cs431-cotter1 file systems tanenbaum chapter 4 silberschatz chapters 10, 11, 12

cs431-cotter 1

File SystemsFile Systems

Tanenbaum Chapter 4

Silberschatz Chapters 10, 11, 12

cs431-cotter 2

Essential requirements for long-term information storage:

• It must be possible to store a very large amount of information.

• The information must survive the termination of the process using it.

• Multiple processes must be able to access the information concurrently.

File Systems

Tanenbaum, Modern Operating Systems 3 e, (c) 2008 Prentice-Hall, Inc. All rights reserved. 0-13-6006639

cs431-cotter 3

File Structure

• None:– File can be a sequence of words or bytes

• Simple record structure:– Lines– Fixed Length– Variable Length

• Complex Structure:– Formatted documents– Relocatable load files

• Who decides?

cs431-cotter 4

Think of a disk as a linear sequence of fixed-size blocks and supporting reading and writing of blocks. Questions that quickly arise:

• How do you find information?• How do you keep one user from reading another’s data?• How do you know which blocks are free?

File Systems


cs431-cotter 5Figure 4-1. Some typical file extensions.

File Naming


cs431-cotter 6

File Access Methods

• Sequential Access– Based on a magnetic tape model– read next, write next– reset

• Direct Access– Based on fixed length logical records– read n, write n– position to n– relative or absolute block numbers

cs431-cotter 7

Figure 4-2. Three kinds of files. (a) Byte sequence. (b) Record sequence. (c) Tree.

File Structure


cs431-cotter 8Figure 4-3. (a) An executable file. (b) An archive.

File Types


cs431-cotter 9

Figure 4-4a. Some possible file attributes.

File Attributes


cs431-cotter 10

The most common system calls relating to files:

File Operations


• Append• Seek• Get Attributes• Set Attributes• Rename

• Create• Delete• Open • Close• Read• Write

cs431-cotter 11Figure 4-5. A simple program to copy a file.

Example Program Using File System Calls (1)


. . .

cs431-cotter 12

Figure 4-5. A simple program to copy a file.

Example Program Using File System Calls (2)


cs431-cotter 13

Directory Structure

• Collection of nodes containing information on all files

F1F2

F3

F4 F5

cs431-cotter 14

Information in a Device Directory

• File name:• File Type:• Address:• Current Length• Maximum Length• Date Last accessed (for archiving)• Date Last updated (for dumping)• Owner ID• Protection information

cs431-cotter 15

Directory Operations

• Search for a file• Create a file• Delete a file• List a directory• Rename a file• Traverse the file system

cs431-cotter 16

Objectives for a Directory System

• Make it efficient– It should be easy to locate a file quickly

• Make file (and directory) naming convenient– Allow 2 users to have the same name for different files– Allow the same file to have more than 1 name

• Allow logical grouping of files– All word processing files together– All c++ files together– etc.

cs431-cotter 17

Alternative Directory Structures

• Single-Level Directory

• Issues:– Naming– Grouping

cat bo a test data mail cont hex word calc

cs431-cotter 18

Alternative Directory Structures

• Two-Level Directory

User1 User2 User3

cs431-cotter 19

Tree-Structured Directory

cs431-cotter 20

Figure 4-8. A UNIX directory tree.

Path Names


cs431-cotter 21

File System Structure• A data structure on a disk that holds files• Disk Storage attributes:

– Allows direct access to any given block of information– Supports re-writing in place (copy block, modify,

rewrite)

• Organized in Layers:– Application Programs– Logical file system– file-organization module– basic file system– I/O control– devices

cs431-cotter 22

File System Organization

• Devices– Peripheral equipment- disk drives, tapes, etc.

• I/O Control– device drivers and interrupt handlers– Communicates directly with peripheral devices– Starts and closes out I/O request

• Basic File System– Requires knowledge of physical device (track,

cylinder, head, etc.)– Manages data transfer at block level (buffering,

placement)

cs431-cotter 23

File System Organization• File-organization Module

– Maps logical block structure of file to physical block structure of disk

– Includes free space manager

• Logical file system– Manages file directory information – Provides security, protection

• Applications– Maintains basic data about the file– Allows users to access file information

cs431-cotter 24

File System Structure

• Access to File– file control block (generic)– file descriptor (UNIX)– file handle (Windows)

• File System Requirements– boot mechanism– file system information– file information– file data

cs431-cotter 25

Allocation MethodsContiguous Allocation

• Each file occupies a set of contiguous blocks on the disk.

• Number of blocks needed identified at file creation– May be increased using file extensions

• Advantages:– Simple to implement– Good for random access of data

• Disadvantages– Files cannot grow– Wastes space

cs431-cotter 26

Contiguous Allocation

0 1 2 3 4

5 6 7 8 9

10 11 12 13 14

15 16 17 18 19

20 21 22 23 24

25 26 27 28 29

30 31 32 33 34

FileA

FileB

FileC

FileE

FileD

File Allocation Table

File Name Start Block Length

FileA

FileBFileCFileDFileE

2 39 5

18 830 226 3

FileA

cs431-cotter 27

Allocation MethodsLinked Allocation

• Each file consists of a linked list of disk blocks.

• Advantages:– Simple to use (only need a starting address)– Good use of free space

• Disadvantages:– Random Access is difficult

ptrdata ptrdata ptrdata Nulldata

cs431-cotter 28

Linked Allocation

0 1 2 3 4

5 6 7 8 9

10 11 12 13 14

15 16 17 18 19

20 21 22 23 24

25 26 27 28 29

30 31 32 33 34

FileB File Allocation Table

File Name Start Block End

... ... ...

......FileB 28

...1

cs431-cotter 29

Linked Allocation

0 1 2 3 4

5 6 7 8 9

10 11 12 13 14

15 16 17 18 19

20 21 22 23 24

25 26 27 28 29

30 31 32 33 34

FileB File Allocation Table

File Name Start Block End

... ... ...

......FileB 28

...1

cs431-cotter 30

Allocation MethodsIndexed Allocation

• Collect all block pointers into an index block.

• Advantages:– Random Access is easy– No external fragmentation

• Disadvantages– Overhead of index block

Index Table

cs431-cotter 31

Indexed Allocation

1

83

14

28

0 1 2 3 4

5 6 7 8 9

10 11 12 13 14

15 16 17 18 19

20 21 22 23 24

25 26 27 28 29

30 31 32 33 34


File Name Index Block

Jeep 24

cs431-cotter 32

Indexed Allocation

1831428

0 1 2 3 4

5 6 7 8 9

10 11 12 13 14

15 16 17 18 19

20 21 22 23 24

25 26 27 28 29

30 31 32 33 34


File Name Index Block

Jeep 24

cs431-cotter 33

UNIX File SystemFile Types

• Regular File

• Directory

• Block-oriented Device

• Character-oriented Device

• Symbolic Link

cs431-cotter 34

UNIX File Ownership

read write execute

owner

group

others

1 0

00

00

1

1

1

(for directories: read, write, search)

-rwx------ 1 rsmith is 785 Feb 11 1994 send_exedrwxr-xr-x 2 rsmith is 512 Apr 16 13:49 sysmon_server

cs431-cotter 35

File Access Model

• open -- read -- write -- close

fdesc = open (*filename, r / w / a / +);n = read (fdesc, buff, 24);n = write (fdesc, buff, 24);lseek (fdesc, 100L, L_SET);

(L_SET, L_CUR, L_END)

close (fdesc);

cs431-cotter 36

Concurrent File Access

• From separate processes– OPEN generates a new file descriptor.– Allows independent access to the same file.– No inherent blocking in access

• From parent / child processes– fork duplicates all active file descriptors. – all processes access a common “current” pointer

• File locking– flock, lockf

cs431-cotter 37

UNIX File Sharing- unrelated processes -

fd flags ptrfd 0

process table entry

fd flags ptrfd 0

process table entry

File Table

file status flagscurrent file offsetv-node ptr


V-node Table

v-node infoI-node infofile size

cs431-cotter 38

UNIX File Sharing - related processes -

fd flags ptrfd 0

process table entry

fd flags ptrfd 0

process table entryfile status flagscurrent file offsetv-node ptr

File Table


V-node Table

v-node infoI-node infofile size

cs431-cotter 39

direct blocks

UNIX i-node

modeowners(2)

timestamps(3)size block

count

single indir

triple indir

double indir

data

data

data

cs431-cotter 40

direct blocks

UNIX i-node

modeowners(2)


count

single indir

triple indir

double indir

data

data

data

data

data::

cs431-cotter 41

direct blocks

UNIX i-node

modeowners(2)


count

single indir

triple indir

double indir

data

data

data

data

data::

::

::

::

data

data

data

data

cs431-cotter 42

I-node File Sizes

1 KB block 2KB block 4KB block

Direct 12 k bytes 24 k bytes 48 k bytes

Single indirect

256 k bytes 1 M bytes 4 M bytes

Double indirect

64 M bytes 512 M bytes

2 G bytes

Triple indirect

16 G bytes 256 G bytes

2 T bytes

cs431-cotter 43

Figure 4-20. Percentage of files smaller than a given size (in bytes).

Disk Space Management Block Size (1)


cs431-cotter 44

Disk Drive Layout(simplified)

DiskDrive

partition partition partition

I-list directory blocks and data blocks

I-node I-node I-node I-node

Block Group

Partition Block group Block group Block group

cs431-cotter 45

Sample filesystem: Creating the directory “testdir”

i-nodei-node1267

i-nodei-node2549

i-listdirectoryblock

cs431-cotter 46

i-nodei-node1267

i-nodei-node2549


1267 •


cs431-cotter 47

i-nodei-node1267

i-nodei-node2549


2549 testdir

1267 •


cs431-cotter 48

i-nodei-node1267

i-nodei-node2549


2549 testdir

1267 •


cs431-cotter 49

i-nodei-node1267

i-nodei-node2549


directoryblock

2549 •

1267 • •

2549 testdir

1267 •


cs431-cotter 50

Sample filesystem after creating the directory “testdir”

i-nodei-node1267

i-nodei-node2549


directoryblock

2549 •

1267 • •

2549 testdir

1267 •

cs431-cotter 51

Viewing directory structure

• ls –ai /– 2 .– :– 1730 home

• ls –ai /home– 1730 .– 2 . .– :– 199 rcotter

• ls –ai /home/rcotter– 199 .– 1730 . .– :– 241224 cs431

• ls –ai ~/cs431– 241224 .– 199 . .– :– 245436 exec.cpp

cs431-cotter 52

Shared Files (1)


Figure 4-16. File system containing a shared file.

ln –s shared_file new_link

cs431-cotter 53

Figure 4-17. (a) Situation prior to linking. (b) After the link is created. (c) After the original owner removes the file.

Shared Files (2)


cs431-cotter 54

Linux File System Structure• Linux uses a Virtual File System (VFS)

– Defines a file object– Provides an interface to manipulate that object

• Designed around OO principles– File system object– File object– Inode object (index node)

• Primary File System - ext2fs– Supports (or maps) several other systems

(MSDOS, NFS (network drives), VFAT (W95), HPFS (OS/2), etc.

cs431-cotter 55

Virtual Filesystem• A kernel software layer that handles all system calls

related to a standard UNIX filesystem.• Supports:

– Disk-based filesystems• IDE Hard drives (UNIX, LINUX, SMB, etc.)• SCSI Hard drives• floppy drives

– Network filesystems• remotely connected filesystems

– Special filesystems• /proc

cs431-cotter 56

VF Exampleinf = open (“/floppy/test”,

O_RDONLY, 0);

outf = open (“/tmp/test”, O_WRONLY|O_CREATE|O_TRUNC, 0600);

do {

cnt = read(inf, buf, 4096);

write (outf, buf, cnt);

} while (cnt);

close (outf);

close (inf);

ext2 MS-DOS

VFS

cp

cs431-cotter 57

Virtual file system Model

• Superblock object– metadata about a mounted filesystem

• inode object– general information about a specific file

• file object– information about process and open file interaction

• dentry object– directory entry information

cs431-cotter 58

Virtual Filesystem and Processes

disk file

Process 1

Process 2

Process 3

file object

dentryobject

dentryobject

file object

file object

superblock inode object

cs431-cotter 59

Ext2fs History

• Minux

• extfs

• ext2fs - 1994

cs431-cotter 60

Ext2fs Blocks

• 1k default

• 2k, 4k possible

• choice based on file sizes

cs431-cotter 61

Ext2fs Disk Data Structures

• Each Block Group contains:– A copy of the Super Block– A copy of the group of block group descriptors– A data block bitmap (for this group)– An inode bitmap (for this group) – A group of inodes (for this group)– Data blocks

BootBlock

Block Group 0 Block Group N

SuperBlock

GroupDescriptors

Data blockBitmap

InodeBitmap

InodeTable

Data Blocks

cs431-cotter 62

Disk Data Structures

• SuperBlock– Contains metadata about the block group– # free blocks, – # free inodes, – block size, – times (last mount, last write), – check status, – blocks to pre-allocate, – alignments, etc. (~38 items)

cs431-cotter 63


• Block Groups– Block group limited to 8 * block size by data

block bitmap.– 1k block limited to 8,192 blocks or 8 mbytes

per block group. 4k block = 128 mbytes

• Group descriptor - 24 bytes per group– block numbers for bitmaps, start of tables, etc.

cs431-cotter 64


• Inode Table– Inodes - 128 bytes– File type and access rights– owner– length– timestamps (last access, last change, etc.)– hard links counter– pointers to data blocks,– etc.

cs431-cotter 65


• Items cached in memory on filesystem mount– Superblock - always cached– Group descriptor - always cached– Block bitmap - fixed limit - most recent only– Inode bitmap - fixed limit - most recent only– Inode - dynamic– Data block - dynamic

cs431-cotter 66

File Blocks Grouped

• Physically and logically

• Inodes distributed throughout disk

• Data blocks allocated close to inode

• Pre-allocation of free blocks to minimize frag.

cs431-cotter 67

Summary Features

• Fast Symbolic Links

• pre-allocation of data blocks

• file-updating strategy

• immutable files or append only files

cs431-cotter 68

Journal File Systems

• Designed to speed file system recovery from system crashes.

• Traditional systems use an fs recovery tool (e.g. fsck) to verify system integrity

• As file system grows, recovery time grows. • Journal File Systems track changes in meta-data

to allow rapid recovery from crashes

cs431-cotter 69

Journal File System

• Uses a version of a transaction log to track changes to file system directory records.

• If a system crash occurs, the transaction log can be reviewed back to a check point to verify any pending work (either undo or redo).

• For large systems, recovery time goes from hours or days to seconds.

cs431-cotter 70

Journal File System - Examples

• XFS– Commercial port by SGI (IRIX OS)– Based on 64 bit system (ported to 32 bit)– Supports large systems 2TB -> 9 million TB– Uses B+trees for improved performance– Journals file system meta-data – Supports Quotas, ACLs.– Supports filesystem extents (contiguous blocks)

cs431-cotter 71


• JFS– Commercial Port by IBM. Used in OS/2– 64 bit system ported to 32 bits.– Journal support of meta-data.– System Size 2TB -> 32PB– Built to scale on SMP architectures– Uses B+trees for improved performance– Supports use of extents.

cs431-cotter 72


• ReiserFS– Designed originally for Linux (32 bit system)– Supports file systems to 2TB -> 16TB– Journal support of meta-data– Btree structure supports large file counts– Supports block packing for small files– No fixed inode allocation - more flexible.

cs431-cotter 73


• Ext3fs– Simple extension of ext2fs that adds journaling– All virtual file system operations are journaled.– Other limitations of ext2fs remain.

cs431-cotter 74


• ext4fs– Next generation of file system development, incorporating

improvements and changes to ext3fs.– Journalling filesystem

• Improvements is journal checksumming– Larger file systems – 1 MTB (exabyte)

• Larger files – 16 TB– Delayed block allocation, multi-block allocator

• Allows files to be written as contiguous sequences of blocks: improves read performance of streaming data files.

• Poses the risk of lost data if system crashes before data is written.– Use of extents

• Large (<= 128 MB) contiguous physical blocks.

cs431-cotter 75

Free Space Management• Bit Vector management

• One bit for each block– 0 = free; 1 = occupied

• Use bit manipulation commands to find free block

• Bit vector requires space– block size = 4096 = 2 12

– disk size = 1 gigabyte = 2 30

– bits = 2 (30-12) = 2 18 = 32k bytes

0 1.....11 0

cs431-cotter 76

Free Space Management

• Bit vector (advantages):– Easy to find contiguous blocks

• Bit vector (disadvantages):– Wastes space (bits allocated to unavailable blocks)

• Issues:– Must keep bit vector on disk (reliability)– Must keep bit vector in memory (speed)– cannot allow memory != disk (memory = 1, disk = 0)

cs431-cotter 77


• Linked List management– Use linked list to identify free space

• Advantages:– no wasted space

• Disadvantages:– harder to identify contiguous space.

• Issues:– Must protect pointer to free list

cs431-cotter 78


• Grouping of blocks

Bootblock

Filesystem

descriptor

Filedescriptors

cs431-cotter 79

Figure 4-29. (a) I-nodes placed at the start of the disk. (b) Disk divided into cylinder groups, each with its own blocks and i-nodes.

Reducing Disk Arm Motion


cs431-cotter 80Figure 4-22. (a) Storing the free list on a linked list. (b) A bitmap.

Keeping Track of Free Blocks (1)


cs431-cotter 81

Figure 4-24. Quotas are kept track of on a per-user basis in a quota table.

Disk Quotas


cs431-cotter 82

Backups to tape are generally made to handle one of two potential problems:

• Recover from disaster.• Recover from stupidity.

File System Backups (1)


cs431-cotter 83

Figure 4-25. A file system to be dumped. Squares are directories, circles are files. Shaded items have been modified since last dump. Each directory and file is labeled by its i-node number.



cs431-cotter 84

Figure 4-26. Bitmaps used by the logical dumping algorithm.



cs431-cotter 85

Figure 4-27. File system states. (a) Consistent. (b) Missing block. (c) Duplicate block in free list. (d) Duplicate data block.

File System Consistency


cs431-cotter 86

Figure 4-28. The buffer cache data structures.

Caching (1)


cs431-cotter 87

• Some blocks, such as i-node blocks, are rarely referenced two times within a short interval.

• Consider a modified LRU scheme, taking two factors into account:

•Is the block likely to be needed again soon?•Is the block essential to the consistency of the file system?

Caching (2)


cs431-cotter 88

Example File Systems

• CD-ROM– ISO 9660– Rock Ridge– Joliet

• MS-DOS

cs431-cotter 89

Figure 4-30. The ISO 9660 directory entry.

The ISO 9660 File System


cs431-cotter 90

Rock Ridge extension fields:

• PX - POSIX attributes.• PN - Major and minor device numbers.• SL - Symbolic link.• NM - Alternative name.• CL - Child location.• PL - Parent location.• RE - Relocation.• TF - Time stamps.

Rock Ridge Extensions


cs431-cotter 91

Joliet extension fields:

• Long file names.• Unicode character set.• Directory nesting deeper than eight levels.• Directory names with extensions

Joliet Extensions


cs431-cotter 92

Figure 4-31. The MS-DOS directory entry.

The MS-DOS File System (1)


cs431-cotter 93

Figure 4-32. Maximum partition size for different block sizes. The empty boxes represent forbidden combinations.

The MS-DOS File System (2)


cs431-cotter 94

Summary

• File Structures

• Directory Structures

• File System Implementation

• File System Management

• Example File Systems

cs431-cotter 95

Questions• Two primary access methods are used to manage files.

Discuss these two methods, including which types of physical devices each is normally associated with.

• What is the difference between a tree structured file directory and an acyclic graph directory? Please give an example Operating System used today that uses each.

• Discuss 3 ways in which files might be allocated to file stores (disks). What are the advantages or disadvantages of each?

• How might a file system manage its free space so that it can easily assign space for a new file when needed? Give at least 2 approaches.

• What can an OS designer to improve the efficiency of a file system? What factors should be considered? How do those affect the efficiency of the system?

• Discuss at least 2 mechanisms that might be used to ensure the recovery of information stored in a file system.

cs431-cotter1 file systems tanenbaum chapter 4 silberschatz chapters 10, 11, 12

Documents

file structure tanenbaum

file attributes tanenbaum

file operations tanenbaum

file creat

executable file

file naming tanenbaum

file types tanenbaum

cs431cotter3 file structure