Chapter 1 Introduction

advertisement

Chapter 7

UNIX and LINUX

Outline

• Overview

• Processes in UNIX

• Memory management in UNIX

• I/O in UNIX

• UNIX file system

2

Overview

• Why some programmers like UNIX better than

Windows?

– Although GUI may be easy for beginners, they provide little flexibility and no insight into how the system works!

– UNIX has GUI, too!

• History of UNIX

– UNICS in Bell Labs

– PDP-11 UNIX

– Portable UNIX, portable C compiler

– Berkeley UNIX (BSD)

– UNIX standards: POSIX from IEEE

– MINIX and Linux

3

UNIX Goals (Philosophy)

• Designed by programmers, for programmers

• Simple, elegant and consistent

• Power and flexibility

– A small number of basic elements

– One program should do just one thing

– An infinite combination to suit the applications

• No useless redundancy

– Simple interface

4

Interfaces to UNIX

User interface

Library interface

System call interface

Users

Standards utility programs

(shell, editors, compilers, etc)

Standard library

(e.g. open, close, read, write, fork, etc)

UNIX operating system

(process management, memory management, the file system, I/O, etc)

Hardware

(CPU, memory, disks, terminals, etc)

User mode

Kernel mode

5

The UNIX Shell: A Command

Line Interface

• Wait for a command line

• Extract the first word from the command line, use it as program name, run the program

• Pass arguments

• Flags: argument controlling the operation or specify an optional value

• Wild cards and magic characters

• Standard input/output

• Filter and pipe symbol

– grep ter *.t | sort | head –20 | tail –5 > foo

• Background execution and shell scripts

6

UNIX Utility Programs

• File and directory manipulation commands: cp, mv

• Filters: grep, head, sort, tail

• Program development tools: cc

• Text processing: vi, edit

• System administration: ps

• Miscellaneous: time

7

Structure of UNIX Kernel

System calls Interrupts and traps

Terminal handling Sockets

File naming

Network protocols

File systems

Mapping

Page faults

Signal handling

Virtual memory

Raw tty

Cooked tyy

Line disciplines

Routing

Buffer cache

Page cache

Process creation and termination

Process scheduling

Character devices

Network device drivers

Disk device drivers Process dispatching

Hardware

8

Outline

• Overview

• Processes in UNIX

• Memory management in UNIX

• I/O in UNIX

• UNIX file system

9

Basic Concepts

• UNIX is a multiprogramming system

– User processes

– Daemons

• Create new processes by system call fork()

– Parent process and child process

– PID, parent process gets child PID from fork()

• Pipes between two processes

– Shell pipelines are implemented using pipes

• Signals: soft interrupt

10

Process Management System

Calls in UNIX

pid=fork() Create a child process identical to the parent pid=waitpid(pid, &statloc, opts) Wait for a child to terminate s=execve(name, argv, envp) Replace a process’ core image exit(status) Terminate process execution and return status s=sigaction(sig, &act, &oldact) Define action to take on signals s=sigreturn(&context) Return from a signal s=sigprocmask(how, &set, &old) Examine or change the signal mask s=sigpending(set) s=sigsuspend(sigmask) s=kil(pid, sig)

Get the set of blocked signals

Replace the signal mask and suspend the process

Send a signal to a process residual=alarm(seconds) s=pause()

Sent the alarm clock

Suspend the caller until the next signal

11

Thread Management

• Not every UNIX has threads package

– Threads package becomes popular

• Threads can be implemented in either user space or kernel

– POSIX does not specify

12

Thread Management System Calls

Thread Call pthread_create pthread_exit pthread_join

Description

Create a new thread in the caller’s address space

Terminate the calling thread

Wait for a thread to terminate pthread_mutex_init Create a new mutex pthread_mutex_destroy Destroy a mutex pthread_mutex_lock Lock a mutex pthread_mutex_unlock Unlock a mutex pthread_cond_init Create a condition variable pthread_cond_destroy Destroy a condition variable pthread_cond_wait pthread_cond_signal

Wait on a condition variable

Release one thread waiting on a condition variable

13

Implementation of Processes

• Two key data structures for processes

• Process table: always in main memory

– Scheduling parameters, memory image, signals, miscellaneous

– A process has one entry in the table

• User structure: in memory only when the process is in memory

– Machine registers, system call state, file descriptor table, accounting info, kernel stack

– A block per process, adjacent to the stack segment

14

Executing “ ls” Typed to the Shell

PID=501 PID=748 PID=748 sh New process sh Same process sh

3. Exec call

1. Fork call

2. New sh

Fork code created

1.

Allocate child’s process table entry

2.

Fill child’s entry from parent

3.

Allocate child’s stack and user area

4.

Fill child’s user area from parent

5.

Allocate PID for child

6.

Set up child to share parent’s text

Exec code

4. Sh overlaid with ls

1.

Find the executable program

2.

Verify the execute permission

3.

Read and verify the header

4.

Copy arguments, environ to kernel

5.

Free the old address space

6.

Allocate new address space

7.

Copy page tables for data and stack

8.

Set up sharing of open files

9.

Copy parent’s registers to child

7.

Copy arguments, environ to stack

8.

Reset signals

9.

Initialize registers

15

Threads in UNIX

• Threads in user space: a user space library

• Threads in kernel: may lead to many problems

– How to maintain the correct traditional UNIX semantics?

– File I/O

– Signal handling

– All solutions to these problems cause something to break somewhere

16

Threads in Linux

• System call clone

– pid=clone(func, stack_ptr, sharing_flgs, arg)

• Sharing_flgs determine whether the thread is in current process or in a new process

– In the same process: changes visible to other threads

• The new thread executes func

• The new thread has its own stack

17

Scheduling in UNIX

• A two level algorithm

– Low-level algorithm: pick the process to run

– High-level algorithm: move processes between memory and disk

• Low-level algorithm uses multiple queues

– Priority=CPU_usage+nice+base

18

Multilevel Queue Structure

Highest

-4

Waiting for disk I/O

-3 Waiting for disk buffer

-2 Waiting for terminal input

2

3

Lowest

-1 Waiting for terminal output

0 Waiting for child to exist

0

1

User priority 0

User priority 1

User priority 2

User priority 3

Process waiting in kernel mode

Process waiting in user mode

19

Scheduling in Linux

• Scheduling is based on threads

• Three classes of threads

– Real-item FIFO, non-preemptive, highest priority

– Real-time round robin, preemptive, high priority

– Timesharing

• Scheduling based on priority and quantum

20

Booting UNIX

• Read & run first sector of the boot disk

– MBR has a small program (up to 512 bytes)

– Load program boot

• Boot reads the root directory of boot device

• Boot reads OS kernel, boot ends

• Kernel starts, set parameters & environment

• Initialization, build kernel data structure

• System auto-configuration, detect devices

• Load device drivers

• Start process 0

21

Process 0

• Continue initialization

• Program the real-time clock

• Mount the root file system

• Create init (process 1) and page daemon

(process 2)

22

Process 1

• Check flags of single-user/multi-user mode

• Single user mode

– Fork off a process running shell

– Wait for that process to exit

• Multi-user mode

– Fork off a process

• Run system initialization shell script /etc/rc

– Read /etc/ttys, fork off processes for terminals

• Run gtty

23

Sequence of Processes in Booting

Process 0 gett y

Process 1 ini t

Terminal 0

Login: login

Terminal 1

Password:

Page daemo n

Process 2 sh

Terminal 2

% cp f1 f2 cp

24

Loading Drivers

• Dynamically loading

– One binary code can be used everywhere

– Drivers are loaded dynamically, even through a network

• Non-dynamically loading

– Only the system administrator can make kernel binary code

– No one can insert any component into the kernel

25

Outline

• Overview

• Processes in UNIX

• Memory management in UNIX

• I/O in UNIX

• UNIX file system

26

Three Segments in UNIX

Processes

• Text segment: program’s executable code

– Read-only, size fixed after process is created

• Data segment: program’s variables, strings, and other data

– Initialized data and un-initialized data (BSS)

– Un-initialized global variables are in BSS, save space

– Size can change

• Stack segment

27

Sharing Between Processes

• Text segment & mapped file can be shared

28

Memory Management System Calls

• POSIX does not specify

• Common system calls s=brk(addr)

System call description

Change data segment size a=mmap(addr, len, prot, flags, fd, offset) Map a file in s=unmap(addr, len) Unmap a file

29

Swapping

• High level scheduling: swapper

• What cause swapping?

– Fork() needs memory for a child process

– Brk() needs to expand a data segment

– Stack needs to be expanded

• Which process will be swapped out?

– The ones blocked/waiting for I/O

– The ones with high (priority + residence time)

• The ones consuming much CPU/staying long in mem

30

Swapping Back Processes

• Check processes on disk every few seconds

– Find processes are ready

– Select the one staying on disk longest

• If have enough free memory (easy swap), bring that process in

• Otherwise, swap out some processes in main memory

• No process is swapped out until it stays in memory for 2 seconds

31

Paging in UNIX

• Do not use the working set model

• Done by kernel & page daemon (process 2)

• Never page out kernel and core map

– Core map records info about contents of the page frames

32

Page Replacement Algorithm

• Run every 250 msec by page daemon

• If insufficient free page frames, page daemon transfers pages to disk

• A modified version of the clock algorithm

33

Memory Management in Linux

• A three-level paging scheme

– Directory, middle, page + offset

• Buddy algorithm

32

32

64

32

16

16

32 32

16

8

8

16

8

8

8

8

8

8

32 32

32 32

8

4

4

8

8

8

4

4

8

8

8

4

4

16

34

Outline

• Overview

• Processes in UNIX

• Memory management in UNIX

• I/O in UNIX

• UNIX file system

35

Input/Output in UNIX

• Integrate devices into file system as special files

– Each I/O device is assigned a path name

• Block special files and character special files

36

Networking and Sockets

• Socket: interface to network

• Types of networking supported by sockets

– Reliable connection-oriented byte stream

– Reliable connection-oriented packet stream

– Unreliable packet transmission

Sending process

Receiving process

User space kernel space Socket

37

I/O System Calls in UNIX

• POSIX specifies function calls for terminal

• Function call ioctl is used in many UNIX

Function call Description s=cfsetospeed(&termios, speed) Set the output speed s=cfsetispeed(&termios, speed) Set the input speed s=cfgetospeed(&termios, speed) Get the output speed s=cfgetispeed(&termios, speed) Get the input speed s=tcsetattr(fd, opt, &termios) s=tcgetattr(fd, &termios)

Set the attributes

Get the attributes

38

Outline

• Overview

• Processes in UNIX

• Memory management in UNIX

• I/O in UNIX

• UNIX file system

39

UNIX File

• A sequence of bytes

– No distinction between ASCII, binary, or any other kinds of files

• Directories are stored as files

• Important directories in UNIX systems

Directory bin dev etc lib usr

Contents

Binary (executable) programs

Special files for I/O devices

Miscellaneous system files

Libraries

User directories

40

Mounting and Locking

• Only one file system (tree)

– One disk is mounted in another disk’s file tree

• UNIX allows user to lock files

– From single byte to an entire file

– Shared locks and exclusive locks

– A process specifies whether it wants to block if the lock cannot be placed

– Shared locked regions may overlap

41

System Calls About Files

System call fd=create(name, mode) fd=open(file, how, …) s=close(fd) n=read(fd, buffer, nbytes) n=write(fd, buffer, nbytes)

Description

One way to create a new file

Open a file for reading, writing or both

Close an open file

Read data from a file into a buffer

Write data from a buffer into a file position=lseek(fd, offset, whence) Move the file pointer s=stat(name, &buf) Get a file’s status information s=fstat(fd, &buf) Get a file’s status information s=pipe(&fd[0]) s=fcntl(fd, cmd, …)

Create a pipe

File locking and other operations

42

System Calls About Directories

System call s=mkdir(path, mode)

Description

Create a new directory s=rmdir(path) Remove a directory s=link(oldpath, newpath) Create a link to an existing file s=unlink(path) s=chdir(path) dir=opendir(path) s=closedir(dir) dirent=readdir(dir)

Rewinddir(dir)

Unlink a file

Change the working directory

Open a directory for reading

Close a directory

Read one directory entry

Rewind a directory so it can be reread

43

Disk Layout of Classical UNIX

• Boot block: contain code to boot computer

• Superblock: contain critical info about the layout of the file system

– # of i-nodes, # of disk blocks, the start of the list of free disk blocks, …

• I-nodes

• Data blocks

Boot block Super blk I-nodes Data blocks

44

Reading A File

• System call n=read(fd, buffer, nbytes)

• Start with file descriptor fd, find i-node

– Put pointer to the i-node in file descriptor?

• Only one current position for one file, recorded in inode

• Two processes open one file with different current position

– Put current position info in file descriptor?

• One file can have multiple current positions

• P1 and P2 open f at the same time, P2 want to write after P1 finishes

• P2 cannot get the correct current position

45

Open File description Table

46

Berkeley Fast File System

• Allow file names up to 255 characters

• Divide the disk up into cylinder groups

– Each with its own superblock, i-nodes and data blocks

– Whenever possible, blocks are allocated in the cylinder group containing the i-node

• Two block sizes

– Large files use large block size

47

Summary

• Three interfaces to UNIX:

– Shell, C library, and system calls

• Key concepts in UNIX

– Process, memory model, I/O, and file system

• Process management in UNIX

– Process table and user structure

• Memory model: text, data and stack

• I/O and file system

48

Download