1. Chapter 4 : File Systems What is a file system?
Objectives & user requirements
Characteristics of files & directories
File system implementation
Directory implementation
Free blocks management
Increasing file system performance

2. File System
The collection of algorithms and data structures which perform the translation from logical file operations (system calls) to actual physical storage of information

3. Objectives of a File System
Provide storage of data and manipulation
Guarantee consistency of data and minimise errors
Optimise performance (system and user)
Eliminate data loss (data destruction)
Support variety of I/O devices
Provide a standard user interface
Support multiple users

4. User Requirements Access files using a symbolic name
Capability to create, delete and change files
Controlled access to system and other users’ files
Control own access rights
Capability of restructuring files
Capability to move data between files
Backup and recovery of files

5. Files Naming Name formation
Extensions (Some typical extensions are shown below)

6. Files (Cont.) Structuring
(a) Byte sequence (as in DOS, Windows & UNIX)
(b) Record sequence (as in old systems)
(c) Tree structure (as in some mainframe Oses)

7. Files (Cont.) File types File access Regular (ASCII, binary)
Directories
Character special files
Block special files
File access
Sequential access
Random access

8. Files (Cont.) File attributes
Read, write, execute, archive, hidden, system etc.
Creation, last access, last modification

9. Files (Cont.) Create Delete Open Close Read Write Append Seek
File operations
Create
Delete
Open
Close
Read
Write
Append
Seek
Get attributes
Set Attributes
Rename

10. Directories Where to store attributes Path names Operations
In directory entry (DOS, Windows)
In a separate data structure (UNIX)
Path names
Absolute path name
Relative path name
Working (current) directory
Operations
Create, delete, rename, open directory, close directory, read directory, link (mount), unlink

11. Directories & Files (UNIX)/ d1 f4 f5 f7 d3 d4 d5 d6 f6
Disk A
Disk B
Root Directory
Linked Branch
Working Directory
Working directory : d2
Absolute path to file f2 : /d1/d2/f2
Relative path to file f2 : f2

12. Physical Disk Space Management
Sector
Heads
Cylinder
Track
Each plate is composed of sectors or physical blocks which are laid along concentric tracks
Sectors are at least 512 bytes in size
Sectors under the head and accessed without a head movement form a cylinder

13. File System Implementation
Contiguous allocation
Linked list allocation
Linked list allocation using an index (DOS file allocation table - FAT)
i-nodes (UNIX)

14. Contiguous Allocation
The file is stored as a contiguous block of data allocated at file creation
(a) Contiguous allocation of disk space for 7 files
(b) State of the disk after files D and E have been removed

15. Contiguous Allocation (Cont.)
FAT (file allocation table) contains file name, start block, length
Advantages
Simple to implement (start block & length is enough to define a file)
Fast access as blocks follow each other
Disadvantages
Fragmentation
Re-allocation (compaction)

16. Linked List Allocation
The file is stored as a linked list of blocks

17. Linked List Allocation (Cont.)
Each block contains a pointer to the next block
FAT (file allocation table) contains file name, first block address
Advantages
Fragmentation is eliminated
Block size is not a power of 2 because of pointer space
Disadvantages
Random access is very slow as links have to be followed

18. Linked list allocation using an index (DOS FAT)
FAT (File allocation table) 3 7 5 1 2 4 6
Disk size
EOF
Free 5
Free
File blocks 7
Bad
First block address is in directory entry 1 … n
Free

19. Linked list allocation using an index (Cont.)
The DOS (Windows) FAT is arranged this way
All block pointers are in FAT so that don’t take up space in actual block
Random access is faster since FAT is always in memory
16-bit DOS FAT length is ( )*2 = bytes

20. Problem
16-bit DOS FAT can only accommodate pointers (ie., a maximum of 64 MB disk)
How can we handle large disks such as a 4 GB disk?
Clustering

21. i (index)-nodes (UNIX)
File mode
Number of links
UID
GID
File size
Time created
Time last accessed
Time last modified
10 disk block numbers
Single indirect block
Triple indirect block
Double indirect block
Indirect blocks
Data blocks

22. i-nodes (Cont.)
Assume each block is 1 KB in size and 32 bits (4 bytes) are used as block numbers
Each indirect block holds 256 block numbers
First 10 blocks : file size <= 10 KB
Single indirect : file size <= = 266 KB
Double indirect : file size <= 256* = KB = MB
Triple indirect : file size <= 256*256* = KB = ~16 GB

23. Directory Implementation
DOS (Windows) directory structure
UNIX directory structure

24. DOS (Windows) Directory Structure (32 bytes)
File name
Ext A
Reserved T P D
Size
8 bytes 3 1 10 2 4
Time of creation
Date of creation
Attributes (A,D,V,S,H,R)
Pointer to first data block

25. UNIX Directory Structure (16 bytes)
I-node #
File name
2 bytes
14 bytes

26. The Windows 98 Directory Structure
Extended MS DOS Directory Entry
An entry for (part of) a long file name

27. The Windows 98 Directory Structure
An example of how a long name is stored in Windows 98

28. Path Name Lookup : /usr/ast/mbox
245
Root (/) i-node 1 . .. 4
bin 7
dev 14
lib 9
etc 6
usr
Root directory file block 245
132
i-node 6 of /usr 6 . 1 .. 19
prog 30
stu 51
html 26
ast 45
genc
/usr directory file block 132 26 . 6 .. 60
mbox 92
books 81
src
/usr/ast directory file block 406
i-node 60 of /usr/ast/mbox
406
i-node 26 of /usr/ast
Blocks of file

29. Two ways of handling long file names in a Directory
In-line In a heap

30. Shared Files / d1 d2 f1 f2 f3
File f2 is shared by two paths (users!) and there is one physical copy.
The directories d1 & d2 point to the same i-node with link count equal to 2
Deletion is done by decrementing the link count. When it reaches zero the file is deleted physically

31. Disk Space Management
Block size
Dark line (left hand scale) gives data rate of a disk
Dotted line (right hand scale) gives disk space efficiency
All files 2KB

32. How to Keep Track of Free Disk Blocks
Linked list of disk blocks
Bit maps
Indexing as used in DOS FAT

33. Linked List of Disk Blocks
Allocation is simple.
Delete block number from free blocks list

34. Bit Maps
The bit map is implemented by reserving a bit string whose length equals the number of blocks
A ‘1’ may indicate that the block is used and ‘0’ for free blocks
If the disk is nearly full then the bit map method may not be as fast as the linked list method

35. Increasing File System Performance
Disks (floopies, hard disks, CD ROMS) are still slow when compared to the memory
Use of a memory cache may speed the disk transfers between disk and process
Blocks are read into the cache first. Subsequent accesses are through the cache
Blocks are swapped in & out using replacement algorithms such as FIFO, LRU
System crashes may cause data loss if modified blocks are not written back to disk

36. Where to Put the Current “File Position” Field
The file position field is a 16 or 32 bit variable which holds the address of the next byte to be read or written in a file
Put it in the i-node
Put it in process table

37. File Position Field in i-node
If two or more processes share the same file, then they must have a different file position
Since i-node is unique for a file, the file position can not be put in the i-node

38. File Position Field in Process Table
When a process forks, both the parent and the child must have the same file position
Since the parent and the child have different process tables they can not share the same file position
So, we can not put in process table

39. Solution Use an intermediate table for file positions parent child
Process tables
position
File positions table
parent
child
i-node of
file