#include <sys/wait.h>
#include <fcntl.h>
#include <sys/prctl.h>
+#include <sys/mount.h>
#include "manager.h"
#include "log.h"
-#include "mount-setup.h"
-#include "hostname-setup.h"
-#include "loopback-setup.h"
-#include "kmod-setup.h"
-#include "locale-setup.h"
-#include "selinux-setup.h"
-#include "ima-setup.h"
-#include "machine-id-util.h"
#include "load-fragment.h"
#include "fdset.h"
#include "special.h"
#include "def.h"
#include "virt.h"
#include "watchdog.h"
+#include "path-util.h"
+#include "switch-root.h"
+#include "capability.h"
+
+#include "mount-setup.h"
+#include "loopback-setup.h"
+#include "kmod-setup.h"
+#include "hostname-setup.h"
+#include "machine-id-setup.h"
+#include "locale-setup.h"
+#include "hwclock.h"
+#include "selinux-setup.h"
+#include "ima-setup.h"
static enum {
ACTION_RUN,
static int arg_crash_chvt = -1;
static bool arg_confirm_spawn = false;
static bool arg_show_status = true;
+static bool arg_switched_root = false;
#ifdef HAVE_SYSV_COMPAT
static bool arg_sysv_console = true;
#endif
-static bool arg_mount_auto = true;
-static bool arg_swap_auto = true;
static char **arg_default_controllers = NULL;
static char ***arg_join_controllers = NULL;
static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
static usec_t arg_runtime_watchdog = 0;
static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
+static struct rlimit *arg_default_rlimit[RLIMIT_NLIMITS] = {};
+static uint64_t arg_capability_bounding_set_drop = 0;
+static nsec_t arg_timer_slack_nsec = (nsec_t) -1;
static FILE* serialization = NULL;
assert(u);
- if (!(c = strdup(u)))
+ c = strdup(u);
+ if (!c)
return -ENOMEM;
free(arg_default_unit);
arg_default_unit = c;
+
return 0;
}
assert(word);
- if (startswith(word, "systemd.unit="))
- return set_default_unit(word + 13);
+ if (startswith(word, "systemd.unit=")) {
+
+ if (!in_initrd())
+ return set_default_unit(word + 13);
+
+ } else if (startswith(word, "rd.systemd.unit=")) {
- else if (startswith(word, "systemd.log_target=")) {
+ if (in_initrd())
+ return set_default_unit(word + 16);
+
+ } else if (startswith(word, "systemd.log_target=")) {
if (log_set_target_from_string(word + 19) < 0)
log_warning("Failed to parse log target %s. Ignoring.", word + 19);
arg_sysv_console = r;
#endif
- } else if (startswith(word, "systemd.")) {
+ } else if (startswith(word, "systemd.") ||
+ (in_initrd() && startswith(word, "rd.systemd."))) {
log_warning("Unknown kernel switch %s. Ignoring.", word);
log_info("Supported kernel switches:\n"
"systemd.unit=UNIT Default unit to start\n"
+ "rd.systemd.unit=UNIT Default unit to start when run in initrd\n"
"systemd.dump_core=0|1 Dump core on crash\n"
"systemd.crash_shell=0|1 Run shell on crash\n"
"systemd.crash_chvt=N Change to VT #N on crash\n"
"systemd.default_standard_output=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
" Set default log output for services\n"
"systemd.default_standard_error=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
- " Set default log error output for services\n");
+ " Set default log error output for services\n"
+ "systemd.setenv=ASSIGNMENT Set an environment variable for all spawned processes\n");
} else if (streq(word, "quiet")) {
arg_show_status = false;
#ifdef HAVE_SYSV_COMPAT
arg_sysv_console = false;
#endif
- } else {
+ } else if (!in_initrd()) {
unsigned i;
/* SysV compatibility */
#endif
{ "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
{ "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
- { "Manager", "MountAuto", config_parse_bool, 0, &arg_mount_auto },
- { "Manager", "SwapAuto", config_parse_bool, 0, &arg_swap_auto },
{ "Manager", "DefaultControllers", config_parse_strv, 0, &arg_default_controllers },
{ "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
{ "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
{ "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
{ "Manager", "RuntimeWatchdogSec", config_parse_usec, 0, &arg_runtime_watchdog },
{ "Manager", "ShutdownWatchdogSec", config_parse_usec, 0, &arg_shutdown_watchdog },
+ { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
+ { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
+ { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU]},
+ { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE]},
+ { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA]},
+ { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK]},
+ { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE]},
+ { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS]},
+ { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE]},
+ { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS]},
+ { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC]},
+ { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK]},
+ { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS]},
+ { "Manager", "DefaultLimitSIGPENDING",config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING]},
+ { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE]},
+ { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE]},
+ { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO]},
+ { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME]},
{ NULL, NULL, NULL, 0, NULL }
};
ARG_SHOW_STATUS,
ARG_SYSV_CONSOLE,
ARG_DESERIALIZE,
+ ARG_SWITCHED_ROOT,
ARG_INTROSPECT,
ARG_DEFAULT_STD_OUTPUT,
ARG_DEFAULT_STD_ERROR
{ "test", no_argument, NULL, ARG_TEST },
{ "help", no_argument, NULL, 'h' },
{ "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
- { "dump-core", no_argument, NULL, ARG_DUMP_CORE },
- { "crash-shell", no_argument, NULL, ARG_CRASH_SHELL },
- { "confirm-spawn", no_argument, NULL, ARG_CONFIRM_SPAWN },
+ { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
+ { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
+ { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
{ "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
#ifdef HAVE_SYSV_COMPAT
{ "sysv-console", optional_argument, NULL, ARG_SYSV_CONSOLE },
#endif
{ "deserialize", required_argument, NULL, ARG_DESERIALIZE },
+ { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
{ "introspect", optional_argument, NULL, ARG_INTROSPECT },
{ "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
{ "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
break;
case ARG_DUMP_CORE:
- arg_dump_core = true;
+ r = optarg ? parse_boolean(optarg) : 1;
+ if (r < 0) {
+ log_error("Failed to parse dump core boolean %s.", optarg);
+ return r;
+ }
+ arg_dump_core = r;
break;
case ARG_CRASH_SHELL:
- arg_crash_shell = true;
+ r = optarg ? parse_boolean(optarg) : 1;
+ if (r < 0) {
+ log_error("Failed to parse crash shell boolean %s.", optarg);
+ return r;
+ }
+ arg_crash_shell = r;
break;
case ARG_CONFIRM_SPAWN:
- arg_confirm_spawn = true;
+ r = optarg ? parse_boolean(optarg) : 1;
+ if (r < 0) {
+ log_error("Failed to parse confirm spawn boolean %s.", optarg);
+ return r;
+ }
+ arg_confirm_spawn = r;
break;
case ARG_SHOW_STATUS:
-
- if (optarg) {
- if ((r = parse_boolean(optarg)) < 0) {
- log_error("Failed to show status boolean %s.", optarg);
- return r;
- }
- arg_show_status = r;
- } else
- arg_show_status = true;
+ r = optarg ? parse_boolean(optarg) : 1;
+ if (r < 0) {
+ log_error("Failed to parse show status boolean %s.", optarg);
+ return r;
+ }
+ arg_show_status = r;
break;
+
#ifdef HAVE_SYSV_COMPAT
case ARG_SYSV_CONSOLE:
-
- if (optarg) {
- if ((r = parse_boolean(optarg)) < 0) {
- log_error("Failed to SysV console boolean %s.", optarg);
- return r;
- }
- arg_sysv_console = r;
- } else
- arg_sysv_console = true;
+ r = optarg ? parse_boolean(optarg) : 1;
+ if (r < 0) {
+ log_error("Failed to parse SysV console boolean %s.", optarg);
+ return r;
+ }
+ arg_sysv_console = r;
break;
#endif
break;
}
+ case ARG_SWITCHED_ROOT:
+ arg_switched_root = true;
+ break;
+
case ARG_INTROSPECT: {
const char * const * i = NULL;
" --unit=UNIT Set default unit\n"
" --system Run a system instance, even if PID != 1\n"
" --user Run a user instance\n"
- " --dump-core Dump core on crash\n"
- " --crash-shell Run shell on crash\n"
- " --confirm-spawn Ask for confirmation when spawning processes\n"
+ " --dump-core[=0|1] Dump core on crash\n"
+ " --crash-shell[=0|1] Run shell on crash\n"
+ " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
" --show-status[=0|1] Show status updates on the console during bootup\n"
#ifdef HAVE_SYSV_COMPAT
" --sysv-console[=0|1] Connect output of SysV scripts to console\n"
const char *shutdown_verb = NULL;
dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
static char systemd[] = "systemd";
- bool is_reexec = false;
+ bool skip_setup = false;
int j;
bool loaded_policy = false;
bool arm_reboot_watchdog = false;
+ bool queue_default_job = false;
+ char *switch_root_dir = NULL, *switch_root_init = NULL;
#ifdef HAVE_SYSV_COMPAT
if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
/* Determine if this is a reexecution or normal bootup. We do
* the full command line parsing much later, so let's just
* have a quick peek here. */
-
for (j = 1; j < argc; j++)
if (streq(argv[j], "--deserialize")) {
- is_reexec = true;
+ skip_setup = true;
+ break;
+ }
+
+ /* If we have switched root, do all the special setup
+ * things */
+ for (j = 1; j < argc; j++)
+ if (streq(argv[j], "--switched-root")) {
+ skip_setup = false;
break;
}
log_set_max_level(LOG_INFO);
if (getpid() == 1) {
+ if (in_initrd()) {
+ char *rd_timestamp = NULL;
+
+ dual_timestamp_get(&initrd_timestamp);
+ asprintf(&rd_timestamp, "%llu %llu",
+ (unsigned long long) initrd_timestamp.realtime,
+ (unsigned long long) initrd_timestamp.monotonic);
+ if (rd_timestamp) {
+ setenv("RD_TIMESTAMP", rd_timestamp, 1);
+ free(rd_timestamp);
+ }
+ }
+
arg_running_as = MANAGER_SYSTEM;
- log_set_target(detect_container(NULL) > 0 ? LOG_TARGET_CONSOLE : LOG_TARGET_JOURNAL_OR_KMSG);
+ log_set_target(detect_container(NULL) > 0 ? LOG_TARGET_JOURNAL : LOG_TARGET_JOURNAL_OR_KMSG);
- if (!is_reexec) {
+ if (!skip_setup) {
if (selinux_setup(&loaded_policy) < 0)
goto finish;
if (ima_setup() < 0)
log_open();
- if (label_init() < 0)
+ if (label_init(NULL) < 0)
goto finish;
- if (!is_reexec)
+ if (!skip_setup)
if (hwclock_is_localtime() > 0) {
int min;
/* Parse the data passed to us. We leave this
* variables set, but the manager later on will not
* pass them on to our children. */
- parse_initrd_timestamp(&initrd_timestamp);
+ if (!in_initrd())
+ parse_initrd_timestamp(&initrd_timestamp);
/* Unset some environment variables passed in from the
* kernel that don't really make sense for us. */
unsetenv("SHLVL");
unsetenv("_");
+ /* When we are invoked by a chroot-like tool such as
+ * nspawn, these might be set, but make little sense
+ * to pass on */
+ unsetenv("USER");
+ unsetenv("LOGNAME");
+
/* All other variables are left as is, so that clients
* can still read them via /proc/1/environ */
}
/* Reset the console, but only if this is really init and we
* are freshly booted */
if (arg_running_as == MANAGER_SYSTEM && arg_action == ACTION_RUN) {
- console_setup(getpid() == 1 && !is_reexec);
+ console_setup(getpid() == 1 && !skip_setup);
make_null_stdio();
}
log_full(arg_running_as == MANAGER_SYSTEM ? LOG_INFO : LOG_DEBUG,
PACKAGE_STRING " running in %s mode. (" SYSTEMD_FEATURES "; " DISTRIBUTION ")", manager_running_as_to_string(arg_running_as));
- if (arg_running_as == MANAGER_SYSTEM && !is_reexec) {
+ if (arg_running_as == MANAGER_SYSTEM && !skip_setup) {
locale_setup();
if (arg_show_status || plymouth_running())
if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0)
watchdog_set_timeout(&arg_runtime_watchdog);
+ if (arg_timer_slack_nsec != (nsec_t) -1)
+ if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
+ log_error("Failed to adjust timer slack: %m");
+
+ if (arg_capability_bounding_set_drop) {
+ r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
+ if (r < 0) {
+ log_error("Failed to drop capability bounding set: %s", strerror(-r));
+ goto finish;
+ }
+ r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
+ if (r < 0) {
+ log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
+ goto finish;
+ }
+ }
+
r = manager_new(arg_running_as, &m);
if (r < 0) {
log_error("Failed to allocate manager object: %s", strerror(-r));
#ifdef HAVE_SYSV_COMPAT
m->sysv_console = arg_sysv_console;
#endif
- m->mount_auto = arg_mount_auto;
- m->swap_auto = arg_swap_auto;
m->default_std_output = arg_default_std_output;
m->default_std_error = arg_default_std_error;
m->runtime_watchdog = arg_runtime_watchdog;
m->shutdown_watchdog = arg_shutdown_watchdog;
+ manager_set_default_rlimits(m, arg_default_rlimit);
+
if (dual_timestamp_is_set(&initrd_timestamp))
m->initrd_timestamp = initrd_timestamp;
manager_set_show_status(m, arg_show_status);
+ /* Remember whether we should queue the default job */
+ queue_default_job = !serialization || arg_switched_root;
+
before_startup = now(CLOCK_MONOTONIC);
r = manager_startup(m, serialization, fds);
if (r < 0)
log_error("Failed to fully start up daemon: %s", strerror(-r));
+ /* This will close all file descriptors that were opened, but
+ * not claimed by any unit. */
if (fds) {
- /* This will close all file descriptors that were opened, but
- * not claimed by any unit. */
-
fdset_free(fds);
fds = NULL;
}
if (serialization) {
fclose(serialization);
serialization = NULL;
- } else {
+ }
+
+ if (queue_default_job) {
DBusError error;
Unit *target = NULL;
Job *default_unit_job;
break;
case MANAGER_REEXECUTE:
+
if (prepare_reexecute(m, &serialization, &fds) < 0)
goto finish;
log_notice("Reexecuting.");
goto finish;
+ case MANAGER_SWITCH_ROOT:
+ /* Steal the switch root parameters */
+ switch_root_dir = m->switch_root;
+ switch_root_init = m->switch_root_init;
+ m->switch_root = m->switch_root_init = NULL;
+
+ if (!switch_root_init)
+ if (prepare_reexecute(m, &serialization, &fds) < 0)
+ goto finish;
+
+ reexecute = true;
+ log_notice("Switching root.");
+ goto finish;
+
case MANAGER_REBOOT:
case MANAGER_POWEROFF:
case MANAGER_HALT:
if (m)
manager_free(m);
+ for (j = 0; j < RLIMIT_NLIMITS; j++)
+ free (arg_default_rlimit[j]);
+
free(arg_default_unit);
strv_free(arg_default_controllers);
free_join_controllers();
dbus_shutdown();
-
label_finish();
if (reexecute) {
- const char *args[15];
- unsigned i = 0;
- char sfd[16];
+ const char **args;
+ unsigned i, args_size;
- assert(serialization);
- assert(fds);
+ /* Close and disarm the watchdog, so that the new
+ * instance can reinitialize it, but doesn't get
+ * rebooted while we do that */
+ watchdog_close(true);
- args[i++] = SYSTEMD_BINARY_PATH;
+ if (switch_root_dir) {
+ r = switch_root(switch_root_dir);
+ if (r < 0)
+ log_error("Failed to switch root, ignoring: %s", strerror(-r));
+ }
- args[i++] = "--log-level";
- args[i++] = log_level_to_string(log_get_max_level());
+ args_size = MAX(6, argc+1);
+ args = newa(const char*, args_size);
- args[i++] = "--log-target";
- args[i++] = log_target_to_string(log_get_target());
+ if (!switch_root_init) {
+ char sfd[16];
- if (arg_running_as == MANAGER_SYSTEM)
- args[i++] = "--system";
- else
- args[i++] = "--user";
+ /* First try to spawn ourselves with the right
+ * path, and with full serialization. We do
+ * this only if the user didn't specify an
+ * explicit init to spawn. */
- if (arg_dump_core)
- args[i++] = "--dump-core";
+ assert(serialization);
+ assert(fds);
- if (arg_crash_shell)
- args[i++] = "--crash-shell";
+ snprintf(sfd, sizeof(sfd), "%i", fileno(serialization));
+ char_array_0(sfd);
- if (arg_confirm_spawn)
- args[i++] = "--confirm-spawn";
+ i = 0;
+ args[i++] = SYSTEMD_BINARY_PATH;
+ if (switch_root_dir)
+ args[i++] = "--switched-root";
+ args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user";
+ args[i++] = "--deserialize";
+ args[i++] = sfd;
+ args[i++] = NULL;
- if (arg_show_status)
- args[i++] = "--show-status=1";
- else
- args[i++] = "--show-status=0";
+ assert(i <= args_size);
+ execv(args[0], (char* const*) args);
+ }
-#ifdef HAVE_SYSV_COMPAT
- if (arg_sysv_console)
- args[i++] = "--sysv-console=1";
- else
- args[i++] = "--sysv-console=0";
-#endif
+ /* Try the fallback, if there is any, without any
+ * serialization. We pass the original argv[] and
+ * envp[]. (Well, modulo the ordering changes due to
+ * getopt() in argv[], and some cleanups in envp[],
+ * but let's hope that doesn't matter.) */
- snprintf(sfd, sizeof(sfd), "%i", fileno(serialization));
- char_array_0(sfd);
+ if (serialization) {
+ fclose(serialization);
+ serialization = NULL;
+ }
- args[i++] = "--deserialize";
- args[i++] = sfd;
+ if (fds) {
+ fdset_free(fds);
+ fds = NULL;
+ }
+ for (j = 1, i = 1; j < argc; j++)
+ args[i++] = argv[j];
args[i++] = NULL;
+ assert(i <= args_size);
- assert(i <= ELEMENTSOF(args));
-
- /* Close and disarm the watchdog, so that the new
- * instance can reinitialize it, but doesn't get
- * rebooted while we do that */
- watchdog_close(true);
+ if (switch_root_init) {
+ args[0] = switch_root_init;
+ execv(args[0], (char* const*) args);
+ log_warning("Failed to execute configured init, trying fallback: %m");
+ }
+ args[0] = "/sbin/init";
execv(args[0], (char* const*) args);
- log_error("Failed to reexecute: %m");
+ if (errno == ENOENT) {
+ log_warning("No /sbin/init, trying fallback");
+
+ args[0] = "/bin/sh";
+ args[1] = NULL;
+ execv(args[0], (char* const*) args);
+ log_error("Failed to execute /bin/sh, giving up: %m");
+ } else
+ log_warning("Failed to execute /sbin/init, giving up: %m");
}
if (serialization)
~ianmdlvl / elogind.git / blobdiff