Perhaps one of the most important decisions an embedded developer makes is which file system(s) to deploy. Some file systems optimize for performance, whereas others optimize for size. Still others optimize for data recovery after device or power failure. This chapter introduces the major file systems in use on Linux systems and examines the characteristics of each as they apply to embedded designs. It is not the intent of this chapter to examine the internal technical details of each file system. Instead, this chapter examines the operational characteristics and development issues related to each file system presented. References in Section 9.11.1, 'Suggestions for Additional Reading,' are provided at the end of the chapter for the interested reader.
Starting with the most popular file system in use on earlier Linux desktop distributions, we introduce concepts using the Second Extended File System (ext2) to lay some foundation for further discussion. Next we look at its successor, the Third Extended File System (ext3), which is the default file system for many popular desktop Linux distributions being shipped today.
After introducing some fundamentals, we examine a variety of specialized file systems, including those optimized for data recovery and for storage space, and those designed for use on Flash memory devices. The Network File System (NFS) is presented, followed by a discussion of the more important Pseudo File Systems, including the proc file system and sysfs.
9.1. Linux File System Concepts
Before delving into the details of the individual file systems, let's look at the big picture of how data is stored on a Linux system. In our study of device drivers in Chapter 8, 'Device Driver Basics,' we looked at the structure of a character device. In general, character devices store and retrieve data in serial streams. The most basic example of a character device is a serial port or magnetic tape drive. In contrast, block devices store and retrieve data in equal-sized chucks of data at a time. For example, a typical IDE hard disk controller can transfer 512 bytes of data at a time to and from a specific, addressable location on the physical media. File systems are based on block devices.
9.1.1. Partitions
Before we begin our discussion of file systems, we start by introducing partitions, the logical division of a physical device upon which a file system exists. At the highest level, data is stored on physical devices in partitions. A partition is a logical division of the physical medium (hard disk, Flash memory) whose data is organized following the specifications of a given partition type. A physical device can have a single partition covering all its available space, or it can be divided into multiple partitions to suit a particular task. A partition can be thought of as a logical disk onto which a complete file system can be written.
Figure 9-1 shows the relationship between partitions and file systems.
Figure 9-1. Partitions and file systems
Linux uses a utility called fdisk to manipulate partitions on block devices. A recent fdisk utility found on many Linux distributions has knowledge of more than 90 different partition types. In practice, only a few are commonly used on Linux systems. Some common partition types include Linux, FAT32, and Linux Swap.
Listing 9-1 displays the output of the fdisk utility targeting a CompactFlash device connected to a USB port. On this particular target system, the USB subsystem assigned the CompactFlash physical device to the device node /dev/sdb.
Listing 9-1. Displaying Partition Information Using fdisk
# fdisk /dev/sdb
Command (m for help): p
Disk /dev/sdb: 49 MB, 49349120 bytes
4 heads, 32 sectors/track, 753 cylinders
Units = cylinders of 128 * 512 = 65536 bytes
Device Boot Start End Blocks Id System
/dev/sdb1 * 1 180 11504 83 Linux
/dev/sdb2 181 360 11520 83 Linux
/dev/sdb3 361 540 11520 83 Linux
/dev/sdb4 541 753 13632 83 Linux
For this discussion, we have created four partitions on the device using the fdisk utility. One of them is marked bootable, as indicated by the asterisk in the column labeled Boot. This is simply the setting of a flag in the data structure that represents the partition table on the device. As you can see from the listing, the logical unit of storage used by fdisk is a cylinder.[69] On this device, a cylinder contains 64KB. On the other hand, Linux represents the smallest unit of storage as a logical block. You can deduce from this listing that a block is a unit of 1024 bytes.
After the CompactFlash has been partitioned in this manner, each device representing a partition can be formatted with a file system of your choice. When a partition is formatted with a given file system type, Linux can mount the corresponding file system from that partition.
9.2. ext2
Building on the example of Listing 9-1, we need to format the partitions created with fdisk. To do so, we use the Linux mke2fs utility. mke2fs is similar to the familiar DOS format command. This utility makes a file system of type ext2 on the specified partition. mke2fs is specific to the ext2 file system; other file systems have their own versions of these utilities. Listing 9-2 captures the output of this process.
Listing 9-2. Formatting a Partition Using mke2fs
# mke2fs /dev/sdb1 -L CFlash_Boot_Vol
mke2fs 1.37 (21-Mar-2005)
Filesystem label=CFlash_Boot_Vol
OS type: Linux
Block size=1024 (log=0)
Fragment size=1024 (log=0)
2880 inodes, 11504 blocks
575 blocks (5.00%) reserved for the super user
First data block=1
Maximum filesystem blocks=11796480
2 block groups
8192 blocks per group, 8192 fragments per group
1440 inodes per group
Superblock backups stored on blocks:
8193
Writing inode tables: done
Writing superblocks and filesystem accounting information: done
