CS代考 COMP2017/COMP9017 Dr. – cscodehelp代写
Function pointers, Signals and low level file I/O
COMP2017/COMP9017 Dr.
FACULTY OF ENGINEERING
Copyright By cscodehelp代写 加微信 cscodehelp
Function Pointers
› So far all variables are exposed as having an address
› Compiled binary code is no different if ( != 0 ) {
execute statement; } else {
execute other statement; }
Function Pointers
int x = 33; if (x == 33) {
x = 480+7; }
movl $33, -4(%rbp) cmpl $33, -4(%rbp) jne .L2
movl $487, -4(%rbp) .L2:
addl $768, -4(%rbp)
› Same with loops
› rbp is the stack frame pointer on x86_64
Function Pointers
int x = 33; if (x == 33) {
x = 480+7;
movl $33, -4(%rbp) cmpl $33, -4(%rbp) jne .L4
movl $487, -4(%rbp) movl $0, %eax
addl $768, -4(%rbp)
› Call? If not a jump, how do we get back?
Function Pointers
int x = 33; if (x == 33) {
x = 480+7;
movl $33, -4(%rbp) cmpl $33, -4(%rbp) jne .L4
movl $487, -4(%rbp) movl $0, %eax
addl $768, -4(%rbp)
› Call? If not a jump, how do we get back?
› Stack is being managed here. Callee or caller will setup and teardown the stack
Function Pointers
int (*fptr)() = foo;
int x = 33;
if (x == 33) {
x = 480+7;
subq $16, %rsp
leaq foo(%rip), %rax movq %rax, -16(%rbp) movl $33, -4(%rbp) cmpl $33, -4(%rbp) jne .L4
› If we jump, or call, all we need is an address
Function Pointers
int (*fptr)() = foo;
int x = 33; if (x == 33) {
x = 480+7;
movl $33, -4(%rbp) cmpl $33, -4(%rbp) jne .L4
movl $487, -4(%rbp) movq -16(%rbp), %rdx movl $0, %eax
call *%rdx
addl $768, -4(%rbp)
› Call a function, jump to address is (almost) the same process
Function pointer
› A function pointer is an address that refers to an area of memory with executable code
› Typically the first instruction of the function call^
› Are useful for conventional programming patterns
› Examples
– Do something, and when you are finished call this function
– Do something, and if it goes wrong, call this function
– I am a data source, give me an function to send new bits of data to
– I want to sort a list of objects, here is the address of a function to perform comparison of two elements
^ Depends on callee/caller conventions
› The declaration of the function pointer parameter looks like: type (*f)(param declaration…)
› and the call of the function looks like: f(params…)
Function pointer: Example
› Call functionA if x is true, or functionB otherwise
Function pointer: Example
› Call functionA if x is true, or functionB otherwise if (x)
› What if we don’t know what funcA and funcB are at compile time?
void do_process(int x, funcA?, funcB?) { if (x)
funcA(x); // print X else
funcB(x); // delete elem X
Function pointer: Example
› What if we don’t know what funcA and funcB are at compile time?
void deleteX(int x); void printX(int x);
void do_process(int x, void __funcA__(int), void __funcB__(int))
funcA(x); // print X else
funcB(x); // delete elem X
Function pointer: Example
› What if we don’t know what funcA and funcB are at compile time?
void deleteX(int x); void printX(int x);
void do_process(int x, void (*funcA)(int), void (*funcB)(int))
funcA(x); // print X else
funcB(x); // delete elem X }
Function pointer: Example
› Write less code. Allow option to change implementation choices at runtime. E.g. heuristics, look and feel, plugins
void printX_1(int x) { printf(“%d”, x); }
void printX_2(int x) { printf(“%d
”, x); } void printX_3(int x) { printf(“x: %d
”, x); }
// delegate which fn pointer
if (user_style == PRETTY)
print_style = printX_3;
// generic code
do_process(value1, print_style, remove_style); do_process(value2, print_style, remove_style); do_process(value3, print_style, remove_style);
COMP2017/COMP9017
FACULTY OF ENGINEERING
› a process can communicate with another using a signal
› these are a form of software interrupt
› execution is interrupted and a function call is made at that point to a user specified function
› when the function returns, execution is resumed
› signals can be generated by one process to another using the kill system call
› signals are also generated by the operating system, eg when an access outside memory bounds is attempted (Segmentation Fault)
SIGTRAP 5 Trace or Breakpoint
Trap SIGABRT 6 Abort
SIGEMT 7 Emulation Trap SIGFPE 8 Arithmetic Exception
SIGKILL 9 Kill
SIGBUS 10 Bus Error SIGSEGV 11 Segmentation
Fault SIGSYS 12 Bad System Call
SIGPIPE 13 Broken Pipe SIGALRM 14 Alarm Clock SIGTERM 15 Terminated SIGUSR1 16 User Signal 1 SIGUSR2 17 User Signal 2
1 Hangup 2 Interrupt
4 Illegal Instruction
› You can send a signal to a running process from the command line using the kill command
› Eg kill -9 12345
Will send the SIGKILL signal to process 12345.
› Some signals can be caught and handled by a user supplied function
› Some signals (such as SIGKILL) cannot be caught and caused the process to be terminated
› You can send a signal to a running process using the kill system call function
#include
int kill (pid_t pid, int sig);
Where pid is the process ID of the process to be signaled and sig is the signal to be sent.
Catching Signals
› You can “catch” a signal by specifying a function that is called when the signal is received
› This is done using the signal function:
#include
sighandler_t signal(int signum, sighandler_t handler); void (*signal(int sig, void (*catch)(int)))(int);
This complicated looking declaration means that signal is called with 2 arguments: the first is the signal to catch, the second is a pointer to the function that will be called when the signal is received. The signal function returns a pointer to the function that previously caught the signal ….phew.
Signal: Catch SIGINT
volatile int interrupted = 0; void impatient(int sigval) {
interrupted = 1; }
int main() {
signal(SIGINT, impatient); printf(”Now we wait…
”); while (!interrupted)
usleep(10);
printf(“Oh..you didn’t like waiting
”); printf(“Program terminated
”);
Does it work?
› Most C functions report errors via return values, or their parameters
› However, there is still an error reporting mechanism using a global variable called errno
› Failed system calls typically set errno to be an integer value representing the type of error.
› A companion function, strerror and perror, will print a textual description of the errno code.
› The
#include
int main() {
FILE *fp = fopen(“doesn’t exist”, “r”); printf(“errno: %d
”, errno);
› errno is set by the last function call that will set errno.
› There is only one errno value
› It can be overridden by subsequent function calls
› It is important to save this value immediately following
Low level file I/O
COMP2017/COMP9017
FACULTY OF ENGINEERING
Low Level File Descriptors
› Low level I/O is performed on file descriptors that are small integers indicating an open file
› When a process is started file descriptor 0 is standard input, 1 is standard output, 2 is standard error output (UNIX)
› System call functions operate on file descriptors
Low Level File Descriptors
› low level I/O functions in C wrap system calls: – creat, open, close
– read, write
› eg read 100 characters from standard input into array
ssize_t result = read(0, buffer, 100);
$ man 2 open
“If I was to change anything in Unix it would be to spell creat with an e”
Low Level File Descriptors
On error, -1 is returned, and errno is set appropriately. In this case, it is left unspecified whether the file position (if any) changes.
› This may be interrupted by a signal. The way to check is to use errno
ssize_t result = read(…); if (result < 0)
error_val = errno;
if (EINTR == error_val) // reattempt
› These operations are blocking. There may be a need to interrupt them upon a new event.
Working with read and write
› Error checking
- errno is set to an error value
- signal can be sent by operating system
#include
signal(SIGINT, interrupted); char buffer[100];
ssize_t result = read(0, buffer, 100); // check for errors
int error_val = errno;
if (0 != error_val) {
printf(“read() was interrupted by signal
”); }
Does it work?
Catching Signals
› You can “catch” a signal by specifying a function that is called when the signal is received
› This is done using the sigaction() function:
#include
int sigaction(int signum, const struct sigaction *act,
struct sigaction *oldact);
struct sigaction {
void (*sa_handler)(int);
void (*sa_sigaction)(int, siginfo_t *, void *);
sigset_t sa_mask;
int sa_flags;
void (*sa_restorer)(void);
Working with read and write
› Error checking
– errno is set to an error value
– signal can be sent by operating system
#include
// setup new handler new_sig_int.sa_handler = interrupted; new_sig_int.sa_flags = 0;
// install the new handler
sigaction(SIGINT, &new_sig_int, NULL);
char buffer[100];
ssize_t result = read(0, buffer, 100);
// check for errors
int error_val = errno;
if (error_val != 0) {
printf(“read() was interrupted by signal
”);
Low Level File Descriptors
› Extra attention is needed when working with files at this level – Buffering
– Sharing vs exclusive access (resource locking)
– Errors and interruptions
– Notifications (Linux)
– Resource limit setting – Performance
› fcntl – manipulate file descriptors
› Valuable to have very fine control of file operations
程序代写 CS代考 加微信: cscodehelp QQ: 2235208643 Email: kyit630461@163.com
