Lecture 5
●
Lecture files in /home/hwang/cs375/lecture05
on csserver.
cp -r /home/hwang/cs375/lecture05 .
scp -r user@csserver.evansville.edu:/home/hwang/cs375/lecture05 .
●
Project 1 posted, due next Thursday
●
Questions?
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
1
Outline
●
Exercises from last lecture
●
Working with files
●
File Descriptors
●
Read, Write, Open and Close
●
File Management
●
File Information
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
2
Working with Files
●
●
C, C++, Java, etc. each provide some means of
file input/output. Each of these languages build
upon the lower-level file access routines
provided by the OS.
The system routines provided by UNIX for
file/device I/O are: open, close, read, write,
and ioctl. They are documented in section 2 of
the man pages, e.g. man 2 write.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
3
Why Use Low-Level Routines?
●
●
●
Normally we use the file access methods
provided by our programming language (I/O
streams in C++, for example). They are more
efficient and portable.
BUT, interprocess communication (IPC)
requires that we work with the OS routines.
Writing kernel code (i.e., device drivers) ALSO
requires knowledge of the low-level routines.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
4
File Descriptors
●
●
●
Information about every open file is kept in a
system-wide file table. (A file may be opened
multiple times with different access modes;
there would then be multiple entries in the
system file table.)
Every process has a file table that contains
indexes into the system file table.
Indexes into the per-process file table are called
file descriptors.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
5
File Descriptors
Per Process File Table
File
Descriptors
Thursday, September 10
System File Table
0
1
2
3
4
5
6
CS 375 UNIX System Programming - Lecture 5
6
File Descriptors
●
●
●
Programs usually (automatically) have standard
input, standard output, and standard error open.
These are associated with file descriptors 0, 1,
and 2, respectively.
You can associate other file descriptors with
other files and devices by using the open
system routine.
System I/O routines require a file descriptor as
an argument in the routine call.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
7
Example: Byte I/O
// File: bytcat.cpp
// byte at a time cat from stdin using system calls
#include <unistd.h>
int main()
{
char c;
ssize_t nread, nwrite;
while ((nread = read(0, &c, sizeof(c))) != ­1) {
if (nread == 0)
break; // break on end­of­file
if((nwrite = write(1, &c, nread)) == ­1)
break; // break on write error
}
if (nread == ­1 || nwrite == ­1)
return 1; // error occurred
return 0;
}
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
8
Example: Block I/O
// File: blkcat.cpp
// block cat from stdin using system calls
#include <unistd.h>
int main()
{
const int BLKSZ = 512;
char c[BLKSZ];
ssize_t nread, nwrite;
while ((nread = read(0, c, BLKSZ)) != ­1) {
if (nread == 0)
break; // break on end­of­file
if((nwrite = write(1, c, nread)) == ­1)
break; // break on write error
}
if (nread == ­1 || nwrite == ­1)
return 1; // error occurred
return 0;
}
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
9
Read and Write
●
●
There is no guarantee that read will read the
number of bytes requested. (Fewer bytes will
be read when near the end of the file. Also,
only a single line of data is read when reading
from a terminal, not 512 bytes.) Notice the
request is made to write nread bytes and not
BLKSZ bytes.
write may also write fewer bytes than
requested, but this usually can be attributed to
more serious problems.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
10
Open and Close
●
The open call is used to associate a file
descriptor with a physical filename.
// Required header files
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
// open returns a file descriptor (an int)
// Use this form
int open(const char *pathname, int flags);
// or this one.
int open(const char *pathname, int flags, mode_t mode);
●
The close call closes a file descriptor.
int close(int fd);
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
11
Open and Close
●
●
●
flags should be O_RDONLY, O_WRONLY, or
O_RDWR possibly bitwise or'd (|) with
O_CREAT, O_EXCL, O_TRUNC, etc.
mode (optional) specifies permissions (see the
man page).
There is also a creat (not create!) call:
int creat(const char *pathname, mode_t mode);
// this is equivalent to int open(const char *pathname, O_CREAT|O_WRONLY|O_TRUNC, mode_t mode);
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
12
Using lseek
●
lseek is used to reposition the file position
pointer associated with a file descriptor:
#include <sys/types.h>
#include <unistd.h>
off_t lseek(int fd, off_t offset, int whence);
●
offset is the # of bytes to move. whence is
either SEEK_SET, SEEK_CUR, or SEEK_END
to move relative to the start of file, the current
position, or the end of the file (offset may be
negative). Returns offset from the start or -1.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
13
Duplicating File Descriptors
●
dup and dup2 are used to duplicate a file
descriptor. The duplicated descriptor will have
the same file pointer in the file table. These
calls are useful for interprocess communication.
#include <unistd.h>
int dup(int oldfd); # use the lowest avail.
int dup2(int oldfd, int newfd)
# make newfd a copy of oldfd
●
They return the new file descriptor (or -1 on
error).
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
14
File Information
●
stat, fstat, and lstat are used to obtain all file
information (type, owner, group, permissions,
times, # of links, etc):
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
int stat(const char *file_name, struct stat *buf);
int fstat(int filedes, struct stat *buf);
int lstat(const char *file_name, struct stat *buf);
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
15
File Information
●
●
●
stat and lstat differ in regard to symlinks. stat
returns info on the file pointed to while lstat
returns info on the symlink. fstat acts like stat.
Each of the routines require a pointer to a stat
structure. The header files define the stat
structure and also define several useful macros.
Use "man 2 stat" to get the man page. ("man
stat" will get the command-line stat.)
Refer to the getstat.cpp program.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
16
File Information
●
There is also a stat command line utility that will
display file information:
$ stat /etc/passwd
File: `/etc/passwd'
Size: 32889 Blocks: 72 IO Block: 4096 regular file
Device: 801h/2049d Inode: 1712752 Links: 1
Access: (0644/­rw­r­­r­­) Uid: (0/root) Gid: (0/root)
Access: 2010­09­08 08:17:16.000000000 ­0500
Modify: 2010­08­27 08:16:42.000000000 ­0500
Change: 2010­08­27 08:16:42.000000000 ­0500
Birth: ­
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
17
Other File Related Routines
chmod(path, mode) chown(path,owner,group)
unlink(path)
link(path1, path2)
symlink(path1, path2)
mkdir(path, mode)
rmdir(path)
chdir(path)
getcwd(buffer, size)
Thursday, September 10
# change file perms
# change file owner (root)
# delete file, # (decr # of links by 1)
# create a hard link to # a file
# create a symbolic link
# create a new directory
# delete directory
# change process directory
# get current working
# directory
CS 375 UNIX System Programming - Lecture 5
18
In-class Exercises
●
●
●
Modify getstat.cpp so that it also displays the
inode #, # of links, and the file size. Reminder:
the man page for stat(2) has information about
the stat structure and various macros.
Further modify getstat.cpp to display file
permissions in both octal and symbolic form.
Further modify getstat.cpp to display all the
information shown by the stat command.
Thursday, September 10
CS 375 UNIX System Programming - Lecture 5
19