X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=c91f6cce21faf264a2f1680b0e337d1a53ba26a6;hp=1f3bda5b4aded530b46484e6202cc0b0f3cd50e1;hb=a8828ed93878b4b4866d40ebfb660e54995ff72e;hpb=f274ece0f76b5709408821e317e87aef76123db6 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 1f3bda5b4..c91f6cce2 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -33,16 +33,21 @@ #include #include #include -#include #include #include #include #include #include #include +#include +#include +#if HAVE_SELINUX +#include +#endif -#include - +#include "sd-daemon.h" +#include "sd-bus.h" +#include "sd-id128.h" #include "log.h" #include "util.h" #include "mkdir.h" @@ -53,9 +58,16 @@ #include "strv.h" #include "path-util.h" #include "loopback-setup.h" -#include "sd-id128.h" #include "dev-setup.h" #include "fdset.h" +#include "build.h" +#include "fileio.h" +#include "bus-util.h" +#include "bus-error.h" +#include "ptyfwd.h" +#include "bus-kernel.h" +#include "env-util.h" +#include "def.h" typedef enum LinkJournal { LINK_NO, @@ -66,8 +78,11 @@ typedef enum LinkJournal { static char *arg_directory = NULL; static char *arg_user = NULL; -static char **arg_controllers = NULL; -static char *arg_uuid = NULL; +static sd_id128_t arg_uuid = {}; +static char *arg_machine = NULL; +static char *process_label = NULL; +static char *file_label = NULL; +static const char *arg_slice = NULL; static bool arg_private_network = false; static bool arg_read_only = false; static bool arg_boot = false; @@ -95,23 +110,39 @@ static uint64_t arg_retain = (1ULL << CAP_SYS_PTRACE) | (1ULL << CAP_SYS_TTY_CONFIG) | (1ULL << CAP_SYS_RESOURCE) | - (1ULL << CAP_SYS_BOOT); + (1ULL << CAP_SYS_BOOT) | + (1ULL << CAP_AUDIT_WRITE) | + (1ULL << CAP_AUDIT_CONTROL) | + (1ULL << CAP_MKNOD); +static char **arg_bind = NULL; +static char **arg_bind_ro = NULL; +static char **arg_setenv = 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" + " --uuid=UUID Set a specific machine UUID for the container\n" + " -M --machine=NAME Set the machine name for the container\n" + " -S --slice=SLICE Place the container in the specified slice\n" + " -L --file-label=LABEL Set the MAC file label to be used by tmpfs file systems in container\n" + " -Z --process-label=LABEL Set the MAC label to be used by processes in 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" + " --drop-capability=CAP Drop the specified capability from the default set\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" + " --setenv=NAME=VALUE Pass an environment variable to PID 1\n", program_invocation_short_name); return 0; @@ -120,45 +151,62 @@ 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_DROP_CAPABILITY, + ARG_LINK_JOURNAL, + ARG_BIND, + ARG_BIND_RO, + ARG_SETENV, }; 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' }, { "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 }, + { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY }, { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, - { NULL, 0, NULL, 0 } + { "bind", required_argument, NULL, ARG_BIND }, + { "bind-ro", required_argument, NULL, ARG_BIND_RO }, + { "machine", required_argument, NULL, 'M' }, + { "slice", required_argument, NULL, 'S' }, + { "setenv", required_argument, NULL, ARG_SETENV }, + { "process-label", required_argument, NULL, 'Z' }, + { "file-label", required_argument, NULL, 'L' }, + {} }; - int c; + int c, r; 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:bL:M:jS:Z:", options, NULL)) >= 0) { switch (c) { case 'h': - help(); + return help(); + + case ARG_VERSION: + puts(PACKAGE_STRING); + puts(SYSTEMD_FEATURES); return 0; case 'D': free(arg_directory); arg_directory = canonicalize_file_name(optarg); if (!arg_directory) { - log_error("Failed to canonicalize root directory."); + log_error("Invalid root directory: %m"); return -ENOMEM; } @@ -166,21 +214,9 @@ 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; - } - - 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); + arg_user = strdup(optarg); + if (!arg_user) + return log_oom(); break; @@ -193,14 +229,53 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_UUID: - arg_uuid = optarg; + r = sd_id128_from_string(optarg, &arg_uuid); + if (r < 0) { + log_error("Invalid UUID: %s", optarg); + return r; + } + break; + + case 'S': + arg_slice = strdup(optarg); + if (!arg_slice) + return log_oom(); + + 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 'L': + file_label = strdup(optarg); + if (!file_label) + return log_oom(); + + break; + + case 'Z': + process_label = strdup(optarg); + if (!process_label) + return log_oom(); + break; case ARG_READ_ONLY: arg_read_only = true; break; - case ARG_CAPABILITY: { + case ARG_CAPABILITY: + case ARG_DROP_CAPABILITY: { char *state, *word; size_t length; @@ -219,7 +294,11 @@ static int parse_argv(int argc, char *argv[]) { } free(t); - arg_retain |= 1ULL << (uint64_t) cap; + + if (c == ARG_CAPABILITY) + arg_retain |= 1ULL << (uint64_t) cap; + else + arg_retain &= ~(1ULL << (uint64_t) cap); } break; @@ -245,12 +324,64 @@ 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; + + 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 log_oom(); + + r = strv_extend(x, b); + if (r < 0) + return log_oom(); + + break; + } + + case ARG_SETENV: { + char **n; + + if (!env_assignment_is_valid(optarg)) { + log_error("Environment variable assignment '%s' is not valid.", optarg); + return -EINVAL; + } + + n = strv_env_set(arg_setenv, optarg); + if (!n) + return log_oom(); + + strv_free(arg_setenv); + arg_setenv = n; + break; + } + case '?': return -EINVAL; default: - log_error("Unknown option code %c", c); - return -EINVAL; + assert_not_reached("Unhandled option"); } } @@ -274,7 +405,7 @@ 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 @@ -287,17 +418,13 @@ static int mount_all(const char *dest) { int r = 0; for (k = 0; k < ELEMENTSOF(mount_table); k++) { - char _cleanup_free_ *where = NULL; + _cleanup_free_ char *where = NULL; + _cleanup_free_ char *options = 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) { @@ -313,13 +440,24 @@ static int mount_all(const char *dest) { if (mount_table[k].what && t > 0) continue; - mkdir_p_label(where, 0755); + mkdir_p(where, 0755); + +#ifdef HAVE_SELINUX + if (file_label && (streq_ptr(mount_table[k].what, "tmpfs") || + streq_ptr(mount_table[k].what, "devpts"))) + options = strjoin(mount_table[k].options, ",context=\"", file_label, "\"", NULL); + else +#endif + options = strjoin(mount_table[k].options, NULL); + + if (!options) + return log_oom(); if (mount(mount_table[k].what, where, mount_table[k].type, mount_table[k].flags, - mount_table[k].options) < 0 && + options) < 0 && mount_table[k].fatal) { log_error("mount(%s) failed: %m", where); @@ -332,6 +470,66 @@ static int mount_all(const char *dest) { return r; } +static int mount_binds(const char *dest, char **l, unsigned long flags) { + char **x, **y; + + STRV_FOREACH_PAIR(x, y, l) { + char *where; + struct stat source_st, dest_st; + int r; + + if (stat(*x, &source_st) < 0) { + log_error("failed to stat %s: %m", *x); + return -errno; + } + + where = strappenda(dest, *y); + r = stat(where, &dest_st); + if (r == 0) { + if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) { + log_error("The file types of %s and %s do not match. Refusing bind mount", + *x, where); + return -EINVAL; + } + } else if (errno == ENOENT) { + r = mkdir_parents_label(where, 0755); + if (r < 0) { + log_error("Failed to bind mount %s: %s", *x, strerror(-r)); + return r; + } + } else { + log_error("Failed to bind mount %s: %s", *x, strerror(errno)); + return -errno; + } + /* Create the mount point, but be conservative -- refuse to create block + * and char devices. */ + if (S_ISDIR(source_st.st_mode)) + mkdir_label(where, 0755); + else if (S_ISFIFO(source_st.st_mode)) + mkfifo(where, 0644); + else if (S_ISSOCK(source_st.st_mode)) + mknod(where, 0644 | S_IFSOCK, 0); + else if (S_ISREG(source_st.st_mode)) + touch(where); + else { + log_error("Refusing to create mountpoint for file: %s", *x); + return -ENOTSUP; + } + + 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) { _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL; char *z, *y; @@ -393,7 +591,7 @@ static int setup_timezone(const char *dest) { } static int setup_resolv_conf(const char *dest) { - char *where; + char _cleanup_free_ *where = NULL; assert(dest); @@ -407,16 +605,13 @@ static int setup_resolv_conf(const char *dest) { /* 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) - mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); - - free(where); + copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW); return 0; } static int setup_boot_id(const char *dest) { - char _cleanup_free_ *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; sd_id128_t rnd; char as_uuid[37]; int r; @@ -442,7 +637,7 @@ 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)); return r; @@ -451,8 +646,8 @@ static int setup_boot_id(const char *dest) { 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); return r; @@ -466,63 +661,67 @@ static int copy_devnodes(const char *dest) { "full\0" "random\0" "urandom\0" - "tty\0" - "ptmx\0"; + "tty\0"; const char *d; int r = 0; - mode_t _cleanup_umask_ u; + _cleanup_umask_ mode_t u; assert(dest); u = umask(0000); NULSTR_FOREACH(d, devnodes) { + _cleanup_free_ char *from = NULL, *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_oom(); - - if (r == 0) - r = -ENOMEM; - - break; - } + from = strappend("/dev/", d); + to = strjoin(dest, "/dev/", d, NULL); + if (!from || !to) + return log_oom(); if (stat(from, &st) < 0) { if (errno != ENOENT) { log_error("Failed to stat %s: %m", from); - if (r == 0) - r = -errno; + return -errno; } } 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); - if (r == 0) - r = -EIO; + return -EIO; } else if (mknod(to, st.st_mode, st.st_rdev) < 0) { log_error("mknod(%s) failed: %m", dest); - if (r == 0) - r = -errno; + return -errno; } } 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 _cleanup_free_ *to = NULL; + _cleanup_free_ char *to = NULL; int r; - mode_t _cleanup_umask_ u; + _cleanup_umask_ mode_t u; assert(dest); assert(console); @@ -568,14 +767,17 @@ static int setup_dev_console(const char *dest, const char *console) { } static int setup_kmsg(const char *dest, int kmsg_socket) { - char _cleanup_free_ *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; int r, fd, k; - mode_t _cleanup_umask_ 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); @@ -616,12 +818,6 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { 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; @@ -646,36 +842,19 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { } 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); - } + if (sethostname(arg_machine, strlen(arg_machine)) < 0) + return -errno; - return r; + return 0; } static int setup_journal(const char *directory) { - sd_id128_t machine_id; - char _cleanup_free_ *p = NULL, *b = NULL, *q = NULL, *d = NULL; + sd_id128_t machine_id, this_id; + _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) return log_oom(); @@ -699,6 +878,24 @@ static int setup_journal(const char *directory) { return r; } + r = sd_id128_get_machine(&this_id); + if (r < 0) { + log_error("Failed to retrieve machine ID: %s", strerror(-r)); + return r; + } + + if (sd_id128_equal(machine_id, this_id)) { + log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR, + "Host and machine ids are equal (%s): refusing to link journals", id); + if (arg_link_journal == LINK_AUTO) + return 0; + return + -EEXIST; + } + + if (arg_link_journal == LINK_NO) + return 0; + free(p); p = strappend("/var/log/journal/", id); q = strjoin(directory, "/var/log/journal/", id, NULL); @@ -799,277 +996,150 @@ static int setup_journal(const char *directory) { return 0; } -static int drop_capabilities(void) { - return capability_bounding_set_drop(~arg_retain, false); -} +static int setup_kdbus(const char *dest, const char *path) { + const char *p; -static int is_os_tree(const char *path) { - int r; - char *p; - /* We use /bin/sh as flag file if something is an OS */ + if (!path) + return 0; - if (asprintf(&p, "%s/bin/sh", path) < 0) - return -ENOMEM; + p = strappenda(dest, "/dev/kdbus"); + if (mkdir(p, 0755) < 0) { + log_error("Failed to create kdbus path: %m"); + return -errno; + } - r = access(p, F_OK); - free(p); + if (mount(path, p, "bind", MS_BIND, NULL) < 0) { + log_error("Failed to mount kdbus domain path: %m"); + return -errno; + } - return r < 0 ? 0 : 1; + return 0; } -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); - fd_nonblock(master, 1); +static int drop_capabilities(void) { + return capability_bounding_set_drop(~arg_retain, false); +} - signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC); - if (signal_fd < 0) { - log_error("signalfd(): %m"); - r = -errno; - goto finish; - } +static int register_machine(pid_t pid) { + _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_bus_unref_ sd_bus *bus = NULL; + int r; - ep = epoll_create1(EPOLL_CLOEXEC); - if (ep < 0) { - log_error("Failed to create epoll: %m"); - r = -errno; - goto finish; + r = sd_bus_open_system(&bus); + if (r < 0) { + log_error("Failed to open system bus: %s", strerror(-r)); + return r; } - /* 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; - } + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + "/org/freedesktop/machine1", + "org.freedesktop.machine1.Manager", + "CreateMachine", + &error, + NULL, + "sayssusa(sv)", + arg_machine, + SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), + "nspawn", + "container", + (uint32_t) pid, + strempty(arg_directory), + !isempty(arg_slice), "Slice", "s", arg_slice); + if (r < 0) { + log_error("Failed to register machine: %s", bus_error_message(&error, r)); + return r; } - zero(stdout_ev); - stdout_ev.events = EPOLLOUT|EPOLLET; - stdout_ev.data.fd = STDOUT_FILENO; - - zero(master_ev); - master_ev.events = EPOLLIN|EPOLLOUT|EPOLLET; - master_ev.data.fd = master; + return 0; +} - zero(signal_ev); - signal_ev.events = EPOLLIN; - signal_ev.data.fd = signal_fd; +static int terminate_machine(pid_t pid) { + _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; + _cleanup_bus_unref_ sd_bus *bus = NULL; + const char *path; + int r; - if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 || - 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 register fds in epoll: %m"); - r = -errno; - goto finish; + r = sd_bus_default_system(&bus); + if (r < 0) { + log_error("Failed to open system bus: %s", strerror(-r)); + return r; } - for (;;) { - struct epoll_event ev[16]; - ssize_t k; - int i, nfds; - - nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1); - if (nfds < 0) { - - if (errno == EINTR || errno == EAGAIN) - continue; - - log_error("epoll_wait(): %m"); - r = -errno; - goto finish; - } - - assert(nfds >= 1); - - for (i = 0; i < nfds; i++) { - if (ev[i].data.fd == STDIN_FILENO) { - - if (ev[i].events & (EPOLLIN|EPOLLHUP)) - stdin_readable = true; - - } else if (ev[i].data.fd == STDOUT_FILENO) { - - if (ev[i].events & (EPOLLOUT|EPOLLHUP)) - stdout_writable = true; - - } else if (ev[i].data.fd == master) { - - if (ev[i].events & (EPOLLIN|EPOLLHUP)) - master_readable = true; - - if (ev[i].events & (EPOLLOUT|EPOLLHUP)) - master_writable = true; - - } else if (ev[i].data.fd == signal_fd) { - struct signalfd_siginfo sfsi; - ssize_t n; - - n = read(signal_fd, &sfsi, sizeof(sfsi)); - if (n != sizeof(sfsi)) { - - if (n >= 0) { - log_error("Failed to read from signalfd: invalid block size"); - r = -EIO; - goto finish; - } - - if (errno != EINTR && errno != EAGAIN) { - log_error("Failed to read from signalfd: %m"); - r = -errno; - goto finish; - } - } else { - - if (sfsi.ssi_signo == SIGWINCH) { - struct winsize ws; - - /* 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; - } - } - } - } - - while ((stdin_readable && in_buffer_full <= 0) || - (master_writable && in_buffer_full > 0) || - (master_readable && out_buffer_full <= 0) || - (stdout_writable && out_buffer_full > 0)) { - - if (stdin_readable && in_buffer_full < LINE_MAX) { - - 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; - else { - log_error("read(): %m"); - r = -errno; - goto finish; - } - } else - in_buffer_full += (size_t) k; - } - - if (master_writable && 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; - else { - log_error("write(): %m"); - r = -errno; - goto finish; - } - - } else { - assert(in_buffer_full >= (size_t) k); - memmove(in_buffer, in_buffer + k, in_buffer_full - k); - in_buffer_full -= k; - } - } - - if (master_readable && out_buffer_full < LINE_MAX) { - - 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; - else { - log_error("read(): %m"); - r = -errno; - goto finish; - } - } else - out_buffer_full += (size_t) k; - } - - if (stdout_writable && out_buffer_full > 0) { - - k = write(STDOUT_FILENO, out_buffer, out_buffer_full); - if (k < 0) { + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + "/org/freedesktop/machine1", + "org.freedesktop.machine1.Manager", + "GetMachineByPID", + &error, + &reply, + "u", + (uint32_t) pid); + if (r < 0) { + /* Note that the machine might already have been + * cleaned up automatically, hence don't consider it a + * failure if we cannot get the machine object. */ + log_debug("Failed to get machine: %s", bus_error_message(&error, r)); + return 0; + } - if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO) - stdout_writable = false; - else { - log_error("write(): %m"); - r = -errno; - goto finish; - } - - } else { - assert(out_buffer_full >= (size_t) k); - memmove(out_buffer, out_buffer + k, out_buffer_full - k); - out_buffer_full -= k; - } - } - } + r = sd_bus_message_read(reply, "o", &path); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_call_method( + bus, + "org.freedesktop.machine1", + path, + "org.freedesktop.machine1.Machine", + "Terminate", + &error, + NULL, + NULL); + if (r < 0) { + log_debug("Failed to terminate machine: %s", bus_error_message(&error, r)); + return 0; } -finish: - if (ep >= 0) - close_nointr_nofail(ep); + return 0; +} - if (signal_fd >= 0) - close_nointr_nofail(signal_fd); +static bool audit_enabled(void) { + int fd; - return r; + 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, n_fd_passed; + _cleanup_close_ int master = -1, kdbus_fd = -1, sync_fd = -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; + _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 }; + _cleanup_fdset_free_ FDSet *fds = NULL; + _cleanup_free_ char *kdbus_domain = NULL; + const char *ns; 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; @@ -1081,12 +1151,26 @@ 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(basename(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; @@ -1097,13 +1181,20 @@ int main(int argc, char *argv[]) { goto finish; } + if (arg_boot && 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; } @@ -1119,29 +1210,6 @@ int main(int argc, char *argv[]) { fdset_close_others(fds); log_open(); - 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; - } - - if (asprintf(&newcg, "%s/nspawn-%lu", oldcg, (unsigned long) getpid()) < 0) { - log_error("Failed to allocate cgroup path."); - 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; - } - - 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)); - } - master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); if (master < 0) { log_error("Failed to acquire pseudo tty: %m"); @@ -1154,29 +1222,27 @@ int main(int argc, char *argv[]) { goto finish; } - log_info("Spawning namespace container on %s (console is %s).", arg_directory, console); - - if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0) - ioctl(master, TIOCSWINSZ, &ws); + log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_directory); if (unlockpt(master) < 0) { log_error("Failed to unlock tty: %m"); goto finish; } - if (tcgetattr(STDIN_FILENO, &saved_attr) >= 0) { - saved_attr_valid = true; - - raw_attr = saved_attr; - cfmakeraw(&raw_attr); - raw_attr.c_lflag &= ~ECHO; - } + ns = strappenda("machine-", arg_machine); + kdbus_fd = bus_kernel_create_domain(ns, &kdbus_domain); + if (r < 0) + log_debug("Failed to create kdbus domain: %s", strerror(-r)); + else + log_debug("Successfully created kdbus domain as %s", kdbus_domain); 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: %m"); 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); @@ -1184,11 +1250,10 @@ int main(int argc, char *argv[]) { for (;;) { siginfo_t status; - if (saved_attr_valid) { - if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { - log_error("Failed to set terminal attributes: %m"); - goto finish; - } + sync_fd = eventfd(0, EFD_CLOEXEC); + if (sync_fd < 0) { + log_error("Failed to create event fd: %m"); + goto finish; } pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL); @@ -1203,13 +1268,12 @@ int main(int argc, char *argv[]) { if (pid == 0) { /* child */ - const char *home = NULL; uid_t uid = (uid_t) -1; gid_t gid = (gid_t) -1; - unsigned n_env = 0; + unsigned n_env = 2; const char *envp[] = { - "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", + "PATH=" DEFAULT_PATH_SPLIT_USR, "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ NULL, /* TERM */ NULL, /* HOME */ @@ -1220,9 +1284,12 @@ int main(int argc, char *argv[]) { NULL, /* LISTEN_PID */ NULL }; + char **env_use; + eventfd_t x; - envp[2] = strv_find_prefix(environ, "TERM="); - n_env = 3; + envp[n_env] = strv_find_prefix(environ, "TERM="); + if (envp[n_env]) + n_env ++; close_nointr_nofail(master); master = -1; @@ -1292,6 +1359,9 @@ int main(int argc, char *argv[]) { if (copy_devnodes(arg_directory) < 0) goto child_fail; + if (setup_ptmx(arg_directory) < 0) + goto child_fail; + dev_setup(arg_directory); if (setup_dev_console(arg_directory, console) < 0) @@ -1315,6 +1385,15 @@ int main(int argc, char *argv[]) { if (setup_journal(arg_directory) < 0) goto child_fail; + if (mount_binds(arg_directory, arg_bind, 0) < 0) + goto child_fail; + + if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) + goto child_fail; + + if (setup_kdbus(arg_directory, kdbus_domain) < 0) + goto child_fail; + if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); goto child_fail; @@ -1407,8 +1486,8 @@ int main(int argc, char *argv[]) { goto child_fail; } - if (arg_uuid) { - if (asprintf((char**)(envp + n_env++), "container_uuid=%s", arg_uuid) < 0) { + 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) { log_oom(); goto child_fail; } @@ -1422,7 +1501,7 @@ int main(int argc, char *argv[]) { } if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) || - (asprintf((char **)(envp + n_env++), "LISTEN_PID=%lu", (unsigned long) getpid()) < 0)) { + (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) { log_oom(); goto child_fail; } @@ -1430,6 +1509,28 @@ int main(int argc, char *argv[]) { setup_hostname(); + eventfd_read(sync_fd, &x); + close_nointr_nofail(sync_fd); + sync_fd = -1; + + if (!strv_isempty(arg_setenv)) { + char **n; + + n = strv_env_merge(2, envp, arg_setenv); + if (!n) { + log_oom(); + goto child_fail; + } + + env_use = n; + } else + env_use = (char**) envp; + +#if HAVE_SELINUX + if (process_label) + if (setexeccon(process_label) < 0) + log_error("setexeccon(\"%s\") failed: %m", process_label); +#endif if (arg_boot) { char **a; size_t l; @@ -1441,18 +1542,18 @@ int main(int argc, char *argv[]) { memcpy(a + 1, argv + optind, l * sizeof(char*)); a[0] = (char*) "/usr/lib/systemd/systemd"; - execve(a[0], a, (char**) envp); + execve(a[0], a, env_use); a[0] = (char*) "/lib/systemd/systemd"; - execve(a[0], a, (char**) envp); + execve(a[0], a, env_use); a[0] = (char*) "/sbin/init"; - execve(a[0], a, (char**) envp); + execve(a[0], a, env_use); } else if (argc > optind) - execvpe(argv[optind], argv + optind, (char**) envp); + execvpe(argv[optind], argv + optind, env_use); else { chdir(home ? home : "/root"); - execle("/bin/bash", "-bash", NULL, (char**) envp); + execle("/bin/bash", "-bash", NULL, env_use); } log_error("execv() failed: %m"); @@ -1464,70 +1565,74 @@ int main(int argc, char *argv[]) { fdset_free(fds); fds = NULL; - if (process_pty(master, pid, &mask) < 0) + r = register_machine(pid); + if (r < 0) goto finish; - if (saved_attr_valid) - tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); + eventfd_write(sync_fd, 1); + close_nointr_nofail(sync_fd); + sync_fd = -1; - r = wait_for_terminate(pid, &status); - if (r < 0) { + k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); + if (k < 0) { + r = EXIT_FAILURE; + break; + } + + putc('\n', stdout); + + /* Kill if it is not dead yet anyway */ + terminate_machine(pid); + + /* Redundant, but better safe than sorry */ + kill(pid, SIGKILL); + + k = wait_for_terminate(pid, &status); + pid = 0; + + 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); - r = status.si_status; + log_error("Container %s failed with error code %i.", arg_machine, status.si_status); break; } - log_debug("Container exited successfully."); + log_debug("Container %s exited successfully.", arg_machine); break; } else if (status.si_code == CLD_KILLED && status.si_status == SIGINT) { - log_info("Container has been shut down."); + log_info("Container %s has been shut down.", arg_machine); r = 0; break; } else if (status.si_code == CLD_KILLED && status.si_status == SIGHUP) { - log_info("Container is being rebooted."); + 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 terminated by signal %s.", signal_to_string(status.si_status)); + log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status)); r = EXIT_FAILURE; break; } else { - log_error("Container failed due to unknown reason."); + log_error("Container %s failed due to unknown reason.", arg_machine); 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); + if (pid > 0) + kill(pid, SIGKILL); free(arg_directory); - strv_free(arg_controllers); - free(oldcg); - free(newcg); - - fdset_free(fds); + free(arg_machine); + free(arg_setenv); return r; }