinfo/EST5EDT /home/chroot/ports-amd64/etc/localtime” so the chroot has the same time as the “host” system. In my /etc/profile.local, I have the ability to override the system hostname in the PS1 prompt, so I edited the profile.local file to have “export HOSTNAME=amd64jail“. That way, I can easily distinguish between shells in the chroot and shells directly to the host. Then, I edited /home/chroot/ports-amd64/etc/group and added the portsamd64 user to the wheel group. This way, portsamd64 can use sudo within the chroot but not outside of it.

Then, I set up /etc/ssh/sshd_config with the following configuration:

Match User portsamd64
        ChrootDirectory /home/chroot/ports-amd64

and restarted sshd.

Now I was able to ssh to my chroot as portsamd64. When logged in, I see the other processes that are running on the system, but cannot affect the parent of the chroot’s filesystem. I created /var/run/dev.db by running “sudo dev_mkdb” and I created /var/run/ld.so.hints by running “sudo ldconfig /usr/lib /usr/local/lib /usr/X11R6/lib“. So, I just checked out the ports tree and built some ports.

I’m sure there are potential security problems with this configuration. I am not using this for security, but for convenience.

I installed OpenBSD 4.4-current normally on wd0a (wd0b = swap, wd0d = /tmp, wd0e = /var, wd0f = /usr and wd0g = /home).

Here is how I set the machine up with OpenBSD 4.3 installed on wd0h.

While booted to OpenBSD 4.4-current (on wd0a), I created and formatted /dev/wd0h and copied OpenBSD 4.3′s bsd.rd to it. Then I rebooted. At the boot> prompt, I typed hd0h:/bsd.rd. Then, I went through the installation as I normally would. I set partition mount points up as I normally would (d = /tmp, e = none, f = none, g = none, h = none, i = /var, j = /usr). Note that I am using the same /tmp. I also want to use the same /home, but I don’t want /home formatted, so I will add it to wd0h:/etc/fstab later. Also note that I typed ‘none’ for wd0h. Then, when I’m told that “The next step *DESTROYS* all existing data on these partitions!“, I typed ![enter] to escape to a shell and ran the following commands:

sed -e 's/wd0a/wd0h/' /tmp/filesystems > /tmp/fs.mwe
cp /tmp/fs.mwe /tmp/filesystems
exit

This tells the OpenBSD installer that my / filesystem is wd0h instead of wd0a. At this point, I typed yes and wd0h and my other new partitions were formatted. Then, I watched the install proceed normally. The bootloader was overwritten. If that’s a problem, it’s easily fixed by booting back into -current (or whatever’s on wd0a) and running /usr/mdec/installboot -v /boot /usr/mdec/biosboot wd0.

To change which installation is going to boot, mount /dev/wd0a, if necessary, and edit wd0a:/etc/boot.conf:

set device wd0h

Whenever I want to boot back into wd0a, I remount /dev/wd0a, edit my boot.conf and reboot.

After following Mark’s directions, stick the flash drive on a machine you’d like to install/upgrade OpenBSD on and boot to a USB drive. It should feel just like the process of using a CD or bsd.rd directly.

Updating the flash drive is easy as well: mount sd0a and download a new bsd.rd (as “bsd”) to it.

After quite a bit of googling, I found my solution.

A post to Mozilla’s bug tracker explained that this is a problem with GTK, not Firefox. To restore the ability to print, I had to create a new ~/.gtkrc-2.0 file and add this line:

gtk-print-backends = "file,lpr,cups"

Now I can print again. Yay!

From the 4.3 errata page:

Some kinds of IPv6 usage would leak kernel memory (in particular, this path was exercised by the named(8) patch for port randomization). Since INET6 is enabled by default, this condition affects all systems.

I have made binpatches for 4.2 and 4.3 for i386, amd64 and sparc64.

From the 4.3 errata page:

A vulnerability has been found with BIND. An attacker could use this vulnerability to poison the cache of a recursive resolving name server. (CVE-2008-1447)

I have made binpatches for 4.2 and 4.3 for i386, amd64 and sparc64.

If you put xdm_flags="" in your /etc/rc.conf.local, xdm will start automatically during the boot process. This is probably what you want nine times out of ten. But, that tenth time, it’s annoying when X starts and you didn’t want it to.

I put the following in my /etc/rc.local instead of putting xdm_flags="" in my /etc/rc.conf.local:

echo "Press any key within 10 seconds to bypass xdm..."
old_tty=`stty -g`
stty -echo -icanon min 0 time 100
_KEY=$(dd count=1 bs=1 2>/dev/null)
stty "${old_tty}"

if [ "${_KEY}" = "" ]; then
   echo "starting xdm..."
   /usr/X11R6/bin/xdm
else
   echo Bypassing xdm
fi

The first stty command waits 10 seconds for you to press any key. The value of that key is stored in ${_KEY}. If you pressed a key (${_KEY} != “”), rc.local does not start xdm and boots to a normal non-X login prompt. If you do not press a key, xdm starts and you’re greeted with the xdm login screen.

I hope that helps someone.

## OpenLDAP Server
if [ "$1" = "" -o "$1" = "openldap" ]; then
slapd_flags=”-u _openldap”
if [ "$slapd_flags" != "NO" -a -x /usr/local/libexec/slapd ]; then
      install -d -o _openldap /var/run/openldap
      /usr/local/libexec/slapd $slapd_flags
      echo -n ‘ slapd’
fi
fi

That way, I can stop OpenLDAP by running `sudo pkill slapd` and restart it with `sudo sh /etc/rc.local openldap`. The boot process is not affected because $1 == “” so, all daemons in my rc.local are started.