Unix Pipes

Pipe sets up communication channel between two (related) processes.

37

Two processes connected by a pipe

One process writes to the pipe, the other reads from the pipe.

Looks exactly the same as reading from/to a file. System call:

int fd[2] ;pipe(&fd[0]) ;

fd[0] now holds descriptor to read from pipefd[1] now holds descriptor to write into pipe

Simple Example

#include <unistd.h> /* for Solaris */#include <fcntl.h>#include <stdio.h>char *message = "This is a message!!!" ;main(){ char buf[1024] ; int fd[2]; pipe(fd); /*create pipe*/ if (fork() != 0) { /* I am the parent */ write(fd[1], message, strlen (message) + 1) ; } else { /*Child code */ read(fd[0], buf, 1024) ; printf("Got this from MaMa!!: %s\n", buf) ; }}

Create a Pipeline

Sometimes useful to connect a set of processes in a pipeline.

Process

c

Process

DPipePipe

Process A writes to pipe AB, Process B reads from AB and writes to BC

Process C reads from BC and writes to CD …..

Inherited File Descriptors

Process created with three files: stdin, stdout, and stderr (file descriptors 0, 1, 2).

write(1, “EEEE”, 10) ; // goes to stdout (terminal) or printf

read(0, buf, 4) ; //read from terminal

Redirecting stdin and stdout

Can due on command line: ./a.out > tmp // write output to tmp rather than

terminal. ./a.out < foo // read from foo not terminal.

When processing in a pipeline useful to redirect stdout of pipe n to stdin of pipe n+1.

Key: When Unix allocates file descriptor, it always chooses the lowest available.

If close stdin and open new file, e.g., close(0), fd = open(“foo”, ….) then fd = 0 and all subsequent reads from stdin will read instead from “foo”.

main()

{

int fd ;

printf("this goes to stdout\n");

close(1); /* close stdout */

fd = open("foo",O_CREAT | O_WRONLY); /* fd = 1! */

printf("fd = %d\n", fd) ; //goes to file "foo"

printf("Should not see this either!\n") ;//ditto

}

Dup System Call

int dupfd = dup(fd); Duplicates a file descriptor -- “aliased”

Both point to same file, that being the argument to dup.

Reads/writes from fd and dupfd going to the same file.

Useful for situation such as: want to write some output to standard out,

then to a file, then back to standard out -- all using printf.

Redirecting I/O for pipeline.

pipeline (process1, process2) /* program names */

char *process1, *process2 ;

{ char buf[1024] ;

int fd[2];

pipe(&fd[0]); /*create pipe*/

if (fork() != 0) { /* parent */

close(fd[0]); //don't need.

close(STD_OUTPUT);

dup(fd[1]); /* sets stdout to this pipe end */

close(fd[1]) ; //don’t need. Already set stdout to pipe.

execl(process2,process2,0);

}

else {

/* child */

close(fd[1]);

close(STD_INPUT);

dup(fd[0]); /* replace stdin */

close(fd[0]);

execl(process1,process1,0);

}

fork()

exec exec

stdout

stdin

Single and Multithreaded Processes

Benefits

Responsiveness: Separate thread to handle user input.

Web server spawns separate thread to handle incoming request.

Resource Sharing: e.g., one code, data segment.

Economy: Much cheaper to create and switch than processes.

Utilization of MP Architectures: Assign each thread to a separate processor if available.

Lightweight process. Threads share all process resources.

State: Thread ID, program counter, register set, and stack.

User-level threads and kernel-level threads.

User Threads Thread management done by user-level

threads library. Advantages:

Very fast: Does not involve kernel in creation, scheduling, or switching.

Disadvantages: When a thread blocks, whole process blocks.

Kernel Threads Supported by the Kernel.

Kernel creates, schedules, and switches threads. Advantage:

When one thread blocks, the whole process does not have to block.

Thus can overlap I/O and computation. Disadvantage:

Slower since kernel is involved.

Multithreading Models Many-to-One

One-to-One

Many-to-Many

Many-to-One Many user-level threads mapped to

single kernel thread.

Used on systems that do not support kernel threads.

Many-to-One Model

One-to-One Each user-level thread maps to

kernel thread.

Examples- Windows 95/98/NT/2000- OS/2

One-to-one Model

Many-to-Many Model Allows many user level threads to be

mapped to many kernel threads. Allows the operating system to

create a sufficient number of kernel threads.

Solaris 2 Windows NT/2000 with the ThreadFiber package

Many-to-Many Model

Threading Issues Semantics of fork() and exec()

system calls. Thread cancellation.

Pthreads a POSIX standard (IEEE 1003.1c)

API for thread creation and synchronization.

API specifies behavior of the thread library, implementation is up to development of the library.

Common in UNIX operating systems.

Solaris 2 Threads

Solaris Process

Windows 2000 Threads Implements the one-to-one

mapping. Each thread contains

- a thread id- register set- separate user and kernel stacks- private data storage area