fcntl how to know which process hold lock file? - c

I'm new with fcntl locking and following this example to create a sample lock in linux using c code: http://www.informit.com/articles/article.aspx?p=23618&seqNum=4
I wonder how can we can print out which process hold the lock file and which process is waiting for lock. I consider using l_pid to figure out the process id which is holding the lock but i'm not sure the right way to do it.
What is the best way to print out which process is holding the lock?

As the man 2 fcntl page describes, you can use the F_GETLK to obtain the process ID that has the conflicting lock (if the conflicting lock is a process-associated one). So, for example,
/* Return 0 if descriptor locked exclusively, positive PID if
a known process holds a conflicting lock, or -1 if the
descriptor cannot be locked (and errno has the reason).
*/
static pid_t lock_exclusively(const int fd)
{
struct flock lock;
int err = 0;
if (fd == -1) {
errno = EINVAL;
return -1;
}
lock.l_type = F_WRLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
if (!fcntl(fd, F_SETLK, &lock))
return 0;
/* Remember the cause of the failure */
err = errno;
lock.l_type = F_WRLCK;
lock.l_whence = SEEK_SET;
lock.l_start = 0;
lock.l_len = 0;
lock.l_pid = 0;
if (fcntl(fd, F_GETLK, &lock) == 0 && lock.l_pid > 0)
return lock.l_pid;
errno = err;
return -1;
}
Do note that fd must be open for reading and writing. I recommend using open(path, O_RDWR | O_NOCTTY) or open(path, O_WRONLY | O_NOCTTY). Closing any file descriptor to the same file will release the lock.
Some may say that re-setting the lock memebers before the second fcntl() call is unnecessary, but I'd rather err on the side of caution here.
As to how to report it, I would simply use
int fd;
pid_t p;
fd = open(path, O_RDWR | O_NOCTTY);
if (fd == -1) {
fprintf(stderr, "%s: Cannot open file: %s.\n",
path, strerror(errno));
exit(EXIT_FAILURE);
}
p = lock_exclusively(fd);
if (p < 0) {
fprintf(stderr, "%s: Cannot lock file: %s.\n",
path, strerror(errno));
exit(EXIT_FAILURE);
} else
if (p > 0) {
fprintf(stderr, "%s: File is already locked by process %ld.\n",
path, (long)p);
exit(EXIT_FAILURE);
}
/* fd is now open and exclusive-locked. */
The user can always run e.g. ps -o cmd= -p PID to see what command that is (or you can try reading /proc/PID/cmdline in Linux).

From the example code:
printf ("locking\n");
/* Initialize the flock structure. */
memset (&lock, 0, sizeof(lock));
lock.l_type = F_WRLCK;
/* Place a write lock on the file. */
fcntl (fd, F_SETLKW, &lock);
printf ("locked; hit Enter to unlock... ");
You need to change the fcntl (fd, F_SETLKW, &lock); to:
if (fcntl (fd, F_SETLK, &lock) == -1) {
printf ("File is locked by pid %i\n", lock.l_pid);
return 0;
}
The F_SETLKW command blocks if it cannot get the lock. F_SETLK will return if it cannot get the lock. Really the code should also check errno == EACCESS or errno == EAGAIN after getting the -1 return value.

Related

ncurses newterm following openpty

I am trying to figure out how to do the following:
create a new pseudo-terminal
open a ncurses screen running inside the (slave) pseudo terminal
fork
A) forward I/O from the terminal the program is running in (bash) to the new (slave) terminal OR
B) exit leaving the ncurses program running in the new pty.
Can anyone provide pointers to what I might be doing wrong or that would make sense of some of this or even better an example program using newterm() with either posix_openpt(), openpty() or forkpty().
The code I have is roughly (details simplified or omitted):
openpty(master,slave,NULL,NULL,NULL);
pid_t res = fork();
if(res == -1)
std::exit(1);
if(res == 0) //child
{
FILE* scrIn = open(slave,O_RDWR|O_NONBLOCK);
FILE* scrOut = open(slave,O_RDWR|O_NONBLOCK);
SCREEN* scr = newterm(NULL,scrIn,scrOut);
}
else //parent
{
if (!optionA)
exit(0); // but leave the child running and using the slave
for(;;)
{
// forward IO to slave
fd_set read_fd;
fd_set write_fd;
fd_set except_fd;
FD_ZERO(&read_fd);
FD_ZERO(&write_fd);
FD_ZERO(&except_fd);
FD_SET(masterTty, &read_fd);
FD_SET(STDIN_FILENO, &read_fd);
select(masterTty+1, &read_fd, &write_fd, &except_fd, NULL);
char input[2];
char output[2];
input[1]=0;
output[1]=0;
if (FD_ISSET(masterTty, &read_fd))
{
if (read(masterTty, &output, 1) != -1)
{
write(STDOUT_FILENO, &output, 1);
}
}
if (FD_ISSET(STDIN_FILENO, &read_fd))
{
read(STDIN_FILENO, &input, 1);
write(masterTty, &input, 1);
}
}
}
}
I have various debug routines logging results from the parent and child to files.
There are several things relating to terminals that I do not understand.
I have seen several behaviours I don't understand depending on what variations I try.
Things I don't understand:
If I instruct the parent process exits the child terminates without anything interesting being logged by the child.
If I try closing stdin, stdout and using dup() or dup2() to make the pty the replace stdin
the curses window uses the original stdin and stdout and uses the original pty not the new one based on the output of ptsname().
(the parent process successful performs IO with the child but in the terminal it was lauched from not the new pty)
If I open the new pty using open() then I get a segfault inside the ncurses newterm() call as below:
Program terminated with signal 11, Segmentation fault.
#0 0x00007fbd0ff580a0 in fileno_unlocked () from /lib64/libc.so.6
Missing separate debuginfos, use: debuginfo-install glibc-2.17-317.el7.x86_64 ncurses-libs-5.9-14.20130511.el7_4.x86_64
(gdb) where
#0 0x00007fbd0ff580a0 in fileno_unlocked () from /lib64/libc.so.6
#1 0x00007fbd106eced9 in newterm () from /lib64/libncurses.so.5
... now in my program...
I am trying to understand the pty system calls here. Using a program like screen or tmux does not help with this (also the source is not sufficiently annotated to fill in the gaps in my understanding).
Some other datums:
I am targeting GNU/Linux
I have also tried using forkpty
I looked at source for openpty, forkpty, login_tty, openpt, grantpt & posix_openpt
(e.g. https://github.com/coreutils/gnulib/blob/master/lib/posix_openpt.c)
I don't have access to a copy of APUE
though I have looked at the pty example.
Although the ncurses documentation for newterm() mentions talking to multiple terminals simultaneously I have not found an example program that does this.
I am still not clear on:
what login_tty / grantpt actually do.
If you opened the pty yourself why wouldn't you already have the correct capabilities?
why I might prefer openpty to posix_openpt or visa-versa.
Note: This is a different question to attach-a-terminal-to-a-process-running-as-a-daemon-to-run-an-ncurses-ui which describes a use case and looks for a solution where this question assumes a particular but incorrect/incomplete implementation for that use case.
Let's look at one possible implementation of pseudoterminal_run(), which creates a new pseudoterminal, forks a child process to run with that pseudoterminal as the controlling terminal with standard input, output, and error directed to that pseudoterminal, and executes a specified binary.
Here's the header file, pseudoterminal.h:
#ifndef PSEUDOTERMINAL_H
#define PSEUDOTERMINAL_H
int pseudoterminal_run(pid_t *const, /* Pointer to where child process ID (= session and process group ID also) is saved */
int *const, /* Pointer to where pseudoterminal master descriptor is saved */
const char *const, /* File name or path of binary to be executed */
char *const [], /* Command-line arguments to binary */
const struct termios *const, /* NULL or pointer to termios settings for the pseudoterminal */
const struct winsize *const); /* NULL or pointer to pseudoterminal size */
#endif /* PSEUDOTERMINAL_H */
Here is the corresponding implementation, pseudoterminal.c:
#define _POSIX_C_SOURCE 200809L
#define _XOPEN_SOURCE 600
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
#include <fcntl.h>
#include <termios.h>
#include <signal.h>
#include <string.h>
#include <errno.h>
/* Helper function: Moves fd so that it does not overlap standard streams.
* If an error occurs, will close fd.
*/
static int not_stdin_stdout_stderr(int fd)
{
unsigned int close_mask = 0;
if (fd == -1) {
errno = EBADF;
return -1;
}
while (1) {
if (fd == STDIN_FILENO)
close_mask |= 1;
else
if (fd == STDOUT_FILENO)
close_mask |= 2;
else
if (fd == STDERR_FILENO)
close_mask |= 4;
else
break;
fd = dup(fd);
if (fd == -1) {
const int saved_errno = errno;
if (close_mask & 1) close(STDIN_FILENO);
if (close_mask & 2) close(STDOUT_FILENO);
if (close_mask & 4) close(STDERR_FILENO);
errno = saved_errno;
return -1;
}
}
if (close_mask & 1) close(STDIN_FILENO);
if (close_mask & 2) close(STDOUT_FILENO);
if (close_mask & 4) close(STDERR_FILENO);
return fd;
}
static int run_slave(int master,
const char * binary,
char *const args[],
const struct termios *termp,
const struct winsize *sizep)
{
int slave;
/* Close standard streams. */
close(STDIN_FILENO);
close(STDOUT_FILENO);
close(STDERR_FILENO);
/* Fix ownership and permissions for the slave side. */
if (grantpt(master) == -1)
return errno;
/* Unlock the pseudoterminal pair */
if (unlockpt(master) == -1)
return errno;
/* Obtain a descriptor to the slave end of the pseudoterminal */
do {
#if defined(TIOCGPTPEER)
slave = ioctl(master, TIOCGPTPEER, O_RDWR);
if (slave == -1) {
if (errno != EINVAL &&
#if defined(ENOIOCTLCMD)
errno != ENOIOCTLCMD &&
#endif
errno != ENOSYS)
return errno;
} else
break;
#endif
const char *slave_pts = ptsname(master);
if (!slave_pts)
return errno;
slave = open(slave_pts, O_RDWR);
if (slave == -1)
return errno;
else
break;
} while (0);
#if defined(TIOCSCTTY)
/* Make sure slave is our controlling terminal. */
ioctl(slave, TIOCSCTTY, 0);
#endif
/* Master is no longer needed. */
close(master);
/* Duplicate slave to standard streams. */
if (slave != STDIN_FILENO)
if (dup2(slave, STDIN_FILENO) == -1)
return errno;
if (slave != STDOUT_FILENO)
if (dup2(slave, STDOUT_FILENO) == -1)
return errno;
if (slave != STDERR_FILENO)
if (dup2(slave, STDERR_FILENO) == -1)
return errno;
/* If provided, set the termios settings. */
if (termp)
if (tcsetattr(STDIN_FILENO, TCSANOW, termp) == -1)
return errno;
/* If provided, set the terminal window size. */
if (sizep)
if (ioctl(STDIN_FILENO, TIOCSWINSZ, sizep) == -1)
return errno;
/* Execute the specified binary. */
if (strchr(binary, '/'))
execv(binary, args); /* binary is a path */
else
execvp(binary, args); /* binary is a filename */
/* Failed! */
return errno;
}
/* Internal exit status used to verify child failure. */
#ifndef PSEUDOTERMINAL_EXIT_FAILURE
#define PSEUDOTERMINAL_EXIT_FAILURE 127
#endif
int pseudoterminal_run(pid_t *const childp,
int *const masterp,
const char *const binary,
char *const args[],
const struct termios *const termp,
const struct winsize *const sizep)
{
int control[2] = { -1, -1 };
int master;
pid_t child;
int cause;
char *const cause_end = (char *)(&cause) + sizeof cause;
char *cause_ptr = (char *)(&cause);
/* Verify required parameters exist. */
if (!childp || !masterp || !binary || !*binary || !args || !args[0]) {
errno = EINVAL;
return -1;
}
/* Acquire a new pseudoterminal */
master = posix_openpt(O_RDWR | O_NOCTTY);
if (master == -1)
return -1;
/* Make sure master does not shadow standard streams. */
master = not_stdin_stdout_stderr(master);
if (master == -1)
return -1;
/* Control pipe passes exec error back to this process. */
if (pipe(control) == -1) {
const int saved_errno = errno;
close(master);
errno = saved_errno;
return -1;
}
/* Write end of the control pipe must not shadow standard streams. */
control[1] = not_stdin_stdout_stderr(control[1]);
if (control[1] == -1) {
const int saved_errno = errno;
close(control[0]);
close(master);
errno = saved_errno;
return -1;
}
/* Write end of the control pipe must be close-on-exec. */
if (fcntl(control[1], F_SETFD, FD_CLOEXEC) == -1) {
const int saved_errno = errno;
close(control[0]);
close(control[1]);
close(master);
errno = saved_errno;
return -1;
}
/* Fork the child process. */
child = fork();
if (child == -1) {
const int saved_errno = errno;
close(control[0]);
close(control[1]);
close(master);
errno = saved_errno;
return -1;
} else
if (!child) {
/*
* Child process
*/
/* Close read end of control pipe. */
close(control[0]);
/* Note: This is the point where one would change real UID,
if one wanted to change identity for the child process. */
/* Child runs in a new session. */
if (setsid() == -1)
cause = errno;
else
cause = run_slave(master, binary, args, termp, sizep);
/* Pass the error back to parent process. */
while (cause_ptr < cause_end) {
ssize_t n = write(control[1], cause_ptr, (size_t)(cause_end - cause_ptr));
if (n > 0)
cause_ptr += n;
else
if (n != -1 || errno != EINTR)
break;
}
exit(PSEUDOTERMINAL_EXIT_FAILURE);
}
/*
* Parent process
*/
/* Close write end of control pipe. */
close(control[1]);
/* Read from the control pipe, to see if child exec failed. */
while (cause_ptr < cause_end) {
ssize_t n = read(control[0], cause_ptr, (size_t)(cause_end - cause_ptr));
if (n > 0) {
cause_ptr += n;
} else
if (n == 0) {
break;
} else
if (n != -1) {
cause = EIO;
cause_ptr = cause_end;
break;
} else
if (errno != EINTR) {
cause = errno;
cause_ptr = cause_end;
}
}
/* Close read end of control pipe as well. */
close(control[0]);
/* Any data received indicates an exec failure. */
if (cause_ptr != (const char *)(&cause)) {
int status;
pid_t p;
/* Partial error report is an I/O error. */
if (cause_ptr != cause_end)
cause = EIO;
/* Make sure the child process is dead, and reap it. */
kill(child, SIGKILL);
do {
p = waitpid(child, &status, 0);
} while (p == -1 && errno == EINTR);
/* If it did not exit with PSEUDOTERMINAL_EXIT_FAILURE, cause is I/O error. */
if (!WIFEXITED(status) || WEXITSTATUS(status) != PSEUDOTERMINAL_EXIT_FAILURE)
cause = EIO;
/* Close master pseudoterminal. */
close(master);
errno = cause;
return -1;
}
/* Success. Save master fd and child PID. */
*masterp = master;
*childp = child;
return 0;
}
To detect errors in the child process before the binary is executed (including errors in executing a binary), the above uses a close-on-exec pipe between the child and the parent to pass errors. In the success case, the pipe write end is closed by the kernel when execution of a new binary starts.
Otherwise the above is a straightforward implementation.
In particular:
posix_openpt(O_RDWR | O_NOCTTY) creates the pseudoterminal pair, and returns the descriptor for the master side. The O_NOCTTY flag is used because we do not want the current process to have that pseudoterminal as the controlling terminal.
in the child process, setsid() is used to start a new session, with both session ID and process group ID matching the child process ID. This way, the parent process can for example send a signal to each process in that group; and when the child opens the pseudoterminal slave side, it should become the controlling terminal for the child process. (The code does do an ioctl(slave_fd, TIOCSCTTY, 0) to ensure that, if TIOCSCTTY is defined.)
grantpt(masterfd) changes the owner user of the slave pseudoterminal to match current real user, so that only the current real user (and privileged users like root) can access the slave side of the pseudoterminal.
unlockpt(masterfd) allows access to the slave side of the pseudoterminal. It must be called before the slave side can be opened.
slavefd = ioctl(masterfd, TIOCGPTPEER, O_RDWR) is used to open the slave side pseudoterminal if available. If not available, or it fails, then slavefd = open(ptsname(masterfd), O_RDWR) is used instead.
The following example.c is an example using the above pseudoterminal.h, which runs a specified binary in a new pseudoterminal, proxying the data between the child process pseudoterminal and the parent process terminal. It logs all reads and writes to a log file you specify as the first command line parameter. The rest of the command line parameters form the command run in the child process.
#define _POSIX_C_SOURCE 200809L
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/wait.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <poll.h>
#include <termios.h>
#include <signal.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>
#include "pseudoterminal.h"
static struct termios master_oldterm, master_newterm, slave_newterm;
static struct winsize slave_size;
static int tty_fd = -1;
static int master_fd = -1;
static void handle_winch(int signum)
{
/* Silence warning about signum not being used. */
(void)signum;
if (tty_fd != -1 && master_fd != -1) {
const int saved_errno = errno;
struct winsize temp_size;
if (ioctl(tty_fd, TIOCGWINSZ, &temp_size) == 0)
if (ioctl(master_fd, TIOCSWINSZ, &temp_size) == 0)
slave_size = temp_size;
errno = saved_errno;
}
}
static int install_winch(void)
{
struct sigaction act;
memset(&act, 0, sizeof act);
sigemptyset(&act.sa_mask);
act.sa_handler = handle_winch;
act.sa_flags = SA_RESTART;
return sigaction(SIGWINCH, &act, NULL);
}
int main(int argc, char *argv[])
{
pid_t child_pid = 0;
int child_status = 0;
FILE *log = NULL;
if (argc < 3 || !strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) {
const char *argv0 = (argc > 0 && argv && argv[0] && argv[0][0]) ? argv[0] : "(this)";
fprintf(stderr, "\n");
fprintf(stderr, "Usage: %s [ -h | --help ]\n", argv0);
fprintf(stderr, " %s LOGFILE COMMAND [ ARGS ... ]\n", argv0);
fprintf(stderr, "\n");
fprintf(stderr, "This program runs COMMAND in a pseudoterminal, logging all I/O\n");
fprintf(stderr, "to LOGFILE, and proxying them to the current terminal.\n");
fprintf(stderr, "\n");
return EXIT_SUCCESS;
}
if (isatty(STDIN_FILENO))
tty_fd = STDIN_FILENO;
else
if (isatty(STDOUT_FILENO))
tty_fd = STDOUT_FILENO;
else
if (isatty(STDERR_FILENO))
tty_fd = STDERR_FILENO;
else {
fprintf(stderr, "This program only runs in a terminal or pseudoterminal.\n");
return EXIT_FAILURE;
}
if (tcgetattr(tty_fd, &master_oldterm) == -1) {
fprintf(stderr, "Cannot obtain termios settings: %s.\n", strerror(errno));
return EXIT_FAILURE;
}
if (ioctl(tty_fd, TIOCGWINSZ, &slave_size) == -1) {
fprintf(stderr, "Cannot obtain terminal window size: %s.\n", strerror(errno));
return EXIT_FAILURE;
}
if (install_winch() == -1) {
fprintf(stderr, "Cannot install SIGWINCH signal handler: %s.\n", strerror(errno));
return EXIT_FAILURE;
}
/* For our own terminal, we want RAW (nonblocking) I/O. */
memcpy(&master_newterm, &master_oldterm, sizeof (struct termios));
master_newterm.c_iflag &= ~(IGNBRK | BRKINT | PARMRK | ISTRIP | INLCR | IGNCR | ICRNL | IXON);
master_newterm.c_oflag &= ~OPOST;
master_newterm.c_lflag &= ~(ECHO | ECHONL | ICANON | ISIG | IEXTEN);
master_newterm.c_cflag &= ~(CSIZE | PARENB);
master_newterm.c_cflag |= CS8;
master_newterm.c_cc[VMIN] = 0;
master_newterm.c_cc[VTIME] = 0;
/* We'll use the same for the new terminal also. */
memcpy(&slave_newterm, &master_newterm, sizeof (struct termios));
/* Open log file */
log = fopen(argv[1], "w");
if (!log) {
fprintf(stderr, "%s: %s.\n", argv[1], strerror(errno));
return EXIT_FAILURE;
}
/* Execute binary in pseudoterminal */
if (pseudoterminal_run(&child_pid, &master_fd, argv[2], argv + 2, &slave_newterm, &slave_size) == -1) {
fprintf(stderr, "%s: %s.\n", argv[2], strerror(errno));
return EXIT_FAILURE;
}
fprintf(log, "Pseudoterminal has %d rows, %d columns (%d x %d pixels)\n",
slave_size.ws_row, slave_size.ws_col, slave_size.ws_xpixel, slave_size.ws_ypixel);
fflush(log);
/* Ensure the master pseudoterminal descriptor is nonblocking. */
fcntl(tty_fd, F_SETFL, O_NONBLOCK);
fcntl(master_fd, F_SETFL, O_NONBLOCK);
/* Pseudoterminal proxy. */
{
struct pollfd fds[2];
const size_t slavein_size = 8192;
unsigned char slavein_data[slavein_size];
size_t slavein_head = 0;
size_t slavein_tail = 0;
const size_t slaveout_size = 8192;
unsigned char slaveout_data[slaveout_size];
size_t slaveout_head = 0;
size_t slaveout_tail = 0;
while (1) {
int io = 0;
if (slavein_head < slavein_tail) {
ssize_t n = write(master_fd, slavein_data + slavein_head, slavein_tail - slavein_head);
if (n > 0) {
slavein_head += n;
io++;
fprintf(log, "Wrote %zd bytes to child pseudoterminal.\n", n);
fflush(log);
} else
if (n != -1) {
fprintf(log, "Error writing to child pseudoterminal: write() returned %zd.\n", n);
fflush(log);
} else
if (errno != EINTR && errno != EAGAIN && errno != EWOULDBLOCK) {
fprintf(log, "Error writing to child pseudoterminal: %s.\n", strerror(errno));
fflush(log);
}
}
if (slavein_head > 0) {
if (slavein_tail > slavein_head) {
memmove(slavein_data, slavein_data + slavein_head, slavein_tail - slavein_head);
slavein_tail -= slavein_head;
slavein_head = 0;
} else {
slavein_tail = 0;
slavein_head = 0;
}
}
if (slaveout_head < slaveout_tail) {
ssize_t n = write(tty_fd, slaveout_data + slaveout_head, slaveout_tail - slaveout_head);
if (n > 0) {
slaveout_head += n;
io++;
fprintf(log, "Wrote %zd bytes to parent terminal.\n", n);
fflush(log);
} else
if (n != -1) {
fprintf(log, "Error writing to parent terminal: write() returned %zd.\n", n);
fflush(log);
} else
if (errno != EINTR && errno != EAGAIN && errno != EWOULDBLOCK) {
fprintf(log, "Error writing to parent terminal: %s.\n", strerror(errno));
fflush(log);
}
}
if (slaveout_head > 0) {
if (slaveout_tail > slaveout_head) {
memmove(slaveout_data, slaveout_data + slaveout_head, slaveout_tail - slaveout_head);
slaveout_tail -= slaveout_head;
slaveout_head = 0;
} else {
slaveout_tail = 0;
slaveout_head = 0;
}
}
if (slavein_tail < slavein_size) {
ssize_t n = read(tty_fd, slavein_data + slavein_tail, slavein_size - slavein_tail);
if (n > 0) {
slavein_tail += n;
io++;
fprintf(log, "Read %zd bytes from parent terminal.\n", n);
fflush(log);
} else
if (!n) {
/* Ignore */
} else
if (n != -1) {
fprintf(log, "Error reading from parent terminal: read() returned %zd.\n", n);
fflush(log);
} else
if (errno != EINTR && errno != EAGAIN && errno != EWOULDBLOCK) {
fprintf(log, "Error reading from parent terminal: %s.\n", strerror(errno));
fflush(log);
}
}
if (slaveout_tail < slaveout_size) {
ssize_t n = read(master_fd, slaveout_data + slaveout_tail, slaveout_size - slaveout_tail);
if (n > 0) {
slaveout_tail += n;
io++;
fprintf(log, "Read %zd bytes from child pseudoterminal.\n", n);
fflush(log);
} else
if (!n) {
/* Ignore */
} else
if (n != -1) {
fprintf(log, "Error reading from child pseudoterminal: read() returned %zd.\n", n);
fflush(log);
} else
if (errno != EINTR && errno != EAGAIN && errno != EWOULDBLOCK) {
fprintf(log, "Error reading from child pseudoterminal: %s.\n", strerror(errno));
fflush(log);
}
}
/* If we did any I/O, retry. */
if (io > 0)
continue;
/* If child process has exited and its output buffer is empty, we're done. */
if (child_pid <= 0 && slaveout_head >= slaveout_tail)
break;
/* Check if the child process has exited. */
if (child_pid > 0) {
pid_t p = waitpid(child_pid, &child_status, WNOHANG);
if (p == child_pid) {
child_pid = -child_pid;
continue;
}
}
/* If both buffers are empty, we proxy also the termios settings. */
if (slaveout_head >= slaveout_tail && slavein_head >= slavein_tail)
if (tcgetattr(master_fd, &slave_newterm) == 0)
if (tcsetattr(tty_fd, TCSANOW, &slave_newterm) == 0)
master_newterm = slave_newterm;
/* Wait for I/O to become possible. */
/* fds[0] is parent terminal */
fds[0].fd = tty_fd;
fds[0].events = POLLIN | (slaveout_head < slaveout_tail ? POLLOUT : 0);
fds[0].revents = 0;
/* fds[1] is child pseudoterminal */
fds[1].fd = master_fd;
fds[1].events = POLLIN | (slavein_head < slaveout_head ? POLLOUT : 0);
fds[1].revents = 0;
/* Wait up to a second */
poll(fds, 2, 1000);
}
}
/* Report child process exit status to log. */
if (WIFEXITED(child_status)) {
if (WEXITSTATUS(child_status) == EXIT_SUCCESS)
fprintf(log, "Child process exited successfully.\n");
else
fprintf(log, "Child process exited with exit status %d.\n", WEXITSTATUS(child_status));
} else
if (WIFSIGNALED(child_status))
fprintf(log, "Child process died from signal %d.\n", WTERMSIG(child_status));
else
fprintf(log, "Child process lost.\n");
fflush(log);
fclose(log);
/* Discard pseudoterminal. */
close(master_fd);
/* Return original parent terminal settings. */
tcflush(tty_fd, TCIOFLUSH);
tcsetattr(tty_fd, TCSANOW, &master_oldterm);
return EXIT_SUCCESS;
}
Whenever the parent process receives a WINCH (window size change) signal, the new terminal window size is obtained from the parent terminal, then set to the child pseudoterminal.
For simplicity (and not providing code that can be used as-is), the example attempts nonblocking reads and writes whenever possible, and only polls (waits until input becomes available, or buffered data can be written) if all four fail. Also, if the buffers are empty then, it copies the terminal settings from the child pseudoterminal to the parent terminal.
Compile using e.g.
gcc -Wall -Wextra -O2 -c pseudoterminal.c
gcc -Wall -Wextra -O2 -c example.c
gcc -Wall -Wextra -O2 example.o pseudoterminal.o -o example
and run e.g. ./example nano.log nano test-file. This runs nano in a sub-pseudoterminal, reflecting everything in it to the parent terminal, and essentially acts as if you had simply ran nano test-file. (Press Ctrl+X to exit.)
However, every read and write is logged to the nano.log file. For simplicity, only the length is currently logged, but you can surely write a dumper function to also log the contents. (Because these contain control characters, you'll want to either escape all control characters, or dump the data in hexadecimal format.)
It is interesting to note that when the child process (last process with the pseudoterminal as their controlling terminal) exits, trying to read from the pseudoterminal master returns -1 with errno == EIO. This means that before treating that as a fatal error, one should reap processes in the child process process group (waitpid(-child_pid, &status, WNOHANG)); and if that returns -1 with errno = ECHILD, it means the EIO was caused by no process having the pseudoterminal slave open.
If we compare this to tmux or screen, we have implemented only a crude version of the part when "attached" to a running session. When the user (parent process, running in the parent terminal) "detaches" from a session, tmux and screen both leave a process collecting the output of the running command. (They do not just buffer everything, they tend to record the effects of the running command to a virtual terminal buffer – rows × columns array of printable glyphs and their attributes –, so that a limited/fixed amount of memory is needed to recover the terminal contents when re-attaching to it later on.)
When re-attaching to a session, the screen/tmux command connects to the existing process (usually using an Unix domain socket, which allows verifying the peer user ID, and also passing the descriptor (to the pseudoterminal master) between processes, so the new process can take the place of the old process, and the old process can exit.
If we set the TERM environment variable to say xterm-256color before executing the child binary, we could interpret everything we read from the pseudoterminal master side in terms of how 256-color xterm does, and e.g. draw the screen using e.g. GTK+ – that is how we'd write our own terminal emulator.
I am trying to figure out how to do the following:
create a new pseudo-terminal
open a ncurses screen running inside the (slave) pseudo terminal
fork
A) forward I/O from the terminal the program is running in (bash) to the new (slave) terminal OR
B) exit leaving the ncurses program running in the new pty.
You seem to have a fundamental misconception of pseudoterminal pairs, and especially the importance of a process being the pseudoterminal master. Without the master, and a process managing the master side, there is literally no pseudoterminal pair: when the master is closed, the kernel forcibly removes the slave too, invalidating the file descriptors the slave has open to the slave side of the pseudoterminal pair.
Above, you completely ignore the role of the master, and wonder why what you want is not working.
My answer shows to accomplish 4.A), with any binary running as the slave, with the program itself being the master, proxying data between the slave pseudoterminal and the master terminal.
Reversing the role, with your "main program" telling some other binary to be the master terminal, is simple: write your own "main program" as a normal ncurses program, but run it using my example program to manage the master side of the pseudoterminal pair. This way signal propagation et cetera works correctly.
If you want to swap the roles, with the pseudoterminal slave being the parent process and the pseudoterminal master being the child process, you need to explain exactly why, when the entire interface has been designed for the opposite.
No, there is no "just a generic master pseudoterminal program or library you can use for this". The reason is that such makes no sense. Whenever you need a pseudoterminal pair, the master is the reason you want one. Any standard stream using human-readable text producing or consuming program is a valid client, using the slave end. They are not important, only the master is.
Can anyone provide pointers to what I might be doing wrong or that would make sense of some of this
I tried that, but you didn't appreciate the effort. I am sorry I tried.
or even better an example program using newterm() with either posix_openpt(), openpty() or forkpty().
No, because your newterm() makes absolutely no sense.
Glärbo's answers have helped me understand the problems enough that after some experimentation I believe I can answer my remaining questions directly.
The important points are:
The master side of the pty must remain opened
The file descriptor for the slave must be opened in the same mode as originally created.
without setsid() on the slave it remains connected to the original controlling terminal.
You need to be careful with ncurses calls when using newterm rather tha initscr
The master side of the pty must remain opened
Me: "If I instruct the parent process exits the child terminates without anything interesting being logged by the child."
Glärbo: "Without the master, and a process managing the master side, there is literally no pseudoterminal pair: when the master is closed, the kernel forcibly removes the slave too, invalidating the file descriptors the slave has open to the slave side of the pseudoterminal pair."
The file descriptor for the slave must be opened in the same mode as originally created.
My incorrect pseudo code (for the child side of the fork):
FILE* scrIn = open(slave,O_RDWR|O_NONBLOCK);
FILE* scrOut = open(slave,O_RDWR|O_NONBLOCK);
SCREEN* scr = newterm(NULL,scrIn,scrOut);
Works if replaced with (error checking omitted):
setsid();
close(STDIN_FILENO);
close(STDOUT_FILENO);
const char* slave_pts = pstname(master);
int slave = open(slave_pts, O_RDWR);
ioctl(slave(TIOCTTY,0);
close(master);
dup2(slave,STDIN_FILENO);
dup2(slave,STDOUT_FILENO);
FILE* slaveFile = fdopen(slavefd,"r+");
SCREEN* scr = newterm(NULL,slaveFile,slaveFile);
(void)set_term(scr);
printw("hello world\n"); // print to the in memory represenation of the curses window
refresh(); // copy the in mem rep to the actual terminal
I think a bad file or file descriptor must have crept through somewhere without being checked. This explains the segfault inside fileno_unlocked().
Also I had tried in some experiments opening the slave twice. Once for reading and once for writing. The mode would have conflicted with the mode of the original fd.
Without setsid() on the child side (with the slave pty) the child process still has the original controlling terminal.
setsid() makes the process a session leader. Only the session leader can change its controlling terminal.
ioctl(slave(TIOCTTY,0) - make slave the controlling terminal
You need to be careful with ncurses calls when using newterm() rather tha initscr()
Many ncurses functions have an implicit "intscr" argument which refers to a screen or window created for the controlling terminals STDIN and STDOUT. They doen't work unless replaced with the equivalent ncurses() functions for a specified WINDOW. You need to call newwin() to create a WINDOW, newterm() only gives you a screen.
In fact I am still wrestling with this kind of issue such as a call to subwin() which fails when the slave pty is used but not with the normal terminal.
It is also noteworthy that:
You need to handle SIGWINCH in the process connected to an actual terminal and pass that to the slave if it needs to knows the terminal size has changed.
You probably need a pipe to daemon to pass additional information.
I left stderr connect to the original terminal above for convenience of debugging. That would be closed in practice.
attach a terminal to a process running as a daemon (to run an ncurses UI) does a better job of describing the use case than the specific issues troubleshooted here.

Wrong write to file output order of synchronized processes?

I have the following problem.
I have two processes that are being synchronized with semaphores and the idea is this:
process 1 writes something to the txt file
process 2 writes something to the txt file
process 1 writes something to the test file
I have included this sample code that demonstrates the problem:
// semaphore names
#define NAME1 "/s1"
#define NAME2 "/s2"
int main()
{
/* semaphores for process synchronization */
sem_t *sm1;
sm1 = sem_open( NAME1, O_CREAT, 0666, 0);
sem_t *sm2;
sm2 = sem_open(NAME2, O_CREAT, 0666, 0);
/* processes*/
int proc1;
int proc2;
/* file lock struct */
struct flock fl = {F_WRLCK, SEEK_SET, 0, 0, 0 };
/* create a text file */
FILE *fp;
int fd;
fp = fopen("output.txt", "w"); // create and close the file
fclose(fp);
if((fd = open("output.txt", O_RDWR)) == -1) { // open the file again to get file descriptor
perror("open");
exit(1);
}
fp = fdopen(fd, "w");
/* first process */
if ((proc1 = fork()) < 0) {
perror("fork");
exit(2);
}
else if(proc1 == 0) {
fl.l_type = F_WRLCK; // set the lock type and pid of the forked process
fl.l_pid = getpid();
if (fcntl(fd, F_SETLKW, &fl) == -1) { // lock the file before writing to it
perror("fcntl");
exit(1);
}
fprintf(fp, "proc1 - action1\n"); // write to the file
fl.l_type = F_UNLCK;
if (fcntl(fd, F_SETLK, &fl) == -1) { // unlock the file so other processes can write to it
perror("fcntl");
exit(1);
}
fprintf(stdout, "proc1 - action1\n");
sem_post(sm1); // let the second process run
sem_wait(sm2); // wait till the second process is done
// write one more thing the same way to the text file after the second process is done
fl.l_type = F_WRLCK;
fl.l_pid = getpid();
if (fcntl(fd, F_SETLKW, &fl) == -1) {
perror("fcntl");
exit(1);
}
fprintf(fp, "proc1 - action2\n");
fl.l_type = F_UNLCK;
if (fcntl(fd, F_SETLK, &fl) == -1) {
perror("fcntl");
exit(1);
}
fprintf(stdout, "proc1 - action2\n");
exit(0);
}
/* second process */
if ((proc2 = fork()) < 0) {
perror("fork");
exit(2);
}
else if(proc2 == 0) {
sem_wait(sm1); // waits for proc1 to perform it's first action
// write something to the text file and let proc1 write it's second action
fl.l_type = F_WRLCK;
fl.l_pid = getpid();
if (fcntl(fd, F_SETLKW, &fl) == -1) {
perror("fcntl");
exit(1);
}
fprintf(fp, "proc2 - action1\n");
fl.l_type = F_UNLCK;
if (fcntl(fd, F_SETLK, &fl) == -1) {
perror("fcntl");
exit(1);
}
fprintf(stdout, "proc2 - action1\n");
sem_post(sm2);
exit(0);
}
// wait for both processes to finish
waitpid(proc1, NULL, 0);
waitpid(proc2, NULL, 0);
sem_close(sm1);
sem_unlink(NAME1);
sem_close(sm2);
sem_unlink(NAME2);
return 0;
}
I have included the fprintf with stdout lines so that you can see that the output in the terminal is correct:
-proc1 - action1
-proc2 - action1
-proc1 - action2
Just like they are being synchronized. However, the output in the output.txt file is this:
-proc2 - action1
-proc1 - action1
-proc1 - action2
Why is this happening?
Also, before the process writes to the file, I always lock it so that no other process can access it and then unlock it again.
I'm not sure if I'm doing this right so I'd appreciate any advice I can get!
Thanks a lot!
You have buffered IO by default - so the buffer isn't actually flushed to the file until you close it if it is smaller output than the buffer size (otherwise you get a buffer's worth at a time when it is full... often 8kb or so). Note also that each process gets its own, separate buffer.
To fix, flush your output after printing, and before writing, fseek() to SEEK_END to ensure you are at the end of the file.
fprintf(fp, "proc1 - action1\n"); // write to the file
fflush(fp);
BTW, i didn't check for the validity of your semaphore code, just noticed you were missing this crucial piece of information. In general, however, if you are using file locking, the semaphore is redundant...
fp = fopen("output.txt", "w"); will truncate the file and is not synchronized, so process #2 may truncate the file even if process #1 has already written something to it. Same for fp = fdopen(fd, "w");
Also, you are heading for trouble if you keep the file open while the other process writes to it. At least you should SEEK to the (new) end of the file after the second process finished its writing or you may overwrite what process #2 did. Better close the file before the other process runs and reopen afterwards. This will also make sure any buffered data is flushed.

fcntl F_GETLK always return true

i'm trying to make a single instance daemon using a lock file but fcntl() doesn't seem to work as expected...
int creat_lock_file (char * pid_fn)
{
struct flock pid_lck = {F_WRLCK, SEEK_SET, 0, 0, 0 };
/* Open/Create pid file */
int pid_fd = open (pid_fn, O_CREAT | O_WRONLY, 0640);
if (pid_fd == -1)
{
syslog (LOG_ERR, "Unable to open PID file > [Errno: %s]", strerror(errno));
return -1;
}
/* Place write lock on pid file */
if (fcntl(pid_fd, F_SETLK, &pid_lck) == -1) {
/* Unhandled error ocured */
syslog (LOG_ERR, "Unhandled error ocured while locking PID file > [Errno: %s]", strerror(errno));
close (pid_fd);
return -1;
}
/* Write PID to lock file */
char pid_lock_buf[11];
sprintf (pid_lock_buf, "%ld\n", (long) getpid ());
write (pid_fd, pid_lock_buf, strlen (pid_lock_buf)+1);
return 0;
}
int get_lock_file_status (char * pid_fn)
{
struct flock pid_lck = {F_WRLCK, SEEK_SET, 0, 0, 0 };
/* Open/Create pid file */
int pid_fd = open (pid_fn, O_CREAT | O_WRONLY, 0640);
if (pid_fd == -1)
{
syslog (LOG_ERR, "Unable to open PID file > [Errno: %s]", strerror(errno));
return -1;
}
/* try locking pid file*/
if(fcntl(pid_fd, F_GETLK, &pid_lck) == -1)
{
if(errno == EAGAIN || errno == EACCES) /* file already locked, close fd and return -1 */
{
close (pid_fd);
return -1;
}
else /* Unhandled error ocured */
{
syslog (LOG_ERR, "Unhandled error ocured while locking PID file > [Errno: %s]", strerror(errno));
close (pid_fd);
return -1;
}
}
close (pid_fd);
return 0;
}
so i call get_lock_file_status and exit if it returns -1 to make sure no other instance is running than i do some stuff (fork chdir etc) and call creat_lock_file to crate and lock a pid file after successfully creating a daemon...
when complied and run the program works as expected runs creates lock file and writes pid to it but when a second instance is started the second instance simply opens the same lock file and writes it's own pid to it!
what am i doing wrong? shouldn't the second instance return -1 in get_lock_file_status?
You're checking the result of F_GETLK in the wrong way. fcntl(2) with F_GETLK only returns -1 on errors. The correct way to check if it would be possible to acquire the lock is to check if the l_type field of the struct flock is set to F_UNLCK, like in the following:
/* try locking pid file*/
if(fcntl(pid_fd, F_GETLK, &pid_lck) == -1) {
syslog (LOG_ERR, "Unhandled error ocured while locking PID file > [Errno: %s]", strerror(errno));
close(pid_fd);
return -1;
}
close(pid_fd);
return (pid_lck.l_type == F_UNLCK) ? 0 : -1;
It should be possible to roll creat_lock_file() and get_lock_file_status() into a single function that opens the file (and creates it if it doesn't exist), tries to set a lock on it, and returns whether the locking was successful (e.g., either a file descriptor or -1).
You should truncate(2) the PID file to zero bytes before writing the PID into it by the way. Say the PID of your process is 5 and the old PID in the PID file is 123. Writing "5" would then make the contents of the PID file "523". Truncating the file to zero bytes solves this problem. (O_TRUNC won't work since you'd be clearing the file when opening it to test if the lock is set.)
Removing the PID file with unlink(2) when the program exits is pretty common too. That way the non-existence of the file indicates the daemon isn't running (though it's not foolproof as the process or system could crash).

linux system c quoting

On an embedded system with Linux as OS I want to call a 3rd party binary to retrieve data from a database and to append this data to an existing file.
The command string which is handed over to the system() function looks like:
"export_from_db >> /tmp/myFile"
Unfortunately, this doesn't work. /tmp/myFile never gets created! If I omit the redirection, then the database export is printed to stdout.
I wonder if system() and the redirection via ">>" go together well? On the prompt I successfully tested this command "export_fom_db >> /tmp/myFile"! Does anybody know how to achieve it using system()? Is there some kind of quoting necessary?
hm.. Actually, it seems ok for me.. this is exactly what system() is for - to execute a line under current shell. Does that embedded linux's shell support >> operator? Have you tried it manually in the terminal?
Another thought is that your application may be run under some other user account and that account could have some weird configuration, like having some csh or ksh instead of bash (or viceversa, depending on what you like). Check what user actually owns the process and check the /etc/passwd for the shell setup.
Also, there is small possibility that the user account that the app runs under simply does not have rights to write to /tmp :) be sure to check that too
Also... there is small possibility that on yours 'embedded linux' simply has the system() implemented in a simplistic way, that just invokes the application with given parameters and skips all other shell-wise operators. This could have been done to save on resources, as system() might be though to be rarely used, or just though to be "too heavy" by your linux designers.. it depends on the distro.. If you tell us which one it is, then people with more knowledge will probably be able to say if this is the case.
On an embedded system, you are better off implementing the system() yourself. Consider the following code (untested!):
#include <unistd.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <errno.h>
/* Helper function: Open the specified file at the desired descriptor.
*/
static int openfd(const int descriptor,
const char *const filename,
const int flags, const mode_t mode)
{
int fd, result;
if (!filename || descriptor == -1)
return errno = EINVAL;
/* Close existing descriptor. Ignore errors. Hopefully it is reused. */
do {
result = close(descriptor);
} while (result == -1 && errno == EINTR);
/* Open the desired file. */
do {
fd = open(filename, flags, mode);
} while (fd == -1 && errno == EINTR);
if (fd == -1)
return errno;
/* Did we get lucky, and get the correct descriptor already? */
if (fd == descriptor)
return 0;
/* Move the descriptor. */
do {
result = dup2(fd, descriptor);
} while (result == -1 && errno == EINTR);
if (result == -1) {
const int saved_errno = errno;
do {
result = close(fd);
} while (result == -1 && errno == EINTR);
return errno = saved_errno;
}
/* Close the temporary descriptor. */
do {
result = close(fd);
} while (result == -1 && errno == EINTR);
if (result == -1) {
const int saved_errno = errno;
do {
result = close(descriptor);
} while (result == -1 && errno == EINTR);
return errno = saved_errno;
}
return 0;
}
/* Start command on the background.
* Note: args[1] is the first argument, args[0] is the command name.
* NULL input/output/error redirects from/to /dev/null.
* Empty string for input/output/error does no redirections;
* the command the uses the same input/output/error.
* For non-empty output or error, specify the 'man 2 open' O_ flags too.
*
* Returns (pid_t)0 with errno set if an error occurs,
* otherwise the PID of the child process started.
*/
pid_t run(const char *file,
char *const args[],
const char *const input,
const char *const output, const int output_flags,
const char *const error, const int error_flags)
{
pid_t child;
int result, flags;
if (!cmd || !arg || !arg[0]) {
errno = EINVAL;
return (pid_t)0;
}
child = fork();
if (child == (pid_t)-1)
return (pid_t)0;
if (child)
return child;
/* This is the child process. */
if (input && *input)
result = openfd(STDIN_FILENO, input, O_RDONLY | O_NOCTTY, 0);
else
if (!input)
result = openfd(STDIN_FILENO, "/dev/null", O_RDONLY | O_NOCTTY, 0);
else
result = 0;
if (result)
exit(127);
if (output && *output)
result = openfd(STDOUT_FILENO, output, output_flags, 0666);
else
if (!output)
result = openfd(STDOUT_FILENO, "/dev/null", O_WRONLY | O_NOCTTY, 0);
else
result = 0;
if (result)
exit(127);
if (error && *error)
result = openfd(STDERR_FILENO, error, error_flags, 0666);
else
if (!error)
result = openfd(STDERR_FILENO, "/dev/null", O_WRONLY | O_NOCTTY, 0);
else
result = 0;
if (result)
exit(127);
execvp(file, args);
exit(127);
}
The run() only starts the command, you'll need to wait for it to complete. Note that your main program can do meaningful work at the same time, unless it needs the exit status (or the files the command is supposed to create) right away. Example use:
/* Command to run. NULL terminates the list. */
char *const cmd[] = { "ls", "-l", NULL };
pid_t child, p;
int status;
child = run(cmd[0], cmd,
NULL /* < /dev/null */,
"/tmp/some-log-file", O_WRONLY | O_CREAT | O_APPEND,
"", 0 /* No redirection for standard error */);
if (!child) {
fprintf(stderr, "Cannot run '%s': %s.\n", cmd[0], strerror(errno));
exit(1);
}
do {
status = 0;
p = waitpid(child, &status, 0);
} while (p == (pid_t)-1 && errno == EINTR);
if (p == (pid_t)-1) {
fprintf(stderr, "Lost '%s': %s.\n", cmd[0], strerror(errno));
exit(1);
}
if (WIFEXITED(status)) {
if (!WEXITSTATUS(status))
printf("Command executed successfully.\n");
else
printf("Command failed with exit status %d.\n", WEXITSTATUS(status));
} else
if (WSIGNALED(status))
printf("Command died from signal %s.\n", strsignal(WTERMSIG(status)));
else
printf("Command died unexpectedly.\n");
although the last part is often abbreviated to
if (WIFEXITED(status) && !WEXITSTATUS(status))
printf("'%s': Successful.\n", cmd[0]);
else
printf("'%s': Failed.\n", cmd[0]);
Note that if you process the output anyway, you probably should use a pipe (either popen() or an extended version of the above function) instead.
Hope you find this useful.

filelock allows opening of files

i executed the following c code on my ubuntu machine...i have read about fcntl()'s use to lock a file and even reading will not be allowed if F_WRLCK opton is set...so i started this program and before relinquishing the lock by pressing enter i tried to open the file in two ways- by directly double clicking on file1.cpp and by running a different c program in a new terminal...both the time the file was opened...so how did fcntl() allow opening of these files when F_WRLCK is set...
int main(int argc, char *argv[])
{
struct flock fl = {F_WRLCK, SEEK_SET, 0, 0, 0 };
int fd;
fl.l_pid = getpid();
if ((fd = open("/home/file1.cpp", O_WRONLY)) == -1)
{
perror("open");
exit(1);
}
if (fcntl(fd, F_SETLKW, &fl) == -1)
{
perror("fcntl");
exit(1);
}
printf("Press <RETURN> to release lock: ");
getchar();
fl.l_type = F_UNLCK; /* set to unlock same region */
if (fcntl(fd, F_SETLK, &fl) == -1)
{
perror("fcntl");
exit(1);
}
printf("Unlocked.\n");
close(fd);
return 0;
}
fcntl locks are purely advisory locks. Their only effect is to cause fcntl F_SETLK calls to block when the lock cannot be obtained. They have absolutely no effect on IO operations. It's up to your program to obtain the locks it needs before performing IO when synchronization is necessary.
This is completely analogous to using mutexes to protect objects in memory. The mutex will not prevent you from reading or writing memory addresses; it's just a protocol you use to make sure your program only reads and writes and proper times.

Resources