X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=b91b0b8a91f4bccc9011028e20a101bb9506fbf7;hp=8765b0185fbff5cde3bb269eb63e2d67f11d39bb;hb=77b6e19458f37cfde127ec6aa9494c0ac45ad890;hpb=04bc4a3f47074d22035831965e97b0990fcf6f63 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 8765b0185..b91b0b8a9 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -33,6 +33,7 @@ #include #include #include +#include #include #include #include @@ -41,11 +42,16 @@ #include #include +#ifdef HAVE_XATTR +#include +#endif + #include #include "log.h" #include "util.h" #include "mkdir.h" +#include "macro.h" #include "audit.h" #include "missing.h" #include "cgroup-util.h" @@ -54,6 +60,13 @@ #include "loopback-setup.h" #include "sd-id128.h" #include "dev-setup.h" +#include "fdset.h" +#include "build.h" +#include "fileio.h" + +#ifndef TTY_GID +#define TTY_GID 5 +#endif typedef enum LinkJournal { LINK_NO, @@ -66,6 +79,7 @@ static char *arg_directory = NULL; static char *arg_user = NULL; static char **arg_controllers = NULL; static char *arg_uuid = NULL; +static char *arg_machine = NULL; static bool arg_private_network = false; static bool arg_read_only = false; static bool arg_boot = false; @@ -92,23 +106,35 @@ static uint64_t arg_retain = (1ULL << CAP_SYS_NICE) | (1ULL << CAP_SYS_PTRACE) | (1ULL << CAP_SYS_TTY_CONFIG) | - (1ULL << CAP_SYS_RESOURCE); + (1ULL << CAP_SYS_RESOURCE) | + (1ULL << CAP_SYS_BOOT) | + (1ULL << CAP_AUDIT_WRITE) | + (1ULL << CAP_AUDIT_CONTROL); +static char **arg_bind = NULL; +static char **arg_bind_ro = NULL; 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" - " -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", + " -h --help Show this help\n" + " --version Print version string\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\n" + " cgroup hierarchies\n" + " --uuid=UUID Set a specific machine UUID for the container\n" + " -M --machine=NAME Set the machine name 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\n" + " capability\n" + " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" + " -j Equivalent to --link-journal=host\n" + " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" + " the container\n" + " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n", program_invocation_short_name); return 0; @@ -117,15 +143,19 @@ static int help(void) { static int parse_argv(int argc, char *argv[]) { enum { - ARG_PRIVATE_NETWORK = 0x100, + ARG_VERSION = 0x100, + ARG_PRIVATE_NETWORK, ARG_UUID, ARG_READ_ONLY, ARG_CAPABILITY, - ARG_LINK_JOURNAL + ARG_LINK_JOURNAL, + ARG_BIND, + ARG_BIND_RO }; static const struct option options[] = { { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, { "directory", required_argument, NULL, 'D' }, { "user", required_argument, NULL, 'u' }, { "controllers", required_argument, NULL, 'C' }, @@ -135,6 +165,9 @@ static int parse_argv(int argc, char *argv[]) { { "read-only", no_argument, NULL, ARG_READ_ONLY }, { "capability", required_argument, NULL, ARG_CAPABILITY }, { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, + { "bind", required_argument, NULL, ARG_BIND }, + { "bind-ro", required_argument, NULL, ARG_BIND_RO }, + { "machine", required_argument, NULL, 'M' }, { NULL, 0, NULL, 0 } }; @@ -143,7 +176,7 @@ static int parse_argv(int argc, char *argv[]) { assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:u:C:bj", options, NULL)) >= 0) { + while ((c = getopt_long(argc, argv, "+hD:u:C:bM:j", options, NULL)) >= 0) { switch (c) { @@ -151,6 +184,11 @@ static int parse_argv(int argc, char *argv[]) { help(); return 0; + case ARG_VERSION: + puts(PACKAGE_STRING); + puts(SYSTEMD_FEATURES); + return 0; + case 'D': free(arg_directory); arg_directory = canonicalize_file_name(optarg); @@ -163,22 +201,19 @@ static int parse_argv(int argc, char *argv[]) { case 'u': free(arg_user); - if (!(arg_user = strdup(optarg))) { - log_error("Failed to duplicate user name."); - return -ENOMEM; - } + arg_user = strdup(optarg); + if (!arg_user) + return log_oom(); break; case 'C': strv_free(arg_controllers); arg_controllers = strv_split(optarg, ","); - if (!arg_controllers) { - log_error("Failed to split controllers list."); - return -ENOMEM; - } - strv_uniq(arg_controllers); + if (!arg_controllers) + return log_oom(); + cg_shorten_controllers(arg_controllers); break; case ARG_PRIVATE_NETWORK: @@ -190,9 +225,27 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_UUID: + if (!id128_is_valid(optarg)) { + log_error("Invalid UUID: %s", optarg); + return -EINVAL; + } + arg_uuid = optarg; break; + case 'M': + if (!hostname_is_valid(optarg)) { + log_error("Invalid machine name: %s", optarg); + return -EINVAL; + } + + free(arg_machine); + arg_machine = strdup(optarg); + if (!arg_machine) + return log_oom(); + + break; + case ARG_READ_ONLY: arg_read_only = true; break; @@ -242,6 +295,43 @@ static int parse_argv(int argc, char *argv[]) { break; + case ARG_BIND: + case ARG_BIND_RO: { + _cleanup_free_ char *a = NULL, *b = NULL; + char *e; + char ***x; + int r; + + x = c == ARG_BIND ? &arg_bind : &arg_bind_ro; + + e = strchr(optarg, ':'); + if (e) { + a = strndup(optarg, e - optarg); + b = strdup(e + 1); + } else { + a = strdup(optarg); + b = strdup(optarg); + } + + if (!a || !b) + return log_oom(); + + if (!path_is_absolute(a) || !path_is_absolute(b)) { + log_error("Invalid bind mount specification: %s", optarg); + return -EINVAL; + } + + r = strv_extend(x, a); + if (r < 0) + return r; + + r = strv_extend(x, b); + if (r < 0) + return r; + + break; + } + case '?': return -EINVAL; @@ -271,7 +361,8 @@ static int mount_all(const char *dest) { { 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", NULL, NULL, MS_BIND, true }, + { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, 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", NULL, NULL, MS_BIND, false }, /* Bind mount first */ @@ -281,24 +372,18 @@ static int mount_all(const char *dest) { unsigned k; int r = 0; - char *where; for (k = 0; k < ELEMENTSOF(mount_table); k++) { + _cleanup_free_ char *where = NULL; int t; - if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) { - log_oom(); - - if (r == 0) - r = -ENOMEM; - - break; - } + where = strjoin(dest, "/", mount_table[k].where, NULL); + if (!where) + return log_oom(); 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; @@ -310,7 +395,7 @@ static int mount_all(const char *dest) { if (mount_table[k].what && t > 0) continue; - mkdir_p_label(where, 0755); + mkdir_p(where, 0755); if (mount(mount_table[k].what, where, @@ -324,42 +409,100 @@ static int mount_all(const char *dest) { if (r == 0) r = -errno; } - - free(where); } return r; } +static int mount_binds(const char *dest, char **l, unsigned long flags) { + char **x, **y; + + STRV_FOREACH_PAIR(x, y, l) { + _cleanup_free_ char *where = NULL; + + where = strjoin(dest, "/", *y, NULL); + if (!where) + return log_oom(); + + mkdir_p_label(where, 0755); + + if (mount(*x, where, "bind", MS_BIND, NULL) < 0) { + log_error("mount(%s) failed: %m", where); + return -errno; + } + + if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) { + log_error("mount(%s) failed: %m", where); + return -errno; + } + } + + return 0; +} + static int setup_timezone(const char *dest) { - char *where; + _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL; + char *z, *y; + int r; assert(dest); /* Fix the timezone, if possible */ + r = readlink_malloc("/etc/localtime", &p); + if (r < 0) { + log_warning("/etc/localtime is not a symlink, not updating container timezone."); + return 0; + } + + z = path_startswith(p, "../usr/share/zoneinfo/"); + if (!z) + z = path_startswith(p, "/usr/share/zoneinfo/"); + if (!z) { + log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone."); + return 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); + r = readlink_malloc(where, &q); + if (r >= 0) { + y = path_startswith(q, "../usr/share/zoneinfo/"); + if (!y) + y = path_startswith(q, "/usr/share/zoneinfo/"); - free(where); - where = strappend(dest, "/etc/timezone"); - if (!where) + /* Already pointing to the right place? Then do nothing .. */ + if (y && streq(y, z)) + return 0; + } + + check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL); + if (!check) 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); + if (access(check, F_OK) < 0) { + log_warning("Timezone %s does not exist in container, not updating container timezone.", z); + return 0; + } + + what = strappend("../usr/share/zoneinfo/", z); + if (!what) + return log_oom(); - free(where); + unlink(where); + if (symlink(what, where) < 0) { + log_error("Failed to correct timezone of container: %m"); + return 0; + } return 0; } static int setup_resolv_conf(const char *dest) { - char *where; + char _cleanup_free_ *where = NULL; + _cleanup_close_ int fd = -1; assert(dest); @@ -371,16 +514,24 @@ static int setup_resolv_conf(const char *dest) { 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); + fd = open(where, O_WRONLY|O_CREAT|O_EXCL|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW, 0644); + + /* We don't really care for the results of this really. If it + * fails, it fails, but meh... */ + if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) < 0) + log_warning("Failed to bind mount /etc/resolv.conf: %m"); + else + if (mount("/etc/resolv.conf", where, "bind", + MS_BIND|MS_REMOUNT|MS_RDONLY, NULL) < 0) { + log_error("Failed to remount /etc/resolv.conf readonly: %m"); + return -errno; + } return 0; } static int setup_boot_id(const char *dest) { - char *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; sd_id128_t rnd; char as_uuid[37]; int r; @@ -391,21 +542,14 @@ static int setup_boot_id(const char *dest) { * the container gets a new one */ from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id"); - if (!from) { - r = log_oom(); - goto finish; - } - to = strappend(dest, "/proc/sys/kernel/random/boot_id"); - if (!to) { - r = log_oom(); - goto finish; - } + if (!from || !to) + return log_oom(); r = sd_id128_randomize(&rnd); if (r < 0) { log_error("Failed to generate random boot id: %s", strerror(-r)); - goto finish; + return r; } snprintf(as_uuid, sizeof(as_uuid), @@ -413,24 +557,19 @@ static int setup_boot_id(const char *dest) { SD_ID128_FORMAT_VAL(rnd)); char_array_0(as_uuid); - r = write_one_line_file(from, as_uuid); + r = write_string_file(from, as_uuid); if (r < 0) { log_error("Failed to write boot id: %s", strerror(-r)); - goto finish; + 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); + } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL)) + log_warning("Failed to make boot id read-only: %m"); unlink(from); - -finish: - free(from); - free(to); - return r; } @@ -442,13 +581,11 @@ static int copy_devnodes(const char *dest) { "full\0" "random\0" "urandom\0" - "tty\0" - "ptmx\0" - "rtc0\0"; + "tty\0"; const char *d; int r = 0; - mode_t u; + _cleanup_umask_ mode_t u; assert(dest); @@ -456,18 +593,13 @@ static int copy_devnodes(const char *dest) { NULSTR_FOREACH(d, devnodes) { struct stat st; - char *from = NULL, *to = NULL; + _cleanup_free_ char *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; @@ -485,7 +617,7 @@ static int copy_devnodes(const char *dest) { } 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; @@ -495,21 +627,31 @@ static int copy_devnodes(const char *dest) { if (r == 0) r = -errno; } - - free(from); - free(to); } - umask(u); - return r; } +static int setup_ptmx(const char *dest) { + _cleanup_free_ char *p = NULL; + + p = strappend(dest, "/dev/ptmx"); + if (!p) + return log_oom(); + + if (symlink("pts/ptmx", p) < 0) { + log_error("Failed to create /dev/ptmx symlink: %m"); + return -errno; + } + + return 0; +} + static int setup_dev_console(const char *dest, const char *console) { struct stat st; - char *to = NULL; + _cleanup_free_ char *to = NULL; int r; - mode_t u; + _cleanup_umask_ mode_t u; assert(dest); assert(console); @@ -518,25 +660,21 @@ static int setup_dev_console(const char *dest, const char *console) { if (stat(console, &st) < 0) { log_error("Failed to stat %s: %m", console); - r = -errno; - goto finish; + return -errno; } else if (!S_ISCHR(st.st_mode)) { - log_error("/dev/console is not a char device."); - r = -EIO; - goto finish; + log_error("/dev/console is not a char device"); + return -EIO; } r = chmod_and_chown(console, 0600, 0, 0); if (r < 0) { log_error("Failed to correct access mode for TTY: %s", strerror(-r)); - goto finish; + return r; } - if (asprintf(&to, "%s/dev/console", dest) < 0) { - r = log_oom(); - goto finish; - } + 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 @@ -547,32 +685,29 @@ static int setup_dev_console(const char *dest, const char *console) { if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) { log_error("mknod() for /dev/console failed: %m"); - r = -errno; - goto finish; + return -errno; } if (mount(console, to, "bind", MS_BIND, NULL) < 0) { log_error("Bind mount for /dev/console failed: %m"); - r = -errno; - goto finish; + return -errno; } -finish: - free(to); - umask(u); - - return r; + return 0; } static int setup_kmsg(const char *dest, int kmsg_socket) { - char *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; int r, fd, k; - mode_t u; + _cleanup_umask_ mode_t u; union { struct cmsghdr cmsghdr; uint8_t buf[CMSG_SPACE(sizeof(int))]; - } control; - struct msghdr mh; + } control = {}; + struct msghdr mh = { + .msg_control = &control, + .msg_controllen = sizeof(control), + }; struct cmsghdr *cmsg; assert(dest); @@ -587,47 +722,32 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { * 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) { - r = log_oom(); - goto finish; - } - - if (asprintf(&to, "%s/proc/kmsg", dest) < 0) { - r = log_oom(); - goto finish; - } + 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"); - r = -errno; - goto finish; + return -errno; } r = chmod_and_chown(from, 0600, 0, 0); if (r < 0) { log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r)); - goto finish; + return r; } if (mount(from, to, "bind", MS_BIND, NULL) < 0) { log_error("Bind mount for /proc/kmsg failed: %m"); - r = -errno; - goto finish; + return -errno; } fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC); if (fd < 0) { log_error("Failed to open fifo: %m"); - r = -errno; - goto finish; + return -errno; } - 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; @@ -643,97 +763,76 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { if (k < 0) { log_error("Failed to send FIFO fd: %m"); - r = -errno; - goto finish; + return -errno; } /* And now make the FIFO unavailable as /dev/kmsg... */ unlink(from); - -finish: - free(from); - free(to); - umask(u); - - return r; + 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; + if (sethostname(arg_machine, strlen(arg_machine)) < 0) + return -errno; - free(hn); - } - - return r; + return 0; } static int setup_journal(const char *directory) { sd_id128_t machine_id; - char *p = NULL, *b = NULL, *l, *q = NULL, *d = NULL; + _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL; + char *id; int r; if (arg_link_journal == LINK_NO) return 0; p = strappend(directory, "/etc/machine-id"); - if (!p) { - r = log_oom(); - goto finish; - } + if (!p) + return log_oom(); 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)); + if (r == -ENOENT && arg_link_journal == LINK_AUTO) + return 0; + else if (r < 0) { + log_error("Failed to read machine ID from %s: %s", p, strerror(-r)); return r; } - l = strstrip(b); - if (isempty(l) && arg_link_journal == LINK_AUTO) { - r = 0; - goto finish; - } + id = strstrip(b); + if (isempty(id) && arg_link_journal == LINK_AUTO) + return 0; - /* Verify validaty */ - r = sd_id128_from_string(l, &machine_id); + /* Verify validity */ + r = sd_id128_from_string(id, &machine_id); if (r < 0) { - log_error("Failed to parse machine ID: %s", strerror(-r)); - goto finish; + log_error("Failed to parse machine ID from %s: %s", p, strerror(-r)); + return r; } free(p); - p = strappend("/var/log/journal/", l); - q = strjoin(directory, "/var/log/journal/", l, NULL); - if (!p || !q) { - r = log_oom(); - goto finish; + p = strappend("/var/log/journal/", id); + q = strjoin(directory, "/var/log/journal/", id, NULL); + if (!p || !q) + return log_oom(); + + if (path_is_mount_point(p, false) > 0) { + if (arg_link_journal != LINK_AUTO) { + log_error("%s: already a mount point, refusing to use for journal", p); + return -EEXIST; + } + + return 0; } - if (path_is_mount_point(p, false) > 0 || - path_is_mount_point(q, false) > 0) { + if (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; + log_error("%s: already a mount point, refusing to use for journal", q); + return -EEXIST; } - r = 0; - goto finish; + return 0; } r = readlink_and_make_absolute(p, &d); @@ -742,116 +841,154 @@ static int setup_journal(const char *directory) { arg_link_journal == LINK_AUTO) && path_equal(d, q)) { - mkdir_p(q, 0755); - - r = 0; - goto finish; + r = mkdir_p(q, 0755); + if (r < 0) + log_warning("failed to create directory %s: %m", q); + return 0; } if (unlink(p) < 0) { log_error("Failed to remove symlink %s: %m", p); - r = -errno; - goto finish; + return -errno; } } 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 { + if (errno == ENOTDIR) { + log_error("%s already exists and is neither a symlink nor a directory", p); + return r; + } else { log_error("Failed to remove %s: %m", p); - r = -errno; + return -errno; } - - goto finish; } } else if (r != -ENOENT) { log_error("readlink(%s) failed: %m", p); - goto finish; + return r; } 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; + return -errno; } - mkdir_p(q, 0755); - - r = 0; - goto finish; + r = mkdir_p(q, 0755); + if (r < 0) + log_warning("failed to create directory %s: %m", q); + return 0; } if (arg_link_journal == LINK_HOST) { r = mkdir_p(p, 0755); if (r < 0) { log_error("Failed to create %s: %m", p); - goto finish; + return r; } - } else if (access(p, F_OK) < 0) { - r = 0; - goto finish; - } + } else if (access(p, F_OK) < 0) + return 0; if (dir_is_empty(q) == 0) { log_error("%s not empty.", q); - r = -ENOTEMPTY; - goto finish; + return -ENOTEMPTY; } r = mkdir_p(q, 0755); if (r < 0) { log_error("Failed to create %s: %m", q); - goto finish; + return r; } 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; + return -errno; } - r = 0; + return 0; +} -finish: - free(p); - free(q); - free(d); - free(b); - return r; +static int setup_cgroup(const char *path) { + char **c; + int r; -} + r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, path, 1); + if (r < 0) { + log_error("Failed to create cgroup: %s", strerror(-r)); + return r; + } -static int drop_capabilities(void) { - return capability_bounding_set_drop(~arg_retain, false); + STRV_FOREACH(c, arg_controllers) { + r = cg_create_and_attach(*c, path, 1); + if (r < 0) + log_warning("Failed to create cgroup in controller %s: %s", *c, strerror(-r)); + } + + return 0; } -static int is_os_tree(const char *path) { - int r; - char *p; - /* We use /bin/sh as flag file if something is an OS */ +static int save_attributes(const char *cgroup, pid_t pid, const char *uuid, const char *directory) { +#ifdef HAVE_XATTR + _cleanup_free_ char *path = NULL; + char buf[DECIMAL_STR_MAX(pid_t)]; + int r = 0, k; - if (asprintf(&p, "%s/bin/sh", path) < 0) - return -ENOMEM; + assert(cgroup); + assert(pid >= 0); + assert(arg_directory); - r = access(p, F_OK); - free(p); + assert_se(snprintf(buf, sizeof(buf), "%lu", (unsigned long) pid) < (int) sizeof(buf)); + + r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, cgroup, NULL, &path); + if (r < 0) { + log_error("Failed to get path: %s", strerror(-r)); + return r; + } + + r = setxattr(path, "trusted.init_pid", buf, strlen(buf), XATTR_CREATE); + if (r < 0) + log_warning("Failed to set %s attribute on %s: %m", "trusted.init_pid", path); - return r < 0 ? 0 : 1; + if (uuid) { + k = setxattr(path, "trusted.machine_id", uuid, strlen(uuid), XATTR_CREATE); + if (k < 0) { + log_warning("Failed to set %s attribute on %s: %m", "trusted.machine_id", path); + if (r == 0) + r = k; + } + } + + k = setxattr(path, "trusted.root_directory", directory, strlen(directory), XATTR_CREATE); + if (k < 0) { + log_warning("Failed to set %s attribute on %s: %m", "trusted.root_directory", path); + if (r == 0) + r = k; + } + return r; +#else + return 0; +#endif +} + +static int drop_capabilities(void) { + return capability_bounding_set_drop(~arg_retain, false); } -static int process_pty(int master, sigset_t *mask) { +static int process_pty(int master, pid_t pid, sigset_t *mask) { char in_buffer[LINE_MAX], out_buffer[LINE_MAX]; size_t in_buffer_full = 0, out_buffer_full = 0; struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev; bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false; int ep = -1, signal_fd = -1, r; + bool tried_orderly_shutdown = false; + + assert(master >= 0); + assert(pid > 0); + assert(mask); fd_nonblock(STDIN_FILENO, 1); fd_nonblock(STDOUT_FILENO, 1); @@ -871,9 +1008,19 @@ static int process_pty(int master, sigset_t *mask) { goto finish; } - zero(stdin_ev); - stdin_ev.events = EPOLLIN|EPOLLET; - stdin_ev.data.fd = STDIN_FILENO; + /* We read from STDIN only if this is actually a TTY, + * otherwise we assume non-interactivity. */ + if (isatty(STDIN_FILENO)) { + zero(stdin_ev); + stdin_ev.events = EPOLLIN|EPOLLET; + stdin_ev.data.fd = STDIN_FILENO; + + if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0) { + log_error("Failed to register STDIN in epoll: %m"); + r = -errno; + goto finish; + } + } zero(stdout_ev); stdout_ev.events = EPOLLOUT|EPOLLET; @@ -887,11 +1034,19 @@ static int process_pty(int master, sigset_t *mask) { signal_ev.events = EPOLLIN; signal_ev.data.fd = signal_fd; - if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0 || - epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 || - epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 || + if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0) { + if (errno != EPERM) { + log_error("Failed to register stdout in epoll: %m"); + r = -errno; + goto finish; + } + /* stdout without epoll support. Likely redirected to regular file. */ + stdout_writable = true; + } + + if (epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 || epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) { - log_error("Failed to regiser fds in epoll: %m"); + log_error("Failed to register fds in epoll: %m"); r = -errno; goto finish; } @@ -959,6 +1114,14 @@ static int process_pty(int master, sigset_t *mask) { /* The window size changed, let's forward that. */ if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0) ioctl(master, TIOCSWINSZ, &ws); + } else if (sfsi.ssi_signo == SIGTERM && arg_boot && !tried_orderly_shutdown) { + + log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination."); + + /* This only works for systemd... */ + tried_orderly_shutdown = true; + kill(pid, SIGRTMIN+3); + } else { r = 0; goto finish; @@ -1056,25 +1219,42 @@ finish: return r; } +static bool audit_enabled(void) { + int fd; + + fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT); + if (fd >= 0) { + close_nointr_nofail(fd); + return true; + } + + return false; +} + int main(int argc, char *argv[]) { pid_t pid = 0; int r = EXIT_FAILURE, k; - char *oldcg = NULL, *newcg = NULL; - char **controller = NULL; - int master = -1; + _cleanup_free_ char *newcg = NULL; + _cleanup_close_ int master = -1; + int n_fd_passed; const char *console = NULL; struct termios saved_attr, raw_attr; sigset_t mask; bool saved_attr_valid = false; struct winsize ws; int kmsg_socket_pair[2] = { -1, -1 }; + FDSet *fds = NULL; log_parse_environment(); log_open(); - r = parse_argv(argc, argv); - if (r <= 0) + k = parse_argv(argc, argv); + if (k < 0) + goto finish; + else if (k == 0) { + r = EXIT_SUCCESS; goto finish; + } if (arg_directory) { char *p; @@ -1086,12 +1266,26 @@ int main(int argc, char *argv[]) { arg_directory = get_current_dir_name(); if (!arg_directory) { - log_error("Failed to determine path"); + log_error("Failed to determine path, please use -D."); goto finish; } path_kill_slashes(arg_directory); + if (!arg_machine) { + arg_machine = strdup(path_get_file_name(arg_directory)); + if (!arg_machine) { + log_oom(); + goto finish; + } + + hostname_cleanup(arg_machine, false); + if (isempty(arg_machine)) { + log_error("Failed to determine machine name automatically, please use -M."); + goto finish; + } + } + if (geteuid() != 0) { log_error("Need to be root."); goto finish; @@ -1102,37 +1296,49 @@ int main(int argc, char *argv[]) { goto finish; } + if (audit_enabled()) { + log_warning("The kernel auditing subsystem is known to be incompatible with containers.\n" + "Please make sure to turn off auditing with 'audit=0' on the kernel command\n" + "line before using systemd-nspawn. Sleeping for 5s...\n"); + sleep(5); + } + if (path_equal(arg_directory, "/")) { log_error("Spawning container on root directory not supported."); goto finish; } - if (is_os_tree(arg_directory) <= 0) { - log_error("Directory %s doesn't look like an OS root directory. Refusing.", arg_directory); + if (path_is_os_tree(arg_directory) <= 0) { + log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); goto finish; } - 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; + log_close(); + n_fd_passed = sd_listen_fds(false); + if (n_fd_passed > 0) { + k = fdset_new_listen_fds(&fds, false); + if (k < 0) { + log_error("Failed to collect file descriptors: %s", strerror(-k)); + goto finish; + } } + fdset_close_others(fds); + log_open(); - if (asprintf(&newcg, "%s/nspawn-%lu", oldcg, (unsigned long) getpid()) < 0) { - log_error("Failed to allocate cgroup path."); + k = cg_get_machine_path(arg_machine, &newcg); + if (k < 0) { + log_error("Failed to determine machine cgroup path: %s", strerror(-k)); goto finish; } - k = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, newcg, 0); - if (k < 0) { - log_error("Failed to create cgroup: %s", strerror(-k)); - goto finish; - } + k = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, newcg, true); + if (k <= 0 && k != -ENOENT) { + log_error("Container already running."); + + free(newcg); + newcg = NULL; - 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)); + goto finish; } master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); @@ -1157,277 +1363,407 @@ int main(int argc, char *argv[]) { goto finish; } - if (tcgetattr(STDIN_FILENO, &saved_attr) < 0) { - log_error("Failed to get terminal attributes: %m"); - goto finish; - } - - saved_attr_valid = true; + if (tcgetattr(STDIN_FILENO, &saved_attr) >= 0) { + saved_attr_valid = true; - raw_attr = saved_attr; - cfmakeraw(&raw_attr); - raw_attr.c_lflag &= ~ECHO; - - if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { - log_error("Failed to set terminal attributes: %m"); - goto finish; + raw_attr = saved_attr; + cfmakeraw(&raw_attr); + raw_attr.c_lflag &= ~ECHO; } if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { - log_error("Failed to create kmsg socket pair"); + log_error("Failed to create kmsg socket pair."); goto finish; } + sd_notify(0, "READY=1"); + assert_se(sigemptyset(&mask) == 0); 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"); - - goto finish; - } - - if (pid == 0) { - /* child */ + for (;;) { + siginfo_t status; + int pipefd[2], pipefd2[2]; - 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 - }; + if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) { + log_error("pipe2(): %m"); + goto finish; + } - envp[2] = strv_find_prefix(environ, "TERM="); + if (pipe2(pipefd2, O_NONBLOCK|O_CLOEXEC) < 0) { + log_error("pipe2(): %m"); + close_pipe(pipefd); + goto finish; + } - close_nointr_nofail(master); + 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"); - close_nointr(STDIN_FILENO); - close_nointr(STDOUT_FILENO); - close_nointr(STDERR_FILENO); + goto finish; + } - close_all_fds(&kmsg_socket_pair[1], 1); + if (pid == 0) { + /* child */ + const char *home = NULL; + uid_t uid = (uid_t) -1; + gid_t gid = (gid_t) -1; + unsigned n_env = 2; + 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, /* LISTEN_FDS */ + NULL, /* LISTEN_PID */ + NULL + }; + + envp[n_env] = strv_find_prefix(environ, "TERM="); + if (envp[n_env]) + n_env ++; + + /* Wait for the parent process to log our PID */ + close_nointr_nofail(pipefd[1]); + fd_wait_for_event(pipefd[0], POLLHUP, -1); + close_nointr_nofail(pipefd[0]); + + close_nointr_nofail(master); + master = -1; + + if (saved_attr_valid) { + if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { + log_error("Failed to set terminal attributes: %m"); + goto child_fail; + } + } - reset_all_signal_handlers(); + close_nointr(STDIN_FILENO); + close_nointr(STDOUT_FILENO); + close_nointr(STDERR_FILENO); - assert_se(sigemptyset(&mask) == 0); - assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); + close_nointr_nofail(kmsg_socket_pair[0]); + kmsg_socket_pair[0] = -1; - 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; + reset_all_signal_handlers(); - if (setsid() < 0) { - log_error("setsid() failed: %m"); - goto child_fail; - } + assert_se(sigemptyset(&mask) == 0); + assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); - if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) { - log_error("PR_SET_PDEATHSIG failed: %m"); - goto child_fail; - } + k = open_terminal(console, O_RDWR); + if (k != STDIN_FILENO) { + if (k >= 0) { + close_nointr_nofail(k); + k = -EINVAL; + } - /* 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; - } + log_error("Failed to open console: %s", strerror(-k)); + 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 (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || + dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) { + log_error("Failed to duplicate console: %m"); + 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."); + if (setsid() < 0) { + log_error("setsid() failed: %m"); goto child_fail; } - if (mount_all(arg_directory) < 0) - goto child_fail; + if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) { + log_error("PR_SET_PDEATHSIG failed: %m"); + goto child_fail; + } - if (copy_devnodes(arg_directory) < 0) - goto child_fail; + if (setup_cgroup(newcg) < 0) + goto child_fail; - dev_setup(arg_directory); + close_pipe(pipefd2); - if (setup_dev_console(arg_directory, console) < 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; + } - if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 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; + } - close_nointr_nofail(kmsg_socket_pair[1]); + 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 (setup_boot_id(arg_directory) < 0) - goto child_fail; + if (mount_all(arg_directory) < 0) + goto child_fail; - if (setup_timezone(arg_directory) < 0) - goto child_fail; + if (copy_devnodes(arg_directory) < 0) + goto child_fail; - if (setup_resolv_conf(arg_directory) < 0) - goto child_fail; + if (setup_ptmx(arg_directory) < 0) + goto child_fail; - if (setup_journal(arg_directory) < 0) - goto child_fail; + dev_setup(arg_directory); - if (chdir(arg_directory) < 0) { - log_error("chdir(%s) failed: %m", arg_directory); - goto child_fail; - } + if (setup_dev_console(arg_directory, console) < 0) + goto child_fail; - if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) { - log_error("mount(MS_MOVE) failed: %m"); - goto child_fail; - } + if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) + goto child_fail; - if (chroot(".") < 0) { - log_error("chroot() failed: %m"); - goto child_fail; - } + close_nointr_nofail(kmsg_socket_pair[1]); + kmsg_socket_pair[1] = -1; - if (chdir("/") < 0) { - log_error("chdir() failed: %m"); - goto child_fail; - } + if (setup_boot_id(arg_directory) < 0) + goto child_fail; - umask(0022); + if (setup_timezone(arg_directory) < 0) + goto child_fail; - loopback_setup(); + if (setup_resolv_conf(arg_directory) < 0) + goto child_fail; - if (drop_capabilities() < 0) { - log_error("drop_capabilities() failed: %m"); - goto child_fail; - } + if (setup_journal(arg_directory) < 0) + goto child_fail; - if (arg_user) { + if (mount_binds(arg_directory, arg_bind, 0) < 0) + goto child_fail; - if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) { - log_error("get_user_creds() failed: %m"); + if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) goto child_fail; - } - if (mkdir_parents_label(home, 0775) < 0) { - log_error("mkdir_parents_label() failed: %m"); + if (chdir(arg_directory) < 0) { + log_error("chdir(%s) failed: %m", arg_directory); goto child_fail; } - if (mkdir_safe_label(home, 0775, uid, gid) < 0) { - log_error("mkdir_safe_label() failed: %m"); + if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) { + log_error("mount(MS_MOVE) failed: %m"); goto child_fail; } - if (initgroups((const char*)arg_user, gid) < 0) { - log_error("initgroups() failed: %m"); + if (chroot(".") < 0) { + log_error("chroot() failed: %m"); goto child_fail; } - if (setresgid(gid, gid, gid) < 0) { - log_error("setregid() failed: %m"); + if (chdir("/") < 0) { + log_error("chdir() failed: %m"); goto child_fail; } - if (setresuid(uid, uid, uid) < 0) { - log_error("setreuid() failed: %m"); + umask(0022); + + loopback_setup(); + + if (drop_capabilities() < 0) { + log_error("drop_capabilities() 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 (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; + } + } - if (arg_uuid) { - if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) { + 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) || + (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) { log_oom(); goto child_fail; } - } - setup_hostname(); + if (arg_uuid) { + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", arg_uuid) < 0) { + log_oom(); + goto child_fail; + } + } + + if (fdset_size(fds) > 0) { + k = fdset_cloexec(fds, false); + if (k < 0) { + log_error("Failed to unset O_CLOEXEC for file descriptors."); + goto child_fail; + } + + if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) || + (asprintf((char **)(envp + n_env++), "LISTEN_PID=%lu", (unsigned long) 1) < 0)) { + log_oom(); + goto child_fail; + } + } + + setup_hostname(); + + if (arg_boot) { + char **a; + size_t l; - if (arg_boot) { - char **a; - size_t l; + /* Automatically search for the init system */ - /* Automatically search for the init system */ + l = 1 + argc - optind; + a = newa(char*, l + 1); + memcpy(a + 1, argv + optind, l * sizeof(char*)); - 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*) "/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); + } - a[0] = (char*) "/lib/systemd/systemd"; - execve(a[0], a, (char**) envp); + log_error("execv() failed: %m"); - 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); + child_fail: + _exit(EXIT_FAILURE); } - log_error("execv() failed: %m"); + log_info("Init process in the container running as PID %lu.", (unsigned long) pid); + close_nointr_nofail(pipefd[0]); + close_nointr_nofail(pipefd[1]); - child_fail: - _exit(EXIT_FAILURE); - } + /* Wait for the child process to establish cgroup hierarchy */ + close_nointr_nofail(pipefd2[1]); + fd_wait_for_event(pipefd2[0], POLLHUP, -1); + close_nointr_nofail(pipefd2[0]); - if (process_pty(master, &mask) < 0) - goto finish; + save_attributes(newcg, pid, arg_uuid, arg_directory); - if (saved_attr_valid) { - tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - saved_attr_valid = false; - } + fdset_free(fds); + fds = NULL; - r = wait_for_terminate_and_warn(argc > optind ? argv[optind] : "bash", pid); + if (process_pty(master, pid, &mask) < 0) + goto finish; - if (r < 0) - r = EXIT_FAILURE; + if (saved_attr_valid) + tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); + + k = wait_for_terminate(pid, &status); + if (k < 0) { + r = EXIT_FAILURE; + break; + } + + if (status.si_code == CLD_EXITED) { + r = status.si_status; + if (status.si_status != 0) { + log_error("Container failed with error code %i.", status.si_status); + break; + } + + 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) { + + 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) tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - if (master >= 0) - close_nointr_nofail(master); - close_pipe(kmsg_socket_pair); - if (oldcg) - cg_attach(SYSTEMD_CGROUP_CONTROLLER, oldcg, 0); - if (newcg) cg_kill_recursive_and_wait(SYSTEMD_CGROUP_CONTROLLER, newcg, true); free(arg_directory); + free(arg_machine); strv_free(arg_controllers); - free(oldcg); - free(newcg); + + fdset_free(fds); return r; }