Issue with CTRL+D on Windows Subsystem for Linux - c

I was working on an assignment using Windows Subsystem for Linux. Below is the C Code that was used to write a mini-shell for this assignment.
I ran into an interesting issue using WSL. On line 35, you can see that I call the read function to read in the buffer and it does a check for null. When pressing Ctrl+D, while using WSL, it will go into the if statement and prints the print message on line 36 infinitely and does not stop until I used Ctrl+C to exit. When running this program on a Linux machine, it behaves appropriately and prints once, and brings us to the top of the loop.
Any ideas as to what this bug could be?
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/wait.h>
#include <errno.h>
#include <string.h>
#include <sys/types
#include <unistd.h>
#include <error.h>
char prompt[] = "$ ";
static int
Fork()
{
pid_t pid;
if ((pid = fork()) < 0)
error(EXIT_FAILURE, errno, "fork error");
return(pid);
}
int
main(void)
{
long MAX = sysconf(_SC_LINE_MAX);
char buf[MAX];
pid_t pid;
int status, n;
do {
write(STDOUT_FILENO, prompt, strlen(prompt));
fflush(NULL);
memset(buf, 0, MAX);
if((n = read(STDIN_FILENO, buf, MAX)) == 0) {
printf("use exit to exit shell\n");
continue;
}
buf[strlen(buf) - 1] = '\0'; // chomp '\n'
if (strncmp(buf, "exit", MAX) == 0) { // match
break;
}
pid = Fork();
if (pid == 0) { // child
execlp(buf, buf, (char *)NULL);
error(EXIT_FAILURE, errno, "exec failure");
}
// parent
if ((pid = waitpid(pid, &status, 0)) < 0)
error(EXIT_FAILURE, errno, "waitpid error");
} while(1);
exit(EXIT_SUCCESS);
}
The program is in C but there are no options available to insert C Code snippets.

Documentation on read() (Linux manpages v 3.54) does not specify that end of file (ctrl/D) causes read to return anything besides of 0. On the contrary, it says that return value zero indicates end of file. So you're relying upon undefined behavior.
Somehow on your Linux ctrl/D causes error, thus read() returns -1. Your program in this case exits the loop. Or, ctrl/D is read literally, then read() returns 1.

Different OSs use different keystrokes for EOF.

Related

execvp() function returns -1 all the time

In my code, I'm trying to use the function execvp() to execute a command that I get in my shell but the function always returns -1 that indicates unsuccess, when I replace the function first argument by (for example) "ps" it works fine but when it is (command) it doesn't work, I've checked that command is fine by printing it after getting it from the input line and it is a fine string with no problems, but the function keeps returning me an error!!
#include <sys/types.h>
#include <sys/wait.h>
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>
#define BUFFER_SIZE 100
int main(void)
{
close(2);
dup(1);
char command[BUFFER_SIZE];
while (1)
{
char *arg[3];
fprintf(stdout, "my-shell> ");
memset(command, 0, BUFFER_SIZE);
fgets(command, BUFFER_SIZE, stdin);
if(strncmp(command, "exit", 4) == 0)
{
break;
}
arg[0] = command;
arg[1] = "\0";
arg[2] = "\0";
i = execvp(command,arg);
printf("%d",i);
}
return 0;
}
I expect that the problem is in the way that command is passed in the function but after trying so much edites to the code, I still can't figure out what the problem really is!
There are 3 major problems and 1 minor one that can be picked out of the code shown (plus what I take to be an artefact of reducing your full code to the code in the question, plus some oddities):
The fgets() function includes the newline in the returned string unless the line is too long (a separate problem). You need to zap that newline:
command[strcspn(command, "\n")] = '\0';
The code does not parse the line that's entered, so only single word commands can sensibly be entered. To fix that, you'd have to be prepared to split the line into words using an appropriate algorithm, removing quotes where appropriate, expanding variables and so on. That will be part of the later stages of developing your shell.
The second argument to execvp() needs to be a NULL-terminated list of strings. You only provide the command name and two empty strings without the null terminator, which gives undefined behaviour.
The minor problem is that using "\0" instead of just "" is pointless.
The artefact is that there is no fork() in the code, so if the command is executed successfully, the 'shell' is replaced by the command and exits when the replacement exits.
The close(2); dup(1); sequence is weird — it means standard error refers to the same file descriptor as standard output. Those lines really aren't needed (or desirable). Leave the errors separate from standard output.
The memset() is superfluous too. Using fprintf(stdout, "my-shell> "); is a funny way of writing printf("my-shell> ");. Using strncmp(command, "exit", 4) means that if the user types exit-or-continue, you'll treat it the same as exit, which is far from ideal.
Putting most of those numerous changes into effect (omitting parsing the command line into separate arguments) leaves:
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/wait.h>
#include <unistd.h>
#define BUFFER_SIZE 100
int main(void)
{
char command[BUFFER_SIZE];
while (1)
{
printf("my-shell> ");
fflush(stdout);
if (fgets(command, BUFFER_SIZE, stdin) != command)
break;
command[strcspn(command, "\n")] = '\0';
if(strcmp(command, "exit") == 0)
{
break;
}
int pid = fork();
if (pid < 0)
{
fprintf(stderr, "failed to fork()\n");
exit(EXIT_FAILURE);
}
if (pid == 0)
{
/* Child - execute command */
/* Should break line into command plus arguments */
char *arg[2];
arg[0] = command;
arg[1] = NULL;
execvp(command, arg);
fprintf(stderr, "failed to execute command '%s'\n", command);
exit(EXIT_FAILURE);
}
/* Parent - wait for child to finish */
int corpse;
int status;
while ((corpse = wait(&status)) > 0)
{
if (corpse == pid)
break;
printf("PID %d exited with status 0x%.4X\n", corpse, status);
}
}
return 0;
}

With vs without the C pipe() function- what's causing this behavior?

I wrote a simple script (taken from a tutorial) which writes data to one end of a pipe in a child process, and reads it from the other end of the pipe in the parent process:
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
int main()
{
pid_t pid;
int mypipefd[2];
int ret;
char buf[20];
ret = pipe(mypipefd);
if (ret == -1) {
printf("Pipe failed.\n");
exit(1);
}
if ((pid = fork()) == -1) {
printf("Fork failed.\n");
exit(1);
} else if (pid == 0) {
printf("Child process.\n");
char msg[] = "Hello there!";
write(mypipefd[1], msg, strlen(msg) + 1);
} else {
printf("Parent process.\n");
read(mypipefd[0], buf, 15);
printf("Buf: %s\n", buf);
}
return 0;
}
This works fine and outputs the results I expect:
Parent process.
Child process.
Buf: Hello there!
[ project ] $
Then as I got more familiar with the code, I wondered why we need to use mypipefd[2] and pipe() to achieve this goal, or whether mypipefd[1] by itself would work. So I tried it out with the following code:
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
int main()
{
pid_t pid;
int my_array[1];
char buf[20];
if ((pid = fork()) == -1) {
printf("Fork failed.\n");
exit(1);
} else if (pid == 0) {
printf("Child process.\n");
char msg[] = "Hello there!\n";
write(my_array[0], msg, strlen(msg) + 1);
} else {
// wait(NULL);
printf("Parent process.\n");
read(my_array[0], buf, 15);
printf("Buf: %s\n", buf);
}
return 0;
}
This code outputs the same text, but it hangs after it finishes printing.
Parent process.
Child process.
Buf: Hello there!
No prompt, this time. I even tried un-commenting that call to wait(NULL), on the off-chance that the root cause was a conflict between parent and child processes. No such luck.
What's going on here? Why am I unable to read and write to a length-of-one array in this way without the program hanging? What exactly is the compiler stuck on?
A pipe, on computers as well as in real life, have two ends. And like pipes in real life, data flows from one end of the pipe (the write end) to the other (the read end).
The pipe function gives you those two ends by writing them to an array of two file-descriptors. The first element of the pair is read-only, and the second is write-only.
The pipe() function accepts an array of 2 integer as an input argument.
#include <unistd.h>
int pipe(int pipefd[2]);
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <fcntl.h> /* Obtain O_* constant definitions */
#include <unistd.h>
int pipe2(int pipefd[2], int flags);
It then generates a new pipe object, and initializes the pipefd array with file descriptors for read and write operation.
What you try to do is call read() and write() using some arbitrary, uninitialized ints (or file descriptor). Meaning the OS did not allocate a pipe object and did not provide you with file descriptors (the pipe's API) to be use with read() and write().
This (calling read() or write() with uninitialized file descriptor) will result in "undefined behavior".
"I find that a good working definition of "undefined behaviur" is "works for me, works for you, works during development and QA, but blows up in your most important customer's face"" --- Scott Meyers

Redirection of stdin and stdout via pipes in C works for external programmes but not for recursive call

I am trying to communicate with forked child processes via pipe redirection of stdin and stdout in C. I already managed to get this to work for shell commands (like ls, for example) executed in child processes. However, I wasn't able to recursively execute the same program and redirect the output (printed by printf(), fprintf() to stdout, ...) via the pipes from the child process to the parent (in this test to stdout of the parent), although this works fine for ls or similar commands.
Here's how I tried to approach this:
I create a pipe, the reading end is for the parent, the child process should write to the writing end.
The Process forks, both processes close the unused end, respectively.
The writing end of the pipe is redirected to STDOUT_FILENO and closed
The child process executes the program recursively (it is called ./to2)
As mentioned, this does work if I execute ls in the child process, but not if I try to call the same program recursively. Here's my test program where I tried to get this to work:
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <string.h>
#include <sys/socket.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/wait.h>
#include <netdb.h>
#include <errno.h>
#include <time.h>
#include <signal.h>
#include <fcntl.h>
static void usage(void){
fprintf(stderr,"RIP");
exit(EXIT_FAILURE);
}
int main(int argc, char *argv[]){
if(argc > 1){
dprintf(STDOUT_FILENO,"Please work\n");
printf("\n THIS IS A MESSAGE FROM THE CHILD \n");
fputs("Pretty Please!\n",stdout);
fflush(stdout);
exit(EXIT_SUCCESS);
}
int p1[2];
if(-1 == pipe(p1)) {
fprintf(stderr,"pipe\n");
fprintf(stderr,"%s\n",strerror(errno));
usage();
}
int f = fork();
if(f == 0){
close(p1[0]);
if(dup2(p1[1],STDOUT_FILENO) < 0){
fprintf(stderr,"dup2\n");
usage();
}
close(p1[1]);
//I want this to work:
//execlp("./to2", "./to2", "-e");
//This works fine:
execlp("ls", "ls");
exit(EXIT_SUCCESS);
} else if (f == -1) {
usage();
} else {
close(p1[1]);
int w = -1;
if(-1 == wait(&w)) usage();
char b[12];
memset(b,0,12);
read(p1[0],&b,12);
char reading_buf[1];
while(read(p1[0], reading_buf, 1) > 0){
write(1, reading_buf, STDOUT_FILENO);
}
close(p1[0]);
}
}
For testing purposes, the function is called recursively with additional arguments, while the parent program is called without additional arguments (hence the if(argc>1)).
In the final program, endless recursion is being avoided by other means.
Did I understand something wrongly? I am pretty confused by the fact that the only thing that doesn't seem to work is redirecting the output of my own
program...
Thank you very much in advance, any help or ideas are greatly appreciated.
The primary problem is precisely as outlined in the comments — you are not calling execlp() correctly (nor ls in the alternative). You must make the last argument on those function calls into an explicit null pointer, as shown in this code, which is a mostly mildly edited version of what's in the question:
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/wait.h>
#include <unistd.h>
static void usage(void)
{
fprintf(stderr, "RIP\n");
exit(EXIT_FAILURE);
}
int main(int argc, char *argv[])
{
if (argc > 1)
{
dprintf(STDOUT_FILENO, "Please work\n");
printf("THIS IS A MESSAGE FROM THE CHILD\n");
fputs("Pretty Please!\n", stdout);
fflush(stdout);
exit(EXIT_SUCCESS);
}
int p1[2];
if (-1 == pipe(p1))
{
fprintf(stderr, "pipe: %s\n", strerror(errno));
usage();
}
int f = fork();
if (f == 0)
{
close(p1[0]);
if (dup2(p1[1], STDOUT_FILENO) < 0)
{
fprintf(stderr, "dup2: %s\n", strerror(errno));
usage();
}
close(p1[1]);
execlp(argv[0], argv[0], "-e", (char *)0);
fprintf(stderr, "failed to exec %s again\n", argv[0]);
exit(EXIT_FAILURE);
}
else if (f == -1)
{
usage();
}
else
{
close(p1[1]);
char b[13];
memset(b, 0, 13);
if (read(p1[0], &b, 12) < 0)
{
fprintf(stderr, "Failed to read from pipe (%s)\n", strerror(errno));
exit(EXIT_FAILURE);
}
int len = strcspn(b, "\n");
printf("M1 [%.*s]\n", len, b);
char reading_buf[1];
while (read(p1[0], reading_buf, 1) > 0)
{
write(1, reading_buf, STDOUT_FILENO);
}
close(p1[0]);
int w = -1;
if (-1 == wait(&w))
usage();
}
return 0;
}
Two important changes should be highlighted:
This code echoes the first line of data — the one written by dprintf() — whereas the original code just read it and discarded it.
The wait() call is after the input, not before. If the child had more data to write than a set of fixed messages, it could block waiting for the parent to read some of the data, while the parent is blocked waiting for the child to exit. This would be a deadlock.
The usage() function is not appropriately named — it doesn't report how to run the program. I also exit with a failure status, not success, if the child process fails the execlp().
Under peculiar circumstances, the wait() call might report on the exit status from some child other than the one that was forked. It is generally best to use a loop to reap such children. However, the circumstances required are extremely peculiar — the process which launched the parent with an exec*() function must have previously created some children for which it didn't wait, so that they are inherited by the parent process (because the PID doesn't change across an exec*() call).

result of child process's exection of some system command can't send to the father process with pipe

Maybe this is not a compact title, I am very sorry about that:). I try redirecting stdin/stdout of a child process to its parent process with pipes. The child process execute a system command from the father process input and return the exec result to the father process with a pipe. Here I implemented "cat -n" and "tr /a-z/ /A-Z/", the former works fine, but later haven't return any results. What has caused this? Thank you.
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/ipc.h>
#include <sys/shm.h>
#include <unistd.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>
#include <signal.h>
#include <assert.h>
#include <sys/sem.h>
#define ERR_EXIT(m) \
do { \
perror(m); \
exit(EXIT_FAILURE); \
} while( 0)
int main(int argc, char *argv[])
{
int chi_pipe[2], par_pipe[2];
if (pipe(chi_pipe) == -1 || pipe(par_pipe) == -1)
ERR_EXIT("pipe error");
/* Set O_NONBLOCK flag for the read end (pfd[0]) of the pipe. */
if (fcntl(chi_pipe[0], F_SETFL, O_NONBLOCK) == -1) {
fprintf(stderr, "Call to fcntl failed.\n"); exit(1);
}
/* Set O_NONBLOCK flag for the read end (pfd[0]) of the pipe. */
if (fcntl(chi_pipe[1], F_SETFL, O_NONBLOCK) == -1) {
fprintf(stderr, "Call to fcntl failed.\n"); exit(1);
}
pid_t pid;
pid = fork();
if (pid == -1)
ERR_EXIT("fork error");
if (pid == 0)
{
close(chi_pipe[0]); // I don't read in channel 1
close(par_pipe[1]); // I don't write in channel 2
dup2(chi_pipe[1], STDOUT_FILENO);
close(STDIN_FILENO);
dup2(par_pipe[0], STDIN_FILENO);
execlp("cat", "cat" , "-n", NULL);
//execlp("tr", "tr" , "/a-z/", "/A-Z/", NULL);
sleep(10);
close(chi_pipe[1]);
close(par_pipe[0]);
_exit(0);
}
close(par_pipe[0]);
close(chi_pipe[1]);
while(1) {
char input[1024];
memset(input, 0 , 1024);
fgets(input, 1024 ,stdin);
write(par_pipe[1], input, strlen(input));
char buf[3*1024];
int count = 0;
while (count <= 0)
count=read(chi_pipe[0], buf, 1024*3);
if (count >= 1)
{
printf("buf=%s", buf);
printf("\n");
}
}
close(par_pipe[1]);
close(chi_pipe[0]);
return 0;
}
A couple of points:
You are suffering from the need to perform non-blocking I/O. You are reading a line from a file, then writing it to a pipe. But there is no guarantee tr will conveniently write that line back translated. It might wait for the next line to come in. There is no line discipline in place. What you need to do is read from your file, write to tr (if the pipe is not full) and read from tr (if bytes are ready) at the same time. Or, more accurately, according to availability of data on the fd (to read) or the availability of space in the pipe (to write). Otherwise you will run into deadlock problems. tr isn't writing because it would rather read more first, and it hasn't got EOF. You aren't reading from tr because it hasn't written yet, so you aren't reading any more from the file either. To do this, you want to use select() (or poll()).
The only way execlp will return is if the exec fails; in that case you don't want to exit(0) as it's necessarily an error.

C Named pipe (fifo). Parent process gets stuck

I want to make a simple program, that fork, and the child writes into the named pipe and the parent reads and displays from the named pipe.
The problem is that it enters the parent, does the first printf and then it gets weird, it doesn't do anything else, does not get to the second printf, it just ways for input in the console.
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
void main()
{
char t[100];
mkfifo("myfifo",777);
pid_t pid;
pid = fork();
if (pid==0)
{
//execl("fifo2","fifo2",(char*)0);
char r[100];
printf("scrie2->");
scanf("%s",r);
int fp;
fp = open("myfifo",O_WRONLY);
write(fp,r,99);
close(fp);
printf("exit kid \n");
exit(0);
} else
{
wait(0);
printf("entered parent \n"); // <- this it prints
// whats below this line apparently its not being executed
int fz; printf("1");
fz = open("myfifo",O_RDONLY); printf("2");
printf("fd: %d",fz);
char p[100];
int size;
printf("------");
//struct stat *info;
//stat("myfifo",info); printf("%d",(*info).st_size);
read(fz,p,99);
close(fz);
printf("%s",p);
printf("exit"); exit(0);
}
}
You really should be checking the return value on function calls for errors, especially mkfifo() and open().
Your call to wait() is going to cause problems in its current location. Opening a FIFO for reading normally blocks until some other process opens the same FIFO for writing, and vice versa1. The parent is waiting for the child to terminate and the child is waiting for a reader process, i.e., the parent, to connect to the FIFO.
1 - see note on open() below for using O_NONBLOCK with a FIFO
Moving the wait() call to just before the parent process exits along with changing the mode in the call to mkfifo() to 0666 seems to resolve some of your immediate problems.
It is also good practice to remove the FIFO when you are finished with it.
unlink("myfifo");
From the open() function documentation in IEEE Std 1003.1-2004:
When opening a FIFO with O_RDONLY or O_WRONLY set:
If O_NONBLOCK is set, an open() for reading-only shall return without delay. An open() for writing-only shall return an error if no process currently has the file open for reading.
If O_NONBLOCK is clear, an open() for reading-only shall block the calling thread until a thread opens the file for writing. An open() for writing-only shall block the calling thread until a thread opens the file for reading.
The following example is a combination of the code in your original question and the FIFO page of Beej's Guide to Unix IPC:
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>
#include <fcntl.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/wait.h>
#include <unistd.h>
#define FIFO_NAME "myfifo"
int main(void)
{
char buf[256];
int num, fd;
pid_t pid;
if (mkfifo(FIFO_NAME, 0666) < 0)
perror("mkfifo");
pid = fork();
if (pid == 0)
{
printf("child - waiting for readers...\n");
if ((fd = open(FIFO_NAME, O_WRONLY)) < 0)
perror("child - open");
printf("child - got a reader -- type some stuff\n");
while (fgets(buf, sizeof(buf), stdin), !feof(stdin))
{
if ((num = write(fd, buf, strlen(buf))) < 0)
perror("child - write");
else
printf("child - wrote %d bytes\n", num);
}
close(fd);
exit(0);
}
else
{
printf("parent - waiting for writers...\n");
if ((fd = open(FIFO_NAME, O_RDONLY)) < 0)
perror("parent - open");
printf("parent - got a writer\n");
do
{
if ((num = read(fd, buf, sizeof(buf))) < 0)
perror("parent - read");
else
{
buf[num] = '\0';
printf("parent - read %d bytes: \"%s\"\n", num, buf);
}
} while (num > 0);
close(fd);
wait(0);
}
unlink(FIFO_NAME);
return 0;
}
This example was tested in Linux. Press Ctrl-D to terminate the program.
First of all, try fprintf to stderr instead of printf (to stdout)
The stderr is unbuffered.
Then you can tell what actually gets printed and what does not.
or at least add fflush before waiting for anything.

Resources