Partitions

The MBR, boot sectors and partition table

The information about how a hard disk has been partitioned is stored in its first sector (that is, the first sector of the first track on the first disk surface). The first sector is the master boot record (MBR) of the disk; this is the sector that the BIOS reads in and starts when the machine is first booted. The master boot record contains a small program that reads the partition table, checks which partition is active (that is, marked bootable), and reads the first sector of that partition, the partition's boot sector (the MBR is also a boot sector, but it has a special status and therefore a special name). This boot sector contains another small program that reads the first part of the operating system stored on that partition (assuming it is bootable), and then starts it.

The partitioning scheme is not built into the hardware, or even into the BIOS. It is only a convention that many operating systems follow. Not all operating systems do follow it, but they are the exceptions. Some operating systems support partitions, but they occupy one partition on the hard disk, and use their internal partitioning method within that partition. The latter type exists peacefully with other operating systems (including Linux), and does not require any special measures, but an operating system that doesn't support partitions cannot co-exist on the same disk with any other operating system.

As a safety precaution, it is a good idea to write down the partition table on a piece of paper, so that if it ever corrupts you don't have to lose all your files. (A bad partition table can be fixed with fdisk). The relevant information is given by the fdisk -l command:

$ fdisk -l /dev/hda

Disk /dev/hda: 15 heads, 57 sectors, 790 cylinders
Units = cylinders of 855 * 512 bytes

   Device Boot  Begin   Start     End  Blocks   Id  System
/dev/hda1           1       1      24   10231+  82  Linux swap
/dev/hda2          25      25      48   10260   83  Linux native
/dev/hda3          49      49     408  153900   83  Linux native
/dev/hda4         409     409     790  163305    5  Extended
/dev/hda5         409     409     744  143611+  83  Linux native
/dev/hda6         745     745     790   19636+  83  Linux native
$

Extended and logical partitions

The original partitioning scheme for PC hard disks allowed only four partitions. This quickly turned out to be too little in real life, partly because some people want more than four operating systems (Linux, MS-DOS, OS/2, Minix, FreeBSD, NetBSD, or Windows/NT, to name a few), but primarily because sometimes it is a good idea to have several partitions for one operating system. For example, swap space is usually best put in its own partition for Linux instead of in the main Linux partition for reasons of speed (see below).

To overcome this design problem, extended partitions were invented. This trick allows partitioning a primary partition into sub-partitions. The primary partition thus subdivided is the extended partition; the subpartitions are logical partitions. They behave like primary [1] partitions, but are created differently. There is no speed difference between them.

The partition structure of a hard disk might look like that in Figure 4-2. The disk is divided into three primary partitions, the second of which is divided into two logical partitions. Part of the disk is not partitioned at all. The disk as a whole and each primary partition has a boot sector.

Partition types

The partition tables (the one in the MBR, and the ones for extended partitions) contain one byte per partition that identifies the type of that partition. This attempts to identify the operating system that uses the partition, or what it uses it for. The purpose is to make it possible to avoid having two operating systems accidentally using the same partition. However, in reality, operating systems do not really care about the partition type byte; e.g., Linux doesn't care at all what it is. Worse, some of them use it incorrectly; e.g., at least some versions of DR-DOS ignore the most significant bit of the byte, while others don't.

There is no standardization agency to specify what each byte value means, but some commonly accepted ones are included in in Table 4-1. The same list is available in the Linux fdisk program.

Partitioning a hard disk

There are many programs for creating and removing partitions. Most operating systems have their own, and it can be a good idea to use each operating system's own, just in case it does something unusual that the others can't. Many of the programs are called fdisk, including the Linux one, or variations thereof. Details on using the Linux fdisk are given on its man page. The cfdisk command is similar to fdisk, but has a nicer (full screen) user interface.

When using IDE disks, the boot partition (the partition with the bootable kernel image files) must be completely within the first 1024 cylinders. This is because the disk is used via the BIOS during boot (before the system goes into protected mode), and BIOS can't handle more than 1024 cylinders. It is sometimes possible to use a boot partition that is only partly within the first 1024 cylinders. This works as long as all the files that are read with the BIOS are within the first 1024 cylinders. Since this is difficult to arrange, it is a very bad idea to do it; you never know when a kernel update or disk defragmentation will result in an unbootable system. Therefore, make sure your boot partition is completely within the first 1024 cylinders.

Some newer versions of the BIOS and IDE disks can, in fact, handle disks with more than 1024 cylinders. If you have such a system, you can forget about the problem; if you aren't quite sure of it, put it within the first 1024 cylinders.

Each partition should have an even number of sectors, since the Linux filesystems use a 1 kilobyte block size, i.e., two sectors. An odd number of sectors will result in the last sector being unused. This won't result in any problems, but it is ugly, and some versions of fdisk will warn about it.

Changing a partition's size usually requires first backing up everything you want to save from that partition (preferably the whole disk, just in case), deleting the partition, creating new partition, then restoring everything to the new partition. If the partition is growing, you may need to adjust the sizes (and backup and restore) of the adjoining partitions as well.

Since changing partition sizes is painful, it is preferable to get the partitions right the first time, or have an effective and easy to use backup system. If you're installing from a media that does not require much human intervention (say, from CD-ROM, as opposed to floppies), it is often easy to play with different configuration at first. Since you don't already have data to back up, it is not so painful to modify partition sizes several times.

There is a program for MS-DOS, called fips, which resizes an MS-DOS partition without requiring the backup and restore, but for other filesystems it is still necessary.

Device files and partitions

Each partition and extended partition has its own device file. The naming convention for these files is that a partition's number is appended after the name of the whole disk, with the convention that 1-4 are primary partitions (regardless of how many primary partitions there are) and 5-8 are logical partitions (regardless of within which primary partition they reside). For example, /dev/hda1 is the first primary partition on the first IDE hard disk, and /dev/sdb7 is the third extended partition on the second SCSI hard disk. The device list in XXX (device list) gives more information.