4 - OpenBSD 4.1 Installation Guide

Table of Contents
4.1 - Overview of the OpenBSD installation procedure
4.2 - Pre-installation checklist
4.3 - Creating bootable OpenBSD install media
4.3.1 - Creating floppies on Unix
4.3.2 - Creating floppies on Windows or DOS
4.3.3 - Creating a boot CD
4.4 - Booting OpenBSD install media
4.5 - Performing an install
4.5.1 - Starting the install
4.5.2 - Setting up disks
4.5.3 - Setting the system hostname
4.5.4 - Configuring the network
4.5.5 - Choosing installation media
4.5.6 - Choosing file sets
4.5.7 - Finishing up
4.6 - What files are needed for installation?
4.7 - How much space do I need for an OpenBSD installation?
4.8 - Multibooting OpenBSD/i386
4.9 - Sending your dmesg to dmesg@openbsd.org after the install
4.10 - Adding a file set after install
4.11 - What is 'bsd.rd'?
4.12 - Common installation problems
4.12.1 - My Compaq only recognizes 16M RAM
4.12.2 - My i386 won't boot after install
4.12.3 - My machine booted, but hung at the ssh-keygen process
4.12.4 - I got the message "Failed to change directory" when doing an install
4.12.5 - My fdisk partition table is trashed or blank!
4.13 - Customizing the install process
4.14 - How can I install a number of similar systems?
4.15 - How can I get a dmesg(8) to report an install problem?
4.1 - Overview of the OpenBSD installation procedure
OpenBSD has a robust and adaptable text-based installation procedure, and can be installed from a single floppy disk. Most platforms follow a similar installation procedure; however there are some differences in the details. In all cases, you are urged to read the platform-specific INSTALL document in the platform directory on the CD-ROM or FTP sites (for example, i386/INSTALL.i386, mac68k/INSTALL.mac68k or sparc/INSTALL.sparc).
The OpenBSD installation process uses a special kernel with a number of utilities and install scripts embedded in a preloaded RAM disk. After this kernel is booted, the operating system is extracted from a number of compressed tar(1) (.tgz) files from a source other than this preloaded RAM disk. There are several ways to boot this install kernel:
Floppy disk: Floppy disk images are provided which can be used to create an install floppy on another Unix-like system, or on a DOS/Windows system. Typical file names are floppy41.fs, though several platforms have multiple floppy images available.
CD-ROM: On several platforms a CD-ROM image (cd41.iso) is provided allowing creation of a bootable CD-ROM. This just contains the installation kernel - install files must still be retrieved via FTP or other source. You can, of course, build your own CD-ROM with whatever files and tools you desire.
Existing partition: The RAM disk kernel can be booted off an already existing partition for an upgrade or reinstall.
Network: Some platforms support booting over a network (for example using PXE or other network boot).
Writing a file system image to disk (miniroot): a filesystem image that can be written to an existing partition, and then can be booted.
Bootable Tape: Some platforms support booting from tape. These tapes can be made following the INSTALL.platform instructions.
Not every platform supports all boot options:
alpha: Floppy, CD-ROM, network, writing a floppy image to hard disk.
amd64: Floppy, CD-ROM, network.
armish: Varies by machine.
hp300: CD-ROM, network.
hppa: Network.
i386: Floppy, CD-ROM, network.
landisk: miniroot, installed using another computer.
mac68k: Booted using utilities running on Mac OS. See INSTALL.mac68k for details.
macppc: CD-ROM, network.
mvme68k: Network, bootable tape.
mvme88k: Network, bootable tape.
sparc: Floppy, CD-ROM, network, writing image to existing swap partition, bootable tape.
sparc64: Floppy (U1/U2 only), CD-ROM, network, writing image to existing partition.
vax: Floppy, network.
zaurus: Boot bsd.rd from Linux partition. See INSTALL.zaurus for details.
All platforms can also use a bsd.rd to reinstall or upgrade.
Once the install kernel is booted, you have several options of where to get the install file sets. Again, not every platform supports every option.
CD-ROM: Of course, we prefer you use the Official CD-ROM set, but for special needs, you can also make your own.
FTP: Either one of the OpenBSD FTP mirror sites or your own local FTP server holding the file sets.
HTTP: Either one of the OpenBSD HTTP mirror sites or your own local web server holding the file sets.
Local disk partition: In many cases, you can install file sets from another partition on a local hard disk. For example, on i386, you can install from a FAT partition or a CD-ROM formatted in ISO9660, Rock Ridge or Joliet format. In some cases, you will have to manually mount the file system before using it.
NFS: Some platforms support using NFS mounts for the file sets.
Tape: File sets can also be read from a supported tape. Details on creating the tape are in the INSTALL.platform document.
4.2 - Pre-installation checklist
Before you start your install, you should have some idea what you want to end up with. You will want to know the following items, at least:
Machine name
Hardware installed and available
Verify compatibility with your platform's hardware compatibility page
If ISA, you also need to know hardware settings, and confirm they are as OpenBSD requires.
Install method to be used (CD-ROM, FTP, etc.)
Should an important bug be found, how will the system be patched?
If done locally, you will need to have sufficient space available for the source tree and building it.
Otherwise, you will need access to another machine to build a patched release on.
Desired disk layout
Does existing data need to be saved elsewhere?
Will OpenBSD coexist on this system with another OS? If so, how both will be booted? Will you need to install a "boot manager"?
Will the entire disk be used for OpenBSD, or do you want to keep an existing partition/OS (or space for a future one)?
How do you wish to sub-partition the OpenBSD part of your disk?
Network settings, if not using DHCP:
Domain name
Domain Name Server(s) (DNS) address
IP addresses and subnet masks for each NIC
Gateway address
Will you be running the X Window System?
4.3 - Creating bootable OpenBSD install media
As examples, we will look at the installation images available for the i386 and sparc platforms.
The i386 platform has six separate installation disk images to choose from:
floppy41.fs (Desktop PC) supports many PCI and ISA NICs, IDE and simple SCSI adapters and some PCMCIA support. Most users will use this image if booting from a floppy
floppyB41.fs (Servers) supports many RAID controllers, and some of the less common SCSI adapters. However, support for many standard SCSI adapters and many EISA and ISA NICS has been removed.
floppyC41.fs (Laptops) supports the CardBus and PCMCIA devices found in many laptops.
cdrom41.fs is, in effect a combination of all three boot disks. It can be used to make a bootable 2.88M floppy, or more commonly, as a boot image for a custom recordable CD.
cd41.iso is an ISO9660 image that can be used to create a bootable CD with most popular CD-ROM creation software on most platforms. This image has the widest selection of drivers, and is usually the recommended choice if your hardware can boot from a CDROM.
cdemu41.iso is an ISO9660 image, using "floppy emulation" booting, using the 2.88M image, cdrom41.fs. It is hoped that few people will need this image -- most people will use cd41.iso, only use cdemu41.iso if cd41.iso doesn't work for you.
Yes, there may be situations where one install disk is required to support your SCSI adapter and another disk is required to support your network adapter. Fortunately, this is a rare event, and can usually be worked around.
The sparc platform has three separate installation disk images to choose from:
floppy41.fs: Supports systems with a floppy disk.
cd41.iso An ISO image usable to make your own CD for booting SPARC systems with a CD-ROM.
miniroot41.fs Can be written to a swap partition and booted.
4.3.1 - Creating floppies on Unix
To create a formatted floppy, use the fdformat(1) command to both format and check for bad sectors.
# fdformat /dev/rfd0c
Format 1440K floppy `/dev/rfd0c'? (y/n): y
If your output is like the above example, then the disk is OK. However, if you do not see ALL "V"'s then the disk is most likely bad, and you should try a new one.
Note that some Unix-like systems have different commands for formatting floppies. Refer to your system's documentation for the exact procedure.
Once you have a clean, formatted floppy it is time to write the installation image to floppy. For this, you can use the dd(1) utility. An example usage of dd(1) is below:
# dd if=floppy41.fs of=/dev/rfd0c bs=32k
Once the image is written, check to make sure that the copied image is the same as the original with the cmp(1) command. If the diskette is identical to the image, you will just see another prompt.
# cmp /dev/rfd0c floppy41.fs
4.3.2 - Creating floppies on Windows or DOS
This section describes how to write the installation images to floppy disk under Windows or DOS. You can get the tools mentioned below from the tools directory on any of the FTP mirrors, or from the 4.1/tools directory on CD1 of the OpenBSD CD set.
To prepare a floppy in MS-DOS or Windows, first use the native formatting tools to format the disk.
To write the installation image to the prepared floppy you can use rawrite, fdimage, or ntrw. rawrite will not work on Windows NT, 2000 or XP.
Note that FDIMAGE.EXE and RAWRITE.EXE are both MS-DOS applications, and thus are limited to MS-DOS's "8.3" file naming convention. As floppyB41.fs and floppyC41.fs have longer file names, you will have to find out how your system stored the file in "8.3 format" before using FDIMAGE.EXE or RAWRITE.EXE to make your boot floppies.
Example usage of rawrite:
C:> rawrite
RaWrite 1.2 - Write disk file to raw floppy diskette

Enter source file name: floppy41.fs
Enter destination drive: a
Please insert a formatted diskette into drive A: and press -ENTER- : Enter
Example usage of fdimage:
C:> fdimage -q floppy41.fs a:
Example usage of ntrw:
C:> ntrw floppy41.fs a:
3.5", 1.44MB, 512 bytes/sector
bufsize is 9216
1474560 bytes written
4.3.3 - Making a CD-ROM
You can create a CD-ROM using either the cd41.iso file or, in the case of the i386 and amd64 platforms, you can also use the cdrom41.fs as the bootable floppy image that is used to boot an i386 system from CD-ROM. The exact details here are left to the reader to determine with the tools they have at their disposal.
Some of the tools in OpenBSD are:
cdrecord, part of the cdrtools collection in the OpenBSD Packages and Ports System.
cdio(1)'s "track at once" (tao) recording option.
4.4 - Booting OpenBSD install media
Booting i386/amd64
Booting an install image on the i386 and amd64 PC platforms is nothing new to most people. If you are using a floppy disk, simply insert the floppy into the floppy drive and boot the system. The install image will then load, provided floppy boot is enabled in your BIOS. If you want to boot from CD, you must go into your system's BIOS and set the boot options to allow booting from CD. Some older BIOSes do not have this option, and you must use a floppy for booting your installation image. Don't worry though; even if you boot from floppy you can still install from the CD.
You can also install by booting bsd.rd from an existing OpenBSD partition, or over the network using the PXE boot process.
Booting sparc/sparc64
NOTE: On the sparc64 platform, only the SBus machines (Ultra 1, Ultra 2) are bootable from floppy.
To boot from floppy, place the floppy disk with the OpenBSD installation image on it into the floppy drive. Then use the following command to boot from the floppy:
ok boot floppy
To boot from CD-ROM, place the OpenBSD CD-ROM disk into the drive. If your Sun only has one CD-ROM drive, then just go to the boot prompt, where you can 'boot cdrom':
ok boot cdrom
Of course, this will only work in new command mode. If you are at the old command mode prompt (a right arrow), type 'n' for the new command mode. (If you are using an old sparc that is pre-sun4c, you probably don't have a new command mode. In this case, you need to experiment.) If you have multiple CD-ROM devices, you need to boot from the correct one. Try probe-scsi from the new command mode.
ok probe-scsi

Target 0
Target 1
Unit 0 Removable Disk QUANTUM EMPIRE_1080S
Target 3
Unit 0 Removable Disk Joe's CD-ROM
Figure out which disk is the CD-ROM you want to boot from. Note the target number.
ok boot /sbus/esp/sd@X,0
4.5 - Performing an install
4.5.1 - Starting the install
Whatever your means of booting is, it is now time to use it. During the boot process, the kernel and all of the programs used to install OpenBSD are loaded into memory. The most common problem when booting is a bad floppy disk or a drive alignment problem. The boot floppy is quite tightly packed -- any bad spot will cause problems.
At almost any point during the OpenBSD install process, you can terminate the current install attempt by hitting CTRL-C and can restart it without rebooting by running install at the shell prompt.
When your boot is successful, you will see a lot of text messages scroll by. This text, on many architectures in white on blue, is the dmesg, the kernel telling you what devices have been found, and where. Don't worry about remembering this text, as a copy is saved as /var/run/dmesg.boot.
Then, you will see the following:
rootdev=0x1100 rrootdev=0x2f00 rawdev=0x2f02
erase ^?, werase ^W, kill ^U, intr ^C, status ^T
(I)nstall, (U)pgrade or (S)hell? i
And with that, we reach our first question. Most of the time, you have the three options shown:
Install: load OpenBSD onto the system, overwriting whatever may have been there. Note that it is possible to leave some partitions untouched in this process, such as a /home, but otherwise, assume everything else is overwritten.
Upgrade: Install a new set of install files on this machine, but do not overwrite any configuration information, user data, or additional programs. No disk formatting is done, nor are the /etc or /var directories overwritten. A few important notes:
You will not be given the option of installing the etc41.tgz file. After the install, you will have to manually merge the changes of etc41.tgz into your system before you can expect it to be fully functional. This is an important step which must be done, as otherwise certain key services (such as pf(4)) may not start.
The Upgrade process is not designed to skip releases! While this will often work, it is not supported. For OpenBSD 4.1, upgrading 4.0 to 4.1 is the only supported upgrade. If you have to upgrade from an older version, upgrade to intermediate versions first, or if the system is very out-of-date, consider a complete reinstall.
More information on upgrading between releases can be found here.
Shell: Sometimes, you need to perform repairs or maintenance to a system which will not (or should not) boot to a normal kernel. This option will allow you to do maintenance to the system. A number of important utilities are available on the boot media.
On occasion, you will not see the "Upgrade" option listed. After a flag day event, it is not possible to directly upgrade; one must reinstall the system from scratch.
In this example, we will do an install, but the upgrade process is similar.
Welcome to the OpenBSD/i386 4.1 install program.

This program will help you install OpenBSD. At any prompt except password
prompts you can escape to a shell by typing '!'. Default answers are shown
in []'s and are selected by pressing RETURN. At any time you can exit this
program by pressing Control-C, but exiting during an install can leave your
system in an inconsistent state.

Terminal type: [vt220] Enter
kbd(8) mapping? ('L' for list) [none] Enter
In most cases, the default terminal type is appropriate; however if you are using a serial console for install, don't just take the default, respond appropriately.
If you do not select a keyboard encoding table, a US keyboard layout will be assumed.
IS YOUR DATA BACKED UP? As with anything that modifies disk contents, this
program can cause SIGNIFICANT data loss.

It is often helpful to have the installation notes handy. For complex disk
configurations, relevant disk hardware manuals and a calculator are useful.

Proceed with install? [no] y
If you take the default here, the install process will terminate and drop you to a shell prompt.
The installation notes referred to here are on the install CDs and FTP servers, in the file INSTALL.<plat>, where <plat> is your platform, for instance, i386.
4.5.2 - Setting up disks
Setting up disks in OpenBSD varies a bit between platforms. For i386, amd64, macppc, zaurus and armish, disk setup is done in two stages. First, the OpenBSD slice of the hard disk is defined using fdisk(8), then that slice is subdivided into OpenBSD partitions using disklabel(8).
Some users may be a little confused by the terminology used here. It will appear we are using the word "partition" in two different ways. This observation is correct. There are two layers of partitioning in the above OpenBSD platforms, the first, one could consider the Operating System partitioning, which is how multiple OSs on one computer mark out their own space on the disk, and the second one is how the OpenBSD partition is sub-partitioned into individual filesystems. The first layer is visible as a disk partition to DOS, Windows, and any other OS that uses this disk layout system, the second layer of partitioning is visible only to OpenBSD and those OSs which can directly read an OpenBSD filesystem.
Cool! Let's get to it...

You will now initialize the disk(s) that OpenBSD will use. To enable all
available security features you should configure the disk(s) to allow the
creation of separate filesystems for /, /tmp, /var, /usr, and /home.

Available disks are: wd0.
Which one is the root disk? (or done) [wd0] Enter
The root disk is the disk the system will boot from, and normally where swap space resides. IDE disks will show up as wd0, wd1, etc., SCSI disks and RAID devices will show up as sd0, sd1, and so on. All the disks OpenBSD can find are listed here -- if you have drives which are not showing up, you have unsupported or improperly configured hardware.
Do you want to use *all* of wd0 for OpenBSD? [no] Enter
If you say "yes" to this question, the entire disk will be allocated to OpenBSD. This will result in a standard Master Boot Record and partition table being written out to disk -- one partition, the size of the entire hard disk, set to the OpenBSD partition type, and flagged as the bootable partition. This will be a common choice for most production uses of OpenBSD; however, there are some systems this should not be done on. Many Compaq systems, many laptops, some Dell and other systems use a "maintenance" or "Suspend to Disk" partition, which should be kept intact. If your system has any other partitions of any type you do not wish to erase, do not select "yes" to the above question. On the other hand, if your system has a brand new disk that has never been used, you will probably want to say "yes" here (or use the "update" option of fdisk), so you do get a valid master boot record and signature in place.
For the sake of this example, we will assume the disk is to be split between OpenBSD and a pre-existing Windows 2000 partition, so we take the default of "no", which will take us into the fdisk(8) program. You can also get more information on fdisk(8) here.
Important Note: Users with a large hard disk (larger than was commonly available when your computer was made) will want to see this section before going any further.
You will now create a single MBR partition to contain your OpenBSD data. This
partition must have an id of 'A6'; must *NOT* overlap other partitions; and
must be marked as the only active partition.

The 'manual' command describes all the fdisk commands in detail.

Disk: wd0 geometry: 2586/240/63 [39100320 Sectors]
Offset: 0 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
*0: 0B 0 1 1 - 202 239 63 [ 63: 3069297 ] Win95 FAT-32
1: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
2: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
3: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
Enter 'help' for information
fdisk: 1> help
help Command help list
manual Show entire OpenBSD man page for fdisk
reinit Re-initialize loaded MBR (to defaults)
setpid Set the identifier of a given table entry
disk Edit current drive stats
edit Edit given table entry
flag Flag given table entry as bootable
update Update machine code in loaded MBR
select Select extended partition table entry MBR
swap Swap two partition entries
print Print loaded MBR partition table
write Write loaded MBR to disk
exit Exit edit of current MBR, without saving changes
quit Quit edit of current MBR, saving current changes
abort Abort program without saving current changes
fdisk: 1>
A few commands are worthy of elaboration:
r or reinit: Clears existing partition table, makes one big OpenBSD partition, flags it active, and installs the OpenBSD MBR code. Equivalent to saying "yes" to the "use *all* of ..." question.
p or print: Displays the current partition table in sectors. "p m" will show the partition table in megabytes, "p g" will show it in gigabytes.
e or edit: edit or alter a table entry.
f or flag: Marks a partition as the active partition, the one that will be booted from.
u or update: Updates the MBR with the OpenBSD boot code, similar to "reinit", except it doesn't alter the existing partition table.
exit and quit: Careful on these, as some users are used to "exit" and "quit" having opposite meanings.
It is worth pointing out once again, an error here will result in significant data loss. If you are going to do this on a drive with important data, it might be worth practicing on a "disposable" drive, in addition to having a good backup.
Our drive here has a 1.5G partition for Windows 2000 (using the FAT filesystem). Looking at the info from the above display, we can see that the Windows partition occupies through cylinder 202 on the drive. So, we are going to allocate the rest of the disk to OpenBSD, starting at cylinder 203. You could also calculate OpenBSD's starting sector of 3069360 by adding the existing partition's starting sector (63) and its size (3069297).
You can edit the drive layout in either Cylinder/Heads/Sectors form or just raw sectors. Which is easier depends upon what you are doing; in this case, working around an existing partition, using CHS format will probably be easier. If you are creating the first partition on the disk, just using raw sectors may be easier.
fdisk: 1> e 1
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
1: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
Partition id ('0' to disable) [0 - FF]: [0] (? for help) a6
Do you wish to edit in CHS mode? [n] y
BIOS Starting cylinder [0 - 2585]: [0] 203
BIOS Starting head [0 - 239]: [0] Enter
BIOS Starting sector [1 - 63]: [0] 1
BIOS Ending cylinder [0 - 2585]: [0] 2585
BIOS Ending head [0 - 239]: [0] 239
BIOS Ending sector [1 - 63]: [0] 63
fdisk:*1> p
Disk: wd0 geometry: 2586/240/63 [39100320 Sectors]
Offset: 0 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
*0: 0B 0 1 1 - 202 239 63 [ 63: 3069297 ] Win95 FAT-32
1: A6 203 0 1 - 2585 239 63 [ 3069360: 36030960 ] OpenBSD
2: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
3: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
fdisk:*1> p m
Disk: wd0 geometry: 2586/240/63 [19092 Megabytes]
Offset: 0 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
*0: 0B 0 1 1 - 202 239 63 [ 63: 1499M] Win95 FAT-32
1: A6 203 0 1 - 2585 239 63 [ 3069360: 17593M] OpenBSD
2: 00 0 0 0 - 0 0 0 [ 0: 0M] unused
3: 00 0 0 0 - 0 0 0 [ 0: 0M] unused
On platforms which use fdisk, it is important that the first partition skips the first track of the disk, in this case, starting on sector 63. This will vary from machine to machine and disk system to disk system. If an OpenBSD partition is created starting at offset 0, this partition table will end up being overwritten by the OpenBSD partition's Partition Boot Record. The system may still be bootable, but it will be very difficult to maintain, and this configuration is not recommended or supported.
Note that the prompt changed to include an asterisk ('*') to indicate you have unsaved changes. As we can see from the output of p m we have not altered our Windows partition, we have successfully allocated the rest of the drive for OpenBSD, and the partitions do not overlap. We are in business. Almost.
What we haven't done is flagged the partition as active so the machine will boot OpenBSD on the next reboot:
fdisk:*1> f 1
Partition 1 marked active.
fdisk:*1> p
Disk: wd0 geometry: 2586/240/63 [39100320 Sectors]
Offset: 0 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
0: 0B 0 1 1 - 202 239 63 [ 63: 3069297 ] Win95 FAT-32
*1: A6 203 0 1 - 2585 239 63 [ 3069360: 36030960 ] OpenBSD
2: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
3: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
And now, we are ready to save our changes:
fdisk:*1> w
Writing MBR at offset 0.
wd0: no disk label
fdisk: 1> q
Creating a disklabel
The next step is to use disklabel(8) to slice up the OpenBSD partition. More details on using disklabel(8) can be found in FAQ 14, disklabel.
Here is the partition information you chose:

Disk: wd0 geometry: 2586/240/63 [39100320 Sectors]
Offset: 0 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
0: 0B 0 1 1 - 202 239 63 [ 63: 3069297 ] Win95 FAT-32
*1: A6 203 0 1 - 2585 239 63 [ 3069360: 36030960 ] OpenBSD
2: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused
3: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused

You will now create an OpenBSD disklabel inside the OpenBSD MBR
partition. The disklabel defines how OpenBSD splits up the MBR partition
into OpenBSD partitions in which filesystems and swap space are created.

The offsets used in the disklabel are ABSOLUTE, i.e. relative to the
start of the disk, NOT the start of the OpenBSD MBR partition.

disklabel: no disk label
WARNING: Disk wd0 has no label. You will be creating a new one.

# using MBR partition 1: type A6 off 3069360 (0x2ed5b0) size 36030960 (0x225c9f0)

Treating sectors 3069360-39100320 as the OpenBSD portion of the disk.
You can use the 'b' command to change this.

Initial label editor (enter '?' for help at any prompt)
> ?
Available commands:
? [cmnd] - this message or command specific help.
a [part] - add new partition.
b - set OpenBSD disk boundaries.
c [part] - change partition size.
D - set label to default.
d [part] - delete partition.
e - edit drive parameters.
g [b|d|u] - use [b]ios, [d]isk or [u]ser geometry.
M - show entire OpenBSD man page for disklabel.
m [part] - modify existing partition.
n [part] - set the mount point for a partition.
p [unit] - print label.
q - quit and save changes.
r - recalculate free space.
s [path] - save label to file.
u - undo last change.
w - write label to disk.
X - toggle expert mode.
x - exit without saving changes.
z - zero out partition table.
Numeric parameters may use suffixes to indicate units:
'b' for bytes, 'c' for cylinders, 'k' for kilobytes, 'm' for megabytes,
'g' for gigabytes or no suffix for sectors (usually 512 bytes).
'%' for percent of total disk size, '&' for percent of free space.
Non-sector units will be rounded to the nearest cylinder.
Entering '?' at most prompts will give you (simple) context sensitive help.
Again, a few of these commands could use a little elaboration:
p - displays (prints) the current disklabel to the screen, and you can use the modifiers k, m or g for kilobytes, megabytes or gigabytes.
D - Clears any existing disklabel, creates a new default disklabel which covers just the current OpenBSD partition. This can be useful if the disk previously had a disklabel on it, and the OpenBSD partition was recreated to a different size -- the old disk label may not get deleted, and may cause confusion.
m - Modifies an existing entry in a disklabel. Do not over estimate what this will do for you. While it may alter the size of a disklabel partition, it will NOT alter the filesystem on the drive. Using this option and expecting it to resize existing partitions is a good way of losing large amounts of data.
Slicing up your disk properly is important. The answer to the question, "How should I partition my system?" is "Exactly how you need it". This will vary from application to application. There is no universal answer. If you are unsure of how you want to partition your system, see this discussion.
In this system, we have over 17G available for OpenBSD. That's a lot of space, and it isn't likely we will need most of it. So, we will deliberately not use absolute minimum sizes. We would rather have a few hundred megabytes of unused space than a kilobyte too little.
On the root disk, the two partitions 'a' and 'b' must be created. The installation process will not proceed until these two partitions are available. 'a' will be used for the root filesystem (/) and 'b' will be used as swap space.
After a little thought, we decide to create just enough partitions to allow the creation of the recommended separate filesystems (/, /tmp, /var, /usr, /home) along with a swap partition:
wd0a: / (root) - 150M. Should be more than enough.
wd0b: (swap) - 300M.
wd0d: /tmp - 120M. /tmp is used for building some software, 120M will probably be enough for most things.
wd0e: /var - 80M. If this were to be a web or mail server, we'd have made this partition much larger, but, that's not what we are doing.
wd0g: /usr - 6G. We want this partition to be large enough to load a few user applications, plus be able to update and rebuild the system by source if desired or needed. The Ports tree will be here as well, which will take almost 160M of this space before ports are built. If one was planning on building many applications from source using ports rather than pre-built packages, you might want a lot more space here.
wd0h: /home - 4G. This will allow plenty of user file space.
Now, if you add those up, you will see over 6G of space is unused! Unused space won't hurt anything, and it gives us flexibility to enlarge things in the future if need be. Need more /tmp? Create a new partition in the unused space, format the new partition with newfs(8), and change /etc/fstab to mount the new partition onto /tmp. Problem solved.
> p m
device: /dev/rwd0c
type: ESDI
disk: ESDI/IDE disk
label: ST320011A
bytes/sector: 512
sectors/track: 63
tracks/cylinder: 16
sectors/cylinder: 1008
cylinders: 16383
total sectors: 39102336
free sectors: 36030960
rpm: 3600

16 partitions:
# size offset fstype [fsize bsize cpg]
a: 17593.2M 1498.7M unused 0 0
c: 19092.9M 0.0M unused 0 0
i: 1498.7M 0.0M MSDOS
> d a
> a a
offset: [3069360] Enter
size: [36030960] 150m
Rounding to nearest cylinder: 307440
FS type: [4.2BSD] Enter
mount point: [none] /
> a b
offset: [3376800] Enter
size: [35723520] 300m
Rounding to nearest cylinder: 614880
FS type: [swap] Enter
> a d
offset: [3991680] Enter
size: [35108640] 120m
Rounding to nearest cylinder: 245952
FS type: [4.2BSD] Enter
mount point: [none] /tmp
> a e
offset: [4237632] Enter
size: [34862688] 80m
Rounding to nearest cylinder: 164304
FS type: [4.2BSD] Enter
mount point: [none] /var
> a g
offset: [4401936] Enter
size: [34698384] 6g
Rounding to nearest cylinder: 12582864
FS type: [4.2BSD] Enter
mount point: [none] /usr
> a h
offset: [16984800] Enter
size: [22115520] 4g
Rounding to nearest cylinder: 8388576
FS type: [4.2BSD] Enter
mount point: [none] /home
> p m
device: /dev/rwd0c
type: ESDI
disk: ESDI/IDE disk
label: ST320011A
bytes/sector: 512
sectors/track: 63
tracks/cylinder: 16
sectors/cylinder: 1008
cylinders: 16383
total sectors: 39102336
free sectors: 22115520
rpm: 3600

16 partitions:
# size offset fstype [fsize bsize cpg]
a: 150.1M 1498.7M 4.2BSD 2048 16384 16 # /
b: 300.2M 1648.8M swap
c: 19092.9M 0.0M unused 0 0
d: 120.1M 1949.1M 4.2BSD 2048 16384 16 # /tmp
e: 80.2M 2069.2M 4.2BSD 2048 16384 16 # /var
g: 6144.0M 2149.4M 4.2BSD 2048 16384 16 # /usr
h: 4096.0M 8293.4M 4.2BSD 2048 16384 16 # /home
i: 1498.7M 0.0M MSDOS
> q
Write new label?: [y] Enter
You will note there is a c partition we seem to have ignored. This partition is your entire hard disk; don't attempt to alter it. You will also note the i partition wasn't defined by us; this is the pre-existing Windows 2000 partition. Partitions are not assigned any particular letters -- with the exception of a (root), b (swap) and c (entire disk), the rest of the partitions (through letter p) are available for use as you desire.
If you look closely at the output of the disklabel, you will note that your drive RPM rating is probably wrong. This is historical; the drive speed is not used in any way by the system. Do not worry about it.
Configuring your mount points and formatting your filesystems
Now comes the final configuration of your mount points. If you configured the mount points through disklabel(8), this step consists of just verifying your selections; otherwise, you can specify them now.
Mount point for wd0d (size=122976k)? (or 'none' or 'done') [/tmp] Enter
Mount point for wd0e (size=82152k)? (or 'none' or 'done') [/var] Enter
Mount point for wd0g (size=6291432k)? (or 'none' or 'done') [/usr] Enter
Mount point for wd0h (size=4194288k)? (or 'none' or 'done') [/home] Enter
Mount point for wd0d (size=122976k)? (or 'none' or 'done') [/tmp] done
No more disks to initialize.

OpenBSD filesystems:
wd0a /
wd0d /tmp
wd0e /var
wd0g /usr
wd0h /home

The next step *DESTROYS* all existing data on these partitions!
Are you really sure that you're ready to proceed? [no] y
/dev/rwd0a: 307440 sectors in 305 cylinders of 16 tracks, 63 sectors
150.1MB in 1 cyl groups (306 c/g, 150.61MB/g, 19328 i/g)
/dev/rwd0d: 245952 sectors in 244 cylinders of 16 tracks, 63 sectors
120.1MB in 1 cyl groups (244 c/g, 120.09MB/g, 15360 i/g)
/dev/rwd0e: 164304 sectors in 163 cylinders of 16 tracks, 63 sectors
80.2MB in 1 cyl groups (164 c/g, 80.72MB/g, 10368 i/g)
/dev/rwd0g: 12582864 sectors in 12483 cylinders of 16 tracks, 63 sectors
6144.0MB in 39 cyl groups (328 c/g, 161.44MB/g, 20608 i/g)
/dev/rwd0h: 8388576 sectors in 8322 cylinders of 16 tracks, 63 sectors
4096.0MB in 26 cyl groups (328 c/g, 161.44MB/g, 20608 i/g)
/dev/wd0a on /mnt type ffs (rw, asynchronous, local, ctime=Sat Oct 7 19:49:44 2
/dev/wd0h on /mnt/home type ffs (rw, asynchronous, local, nodev, nosuid, ctime=T
hu Apr 19 00:54:25 2007)
/dev/wd0d on /mnt/tmp type ffs (rw, asynchronous, local, nodev, nosuid, ctime=Th
u Apr 19 00:54:25 2007)
/dev/wd0g on /mnt/usr type ffs (rw, asynchronous, local, nodev, ctime=Thu Apr 19
00:54:25 2007)
/dev/wd0e on /mnt/var type ffs (rw, asynchronous, local, nodev, nosuid, ctime=Th
u Apr 19 00:54:25 2007)
You may wonder why the installer again asks for mount points. This allows you to recover from any errors or omissions in the mount points specified during the creation of the disklabel. For instance, the installation process will automatically delete any duplicate mount points you enter during the configuration of the disklabel. The disklabel program will allow you to enter such duplicates, and thus they must be checked for after the disklabel program exits. The deleted duplicate mount points will result in partitions without mount points, that you must assign new mount points for if you wish to use the space.
Notice the "Are you really sure that you are ready to proceed?" question defaults to no, so you will have to deliberately tell it to proceed and format your partitions. If you chose no, you would simply be dropped into a shell and could start the install again by typing "install", or just by rebooting again with your boot disk.
At this point all filesystems will be formatted for you. This could take some time depending on the size of the partitions and the speed of the disk.
4.5.3 - Setting the system hostname
Now you must set the system hostname. This value, along with the DNS domain name (specified below), will be saved in the file /etc/myname, which is used during normal boot to set the hostname of the system. If you do not set the domain name of the system, the default value of 'my.domain' will be used.
It is important to set this name now, because it will be used when the cryptographic keys for the system are generated during the first boot after installation. This generation takes place whether the network is configured or not.
System hostname (short form, e.g. 'foo'): puffy
4.5.4 - Configuring the network
Now it is time to configure your network. The network must be configured if you are planning on doing an FTP or NFS based install, considering it will be based upon the information you are about to enter. Here is a walk through of the network configuration section of the install process. In our example, we will attach one interface (fxp0) to a cable modem, which will be configured using DHCP, the other will be to our internal network, and configured statically.
Configure the network? [yes] Enter
Available interfaces are: fxp0 xl0.
Which one do you wish to initialize? (or 'done') [fxp0] xl0
Symbolic (host) name for xl0? [puffy] Enter
The default media for xl0 is
media: Ethernet autoselect (100baseTX full-duplex)
Do you want to change the default media? [no] Enter
IPv4 address for xl0? (or 'dhcp')
Netmask? [] Enter
IPv6 address for xl0? (or 'rtsol' or 'none') [none] Enter
Available interfaces are: fxp0.
Which one do you wish to initialize? (or 'done') [fxp0] Enter
Symbolic (host) name for fxp0? [puffy] Enter
The media options for fxp0 are currently
media: Ethernet autoselect (10baseT half-duplex)
Do you want to change the media options? [no] Enter
IPv4 address for fxp0? (or 'none' or 'dhcp') dhcp
Issuing hostname-associated DHCP request for fxp0.
DHCPDISCOVER on fxp0 to port 67 interval 1
DHCPREQUEST on fxp0 to port 67
bound to -- renewal in 1800 seconds.
IPv6 address for fxp0? (or 'rtsol' or 'none') [none] Enter
No more interfaces to initialize.
DNS domain name? (e.g. 'bar.com') [my.domain] example.com
DNS nameserver? (IP address or 'none') [] Enter
Use the nameserver now? [yes] Enter
Default route? (IP address, 'dhcp' or 'none') [dhcp] Enter
Edit hosts with ed? [no] Enter
Do you want to do any manual network configuration? [no] Enter
NOTE: Only one interface can easily be configured using DHCP during an install. If you attempt to configure more than one interface using DHCP you will encounter errors. You have to manually configure the additional interfaces after the installation.
Now, we set the password for the root account:
Password for root account? (will not echo) pAssWOrd
Password for root account? (again) pAssWOrd
Use a secure password for the root account. You will create other user accounts after the system is booted. From passwd(1):
The new password should be at least six characters long and not purely
alphabetic. Its total length must be less than _PASSWORD_LEN (currently
128 characters). A mixture of both lower and uppercase letters, numbers,
and meta-characters is encouraged.
4.5.5 - Choosing installation media
After your network is set up, the install script will give you a chance to make manual adjustments to the configuration.
Next, you will get a chance to choose your installation media. The options are listed below.
Let's install the sets!
Location of sets? (cd disk ftp http or 'done') [cd] Enter
Available CD-ROMs are: cd0.
In this example we are installing from CD-ROM. This will bring up a list of devices on your computer identified as a CD-ROM. Most people will only have one. If you need to, make sure you pick the device which you will use to install OpenBSD from.
NOTE: Not all platforms support all installation options. In this case, the OpenBSD/i386 platform does not support NFS installs, so they are not shown on this list.
Available CD-ROMs are: cd0.
Which one contains the install media? (or 'done') [cd0] Enter
Pathname to the sets? (or 'done') [4.1/i386] Enter
Here, you are prompted for which directory the installation files are, which is 4.1/i386/ on the official CD-ROM.
4.5.6 - Choosing file sets.
Now it's time to choose which file sets you will be installing. You can get a description of these files in the next section. The files that the install program finds will be shown to you on the screen. Your job is just to specify which files you want. By default all the non-X file sets are selected; however, some advanced users may wish to limit this to the bare minimum required to run OpenBSD, which would be base41.tgz, etc41.tgz and bsd. Most people will want to install the default file sets or all file sets. The example below is that of a full install.
Select sets by entering a set name, a file name pattern or 'all'. De-select
sets by prepending a '-' to the set name, file name pattern or 'all'. Selected
sets are labeled '[x]'.

[X] bsd
[X] bsd.rd
[ ] bsd.mp
[X] base41.tgz
[X] etc41.tgz
[X] misc41.tgz
[X] comp41.tgz
[X] man41.tgz
[X] game41.tgz
[ ] xbase41.tgz
[ ] xetc41.tgz
[ ] xshare41.tgz
[ ] xfont41.tgz
[ ] xserv41.tgz
Set name? (or 'done') [bsd.mp] all

[X] bsd
[X] bsd.rd
[X] bsd.mp
[X] base41.tgz
[X] etc41.tgz
[X] misc41.tgz
[X] comp41.tgz
[X] man41.tgz
[X] game41.tgz
[X] xbase41.tgz
[X] xetc41.tgz
[X] xshare41.tgz
[X] xfont41.tgz
[X] xserv41.tgz
You can do all kinds of nifty things here -- "-x*" would remove all X components, if you changed your mind. In this case, we are going to load all the sets. While the system will run with fewer sets, either the starting default or installing all sets is recommended. More details on selecting sets here.
Once you have successfully picked which file sets you want, you will be prompted to make sure you want to extract these file sets and they will then be installed. A progress bar will be shown that will keep you informed on how much time it will take. The times range greatly depending on what system it is you are installing OpenBSD on, the file sets installed, and the speed of the source media. This part may take from a few minutes to several hours.
Set name? (or 'done') [done] Enter
Ready to install sets? [yes] Enter
Getting bsd ...
100% |**************************************************| 5972 KB 00:03
Getting bsd.rd ...
100% |**************************************************| 4887 KB 00:03
Getting bsd.mp ...
100% |**************************************************| 6020 KB 00:03
Getting base41.tgz ...
100% |**************************************************| 41437 KB 00:34
Getting etc41.tgz ...
100% |**************************************************| 1210 KB 00:01
Getting misc41.tgz ...
100% |**************************************************| 2238 KB 00:02
Getting comp41.tgz ...
100% |**************************************************| 76666 KB 01:02
Getting man41.tgz ...
100% |**************************************************| 7473 KB 00:07
Getting game41.tgz ...
100% |**************************************************| 2548 KB 00:02
Getting xbase41.tgz ...
100% |**************************************************| 10344 KB 00:08
Getting xetc41.tgz ...
100% |**************************************************| 90772 00:00
Getting xshare41.tgz ...
100% |**************************************************| 2024 KB 00:03
Getting xfont41.tgz ...
100% |**************************************************| 32456 KB 00:24
Getting xserv41.tgz ...
100% |**************************************************| 19365 KB 00:15
Location of sets? (cd disk ftp http or 'done') [done] Enter
At this point, you can pull additional files from other sources (including custom file sets) if desired, or hit 'done' if you have installed all the file sets you need.
4.5.7 - Finishing up
Next, you get asked a few questions about settings for your installed system. First is whether sshd(8) should be started on boot. Usually, you will want sshd(8) running, but occasionally you may not. If your application has no need for sshd(8), there is a small theoretical security advantage to not having it running.
Start sshd(8) by default? [yes] y
(If you change your mind later, alter /etc/rc.conf.local or /etc/rc.conf.)
You will be given the option to run OpenNTPD on boot. OpenNTPD is a low-impact way to keep your computer's clock accurately synchronized, and the default configuration is sufficient for many people's use.
Start ntpd(8) by default? [no] y
(If you change your mind later, alter /etc/rc.conf.local or /etc/rc.conf.)
On some platforms, you will now be asked if you plan to run X on this system. If you answer 'Y', /etc/sysctl.conf will be modified to include the line machdep.allowaperture=1 or machdep.allowaperture=2, depending on your platform. Some platforms will not ask this question at all. If you do not intend to run X on this system or are not sure, answer 'N' here, as you can easily change it by editing /etc/sysctl.conf should you need to later. There is a potential security advantage to leaving this aperture driver xf86(4) disabled, as the graphics engine on a modern video card could potentially be used to alter memory beyond the processor's control.
Do you expect to run the X Window System? [no] y
Next, you are asked if you are wanting to use a serial console with this computer, rather than a standard keyboard and monitor. If you chose "yes" and answer a couple other simple questions, /etc/boot.conf and /etc/ttys will be edited appropriately for you. Most users will take the default, no here.
Change the default console to com0? [no] Enter
Your last task is to enter the time zone. Depending on where your machine lives, there are may be several equally valid answers for the question. In the example that follows, we used US/Eastern, but could also have used EST5EDT or US/Michigan and had the same result. Hitting ? at the prompts will guide you through your choices.
Saving configuration files......done.
Generating initial host.random file ......done.
What timezone are you in? ('?' for list) [Canada/Mountain] ?
Africa/ Chile/ GB-Eire Israel NZ-CHAT UCT
America/ Cuba GMT Jamaica Navajo US/
Antarctica/ EET GMT+0 Japan PRC UTC
Arctic/ EST GMT-0 Kwajalein PST8PDT Universal
Asia/ EST5EDT GMT0 Libya Pacific/ W-SU
Atlantic/ Egypt Greenwich MET Poland WET
Australia/ Eire HST MST Portugal Zulu
Brazil/ Etc/ Hongkong MST7MDT ROC posix/
CET Europe/ Iceland Mexico/ ROK posixrules
CST6CDT Factory Indian/ Mideast/ Singapore right/
Canada/ GB Iran NZ Turkey zone.tab
What timezone are you in? ('?' for list) [Canada/Mountain] US
What sub-timezone of 'US' are you in? ('?' for list) ?
Alaska Central Hawaii Mountain Samoa
Aleutian East-Indiana Indiana-Starke Pacific
Arizona Eastern Michigan Pacific-New
Select a sub-timezone of 'US' ('?' for list): Eastern
Setting local timezone to 'US/Eastern'...done.
If you are concerned about very precise time, you may wish to read this.
The last steps are for the system to create the /dev directory (which may take a while on some systems, especially if you have a small amount of RAM), and install the boot blocks.
Making all device nodes...done.
Installing boot block...
boot: /mnt/boot
proto: /usr/mdec/biosboot
device: /dev/rwd0c
/usr/mdec/biosboot: entry point 0
proto bootblock size 512
/mnt/boot is 3 blocks x 16384 bytes
fs block shift 2; part offset 3069360; inode block 216, offset 9256
using MBR partition 1: type 166 (0xa6) offset 3069360 (0x2ed5b0)

CONGRATULATIONS! Your OpenBSD install has been successfully completed!
To boot the new system, enter halt at the command prompt. Once the
system has halted, reset the machine and boot from the disk.
# halt
syncing disks... done

The operating system has halted.
Please press any key to reboot.
OpenBSD is now installed on your system and ready for its first boot, but before you do...
Before you reboot
At this point, your system is installed and ready to be rebooted and configured for service. Before doing this, however, it would be wise to check out the Errata page to see if there are any bugs that would immediately impact you.
A trick you can use for some "before first boot" configuration is to run:
# /mnt/usr/sbin/chroot /mnt
at the shell prompt. This will set your mount points to be what they will be on a normal reboot of your newly installed system. You can now do some basic system configuration, such as adding users, changing mount points, etc.
After you reboot
One of your first things to read after you install your system is afterboot(8).
You may also find the following links useful:
Adding users in OpenBSD
Initial Network Setup
Man Pages of popular/useful commands
OpenBSD man pages on the Web
The OpenBSD Packages and Ports system for installing software
One last thing...
The OpenBSD developers ask you to Send in a copy of your dmesg. This is really appreciated by the developers, and ultimately, all users.
4.6 - What files are needed for installation?
The complete OpenBSD installation is broken up into a number of separate file sets. Not every application requires every file set. Here is an overview of each:
bsd - This is the Kernel. Required
bsd.mp - Multi-processor (SMP) kernel (only some platforms)
bsd.rd - RAM disk kernel
base41.tgz - Contains the base OpenBSD system Required
etc41.tgz - Contains all the files in /etc Required
comp41.tgz - Contains the compiler and its tools, headers and libraries. Recommended
man41.tgz - Contains man pages Recommended
misc41.tgz - Contains misc info, setup documentation
game41.tgz - Contains the games for OpenBSD
xbase41.tgz - Contains the base files for X11
xetc41.tgz - Contains the /etc/X11 and /etc/fonts configuration files
xfont41.tgz - Contains X11's font server and fonts
xserv41.tgz - Contains X11's X servers
xshare41.tgz - Contains manpages, locale settings, includes, etc. for X
The etc41.tgz and xetc41.tgz sets are not installed as part of an upgrade, only as part of a complete install, so any customizations you make will not be lost. You will have to update your /etc, /dev and /var directories manually.
Even if you have no intention of running X, some third party packages require the graphic libraries in xbase41.tgz to be installed on your system. These applications can usually be satisified simply by installing xbase41.tgz, the rest of X is not needed.
4.7 - How much space do I need for an OpenBSD installation?
Obviously, the answer to this question varies tremendously based on your use of the system. However, these numbers can be used as a starting point:
(root) 60MB
/usr 420MB (no X) or 550MB (with X)
/var 25MB
/tmp 50MB
swap 32MB
Those are minimum suggested filesystem sizes for a full system install. The numbers include enough extra space to permit you to run a typical home system that is connected to the Internet, but not much else.
Keep the following facts in mind, however:
These are minimum values. Disk space is relatively cheap now, trying to squeeze your system into the smallest possible disk is rarely worth the effort. For special purpose applications, the above numbers can be made smaller, but you will need to experiment with it.
These numbers do NOT include the ports tree.
If you plan to install a significant amount of third party software, make your /usr partition much larger. How large will depend on your applications, of course.
For a system that handles lots of email or web pages (stored, respectively, in /var/mail and /var/www) you will want to make your /var partition significantly larger, or put them on separate partitions.
For a multiuser system which may generate lots of logs, you will want to make your /var partition significantly larger still, or create a separate log partition (/var/log).
If you plan to rebuild the kernel and system from source, you will want to make the /usr partition significantly larger, 4G is not a bad size.
Compiling some ports from source can take huge amounts of space on your /usr and /tmp partitions. This is another reason we suggest using pre-compiled packages instead.
The /tmp partition is used in the compiling of ports, among other things, so how big you make it depends on what you do with it. 50M may be plenty for most people, but some large applications may require 100M or more of /tmp space.
The 'b' partition of your first drive automatically becomes your system swap partition -- we recommend a minimum of 32MB but if you have disk to spare make it at least 64MB. If you have lots of disk space to spare, make this 256MB, or even 512MB. On the other hand, if you are using a flash device for disk, you probably want no swap partition at all. Many people follow an old rule of thumb that your swap partition should be twice the size of your main system RAM. This rule is nonsense. On a modern system, that's a LOT of swap, most people prefer that their systems never swap. Use what is appropriate for your needs.
Swap and /var spaces are used to store system core dumps on in the event of a crash(8). If this is a consideration for you, your swap space should be slightly larger than the amount of main memory you are likely to ever have in the system. Upon reboot, savecore(8) will attempt to save the contents of the swap partition to a file in /var/crash so again, if this is a priority for you, your /var partition must have enough free space to hold these dump files. Be realistic -- few developers will want to look at your 1GB dump file, so if you aren't planning on investigating a crash locally, this is probably not a concern.
There are several reasons for using separate filesystems, instead of shoving everything into one or two filesystems:
Security: You can mark some filesystems as 'nosuid', 'nodev', 'noexec', 'readonly', etc. This is done by the install process, if you use the above described partitions.
Stability: A user, or a misbehaved program, can fill a filesystem with garbage if they have write permissions for it. Your critical programs, which of course run on a different filesystem, do not get interrupted.
Speed: A filesystem which gets written to frequently may get somewhat fragmented. (Luckily, the ffs filesystem that OpenBSD uses is not prone to heavy fragmentation.)
Integrity: If one filesystem is corrupted for some reason then your other filesystems are still OK.
Size: Many machines have limits on the area of a disk where the boot ROM can load the kernel from. In some cases, this limit may be very small (504M for an older 486), in other cases, a much larger limit (for example, 2G, 8G, or 128G on i386 systems). As the kernel can end up anywhere within the root partition, the entire root partition should be within this area. For more details, see this section. A good guideline might be to keep your / partition completely below 2G, unless you know your platform (and particular machine) can handle more (or less) than that.
Some additional thoughts on partitioning:
For your first attempt at an experimentation system, one big / partition and swap may be easiest until you know how much space you need. By doing this you will be sacrificing some of the default security features of OpenBSD that require separate filesystems for /, /tmp, /var, /usr and /home. However, you probably should not be going into production with your first OpenBSD install.
A system exposed to the Internet or other hostile forces should have a separate /var (and maybe even a separate /var/log) for logging.
A /home partition can be nice. New version of the OS? Wipe and reload everything else, leave your /home partition untouched. Remember to save a copy of your configuration files, though!
A separate partition for anything which may accumulate a large quantity of files that may need to be deleted can be faster to reformat and recreate than to delete. See the building by source FAQ for an example (/usr/obj).
If you wish to rebuild your system from source for any reason, the source will be in /usr/src. If you don't make a separate partition for /usr/src, make sure /usr has sufficient space.
A commonly forgotten fact: you do not have to allocate all space on a drive when you set the system up! Since you will now find it a challenge to buy a new drive smaller than 20G, it can make sense to leave a chunk of your drive unallocated. If you outgrow a partition, you can allocate a new partition from your unused space, duplicate your existing partition to the new partition, change /etc/fstab to point to the new partition, remount, you now have more space.
If you make your partitions too close to the minimum size required, you will probably regret it later, when it is time to upgrade your system.
If you make very large partitions, keep in mind that performing filesystem checks using fsck(8) requires about 1M of RAM per gigabyte of filesystem size, and may be very time-consuming or not even feasible on older, slower systems (please also refer to this section).
If you permit users to write to /var/www (i.e., personal web pages), you might wish to put it on a separate partition, so you can use quotas to restrict the space they use, and if they fill the partition, no other parts of your system will be impacted.
4.8 - Multibooting OpenBSD/i386
Multibooting is having several operating systems on one computer, and some means of selecting the which OS is to boot. It is not a trivial task! If you don't understand what you are doing, you may end up deleting large amounts of data from your computer. New OpenBSD users are strongly encouraged to start with a blank hard drive on a dedicated machine, and then practice your desired configuration on a non-production system before attempting a multiboot configuration on a production machine. FAQ 14 has more information about the OpenBSD boot process.
Only one of the four primary MBR partitions can be used for booting OpenBSD (i.e., extended partitions will not work).
Here are several options to multibooting:
Setting active partitions
This is probably the most overlooked, and yet, sometimes the best solution for multibooting. Simply set the active partition in whatever OS you are currently using to be the one you want to boot by default when you next boot. Virtually every OS offers a program to do this; OpenBSD's is fdisk(8), similar named programs are in Windows 9x and DOS, and many other operating systems. This can be highly desirable for OSs or systems which take a long time to shut down and reboot -- you can set it and start the reboot process, then walk away, grab a cup of coffee, and come back to the system booted the way you want it -- no waiting for the Magic Moment to select the next OS.
Boot floppy
If you have a system that is used to boot OpenBSD infrequently (or don't wish other users of the computer to note anything has changed), consider using a boot floppy. Simply use one of the standard OpenBSD install floppies, and create an /etc/boot.conf file (yes, you will also have to create an /etc directory on the floppy) with the contents:
boot hd0a:/bsd
to cause the system to boot from hard drive 0, OpenBSD partition 'a', kernel file /bsd. Note you can also boot from other drives with a line like: "boot hd2a:/bsd" to boot off the third hard drive on your system. To boot from OpenBSD, slip your floppy in, reboot. To boot from the other OS, eject the floppy, reboot.
In this case, the boot(8) program is loaded from the floppy, looks for and reads /etc/boot.conf. The "boot hd0a:/bsd" line instructs boot(8) where to load the kernel from -- in this case, the first HD the BIOS sees. Keep in mind, only a small file (/boot) is loaded from the floppy -- the system loads the entire kernel off the hard disk, so this only adds about five seconds to the boot process.
Windows NT/2000/XP NTLDR
To multiboot OpenBSD and Windows NT/2000/XP, you can use NTLDR, the boot loader that NT uses. To multi-boot with NT, you need a copy of your OpenBSD Partition Boot Record (PBR). After running installboot, you can copy it to a file using dd(1), following a process similar to:
# dd if=/dev/rsd0a of=openbsd.pbr bs=512 count=1
Note: this is a really good time to remind you that blindly typing commands in you don't understand is a really bad idea. This line will not work directly on most computers. It is left to the reader to adapt it to their machine.
Now boot NT and put openbsd.pbr in C:. Add a line like this to the end of C:BOOT.INI:
When you reboot, you should be able to select OpenBSD from the NT loader menu. There is much more information available about NTLDR at the NTLDR Hacking Guide.
On Windows XP you can also edit the boot information using the GUI; see the XP Boot.ini HOWTO.
Programs that do much of this for you are available, for example, BootPart. This program can be run from Windows NT/2000/XP, and will fetch the OpenBSD PBR, place it on your NT/2000/XP partition, and will add it to C:BOOT.INI
Note: The Windows NT/2000/XP boot loader is only capable of booting OSs from the primary hard drive. You can not use it to load OpenBSD from the second drive on a system.
Other boot loaders
Some other bootloaders OpenBSD users have used successfully include GAG, OS-BS, The Ranish Partition Manager and GRUB.
OpenBSD and Linux (i386)
Please refer to INSTALL.linux, which gives in depth instructions on getting OpenBSD working with Linux.
Time zone issues
OpenBSD expects the computer's real-time clock to be set to UTC (Universal Coordinated Time). Some other OSs expect the real-time clock to be set to local time. Obviously, this can create a bit of a problem if you are using both OSs on the same computer. One or the other is most likely going to have to be adapted. More info on doing this is in FAQ 8 - Why is my clock off by several hours?
4.9 - Sending your dmesg to dmesg@openbsd.org after the install
Just to remind people, it's important for the OpenBSD developers to keep track of what hardware works, and what hardware doesn't work perfectly.
A quote from /usr/src/etc/root/root.mail
If you wish to ensure that OpenBSD runs better on your machines, please do us
a favor (after you have your mail system configured!) and type something like:
# dmesg | mail -s "Sony VAIO 505R laptop, apm works OK" dmesg@openbsd.org
so that we can see what kinds of configurations people are running. As shown,
including a bit of information about your machine in the subject or the body
can help us even further. We will use this information to improve device driver
support in future releases. (Please do this using the supplied GENERIC kernel,
not for a custom compiled kernel, unless you're unable to boot the GENERIC
kernel. If you have a multi-processor machine, dmesg results of both GENERIC.MP
and GENERIC kernels are appreciated.) The device driver information we get from
this helps us fix existing drivers. Thank you!
Make sure you send email from an account that is able to also receive email so developers can contact you if they have something they want you to test or change in order to get your setup working. It's not important at all to send the email from the same machine that is running OpenBSD, so if that machine is unable to receive email, just
$ dmesg | mail your-account@yourmail.dom
and then forward that message to
where your-account@yourmail.dom is your regular email account.
Please send only GENERIC kernel dmesgs. Custom kernels that have device drivers removed are not helpful.
If you have a supported multiprocessor system and normally run the GENERIC.MP kernel, it is helpful to developers to see the dmesg output of both the GENERIC kernel and the GENERIC.MP kernel, so please send both of them in separate emails.
The dmesgs are received on a computer using the spamd spam rejection system. This may cause your dmesg to not be accepted by the mail servers for a period of time. Be patient, after half an hour to an hour or so, it will get through.
The method above is very easy, but if you have chosen not to configure mail on your OpenBSD system, you should still send your dmesg to the developers. Save your dmesg output to a text file.
$ dmesg > ~/dmesg.txt
Then transfer this file (using FTP/scp/floppydisk/carrier-pigeon/...) to the system you normally use for email. Since the dmesg output you send in is processed automatically, be sure to check the following when using alternate email clients/systems:
Configure your email client to send messages as plain text; do not use HTML-formatted email.
Turn off any forced line break feature. Many email clients are configured to insert line breaks at 72 columns (the norm for mailing lists).
Make sure your email client does not reformat messages into "text-flow" nonsense.
Do not send the dmesg output as file attachment. Put the dmesg output into the body of the message.
4.10 - Adding a file set after install
"Oh no! I forgot to add a file set when I did the install!"
Sometimes, you realize you really DID need comp41.tgz (or any other system component) after all, but you didn't realize this at the time you installed your system. Good news: There are two easy ways to add file sets after the initial install:
Using the upgrade process
Simply boot your install media (CD-ROM or Floppy), and choose Upgrade (rather than Install). When you get to the lists of file sets to install, choose the sets you neglected to install first time around, select your source, and let it install them for you.
Using tar(1)
The install file sets are simply compressed tar files, and you can expand them manually from the root of the filesystem:
# cd /
# tar xzvpf comp41.tgz
Do NOT forget the 'p' option in the above command in order to restore the file permissions properly!
One common mistake is to think you can use pkg_add(1) to add a missing file sets. This does not work. pkg_add(1) is the package management tool to install third party software. It handles package files, not generic tar files like the install sets.
If you are installing the xbase file set on your system for the first time using tar(1) and without rebooting, the shared library cache must be updated after the installation using ldconfig(8). To add all the X libraries to the cache:
# ldconfig -m /usr/X11R6/lib
Alternatively, you can just reboot your system, and this will be done automatically by the rc(8) startup script.
4.11 - What is 'bsd.rd'?
bsd.rd is a "RAM Disk" kernel. This file can be very useful; many developers are careful to keep it on the root of their system at all times.
Calling it a "RAM Disk kernel" describes the root filesystem of the kernel -- rather than being a physical drive, the utilities available after the boot of bsd.rd are stored in the kernel, and are run from a RAM-based filesystem. bsd.rd also includes a healthy set of utilities to allow you to do system maintenance and installation.
On some platforms, bsd.rd is actually the preferred installation technique -- you place this kernel on an existing filesystem, boot it, and run the install from it. On most platforms, if you have a running older version of OpenBSD, you can FTP a new version of bsd.rd, reboot from it, and install a new version of OpenBSD without using any removable media at all.
Here is an example of booting bsd.rd on an i386 system:
Using Drive: 0 Partition: 3
reading boot.....
probing: pc0 com0 com1 apm mem[639k 255M a20=on]
disk: fd0 hd0+
>> OpenBSD/i386 BOOT 2.10
boot> boot hd0a:/bsd.rd
. . . normal boot to install . . .
As indicated, you will be brought to the install program, but you can also drop to the shell to do maintenance on your system.
The general rule on booting bsd.rd is to change your boot kernel from /bsd to bsd.rd through whatever means used on your platform.
4.12 - Common installation problems
4.12.1 - My Compaq only recognizes 16M RAM
Some Compaq systems have an issue where the full system RAM is not detected by the OpenBSD second stage boot loader properly, and only 16M may be detected and used by OpenBSD. This can be corrected either by creating/editing /etc/boot.conf file, or by entering commands at the "boot>" prompt before OpenBSD loads. If you had a machine with 64M RAM, but OpenBSD was only detecting the first 16M, the command you would use would be:
machine mem +0x3000000@0x1000000
to add 48M (0x3000000) after the first 16M (0x1000000). Typically, if you had a machine with this problem, you would enter the above command first at the install floppy/CD-ROM's boot> prompt, load the system, reboot, and create an /etc/boot.conf file with the above line in it so all future bootings will recognize all available RAM.
It has also been reported that a ROM update will fix this on some systems.
4.12.2 - My i386 won't boot after install
Your install seemed to go fine, but on first boot, you see no sign of OpenBSD attempting to boot. There are a few common reasons for this problem:
No partition was flagged active in fdisk(8). To fix this, reboot the machine using the boot floppy or media, and "flag" a partition as "active" (bootable). See here and here.
No valid boot loader was ever put on the disk. If you answer "Y" to the "Use entire disk for OpenBSD?" question during the install, or use the "reinit" option of fdisk(8), the OpenBSD boot record is installed on the Master Boot Record of the disk; otherwise, the existing master boot code is untouched. This will be a problem if no other boot record existed. One solution is to boot the install media again, drop to the shell and invoke fdisk(8) to update the MBR code from the command line:
· # fdisk -u wd0
Note: the "update" option within the interactive ("-e") mode of fdisk will not write the signature bytes required to make the disk bootable.
In some rare occasions, something may go wrong with the second stage boot loader install. Reinstalling the second stage boot loader is discussed here.
4.12.3 - My (older, slower) machine booted, but hung at the ssh-keygen steps
It is very likely your machine is running fine, just taking a while to do the ssh key generation process. A SPARCStation2 or a Macintosh Quadra may take several hours to complete the three ssh-keygen(1) steps. Just let it finish; it is only done once per install.
Note that the default key size was increased for OpenBSD 3.8, so the generation times are much longer than they used to be. Users of very slow machines may wish to generate their keys on another computer, place them in a site41.tgz file, and install them with the rest of the file sets.
4.12.4 - I got the message "Failed to change directory" when doing an install
When doing an FTP install of a snapshot during the -beta stage of the OpenBSD development cycle, you may see this:

Do you want to see a list of potential FTP servers? [yes] Enter
Getting the list from (ftp.openbsd.org)... FAILED
Failed to change directory.
Server IP address or hostname?
This is normal and expected behavior during this pre-release part of the cycle. The install program looks for the FTP list on the primary FTP server in a directory that won't be available until the release date, so you get the above message.
Simply use the FTP mirror list to find your favorite FTP mirror, and manually enter its name when prompted.
Note: You should not see this if you are installing -release or from CD-ROM.
4.12.5 - My fdisk partition table is trashed or blank!
Occasionally, a user will find a system will work, but when doing an fdisk wd0, they see a completely blank (or sometimes, garbage) partition table. This is usually caused by having created a partition in fdisk(8) which had an offset of zero sectors, rather than the one track offset it should have (note: this is assuming the i386 or amd64 platform. Other platforms have different offset requirements, some need NO offset). The system then boots using the PBR, not using the MBR.
While this configuration can work, it can be a maintenance problem and should be fixed. To fix this, the disk's file systems must generally be recreated from scratch (though if you REALLY know what you are doing, you may be able to recreate just your disklabel and MBR, and only lose and have to rebuild the first OpenBSD partition on the disk).
4.13 - Customizing the install process
siteXX.tgz file
The OpenBSD install/upgrade scripts allow the selection of a user-created set called "siteXX.tgz", where XX is the release version (e.g. 41). The siteXX.tgz file set is, like the other file sets, a gzip(1) compressed tar(1) archive rooted in '/' and is un-tarred like the other sets with the options xzpf. This set will be installed last, after all other file sets.
This file set allows the user to add to and/or override the files installed in the 'normal' sets and thus customize the installation or upgrade.
You can also create and use hostname-specific install sets, which are named siteXX-<hostname>.tgz, for example, "site41-puffy.tgz". This allows easy per-host customized installations, upgrades, or disaster recovery.
Some example uses of a siteXX.tgz file:
Create a siteXX.tgz file that contains all the file changes you made since first installing OpenBSD. Then, if you have to re-create the system you simply select siteXX.tgz during the re-install and all of your changes are replicated on the new system.
Create a series of machine specific directories that each contain a siteXX.tgz file that contains files specific to those machine types. Installation of machines (e.g. boxes with different graphics cards) of a particular category can be completed by selecting the appropriate siteXX.tgz file.
Put the files you routinely customize in a same or similar way in a siteXX.tgz file -- /etc/skel files, /etc/pf.conf, /var/www/conf/httpd.conf, /etc/rc.conf.local, etc.
install.site/upgrade.site scripts
As the last step in the install/upgrade process, the scripts look in the root directory of the newly installed/upgraded system for install.site or upgrade.site, as appropriate to the current process, and runs this script in an environment chrooted to the installed/upgraded system's root. Remember, the upgrade is done from a booted file system, so your target file system is actually mounted on /mnt. However, because of the chroot, your script can be written as if it is running in the "normal" root of your file system. Since this script is run after all the files are installed, you have almost full functionality of your system (though, in single user mode) when your script runs.
Note that the install.site script would have to be in a siteXX.tgz file, while the upgrade.site script could be put in the root directory before the upgrade, or could be put in a siteXX.tgz file.
The scripts can be used to do many things:
Remove files that are installed/upgraded that you don't want present on the system.
Remove/upgrade/install the packages you want on the installed system.
Do an immediate backup/archive of the new system before you expose it to the rest of the world.
Use rdate(8) to set the system time.
The combination of siteXX.tgz and install.site/upgrade.site files is intended to give users broad customization capabilities without having to build their own custom install sets.
Note: if you will be doing your install from an http server, you will need to add your site*.tgz file(s) to the file index.txt in the source directory in order for them to be listed as an option at install time. This is not needed for FTP or other installs.
4.14 - How can I install a number of similar systems?
Here are some tools you can use when you have to deploy a number of similar OpenBSD systems.
siteXX.tgz and install/upgrade.site files
See the above article.
Restore from dump(8)
On most platforms, the boot media includes the restore(8) program, which can be used to restore a backup made by dump(8). Thus, you could boot from a floppy, CD, or bsd.rd file, then fdisk, disklabel, and restore the desired configuration from tape or other media, and install the boot blocks. More details here.
Disk imaging
Unfortunately, there are no known disk imaging packages which are FFS-aware and can make an image containing only the active file space. Most of the major disk imaging solutions will treat an OpenBSD partition as a "generic" partition, and can make an image of the whole disk. This often accomplishes your goal, but usually with huge amounts of wasted space -- an empty, 10G /home partition will require 10G of space in the image, even if there isn't a single file in it. While you can typically install a drive image to a larger drive, you would not be able to directly use the extra space, and you would not be able to install an image to a smaller drive.
If this is an acceptable situation, you may find the dd command will do what you need, allowing you to copy one disk to another, sector-for-sector. This would provide the same functionality as commercial programs without the cost.
4.15 - How can I get a dmesg(8) to report an install problem?
When reporting a problem, it is critical to include the complete system dmesg(8). However, often when you need to do this, it is because the system is working improperly or won't install so you may not have disk, network, or other resources you need to get the dmesg to the appropriate mail list. There are other ways, however:
Floppy disk: The boot disks and CD-ROM have enough tools to let you record your dmesg to an MSDOS floppy disk for reading on another machine. Place an MSDOS formatted floppy in your disk drive and execute the following commands:
· mount -t msdos /dev/fd0a /mnt
· dmesg >/mnt/dmesg.txt
· umount /mnt
If you have another OpenBSD system, you can also write it to an OpenBSD compatible floppy -- often, the boot floppy has enough room on it to hold the dmesg. In that case, leave off the "-t msdos" above.
Serial Console: Using a serial console and capturing the output on another computer is often the best way to capture diagnostic information - particularly if the computer panics immediately after boot. As well as a second computer, you will need a suitable serial cable (often a null-modem cable), and a terminal emulator program that can capture screen output to file.
General information on setting up a serial console is provided elsewhere in the FAQ; in order to capture a log of the install, the following commands are usually sufficient.
At the boot loader prompt, enter
boot> set tty com0
This will tell OpenBSD to use the first serial port (often called COM1 or COMA in PC documentation) as a serial console. The default baud rate is 9600.
These machines will automatically use a serial console if started without a keyboard present. If you have a keyboard and monitor attached, you can still force the system to use a serial console with the following invocation at the ok prompt.
ok setenv input-device ttya
ok setenv output-device ttya
ok reset
FTP: Under some circumstances, and provided you first set up the network correctly, you may be able to use the ftp(1) client on the boot disk or CD-ROM to send the dmesg to a local FTP server, where you can retrieve it later.

Обновлено: 13.03.2015