Installation is supported from several media types, including: FFS partitions Tape Remote NFS partition CD-ROM FTP HTTP Not all methods are supported on all Sparc Systems and some of them work only with the new single-floppy installation or the miniroot installation, not with the older multi-floppy installation. If you have the OpenBSD CD-ROM distribution (and a CD-ROM drive), you may be able boot from it. Not all sparc systems support booting from CD-ROM and the current boot images is only known to work on Sun4c architecture workstations such as the IPC, SS1 or SS2. If you can boot from the CD-ROM, you are home free and can proceed to the installation steps. If not, you will need to do some setup work to prepare a bootable image, either floppies, a hard drive, or a compatible net boot server. In addition to the bootable image, you also need to consider how to access the binary distribution sets to actually install the system. If you have the OpenBSD CD-ROM distribution you can either access the CD-ROM directly from the bootable image or remotely mounted on another system via NFS. Although you can access the distribution sets directly from the CD-ROM or from one of the FTP mirrors over the internet, you may wish to transfer the sets to a local FTP or NFS server, or copy them to a partition on the target system's disk or onto a SCSI tape. The variety of options listed may seem confusing, but situations vary widely in terms of what peripherals and what sort of network arrangements a user has, the intent is to provide some way that will be practical. Creating a bootable floppy disk using DOS/Windows: First you need to get access to the OpenBSD Bootable floppy images. If you can access the CD-ROM distribution under DOS the bootable disks are in the 2.3/sparc directory, otherwise you you will have to download them from one of the OpenBSD ftp or http mirror sites, using ftp or a web-viewer. In either case, take care to do "binary" transfers, since these are images files and any DOS cr/lf translations or control/z EOF interpretations will result in corrupted transfers. You will also need to go to the "tools" directory and grab a copy of the rawrite.exe utility and its documentation. This program is needed to correctly copy the bootable filesystem image to the floppy, since it's an image of a unix partition containing a ffs filesystem, not a MSDOS format diskette. Once you have installed rawrite.exe, just run it and specify the name of the bootable image, such as "floppy.fs" and the name of the floppy drive, such as "a:". Be sure to use good quality HD (1.44MB) floppies, formatted on the system you're using. The image copy and boot process is not especially tolerant of read errors. Note that, when installing, the boot floppy can be write-protected (i.e. read-only). Creating a bootable floppy disk using SunOS or other Un*x-like system: First, you will need obtain a local copy of the bootable filesystem image as described above. If possible use cksum or md5 to verify the checksums of the images vs. the values in the CKSUM or MD5 files on the mirror site. Next, use the dd(1) utility to copy the file to the floppy drive. Under SunOS, the command would be: dd if=floppy23.fs of=/dev/rfd0c bs=36b If you are using something other than SunOS, you may have to adapt this to conform to local naming conventions for the floppy and options suitable for copying to a "raw" floppy image. The key issue is that the device name used for the floppy *must* be one that refers to the whole 2880 block image, not a partition or compatibility mode, and the copy command needs to be compatible with the requirement that writes to a raw device must be in multiples of 512-byte blocks. The variations are endless and beyond the scope of this document. If you're doing this on the system you intend to boot the floppy on, copying the floppy back to a file and doing a compare or checksum is a good way to verify that the floppy is readable and free of read/write errors. Creating a bootable hard disk using SunOS or other Un*x-like system: If you don't have a floppy drive you can copy the single floppy installation image "floppy23.fs" or the mini-root "miniroot23.fs" onto the hard disk you intend to boot on. Traditionally, the way to do this is to use dd(1) to place the bootable filesystem image in the "swap" partition of the disk (while running in single user mode), and then booting from that partition. Using the "b" partition allows you to boot without overwriting any useful parts of the disk, you can also use another partition, but don't used the "a" or "c" partition without understanding the disklabel issues described below under "incompatible systems". This requires that you be running SunOS, Solaris, OpenBSD or NetBSD which have a compatible view of SunOS disk labels and partitions. Use the dd(1) utility to copy the file to the hard drive. Under SunOS, the command would be: dd if=floppy23.fs of=/dev/rsd0b bs=36b - or - dd if=miniroot23.fs of=/dev/rsd0b bs=36b The blocksize is arbitrary as long as it's a multiple of 512-bytes and within the maximum supported by the driver, i.e. bs=126b may not work for all cases. Again, device/partition names may vary, depending on the OS involved. If you are preparing the hard drive on an incompatible system or don't care about the hard disk contents, you can also install the bootable image starting at the beginning of the disk. This lets you prepare a bootable hard-drive even if don't have a working operating system on your Sparc, but it important to understand that the bootable image installed this way includes a "disk label" which can wipe out any pre-existing disklabels or partitioning for the drive. The floppy image is used only for booting, and can be placed in a partition that will be overwritten during the install process, since it actually runs off a ram-disk image in the kernel. In contrast the miniroot is a normal unix root filesystem and you must place in a partition that will not be overwritten until you've completed the installation process. To copy the floppy image to the whole disk, overwriting labels: dd if=floppy23.fs of=/dev/rsdXc bs=36b Two notes - X should be replaced by the unit number of the target disk, which is most likely *not* the disk/partition that's your current root partition. Again names may vary depending on the OS involved. Second, after doing this, the disklabel will be one that would be appropriate for a floppy, i.e. one partition of 2880 block, and you'll probably want to change that later on. If you're starting with a virgin disk and trying to do this under SunOS, use format(8) and newfs(8) to set up the partitions and mark the intended partition as an normal partition type. If you're using OpenBSD, perhaps on another architecture, OpenBSD will create a "fictitious label" that will let you access the whole disk. To copy the floppy image to the hard disk, preserving SunOS, Solaris NetBSD or OpenBSD labels: dd if=floppy23.fs of=/dev/rsdXc bs=1b skip=1 seek=1 You need to be sure that your version of dd(1) supports the skip and seek operands, otherwise you can try a technique like: dd if=/dev/rsdXc of=/tmp/label bs=1b count=1 dd if=floppy23.fs of=/dev/rsdXc bs=36b dd if=/tmp/label of=/dev/rsdXc bs=1b count=1 In either case, you've created a situation where the disklabel and the filesystem information don't agree about the partition size and geometry, however the results will be usable. Creating a network bootable setup using SunOS or other Un*x-like system: The details of setting up a network bootable environment vary considerably, depending on the networks host. Extract the OpenBSD diskless(8) man page from the man23.tar.gz distribution set or see the copy on the OpenBSD web page. You will also need to reference the relevant man pages or administrators guide for the host system. Basically, you will need to set up reverse-arp (rarpd) and boot parameter (bootpd) information and make the OpenBSD bootblock, kernel/miniroot partition, and a swap file available as required by the netboot setup. The steps necessary to prepare the distribution sets for installation depend on which method of installation you choose. Some methods require a bit of setup first that is explained below. The new single floppy installation allows installing OpenBSD directly from FTP mirror sites over the internet, however you must consider the speed and reliability of your internet connection for this option. It may save much time and frustration to use ftp get/reget to transfer the distribution sets to a local server or disk and perform the installation from there, rather than directly on the internet. To install or upgrade OpenBSD using a tape, you need to do the following: To install OpenBSD from a tape, you need to make a tape that contains the distribution set files, each in "tar" format or in "gzipped tar format". First you will need to transfer the distribution sets to your local system, using ftp or by mounting the CD-ROM containing the release. Then you need to make a tape containing the files. If you're making the tape on a UN*X-like system, the easiest way to do so is make a shell script along the following lines, call it "/tmp/maketape". #! /bin/sh tape=/dev/nrst0 mt -f ${tape} rewind for file in base23 etc23 comp23 game23 man23 misc23 text23 x*23 do dd if=${file}.tar.gz of=${tape} done tar cf ${tape} bsd mt -f ${tape} offline # end of script And then: cd .../2.3/sparc sh -x /tmp/maketape If you're using a system other than OpenBSD or SunOS, the tape name and other requirements may change. To install OpenBSD using a remote partition, mounted via NFS, you must do the following: NOTE: This method of installation is recommended only for those already familiar with using BSD network configuration and management commands. If you aren't, this documentation should help, but is not intended to be all-encompassing. Place the OpenBSD distribution sets you wish to install into a directory on an NFS server, and make that directory mountable by the machine on which you are installing or upgrading OpenBSD. This will probably require modifying the /etc/exports file on of the NFS server and resetting its mount daemon (mountd). (Both of these actions will probably require superuser privileges on the server.) You need to know the the numeric IP address of the NFS server, and, if the server is not on a network directly connected to the machine on which you're installing or upgrading OpenBSD, you need to know the numeric IP address of the router closest to the OpenBSD machine. Finally, you need to know the numeric IP address of the OpenBSD machine itself. Once the NFS server is set up properly and you have the information mentioned above, you can proceed to the next step in the installation or upgrade process. If you're installing OpenBSD from scratch, go to the section on preparing your hard disk, below. If you're upgrading an existing installation, go directly to the section on upgrading. If you are upgrading OpenBSD, you also have the option of installing OpenBSD by putting the new distribution sets somewhere in your existing file system, and using them from there. To do that, you must do the following: Place the distribution sets you wish to upgrade somewhere in your current file system tree. At a bare minimum, you must upgrade the "base" binary distribution, and so must put the "base23" set somewhere in your file system. If you wish, you can do the other sets, as well, but you should NOT upgrade the "etc" distribution; the "etc" distribution contains system configuration files that you should review and update by hand. Once you have done this, you can proceed to the next step in the upgrade process, actually upgrading your system.