/* $OpenBSD: monitor.c,v 1.72 2010/05/10 02:00:50 krw Exp $ */ /* * Copyright (c) 2003 Håkan Olsson. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "conf.h" #include "log.h" #include "monitor.h" #include "policy.h" #include "ui.h" #include "util.h" #include "pf_key_v2.h" struct monitor_state { pid_t pid; int s; char root[MAXPATHLEN]; } m_state; extern char *pid_file; extern void set_slave_signals(void); /* Private functions. */ static void must_read(void *, size_t); static void must_write(const void *, size_t); static void m_priv_getfd(void); static void m_priv_setsockopt(void); static void m_priv_req_readdir(void); static void m_priv_bind(void); static void m_priv_pfkey_open(void); static int m_priv_local_sanitize_path(char *, size_t, int); static int m_priv_check_sockopt(int, int); static int m_priv_check_bind(const struct sockaddr *, socklen_t); static void set_monitor_signals(void); static void sig_pass_to_chld(int); /* * Public functions, unprivileged. */ /* Setup monitor context, fork, drop child privs. */ pid_t monitor_init(int debug) { struct passwd *pw; int p[2]; bzero(&m_state, sizeof m_state); if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, p) != 0) log_fatal("monitor_init: socketpair() failed"); pw = getpwnam(ISAKMPD_PRIVSEP_USER); if (pw == NULL) log_fatalx("monitor_init: getpwnam(\"%s\") failed", ISAKMPD_PRIVSEP_USER); strlcpy(m_state.root, pw->pw_dir, sizeof m_state.root); set_monitor_signals(); m_state.pid = fork(); if (m_state.pid == -1) log_fatal("monitor_init: fork of unprivileged child failed"); if (m_state.pid == 0) { /* The child process drops privileges. */ set_slave_signals(); if (chroot(pw->pw_dir) != 0 || chdir("/") != 0) log_fatal("monitor_init: chroot failed"); if (setgroups(1, &pw->pw_gid) == -1 || setresgid(pw->pw_gid, pw->pw_gid, pw->pw_gid) || setresuid(pw->pw_uid, pw->pw_uid, pw->pw_uid)) log_fatal("monitor_init: can't drop privileges"); m_state.s = p[0]; close(p[1]); LOG_DBG((LOG_MISC, 10, "monitor_init: privileges dropped for child process")); } else { /* Privileged monitor. */ setproctitle("monitor [priv]"); m_state.s = p[1]; close(p[0]); } /* With "-dd", stop and wait here. For gdb "attach" etc. */ if (debug > 1) { log_print("monitor_init: stopped %s PID %d fd %d%s", m_state.pid ? "priv" : "child", getpid(), m_state.s, m_state.pid ? ", waiting for SIGCONT" : ""); kill(getpid(), SIGSTOP); /* Wait here for SIGCONT. */ if (m_state.pid) kill(m_state.pid, SIGCONT); /* Continue child. */ } return m_state.pid; } void monitor_exit(int code) { int status; pid_t pid; if (m_state.pid != 0) { /* When called from the monitor, kill slave and wait for it */ kill(m_state.pid, SIGTERM); do { pid = waitpid(m_state.pid, &status, 0); } while (pid == -1 && errno == EINTR); /* Remove FIFO and pid files. */ unlink(ui_fifo); unlink(pid_file); } close(m_state.s); exit(code); } int monitor_pf_key_v2_open(void) { int err, cmd; cmd = MONITOR_PFKEY_OPEN; must_write(&cmd, sizeof cmd); must_read(&err, sizeof err); if (err < 0) { log_error("monitor_pf_key_v2_open: parent could not create " "PF_KEY socket"); return -1; } pf_key_v2_socket = mm_receive_fd(m_state.s); if (pf_key_v2_socket < 0) { log_error("monitor_pf_key_v2_open: mm_receive_fd() failed"); return -1; } return pf_key_v2_socket; } int monitor_open(const char *path, int flags, mode_t mode) { size_t len; int fd, err, cmd; char pathreal[MAXPATHLEN]; if (path[0] == '/') strlcpy(pathreal, path, sizeof pathreal); else snprintf(pathreal, sizeof pathreal, "%s/%s", m_state.root, path); cmd = MONITOR_GET_FD; must_write(&cmd, sizeof cmd); len = strlen(pathreal); must_write(&len, sizeof len); must_write(&pathreal, len); must_write(&flags, sizeof flags); must_write(&mode, sizeof mode); must_read(&err, sizeof err); if (err != 0) { errno = err; return -1; } fd = mm_receive_fd(m_state.s); if (fd < 0) { log_error("monitor_open: mm_receive_fd () failed"); return -1; } return fd; } FILE * monitor_fopen(const char *path, const char *mode) { FILE *fp; int fd, flags = 0, saved_errno; mode_t mask, cur_umask; /* Only the child process is supposed to run this. */ if (m_state.pid) log_fatal("[priv] bad call to monitor_fopen"); switch (mode[0]) { case 'r': flags = (mode[1] == '+' ? O_RDWR : O_RDONLY); break; case 'w': flags = (mode[1] == '+' ? O_RDWR : O_WRONLY) | O_CREAT | O_TRUNC; break; case 'a': flags = (mode[1] == '+' ? O_RDWR : O_WRONLY) | O_CREAT | O_APPEND; break; default: log_fatal("monitor_fopen: bad call"); } cur_umask = umask(0); (void)umask(cur_umask); mask = S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH; mask &= ~cur_umask; fd = monitor_open(path, flags, mask); if (fd < 0) return NULL; /* Got the fd, attach a FILE * to it. */ fp = fdopen(fd, mode); if (!fp) { log_error("monitor_fopen: fdopen() failed"); saved_errno = errno; close(fd); errno = saved_errno; return NULL; } return fp; } int monitor_stat(const char *path, struct stat *sb) { int fd, r, saved_errno; /* O_NONBLOCK is needed for stat'ing fifos. */ fd = monitor_open(path, O_RDONLY | O_NONBLOCK, 0); if (fd < 0) return -1; r = fstat(fd, sb); saved_errno = errno; close(fd); errno = saved_errno; return r; } int monitor_setsockopt(int s, int level, int optname, const void *optval, socklen_t optlen) { int ret, err, cmd; cmd = MONITOR_SETSOCKOPT; must_write(&cmd, sizeof cmd); if (mm_send_fd(m_state.s, s)) { log_print("monitor_setsockopt: read/write error"); return -1; } must_write(&level, sizeof level); must_write(&optname, sizeof optname); must_write(&optlen, sizeof optlen); must_write(optval, optlen); must_read(&err, sizeof err); must_read(&ret, sizeof ret); if (err != 0) errno = err; return ret; } int monitor_bind(int s, const struct sockaddr *name, socklen_t namelen) { int ret, err, cmd; cmd = MONITOR_BIND; must_write(&cmd, sizeof cmd); if (mm_send_fd(m_state.s, s)) { log_print("monitor_bind: read/write error"); return -1; } must_write(&namelen, sizeof namelen); must_write(name, namelen); must_read(&err, sizeof err); must_read(&ret, sizeof ret); if (err != 0) errno = err; return ret; } int monitor_req_readdir(const char *filename) { int cmd, err; size_t len; cmd = MONITOR_REQ_READDIR; must_write(&cmd, sizeof cmd); len = strlen(filename); must_write(&len, sizeof len); must_write(filename, len); must_read(&err, sizeof err); if (err == -1) must_read(&errno, sizeof errno); return err; } int monitor_readdir(char *file, size_t size) { int fd; size_t len; must_read(&len, sizeof len); if (len == 0) return -1; if (len >= size) log_fatal("monitor_readdir: received bad length from monitor"); must_read(file, len); file[len] = '\0'; fd = mm_receive_fd(m_state.s); return fd; } void monitor_init_done(void) { int cmd; cmd = MONITOR_INIT_DONE; must_write(&cmd, sizeof cmd); } /* * Start of code running with privileges (the monitor process). */ static void set_monitor_signals(void) { int n; for (n = 1; n < _NSIG; n++) signal(n, SIG_DFL); /* Forward some signals to the child. */ signal(SIGTERM, sig_pass_to_chld); signal(SIGHUP, sig_pass_to_chld); signal(SIGUSR1, sig_pass_to_chld); } static void sig_pass_to_chld(int sig) { int oerrno = errno; if (m_state.pid > 0) kill(m_state.pid, sig); errno = oerrno; } /* This function is where the privileged process waits(loops) indefinitely. */ void monitor_loop(int debug) { int msgcode; if (!debug) log_to(0); for (;;) { must_read(&msgcode, sizeof msgcode); switch (msgcode) { case MONITOR_GET_FD: m_priv_getfd(); break; case MONITOR_PFKEY_OPEN: LOG_DBG((LOG_MISC, 80, "monitor_loop: MONITOR_PFKEY_OPEN")); m_priv_pfkey_open(); break; case MONITOR_SETSOCKOPT: LOG_DBG((LOG_MISC, 80, "monitor_loop: MONITOR_SETSOCKOPT")); m_priv_setsockopt(); break; case MONITOR_BIND: LOG_DBG((LOG_MISC, 80, "monitor_loop: MONITOR_BIND")); m_priv_bind(); break; case MONITOR_REQ_READDIR: LOG_DBG((LOG_MISC, 80, "monitor_loop: MONITOR_REQ_READDIR")); m_priv_req_readdir(); break; case MONITOR_INIT_DONE: LOG_DBG((LOG_MISC, 80, "monitor_loop: MONITOR_INIT_DONE")); break; case MONITOR_SHUTDOWN: LOG_DBG((LOG_MISC, 80, "monitor_loop: MONITOR_SHUTDOWN")); break; default: log_print("monitor_loop: got unknown code %d", msgcode); } } exit(0); } /* Privileged: called by monitor_loop. */ static void m_priv_pfkey_open(void) { int fd, err = 0; fd = pf_key_v2_open(); if (fd < 0) err = -1; must_write(&err, sizeof err); if (fd > 0 && mm_send_fd(m_state.s, fd)) { log_error("m_priv_pfkey_open: read/write operation failed"); close(fd); return; } close(fd); } /* Privileged: called by monitor_loop. */ static void m_priv_getfd(void) { char path[MAXPATHLEN]; size_t len; int v, flags, ret; int err = 0; mode_t mode; must_read(&len, sizeof len); if (len == 0 || len >= sizeof path) log_fatal("m_priv_getfd: invalid pathname length"); must_read(path, len); path[len] = '\0'; if (strlen(path) != len) log_fatal("m_priv_getfd: invalid pathname"); must_read(&flags, sizeof flags); must_read(&mode, sizeof mode); if ((ret = m_priv_local_sanitize_path(path, sizeof path, flags)) != 0) { if (ret == 1) log_print("m_priv_getfd: illegal path \"%s\"", path); err = EACCES; v = -1; } else { if ((v = open(path, flags, mode)) == -1) err = errno; } must_write(&err, sizeof err); if (v != -1) { if (mm_send_fd(m_state.s, v) == -1) log_error("m_priv_getfd: sending fd failed"); close(v); } } /* Privileged: called by monitor_loop. */ static void m_priv_setsockopt(void) { int sock, level, optname, v; int err = 0; char *optval = 0; socklen_t optlen; sock = mm_receive_fd(m_state.s); if (sock < 0) { log_print("m_priv_setsockopt: read/write error"); return; } must_read(&level, sizeof level); must_read(&optname, sizeof optname); must_read(&optlen, sizeof optlen); optval = (char *)malloc(optlen); if (!optval) { log_print("m_priv_setsockopt: malloc failed"); close(sock); return; } must_read(optval, optlen); if (m_priv_check_sockopt(level, optname) != 0) { err = EACCES; v = -1; } else { v = setsockopt(sock, level, optname, optval, optlen); if (v < 0) err = errno; } close(sock); sock = -1; must_write(&err, sizeof err); must_write(&v, sizeof v); free(optval); return; } /* Privileged: called by monitor_loop. */ static void m_priv_bind(void) { int sock, v, err = 0; struct sockaddr *name = 0; socklen_t namelen; sock = mm_receive_fd(m_state.s); if (sock < 0) { log_print("m_priv_bind: read/write error"); return; } must_read(&namelen, sizeof namelen); name = (struct sockaddr *)malloc(namelen); if (!name) { log_print("m_priv_bind: malloc failed"); close(sock); return; } must_read((char *)name, namelen); if (m_priv_check_bind(name, namelen) != 0) { err = EACCES; v = -1; } else { v = bind(sock, name, namelen); if (v < 0) { log_error("m_priv_bind: bind(%d,%p,%d) returned %d", sock, name, namelen, v); err = errno; } } close(sock); sock = -1; must_write(&err, sizeof err); must_write(&v, sizeof v); free(name); return; } /* * Help functions, used by both privileged and unprivileged code */ /* * Read data with the assertion that it all must come through, or else abort * the process. Based on atomicio() from openssh. */ static void must_read(void *buf, size_t n) { char *s = buf; size_t pos = 0; ssize_t res; while (n > pos) { res = read(m_state.s, s + pos, n - pos); switch (res) { case -1: if (errno == EINTR || errno == EAGAIN) continue; case 0: monitor_exit(0); default: pos += res; } } } /* * Write data with the assertion that it all has to be written, or else abort * the process. Based on atomicio() from openssh. */ static void must_write(const void *buf, size_t n) { const char *s = buf; size_t pos = 0; ssize_t res; while (n > pos) { res = write(m_state.s, s + pos, n - pos); switch (res) { case -1: if (errno == EINTR || errno == EAGAIN) continue; case 0: monitor_exit(0); default: pos += res; } } } /* Check that path/mode is permitted. */ static int m_priv_local_sanitize_path(char *path, size_t pmax, int flags) { char new_path[PATH_MAX], var_run[PATH_MAX]; /* * We only permit paths starting with * /etc/isakmpd/ (read only) * /var/run/ (rw) */ if (realpath(path, new_path) == NULL || realpath("/var/run", var_run) == NULL) { /* * We could not decide whether the path is ok or not. * Indicate this be returning 2. */ if (errno == ENOENT) return 2; goto bad_path; } strlcat(var_run, "/", sizeof(var_run)); if (strncmp(var_run, new_path, strlen(var_run)) == 0) return 0; if (strncmp(ISAKMPD_ROOT, new_path, strlen(ISAKMPD_ROOT)) == 0 && (flags & O_ACCMODE) == O_RDONLY) return 0; bad_path: return 1; } /* Check setsockopt */ static int m_priv_check_sockopt(int level, int name) { switch (level) { /* These are allowed */ case SOL_SOCKET: case IPPROTO_IP: case IPPROTO_IPV6: break; default: log_print("m_priv_check_sockopt: Illegal level %d", level); return 1; } switch (name) { /* These are allowed */ case SO_REUSEPORT: case SO_REUSEADDR: case IP_AUTH_LEVEL: case IP_ESP_TRANS_LEVEL: case IP_ESP_NETWORK_LEVEL: case IP_IPCOMP_LEVEL: case IPV6_AUTH_LEVEL: case IPV6_ESP_TRANS_LEVEL: case IPV6_ESP_NETWORK_LEVEL: case IPV6_IPCOMP_LEVEL: break; default: log_print("m_priv_check_sockopt: Illegal option name %d", name); return 1; } return 0; } /* Check bind */ static int m_priv_check_bind(const struct sockaddr *sa, socklen_t salen) { in_port_t port; if (sa == NULL) { log_print("NULL address"); return 1; } if (SA_LEN(sa) != salen) { log_print("Length mismatch: %lu %lu", (unsigned long)sa->sa_len, (unsigned long)salen); return 1; } switch (sa->sa_family) { case AF_INET: if (salen != sizeof(struct sockaddr_in)) { log_print("Invalid inet address length"); return 1; } port = ((const struct sockaddr_in *)sa)->sin_port; break; case AF_INET6: if (salen != sizeof(struct sockaddr_in6)) { log_print("Invalid inet6 address length"); return 1; } port = ((const struct sockaddr_in6 *)sa)->sin6_port; break; default: log_print("Unknown address family"); return 1; } port = ntohs(port); if (port != ISAKMP_PORT_DEFAULT && port < 1024) { log_print("Disallowed port %u", port); return 1; } return 0; } static void m_priv_req_readdir() { size_t len; char path[MAXPATHLEN]; DIR *dp; struct dirent *file; struct stat sb; int off, size, fd, ret, serrno; must_read(&len, sizeof len); if (len == 0 || len >= sizeof path) log_fatal("m_priv_req_readdir: invalid pathname length"); must_read(path, len); path[len] = '\0'; if (strlen(path) != len) log_fatal("m_priv_req_readdir: invalid pathname"); off = strlen(path); size = sizeof path - off; if ((dp = opendir(path)) == NULL) { serrno = errno; ret = -1; must_write(&ret, sizeof ret); must_write(&serrno, sizeof serrno); return; } /* report opendir() success */ ret = 0; must_write(&ret, sizeof ret); while ((file = readdir(dp)) != NULL) { strlcpy(path + off, file->d_name, size); if (m_priv_local_sanitize_path(path, sizeof path, O_RDONLY) != 0) continue; fd = open(path, O_RDONLY, 0); if (fd == -1) { log_error("m_priv_req_readdir: open " "(\"%s\", O_RDONLY, 0) failed", path); continue; } if ((fstat(fd, &sb) == -1) || !(S_ISREG(sb.st_mode) || S_ISLNK(sb.st_mode))) { close(fd); continue; } len = strlen(path); must_write(&len, sizeof len); must_write(path, len); mm_send_fd(m_state.s, fd); close(fd); } closedir(dp); len = 0; must_write(&len, sizeof len); }