X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=a1d77244f8c38edab93f1139cff95a3af6355f1a;hb=e866af3acc30fcd1183a028ea3ef552b7237cc55;hp=0903c57225eb04e8bad82c7ee4a431bbc27789c4;hpb=d96c1ecf7bf9dae6b0cb728e41d09999ecce0b4a;p=elogind.git diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 0903c5722..a1d77244f 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -84,13 +84,20 @@ #include "def.h" #include "rtnl-util.h" #include "udev-util.h" +#include "eventfd-util.h" #include "blkid-util.h" #include "gpt.h" +#include "siphash24.h" #ifdef HAVE_SECCOMP #include "seccomp-util.h" #endif +typedef enum ContainerStatus { + CONTAINER_TERMINATED, + CONTAINER_REBOOTED +} ContainerStatus; + typedef enum LinkJournal { LINK_NO, LINK_AUTO, @@ -656,7 +663,7 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { return r; } } else { - log_error("Failed to bind mount %s: %s", *x, strerror(errno)); + log_error("Failed to bind mount %s: %m", *x); return -errno; } /* Create the mount point, but be conservative -- refuse to create block @@ -768,6 +775,15 @@ static int setup_resolv_conf(const char *dest) { return 0; } +static char* id128_format_as_uuid(sd_id128_t id, char s[37]) { + + snprintf(s, 37, + "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", + SD_ID128_FORMAT_VAL(id)); + + return s; +} + static int setup_boot_id(const char *dest) { _cleanup_free_ char *from = NULL, *to = NULL; sd_id128_t rnd = {}; @@ -793,10 +809,7 @@ static int setup_boot_id(const char *dest) { return r; } - snprintf(as_uuid, sizeof(as_uuid), - "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", - SD_ID128_FORMAT_VAL(rnd)); - char_array_0(as_uuid); + id128_format_as_uuid(rnd, as_uuid); r = write_string_file(from, as_uuid); if (r < 0) { @@ -985,7 +998,7 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { /* Store away the fd in the socket, so that it stays open as * long as we run the child */ k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL); - close_nointr_nofail(fd); + safe_close(fd); if (k < 0) { log_error("Failed to send FIFO fd: %m"); @@ -1136,10 +1149,8 @@ static int setup_journal(const char *directory) { } else if (access(p, F_OK) < 0) return 0; - if (dir_is_empty(q) == 0) { - log_error("%s not empty.", q); - return -ENOTEMPTY; - } + if (dir_is_empty(q) == 0) + log_warning("%s is not empty, proceeding anyway.", q); r = mkdir_p(q, 0755); if (r < 0) { @@ -1258,7 +1269,7 @@ static int register_machine(pid_t pid) { return r; } - r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 8, + r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 10, /* Allow the container to * access and create the API * device nodes, so that @@ -1277,7 +1288,18 @@ static int register_machine(pid_t pid) { * container to ever create * these device nodes. */ "/dev/pts/ptmx", "rw", - "char-pts", "rw"); + "char-pts", "rw", + /* Allow the container + * access to all kdbus + * devices. Again, the + * container cannot create + * these nodes, only use + * them. We use a pretty + * open match here, so that + * the kernel API can still + * change. */ + "char-kdbus", "rw", + "char-kdbus/*", "rw"); if (r < 0) { log_error("Failed to add device whitelist: %s", strerror(-r)); return r; @@ -1388,9 +1410,46 @@ static int reset_audit_loginuid(void) { return 0; } +#define HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2) + +static int get_mac(struct ether_addr *mac) { + int r; + + uint8_t result[8]; + size_t l, sz; + uint8_t *v; + + l = strlen(arg_machine); + sz = sizeof(sd_id128_t) + l; + v = alloca(sz); + + /* fetch some persistent data unique to the host */ + r = sd_id128_get_machine((sd_id128_t*) v); + if (r < 0) + return r; + + /* combine with some data unique (on this host) to this + * container instance */ + memcpy(v + sizeof(sd_id128_t), arg_machine, l); + + /* Let's hash the host machine ID plus the container name. We + * use a fixed, but originally randomly created hash key here. */ + siphash24(result, v, sz, HASH_KEY.bytes); + + assert_cc(ETH_ALEN <= sizeof(result)); + memcpy(mac->ether_addr_octet, result, ETH_ALEN); + + /* see eth_random_addr in the kernel */ + mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */ + mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */ + + return 0; +} + static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + struct ether_addr mac; int r; if (!arg_private_network) @@ -1405,9 +1464,14 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { memcpy(iface_name, "vb-", 3); else memcpy(iface_name, "ve-", 3); - strncpy(iface_name+3, arg_machine, IFNAMSIZ - 3); + r = get_mac(&mac); + if (r < 0) { + log_error("Failed to generate predictable MAC address for host0"); + return r; + } + r = sd_rtnl_open(&rtnl, 0); if (r < 0) { log_error("Failed to connect to netlink: %s", strerror(-r)); @@ -1432,13 +1496,7 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "veth"); - if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); - return r; - } - - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "veth"); if (r < 0) { log_error("Failed to open netlink container: %s", strerror(-r)); return r; @@ -1456,6 +1514,12 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { return r; } + r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac); + if (r < 0) { + log_error("Failed to add netlink MAC address: %s", strerror(-r)); + return r; + } + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); if (r < 0) { log_error("Failed to add netlink namespace field: %s", strerror(-r)); @@ -1521,6 +1585,12 @@ static int setup_bridge(const char veth_name[]) { return r; } + r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP); + if (r < 0) { + log_error("Failed to set IFF_UP flag: %s", strerror(-r)); + return r; + } + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name); if (r < 0) { log_error("Failed to add netlink interface name field: %s", strerror(-r)); @@ -1691,13 +1761,7 @@ static int setup_macvlan(pid_t pid) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "macvlan"); - if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); - return r; - } - - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "macvlan"); if (r < 0) { log_error("Failed to open netlink container: %s", strerror(-r)); return r; @@ -1880,9 +1944,9 @@ static int setup_image(char **device_path, int *loop_nr) { static int dissect_image( int fd, - char **root_device, - char **home_device, - char **srv_device, + char **root_device, bool *root_device_rw, + char **home_device, bool *home_device_rw, + char **srv_device, bool *srv_device_rw, bool *secondary) { #ifdef HAVE_BLKID @@ -1893,6 +1957,7 @@ static int dissect_image( _cleanup_blkid_free_probe_ blkid_probe b = NULL; _cleanup_udev_unref_ struct udev *udev = NULL; struct udev_list_entry *first, *item; + bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true; const char *pttype = NULL; blkid_partlist pl; struct stat st; @@ -1982,6 +2047,7 @@ static int dissect_image( udev_list_entry_foreach(item, first) { _cleanup_udev_device_unref_ struct udev_device *q; const char *stype, *node; + unsigned long long flags; sd_id128_t type_id; blkid_partition pp; dev_t qn; @@ -2012,6 +2078,10 @@ static int dissect_image( if (!pp) continue; + flags = blkid_partition_get_flags(pp); + if (flags & GPT_FLAG_NO_AUTO) + continue; + nr = blkid_partition_get_partno(pp); if (nr < 0) continue; @@ -2029,6 +2099,8 @@ static int dissect_image( continue; home_nr = nr; + home_rw = !(flags & GPT_FLAG_READ_ONLY); + free(home); home = strdup(node); if (!home) @@ -2039,6 +2111,8 @@ static int dissect_image( continue; srv_nr = nr; + srv_rw = !(flags & GPT_FLAG_READ_ONLY); + free(srv); srv = strdup(node); if (!srv) @@ -2051,6 +2125,8 @@ static int dissect_image( continue; root_nr = nr; + root_rw = !(flags & GPT_FLAG_READ_ONLY); + free(root); root = strdup(node); if (!root) @@ -2064,6 +2140,9 @@ static int dissect_image( continue; secondary_root_nr = nr; + secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + + free(secondary_root); secondary_root = strdup(node); if (!secondary_root) @@ -2081,21 +2160,29 @@ static int dissect_image( if (root) { *root_device = root; root = NULL; + + *root_device_rw = root_rw; *secondary = false; } else if (secondary_root) { *root_device = secondary_root; secondary_root = NULL; + + *root_device_rw = secondary_root_rw; *secondary = true; } if (home) { *home_device = home; home = NULL; + + *home_device_rw = home_rw; } if (srv) { *srv_device = srv; srv = NULL; + + *srv_device_rw = srv_rw; } return 0; @@ -2105,7 +2192,7 @@ static int dissect_image( #endif } -static int mount_device(const char *what, const char *where, const char *directory) { +static int mount_device(const char *what, const char *where, const char *directory, bool rw) { #ifdef HAVE_BLKID _cleanup_blkid_free_probe_ blkid_probe b = NULL; const char *fstype, *p; @@ -2114,6 +2201,9 @@ static int mount_device(const char *what, const char *where, const char *directo assert(what); assert(where); + if (arg_read_only) + rw = false; + if (directory) p = strappenda(where, directory); else @@ -2156,7 +2246,7 @@ static int mount_device(const char *what, const char *where, const char *directo return -ENOTSUP; } - if (mount(what, p, fstype, arg_read_only ? MS_NODEV|MS_RDONLY : 0, NULL) < 0) { + if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) { log_error("Failed to mount %s: %m", what); return -errno; } @@ -2168,13 +2258,17 @@ static int mount_device(const char *what, const char *where, const char *directo #endif } -static int mount_devices(const char *where, const char *root_device, const char *home_device, const char *srv_device) { +static int mount_devices( + const char *where, + const char *root_device, bool root_device_rw, + const char *home_device, bool home_device_rw, + const char *srv_device, bool srv_device_rw) { int r; assert(where); if (root_device) { - r = mount_device(root_device, arg_directory, NULL); + r = mount_device(root_device, arg_directory, NULL, root_device_rw); if (r < 0) { log_error("Failed to mount root directory: %s", strerror(-r)); return r; @@ -2182,7 +2276,7 @@ static int mount_devices(const char *where, const char *root_device, const char } if (home_device) { - r = mount_device(home_device, arg_directory, "/home"); + r = mount_device(home_device, arg_directory, "/home", home_device_rw); if (r < 0) { log_error("Failed to mount home directory: %s", strerror(-r)); return r; @@ -2190,7 +2284,7 @@ static int mount_devices(const char *where, const char *root_device, const char } if (srv_device) { - r = mount_device(srv_device, arg_directory, "/srv"); + r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw); if (r < 0) { log_error("Failed to mount server data directory: %s", strerror(-r)); return r; @@ -2208,8 +2302,7 @@ static void loop_remove(int nr, int *image_fd) { if (image_fd && *image_fd >= 0) { ioctl(*image_fd, LOOP_CLR_FD); - close_nointr_nofail(*image_fd); - *image_fd = -1; + *image_fd = safe_close(*image_fd); } control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); @@ -2219,18 +2312,352 @@ static void loop_remove(int nr, int *image_fd) { ioctl(control, LOOP_CTL_REMOVE, nr); } +static int spawn_getent(const char *database, const char *key, pid_t *rpid) { + int pipe_fds[2]; + pid_t pid; + + assert(database); + assert(key); + assert(rpid); + + if (pipe2(pipe_fds, O_CLOEXEC) < 0) { + log_error("Failed to allocate pipe: %m"); + return -errno; + } + + pid = fork(); + if (pid < 0) { + log_error("Failed to fork getent child: %m"); + return -errno; + } else if (pid == 0) { + int nullfd; + char *empty_env = NULL; + + if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0) + _exit(EXIT_FAILURE); + + if (pipe_fds[0] > 2) + safe_close(pipe_fds[0]); + if (pipe_fds[1] > 2) + safe_close(pipe_fds[1]); + + nullfd = open("/dev/null", O_RDWR); + if (nullfd < 0) + _exit(EXIT_FAILURE); + + if (dup3(nullfd, STDIN_FILENO, 0) < 0) + _exit(EXIT_FAILURE); + + if (dup3(nullfd, STDERR_FILENO, 0) < 0) + _exit(EXIT_FAILURE); + + if (nullfd > 2) + safe_close(nullfd); + + reset_all_signal_handlers(); + close_all_fds(NULL, 0); + + execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env); + execle("/bin/getent", "getent", database, key, NULL, &empty_env); + _exit(EXIT_FAILURE); + } + + pipe_fds[1] = safe_close(pipe_fds[1]); + + *rpid = pid; + + return pipe_fds[0]; +} + +static int change_uid_gid(char **_home) { + char line[LINE_MAX], *w, *x, *state, *u, *g, *h; + _cleanup_free_ uid_t *uids = NULL; + _cleanup_free_ char *home = NULL; + _cleanup_fclose_ FILE *f = NULL; + _cleanup_close_ int fd = -1; + unsigned n_uids = 0; + size_t sz = 0, l; + uid_t uid; + gid_t gid; + pid_t pid; + int r; + + assert(_home); + + if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) { + /* Reset everything fully to 0, just in case */ + + if (setgroups(0, NULL) < 0) { + log_error("setgroups() failed: %m"); + return -errno; + } + + if (setresgid(0, 0, 0) < 0) { + log_error("setregid() failed: %m"); + return -errno; + } + + if (setresuid(0, 0, 0) < 0) { + log_error("setreuid() failed: %m"); + return -errno; + } + + *_home = NULL; + return 0; + } + + /* First, get user credentials */ + fd = spawn_getent("passwd", arg_user, &pid); + if (fd < 0) + return fd; + + f = fdopen(fd, "r"); + if (!f) + return log_oom(); + fd = -1; + + if (!fgets(line, sizeof(line), f)) { + + if (!ferror(f)) { + log_error("Failed to resolve user %s.", arg_user); + return -ESRCH; + } + + log_error("Failed to read from getent: %m"); + return -errno; + } + + truncate_nl(line); + + wait_for_terminate_and_warn("getent passwd", pid); + + x = strchr(line, ':'); + if (!x) { + log_error("/etc/passwd entry has invalid user field."); + return -EIO; + } + + u = strchr(x+1, ':'); + if (!u) { + log_error("/etc/passwd entry has invalid password field."); + return -EIO; + } + + u++; + g = strchr(u, ':'); + if (!g) { + log_error("/etc/passwd entry has invalid UID field."); + return -EIO; + } + + *g = 0; + g++; + x = strchr(g, ':'); + if (!x) { + log_error("/etc/passwd entry has invalid GID field."); + return -EIO; + } + + *x = 0; + h = strchr(x+1, ':'); + if (!h) { + log_error("/etc/passwd entry has invalid GECOS field."); + return -EIO; + } + + h++; + x = strchr(h, ':'); + if (!x) { + log_error("/etc/passwd entry has invalid home directory field."); + return -EIO; + } + + *x = 0; + + r = parse_uid(u, &uid); + if (r < 0) { + log_error("Failed to parse UID of user."); + return -EIO; + } + + r = parse_gid(g, &gid); + if (r < 0) { + log_error("Failed to parse GID of user."); + return -EIO; + } + + home = strdup(h); + if (!home) + return log_oom(); + + /* Second, get group memberships */ + fd = spawn_getent("initgroups", arg_user, &pid); + if (fd < 0) + return fd; + + fclose(f); + f = fdopen(fd, "r"); + if (!f) + return log_oom(); + fd = -1; + + if (!fgets(line, sizeof(line), f)) { + if (!ferror(f)) { + log_error("Failed to resolve user %s.", arg_user); + return -ESRCH; + } + + log_error("Failed to read from getent: %m"); + return -errno; + } + + truncate_nl(line); + + wait_for_terminate_and_warn("getent initgroups", pid); + + /* Skip over the username and subsequent separator whitespace */ + x = line; + x += strcspn(x, WHITESPACE); + x += strspn(x, WHITESPACE); + + FOREACH_WORD(w, l, x, state) { + char c[l+1]; + + memcpy(c, w, l); + c[l] = 0; + + if (!GREEDY_REALLOC(uids, sz, n_uids+1)) + return log_oom(); + + r = parse_uid(c, &uids[n_uids++]); + if (r < 0) { + log_error("Failed to parse group data from getent."); + return -EIO; + } + } + + r = mkdir_parents(home, 0775); + if (r < 0) { + log_error("Failed to make home root directory: %s", strerror(-r)); + return r; + } + + r = mkdir_safe(home, 0755, uid, gid); + if (r < 0 && r != -EEXIST) { + log_error("Failed to make home directory: %s", strerror(-r)); + return r; + } + + fchown(STDIN_FILENO, uid, gid); + fchown(STDOUT_FILENO, uid, gid); + fchown(STDERR_FILENO, uid, gid); + + if (setgroups(n_uids, uids) < 0) { + log_error("Failed to set auxiliary groups: %m"); + return -errno; + } + + if (setresgid(gid, gid, gid) < 0) { + log_error("setregid() failed: %m"); + return -errno; + } + + if (setresuid(uid, uid, uid) < 0) { + log_error("setreuid() failed: %m"); + return -errno; + } + + if (_home) { + *_home = home; + home = NULL; + } + + return 0; +} + +/* + * Return 0 in case the container is being rebooted, has been shut + * down or exited successfully. On failures a negative value is + * returned. + * + * The status of the container "CONTAINER_TERMINATED" or + * "CONTAINER_REBOOTED" will be saved in the container argument + */ +static int wait_for_container(pid_t pid, ContainerStatus *container) { + int r; + siginfo_t status; + + r = wait_for_terminate(pid, &status); + if (r < 0) + return r; + + switch (status.si_code) { + case CLD_EXITED: + r = status.si_status; + if (r == 0) { + if (!arg_quiet) + log_debug("Container %s exited successfully.", + arg_machine); + + *container = CONTAINER_TERMINATED; + } else { + log_error("Container %s failed with error code %i.", + arg_machine, status.si_status); + r = -1; + } + break; + + case CLD_KILLED: + if (status.si_status == SIGINT) { + if (!arg_quiet) + log_info("Container %s has been shut down.", + arg_machine); + + *container = CONTAINER_TERMINATED; + r = 0; + break; + } else if (status.si_status == SIGHUP) { + if (!arg_quiet) + log_info("Container %s is being rebooted.", + arg_machine); + + *container = CONTAINER_REBOOTED; + r = 0; + break; + } + /* CLD_KILLED fallthrough */ + + case CLD_DUMPED: + log_error("Container %s terminated by signal %s.", + arg_machine, signal_to_string(status.si_status)); + r = -1; + break; + + default: + log_error("Container %s failed due to unknown reason.", + arg_machine); + r = -1; + break; + } + + return r; +} + +static void nop_handler(int sig) {} + int main(int argc, char *argv[]) { _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL; + bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; _cleanup_close_ int master = -1, kdbus_fd = -1, image_fd = -1; - _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 }; + _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }; _cleanup_fdset_free_ FDSet *fds = NULL; int r = EXIT_FAILURE, k, n_fd_passed, loop_nr = -1; const char *console = NULL; char veth_name[IFNAMSIZ]; bool secondary = false; + sigset_t mask, mask_chld; pid_t pid = 0; - sigset_t mask; log_parse_environment(); log_open(); @@ -2339,7 +2766,7 @@ int main(int argc, char *argv[]) { goto finish; } - r = dissect_image(image_fd, &root_device, &home_device, &srv_device, &secondary); + r = dissect_image(image_fd, &root_device, &root_device_rw, &home_device, &home_device_rw, &srv_device, &srv_device_rw, &secondary); if (r < 0) goto finish; } @@ -2392,44 +2819,50 @@ int main(int argc, char *argv[]) { sd_notify(0, "READY=1"); assert_se(sigemptyset(&mask) == 0); + assert_se(sigemptyset(&mask_chld) == 0); + sigaddset(&mask_chld, SIGCHLD); sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); for (;;) { - int parent_ready_fd = -1, child_ready_fd = -1; - siginfo_t status; - eventfd_t x; - - parent_ready_fd = eventfd(0, EFD_CLOEXEC); - if (parent_ready_fd < 0) { - log_error("Failed to create event fd: %m"); + ContainerStatus container_status; + int eventfds[2] = { -1, -1 }; + struct sigaction sa = { + .sa_handler = nop_handler, + .sa_flags = SA_NOCLDSTOP, + }; + + /* Child can be killed before execv(), so handle SIGCHLD + * in order to interrupt parent's blocking calls and + * give it a chance to call wait() and terminate. */ + r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL); + if (r < 0) { + log_error("Failed to change the signal mask: %m"); goto finish; } - child_ready_fd = eventfd(0, EFD_CLOEXEC); - if (child_ready_fd < 0) { - log_error("Failed to create event fd: %m"); + r = sigaction(SIGCHLD, &sa, NULL); + if (r < 0) { + log_error("Failed to install SIGCHLD handler: %m"); goto finish; } - pid = syscall(__NR_clone, - SIGCHLD|CLONE_NEWNS| - (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| - (arg_private_network ? CLONE_NEWNET : 0), NULL); + pid = clone_with_eventfd(SIGCHLD|CLONE_NEWNS| + (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| + (arg_private_network ? CLONE_NEWNET : 0), eventfds); if (pid < 0) { if (errno == EINVAL) log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m"); else log_error("clone() failed: %m"); + r = pid; goto finish; } if (pid == 0) { /* child */ - const char *home = NULL; - uid_t uid = (uid_t) -1; - gid_t gid = (gid_t) -1; + _cleanup_free_ char *home = NULL; unsigned n_env = 2; const char *envp[] = { "PATH=" DEFAULT_PATH_SPLIT_USR, @@ -2449,15 +2882,13 @@ int main(int argc, char *argv[]) { if (envp[n_env]) n_env ++; - close_nointr_nofail(master); - master = -1; + master = safe_close(master); close_nointr(STDIN_FILENO); close_nointr(STDOUT_FILENO); close_nointr(STDERR_FILENO); - close_nointr_nofail(kmsg_socket_pair[0]); - kmsg_socket_pair[0] = -1; + kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]); reset_all_signal_handlers(); @@ -2467,7 +2898,7 @@ int main(int argc, char *argv[]) { k = open_terminal(console, O_RDWR); if (k != STDIN_FILENO) { if (k >= 0) { - close_nointr_nofail(k); + safe_close(k); k = -EINVAL; } @@ -2502,7 +2933,10 @@ int main(int argc, char *argv[]) { goto child_fail; } - if (mount_devices(arg_directory, root_device, home_device, srv_device) < 0) + if (mount_devices(arg_directory, + root_device, root_device_rw, + home_device, home_device_rw, + srv_device, srv_device_rw) < 0) goto child_fail; /* Turn directory into bind mount */ @@ -2537,8 +2971,7 @@ int main(int argc, char *argv[]) { if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) goto child_fail; - close_nointr_nofail(kmsg_socket_pair[1]); - kmsg_socket_pair[1] = -1; + kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); if (setup_boot_id(arg_directory) < 0) goto child_fail; @@ -2564,9 +2997,11 @@ int main(int argc, char *argv[]) { /* Tell the parent that we are ready, and that * it can cgroupify us to that we lack access * to certain devices and resources. */ - eventfd_write(child_ready_fd, 1); - close_nointr_nofail(child_ready_fd); - child_ready_fd = -1; + r = eventfd_send_state(eventfds[1], + EVENTFD_CHILD_SUCCEEDED); + eventfds[1] = safe_close(eventfds[1]); + if (r < 0) + goto child_fail; if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); @@ -2598,61 +3033,9 @@ int main(int argc, char *argv[]) { goto child_fail; } - if (arg_user) { - - /* Note that this resolves user names - * inside the container, and hence - * accesses the NSS modules from the - * container and not the host. This is - * a bit weird... */ - - if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) { - log_error("get_user_creds() failed: %m"); - goto child_fail; - } - - if (mkdir_parents_label(home, 0775) < 0) { - log_error("mkdir_parents_label() failed: %m"); - goto child_fail; - } - - if (mkdir_safe_label(home, 0775, uid, gid) < 0) { - log_error("mkdir_safe_label() failed: %m"); - goto child_fail; - } - - if (initgroups((const char*)arg_user, gid) < 0) { - log_error("initgroups() failed: %m"); - goto child_fail; - } - - if (setresgid(gid, gid, gid) < 0) { - log_error("setregid() failed: %m"); - goto child_fail; - } - - if (setresuid(uid, uid, uid) < 0) { - log_error("setreuid() failed: %m"); - goto child_fail; - } - } else { - /* Reset everything fully to 0, just in case */ - - if (setgroups(0, NULL) < 0) { - log_error("setgroups() failed: %m"); - goto child_fail; - } - - if (setresgid(0, 0, 0) < 0) { - log_error("setregid() failed: %m"); - goto child_fail; - } - - if (setresuid(0, 0, 0) < 0) { - log_error("setreuid() failed: %m"); - goto child_fail; - } - } + r = change_uid_gid(&home); + if (r < 0) + goto child_fail; if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) || (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) || @@ -2662,7 +3045,9 @@ int main(int argc, char *argv[]) { } if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) { - if (asprintf((char**)(envp + n_env++), "container_uuid=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid)) < 0) { + char as_uuid[37]; + + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) { log_oom(); goto child_fail; } @@ -2698,8 +3083,10 @@ int main(int argc, char *argv[]) { #ifdef HAVE_SELINUX if (arg_selinux_context) - if (setexeccon((security_context_t) arg_selinux_context) < 0) + if (setexeccon((security_context_t) arg_selinux_context) < 0) { log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context); + goto child_fail; + } #endif if (!strv_isempty(arg_setenv)) { @@ -2716,9 +3103,10 @@ int main(int argc, char *argv[]) { env_use = (char**) envp; /* Wait until the parent is ready with the setup, too... */ - eventfd_read(parent_ready_fd, &x); - close_nointr_nofail(parent_ready_fd); - parent_ready_fd = -1; + r = eventfd_parent_succeeded(eventfds[0]); + eventfds[0] = safe_close(eventfds[0]); + if (r < 0) + goto child_fail; if (arg_boot) { char **a; @@ -2749,17 +3137,27 @@ int main(int argc, char *argv[]) { log_error("execv() failed: %m"); child_fail: + /* Tell the parent that the setup failed, so he + * can clean up resources and terminate. */ + if (eventfds[1] != -1) + eventfd_send_state(eventfds[1], + EVENTFD_CHILD_FAILED); _exit(EXIT_FAILURE); } fdset_free(fds); fds = NULL; - /* Wait until the child reported that it is ready with - * all it needs to do with priviliges. After we got - * the notification we can make the process join its - * cgroup which might limit what it can do */ - eventfd_read(child_ready_fd, &x); + /* Wait for the child event: + * If EVENTFD_CHILD_FAILED, the child will terminate soon. + * If EVENTFD_CHILD_SUCCEEDED, the child is reporting that + * it is ready with all it needs to do with priviliges. + * After we got the notification we can make the process + * join its cgroup which might limit what it can do */ + r = eventfd_child_succeeded(eventfds[1]); + eventfds[1] = safe_close(eventfds[1]); + if (r < 0) + goto check_container_status; r = register_machine(pid); if (r < 0) @@ -2781,10 +3179,25 @@ int main(int argc, char *argv[]) { if (r < 0) goto finish; + /* Block SIGCHLD here, before notifying child. + * process_pty() will handle it with the other signals. */ + r = sigprocmask(SIG_BLOCK, &mask_chld, NULL); + if (r < 0) + goto finish; + + /* Reset signal to default */ + r = default_signals(SIGCHLD, -1); + if (r < 0) + goto finish; + /* Notify the child that the parent is ready with all - * its setup, and thtat the child can now hand over + * its setup, and that the child can now hand over * control to the code to run inside the container. */ - eventfd_write(parent_ready_fd, 1); + r = eventfd_send_state(eventfds[0], + EVENTFD_PARENT_SUCCEEDED); + eventfds[0] = safe_close(eventfds[0]); + if (r < 0) + goto finish; k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); if (k < 0) { @@ -2798,51 +3211,20 @@ int main(int argc, char *argv[]) { /* Kill if it is not dead yet anyway */ terminate_machine(pid); +check_container_status: /* Redundant, but better safe than sorry */ kill(pid, SIGKILL); - k = wait_for_terminate(pid, &status); + r = wait_for_container(pid, &container_status); pid = 0; - if (k < 0) { + if (r < 0) { r = EXIT_FAILURE; break; - } - - if (status.si_code == CLD_EXITED) { - r = status.si_status; - if (status.si_status != 0) { - log_error("Container %s failed with error code %i.", arg_machine, status.si_status); - break; - } - - if (!arg_quiet) - log_debug("Container %s exited successfully.", arg_machine); - break; - } else if (status.si_code == CLD_KILLED && - status.si_status == SIGINT) { - - if (!arg_quiet) - log_info("Container %s has been shut down.", arg_machine); - r = 0; + } else if (container_status == CONTAINER_TERMINATED) break; - } else if (status.si_code == CLD_KILLED && - status.si_status == SIGHUP) { - if (!arg_quiet) - log_info("Container %s is being rebooted.", arg_machine); - continue; - } else if (status.si_code == CLD_KILLED || - status.si_code == CLD_DUMPED) { - - log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status)); - r = EXIT_FAILURE; - break; - } else { - log_error("Container %s failed due to unknown reason.", arg_machine); - r = EXIT_FAILURE; - break; - } + /* CONTAINER_REBOOTED, loop again */ } finish: