X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=ab7a239ed59f7d7df91425a675e4a2feaa195722;hp=685b4d4e50d4b5c588ebd8374d028b2cb83b046b;hb=89154bd4ac71157e382bacdc4dc3a96990940d36;hpb=dce818b390a857a11f7dd634684500675cf79833 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 685b4d4e5..ab7a239ed 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -38,32 +38,79 @@ #include #include #include +#include +#include #include #include "log.h" #include "util.h" #include "mkdir.h" +#include "macro.h" #include "audit.h" #include "missing.h" #include "cgroup-util.h" #include "strv.h" +#include "path-util.h" #include "loopback-setup.h" +#include "sd-id128.h" +#include "dev-setup.h" + +typedef enum LinkJournal { + LINK_NO, + LINK_AUTO, + LINK_HOST, + LINK_GUEST +} LinkJournal; static char *arg_directory = NULL; static char *arg_user = NULL; static char **arg_controllers = NULL; +static char *arg_uuid = NULL; static bool arg_private_network = false; +static bool arg_read_only = false; +static bool arg_boot = false; +static LinkJournal arg_link_journal = LINK_AUTO; +static uint64_t arg_retain = + (1ULL << CAP_CHOWN) | + (1ULL << CAP_DAC_OVERRIDE) | + (1ULL << CAP_DAC_READ_SEARCH) | + (1ULL << CAP_FOWNER) | + (1ULL << CAP_FSETID) | + (1ULL << CAP_IPC_OWNER) | + (1ULL << CAP_KILL) | + (1ULL << CAP_LEASE) | + (1ULL << CAP_LINUX_IMMUTABLE) | + (1ULL << CAP_NET_BIND_SERVICE) | + (1ULL << CAP_NET_BROADCAST) | + (1ULL << CAP_NET_RAW) | + (1ULL << CAP_SETGID) | + (1ULL << CAP_SETFCAP) | + (1ULL << CAP_SETPCAP) | + (1ULL << CAP_SETUID) | + (1ULL << CAP_SYS_ADMIN) | + (1ULL << CAP_SYS_CHROOT) | + (1ULL << CAP_SYS_NICE) | + (1ULL << CAP_SYS_PTRACE) | + (1ULL << CAP_SYS_TTY_CONFIG) | + (1ULL << CAP_SYS_RESOURCE) | + (1ULL << CAP_SYS_BOOT); static int help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" "Spawn a minimal namespace container for debugging, testing and building.\n\n" - " -h --help Show this help\n" - " -D --directory=NAME Root directory for the container\n" - " -u --user=USER Run the command under specified user or uid\n" - " -C --controllers=LIST Put the container in specified comma-separated cgroup hierarchies\n" - " --private-network Disable network in container\n", + " -h --help Show this help\n" + " -D --directory=NAME Root directory for the container\n" + " -b --boot Boot up full system (i.e. invoke init)\n" + " -u --user=USER Run the command under specified user or uid\n" + " -C --controllers=LIST Put the container in specified comma-separated cgroup hierarchies\n" + " --uuid=UUID Set a specific machine UUID for the container\n" + " --private-network Disable network in container\n" + " --read-only Mount the root directory read-only\n" + " --capability=CAP In addition to the default, retain specified capability\n" + " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" + " -j Equivalent to --link-journal=host\n", program_invocation_short_name); return 0; @@ -72,7 +119,11 @@ static int help(void) { static int parse_argv(int argc, char *argv[]) { enum { - ARG_PRIVATE_NETWORK = 0x100 + ARG_PRIVATE_NETWORK = 0x100, + ARG_UUID, + ARG_READ_ONLY, + ARG_CAPABILITY, + ARG_LINK_JOURNAL }; static const struct option options[] = { @@ -81,6 +132,11 @@ static int parse_argv(int argc, char *argv[]) { { "user", required_argument, NULL, 'u' }, { "controllers", required_argument, NULL, 'C' }, { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, + { "boot", no_argument, NULL, 'b' }, + { "uuid", required_argument, NULL, ARG_UUID }, + { "read-only", no_argument, NULL, ARG_READ_ONLY }, + { "capability", required_argument, NULL, ARG_CAPABILITY }, + { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, { NULL, 0, NULL, 0 } }; @@ -89,7 +145,7 @@ static int parse_argv(int argc, char *argv[]) { assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:u:C:", options, NULL)) >= 0) { + while ((c = getopt_long(argc, argv, "+hD:u:C:bj", options, NULL)) >= 0) { switch (c) { @@ -99,8 +155,9 @@ static int parse_argv(int argc, char *argv[]) { case 'D': free(arg_directory); - if (!(arg_directory = strdup(optarg))) { - log_error("Failed to duplicate root directory."); + arg_directory = canonicalize_file_name(optarg); + if (!arg_directory) { + log_error("Failed to canonicalize root directory."); return -ENOMEM; } @@ -130,6 +187,63 @@ static int parse_argv(int argc, char *argv[]) { arg_private_network = true; break; + case 'b': + arg_boot = true; + break; + + case ARG_UUID: + arg_uuid = optarg; + break; + + case ARG_READ_ONLY: + arg_read_only = true; + break; + + case ARG_CAPABILITY: { + char *state, *word; + size_t length; + + FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) { + cap_value_t cap; + char *t; + + t = strndup(word, length); + if (!t) + return log_oom(); + + if (cap_from_name(t, &cap) < 0) { + log_error("Failed to parse capability %s.", t); + free(t); + return -EINVAL; + } + + free(t); + arg_retain |= 1ULL << (uint64_t) cap; + } + + break; + } + + case 'j': + arg_link_journal = LINK_GUEST; + break; + + case ARG_LINK_JOURNAL: + if (streq(optarg, "auto")) + arg_link_journal = LINK_AUTO; + else if (streq(optarg, "no")) + arg_link_journal = LINK_NO; + else if (streq(optarg, "guest")) + arg_link_journal = LINK_GUEST; + else if (streq(optarg, "host")) + arg_link_journal = LINK_HOST; + else { + log_error("Failed to parse link journal mode %s", optarg); + return -EINVAL; + } + + break; + case '?': return -EINVAL; @@ -155,28 +269,28 @@ static int mount_all(const char *dest) { static const MountPoint mount_table[] = { { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, - { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */ - { "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ - { "/sys", "/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */ - { "/sys", "/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ + { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */ + { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ + { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true }, - { "/dev/pts", "/dev/pts", "bind", NULL, MS_BIND, true }, + { "/dev/pts", "/dev/pts", NULL, NULL, MS_BIND, true }, + { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, #ifdef HAVE_SELINUX - { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND, false }, /* Bind mount first */ - { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ + { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */ + { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ #endif }; unsigned k; int r = 0; - char *where; for (k = 0; k < ELEMENTSOF(mount_table); k++) { + char _cleanup_free_ *where = NULL; int t; if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) { - log_error("Out of memory"); + log_oom(); if (r == 0) r = -ENOMEM; @@ -184,10 +298,9 @@ static int mount_all(const char *dest) { break; } - t = path_is_mount_point(where, false); + t = path_is_mount_point(where, true); if (t < 0) { log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t)); - free(where); if (r == 0) r = t; @@ -195,7 +308,11 @@ static int mount_all(const char *dest) { continue; } - mkdir_p(where, 0755); + /* Skip this entry if it is not a remount. */ + if (mount_table[k].what && t > 0) + continue; + + mkdir_p_label(where, 0755); if (mount(mount_table[k].what, where, @@ -209,36 +326,103 @@ static int mount_all(const char *dest) { if (r == 0) r = -errno; } - - free(where); } + return r; +} + +static int setup_timezone(const char *dest) { + char *where; + + assert(dest); + /* Fix the timezone, if possible */ - if (asprintf(&where, "%s/etc/localtime", dest) >= 0) { + where = strappend(dest, "/etc/localtime"); + if (!where) + return log_oom(); - if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0) - mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0) + mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); - free(where); - } + free(where); + + where = strappend(dest, "/etc/timezone"); + if (!where) + return log_oom(); + + if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0) + mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + + free(where); + + return 0; +} + +static int setup_resolv_conf(const char *dest) { + char *where; + + assert(dest); + + if (arg_private_network) + return 0; + + /* Fix resolv.conf, if possible */ + where = strappend(dest, "/etc/resolv.conf"); + if (!where) + return log_oom(); + + if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0) + mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + + free(where); + + return 0; +} + +static int setup_boot_id(const char *dest) { + char _cleanup_free_ *from = NULL, *to = NULL; + sd_id128_t rnd; + char as_uuid[37]; + int r; + + assert(dest); - if (asprintf(&where, "%s/etc/timezone", dest) >= 0) { + /* Generate a new randomized boot ID, so that each boot-up of + * the container gets a new one */ - if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0) - mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id"); + to = strappend(dest, "/proc/sys/kernel/random/boot_id"); + if (!from || !to) + return log_oom(); - free(where); + r = sd_id128_randomize(&rnd); + if (r < 0) { + log_error("Failed to generate random boot id: %s", strerror(-r)); + return r; } - if (asprintf(&where, "%s/proc/kmsg", dest) >= 0) { - mount("/dev/null", where, "bind", MS_BIND, NULL); - free(where); + 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); + + r = write_one_line_file(from, as_uuid); + if (r < 0) { + log_error("Failed to write boot id: %s", strerror(-r)); + return r; } + if (mount(from, to, "bind", MS_BIND, NULL) < 0) { + log_error("Failed to bind mount boot id: %m"); + r = -errno; + } else + mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + + unlink(from); return r; } -static int copy_devnodes(const char *dest, const char *console) { +static int copy_devnodes(const char *dest) { static const char devnodes[] = "null\0" @@ -247,34 +431,25 @@ static int copy_devnodes(const char *dest, const char *console) { "random\0" "urandom\0" "tty\0" - "ptmx\0" - "kmsg\0" - "rtc0\0"; + "ptmx\0"; const char *d; - int r = 0, k; - mode_t u; - struct stat st; - char *from = NULL, *to = NULL; + int r = 0; + mode_t _cleanup_umask_ u; assert(dest); - assert(console); u = umask(0000); NULSTR_FOREACH(d, devnodes) { - from = to = NULL; + struct stat st; + char _cleanup_free_ *from = NULL, *to = NULL; asprintf(&from, "/dev/%s", d); asprintf(&to, "%s/dev/%s", dest, d); if (!from || !to) { - log_error("Failed to allocate devnode path"); - - free(from); - free(to); - - from = to = NULL; + log_oom(); if (r == 0) r = -ENOMEM; @@ -292,7 +467,7 @@ static int copy_devnodes(const char *dest, const char *console) { } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) { - log_error("%s is not a char or block device, cannot copy.", from); + log_error("%s is not a char or block device, cannot copy", from); if (r == 0) r = -EIO; @@ -302,37 +477,40 @@ static int copy_devnodes(const char *dest, const char *console) { if (r == 0) r = -errno; } - - free(from); - free(to); } - if (stat(console, &st) < 0) { + return r; +} - log_error("Failed to stat %s: %m", console); - if (r == 0) - r = -errno; +static int setup_dev_console(const char *dest, const char *console) { + struct stat st; + char _cleanup_free_ *to = NULL; + int r; + mode_t _cleanup_umask_ u; - goto finish; + assert(dest); + assert(console); - } else if (!S_ISCHR(st.st_mode)) { + u = umask(0000); - log_error("/dev/console is not a char device."); - if (r == 0) - r = -EIO; + if (stat(console, &st) < 0) { + log_error("Failed to stat %s: %m", console); + return -errno; - goto finish; + } else if (!S_ISCHR(st.st_mode)) { + log_error("/dev/console is not a char device"); + return -EIO; } - if (asprintf(&to, "%s/dev/console", dest) < 0) { - - log_error("Out of memory"); - if (r == 0) - r = -ENOMEM; - - goto finish; + r = chmod_and_chown(console, 0600, 0, 0); + if (r < 0) { + log_error("Failed to correct access mode for TTY: %s", strerror(-r)); + return r; } + if (asprintf(&to, "%s/dev/console", dest) < 0) + return log_oom(); + /* We need to bind mount the right tty to /dev/console since * ptys can only exist on pts file systems. To have something * to bind mount things on we create a device node first, that @@ -340,75 +518,268 @@ static int copy_devnodes(const char *dest, const char *console) { * doesn't actually matter here, since we mount it over * anyway). */ - if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) - log_error("mknod for /dev/console failed: %m"); + if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) { + log_error("mknod() for /dev/console failed: %m"); + return -errno; + } if (mount(console, to, "bind", MS_BIND, NULL) < 0) { - log_error("bind mount for /dev/console failed: %m"); + log_error("Bind mount for /dev/console failed: %m"); + return -errno; + } - if (r == 0) - r = -errno; + return 0; +} + +static int setup_kmsg(const char *dest, int kmsg_socket) { + char _cleanup_free_ *from = NULL, *to = NULL; + int r, fd, k; + mode_t _cleanup_umask_ u; + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(int))]; + } control; + struct msghdr mh; + struct cmsghdr *cmsg; + + assert(dest); + assert(kmsg_socket >= 0); + + u = umask(0000); + + /* We create the kmsg FIFO as /dev/kmsg, but immediately + * delete it after bind mounting it to /proc/kmsg. While FIFOs + * on the reading side behave very similar to /proc/kmsg, + * their writing side behaves differently from /dev/kmsg in + * that writing blocks when nothing is reading. In order to + * avoid any problems with containers deadlocking due to this + * we simply make /dev/kmsg unavailable to the container. */ + if (asprintf(&from, "%s/dev/kmsg", dest) < 0 || + asprintf(&to, "%s/proc/kmsg", dest) < 0) + return log_oom(); + + if (mkfifo(from, 0600) < 0) { + log_error("mkfifo() for /dev/kmsg failed: %m"); + return -errno; } - free(to); + r = chmod_and_chown(from, 0600, 0, 0); + if (r < 0) { + log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r)); + return r; + } - if ((k = chmod_and_chown(console, 0600, 0, 0)) < 0) { - log_error("Failed to correct access mode for TTY: %s", strerror(-k)); + if (mount(from, to, "bind", MS_BIND, NULL) < 0) { + log_error("Bind mount for /proc/kmsg failed: %m"); + return -errno; + } - if (r == 0) - r = k; + fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC); + if (fd < 0) { + log_error("Failed to open fifo: %m"); + return -errno; } -finish: - umask(u); + zero(mh); + zero(control); + + mh.msg_control = &control; + mh.msg_controllen = sizeof(control); + + cmsg = CMSG_FIRSTHDR(&mh); + cmsg->cmsg_level = SOL_SOCKET; + cmsg->cmsg_type = SCM_RIGHTS; + cmsg->cmsg_len = CMSG_LEN(sizeof(int)); + memcpy(CMSG_DATA(cmsg), &fd, sizeof(int)); + + mh.msg_controllen = cmsg->cmsg_len; + + /* 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); + + if (k < 0) { + log_error("Failed to send FIFO fd: %m"); + return -errno; + } + + /* And now make the FIFO unavailable as /dev/kmsg... */ + unlink(from); + return 0; +} + +static int setup_hostname(void) { + char *hn; + int r = 0; + + hn = path_get_file_name(arg_directory); + if (hn) { + hn = strdup(hn); + if (!hn) + return -ENOMEM; + + hostname_cleanup(hn); + + if (!isempty(hn)) + if (sethostname(hn, strlen(hn)) < 0) + r = -errno; + + free(hn); + } return r; } -static int drop_capabilities(void) { - static const unsigned long retain[] = { - CAP_CHOWN, - CAP_DAC_OVERRIDE, - CAP_DAC_READ_SEARCH, - CAP_FOWNER, - CAP_FSETID, - CAP_IPC_OWNER, - CAP_KILL, - CAP_LEASE, - CAP_LINUX_IMMUTABLE, - CAP_NET_BIND_SERVICE, - CAP_NET_BROADCAST, - CAP_NET_RAW, - CAP_SETGID, - CAP_SETFCAP, - CAP_SETPCAP, - CAP_SETUID, - CAP_SYS_ADMIN, - CAP_SYS_CHROOT, - CAP_SYS_NICE, - CAP_SYS_PTRACE, - CAP_SYS_TTY_CONFIG - }; +static int setup_journal(const char *directory) { + sd_id128_t machine_id; + char *p = NULL, *b = NULL, *l, *q = NULL, *d = NULL; + int r; - unsigned long l; + if (arg_link_journal == LINK_NO) + return 0; - for (l = 0; l <= cap_last_cap(); l++) { - unsigned i; + p = strappend(directory, "/etc/machine-id"); + if (!p) { + r = log_oom(); + goto finish; + } - for (i = 0; i < ELEMENTSOF(retain); i++) - if (retain[i] == l) - break; + r = read_one_line_file(p, &b); + if (r == -ENOENT && arg_link_journal == LINK_AUTO) { + r = 0; + goto finish; + } else if (r < 0) { + log_error("Failed to read machine ID: %s", strerror(-r)); + return r; + } - if (i < ELEMENTSOF(retain)) - continue; + l = strstrip(b); + if (isempty(l) && arg_link_journal == LINK_AUTO) { + r = 0; + goto finish; + } - if (prctl(PR_CAPBSET_DROP, l) < 0) { - log_error("PR_CAPBSET_DROP failed: %m"); - return -errno; + /* Verify validaty */ + r = sd_id128_from_string(l, &machine_id); + if (r < 0) { + log_error("Failed to parse machine ID: %s", strerror(-r)); + goto finish; + } + + free(p); + p = strappend("/var/log/journal/", l); + q = strjoin(directory, "/var/log/journal/", l, NULL); + if (!p || !q) { + r = log_oom(); + goto finish; + } + + if (path_is_mount_point(p, false) > 0 || + path_is_mount_point(q, false) > 0) { + if (arg_link_journal != LINK_AUTO) { + log_error("Journal already a mount point, refusing."); + r = -EEXIST; + goto finish; } + + r = 0; + goto finish; } - return 0; + r = readlink_and_make_absolute(p, &d); + if (r >= 0) { + if ((arg_link_journal == LINK_GUEST || + arg_link_journal == LINK_AUTO) && + path_equal(d, q)) { + + mkdir_p(q, 0755); + + r = 0; + goto finish; + } + + if (unlink(p) < 0) { + log_error("Failed to remove symlink %s: %m", p); + r = -errno; + goto finish; + } + } else if (r == -EINVAL) { + + if (arg_link_journal == LINK_GUEST && + rmdir(p) < 0) { + + if (errno == ENOTDIR) + log_error("%s already exists and is neither symlink nor directory.", p); + else { + log_error("Failed to remove %s: %m", p); + r = -errno; + } + + goto finish; + } + } else if (r != -ENOENT) { + log_error("readlink(%s) failed: %m", p); + goto finish; + } + + if (arg_link_journal == LINK_GUEST) { + + if (symlink(q, p) < 0) { + log_error("Failed to symlink %s to %s: %m", q, p); + r = -errno; + goto finish; + } + + mkdir_p(q, 0755); + + r = 0; + goto finish; + } + + if (arg_link_journal == LINK_HOST) { + r = mkdir_p(p, 0755); + if (r < 0) { + log_error("Failed to create %s: %m", p); + goto finish; + } + + } else if (access(p, F_OK) < 0) { + r = 0; + goto finish; + } + + if (dir_is_empty(q) == 0) { + log_error("%s not empty.", q); + r = -ENOTEMPTY; + goto finish; + } + + r = mkdir_p(q, 0755); + if (r < 0) { + log_error("Failed to create %s: %m", q); + goto finish; + } + + if (mount(p, q, "bind", MS_BIND, NULL) < 0) { + log_error("Failed to bind mount journal from host into guest: %m"); + r = -errno; + goto finish; + } + + r = 0; + +finish: + free(p); + free(q); + free(d); + free(b); + return r; + +} + +static int drop_capabilities(void) { + return capability_bounding_set_drop(~arg_retain, false); } static int is_os_tree(const char *path) { @@ -437,13 +808,15 @@ static int process_pty(int master, sigset_t *mask) { fd_nonblock(STDOUT_FILENO, 1); fd_nonblock(master, 1); - if ((signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC)) < 0) { + signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC); + if (signal_fd < 0) { log_error("signalfd(): %m"); r = -errno; goto finish; } - if ((ep = epoll_create1(EPOLL_CLOEXEC)) < 0) { + ep = epoll_create1(EPOLL_CLOEXEC); + if (ep < 0) { log_error("Failed to create epoll: %m"); r = -errno; goto finish; @@ -479,7 +852,8 @@ static int process_pty(int master, sigset_t *mask) { ssize_t k; int i, nfds; - if ((nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1)) < 0) { + nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1); + if (nfds < 0) { if (errno == EINTR || errno == EAGAIN) continue; @@ -514,7 +888,8 @@ static int process_pty(int master, sigset_t *mask) { struct signalfd_siginfo sfsi; ssize_t n; - if ((n = read(signal_fd, &sfsi, sizeof(sfsi))) != sizeof(sfsi)) { + n = read(signal_fd, &sfsi, sizeof(sfsi)); + if (n != sizeof(sfsi)) { if (n >= 0) { log_error("Failed to read from signalfd: invalid block size"); @@ -550,7 +925,8 @@ static int process_pty(int master, sigset_t *mask) { if (stdin_readable && in_buffer_full < LINE_MAX) { - if ((k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full)) < 0) { + k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full); + if (k < 0) { if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) stdin_readable = false; @@ -565,7 +941,8 @@ static int process_pty(int master, sigset_t *mask) { if (master_writable && in_buffer_full > 0) { - if ((k = write(master, in_buffer, in_buffer_full)) < 0) { + k = write(master, in_buffer, in_buffer_full); + if (k < 0) { if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) master_writable = false; @@ -584,7 +961,8 @@ static int process_pty(int master, sigset_t *mask) { if (master_readable && out_buffer_full < LINE_MAX) { - if ((k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full)) < 0) { + k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full); + if (k < 0) { if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) master_readable = false; @@ -599,7 +977,8 @@ static int process_pty(int master, sigset_t *mask) { if (stdout_writable && out_buffer_full > 0) { - if ((k = write(STDOUT_FILENO, out_buffer, out_buffer_full)) < 0) { + k = write(STDOUT_FILENO, out_buffer, out_buffer_full); + if (k < 0) { if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) stdout_writable = false; @@ -639,11 +1018,13 @@ int main(int argc, char *argv[]) { sigset_t mask; bool saved_attr_valid = false; struct winsize ws; + int kmsg_socket_pair[2] = { -1, -1 }; log_parse_environment(); log_open(); - if ((r = parse_argv(argc, argv)) <= 0) + r = parse_argv(argc, argv); + if (r <= 0) goto finish; if (arg_directory) { @@ -682,7 +1063,8 @@ int main(int argc, char *argv[]) { goto finish; } - if ((k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg)) < 0) { + k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg); + if (k < 0) { log_error("Failed to determine current cgroup: %s", strerror(-k)); goto finish; } @@ -698,18 +1080,20 @@ int main(int argc, char *argv[]) { goto finish; } - STRV_FOREACH(controller,arg_controllers) { + STRV_FOREACH(controller, arg_controllers) { k = cg_create_and_attach(*controller, newcg, 0); if (k < 0) log_warning("Failed to create cgroup in controller %s: %s", *controller, strerror(-k)); } - if ((master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY)) < 0) { + master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); + if (master < 0) { log_error("Failed to acquire pseudo tty: %m"); goto finish; } - if (!(console = ptsname(master))) { + console = ptsname(master); + if (!console) { log_error("Failed to determine tty name: %m"); goto finish; } @@ -735,8 +1119,8 @@ int main(int argc, char *argv[]) { cfmakeraw(&raw_attr); raw_attr.c_lflag &= ~ECHO; - if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { - log_error("Failed to set terminal attributes: %m"); + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { + log_error("Failed to create kmsg socket pair"); goto finish; } @@ -744,164 +1128,271 @@ int main(int argc, char *argv[]) { sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); - pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL); - 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"); + for (;;) { + siginfo_t status; - goto finish; - } + if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { + log_error("Failed to set terminal attributes: %m"); + goto finish; + } - if (pid == 0) { - /* child */ + pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL); + 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"); - const char *hn; - const char *home = NULL; - uid_t uid = (uid_t) -1; - gid_t gid = (gid_t) -1; - const char *envp[] = { - "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", - "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ - NULL, /* TERM */ - NULL, /* HOME */ - NULL, /* USER */ - NULL, /* LOGNAME */ - NULL - }; + goto finish; + } - envp[2] = strv_find_prefix(environ, "TERM="); + if (pid == 0) { + /* child */ - close_nointr_nofail(master); + const char *home = NULL; + uid_t uid = (uid_t) -1; + gid_t gid = (gid_t) -1; + const char *envp[] = { + "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", + "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ + NULL, /* TERM */ + NULL, /* HOME */ + NULL, /* USER */ + NULL, /* LOGNAME */ + NULL, /* container_uuid */ + NULL + }; + + envp[2] = strv_find_prefix(environ, "TERM="); - close_nointr(STDIN_FILENO); - close_nointr(STDOUT_FILENO); - close_nointr(STDERR_FILENO); + close_nointr_nofail(master); - close_all_fds(NULL, 0); + close_nointr(STDIN_FILENO); + close_nointr(STDOUT_FILENO); + close_nointr(STDERR_FILENO); - reset_all_signal_handlers(); + close_all_fds(&kmsg_socket_pair[1], 1); + + reset_all_signal_handlers(); + + assert_se(sigemptyset(&mask) == 0); + assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); + + if (open_terminal(console, O_RDWR) != STDIN_FILENO || + dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || + dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) + goto child_fail; - assert_se(sigemptyset(&mask) == 0); - assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); + if (setsid() < 0) { + log_error("setsid() failed: %m"); + goto child_fail; + } - if (setsid() < 0) - goto child_fail; + if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) { + log_error("PR_SET_PDEATHSIG failed: %m"); + goto child_fail; + } - if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) - goto child_fail; + /* Mark everything as slave, so that we still + * receive mounts from the real root, but don't + * propagate mounts to the real root. */ + if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { + log_error("MS_SLAVE|MS_REC failed: %m"); + goto child_fail; + } - /* Mark / as private, in case somebody marked it shared */ - if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0) - goto child_fail; + /* Turn directory into bind mount */ + if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) { + log_error("Failed to make bind mount."); + goto child_fail; + } - if (mount_all(arg_directory) < 0) - goto child_fail; + if (arg_read_only) + if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) { + log_error("Failed to make read-only."); + goto child_fail; + } - if (copy_devnodes(arg_directory, console) < 0) - goto child_fail; + if (mount_all(arg_directory) < 0) + goto child_fail; - if (chdir(arg_directory) < 0) { - log_error("chdir(%s) failed: %m", arg_directory); - goto child_fail; - } + if (copy_devnodes(arg_directory) < 0) + goto child_fail; - if (open_terminal("dev/console", O_RDWR) != STDIN_FILENO || - dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || - dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) - goto child_fail; + dev_setup(arg_directory); - if (mount(arg_directory, "/", "bind", MS_BIND|MS_MOVE, NULL) < 0) { - log_error("mount(MS_MOVE) failed: %m"); - goto child_fail; - } + if (setup_dev_console(arg_directory, console) < 0) + goto child_fail; - if (chroot(".") < 0) { - log_error("chroot() failed: %m"); - goto child_fail; - } + if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) + goto child_fail; - if (chdir("/") < 0) { - log_error("chdir() failed: %m"); - goto child_fail; - } + close_nointr_nofail(kmsg_socket_pair[1]); - umask(0022); + if (setup_boot_id(arg_directory) < 0) + goto child_fail; - loopback_setup(); + if (setup_timezone(arg_directory) < 0) + goto child_fail; - if (drop_capabilities() < 0) - goto child_fail; + if (setup_resolv_conf(arg_directory) < 0) + goto child_fail; - if (arg_user) { + if (setup_journal(arg_directory) < 0) + goto child_fail; - if (get_user_creds((const char**)&arg_user, &uid, &gid, &home) < 0) { - log_error("get_user_creds() failed: %m"); + if (chdir(arg_directory) < 0) { + log_error("chdir(%s) failed: %m", arg_directory); goto child_fail; } - if (mkdir_parents(home, 0775) < 0) { - log_error("mkdir_parents() failed: %m"); + if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) { + log_error("mount(MS_MOVE) failed: %m"); goto child_fail; } - if (safe_mkdir(home, 0775, uid, gid) < 0) { - log_error("safe_mkdir() failed: %m"); + if (chroot(".") < 0) { + log_error("chroot() failed: %m"); goto child_fail; } - if (initgroups((const char*)arg_user, gid) < 0) { - log_error("initgroups() failed: %m"); + if (chdir("/") < 0) { + log_error("chdir() failed: %m"); goto child_fail; } - if (setresgid(gid, gid, gid) < 0) { - log_error("setregid() failed: %m"); + umask(0022); + + loopback_setup(); + + if (drop_capabilities() < 0) { + log_error("drop_capabilities() failed: %m"); goto child_fail; } - if (setresuid(uid, uid, uid) < 0) { - log_error("setreuid() failed: %m"); + if (arg_user) { + + 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; + } + } + + if ((asprintf((char**)(envp + 3), "HOME=%s", home ? home: "/root") < 0) || + (asprintf((char**)(envp + 4), "USER=%s", arg_user ? arg_user : "root") < 0) || + (asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) { + log_oom(); goto child_fail; } - } - if ((asprintf((char**)(envp + 3), "HOME=%s", home? home: "/root") < 0) || - (asprintf((char**)(envp + 4), "USER=%s", arg_user? arg_user : "root") < 0) || - (asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user? arg_user : "root") < 0)) { - log_error("Out of memory"); - goto child_fail; - } + if (arg_uuid) { + if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) { + log_oom(); + goto child_fail; + } + } + + setup_hostname(); + + if (arg_boot) { + char **a; + size_t l; - if ((hn = file_name_from_path(arg_directory))) - sethostname(hn, strlen(hn)); + /* Automatically search for the init system */ - if (argc > optind) - execvpe(argv[optind], argv + optind, (char**) envp); - else { - chdir(home ? home : "/root"); - execle("/bin/bash", "-bash", NULL, (char**) envp); + l = 1 + argc - optind; + a = newa(char*, l + 1); + memcpy(a + 1, argv + optind, l * sizeof(char*)); + + a[0] = (char*) "/usr/lib/systemd/systemd"; + execve(a[0], a, (char**) envp); + + a[0] = (char*) "/lib/systemd/systemd"; + execve(a[0], a, (char**) envp); + + a[0] = (char*) "/sbin/init"; + execve(a[0], a, (char**) envp); + } else if (argc > optind) + execvpe(argv[optind], argv + optind, (char**) envp); + else { + chdir(home ? home : "/root"); + execle("/bin/bash", "-bash", NULL, (char**) envp); + } + + log_error("execv() failed: %m"); + + child_fail: + _exit(EXIT_FAILURE); } - log_error("execv() failed: %m"); + if (process_pty(master, &mask) < 0) + goto finish; - child_fail: - _exit(EXIT_FAILURE); - } - if (process_pty(master, &mask) < 0) - goto finish; + if (saved_attr_valid) + tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - if (saved_attr_valid) { - tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - saved_attr_valid = false; - } + r = wait_for_terminate(pid, &status); + if (r < 0) { + r = EXIT_FAILURE; + break; + } + + if (status.si_code == CLD_EXITED) { + if (status.si_status != 0) { + log_error("Container failed with error code %i.", status.si_status); + r = status.si_status; + break; + } - r = wait_for_terminate_and_warn(argc > optind ? argv[optind] : "bash", pid); + log_debug("Container exited successfully."); + break; + } else if (status.si_code == CLD_KILLED && + status.si_status == SIGINT) { + log_info("Container has been shut down."); + r = 0; + break; + } else if (status.si_code == CLD_KILLED && + status.si_status == SIGHUP) { + log_info("Container is being rebooted."); + continue; + } else if (status.si_code == CLD_KILLED || + status.si_code == CLD_DUMPED) { - if (r < 0) - r = EXIT_FAILURE; + log_error("Container terminated by signal %s.", signal_to_string(status.si_status)); + r = EXIT_FAILURE; + break; + } else { + log_error("Container failed due to unknown reason."); + r = EXIT_FAILURE; + break; + } + } finish: if (saved_attr_valid) @@ -910,6 +1401,8 @@ finish: if (master >= 0) close_nointr_nofail(master); + close_pipe(kmsg_socket_pair); + if (oldcg) cg_attach(SYSTEMD_CGROUP_CONTROLLER, oldcg, 0);