Special Installation Topics

Partitioning with Disk Druid during installation

During installation, you are given the opportunity to change how your hard disk is partitioned. Red Hat recommends using the Disk Druid. The Disk Druid screen is divided into two sections. The top shows general information about each hard disk. The bottom shows details of each partition. Figure 2-2 shows an example of the Disk Druid window.

Figure 2-2: Partition your disk during installation from the Disk Setup window.

For each of the hard disk partitions, you can see:

• Device — The device name is the name representing the hard disk partition in the /dev directory. Each disk partition device begins with two letters: hd for IDE disks, sd for SCSI disks, ed for ESDI disks, or xd for XT disks. After that is a single letter representing the number of the disk (disk 1 is a, disk 2 is b, disk 3 is c, and so on). The partition number for that disk (1, 2, 3, and so on) follows that.

• Mount Point/Raid/Volume — The directory where the partition is connected into the Linux file system (if it is). You must assign the root partition (/) to a native Linux partition before you can proceed. If you are using RAID or LVM, the name of the RAID device or LVM volume appears here.

• Type — The type of file system that is installed on the disk partition. In most cases, the file system will be Linux (ext3), Win VFAT (vfat), or Linux swap. However, you can also use the previous Linux file system (ext2), physical volume (LVM), or software RAID.

• Format — Indicates whether (check mark) or not (no check mark) the installation process should format the hard disk partition. Partitions marked with a check are erased! So, on a multiboot system, be sure your Windows partitions, as well as other partitions containing data are not checked!

• Size (MB) — The amount of disk space allocated for the partition. If you selected to let the partition grow to fill the existing space, this number may be much larger than the requested amount.

• Start/End — Represents the partition's starting and ending cylinders on the hard disk.

In the top section, you can see each of the hard disks that are connected to your computer. The drive name is shown first. The Geometry section (Geom) shows the numbers of cylinders, heads, and sectors, respectively, on the disk. That's followed by the model name of the disk. The total amount of disk space, the amount used, and the amount free are shown in megabytes.

Partitioning with fdisk

The fdisk utility does the same job as Disk Druid, but it's no longer offered as an option during Red Hat installation. (If you are old school, however, you could press Ctrl+Alt+F2 during the installation process and run fdisk from the shell to partition your disk.)

The following procedures are performed from the command line as root user.

The fdisk command is one that is available on many different operating systems (although it looks and behaves differently on each). In Linux, fdisk is a menu-based command. To use fdisk to list all your partitions, type the following (as root user):

# fdisk –l
???
Disk /dev/hda: 40.0 GB, 40020664320 bytes
255 heads, 63 sectors/track, 4865 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
   
   Device Boot    Start       End    Blocks   Id  System
/dev/hda1   *         1        13    104391   83  Linux
/dev/hda2            14      4833  38716650   83  Linux
/dev/hda3          4834      4865    257040   82  Linux swap

To see how each partition is being used on your current system, type the following:

# df –h
Filesystem            Size  Used Avail Use% Mounted on
/dev/hda2              37G  5.4G   30G  16% /
/dev/hda1              99M  8.6M   86M  10% /boot
none                   61M     0   61M   0% /dev/shm

From the output of df, you can see that the root of your Linux system (/) is on the /dev/hda2 partition and that the /dev/hda1 partition is used for /boot.

To use fdisk to change your partitions, begin (as root user) by typing:

# fdisk device

where device is replaced by the name of the device you want to work with. For example, here are some of your choices:

After you have started fdisk, type m to see the options. Here is what you can do with fdisk:

• Delete a partition — Type d and you are asked to enter a partition number on the current hard disk. Type the partition number and press Enter. For example, /dev/sda2 would be partition number 2. (The deletion won't take effect until you write the change. Until then, it's not too late to back out.)

• Create a partition — If you have free space, you can add a new partition. Type n and you are asked to enter l for a logical partition (5 or over) or p for a primary partition (1–4). Enter a partition number from the available range. Then choose the first cylinder number from those available. (The output from fdisk –l shown earlier will show you cylinders being used under the Start and End columns.)

  Next, enter the cylinder number the partition will end with (or type the specific number of megabytes or kilobytes you want: for example, +50M or +1024K). You just created an ext3 Linux partition. Again, this change isn't permanent until you write the changes.

• Change the partition type — Press t to choose the type of file system. Enter the partition number of the partition number you want to change. Type the number representing the file system type you want to use in hexadecimal code. (Type L at this point to see a list of file system types and codes.) For a Linux file system, use the number 83; use 82 for a Linux swap partition. For a windows FAT32 file system, you can use the letter b.

• Display the partition table — Throughout this process, feel free to type p to display (print on the screen) the partition table as it now stands.

• Saving and quitting — If you don't like a change you make to your partitions, press q to exit without saving. Nothing will have changed on your partition table.

  Before you write your changes, display the partition table again and make sure that it is what you want it to be. To write your changes to the partition table, press w. You are warned about how dangerous it is to change partitions and asked to confirm the change.

An alternative to the fdisk command is sfdisk. The sfdisk command is command-line–oriented. Type the full command line to list or change partitions. (See the sfdisk man page for details.)

Tips for creating partitions

Changing your disk partitions to handle multiple operating systems can be very tricky. Part of the reason is that each different operating system has its own ideas about how partitioning information should be handled, as well as different tools for doing it. Here are some tips to help you get it right.

• If you are creating a dual-boot system, particularly for Windows ME or Windows XP, try to install the Windows operating system first. Otherwise, the Windows installation may make the Linux partitions inaccessible.

• The fdisk man page recommends that you use partitioning tools that come with an operating system to create partitions for that operating system. For example, the DOS fdisk knows how to create partitions that DOS will like, and the Red Hat Linux fdisk will happily make your Linux partitions. Once your hard disk is set up for dual boot, however, you should probably not go back to Windows-only partitioning tools. Use Linux fdisk or a product made for multi-boot systems (such as Partition Magic).

• You can have up to 63 partitions on an IDE hard disk. A SCSI hard disk can have up to 15 partitions. You won't need nearly that many partitions.

If you are using Red Hat Linux as a desktop system, you probably don't need a lot of different partitions within your Red Hat Linux system. There are, however, some very good reasons for having multiple partitions for Red Hat Linux systems that are shared by a lot of users or are public Web servers or file servers. Multiple partitions within Red Hat Linux offer the following advantages:

• Protection from attacks — Denial-of-service attacks sometimes take action that tries to fill up your hard disk. If public areas, such as /var, are on separate partitions, a successful attack can fill up a partition without shutting down the whole computer. Because /var is the default location for Web and FTP servers, and therefore might hold a lot of data, often entire hard disks are assigned to the /var file system alone.

• Protection from corrupted file systems — If you have only one file system (/), corruption of that file system can cause the whole Red Hat Linux system to be damaged. Corruption of a smaller partition can be easier to correct and can often allow the computer to stay in service while the corruption is fixed.

Here are some directories that you may want to consider making into separate file system partitions.

• /boot — Sometimes the BIOS in older PCs can access only the first 1024 cylinders of your hard disk. To make sure that the information in your /boot directory is accessible to the BIOS, create a separate disk partition (of only a few MB) for /boot and make sure that it exists below cylinder 1024. Then, the rest of your Linux system can exist outside of that 1024-cylinder boundary if you like. Even with several boot images, there is no need for /boot to be larger than 64MB. For newer hard disks, you can sometimes avoid this problem by selecting "linear mode" during installation. Then the boot partition can be anywhere on the disk.

• /usr — This directory structure contains most of the applications and utilities available to Red Hat Linux users. Having /usr on a separate partition lets you mount that file system as read-only after the operating system has been installed. This prevents attackers from replacing or removing important system applications with their own versions that may cause security problems. A separate /usr partition is also useful if you have diskless workstations on your local network. Using NFS, you can share /usr over the network with those workstations.

• /var — Your FTP (/var/ftp) and Web-server (/var/www) directories are, by default, stored under /var. Having a separate /var partition can prevent an attack on those facilities from corrupting or filling up your entire hard disk.

• /home — Because your user account directories are located in this directory, having a separate /home account can prevent an indiscriminate user from filling up the entire hard disk. (Disk quotas, see Chapter 10, represent another way of controlling disk use.)

• /tmp — Protecting /tmp from the rest of the hard disk by placing it on a separate partition can ensure that applications that need to write to temporary files in /tmp are able to complete their processing, even if the rest of the disk fills up.

Although people who use Red Hat Linux casually rarely see a need for lots of partitions, those who maintain and occasionally have to recover large systems are thankful when the system they need to fix has several partitions. Multiple partitions can localize deliberate damage (such as denial-of-service attacks), problems from errant users, and accidental file system corruption.

Before you begin

Read this section before you begin resizing your Windows partitions!

FIPS is a well-tested utility for resizing DOS FAT and VFAT file systems so that space can be made available to install Linux. To find out what kind of file system is being used on your Windows computer, do the following while Windows is running:

1. Double-click the My Computer icon.

2. Right-click the hard-disk (C:) icon and select Properties.

3. Look at the File system line. If it shows the file system type as FAT or FAT32 the disk partition can be resized with FIPS.

If you are using a Windows 95, Windows 98, or Windows ME computer, providing that you have enough free disk space on the partition, you can probably use FIPS to resize it. If, however, you have a Windows 2000, Windows NT, or Windows XP computer with an NTFS file system, you will have to use some other tools to resize it. Here are some of your choices:

• parted command — A Red Hat Linux utility called parted can be used to resize NTFS file systems. Because it runs in Linux, you need to be able to run it from a boot disk if you currently have a Windows-only system. You can download a boot disk image from the following location and create a boot disk with that image using the boot-disk–creation descriptions in this chapter:

  ftp://ftp.gnu.org/gnu/parted/bootdisk/

  The README file in this directory describes which boot image to use and how to use it. Documentation of this utility is available from /usr/share/doc/parted*/USER.

• Partition Magic — I've heard good reports from people using Partition Magic (www.partitionmagic.com/partitionmagic) to resize NTFS partitions. Partition Magic also helps you create new partitions and manage them. It supports Windows XP Professional/Home, Windows 95b-98SE, Windows Me, Windows 2000 Professional, and NT 4.0 workstation (SP6a). The cost is currently $70.

• Acronis OS Selector — This is another well-regarded product for managing, creating, and resizing partitions. It supports a variety of file system types, including FAT12, FAT16, FAT32, NTFS, and Linux partition types (ext2, ext3 and Linux ReiserFS). It also supports the same Windows platforms that Partition Magic does. The cost is currently $45 from www.acronis.com.

As I mentioned earlier, if your Windows partition is FAT or VFAT you can continue with the procedure in the next session to resize your hard disk.

Backing up and repartitioning

One way to divide a disk that is currently totally used for DOS (DOS, Windows 9x/2000, and so on) is not very complicated, but it can take some time. The method is to do a full system backup of your data, erase the whole disk, repartition the disk, restore your Windows 9x/2000 operating system to the hard disk (on a smaller partition), and add Red Hat Linux to the hard disk (using new partitions). Whew!

The problem with the backup/repartition approach is that it can be a pain. If your backup device is a 1.44MB floppy disk and your hard disk holds a gigabyte or more, a full backup is not a pleasant prospect. Alternatively, there is a DOS utility called FIPS.

Using the FIPS utility

The FIPS utility came about to solve the problem of a would-be Linux user with a monolithic DOS/Windows hard disk. With FIPS, you can change the size of your DOS partition without erasing it. Though this process is risky, and nobody recommends it without major "buyer beware" warnings, many use it safely and save a lot of time and effort.

FIPS works by changing the values in the partition table and boot sector. Space is gained by changing the partition table that is used to create a new primary DOS partition. After that, the new DOS partition can be converted to Linux partitions.

Split your DOS (Windows) disk into separate DOS and Linux partitions as follows:

1. Check your Windows/DOS partition.

   In a DOS window, run chkdsk to check for disk errors. The output shows how your disk space is used, as well as errors found and corrected. Next, run scandisk in Windows 9x/2000. Scandisk deals with lost file fragments, invalid filenames, and cross-linked files. It also scans for disk surface errors. The point is to correct hard disk errors encountered.

2. Create a bootable FIPS floppy.

   You need to create a bootable floppy disk. Insert a blank, 3.5-inch floppy disk in drive A. Then, from DOS (or a DOS window), type the following to create a boot floppy:

   format a:/s
   

   After creating the boot floppy, copy the following files from the first Red Hat Linux installation CD (CD #1) to this bootable floppy: restorrb.exe, fips.exe, and errors.txt. (These files are located in the dosutils/fips20 directory.) To make sure that the bootable floppy is working properly, insert the floppy disk into a PC that has a DOS (Windows 9x/2000) partition and reboot the computer. Now make sure that you can access the partition by typing dir C:\ at the prompt.

3. Defragment your hard disk.

   Reboot your PC so that Windows 9x/2000 starts up. To defragment your disk, so that all of your used space is put in order on the disk, open My Computer, right-click your hard disk icon (C:), select Properties, click Tools, and select Defragment Now.

   Defragmenting your disk can be a fairly long process. The result of defragmentation is that all the data on your disk are contiguous, creating a lot of contiguous free space at the end of the partition. There are cases where you will have to do the following special tasks to make this true:

   • If the Windows swap file is not moved during defragmentation, you must remove it. Then, after you defragment your disk again and run FIPS, you will need to restore the swap file. To remove the swap file, open the Control Panel, open the System icon, and then click the Performance tab and select Virtual Memory. To disable the swap file, click Disable Virtual Memory.

   • If your DOS partition has hidden files that are on the space you are trying to free up, you need to find them. In some cases, you won't be able to delete them. In other cases, such as swap files created by a program, you can safely delete those files. This is a bit tricky because some files should not be deleted, such as DOS system files. You can use the attrib -s -h command from the root directory to deal with hidden files.

4. Reboot (FIPS boot disk).

   Before you reboot to the FIPS disk, you want to make sure that nothing is written to the hard disk when DOS/Windows shuts down. Look for programs in config.sys and autoexec.bat that write to disk. When you are satisfied, insert the boot floppy and reboot.

5. Run FIPS.

   With your computer now booted in DOS (with C: drive accessible), run the FIPS program by typing:

   # fips
   

   You can quit FIPS at any time by pressing Ctrl+C. If you have more than one hard disk, FIPS will ask which one you want to use. FIPS then displays the partition table and asks you the number of the partition you want to split:

   • Type the number of the partition you want to split. FIPS checks the partition for free space and asks if you want to make a copy of your root and boot sector before you go on (this is recommended).

   • Type Y to backup your root and boot sector. FIPS asks if you have your FIPS bootable floppy in the drive.

   • Make sure the floppy is in the drive and type Y. FIPS copies a file called rootboot.000 to your A: drive. Then FIPS determines how much free space is available on the partition. If FIPS can't find at least one cylinder free, it probably means that you need to remove a mirror or image file (after you do that, you will have to defragment the drive again before you can rerun FIPS). If all goes well, you will see output that looks similar to the following:

      Old partition       Cylinder       New Partition
           2753.3 MB            351           35393.2 MB

     Note?

     If the New Partition doesn't show enough disk space, you may need to go back and remove some files until there is enough free space.

   • Use Right Arrow and Left Arrow keys to choose what cylinder the new partition should start on. As you press those keys, you can see the sizes of the old and new partitions change. When the partitions are the sizes you want, press Enter.

   • When the new partition table appears, you can accept it or re-edit the table.

   • Press C to continue if everything is okay. You are asked if you are ready to write the new partition table to disk.

   • Press Y to proceed. When FIPS is done, it will say Bye! and exit.

6. Reboot (FIPS boot disk) and test.

   Reboot from the DOS/FIPS disk again. Then run the fips command again, but this time in test mode as follows:

   # fips -t
   

   If the partition table looks okay, exit FIPS by pressing Ctrl+C. You can also use a program such as chkdsk to be sure the old partition is still working.

   Caution?

   If the partitions are not correct, you can restore your original partitioning by running the restorrb command from the FIPS floppy disk.

7. Restart your computer.

   At this point, you can restart your computer as you would normally (remove the FIPS boot disk). Try using the Windows 9x/2000 system to make sure that everything is working okay.

With partitioning done, you can install Red Hat Linux using the reclaimed space. (Don't assign the free disk space to anything yet. Just go ahead and install Red Hat Linux and assign the Linux partitions you need during installation.)

If you have a problem with FIPS, its creator asks that you make a transcript of the FIPS session (using the -d option) and send it to schaefer@rbg.informatik.th-darmstadt.de. The transcript will appear in the fipsinfo.dbg file.

Creating a floppy disk in Linux

Using the dd command, you can copy data from one device to another in Linux (or other UNIX system). To copy a disk image from the installation CD to a disk, you need:

• A blank 3.5-inch floppy disk (formatted).

• Write permission to the disk drive (probably drive A, which in Red Hat Linux is /dev/fd0). The root user should have write permission, but root would have to open write permission to allow another user to write to floppy disk.

• An available copy of the disk image to be copied. (The best way to get access to this in Red Hat Linux is by inserting the CD and typing mount /dev/cdrom, then typing cd /mnt/cdrom/images.)

To copy the bootdisk.img file from the images directory on the CD, type the following:

# dd if=bootdisk.img of=/dev/fd0 bs=1440k

Replace bootdisk.img with the name of any other image file you want to copy.

Creating a disk in DOS

The Red Hat CD has both image files and DOS utilities needed to create the installation disks while you are in DOS or Windows. From a DOS prompt, do the following:

1. Insert Red Hat CD #1 into the drive and change to the CD's drive letter (D: or E:).

2. Insert a blank, formatted 3.5-inch floppy disk into drive A.

3. Change to the dosutils directory (cd \dosutils).

4. Run the rawrite command as follows:

       C:\> rawrite
       Enter disk image source file name: D:\images\bootdisk.img
       Enter target diskette drive: a:
       Please insert a formatted diskette into drive A: and
              press -Enter- :

5. Press Enter. The image is written to the disk. Remove the disk and mark it.

Booting your computer with GRUB

With multiple operating systems installed and several partitions set up, how does your computer know which operating system to start? To select and manage which partition is booted and how it is booted, you need a boot loader. The boot loader that is installed by default with Red Hat Linux is called the GRand Unified Boot loader (GRUB).

GRUB is a GNU software package (www.gnu.org/software/grub) that replaced the LILO boot loader by default in Red Hat Linux 7.2. GRUB offers the following features:

• Support for multiple executable formats.

• Support for Multiboot operating systems (such as Red Hat Linux, FreeBSD, NetBSD, OpenBSD, and other Linux systems).

• Support for non-Multiboot operating systems (such as Windows 95, Windows 98, Windows NT, Windows ME, Windows XP, and OS/2) via a chain-loading function. Chain-loading is the act of loading another boot loader (presumably one that is specific to the proprietary operating system) from GRUB to start the selected operating system.

• Support for multiple file system types.

• Support for automatic decompression of boot images.

• Support for downloading boot images from a network.

For more information on how GRUB works, type man grub or info grub. The info command contains more details about GRUB.

Booting with GRUB

When you install Red Hat Linux, information needed to boot your computer (with one or more operating systems) is automatically set up and ready to go. Simply restart your computer. When you see the GRUB boot screen (it says GRUB at the top and lists bootable partitions below it), do one of the following:

• Default: If you do nothing, the default operating system will boot automatically after a few seconds.

• Select an operating system: Use the up and down arrow keys to select any of the operating systems shown on the screen. Then press Enter to boot that operating system.

• Edit the boot process: If you want to change any of the options used during the boot process, use the arrow keys to select the operating system you want and type e to select it. Follow the next procedure to change your boot options temporarily.

If you want to change your boot options so that they take effect every time you boot your computer, see the section on permanently changing boot options. Changing those options involves editing the /boot/grub/grub.conf file.

Temporarily changing boot options

From the GRUB boot screen, you can select to change or add boot options for the current boot session. First, select the operating system you want (using the arrow keys) and type e (as described earlier). You will see a graphical screen that contains information like the following:

GRUB version 0.93 (639K lower / 128768K upper memory)
   
root (hd0,0)
kernel /boot/vmlinuz-2.4.22-1 ro root=LABEL=/ hdc=ide-scsi
initrd /boot/initrd-2.4.22-1.img
   
Use the ( and ( keys to select which entry is highlighted.
Press 'b' to boot, 'e' to edit the selected command in the
boot sequence, 'c' for a command-line, 'o' to open a new line
after ('O' for before) the selected line, 'd' to remove the
selected line, or escape to go back to the main menu.

There are three lines in the example of the GRUB editing screen that identify the boot process for the operating system you chose. The first line (beginning with root) shows that the entry for the GRUB boot loader is on the first partition of the first hard disk (hd0,0). GRUB represents the hard disk as hd, regardless of whether it is a SCSI, IDE, or other type of disk. You just count the drive number and partition number, starting from zero.

The second line of the example (beginning with kernel) identifies the boot image (/boot/vmlinuz-2.4.22-1) and several options. The options identify the partition as initially being loaded ro (read-only) and the location of the root file system on a partition with the label LABEL=/. The third line (starting with initrd) identifies the location of the initial RAM disk, which contains the minimum files and directories needed during the boot process.

If you are going to change any of the lines related to the boot process, you would probably change only the second line to add or remove boot options. Here is how you do that:

1. Position the cursor on the kernel line and type e.

2. Either add or remove options after the name of the boot image. You can use a minimal set of bash shell command-line editing features to edit the line. You can even use command completion (type part of a filename and press Tab to complete it). Here are a few options you may want to add or delete:

   • Boot to a shell:???If you forgot your root password or if your boot process hangs, you can boot directly to a shell by adding init=/bin/sh to the boot line. (The file system is mounted read-only, so you can copy files out. You need to remount the file system with read/write permission to be able to change files.)

   • Select a run level:???If you want to boot to a particular run level, you can add the word linux, followed by the number of the run level you want. For example, to have Red Hat Linux boot to run level 3 (multiuser plus networking mode), add linux 3 to the end of the boot line. You can also boot to single-user mode (1), multi-user mode (2), or X GUI mode (5). Level 3 is a good choice if your GUI is temporarily broken.

3. Press Enter to return to the editing screen.

4. Type b to boot the computer with the new options. The next time you boot your computer, the new options will not be saved. To add options so they are saved permanently, see the next section.

Permanently changing boot options

You can change the options that take effect each time you boot your computer by changing the GRUB configuration file. In Red Hat Linux, GRUB configuration centers around the /boot/grub/grub.conf file.

The /boot/grub/grub.conf file is created when you install Red Hat Linux. Here is an example of a grub.conf file.

# grub.conf generated by anaconda
#
# Note that you do not have to rerun grub after making
# changes to this file
# NOTICE: You have a /boot partition.  This means that
#         all kernel and initrd paths are relative to /boot/, eg.
#         root (hd0,0)
#         kernel /vmlinuz-version ro root=/dev/hda3
#         initrd /initrd-version.img
#boot=/dev/hda
default=0
timeout=10
splashimage=(hd0,4)/grub/splash.xpm.gz
title Red Hat Linux (2.4.22-1)
     root (hd0,4)
     kernel /vmlinuz-2.4.22-1 ro root=LABEL=/
     initrd /initrd-2.4.22-1.img
title Windows XP
     rootnoverify (hd0,0)
     chainloader +1

The default=0 line indicates that the first partition in this list (in this case Red Hat Linux) will be the one that is booted by default. The line timeout=10 causes GRUB to pause for ten seconds before booting the default partition. (That's how much time you have to press e if you want to edit the boot line, or to press arrow keys to select a different operating system to boot.)

The splashimage line looks in the fifth partition on the first disk (hd0,4) for the boot partition (in this case /dev/hda5, which is the /boot partition). GRUB loads splash.xpm.gz as the image on the splash screen (/boot/grub/splash.xpm.gz). The splash screen appears as the background of the boot screen.

The two bootable partitions in this example are Red Hat Linux and Windows XP. The title lines for each of those partitions are followed by the name that appears on the boot screen to represent each partition.

For the Red Hat Linux system, the root line indicates the location of the boot partition as the second partition on the first disk. So, to find the bootable kernel (vmlinuz-2.4.22-1) and the initrd initial RAM disk boot image that is loaded (initrd-2.4.22-1.img), GRUB looks in the root of hd0,4 (which is represented by /dev/hda5 and is eventually mounted as /boot). Other options on the kernel line set the partition as read-only initially (ro) and set the root file system to /dev/hda6.

For the Windows XP partition, the rootnoverify line indicates that GRUB should not try to moun