File Systems

File systems
 Files
 Directories & naming
 File system implementation
 Example file systems
 Long-term information storage
 Must store large amounts of data
 Gigabytes -> terabytes -> petabytes
 Stored information must survive the termination of
the process using it
 Lifetime can be seconds to years
 Must have some way of finding it!
 Multiple processes must be able to access the
information concurrently

3
 Naming files
 Important to be able to find files after they’re created
 Every file has at least one name
 Name can be
 Human-accessible: “foo.c”, “my photo”, “Go Panthers!”, “Go Banana
Slugs!”
 Machine-usable: 4502, 33481
 Case may or may not matter
 Depends on the file system
 Name may include information about the file’s contents
 Certainly does for the user (the name should make it easy to figure out
what’s in it!)
 Computer may use part of the name to determine the file type

4
 Typical file extensions

CS 1550, cs.pitt.edu Chapter 6 5

(originaly modified by Ethan
 File structures

1 record
1 byte

12A 101 111

sab wm cm avg ejw sab elm br

Sequence of bytes Sequence of records

S02 F01 W02

Tree
6
 File types

Executable
file

Archive

7
 Accessing a file
 Sequential access
 Read all bytes/records from the beginning
 Cannot jump around
 May rewind or back up, however
 Convenient when medium was magnetic tape
 Often useful when whole file is needed
 Random access
 Bytes (or records) read in any order
 Essential for database systems
 Read can be …
 Move file marker (seek), then read or …
 Read and then move file marker

8
File attributes

9
 File operations
 Create: make a new file  Append: like write, but only
 Delete: remove an existing at the end of the file
file  Seek: move the “current”
 Open: prepare a file to be pointer elsewhere in the file
accessed  Get attributes: retrieve
 Close: indicate that a file is attribute information
no longer being accessed  Set attributes: modify
 Read: get data from a file attribute information
 Write: put data to a file  Rename: change a file’s
name

10
Using file system calls

11
Using file system calls, continued

12
 Memory-mapped files

Program Program
text text abc
Data Data xyz

Before mapping After mapping

 Segmented process before mapping files into its address

space
 Process after mapping
 Existing file abc into one segment
 Creating new segment for xyz

13
 More on memory-mapped files
 Memory-mapped files are a convenient abstraction
 Example: string search in a large file can be done just as
with memory!
 Let the OS do the buffering (reads & writes) in the virtual
memory system
 Some issues come up…
 How long is the file?
 Easy if read-only
 Difficult if writes allowed: what if a write is past the end of file?
 What happens if the file is shared: when do changes
appear to other processes?
 When are writes flushed out to disk?
 Clearly, easier to memory map read-only files…
14
 Directories
 Naming is nice, but limited
 Humans like to group things together for
convenience
 File systems allow this to be done with directories
(sometimes called folders)
 Grouping makes it easier to
 Find files in the first place: remember the enclosing
directories for the file
 Locate related files (or just determine which files are
related)

15
 Single-level directory systems
Root
directory

A A B C
foo bar baz blah

 One directory in the file system

 Example directory
 Contains 4 files (foo, bar, baz, blah)
 owned by 3 different people: A, B, and C (owners shown in red)
 Problem: what if user B wants to create a file called foo?

16
 Two-level directory system
Root
directory

A B C

A A B B C C C
foo bar foo baz bar foo blah

 Solves naming problem: each user has her own directory

 Multiple users can use the same file name
 By default, users access files in their own directories
 Extension: allow users to access files in others’ directories

17
 Hierarchical directory system
Root
directory

A B C

A A A B B C C C
Papers foo Photos foo Papers bar foo blah

A A A B B
os.tex sunset Family foo.tex foo.ps

A A A
sunset kids Mom
18
 Unix directory tree

CS 1550, cs.pitt.edu Chapter 6 19

(originaly modified by Ethan
 Operations on directories
 Create: make a new  Readdir: read a directory
directory entry
 Delete: remove a directory  Rename: change the name
(usually must be empty) of a directory
 Opendir: open a directory  Similar to renaming a file
to allow searching it  Link: create a new entry in
 Closedir: close a directory a directory to link to an
(done searching) existing file
 Unlink: remove an entry in
a directory
 Remove the file if this is the
last link to this file

20
 File system implementation issues
 How are disks divided up into file systems?
 How does the file system allocate blocks to files?
 How does the file system manage free space?
 How are directories handled?
 How can the file system improve…
 Performance?
 Reliability?

21
 Carving up the disk

Entire disk
Partition table
Master
Partition 1 Partition 2 Partition 3 Partition 4
boot record

Boot Super Free space Index

Files & directories
block block management nodes

22
 Contiguous allocation for file blocks
A B C D E F

A Free C Free E F

 Contiguous allocation requires all blocks of a file to be

consecutive on disk
 Problem: deleting files leaves “holes”
 Similar to memory allocation issues
 Compacting the disk can be a very slow procedure…

23
 Contiguous allocation
 Data in each file is stored in 0 1 2 3
consecutive blocks on disk
 Simple & efficient indexing
 Starting location (block #) on disk
(start) 4 5 6 7
 Length of the file in blocks (length)
 Random access well-supported
 Difficult to grow files
 Must pre-allocate all needed space
 Wasteful of storage if file isn’t 8 9 10 11
using all of the space
 Logical to physical mapping is easy
blocknum = (pos / 1024) + start;
offset_in_block = pos % 1024;
Start=5
Length=2902

24
 Linked allocation
0 1 2 3
 File is a linked list of disk
4 6
blocks
 Blocks may be scattered
around the disk drive 4 5 6 7
 Block contains both pointer
x x
to next block and data
 Files may be as long as
needed 8 9 10 11
 New blocks are allocated as 0
needed
 Linked into list of blocks in
file Start=9 Start=3
 Removed from list (bitmap) End=4 End=6
of free blocks Length=2902 Length=1500
25
 Finding blocks with linked allocation
 Directory structure is simple
 Starting address looked up from directory
 Directory only keeps track of first block (not others)
 No wasted space - all blocks can be used
 Random access is difficult: must always start at first block!
 Logical to physical mapping is done by
block = start;
offset_in_block = pos % 1020;
for (j = 0; j < pos / 1020; j++) {
block = block->next;
}
 Assumes that next pointer is stored at end of block
 May require a long time for seek to random location in file

26
 Linked allocation using a RAM-based table
0 4
1 -1
 Links on disk are slow
2 -1  Keep linked list in memory
3 -2  Advantage: faster
4 -2
5 -1
 Disadvantages
6 3 B
 Have to copy it to disk at
7 -1 some point
8 -1  Have to keep in-memory and
9 0 A on-disk copy consistent
10 -1
11 -1
12 -1
13 -1
14 -1
15 -1

CS 1550, cs.pitt.edu Chapter 6 27

(originaly modified by Ethan
 Using a block index for allocation
 Store file block addresses in Name index size
an array grades 4 4802
 Array itself is stored in a disk
block
0 1 2 3
 Directory has a pointer to this
disk block
 Non-existent blocks indicated
by -1 6
4 5 6 7
 Random access easy 9
 Limit on file size? 7
0
8 8 9 10 11

CS 1550, cs.pitt.edu Chapter 6 28

(originaly modified by Ethan
 Finding blocks with indexed allocation
 Need location of index table: look up in directory
 Random & sequential access both well-supported:
look up block number in index table
 Space utilization is good
 No wasted disk blocks (allocate individually)
 Files can grow and shrink easily
 Overhead of a single disk block per file
 Logical to physical mapping is done by
block = index[block % 1024];
offset_in_block = pos % 1024;
 Limited file size: 256 pointers per index block, 1 KB
per file block -> 256 KB per file limit

29
 Larger files with indexed allocation
 How can indexed allocation allow files larger than a single
index block?
 Linked index blocks: similar to linked file blocks, but using
index blocks instead
 Logical to physical mapping is done by
index = start;
blocknum = pos / 1024;
for (j = 0; j < blocknum /255); j++) {
index = index->next;
}
block = index[blocknum % 255];
offset_in_block = pos % 1024;
 File size is now unlimited
 Random access slow, but only for very large files
CS 1550, cs.pitt.edu Chapter 6 30
(originaly modified by Ethan
 Two-level indexed allocation
 Allow larger files by creating an index of index blocks
 File size still limited, but much larger
 Limit for 1 KB blocks = 1 KB * 256 * 256 = 226 bytes = 64 MB
 Logical to physical mapping is done by
blocknum = pos / 1024;
index = start[blocknum / 256)];
block = index[blocknum % 256]
offset_in_block = pos % 1024;
 Start is the only pointer kept in the directory

 Overhead is now at least two blocks per file

 This can be extended to more than two levels if larger files

are needed...

CS 1550, cs.pitt.edu Chapter 6 31

(originaly modified by Ethan
 Block allocation with extents
 Reduce space consumed by index pointers
 Often, consecutive blocks in file are sequential on disk
 Store <block,count> instead of just <block> in index
 At each level, keep total count for the index for efficiency
 Lookup procedure is:
 Find correct index block by checking the starting file offset for each
index block
 Find correct <block,count> entry by running through index block,
keeping track of how far into file the entry is
 Find correct block in <block,count> pair
 More efficient if file blocks tend to be consecutive on disk
 Allocating blocks like this allows faster reads & writes
 Lookup is somewhat more complex

CS 1550, cs.pitt.edu Chapter 6 32

(originaly modified by Ethan
 Managing free space: bit vector
 Keep a bit vector, with one entry per file block
 Number bits from 0 through n-1, where n is the number of file blocks
on the disk
 If bit[j] == 0, block j is free
 If bit[j] == 1, block j is in use by a file (for data or index)
 If words are 32 bits long, calculate appropriate bit by:
wordnum = block / 32;
bitnum = block % 32;
 Search for free blocks by looking for words with bits unset
(words != 0xffffffff)
 Easy to find consecutive blocks for a single file
 Bit map must be stored on disk, and consumes space
 Assume 4 KB blocks, 8 GB disk => 2M blocks
 2M bits = 221 bits = 218 bytes = 256KB overhead

CS 1550, cs.pitt.edu Chapter 6 33

(originaly modified by Ethan
 Managing free space: linked list
 Use a linked list to manage free blocks
 Similar to linked list for file allocation
 No wasted space for bitmap
 No need for random access unless we want to find
consecutive blocks for a single file
 Difficult to know how many blocks are free unless
it’s tracked elsewhere in the file system
 Difficult to group nearby blocks together if they’re
freed at different times
 Less efficient allocation of blocks to files
 Files read & written more because consecutive blocks not
nearby

CS 1550, cs.pitt.edu Chapter 6 34

(originaly modified by Ethan
 Issues with free space management
 OS must protect data structures used for free space
management
 OS must keep in-memory and on-disk structures consistent
 Update free list when block is removed: change a pointer in the
previous block in the free list
 Update bit map when block is allocated
 Caution: on-disk map must never indicate that a block is free when it’s
part of a file
 Solution: set bit[j] in free map to 1 on disk before using block[j] in a file
and setting bit[j] to 1 in memory
 New problem: OS crash may leave bit[j] == 1 when block isn’t actually
used in a file
 New solution: OS checks the file system when it boots up…
 Managing free space is a big source of slowdown in file
systems
CS 1550, cs.pitt.edu Chapter 6 35
(originaly modified by Ethan
 What’s in a directory?
 Two types of information
 File names
 File metadata (size, timestamps, etc.)
 Basic choices for directory information
 Store all information in directory
 Fixed size entries
 Disk addresses and attributes in directory entry
 Store names & pointers to index nodes (i-nodes)

attributes
games attributes games
mail attributes mail attributes
news attributes news
attributes
research attributes research
attributes
Storing all information Using pointers to
in the directory index nodes
CS 1550, cs.pitt.edu Chapter 6 36
(originaly modified by Ethan
 Directory structure
 Structure
 Linear list of files (often itself stored in a file)
 Simple to program
 Slow to run
 Increase speed by keeping it sorted (insertions are slower!)
 Hash table: name hashed and looked up in file
 Decreases search time: no linear searches!
 May be difficult to expand
 Can result in collisions (two files hash to same location)
 Tree
 Fast for searching
 Easy to expand
 Difficult to do in on-disk directory
 Name length
 Fixed: easy to program
 Variable: more flexible, better for users

CS 1550, cs.pitt.edu Chapter 6 37

(originaly modified by Ethan
 Handling long file names in a directory

CS 1550, cs.pitt.edu Chapter 6 38

(originaly modified by Ethan
 Sharing files
Root
directory

A B C

A A A B B C C C
Papers foo Photos foo Photos bar foo blah

A A A B
os.tex sunset Family lake

A A ?
sunset kids ???
CS 1550, cs.pitt.edu Chapter 6 39
(originaly modified by Ethan
 Solution: use links
 A creates a file, and inserts into her directory
 B shares the file by creating a link to it
 A unlinks the file
 B still links to the file
 Owner is still A (unless B explicitly changes it)

A A B B

b.tex b.tex
a.tex a.tex

Owner: A Owner: A Owner: A

Count: 1 Count: 2 Count: 1

CS 1550, cs.pitt.edu Chapter 6 40

(originaly modified by Ethan
 Managing disk space

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
CS 1550, cs.pitt.edu Chapter 6 41
(originaly modified by Ethan
 Disk quotas

CS 1550, cs.pitt.edu Chapter 6 42

(originaly modified by Ethan
 Backing up a file system
 A file system to be dumped
 Squares are directories, circles are files
 Shaded items, modified since last dump
 Each directory & file labeled by i-node number

File that has

not changed

CS 1550, cs.pitt.edu Chapter 6 43

(originaly modified by Ethan
 Bitmaps used in a file system dump

CS 1550, cs.pitt.edu Chapter 6 44

(originaly modified by Ethan
 Checking the file system for consistency

Consistent Missing (“lost”) block

Duplicate block in free list Duplicate block in two files

CS 1550, cs.pitt.edu Chapter 6 45

(originaly modified by Ethan
 File system cache
 Many files are used repeatedly
 Option: read it each time from disk
 Better: keep a copy in memory
 File system cache
 Set of recently used file blocks
 Keep blocks just referenced
 Throw out old, unused blocks
 Same kinds of algorithms as for virtual memory
 More effort per reference is OK: file references are a lot less
frequent than memory references
 Goal: eliminate as many disk accesses as possible!
 Repeated reads & writes
 Files deleted before they’re ever written to disk

CS 1550, cs.pitt.edu Chapter 6 46

(originaly modified by Ethan
 File block cache data structures

CS 1550, cs.pitt.edu Chapter 6 47

(originaly modified by Ethan
 Grouping data on disk

CS 1550, cs.pitt.edu Chapter 6 48

(originaly modified by Ethan
 Log-structured file systems
 Trends in disk & memory
 Faster CPUs
 Larger memories
 Result
 More memory -> disk caches can also be larger
 Increasing number of read requests can come from cache
 Thus, most disk accesses will be writes
 LFS structures entire disk as a log
 All writes initially buffered in memory
 Periodically write these to the end of the disk log
 When file opened, locate i-node, then find blocks
 Issue: what happens when blocks are deleted?
CS 1550, cs.pitt.edu Chapter 6 49
(originaly modified by Ethan
 Unix Fast File System indexing scheme
protection mode data
data
owner & group
timestamps
...
size data
data
block count •
• ...
link count •
data
...
Direct pointers
•
•
•
• data
...
• •
single indirect • • data
• •
data
double indirect •
•
• • ...
triple indirect • • •
• • • data
inode
CS 1550, cs.pitt.edu Chapter 6 50
(originaly modified by Ethan
 More on Unix FFS
 First few block pointers kept in directory
 Small files have no extra overhead for index blocks
 Reading & writing small files is very fast!
 Indirect structures only allocated if needed
 For 4 KB file blocks (common in Unix), max file sizes are:
 48 KB in directory (usually 12 direct blocks)
 1024 * 4 KB = 4 MB of additional file data for single indirect
 1024 * 1024 * 4 KB = 4 GB of additional file data for double indirect
 1024 * 1024 * 1024 * 4 KB = 4 TB for triple indirect
 Maximum of 5 accesses for any file block on disk
 1 access to read inode & 1 to read file block
 Maximum of 3 accesses to index blocks
 Usually much fewer (1-2) because inode in memory

CS 1550, cs.pitt.edu Chapter 6 51

(originaly modified by Ethan
 Directories in FFS
 Directories in FFS are just Directory
special files inode number
 Same basic mechanisms record length
 Different internal structure name length
 Directory entries contain
 File name name
 I-node number
 Other Unix file systems inode number
have more complex record length
schemes name length
 Not always simple files…
name

CS 1550, cs.pitt.edu Chapter 6 52

(originaly modified by Ethan
 CD-ROM file system

CS 1550, cs.pitt.edu Chapter 6 53

(originaly modified by Ethan
 Directory entry in MS-DOS

CS 1550, cs.pitt.edu Chapter 6 54

(originaly modified by Ethan
 MS-DOS File Allocation Table

Block size FAT-12 FAT-16 FAT-32

0.5 KB 2 MB
1 KB 4 MB
2 KB 8 MB 128 MB
4 KB 16 MB 256 MB 1 TB
8 KB 512 MB 2 TB
16 KB 1024 MB 2 TB
32 KB 2048 MB 2 TB

CS 1550, cs.pitt.edu Chapter 6 55

(originaly modified by Ethan
 Windows 98 directory entry & file name

Bytes

Checksum

CS 1550, cs.pitt.edu Chapter 6 56

(originaly modified by Ethan
 Storing a long name in Windows 98

 Long name stored in Windows 98 so that it’s backwards

compatible with short names
 Short name in “real” directory entry
 Long name in “fake” directory entries: ignored by older systems
 OS designers will go to great lengths to make new systems
work with older systems…
CS 1550, cs.pitt.edu Chapter 6 57
(originaly modified by Ethan