SYSDA lecture 4

File systems- SYSDA lecture 4

File- a named unit of storage, existing persistently

Can be written, read or updated

e.g. word document

manipulated by system calls which are primitive operations:

create

delete

open

close

read

write

seek

current file position

open file, read 200 bytes, then read 100 bytes

create file- write 100 bytes, write 100 bytes

You can also:

Get attributes

Set attributes

Rename a file

Files contain metadata

Type, size, protection, time, date etc are all held regarding  a specific file but is not considered part of content

Directories

Special type of file

Contains a list of names of some other files, along with references to those files

e.g. home/pictures that contains files in pictures directory

unix ecample= usr/jim/a.text

goes to usr directory, then jim which is in usr and then in jim is the a.text file

pwd- command which tells user of the current directory that they are in

Special files in unix

Direct interfaces to i/o devices

i/o devices including but not limited to things such as keyboards, terminals all have special files

usually liv in the /dev directory

File systems

Fat

NTFS- more reliable than fat!

Extx- default in linux based systems

HFS- MAC OS

Analysing these

File system category- info about system as a whole

Content- information inside ordinary files

Metadata- info about file and location stored

Filename category- info about file naming

Application- journal information

Implementation of a linux based os

You have a super block

Inode table

Block table

Superblock- begins 1024 bytes from start of file system- 1024 bytes long

Contains global info about file system, including block size

Blocks-each file allocated a whole number of blocks

Block size is minimum space allocated for any non-empty file

Inodes- every file has exactly one inode

Means of referring to file or directory

Fixed-size

Inodes contain attributes including file, owner and permissions as well as references to blocks that hold content- useful for finding files

So

Contents of inode 2 could be stored in blocks 1 4 and 5

Directories

Has its own inode

Remember . means current directory

.. is parent

So can see the inodes of the files within the directory

Directory entries contain inodes (metadata) which contain reference to content blocks

Physical organisation

Each directory contains a pointer to next entry in directory

No of bytes from start of this entry to start of next

Deleted files are “skipped over”- may have details remaining even after deletion!

a.text

b.text

c.text

if b deleted then info remains but directory skips it and goes to C

So basically in inode table 2 there is the .directory which has inode number 471

So going to that inode will show the current “.” As 471 and then “..” which is the parent directory-

471 contains ref to passwd which is inode 893

Going to 893 will give start of text of the password file

Allocation

System is divided up into many block groups, used to improve locality of allocation of blocks and inodes

Each group has its own tables of inodes and blocks

Which can be allocated or unallocated state