OpenBSD Upgrade Guide: 6.2 to 6.3

Upgrades are only supported from one release to the release immediately following it. Read through and understand this process before attempting it. For critical or physically remote machines, test it on an identical, local system first.

Start by performing the pre-upgrade steps. Next, boot from the install kernel, bsd.rd: use bootable install media, or place the 6.3 version of bsd.rd in the root of your filesystem and instruct the boot loader to boot this kernel. Once this kernel is booted, choose the (U)pgrade option and follow the prompts. Apply the configuration changes and finish up by upgrading the packages: pkg_add -u.

Alternatively, you can use the manual upgrade process.

You may wish to check the errata page or upgrade to the stable branch to get any post-release fixes.

Before rebooting into the install kernel

Get and verify bsd.rd. Download the ramdisk kernel and the cryptographically signed checksum file for your architecture.
- bsd.rd [alpha] [amd64] [arm64] [armv7] [i386] [landisk] [loongson] [luna88k] [macppc] [octeon] [sparc64]
- SHA256.sig [alpha] [amd64] [arm64] [armv7] [i386] [landisk] [loongson] [luna88k] [macppc] [octeon] [sparc64]
Verify them using signify(1):
```
  $ signify -C -p /etc/signify/openbsd-63-base.pub -x SHA256.sig bsd.rd
  Signature Verified
  bsd.rd: OK
```
Spring cleaning in /usr/share. Remove all outdated manuals. Moreover, the relinking of libraries and kernels at boot time now happens in /usr/share/relink. The old relinking directory is no longer used and can be removed.
```
  # rm -r /usr/share/man
  # rm -r /usr/share/compile
```
RFC 7217 style IPv6 addresses enabled by default. Stateless address autoconfiguration and link local IPv6 addresses historically embedded the layer 2 (ethernet mac) address in the lower 64 bits of the IPv6 address. This has various downsides and RFC 7217 specifies an alternative scheme of how to generate autoconfiguration addresses that are stable between reboots. This is enabled per default and IPv6 link local addresses will change if IPv6 is enabled on an interface. Furthermore, stateless autoconfiguration IPv6 addresses will change if autoconfiguration is enabled on an interface.
If you need the old style stateless address calculated from the layer 2 address (i.e. ethernet mac address) put -soii into the /etc/hostname.if file. See also ifconfig(8).

Configuration and syntax changes

ifconfig <if> deletetunnel option deprecated. The ifconfig(8) deletetunnel option was replaced with -tunnel. Adjust your hostname.if(5) configuration and scripts accordingly.
iked(8) and isakmpd(8): ECP group incompatibility. In the final RFC 5903, the computation for the DH shared secret changed. The iked(8) and isakmpd(8) programs were modified to follow the RFC. These changes are not backwards compatible, so if you are using ECP groups, please make sure to upgrade all peers at the same time.
New default setup for touchpads. X selects ws(4) as default driver for touchpads now. In the setup with ws, touchpad-specific input processing is done by wsmouse(4). Touchpad configuration parameters are made available in wsconsctl(8). Using the synaptics(4) input driver is still possible with a custom xorg.conf file.
vmd(8): switch configuration. vmd(8) no longer automatically creates an underlying bridge interface for virtual switches defined in vm.conf(5). This will require users to create their bridges ahead of time, e.g. using /etc/hostname.bridge0. This also requires users to modify their vm.conf(5) files to add an interface line to the switch definition, for example:
```
  switch "uplink" {
      interface bridge0
  }
```

Files to remove

Remove libpthread-stubs. With the integration of the pthread stub functions in libc, libpthread-stubs is no longer needed in xenocara. After upgrading base and packages, users may want to remove the old installed library, either via the sysclean package or by running the following commands:
```
  # cd /usr/X11R6/lib
  # rm libpthread-stubs.a \
       libpthread-stubs.so.2.0 \
       pkgconfig/pthread-stubs.pc
```

Special packages

Buildbot major update. There was a major update to Buildbot 0.9. While the buildslave (renamed to buildworker) workdir is compatible between 0.8 and 0.9, this is not the case for the buildmaster. On top of this, a 0.9 buildworker won't work with a 0.8 buildmaster.
Refer to the upgrade guide for information on how to migrate your configuration. If you want to keep running a 0.8 buildmaster to keep access to your old build logs/history, it's advised to install the 0.8 branch from pip and run it outside of packages.
memcached defaults to localhost only. memcached's documentation says, "you must not expose memcached directly to the internet, or otherwise any untrusted users". In order to provide sane defaults, the normal command line in memcached's rc script has been changed to listen on localhost only. See the pkg-readme file for information about changing this to listen to external network traffic if this is required.
neomutt binary renamed. In neomutt releases since 20171013, binaries and other files have been renamed so that they can coexist with mutt. Users may wish to add a shell alias or a symlink for mutt -> neomutt.
newsbeuter replaced with newsboat. Upgrading newsbeuter using pkg_add will result in newsboat being installed. Upon running newsboat for the first time it will prompt if it should import an existing newsbeuter configuration.
PostgreSQL major update. There was a major update to PostgreSQL 10.1. Use pg_upgrade as described in the postgresql-server pkg-readme or do a dump/restore.

Upgrade without the install kernel

This is NOT the recommended process. Use the install kernel method if at all possible!

Sometimes, you need to do an upgrade of a machine for which the normal upgrade process is not possible. The most common case is a machine in a remote location and there is no easy access to the system console.

Preparation

Place install files in a good location. Make sure you have sufficient space! Running out of space on a remote upgrade could be...unfortunate. Note that using softdeps can exaggerate the situation as deleted and overwritten files do not release their space immediately. Consider disabling the softdep mount option in /etc/fstab and rebooting before undertaking a manual upgrade. Having at least 500MB free on /usr would be recommended.
Become root. While using doas(1) before each command is generally a good practice, the command will likely be broken by the last steps, so you should become root before starting this process. It might be good to verify your access to root using a method other than doas at this point, i.e., direct login or using su(1).
Stop and/or disable any appropriate applications. During this process, all the userland applications will be replaced but may not be runnable, and strange things may happen as a result. You may also have issues with DNS resolution during the first reboot, so PF rules and NFS mounts dependent upon DNS may cause boot-up problems. There may be other applications which you wish to keep from running immediately after the upgrade, stop and disable them as well.
Install new boot blocks. This should actually be done at the end of any upgrade. If this has been neglected, then failure to do this now may break serial console or other things, depending on your platform. Use installboot(8), assuming sd0 is your boot disk:
```
    installboot sd0
```

Upgrading manually

Install new kernels. The extra steps for copying over the primary kernel are done to ensure that there is always a valid kernel on the disk.

If using the multiprocessor kernel:

    cd /usr/rel    # where you put the release files
    ln -f /bsd /obsd && cp bsd.mp /nbsd && mv /nbsd /bsd
    cp bsd.rd /
    cp bsd /bsd.sp

If using the single processor kernel:

    cd /usr/rel    # where you put the release files
    ln -f /bsd /obsd && cp bsd /nbsd && mv /nbsd /bsd
    cp bsd.rd bsd.mp /    # may give a harmless warning

Enable KARL. Store the kernel's checksum:

    sha256 -h /var/db/kernel.SHA256 /bsd

Install new userland. Save a copy of reboot(8), extract and install the release tarballs, reboot. Install base63.tgz last, because the new base system, in particular tar(1), gzip(1) and reboot(8), will not work with the old kernel. Either untar the needed filesets manually

    cp /sbin/reboot /sbin/oreboot
    tar -C / -xzphf xshare63.tgz
    tar -C / -xzphf xserv63.tgz
    tar -C / -xzphf xfont63.tgz
    tar -C / -xzphf xbase63.tgz
    tar -C / -xzphf man63.tgz
    tar -C / -xzphf game63.tgz
    tar -C / -xzphf comp63.tgz
    tar -C / -xzphf base63.tgz    # Install last!
    /sbin/oreboot

or, if you use ksh(1), you can do

    cp /sbin/reboot /sbin/oreboot
    for _f in [!b]*63.tgz base63.tgz; do tar -C / -xzphf "$_f" || break; done
    /sbin/oreboot

Note that tar(1) can expand only one archive per invocation, so a simple glob won't work.

After reboot, update /dev. Run MAKEDEV(8):
```
    cd /dev
    ./MAKEDEV all
```
Update boot loader. Still assuming sd0 is your boot disk:
```
    installboot sd0
```
Update system configuration files. Run sysmerge(8):
```
    sysmerge
```
Update firmware. There may be new firmware for your system. Update it with fw_update(1):
```
    fw_update
```
Finish up. Review the console output from boot (using dmesg -s) and correct any failures as necessary. All the steps following configuration changes above also apply to manual upgrades. Finally, remove /sbin/oreboot and update packages: pkg_add -u. Reboot once more to make sure you run on your own kernel generated by KARL.

[FAQ Index] | [6.1 -> 6.2] [6.3 -> 6.4]

$OpenBSD: upgrade63.html,v 1.16 2019/10/04 10:15:36 fcambus Exp $