#include <linux/fs.h>
#include <sys/un.h>
#include <sys/socket.h>
+#include <linux/netlink.h>
#include <systemd/sd-daemon.h>
+#include <systemd/sd-bus.h>
#include "log.h"
#include "util.h"
#include "fdset.h"
#include "build.h"
#include "fileio.h"
+#include "bus-util.h"
+
+#ifndef TTY_GID
+#define TTY_GID 5
+#endif
typedef enum LinkJournal {
LINK_NO,
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 const char *arg_slice = NULL;
static bool arg_private_network = false;
static bool arg_read_only = false;
static bool arg_boot = false;
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"
- " --version Print version string\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"
+ " -S --slice=SLICE Place the container in the specified slice\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"
{ "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 },
{ "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' },
+ { "slice", required_argument, NULL, 'S' },
{ NULL, 0, NULL, 0 }
};
- 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:bM:jS:", options, NULL)) >= 0) {
switch (c) {
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;
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);
+ 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:
_cleanup_free_ char *a = NULL, *b = NULL;
char *e;
char ***x;
- int r;
x = c == ARG_BIND ? &arg_bind : &arg_bind_ro;
{ 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
int r = 0;
for (k = 0; k < ELEMENTSOF(mount_table); k++) {
- char _cleanup_free_ *where = NULL;
+ _cleanup_free_ char *where = NULL;
int t;
where = strjoin(dest, "/", mount_table[k].where, NULL);
STRV_FOREACH_PAIR(x, y, l) {
_cleanup_free_ char *where = NULL;
+ struct stat source_st, dest_st;
+
+ if (stat(*x, &source_st) < 0) {
+ log_error("failed to stat %s: %m", *x);
+ return -errno;
+ }
where = strjoin(dest, "/", *y, NULL);
if (!where)
return log_oom();
- mkdir_p_label(where, 0755);
+ if (stat(where, &dest_st) == 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 {
+ /* Create the mount point, but be conservative -- refuse to create block
+ * and char devices. */
+ if (S_ISDIR(source_st.st_mode))
+ mkdir_p_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);
}
static int setup_resolv_conf(const char *dest) {
- char *where;
+ char _cleanup_free_ *where = NULL;
assert(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;
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;
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;
"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);
NULSTR_FOREACH(d, devnodes) {
struct stat st;
- char _cleanup_free_ *from = NULL, *to = NULL;
+ _cleanup_free_ char *from = NULL, *to = NULL;
asprintf(&from, "/dev/%s", d);
asprintf(&to, "%s/dev/%s", dest, d);
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);
}
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);
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;
}
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;
+ _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL;
char *id;
int r;
return capability_bounding_set_drop(~arg_retain, false);
}
-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 (asprintf(&p, "%s/bin/sh", path) < 0)
- return -ENOMEM;
-
- r = access(p, F_OK);
- free(p);
-
- return r < 0 ? 0 : 1;
-}
-
static int process_pty(int master, pid_t pid, sigset_t *mask) {
char in_buffer[LINE_MAX], out_buffer[LINE_MAX];
return r;
}
+static int register_machine(void) {
+ _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
+ _cleanup_bus_unref_ sd_bus *bus = NULL;
+ int r;
+
+ r = sd_bus_open_system(&bus);
+ if (r < 0) {
+ log_error("Failed to open system bus: %s", strerror(-r));
+ return r;
+ }
+
+ 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) 0,
+ strempty(arg_directory),
+ 1, "Slice", "s", strempty(arg_slice));
+ if (r < 0) {
+ log_error("Failed to register machine: %s", error.message ? error.message : strerror(-r));
+ return r;
+ }
+
+ return 0;
+}
+
+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, n_fd_passed;
+ _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;
+ _cleanup_fdset_free_ 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;
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;
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;
}
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");
}
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);
for (;;) {
siginfo_t status;
- int pipefd[2];
+ int pipefd[2], pipefd2[2];
- if(pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
+ if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
log_error("pipe2(): %m");
goto finish;
}
+ if (pipe2(pipefd2, O_NONBLOCK|O_CLOEXEC) < 0) {
+ log_error("pipe2(): %m");
+ close_pipe(pipefd);
+ goto finish;
+ }
+
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)
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]);
goto child_fail;
}
+ close_pipe(pipefd2);
+
+ r = register_machine();
+ if (r < 0)
+ goto finish;
+
/* Mark everything as slave, so that we still
* receive mounts from the real root, but don't
* propagate mounts to the real root. */
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)
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;
}
}
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;
}
_exit(EXIT_FAILURE);
}
- log_info("Init process in the container running as PID %d", pid);
+ log_info("Init process in the container running as PID %lu.", (unsigned long) pid);
close_nointr_nofail(pipefd[0]);
close_nointr_nofail(pipefd[1]);
+ /* 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]);
+
fdset_free(fds);
fds = NULL;
if (saved_attr_valid)
tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
- r = wait_for_terminate(pid, &status);
- if (r < 0) {
+ 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);
- r = status.si_status;
break;
}
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);
return r;
}
~ianmdlvl / elogind.git / blobdiff