3 The UNIX Shell

This chapter contains the following information:

Working with the UNIX shell in the Full Installation environment

How to invoke the UNIX shell from the Full Installation

Samples of the types of tasks that can be performed from the UNIX shell environment

How to use the UNIX shell to:
- Create swap space
- Mount file systems
- Restore UNIX file systems (UFS) or Advanced File Systems (AdvFS) from tape backups
- Restore the system image from a damaged disk to a new disk and assign the old device name to the new device
- Change the size of a disk partition using the disklabel command

Returning to the Full Installation procedure from the UNIX shell environment

3.1 What Is the UNIX Shell?

The UNIX shell option is a way to access standard UNIX commands during a Full Installation. The primary purpose of the UNIX shell option is to provide a way to perform disk and file system maintenance before or during a Full Installation. The UNIX shell provides a way to access all UNIX commands that help you recover from serious problems such as / (root) file system corruption and enables you to perform general file system and disk maintenance tasks. It is usually not necessary to access the UNIX shell during an installation, but the option to do so is offered just in case you need it.

The base operating system distribution media (CD-ROM or RIS) contains file systems that are laid out just as the software would be installed on the system and provides direct access to the /, /usr, and /var directories. The distribution media contains a mixture of compressed and uncompressed software subsets. This format makes all commands and utilities in the uncompressed software subsets available in the UNIX shell even if your operating system is not yet fully functional. In effect, the mounted installation media is almost a full operating system environment. The UNIX shell option enables you to use the installation media as a tool for disaster recovery.

You should perform system management activities in the UNIX shell only if you have previous UNIX operating system experience.

3.2 Invoking the UNIX Shell

How you invoke the UNIX shell from the installation procedure depends upon whether you are using the graphical or text-based interface. When you invoke the UNIX shell option, the system is in a Bourne shell in single-user mode with superuser privileges. Refer to the Command and Shell User's Guide for more information about shells and privileges.

After you boot your system from the distribution media to start a Full Installation, you will see an Installation window on systems capable of graphical display. To invoke the UNIX shell, select the Shell Window menu item from the File menu at the top of the dialog box.

On systems that do not have graphics capabilities, there are two ways to invoke the UNIX shell from the text-based Full Installation interface. The first is to choose menu option 3 from the first screen of the text-based interface. The second method is to press [Ctrl/c] (at any time except during software subset loading).

Note

Any installation selections you have made before invoking the UNIX shell from the text-based interface are lost. Follow this procedure to restart the text-based installation and start the selection process again:
# cd /
# restart

3.3 UNIX Shell Capabilities

Using the UNIX shell in the installation environment provides most of the capabilities of a full operating system environment. The difference between using the UNIX shell in the installation environment and a normal operating system environment is that it works without a swap device and with very limited capacity within the memory file systems (MFS) that are available during a Full Installation. These two factors mean that tasks requiring large amounts of memory that create the need to swap, or tasks requiring large amounts of disk space (such as /tmp space), are likely to encounter failures in the installation environment.

Note

The / (root) file system is located on the CD-ROM or the RIS server, and after the installation distribution media is booted, it is mounted with read-only permissions. The memory file systems (MFS) at /var, /dev, and /devices are mounted with read-write permissions. Any changes you make to files in /var, /dev, and /devices are volatile and will not be saved when you halt the installation and return to the console prompt ( >>> ).

Here is a sample of the types of tasks you can perform from the UNIX shell; some of these tasks are documented in this chapter:

Create new file systems with the newfs command for UNIX File Systems (UFS) or with the mkfdmn and mkfset commands for Advanced File Systems (AdvFS).

Restore file systems with the restore command (for UFS) or the vrestore command (for AdvFS).

Modify disk partition tables with the disklabel command on systems that do not have graphics capabilities. Otherwise, use the graphical Disk Configuration Utility, which is accessed either by clicking on the Edit Partitions... button on the File System Layout window, or by invoking the /usr/sbin/diskconfig utility at the shell prompt.

Mount disks and file systems with the mount command.

Fix UFS file systems with the fsck command. The fsck command is not required for AdvFS file systems.

View a file with the ed text editor. By default, the EDITOR environment variable is set to ed. To enable the vi text editor:
- On older systems with console firmware that supports curses capability:
```
# TERM=vt100
^{[Footnote 4]}
 
# export TERM
```
- On newer systems with personal computer-like graphics:
```
# TERM=pccons
# export TERM
```

3.4 Creating Swap Space

If you need to perform a task that requires swap space, you can turn swapping on in the UNIX shell. The only suggestion for turning on swap space in the UNIX shell environment is that you should not use a partition that contains data you want to preserve because it will be overwritten. To avoid this problem, use a disk partition that previously has been used for swap space.

Follow this procedure to turn on swapping in the UNIX shell:

Decide which device (that is, the device name and partition) you want to use for the swap area. Be careful not to choose an area that has data that you want to preserve.

Note

A disk partition that previously was used for swap space will be labeled as swap in the disklabel and can be used without harm.

Change to the /dev/disk directory:
```
# cd /dev/disk
```

Turn the swap device on. In this example, the swap device is dsk0b:
```
# swapon dsk0b
```

Verify that the swap device is turned on:
```
# swapon -s
```

Note

The restart command, which is used to restart the installation from the UNIX shell, is no longer available after swap space has been enabled. To restart the installation, halt the system and reboot from the distribution media.

3.5 Mounting File Systems

The UNIX shell can be used to perform maintenance operations on existing file systems. For instance, if the kernel (the file called vmunix ) in your / (root) file system becomes damaged and you have a good backup copy of the kernel, you can mount your / file system and replace the damaged kernel. If you are using LSM volumes for the / file system, refer to Installation Guide for information about how to restart LSM.

Follow this procedure to mount an existing root file system located on /dev/disk/dsk0a to another mount point

Create a mount point in /var:
```
# mkdir /var/mnt
```

Mount the file system:

For UNIX file systems (UFS), enter:


# fsck -y /dev/disk/dsk0a
# mount /dev/disk/dsk0a /var/mnt

For Advanced File Systems (AdvFS), enter:


# mkdir -p /etc/fdmns/root_domain
# cd /etc/fdmns/root_domain
# ln -s /dev/disk/dsk0a dsk0a
# mount root_domain#root /var/mnt

The existing / file system is accessible at /var/mnt and can be modified at this point.

3.6 Restoring File Systems from Backup

The UNIX shell is ideal for restoring damaged / (root) file systems, but you can use the UNIX shell to restore file systems other than the / file system.

It is recommended to restore file systems from a full operating system environment. If such an environment is unavailable due to the need to restore either /var or /usr, boot your system to single-user mode by using your existing or restored / file system to provide more writable disk space than is available in the installation environment. Swap space can be made available in the UNIX shell as shown in the instructions in Section 3.4.

Section 3.6.1 describes how to restore UFS file systems, and Section 3.6.2 describes how to restore AdvFS file systems. Both sets of procedures assume that you are restoring the file system to the same disk.

3.6.1 Restoring UNIX File Systems (UFS) from Tape Backup

Use the following procedure to restore a UNIX file system to the same disk on which it originally resided. The size of the partition you are restoring must be larger than the size of the dump file.

Note

The disklabel command used in the following procedure writes the default disk partition tables to the disk. Writing a label with customized partition table settings may affect the entire disk. If the disk you are restoring has customized partition table settings, use the instructions shown in Section 3.8 to recreate them. Refer to System Administration or to the disklabel(8) reference page for more information.

Read the disk label on the affected disk:
disklabel -r disk

If the disk has a disk label, proceed to Step 3 to erase it. If the disk does not have a disk label, proceed to Step 4 to create one.

Zero out the disk label:
```
# disklabel -z dsk0
```

Create a new disk partition table with an a partition of at least 128 MB to hold the root file system. Follow the procedures in Section 3.8, and then continue with Step 5.

Create a new / file system by using the following command syntax:
newfs raw_device disk_type
The raw_device parameter specifies the full raw device pathname of the disk device on your system. The / file system must reside on the a partition. So, to create a new file system on an RZ26L disk on partition a, enter the following command:
```
# newfs /dev/rdisk/dsk0a rz26l
```

To prepare to mount the file system, create a mount point in /var:
```
# mkdir /var/mnt
```
Create mount points under the /var or /tmp directories because these are the only directories that are writable from the UNIX shell.

Mount the file system by using the following command syntax:
mount block_device /var/mnt
The block_device parameter specifies the full block device pathname of the disk device. For example, to mount the file system created in the previous step, enter the following command:
```
# mount /dev/disk/dsk0a /var/mnt
```

Verify that the tape device is powered up and is connected to the system.

Note

If the tape device was not powered up when the system initialized, it may not be visible to the system. In that case, shut down the system, power up the tape device, and reboot the system.

Determine whether or not a tape device special file exists. Use the following command to obtain information about all devices on the system:
```
# hwmgr -view device
```
If the output from this command does not indicate the presence of a device similar to /dev/tape/tape0_d0, create the tape device special file as described in Step10. Otherwise, skip to Step 11.

Create a device special file for the tape device:
```
# /sbin/dn_setup -install_tape
```
Messages such as +tape0 list the device names that are created during this operation. The plus sign indicates that the device is being added. When all tape device special files are created, proceed to Step 11 to continue with the recovery of the damaged / file system.

Restore the file system. If you are restoring dump files from a local file system, change to the /var/mnt directory, insert the medium containing the dump file, and enter the restore command with the following command syntax:
restore -Yrf dumpfile
The dumpfile parameter specifies the pathname of the file containing the dump data. For a tape, enter the following commands:
```
# cd /var/mnt
# restore -Yrf /dev/tape/tape_device
```

3.6.2 Restoring Advanced File Systems (AdvFS) from Tape Backup

Use the following procedure to restore an AdvFS file system to the same disk on which it originally resided.

Note

The disklabel command used in this procedure writes the default disk partition tables to the disk. Writing a label with customized partition table settings may affect the entire disk. If the disk you are restoring has customized partition table settings, use the instructions shown in Section 3.8 to recreate them. Refer to System Administration or to the disklabel(8) reference page for more information.

If the disk does not have a label, which could occur if the disk was physically damaged or replaced, write the default disk partition tables and bootstrap programs. The disk partitions and bootstrap programs should be operational. To determine if the disk has a valid label, use the disklabel command with the following syntax:
disklabel -r disk

If the disk has a disk label, proceed to Step 3 to erase it. If the disk does not have a disk label, proceed to Step 4 to create one.

If the disk has a disk label, zero out the disk label:
```
# disklabel -z dsk0
```
Proceed to Step 4.

Create a new disk partition table with an a partition of at least 128 MB to hold the root file system. Follow the procedures in Section 3.8, and then continue with Step 5.

Create a new / (root) file domain by using the following command syntax:
mkfdmn -r raw_disk domain
The raw_device parameter specifies the full raw device pathname of the disk device on your system. For example, to create a new file system on an RZ26L partition a, enter the following command:
```
# mkfdmn -r /dev/rdisk/dsk0a root_domain
```

Create a root fileset in the root_domain file by using the following command:
```
# mkfset domain fileset
```
The domain parameter specifies the name of the domain in which to create the fileset. For example:
```
# mkfset root_domain root
```

To prepare to mount the fileset, create a mount point in /var or /tmp:
```
# mkdir /var/mnt
```
It is recommended that you create mount points under the /var or /tmp directories because these are the only directories that are writable from the UNIX shell.

Mount the root fileset by using the following command syntax:
mount domain#fileset mount_point
The domain#fileset parameter specifies the root file domain and the root fileset. For example, to mount the fileset created in the previous steps, enter the following command:
```
# mount root_domain#root /var/mnt
```

Verify that the tape device is powered up and is connected to the system.

Note

If the tape device was not powered up when the system initialized, it may not be visible to the system. In that case, shut down the system, power up the tape device, and reboot the system.

Determine whether or not a tape device special file exists. Use the following command to obtain information about all devices on the system:
```
# hwmgr -view device
```
If the output from this command does not indicate the presence of a device similar to /dev/tape/tape0_d0, create the tape device special file as described in Step 11. Otherwise, skip to Step 12.

Create a device special file for the tape device:
```
# /sbin/dn_setup -install_tape
```
Messages such as +tape0 list the device names that are created during this operation. The plus sign indicates that the device is being added. When all tape device special files are created, proceed to Step 12 to continue with the recovery of the damaged / file system.

Restore the fileset using the vrestore command. To restore files from a local file system, change to the /var/mnt directory, insert the medium containing the dump file, and enter the vrestore command using the following syntax:
vrestore -vxf dumpfile
The dumpfile parameter specifies the pathname of the file containing the dump data. Enter the following commands for a tape:
```
# cd /var/mnt
# vrestore -vxf /dev/tape/tape_device
```
Note

You can restore a UFS format dump tape to AdvFS (for instance if you are converting a UFS / file system to AdvFS) and you can make a vdump tape on UFS. The restore command you use depends on the format of the tape (dump or vdump). Use vrestore to restore AdvFS dumps performed with the vdump command and restore to restore UFS dumps performed with the dump command. The corresponding restore command is used regardless of the target file system type.

Verify the /etc/fstab and /etc/fdmns directories. The mkfdmn command added /etc/fdmns/root_domain to the temporary / file system in the UNIX shell, which is deleted when you reboot the system.

You can use the UNIX shell to restore other file systems, but it is recommended that you perform file system restores from a full operating system environment. If such an environment is unavailable due to the need to restore either /var or /usr, you should boot your system to single-user mode by using your existing or restored / file system. In the single-user mode, more disk space is available, and swap space can be made available by issuing the following command:


# swapon -a

3.7 Restoring the Root File System Image from a Damaged Disk to a New Disk

This procedure describes how to restore a saved root file system image from a bad or damaged disk to a new disk and rename the new disk with the old system disk name.

The process of introducing a new system disk is more complicated than just updating the /etc/fstab file because the installation kernel may assign a different device name to the devices on the system than the ones that were assigned on the saved system image. As a result, the restored system image will disagree with the kernel used to restore the image, and you cannot assume that it is possible to determine what the mapping is between them.

Note

The following procedure assumes you have restored the system image to a new disk as described in Section 3.6 .

Halt the system:
```
# shutdown -h now
```

Determine the console disk names on the system:
```
>>> show dev
```

Boot the system to single user mode from the new disk with the restored image:
```
>>> boot -fl s console_disk_name
```

In single user mode, mount the / file system as writable:
```
# /sbin/mountroot
```
This process creates a device entry for the newly found disk. You will see several processing messages ending with output that is similar to the following:
```
+dskNa +dskNb +dskNc ...
```
In the previous example, N is the disk number assigned to the disk that is being seen for the first time and should be the disk that was restored. However, if more than one disk is shown in the output, you have added multiple disks since the system disk was last saved. Proceed to Step 5 to determine the disk name associated with the old damaged disk.

Determine the old boot disk name:
```
# /sbin/consvar -g booted_dev
```
Output from this command is similar to the following.
```
booted_dev = dskN
```
In the previous command output, booted_dev = dskN is the old_disk_name.

Exchange the old disk name with the new disk name:
```
# dsfmgr -e  new_disk_name old_disk_name
```
In the command line, substitute new_disk_name with the disk name obtained in Step 4, and substitute old_disk_name with the disk name returned in Step 5.
Note

You will receive an error if the old, damaged disk has been removed and is not active. In that situation, use the move flag (-m) instead:
```
# dsfmgr -m  new_disk_name old_disk_name
```

Reset the boot device:


# /sbin/consvar -s bootdef_dev=new_disk_name

Optionally perform additional maintenance tasks, or if you are finished, boot the system to multiuser mode:
```
# init 3
```

3.8 Changing the Size of a Disk Partition Using the disklabel Command

If you are using the text-based installation interface, use the disklabel command to change the drive identification or the disk partitions on the drive or to replace a damaged label or bootstrap. Refer to the disklabel(8) reference page for more information.

To look at the existing disk partition layout, enter the disklabel command in the following format and replace the variable n with the unit number of the disk. For example, to look at the existing disk partition layout of disk, enter the following command:

# disklabel -r /dev/disk/dskn

The following example uses an rz26l disk. In this example, the size of the b partition is decreased, and the size of the a partition is increased to 128 MB which is the minimum size to hold the / root file system. Follow this procedure to change the size of disk partitions:

Read the disk label on dsk0:
```
# disklabel -r dsk0
```

Set the EDITOR environment variable to use the ed editor:
```
# EDITOR=ed
# export EDITOR
```
Note

If you have a VGA monitor and want to use the vi editor, you first have to set the following variables:
```
# TERM=vt100
# export TERM
# EDITOR=vi
# export EDITOR
```
Then, use the corresponding vi commands instead of the commands shown in this procedure to make the disk label changes. The rest of this procedure uses the ed text editor.

Copy the disk label for dsk0 to a temporary file:
```
# disklabel -r dsk0 > /tmp/old_disklabel
```

Edit the copy of the disk label to protect against introducing errors on the real disk label:
```
# ed /tmp/old_disklabel
```

Display the disk label:
```
1,$p
```

Search for the b partition:
```
/b: 
```
Information similar to the following is displayed:
```
b:  262144  131072  unused  1024  8192  #  (Cyl. 164*- 492*)
```

Change the size of the b partition from 262144 sectors to 131072 sectors :
```
s/262144/131072/p
```
This reduces the size of the b partition from 128 MB to 64 MB. The revised information is displayed:
```
b:  131072  131072  unused  1024  8192  # (Cyl. 164*- 402)
```
There is no need to modify cylinder information; cylinder information automatically is modified when you save and exit the file.

Change the offset of the b partition from 131072 sectors to 262144 sectors :
```
s/131073/262144/2p
```
This changes the offset of the b partition to start at position 262144. The revised information is:
```
b:  131072  262144  unused  1024  8192  # (Cyl. 164*- 402)
```
There is no need to modify cylinder information; cylinder information automatically is modified when you save and exit the file.

Search for the a partition:
```
/a: 
```
Information similar to the following is displayed:
```
a:   131072  0    unused    0     0   # (Cyl. 0 -164*)
```

Because the size of the b partition was reduced by 131072 sectors, the size of the a partition should be increased by 131072 sectors. Change the size of the a partition from 131072 sectors to 262144 sectors:
```
s/131072/262144/p
```
This increases the size of the a partition from 64 MB to 128 MB.

Display the disk label again to verify your changes:
```
1,$p
```

Save your edits and quit the editor:
```
wq
```

Apply the new disk label to the disk:
```
# disklabel -R dsk0 /tmp/old_label
```

Display the newly customized disk label:
```
# disklabel -r dsk0
```

Exit the shell or restart the Full Installation.

3.9 Returning to the Installation Procedure from the UNIX Shell

The procedure for returning to Full Installation from the UNIX shell depends upon whether you are using the text-based or graphical interface.

3.9.1 Text-Based Interface

If you entered the UNIX shell from the text-based interface, enter the following commands to invoke the Full Installation from the UNIX shell:

# cd /
# restart

If you have a system console with graphics capability and you want to restart the installation procedure with the text-based interface instead of the graphical user interface, enter the following command:


# cd /
# restart nogui

Any installation selections you made up to the point you exited to the UNIX shell are lost. When you restart the Full Installation, you have to make your selections again.

3.9.2 Graphical Interface

If you entered the UNIX shell from the File menu in the graphical interface, you can resume the Full Installation at any time just by clicking back in the Full Installation window. You do not have to exit the shell window; however, if you want to exit the shell window, enter exit at the # prompt within the shell window.

If you entered the UNIX shell from the Quit button on the Summary window, use the procedure shown in Section 3.9.1 to restart the Full Installation.