Chapter 10 : Case Study - UNIX

1
Chapter 10 Case Study - UNIX

• History of unix
• Overview of unix
• Processes in unix
• Memory management in unix
• Input/output in unix
• The unix file system
• Security in unix
• Note This case study covers only UNIX. Please
  read chapter 10 of the text book for LINUX.

2
History of UNIX

• Originated from the MULTICS (Multiplexed
  Information and Computing Service) operating
  system by M.I.T, Bell Labs and GE
• There are two main versions
• ATT System V Release 4 (SVR4)
• Originally developed by ATT, now SCO
• BSD (Berkeley Software Distribution)

3
Overview of UNIX

• Supports various architectures
• Structure varies
• Supports preemptive multitasking
• Multiuser environment - generally secure
• Supports multithreaded applications
• Protection/Security is high on modern versions
• Supports symmetric multiprocessing
• Highly scalable/portable to various systems
• Many types/flavours of UNIX exist

4
UNIX Layers

• The layers of a UNIX system.

5
UNIX Utility Programs

• A few of the more common UNIX utility programs
  required by POSIX

6
UNIX Kernel (1)

• Approximate structure of generic UNIX kernel

7
UNIX Kernel (2)

• Bottom layer
• Device drivers for character and block devices
• Process dispatcher which stops the current
  process, saves its state and starts the
  appropriate driver when an interrupt occurs

8
Process Creation in UNIX - fork

• Process creation in UNIX.

9
POSIX Shell

• A highly simplified shell

10
The ls Command

• Steps in executing the command ls type to the
  shell

11
Some UNIX Process Concepts

• Daemons (background processes)
• Cron daemon which wakes up once a minute to check
  scheduled events (eg., disk backup)
• Pipes - syncronized channels between processes to
  pass byte streams
• Signals software interrupts used for
  interprocess communication. Choices catch,
  ignore, or kill process

12
Signals Required By POSIX

• The signals required by POSIX.

13
System Calls for Process Management

• s is an error code
• pid is a process ID
• residual is the remaining time from the previous
  alarm

14
Thread Calls in POSIX

• The principal POSIX thread calls.

15
Thread Calls in POSIX

• Threads were not in the first versions of UNIX
• There are many thread packages in use which are
  standardized in POSIX
• Thread calls are the same for user-space or
  kernel-space
• In kernel-space implementation calls are system
  calls
• In user-space implementation calls are to a
  run-time library

16
Thread Communication - mutexes

• Threads use locks called mutexes for short-time
  locking a resource (say a shared buffer)
• A mutex must be first created (and finally
  destroyed
• Mutual exclusion is implemented by locking a
  mutex before accessing a resource and unlock it
  when they are done (like binary semaphores which
  is 0 or 1, a mutex is either locked or
  unlocked)

17
Thread Communication condition variables

• For long-term synchronization (such as waiting
  for a tape to become free) condition variables
  are used
• Condition variables have to be created first and
  later destroyed like mutexes
• A condition variable is used by having one thread
  wait on it, and another thread signal it. If no
  thread is waiting when a signal is sent, the
  signal is lost

18
UNIX Scheduler (1)

• The UNIX scheduler is based on a multilevel queue
  structure (highest priority queue first,
  round-robin in each queue)
• In this scheme, a process which was blocked and
  waiting for an event joins the appropriate queue
  when blocking is over (a process whose disk I/O
  is finished joins,say, queue 4)

19
UNIX Scheduler (2)

• Once a second the priority of all processes are
  recalculated to avoid starvation using
• priority CPU_usage nice base
• CPU_usage, represents the average number of clock
  ticks per second that the process has had during
  the past few seconds
• Nice is a value between 20 to 20 (default 0).
  Nice system call can be used to set this value
  0-20
• Base is a system parameter in UNIX source code
• The scheduler forces CPU bound (on positive
  queues) get any service that is left over when
  all I/O bound and interactive processes are
  blocked

20
Booting UNIX (1)

• The first sector of the boot disk (master boot
  record) is read in and executed
• This sector loads the boot program
• Boot reads root directory, loads kernel and
  starts its execution
• Kernel reads the rest of the operating system
  (main C-code section)
• C code does some initialization, allocates system
  data structures, loads device drivers and
  handcrafts the first process, process 0

21
Booting UNIX (2)
cp

• The sequences of processes used to boot some
  systems

22
Handling Memory

• Each process has three segments (shown as one
  segment in the figure, but if hardware supports
  they can be separate)
• Text executable code (which is shared in the
  figure)
• Data variables, strings, arrays etc.
• initialized data variables which must be
  initialized to some value when program starts
• Uninitialized data (BSS) not initialized but
  has value 0 as default
• Stack
• Text is fixed in length, data and stack can grow
  and shrink

23
Paging in UNIX (1)

• Prior to 3BSD, UNIX systems used swapping (if
  memory is full, swap processes to disk)
• To run a process, all that is needed is the user
  structure and page table. The pages of the text,
  data and stack segments are brought in on demand

24
Paging in UNIX (2) Core Map

• Main memory kernel, core map, pages
• Core map has an entry for each page and contains
  information about the contents of the page frames

25
Page Replacement Algorithm (1)

• The page replacement algorithm is executed by the
  page daemon (process 2)
• Page daemon wakes up every 250 msec and transfers
  pages to disk if the amount of free memory is
  less than the system parameter lotsfree
  (typically set to ¼ of memory)
• Page daemon uses the two-handed clock algorithm

26
Basic Clock Algorithm (1)

• The pointer (hand) points to the oldest page
• When a page fault occurs,
• if the R bit of the pointed page is 0 (page not
  referred), this page is evicted (the new page
  replaces this page - written to disk first if it
  is dirty)
• if the R bit is 1 (page accessed), R bit is
  cleared and the hand is advanced to the next page

27
Two-handed Clock Algorithm (2)

• Page daemon has to do two passes with one core
  map pointer. Pass 1 clears all R bits, second
  pass removes pages (R bits are set between pass 1
  and 2)
• Page daemon maintains two pointers into the core
  map to speed up the process (one pass instead of
  two) for large memories
• When page daemon runs, it first clears the R bit
  at the front hand, and then checks the R bit at
  the back hand, after which it advances both hands
• Each time the page daemon runs, the hands rotate
  less than a full revolution, the amount depending
  on the number of pages needed to reach lotsfree

28
I/O in UNIX

• All I/O devices are integrated into the file
  system as special files
• These special files are accessed like ordinary
  files (ie., file operations such as read, write,
  open are the same for special files

29
Networking in UNIX (1)

• Sockets are used to establish a connection
  between network nodes

30
Networking in UNIX (2)

• Sockets are created and destroyed dynamically
• Creating a socket returns a file descriptor,
  which is needed for establishing a connection,
  reading data, writing data, and releasing the
  connection
• One party makes a listen call on a local socket,
  which creates a buffer and blocks until data
  arrive
• The other party makes a connect call giving as
  parameters the file descriptor of the local
  socket and the address of a remote socket (a
  sockets has an address in the network like the
  internet)
• Once a connection is established, a socket
  functions like a pipe

31
UNIX I/O (1)

• When a user accesses a special file, the file
  system determines the major and minor device
  numbers and whether it is a block or character
  special file
• Major device number is used to index into either
  bdevsw array for block special or cdevsw for
  character special files
• These structures contain pointers to the
  procedures to open the device, read, write etc.,
  Some of the fields of a typical cdevsw table are
  shown below

32
UNIX I/O (2)

• The UNIX I/O system in BSD

33
UNIX I/O (3)

• For block special files (eg., disks) the blocks
  are cached in a buffer cache
• The buffer cache works for both reads and writes
• Usually dirty (modified) blocks are written to
  the disk in every 30 seconds
• For character special devices, data is buffered
  in a chain of C-lists. A C-list block is 64
  characters long, plus a count and a pointer to
  the next block (BSD method of character buffering)

34
The UNIX File System (1)

• Some important directories found in most UNIX
  systems

35
The UNIX File System (2)

• Before linking.
• After linking.

(a) Before linking. (b) After linking
36
The UNIX File System (3)

• Separate file systems
• After mounting

37
Locking Files in UNIX (1)

• Accessing a file by several processes need some
  critical section management
• This is done by locks
• A lock is defined by a file name, the starting
  byte and the number of bytes
• When placing a lock, the process specifies to
• Block when the existing block is removed, the
  process is unblocked and the lock is placed
• Not to block the system call returns with a
  status code telling whether the lock succeeded or
  not

38
Locking Files (2)

• (a) File with one lock
• (b) Addition of a second lock
• (c) A third lock

39
System Calls for File Management

• s is an error code (-1 if an error has occured)
• fd is a file descriptor (a positive number 0
  standard input)
• position is a file offset

40
The stat System Call

• Fields returned by the stat system call

41
System Calls for Directory Management

• s is an error code
• dir identifies a directory stream
• dirent is a directory entry

42
UNIX File System (1)

• Disk layout in classical UNIX systems

• Block 0 is the boot block
• Block 1 is the superblock which contains
  information about the layout of the file system,
  including the number of i-nodes, number of disk
  blocks, and start of the list of free disk blocks

43
UNIX File System (2)

• Directory entry fields.

Structure of the i-node in System V
44
UNIX File System (3)
File descriptor table is indexed by the fd
parameter and has one entry for each file

• The relation between the file descriptor
  table, the open file description

45
UNIX File System (4)

• A BSD directory with three files.The same
  directory after the file voluminous has been
  removed

• File name can be 255 characters long
• The first 4 fields are fixed length

46
Security in UNIX (1)

• Each UNIX user has a UID (User ID). A UID is an
  integer between 0 and 65536. Files, processes and
  other resources are marked with the UID of their
  owner
• The user with UID 0 is the superuser
• Users can be organized in groups, which are also
  numbered with 16-bit GIDs (Group ID)

47
Security in UNIX (2)

• Some examples of file protection modes

48
System Calls for File Protection

• s is an error code
• uid and gid are the UID and GID, respectively