#include "build.h"
#include "fileio.h"
+#ifndef TTY_GID
+#define TTY_GID 5
+#endif
+
typedef enum LinkJournal {
LINK_NO,
LINK_AUTO,
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;
(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"
- " --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 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;
ARG_UUID,
ARG_READ_ONLY,
ARG_CAPABILITY,
- ARG_LINK_JOURNAL
+ ARG_LINK_JOURNAL,
+ ARG_BIND,
+ ARG_BIND_RO
};
static const struct option options[] = {
{ "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 }
};
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:
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;
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;
{ 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
char _cleanup_free_ *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) {
if (mount_table[k].what && t > 0)
continue;
- mkdir_p_label(where, 0755);
+ mkdir_p(where, 0755);
if (mount(mount_table[k].what,
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) {
_cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
char *z, *y;
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;
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;
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) {
return 0;
}
-static int drop_capabilities(void) {
- return capability_bounding_set_drop(~arg_retain, false);
-}
-
-static int is_os_tree(const char *path) {
+static int setup_cgroup(const char *path) {
+ char **c;
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 = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, path, 1);
+ if (r < 0) {
+ log_error("Failed to create cgroup: %s", strerror(-r));
+ return r;
+ }
- r = access(p, F_OK);
- free(p);
+ 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 r < 0 ? 0 : 1;
+ return 0;
+}
+
+static int drop_capabilities(void) {
+ return capability_bounding_set_drop(~arg_retain, false);
}
static int process_pty(int master, pid_t pid, sigset_t *mask) {
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_free_ char *machine_root = NULL, *newcg = NULL;
+ _cleanup_close_ int master = -1;
+ int n_fd_passed;
const char *console = NULL;
struct termios saved_attr, raw_attr;
sigset_t mask;
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);
+ 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 (is_os_tree(arg_directory) <= 0) {
+ if (path_is_os_tree(arg_directory) <= 0) {
log_error("Directory %s doesn't look like an OS root directory. Refusing.", arg_directory);
goto finish;
}
fdset_close_others(fds);
log_open();
- k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg);
+ k = cg_get_machine_path(&machine_root);
if (k < 0) {
- log_error("Failed to determine current cgroup: %s", strerror(-k));
+ log_error("Failed to determine machine cgroup path: %s", strerror(-k));
goto finish;
}
- if (asprintf(&newcg, "%s/nspawn-%lu", oldcg, (unsigned long) getpid()) < 0) {
+ newcg = strjoin(machine_root, "/", arg_machine, NULL);
+ if (!newcg) {
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;
- }
+ r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, newcg, false);
+ if (r <= 0 && r != -ENOENT) {
+ log_error("Container already running.");
- 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));
+ free(newcg);
+ newcg = NULL;
+
+ goto finish;
}
master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
}
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;
}
siginfo_t status;
int pipefd[2];
- if(pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
+ if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
log_error("pipe2(): %m");
goto finish;
}
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",
"container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
NULL
};
- 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(pipefd[1]);
fd_wait_for_event(pipefd[0], POLLHUP, -1);
goto child_fail;
}
+ if (setup_cgroup(newcg) < 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 (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)
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 (chdir(arg_directory) < 0) {
log_error("chdir(%s) failed: %m", arg_directory);
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=%lu", (unsigned long) 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]);
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);