-/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
-
/***
This file is part of systemd.
#include "copy.h"
#include "dev-setup.h"
#include "env-util.h"
-#include "event-util.h"
#include "fd-util.h"
#include "fdset.h"
#include "fileio.h"
#include "hostname-util.h"
#include "log.h"
#include "loopback-setup.h"
+#include "machine-id-setup.h"
#include "machine-image.h"
#include "macro.h"
#include "missing.h"
#include "nspawn-register.h"
#include "nspawn-settings.h"
#include "nspawn-setuid.h"
+#include "nspawn-stub-pid1.h"
#include "parse-util.h"
#include "path-util.h"
#include "process-util.h"
#ifdef HAVE_SECCOMP
#include "seccomp-util.h"
#endif
+#include "selinux-util.h"
#include "signal-util.h"
#include "socket-util.h"
#include "stat-util.h"
static char *arg_directory = NULL;
static char *arg_template = NULL;
+static char *arg_chdir = NULL;
static char *arg_user = 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;
+static StartMode arg_start_mode = START_PID1;
static bool arg_ephemeral = false;
static LinkJournal arg_link_journal = LINK_AUTO;
static bool arg_link_journal_try = false;
" -x --ephemeral Run container with snapshot of root directory, and\n"
" remove it after exit\n"
" -i --image=PATH File system device or disk image for the container\n"
+ " -a --as-pid2 Maintain a stub init as PID1, invoke binary as PID2\n"
" -b --boot Boot up full system (i.e. invoke init)\n"
+ " --chdir=PATH Set working directory in the container\n"
" -u --user=USER Run the command under specified user or uid\n"
" -M --machine=NAME Set the machine name for the container\n"
" --uuid=UUID Set a specific machine UUID for the container\n"
" capability\n"
" --drop-capability=CAP Drop the specified capability from the default set\n"
" --kill-signal=SIGNAL Select signal to use for shutting down PID 1\n"
- " --link-journal=MODE Link up guest journal, one of no, auto, guest, host,\n"
- " try-guest, try-host\n"
+ " --link-journal=MODE Link up guest journal, one of no, auto, guest, \n"
+ " host, try-guest, try-host\n"
" -j Equivalent to --link-journal=try-guest\n"
" --read-only Mount the root directory read-only\n"
" --bind=PATH[:PATH[:OPTIONS]]\n"
" the container\n"
" --overlay-ro=PATH[:PATH...]:PATH\n"
" Similar, but creates a read-only overlay mount\n"
- " --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
+ " -E --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
" --share-system Share system namespaces with host\n"
" --register=BOOLEAN Register container as machine\n"
" --keep-unit Do not register a scope for the machine, reuse\n"
ARG_TMPFS,
ARG_OVERLAY,
ARG_OVERLAY_RO,
- ARG_SETENV,
ARG_SHARE_SYSTEM,
ARG_REGISTER,
ARG_KEEP_UNIT,
ARG_PRIVATE_USERS,
ARG_KILL_SIGNAL,
ARG_SETTINGS,
+ ARG_CHDIR,
};
static const struct option options[] = {
{ "ephemeral", no_argument, NULL, 'x' },
{ "user", required_argument, NULL, 'u' },
{ "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
+ { "as-pid2", no_argument, NULL, 'a' },
{ "boot", no_argument, NULL, 'b' },
{ "uuid", required_argument, NULL, ARG_UUID },
{ "read-only", no_argument, NULL, ARG_READ_ONLY },
{ "overlay-ro", required_argument, NULL, ARG_OVERLAY_RO },
{ "machine", required_argument, NULL, 'M' },
{ "slice", required_argument, NULL, 'S' },
- { "setenv", required_argument, NULL, ARG_SETENV },
+ { "setenv", required_argument, NULL, 'E' },
{ "selinux-context", required_argument, NULL, 'Z' },
{ "selinux-apifs-context", required_argument, NULL, 'L' },
{ "quiet", no_argument, NULL, 'q' },
{ "private-users", optional_argument, NULL, ARG_PRIVATE_USERS },
{ "kill-signal", required_argument, NULL, ARG_KILL_SIGNAL },
{ "settings", required_argument, NULL, ARG_SETTINGS },
+ { "chdir", required_argument, NULL, ARG_CHDIR },
{}
};
assert(argc >= 0);
assert(argv);
- while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:xp:n", options, NULL)) >= 0)
+ while ((c = getopt_long(argc, argv, "+hD:u:abL:M:jS:Z:qi:xp:n", options, NULL)) >= 0)
switch (c) {
break;
case 'b':
- arg_boot = true;
- arg_settings_mask |= SETTING_BOOT;
+ if (arg_start_mode == START_PID2) {
+ log_error("--boot and --as-pid2 may not be combined.");
+ return -EINVAL;
+ }
+
+ arg_start_mode = START_BOOT;
+ arg_settings_mask |= SETTING_START_MODE;
+ break;
+
+ case 'a':
+ if (arg_start_mode == START_BOOT) {
+ log_error("--boot and --as-pid2 may not be combined.");
+ return -EINVAL;
+ }
+
+ arg_start_mode = START_PID2;
+ arg_settings_mask |= SETTING_START_MODE;
break;
case ARG_UUID:
case ARG_CAPABILITY:
case ARG_DROP_CAPABILITY: {
p = optarg;
- for(;;) {
+ for (;;) {
_cleanup_free_ char *t = NULL;
r = extract_first_word(&p, &t, ",", 0);
break;
}
- case ARG_SETENV: {
+ case 'E': {
char **n;
if (!env_assignment_is_valid(optarg)) {
break;
+ case ARG_CHDIR:
+ if (!path_is_absolute(optarg)) {
+ log_error("Working directory %s is not an absolute path.", optarg);
+ return -EINVAL;
+ }
+
+ r = free_and_strdup(&arg_chdir, optarg);
+ if (r < 0)
+ return log_oom();
+
+ arg_settings_mask |= SETTING_WORKING_DIRECTORY;
+ break;
+
case '?':
return -EINVAL;
if (arg_share_system)
arg_register = false;
- if (arg_boot && arg_share_system) {
+ if (arg_start_mode != START_PID1 && arg_share_system) {
log_error("--boot and --share-system may not be combined.");
return -EINVAL;
}
if (!arg_parameters)
return log_oom();
- arg_settings_mask |= SETTING_BOOT;
+ arg_settings_mask |= SETTING_START_MODE;
}
/* Load all settings from .nspawn files */
return -EINVAL;
}
- if (arg_boot && arg_kill_signal <= 0)
+#ifndef HAVE_LIBIPTC
+ if (arg_expose_ports) {
+ log_error("--port= is not supported, compiled without libiptc support.");
+ return -EOPNOTSUPP;
+ }
+#endif
+
+ if (arg_start_mode == START_BOOT && arg_kill_signal <= 0)
arg_kill_signal = SIGRTMIN+3;
return 0;
}
check = strjoina("/usr/share/zoneinfo/", z);
- check = prefix_root(dest, check);
+ check = prefix_roota(dest, check);
if (laccess(check, F_OK) < 0) {
log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
return 0;
}
static int setup_journal(const char *directory) {
- sd_id128_t machine_id, this_id;
+ sd_id128_t this_id;
_cleanup_free_ char *b = NULL, *d = NULL;
- const char *etc_machine_id, *p, *q;
- char *id;
+ const char *p, *q;
+ bool try;
+ char id[33];
int r;
/* Don't link journals in ephemeral mode */
if (arg_ephemeral)
return 0;
- etc_machine_id = prefix_roota(directory, "/etc/machine-id");
-
- r = read_one_line_file(etc_machine_id, &b);
- if (r == -ENOENT && arg_link_journal == LINK_AUTO)
- return 0;
- else if (r < 0)
- return log_error_errno(r, "Failed to read machine ID from %s: %m", etc_machine_id);
-
- id = strstrip(b);
- if (isempty(id) && arg_link_journal == LINK_AUTO)
+ if (arg_link_journal == LINK_NO)
return 0;
- /* Verify validity */
- r = sd_id128_from_string(id, &machine_id);
- if (r < 0)
- return log_error_errno(r, "Failed to parse machine ID from %s: %m", etc_machine_id);
+ try = arg_link_journal_try || arg_link_journal == LINK_AUTO;
r = sd_id128_get_machine(&this_id);
if (r < 0)
return log_error_errno(r, "Failed to retrieve machine ID: %m");
- if (sd_id128_equal(machine_id, this_id)) {
- log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR,
+ if (sd_id128_equal(arg_uuid, this_id)) {
+ log_full(try ? LOG_WARNING : LOG_ERR,
"Host and machine ids are equal (%s): refusing to link journals", id);
- if (arg_link_journal == LINK_AUTO)
+ if (try)
return 0;
return -EEXIST;
}
- if (arg_link_journal == LINK_NO)
- return 0;
-
r = userns_mkdir(directory, "/var", 0755, 0, 0);
if (r < 0)
return log_error_errno(r, "Failed to create /var: %m");
if (r < 0)
return log_error_errno(r, "Failed to create /var/log/journal: %m");
+ (void) sd_id128_to_string(arg_uuid, id);
+
p = strjoina("/var/log/journal/", id);
q = prefix_roota(directory, p);
if (path_is_mount_point(p, 0) > 0) {
- if (arg_link_journal != LINK_AUTO) {
- log_error("%s: already a mount point, refusing to use for journal", p);
- return -EEXIST;
- }
+ if (try)
+ return 0;
- return 0;
+ log_error("%s: already a mount point, refusing to use for journal", p);
+ return -EEXIST;
}
if (path_is_mount_point(q, 0) > 0) {
- if (arg_link_journal != LINK_AUTO) {
- log_error("%s: already a mount point, refusing to use for journal", q);
- return -EEXIST;
- }
+ if (try)
+ return 0;
- return 0;
+ log_error("%s: already a mount point, refusing to use for journal", q);
+ return -EEXIST;
}
r = readlink_and_make_absolute(p, &d);
if (arg_link_journal == LINK_GUEST) {
if (symlink(q, p) < 0) {
- if (arg_link_journal_try) {
+ if (try) {
log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
return 0;
} else
}
if (arg_link_journal == LINK_HOST) {
- /* don't create parents here -- if the host doesn't have
+ /* don't create parents here — if the host doesn't have
* permanent journal set up, don't force it here */
- r = mkdir(p, 0755);
- if (r < 0) {
- if (arg_link_journal_try) {
+
+ if (mkdir(p, 0755) < 0 && errno != EEXIST) {
+ if (try) {
log_debug_errno(errno, "Failed to create %s, skipping journal setup: %m", p);
return 0;
} else
}
static int drop_capabilities(void) {
- return capability_bounding_set_drop(~arg_retain, false);
+ return capability_bounding_set_drop(arg_retain, false);
}
static int reset_audit_loginuid(void) {
#endif
}
+static int setup_machine_id(const char *directory) {
+ int r;
+ const char *etc_machine_id, *t;
+ _cleanup_free_ char *s = NULL;
+
+ etc_machine_id = prefix_roota(directory, "/etc/machine-id");
+
+ r = read_one_line_file(etc_machine_id, &s);
+ if (r < 0)
+ return log_error_errno(r, "Failed to read machine ID from %s: %m", etc_machine_id);
+
+ t = strstrip(s);
+
+ if (!isempty(t)) {
+ r = sd_id128_from_string(t, &arg_uuid);
+ if (r < 0)
+ return log_error_errno(r, "Failed to parse machine ID from %s: %m", etc_machine_id);
+ } else {
+ if (sd_id128_is_null(arg_uuid)) {
+ r = sd_id128_randomize(&arg_uuid);
+ if (r < 0)
+ return log_error_errno(r, "Failed to generate random machine ID: %m");
+ }
+ }
+
+ r = machine_id_setup(directory, arg_uuid);
+ if (r < 0)
+ return log_error_errno(r, "Failed to setup machine ID: %m");
+
+ return 0;
+}
+
static int mount_devices(
const char *where,
const char *root_device, bool root_device_rw,
FDSet *fds) {
_cleanup_free_ char *home = NULL;
+ char as_uuid[37];
unsigned n_env = 1;
const char *envp[] = {
"PATH=" DEFAULT_PATH_SPLIT_USR,
envp[n_env] = strv_find_prefix(environ, "TERM=");
if (envp[n_env])
- n_env ++;
+ n_env++;
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))
return log_oom();
- if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) {
- char as_uuid[37];
+ assert(!sd_id128_equal(arg_uuid, SD_ID128_NULL));
- if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0)
- return log_oom();
- }
+ if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0)
+ return log_oom();
if (fdset_size(fds) > 0) {
r = fdset_cloexec(fds, false);
return -ESRCH;
}
+ if (arg_chdir)
+ if (chdir(arg_chdir) < 0)
+ return log_error_errno(errno, "Failed to change to specified working directory %s: %m", arg_chdir);
+
+ if (arg_start_mode == START_PID2) {
+ r = stub_pid1();
+ if (r < 0)
+ return r;
+ }
+
/* Now, explicitly close the log, so that we
* then can close all remaining fds. Closing
* the log explicitly first has the benefit
log_close();
(void) fdset_close_others(fds);
- if (arg_boot) {
+ if (arg_start_mode == START_BOOT) {
char **a;
size_t m;
/* Automatically search for the init system */
- m = 1 + strv_length(arg_parameters);
- a = newa(char*, m + 1);
- if (strv_isempty(arg_parameters))
- a[1] = NULL;
- else
- memcpy(a + 1, arg_parameters, m * sizeof(char*));
+ m = strv_length(arg_parameters);
+ a = newa(char*, m + 2);
+ memcpy_safe(a + 1, arg_parameters, m * sizeof(char*));
+ a[1 + m] = NULL;
a[0] = (char*) "/usr/lib/systemd/systemd";
execve(a[0], a, env_use);
} else if (!strv_isempty(arg_parameters))
execvpe(arg_parameters[0], arg_parameters, env_use);
else {
- chdir(home ?: "/root");
+ if (!arg_chdir)
+ /* If we cannot change the directory, we'll end up in /, that is expected. */
+ (void) chdir(home ?: "/root");
+
execle("/bin/bash", "-bash", NULL, env_use);
execle("/bin/sh", "-sh", NULL, env_use);
}
bool interactive,
bool secondary,
int pid_socket,
+ int uuid_socket,
int kmsg_socket,
int rtnl_socket,
int uid_shift_socket,
assert(directory);
assert(console);
assert(pid_socket >= 0);
+ assert(uuid_socket >= 0);
assert(kmsg_socket >= 0);
cg_unified_flush();
if (r < 0)
return r;
+ r = setup_machine_id(directory);
+ if (r < 0)
+ return r;
+
r = setup_journal(directory);
if (r < 0)
return r;
return log_error_errno(errno, "Failed to fork inner child: %m");
if (pid == 0) {
pid_socket = safe_close(pid_socket);
+ uuid_socket = safe_close(uuid_socket);
uid_shift_socket = safe_close(uid_shift_socket);
/* The inner child has all namespaces that are
return -EIO;
}
+ l = send(uuid_socket, &arg_uuid, sizeof(arg_uuid), MSG_NOSIGNAL);
+ if (l < 0)
+ return log_error_errno(errno, "Failed to send machine ID: %m");
+ if (l != sizeof(arg_uuid)) {
+ log_error("Short write while sending machine ID.");
+ return -EIO;
+ }
+
pid_socket = safe_close(pid_socket);
+ uuid_socket = safe_close(uuid_socket);
kmsg_socket = safe_close(kmsg_socket);
rtnl_socket = safe_close(rtnl_socket);
/* Copy over bits from the settings, unless they have been
* explicitly masked by command line switches. */
- if ((arg_settings_mask & SETTING_BOOT) == 0 &&
- settings->boot >= 0) {
- arg_boot = settings->boot;
+ if ((arg_settings_mask & SETTING_START_MODE) == 0 &&
+ settings->start_mode >= 0) {
+ arg_start_mode = settings->start_mode;
strv_free(arg_parameters);
arg_parameters = settings->parameters;
settings->parameters = NULL;
}
+ if ((arg_settings_mask & SETTING_WORKING_DIRECTORY) == 0 &&
+ settings->working_directory) {
+ free(arg_chdir);
+ arg_chdir = settings->working_directory;
+ settings->working_directory = NULL;
+ }
+
if ((arg_settings_mask & SETTING_ENVIRONMENT) == 0 &&
settings->environment) {
strv_free(arg_setenv);
log_parse_environment();
log_open();
+ /* Make sure rename_process() in the stub init process can work */
+ saved_argv = argv;
+ saved_argc = argc;
+
r = parse_argv(argc, argv);
if (r <= 0)
goto finish;
}
}
- if (arg_boot) {
+ if (arg_start_mode == START_BOOT) {
if (path_is_os_tree(arg_directory) <= 0) {
log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory);
r = -EINVAL;
goto finish;
}
+ if (arg_selinux_apifs_context) {
+ r = mac_selinux_apply(console, arg_selinux_apifs_context);
+ if (r < 0)
+ goto finish;
+ }
+
if (unlockpt(master) < 0) {
r = log_error_errno(errno, "Failed to unlock tty: %m");
goto finish;
}
for (;;) {
- _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }, rtnl_socket_pair[2] = { -1, -1 }, pid_socket_pair[2] = { -1, -1 }, uid_shift_socket_pair[2] = { -1, -1 };
+ _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }, rtnl_socket_pair[2] = { -1, -1 },
+ pid_socket_pair[2] = { -1, -1 }, uuid_socket_pair[2] = { -1, -1 }, uid_shift_socket_pair[2] = { -1, -1 };
ContainerStatus container_status;
_cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
static const struct sigaction sa = {
};
int ifi = 0;
ssize_t l;
- _cleanup_event_unref_ sd_event *event = NULL;
+ _cleanup_(sd_event_unrefp) sd_event *event = NULL;
_cleanup_(pty_forward_freep) PTYForward *forward = NULL;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
+ _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL;
char last_char = 0;
r = barrier_create(&barrier);
goto finish;
}
+ if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uuid_socket_pair) < 0) {
+ r = log_error_errno(errno, "Failed to create id socket pair: %m");
+ goto finish;
+ }
+
if (arg_userns)
if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0) {
r = log_error_errno(errno, "Failed to create uid shift socket pair: %m");
kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]);
rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
pid_socket_pair[0] = safe_close(pid_socket_pair[0]);
+ uuid_socket_pair[0] = safe_close(uuid_socket_pair[0]);
uid_shift_socket_pair[0] = safe_close(uid_shift_socket_pair[0]);
(void) reset_all_signal_handlers();
interactive,
secondary,
pid_socket_pair[1],
+ uuid_socket_pair[1],
kmsg_socket_pair[1],
rtnl_socket_pair[1],
uid_shift_socket_pair[1],
kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]);
pid_socket_pair[1] = safe_close(pid_socket_pair[1]);
+ uuid_socket_pair[1] = safe_close(uuid_socket_pair[1]);
uid_shift_socket_pair[1] = safe_close(uid_shift_socket_pair[1]);
/* Wait for the outer child. */
goto finish;
}
+ /* We also retrieve container UUID in case it was generated by outer child */
+ l = recv(uuid_socket_pair[0], &arg_uuid, sizeof(arg_uuid), 0);
+ if (l < 0) {
+ r = log_error_errno(errno, "Failed to read container machine ID: %m");
+ goto finish;
+ }
+ if (l != sizeof(arg_uuid)) {
+ log_error("Short read while reading container machined ID.");
+ r = EIO;
+ goto finish;
+ }
+
log_debug("Init process invoked as PID " PID_FMT, pid);
if (arg_userns) {
/* We failed to wait for the container, or the
* container exited abnormally */
goto finish;
- else if (r > 0 || container_status == CONTAINER_TERMINATED){
+ else if (r > 0 || container_status == CONTAINER_TERMINATED) {
/* The container exited with a non-zero
* status, or with zero status and no reboot
* was requested. */
free(arg_image);
free(arg_machine);
free(arg_user);
+ free(arg_chdir);
strv_free(arg_setenv);
free(arg_network_bridge);
strv_free(arg_network_interfaces);