/* xwrap.cpp - wrappers around existing library functions. * * Functions with the x prefix are wrappers that either succeed or log the * error message. They usually have the same arguments and return value * as the function they wrap. * */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "logging.h" #include "utils.h" FILE *xfopen(const char *pathname, const char *mode) { FILE *fp = fopen(pathname, mode); if (fp == NULL) { PLOGE("fopen: %s", pathname); } return fp; } FILE *xfdopen(int fd, const char *mode) { FILE *fp = fdopen(fd, mode); if (fp == NULL) { PLOGE("fopen"); } return fp; } int xopen2(const char *pathname, int flags) { int fd = open(pathname, flags); if (fd < 0) { PLOGE("open: %s", pathname); } return fd; } int xopen3(const char *pathname, int flags, mode_t mode) { int fd = open(pathname, flags, mode); if (fd < 0) { PLOGE("open: %s", pathname); } return fd; } int xopenat(int dirfd, const char *pathname, int flags) { int fd = openat(dirfd, pathname, flags); if (fd < 0) { PLOGE("openat: %s", pathname); } return fd; } ssize_t xwrite(int fd, const void *buf, size_t count) { int ret = write(fd, buf, count); if (count != ret) { PLOGE("write"); } return ret; } // Read error other than EOF ssize_t xread(int fd, void *buf, size_t count) { int ret = read(fd, buf, count); if (ret < 0) { PLOGE("read"); } return ret; } // Read exact same size as count ssize_t xxread(int fd, void *buf, size_t count) { int ret = read(fd, buf, count); if (count != ret) { PLOGE("read (%d != %d)", count, ret); } return ret; } int xpipe2(int pipefd[2], int flags) { int ret = pipe2(pipefd, flags); if (ret == -1) { PLOGE("pipe2"); } return ret; } int xsetns(int fd, int nstype) { int ret = (int) syscall(__NR_setns, fd, nstype); if (ret == -1) { PLOGE("setns"); } return ret; } int xunshare(int flags) { int ret = (int) syscall(__NR_unshare, flags); if (ret == -1) { PLOGE("unshare"); } return ret; } DIR *xopendir(const char *name) { DIR *d = opendir(name); if (d == NULL) { PLOGE("opendir: %s", name); } return d; } DIR *xfdopendir(int fd) { DIR *d = fdopendir(fd); if (d == NULL) { PLOGE("fdopendir"); } return d; } struct dirent *xreaddir(DIR *dirp) { errno = 0; struct dirent *e = readdir(dirp); if (errno && e == NULL) { PLOGE("readdir"); } return e; } pid_t xsetsid() { pid_t pid = setsid(); if (pid == -1) { PLOGE("setsid"); } return pid; } int xsocket(int domain, int type, int protocol) { int fd = socket(domain, type, protocol); if (fd == -1) { PLOGE("socket"); } return fd; } int xbind(int sockfd, const struct sockaddr *addr, socklen_t addrlen) { int ret = bind(sockfd, addr, addrlen); if (ret == -1) { PLOGE("bind"); } return ret; } int xlisten(int sockfd, int backlog) { int ret = listen(sockfd, backlog); if (ret == -1) { PLOGE("listen"); } return ret; } static int accept4_compat(int sockfd, struct sockaddr *addr, socklen_t *addrlen, int flags) { int fd = accept(sockfd, addr, addrlen); if (fd == -1) { PLOGE("accept"); } else { if (flags & SOCK_CLOEXEC) fcntl(fd, F_SETFD, FD_CLOEXEC); if (flags & SOCK_NONBLOCK) { int i = fcntl(fd, F_GETFL); fcntl(fd, F_SETFL, i | O_NONBLOCK); } } return fd; } int xaccept4(int sockfd, struct sockaddr *addr, socklen_t *addrlen, int flags) { int fd = (int) syscall(__NR_accept4, sockfd, addr, addrlen, flags); if (fd == -1) { if (errno == ENOSYS) return accept4_compat(sockfd, addr, addrlen, flags); PLOGE("accept4"); } return fd; } void *xmalloc(size_t size) { void *p = malloc(size); if (p == NULL) { PLOGE("malloc"); } return p; } void *xcalloc(size_t nmemb, size_t size) { void *p = calloc(nmemb, size); if (p == NULL) { PLOGE("calloc"); } return p; } void *xrealloc(void *ptr, size_t size) { void *p = realloc(ptr, size); if (p == NULL) { PLOGE("realloc"); } return p; } ssize_t xsendmsg(int sockfd, const struct msghdr *msg, int flags) { int sent = sendmsg(sockfd, msg, flags); if (sent == -1) { PLOGE("sendmsg"); } return sent; } ssize_t xrecvmsg(int sockfd, struct msghdr *msg, int flags) { int rec = recvmsg(sockfd, msg, flags); if (rec == -1) { PLOGE("recvmsg"); } return rec; } int xpthread_create(pthread_t *thread, const pthread_attr_t *attr, void *(*start_routine) (void *), void *arg) { errno = pthread_create(thread, attr, start_routine, arg); if (errno) { PLOGE("pthread_create"); } return errno; } int xstat(const char *pathname, struct stat *buf) { int ret = stat(pathname, buf); if (ret == -1) { PLOGE("stat %s", pathname); } return ret; } int xlstat(const char *pathname, struct stat *buf) { int ret = lstat(pathname, buf); if (ret == -1) { PLOGE("lstat %s", pathname); } return ret; } int xdup2(int oldfd, int newfd) { int ret = dup2(oldfd, newfd); if (ret == -1) { PLOGE("dup2"); } return ret; } int xdup3(int oldfd, int newfd, int flags) { int ret = (int) syscall(__NR_dup3, oldfd, newfd, flags); if (ret == -1) { PLOGE("dup3"); } return ret; } ssize_t xreadlink(const char *pathname, char *buf, size_t bufsiz) { ssize_t ret = readlink(pathname, buf, bufsiz); if (ret == -1) { PLOGE("readlink %s", pathname); } else { buf[ret] = '\0'; } return ret; } ssize_t xreadlinkat(int dirfd, const char *pathname, char *buf, size_t bufsiz) { ssize_t ret = syscall(__NR_readlinkat, dirfd, pathname, buf, bufsiz); if (ret == -1) { PLOGE("readlinkat %s", pathname); } else { buf[ret] = '\0'; } return ret; } int xsymlink(const char *target, const char *linkpath) { int ret = symlink(target, linkpath); if (ret == -1) { PLOGE("symlink %s->%s", target, linkpath); } return ret; } int xsymlinkat(const char *target, int newdirfd, const char *linkpath) { int ret = (int) syscall(__NR_symlinkat, target, newdirfd, linkpath); if (ret == -1) { PLOGE("symlinkat %s->%s", target, linkpath); } return ret; } int xlinkat(int olddirfd, const char *oldpath, int newdirfd, const char *newpath, int flags) { int ret = (int) syscall(__NR_linkat, olddirfd, oldpath, newdirfd, newpath, flags); if (ret == -1) { PLOGE("linkat %s->%s", oldpath, newpath); } return ret; } int xmount(const char *source, const char *target, const char *filesystemtype, unsigned long mountflags, const void *data) { int ret = mount(source, target, filesystemtype, mountflags, data); if (ret == -1) { PLOGE("mount %s->%s", source, target); } return ret; } int xumount(const char *target) { int ret = umount(target); if (ret == -1) { PLOGE("umount %s", target); } return ret; } int xumount2(const char *target, int flags) { int ret = umount2(target, flags); if (ret == -1) { PLOGE("umount2 %s", target); } return ret; } int xrename(const char *oldpath, const char *newpath) { int ret = rename(oldpath, newpath); if (ret == -1) { PLOGE("rename %s->%s", oldpath, newpath); } return ret; } int xmkdir(const char *pathname, mode_t mode) { int ret = mkdir(pathname, mode); if (ret == -1 && errno != EEXIST) { PLOGE("mkdir %s %u", pathname, mode); } return ret; } int xmkdirs(const char *pathname, mode_t mode) { int ret = mkdirs(pathname, mode); if (ret == -1) { PLOGE("mkdirs %s", pathname); } return ret; } int xmkdirat(int dirfd, const char *pathname, mode_t mode) { int ret = mkdirat(dirfd, pathname, mode); if (ret == -1 && errno != EEXIST) { PLOGE("mkdirat %s %u", pathname, mode); } return ret; } void *xmmap(void *addr, size_t length, int prot, int flags, int fd, off_t offset) { void *ret = mmap(addr, length, prot, flags, fd, offset); if (ret == MAP_FAILED) { PLOGE("mmap"); } return ret; } ssize_t xsendfile(int out_fd, int in_fd, off_t *offset, size_t count) { ssize_t ret = sendfile(out_fd, in_fd, offset, count); if (count != ret) { PLOGE("sendfile"); } return ret; } pid_t xfork() { int ret = fork(); if (ret == -1) { PLOGE("fork"); } return ret; } int xpoll(struct pollfd *fds, nfds_t nfds, int timeout) { int ret = poll(fds, nfds, timeout); if (ret == -1) { PLOGE("poll"); } return ret; }