SRM Installations

The biggest issue with SRM is that SRM only recognizes BSD-style disk labels when booting. Therefore, you will need to perform a custom installation in order to use BSD-style disk labels from within fdisk. To do this, you'll have to use the custom installation class. You should consult the Official Red Hat Linux Installation Guide and/or the Official Red Hat Linux Reference Guide, since they provide more thorough overviews of disk partitioning and the considerations that must be kept in mind for partitioning, given the role of the machine.

You should also be familiar with how the SRM console lists the devices present in your system. The SRM console references devices based on their type. Here is a list of common devices, and how they are referenced in the SRM console:

Table 1-7. Common Devices and How They are Referenced in the SRM Console

Device	Description
dkaXXX	SCSI device on the first SCSI bus. Further SCSI busses are denoted as dkbXXX...
dqaXXX	IDE device on the first IDE bus. Further IDE busses are denoted as dqbXXX...
dvaXXX	Floppy device
ewaXXX	Ethernet adapter
mkaXXX	SCSI tape device
pkaXXX	SCSI host adapter for the first SCSI bus. Further SCSI busses are denoted as pkbXXX...

Pre-Installation Information

To make the Red Hat Linux/Alpha installation go smoothly, you'll need some information before you begin:

The SRM device of the floppy drive. In most cases, the SRM device will be dva0.
The SRM device address of your hard drive. For example, in this document, we've used the SRM device dkb0 to refer to the hard drive.
The SRM device address of your CD-ROM drive. In this document, we've used device dka500 to refer to the CD-ROM drive.
The version number of the kernel you are installing.

To see the list of devices that the SRM console can see, type in the following command:

>>> show dev

A list of the devices that the SRM console is capable of seeing will appear. You must chose one of these devices from which to boot Red Hat Linux, during the installation process and every time you wish to boot your machine.

In some cases, however, you may have a SCSI adapter, IDE disks, or networking hardware that is compatible with Red Hat Linux, but is either not OEM Digital or Compaq hardware, or is otherwise undetectable by the SRM console. This may not impede your running Red Hat Linux on your Alpha machine, but you are always required to boot from a device that the SRM console recognizes.

SRM Variables

The SRM console has several variables which store their values in NVRAM, therefore keeping them between reboots. To see the value of SRM console variables, type in the command:

>>> show VAR

Replace VAR with the name of the SRM console variable.

To make listing variables easier, you can append an asterisk (*) at any point at the end of a variable name to show all variables that match the text typed so far. For example, the following command will show a list of all SRM console variables that begin with the letter "b":

>>> show b*

A special case of the show command is the show dev command (as previously mentioned), which lists all the devices that the SRM console can locate.

You can explicitly list all the devices of a particular type that SRM can see on your machine using the command:

>>> show dev dk

The previous command will show all SRM devices that begin with "dk" (i.e., all SCSI disks).

To set an SRM console variable, use the command:

>>> set VAR VALUE

In the above command, VAR is the name of the SRM console variable, and VALUE is the value it should be set to. Integer values do not have to be in quotes; however, it is recommended that you enclose all string values in quotes. And, while single-word string values don't need to be in quotes, it's easier to ALWAYS put string variable values in quotes.

To clear a SRM console variable, use the following command:

>>> set VAR ""

In the above command, VAR is the name of the SRM console variable, and the empty quotes ("") denote a blank value. Some integer values cannot be cleared, but can be set to 0. Also note that some SRM console variables are integer in value, but boolean in nature.

Several SRM console variables are worth mentioning:

BOOT_FILE: The BOOT_FILE variable is equivalent to the -file option of the SRM boot command.
BOOT_OSFLAGS: The BOOT_OSFLAGS SRM console variable is equivalent to the value passed to the boot command by the -flags option. It can be set to the value that you would pass to the -flags parameter of an SRM boot command. At the very least, it is convenient to set the root=/dev/XXX value.
BOOTDEF_DEV: The BOOTDEF_DEV SRM console variable sets the default boot device used when the boot command is issued.

SRM Disk Images

Some things have changed for the better since the 5.x releases of Red Hat Linux. Now there is a single, generic kernel for all Alpha machines. This simplifies Alpha installation immensely. The recommended method for booting into the Red Hat Linux installation process is to use a floppy disk to load the kernel and to load the installer directly from the CD-ROM.

All Alpha installations will need the kernel disk, created from the image file /images/generic.img on CD 1. However, if you plan to do an NFS installation, you will also need to use the RAM disk image from /images/ramdisk.img, also on CD 1. Instructions for creating disk images are located in the section called Writing an Image File to a Diskette.

SRM can boot directly from the Red Hat Linux CD that is shipped in the boxed set, as the CD does have a valid aboot block.

Booting the Installer

Now that you've created the kernel disk, you will need to boot into the Red Hat Linux installer. The boot command in the SRM console has a simple format and looks like this:

>>> boot <device> -file <boot file> -flags "<boot flags>"

Please Note: The entire text of all boot flags must be enclosed in double-quotes.

In the SRM console, the first floppy disk device is denoted as dva0. The generic kernel, contained on the boot disk you created, is called vmlinux.gz, and the CD-ROM, containing the Red Hat Linux installation program and files, is your CD device (dka500, in our example). Therefore, the boot command to boot into the installer is as follows:

>>> boot dva0 -file vmlinux.gz -flags "root=/dev/scd0"

If you have an IDE CD-ROM, you will need to know where in the IDE bus it is installed. It will be one of the following devices, based on its position on the IDE busses of your machine:

Table 1-8. Location of an IDE CD-ROM on an IDE Bus

Device	Location
hda	First IDE bus, master device
hdb	First IDE bus, slave device
hdc	Second IDE bus, master device
hdd	Second IDE bus, slave device

If your IDE CD-ROM is the first device on the second IDE bus, the boot command to launch the installer would be:

>>> boot dva0 -file vmlinux.gz -flags "root=/dev/hdc"

SRM Disk Partitioning

Once you get to the fdisk program, press "b" for BSD-style disk labeling.

BSD-style disk labeling is different in certain ways from the standard fdisk partitioning used in ARC/AlphaBIOS installations and on Intel installations:

The partitions are identified by letters (a:, b:, c:...) instead of numbers.
The partition types are different. For example, "83" and "82" no longer correspond to "Linux" and "Linux Swap," respectively. In fact, the Linux partition is referred to as "ext2" in BSD disk labels.
You can have a maximum of eight partitions (a: to h:) per drive.
The first partition should start on cylinder 2, since the first cylinder may not be full-size.
Partition types must be explicitly set after creation. The BSD disk label program uses a default partition type of "unknown" which causes the Disk Setup portion of the installer not to see the partitions.

You will want to create at least three partitions (or more if you chose to put /usr in a separate partition, etc.)

a:: The a: partition holds the aboot secondary loader. It should be about 2MB, and set to type "boot."
b:: The b: partition is your swap partition. It should be approximately twice the size of the amount of physical RAM present in your machine. It should be set to type "swap."
c: - h:: The c: through h: partitions compose your Linux filesystem layout. They should be sized as appropriate and set to type "ext2."

Follow the guidelines in the Official Red Hat Linux Installation Guide and in the section called Disk Partitioning for tips and techniques on how to divide your drive into partitions.

SRM Installation Process

Install Red Hat Linux/Alpha as described in the Official Red Hat Linux Installation Guide.

Post-Installation Boot Setup

Once the installation has been completed, the machine should be capable of booting directly from the SRM console. You should then be able to boot in SRM directly from the hard drive, using the command:

>>> boot <device> -file <boot file> -flags "<boot flags>"

<boot device> is the SRM device on which you installed the aboot secondary boot loader. <boot file> is the uncompressed kernel file that you installed (as described in the section called Pre-Installation Information). You will need to preface this with the number of the partition you have the kernel on (e.g., `3' for the third partition on the boot device, `2' for the second partition, etc.) and include the full path to the kernel.

<boot flags> are the root device (device mounted as /) and any other kernel flags that you need to pass.

For example, if you installed aboot on the first SCSI device on the second SCSI bus (device dkb0), the root of your Linux filesystem is the third partition of your first SCSI drive (/dev/sda3), and you installed version 2.2.12-20 of the Linux kernel on a uni-processor machine, then the boot command to boot your Alpha is:

>>> boot dkb0 -file 3/boot/vmlinuz-2.2.12-20 -flags "root=/dev/sda3"

Please Note: The kernel file you use in the <boot file> parameter may change.

SRM Autoboot

The SRM console has the ability to autoboot. It is controlled by the value of the AUTO_ACTION SRM console variable. AUTO_ACTION can be set to `halt' or `boot' to reflect the default power-on function of the Alpha machine when booted into the SRM console.

Please Note: When setting the value of the AUTO_ACTION SRM console variable, note that your machine will not be able to return to the SRM console unless you press the `HALT' button on the machine. Most desktop Alpha machines do not have a `HALT' button, so you should only set the AUTO_ACTION value to boot if the Alpha machine is a high-availability server that has a `HALT' button on the front panel.

Troubleshooting an SRM Installation

If you encounter problems installing Red Hat Linux/Alpha from SRM, answer the following questions:

Q: When you tried to boot your Alpha after installation, did the boot process fail because your machine can't find any partitions?
A: For SRM installations, you need to use BSD-style disk labeling. You'll need to install Red Hat Linux again. Choose the expert installation mode, in order to use BSD-style disk labels from within fdisk. See the section called SRM Disk Partitioning for more information about using BSD-style disk labels during SRM installations.
Q: When you tried to boot after installation, did you get an error message which indicated that no valid boot block was found?
A: Red Hat Linux/Alpha can't find the secondary boot loader, aboot.
While the installation process puts all the necessary components of the Red Hat Linux Operating System on your machine, the secondary loader may need to be installed manually. The secondary loader initializes the hardware and loads the Linux kernel into memory so it can bootstrap the machine.
Boot the installer. Switch to the command prompt on the TTY2 console with the following key combination: Alt-F2. After you've switched to the command prompt, verify that your root directory is mounted with the following command:
# mount
This should show the partition you selected to be mounted as / currently mounted as /mnt. Temporarily make it the root directory by issuing the command:
# chroot /mnt
Install the secondary boot loader, aboot. Install aboot to the disk from which you will be booting, using the swriteboot command. If you plan to boot your Alpha from the first SCSI hard disk, execute the swriteboot command like this:
bash# /sbin/swriteboot /dev/sda /boot/bootlx
If you plan to boot your Alpha from the second IDE device, the swriteboot command should be executed like this:
bash# /sbin/swriteboot /dev/hdb /boot/bootlx
Please Note: If you are prompted to use the -fX option because swriteboot will overwrite the partition data of partition X, then your a: partition is not large enough, and you will have to re-partition your drive and increase the size of the a: partition.
Once the aboot secondary loader has been installed, type the following command to exit the temporary root device set using the chroot command:
bash# exit
Now switch back to TTY1 using the following keystrokes: Alt-F1 and exit the installer.
Now try to boot your machine as stated in the section called Post-Installation Boot Setup.
Q: Have you checked the Errata page on Red Hat's website?
A: Check Red Hat's Errata page at http://www.redhat.com/errata for a solution to your problem.