Linux with Operating System Concepts

Character
Type
-
Regular file
d
Directory
b
Block device
c
Character device
l
Symbolic link
p
Named pipe
s
Domain socket

---

The Linux File System
◾
399
the inode number dedicated to that file. The symbolic link is a pointer to a hard link stored 
elsewhere. As the user modifies the contents of the directory, this list is modified.
New files require new hard links pointing to newly allocated inodes. The deletion of a 
file causes the hard link to be removed and the numeric entry of hard links to a file to be 
decremented (e.g., in Figure 10.3, deleting the item disk in /dev would result in the hard 
link count being reduced to 4). The inode itself remains allocated to the given file unless 
the hard link count becomes 0.
10.3.5 inode
We now turn to the inode. When the file system is first established, it comes with a set 
number of inodes. The inode is a data structure used to store file information. The informa-
tion that every inode will store consists of
• The file type
• The file’s permissions
• The file’s owner and group
• The file’s size
• The inode number
• A timestamp indicating when the inode was last modified, when the file was created, 
and when the file was last accessed
• A link count (the number of hard links that point at this file)
• The location of the file (i.e., the device storing the file) and pointers to the individual 
file blocks (if this is a regular file), these pointers break down into
• A set number of pointers that point directly to blocks
FIGURE 10.3 
Long listings illustrating file types.

---

400
◾
Linux with Operating System Concepts
• A set number of pointers that point to indirect blocks; each indirect block con-
tains pointers that point directly to blocks
• A set number of pointers that point to doubly indirect blocks, which are blocks 
that have pointers that point to additional indirect blocks
• A set number of pointers that point to triply indirect blocks, which are blocks that 
have pointers that point to additional doubly indirect blocks
Typically, an inode will contain 15 pointers broken down as follows:
• 12 direct pointers
• 1 indirect pointer
• 1 double indirect pointer
• 1 triply indirect pointer
An inode is illustrated in Figure 10.4 (which contains 1 indirect pointer and 1 doubly 
indirect pointer but no triply indirect pointer because of a lack of space).
Let us take a look at how to access a Linux file through the inode. We will make a few 
assumptions. First, our Linux inode will store 12 direct pointers, 1 indirect pointer, 1 dou-
bly indirect pointer, and 1 triply indirect pointer. Blocks of pointers will store 12 pointers 
no matter whether they are indirect, doubly indirect, or triply indirect blocks. We will 
assume that our file consists of 500 blocks, numbered 0 to 499, each block storing 8 KB (the 
File inode
File size (bytes)
Indirect block
Doubly indirect block
Pointers to
additional
disk blocks
Pointers to
indirect
blocks
Pointers to
additional
disk blocks
Block 0 
Block 1 
Block 2 
Block 3 
Block 4
Block 5
Indirect block
Storage device ID
Owner UID
Group UID
Mode
Flags
Timestamp
Hard link count
File block pointers
FIGURE 10.4 
inode structure with pointers to disk blocks. (Adapted from Fox, R. 
Information 
Technology: An Introduction for Today’s Digital World
,
 
FL: CRC Press, 2013.)

---

The Linux File System
◾
401
typical disk block stores between 1 KB and 8 KB depending on the file system utilized). 
Our example file then stores 500 * 8 KB 
=
4000 KB or 4 MB. Here is the breakdown of how 
we access the various blocks.
• Blocks 0–11: direct pointers from the inode.
• Blocks 12–23: pointers from an indirect block pointed to by the inode’s indirect 
pointer.
• For the rest of the file, access is more complicated.
• We follow the inode’s doubly indirect pointer to a doubly indirect block. This 
block contains 12 pointers to indirect blocks. Each indirect block contains 12 
pointers to disk blocks.
− The doubly indirect block’s first pointer points to an indirect block of 12 
pointers, which point to blocks 24–35.
− The doubly indirect block’s second pointer points to another indirect block of 
12 pointers, which point to blocks 36–47.
− …
− The doubly indirect block’s last pointer points to an indirect block of 12 point-
ers, which point to blocks 156–167.
• We follow the inode’s triply indirect pointer to a triply indirect block. This block 
contains 12 pointers to doubly indirect blocks, each of which contains 12 point-
ers to indirect blocks, each of which contain 12 pointers to disk blocks. From the 
triply indirect block, we can reach blocks 168 through 499 (with room to increase 
the file to block 1895).
Earlier, we noted that the disk drive supports random access. The idea is that to track 
down a block, block i, we have a mechanism to locate it. This is done through the inode 
pointers as just described.
The above example is far from accurate. A disk block used to store an indirect, doubly 
indirect, or triply indirect block of pointers would be 8 KB in size. Such a sized block would 
store far more than 12 pointers. A pointer is usually 32 or 64 bits long (4 or 8 bytes). If we 
assume an 8 byte pointer and an 8 KB disk block, then an indirect, doubly indirect, or 
triply indirect block would store 8 KB/8 B pointers 
=
1 K or 1024 pointers rather than 12.
When a file system is created, it comes with a set number of inodes. The actual number 
depends on the size of the file system and the size of a disk block. Typically, there is 1 inode 
for every 2–8 KB of file system space. If we have a 1 TB file system (a common size for a 
hard disk today), we might have as many as 128 K (approximately 128 thousand) inodes. 
The remainder of the file system is made up of disk blocks dedicated to file storage and 
pointers. Unless nearly all of the files in the file system are very small, the number of inodes 
should be more than sufficient for any file system usage.

---

402
◾
Linux with Operating System Concepts
The system administrator is in charge of administering the file system. Rarely, if ever, 
will the system administrator have to worry about inodes or pointers. The Linux file system 
commands instead operate at a higher level of abstraction, allowing administrators and 
users alike to operate on the file entities (files, directories, links, etc.). It is the device driv-
ers, implemented by system programmers, which must deal with inodes and the location 
of disk blocks.
Let us consider, as an example, file creation. First, a new inode must be allocated from 
the available inodes. The next inode in order is used. The inode’s information is filled in, 
consisting of the file type, initial permissions (defaulting from the user’s umask value), 
owner and group of the user, the file system’s device number, and a timestamp for file 
creation. If the file is to store some initial contents, disk blocks are allocated and the direct 
pointers are modified in the inode to point at these blocks. Finally, the content can be saved 
to those blocks. Additionally, the directory is modified to store the new hard link.
If we were instead to move a file within the same partition, then all we have to do is trans-
fer the hard link from one directory to another. Copying a file requires a new inode and 
disk blocks with the original copy’s blocks copied into the new copy’s blocks. Deleting a file 
requires the removal of the hard link from the directory along with a decrement to the hard 
link count. If this reaches zero, the inode is returned to the collection of free inodes. Disk 
blocks themselves are not modified (erased) but instead are returned to free blocks to be 
reused in the future. If a new file is created, the returned inode might be reused at that time.
10.3.6 Linux Commands to Inspect inodes and Files
There are several tools available to inspect inodes. The following commands provide such 
information:
• 
stat
—provides details on specific file usage, the option 
–c %i
displays the file’s 
inode number
• 
ls
—the option –i displays the inodes of all entries in the directory
• 
df –i
—we explore df in more detail in the next section but this command provides 
information on the utilization of the file system, partition by partition. The -i option 
includes details on the number of inodes used.
We wrap up this section by examining the information obtainable from stat. The stat 
command itself will respond with the name of the file, the size of the file, the blocks used 
to store the file, the device storing the file (specified as a device number), the inode of the 
file, the number of hard links to the file, the file’s permissions, UID, GID in both name and 
number, and the last access, modification, and change date and time for the file. The stat 
command has many options. The most significant are listed here:
• -L—follow links to obtain inode information of files stored in other directories (with-
out this, symbolic links in the given directory are ignored)
• -f—used to obtain statistics on an entire file system rather than a file

---

The Linux File System
◾
403
• -c