X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fcore%2Fmain.c;h=4da8ecb86453d8336e998f1c400b34dc3ab74f54;hb=774de5a97fe69da822fde77b88af8d970ab5d0c6;hp=6f6b565d68d9d5424bfaa3ccea2f9986347de8a6;hpb=cd3bd60a2e4bb08526d953a323bbe1a0ace78b9e;p=elogind.git diff --git a/src/core/main.c b/src/core/main.c index 6f6b565d6..4da8ecb86 100644 --- a/src/core/main.c +++ b/src/core/main.c @@ -51,6 +51,7 @@ #include "path-util.h" #include "switch-root.h" #include "capability.h" +#include "killall.h" #include "mount-setup.h" #include "loopback-setup.h" @@ -61,17 +62,19 @@ #include "hwclock.h" #include "selinux-setup.h" #include "ima-setup.h" +#include "sd-daemon.h" static enum { ACTION_RUN, ACTION_HELP, + ACTION_VERSION, ACTION_TEST, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DONE } arg_action = ACTION_RUN; static char *arg_default_unit = NULL; -static ManagerRunningAs arg_running_as = _MANAGER_RUNNING_AS_INVALID; +static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID; static bool arg_dump_core = true; static bool arg_crash_shell = false; @@ -166,12 +169,12 @@ _noreturn_ static void crash(int sig) { pid = fork(); if (pid < 0) - log_error("Failed to fork off crash shell: %s", strerror(errno)); + log_error("Failed to fork off crash shell: %m"); else if (pid == 0) { make_console_stdio(); execl("/bin/sh", "/bin/sh", NULL); - log_error("execl() failed: %s", strerror(errno)); + log_error("execl() failed: %m"); _exit(1); } @@ -348,12 +351,12 @@ static int parse_proc_cmdline_word(const char *word) { if (!eq) { r = unsetenv(cenv); if (r < 0) - log_warning("unsetenv failed %s. Ignoring.", strerror(errno)); + log_warning("unsetenv failed %m. Ignoring."); } else { *eq = 0; r = setenv(cenv, eq + 1, 1); if (r < 0) - log_warning("setenv failed %s. Ignoring.", strerror(errno)); + log_warning("setenv failed %m. Ignoring."); } free(cenv); @@ -493,14 +496,14 @@ static int config_parse_cpu_affinity2( unsigned cpu; if (!(t = strndup(w, l))) - return -ENOMEM; + return log_oom(); r = safe_atou(t, &cpu); free(t); if (!c) if (!(c = cpu_set_malloc(&ncpus))) - return -ENOMEM; + return log_oom(); if (r < 0 || cpu >= ncpus) { log_error("[%s:%u] Failed to parse CPU affinity: %s", filename, line, rvalue); @@ -566,7 +569,7 @@ static int config_parse_join_controllers( s = strndup(w, length); if (!s) - return -ENOMEM; + return log_oom(); l = strv_split(s, ","); free(s); @@ -582,7 +585,7 @@ static int config_parse_join_controllers( arg_join_controllers = new(char**, 2); if (!arg_join_controllers) { strv_free(l); - return -ENOMEM; + return log_oom(); } arg_join_controllers[0] = l; @@ -596,7 +599,7 @@ static int config_parse_join_controllers( t = new0(char**, n+2); if (!t) { strv_free(l); - return -ENOMEM; + return log_oom(); } n = 0; @@ -610,7 +613,7 @@ static int config_parse_join_controllers( if (!c) { strv_free(l); strv_free_free(t); - return -ENOMEM; + return log_oom(); } strv_free(l); @@ -622,7 +625,7 @@ static int config_parse_join_controllers( if (!c) { strv_free(l); strv_free_free(t); - return -ENOMEM; + return log_oom(); } t[n++] = c; @@ -682,7 +685,7 @@ static int parse_config_file(void) { const char *fn; int r; - fn = arg_running_as == MANAGER_SYSTEM ? SYSTEM_CONFIG_FILE : USER_CONFIG_FILE; + fn = arg_running_as == SYSTEMD_SYSTEM ? SYSTEM_CONFIG_FILE : USER_CONFIG_FILE; f = fopen(fn, "re"); if (!f) { if (errno == ENOENT) @@ -725,10 +728,13 @@ static int parse_proc_cmdline(void) { } r = parse_proc_cmdline_word(word); - free(word); - - if (r < 0) + if (r < 0) { + log_error("Failed on cmdline argument %s: %s", word, strerror(-r)); + free(word); goto finish; + } + + free(word); } r = 0; @@ -749,6 +755,7 @@ static int parse_argv(int argc, char *argv[]) { ARG_SYSTEM, ARG_USER, ARG_TEST, + ARG_VERSION, ARG_DUMP_CONFIGURATION_ITEMS, ARG_DUMP_CORE, ARG_CRASH_SHELL, @@ -771,6 +778,7 @@ static int parse_argv(int argc, char *argv[]) { { "user", no_argument, NULL, ARG_USER }, { "test", no_argument, NULL, ARG_TEST }, { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS }, { "dump-core", optional_argument, NULL, ARG_DUMP_CORE }, { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL }, @@ -865,17 +873,21 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_SYSTEM: - arg_running_as = MANAGER_SYSTEM; + arg_running_as = SYSTEMD_SYSTEM; break; case ARG_USER: - arg_running_as = MANAGER_USER; + arg_running_as = SYSTEMD_USER; break; case ARG_TEST: arg_action = ACTION_TEST; break; + case ARG_VERSION: + arg_action = ACTION_VERSION; + break; + case ARG_DUMP_CONFIGURATION_ITEMS: arg_action = ACTION_DUMP_CONFIGURATION_ITEMS; break; @@ -1009,8 +1021,10 @@ static int parse_argv(int argc, char *argv[]) { * instead. */ for (a = argv; a < argv + argc; a++) - if ((r = parse_proc_cmdline_word(*a)) < 0) + if ((r = parse_proc_cmdline_word(*a)) < 0) { + log_error("Failed on cmdline argument %s: %s", *a, strerror(-r)); return r; + } } return 0; @@ -1042,7 +1056,15 @@ static int help(void) { return 0; } -static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds) { +static int version(void) { + puts(PACKAGE_STRING); + puts(DISTRIBUTION); + puts(SYSTEMD_FEATURES); + + return 0; +} + +static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool serialize_jobs) { FILE *f = NULL; FDSet *fds = NULL; int r; @@ -1054,18 +1076,21 @@ static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds) { /* Make sure nothing is really destructed when we shut down */ m->n_reloading ++; - if ((r = manager_open_serialization(m, &f)) < 0) { + r = manager_open_serialization(m, &f); + if (r < 0) { log_error("Failed to create serialization file: %s", strerror(-r)); goto fail; } - if (!(fds = fdset_new())) { + fds = fdset_new(); + if (!fds) { r = -ENOMEM; log_error("Failed to allocate fd set: %s", strerror(-r)); goto fail; } - if ((r = manager_serialize(m, f, fds)) < 0) { + r = manager_serialize(m, f, fds, serialize_jobs); + if (r < 0) { log_error("Failed to serialize state: %s", strerror(-r)); goto fail; } @@ -1075,12 +1100,14 @@ static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds) { goto fail; } - if ((r = fd_cloexec(fileno(f), false)) < 0) { + r = fd_cloexec(fileno(f), false); + if (r < 0) { log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r)); goto fail; } - if ((r = fdset_cloexec(fds, false)) < 0) { + r = fdset_cloexec(fds, false); + if (r < 0) { log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r)); goto fail; } @@ -1099,6 +1126,42 @@ fail: return r; } +static int bump_rlimit_nofile(struct rlimit *saved_rlimit) { + struct rlimit nl; + int r; + + assert(saved_rlimit); + + /* Save the original RLIMIT_NOFILE so that we can reset it + * later when transitioning from the initrd to the main + * systemd or suchlike. */ + if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) { + log_error("Reading RLIMIT_NOFILE failed: %m"); + return -errno; + } + + /* Make sure forked processes get the default kernel setting */ + if (!arg_default_rlimit[RLIMIT_NOFILE]) { + struct rlimit *rl; + + rl = newdup(struct rlimit, saved_rlimit, 1); + if (!rl) + return log_oom(); + + arg_default_rlimit[RLIMIT_NOFILE] = rl; + } + + /* Bump up the resource limit for ourselves substantially */ + nl.rlim_cur = nl.rlim_max = 64*1024; + r = setrlimit_closest(RLIMIT_NOFILE, &nl); + if (r < 0) { + log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r)); + return r; + } + + return 0; +} + static struct dual_timestamp* parse_initrd_timestamp(struct dual_timestamp *t) { const char *e; unsigned long long a, b; @@ -1165,6 +1228,28 @@ static void test_cgroups(void) { sleep(10); } +static int initialize_join_controllers(void) { + /* By default, mount "cpu" + "cpuacct" together, and "net_cls" + * + "net_prio". We'd like to add "cpuset" to the mix, but + * "cpuset" does't really work for groups with no initialized + * attributes. */ + + arg_join_controllers = new(char**, 3); + if (!arg_join_controllers) + return -ENOMEM; + + arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL); + if (!arg_join_controllers[0]) + return -ENOMEM; + + arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL); + if (!arg_join_controllers[1]) + return -ENOMEM; + + arg_join_controllers[2] = NULL; + return 0; +} + int main(int argc, char *argv[]) { Manager *m = NULL; int r, retval = EXIT_FAILURE; @@ -1181,6 +1266,7 @@ int main(int argc, char *argv[]) { bool arm_reboot_watchdog = false; bool queue_default_job = false; char *switch_root_dir = NULL, *switch_root_init = NULL; + static struct rlimit saved_rlimit_nofile = { 0, 0 }; #ifdef HAVE_SYSV_COMPAT if (getpid() != 1 && strstr(program_invocation_short_name, "init")) { @@ -1222,10 +1308,15 @@ int main(int argc, char *argv[]) { saved_argc = argc; log_show_color(isatty(STDERR_FILENO) > 0); - log_show_location(false); - log_set_max_level(LOG_INFO); - if (getpid() == 1) { + if (getpid() == 1 && detect_container(NULL) <= 0) { + + /* Running outside of a container as PID 1 */ + arg_running_as = SYSTEMD_SYSTEM; + make_null_stdio(); + log_set_target(LOG_TARGET_KMSG); + log_open(); + if (in_initrd()) { char *rd_timestamp = NULL; @@ -1239,9 +1330,6 @@ int main(int argc, char *argv[]) { } } - arg_running_as = MANAGER_SYSTEM; - log_set_target(detect_container(NULL) > 0 ? LOG_TARGET_JOURNAL : LOG_TARGET_JOURNAL_OR_KMSG); - if (!skip_setup) { if (selinux_setup(&loaded_policy) < 0) goto finish; @@ -1249,41 +1337,72 @@ int main(int argc, char *argv[]) { goto finish; } - log_open(); - if (label_init(NULL) < 0) goto finish; - if (!skip_setup) + if (!skip_setup) { if (hwclock_is_localtime() > 0) { int min; - r = hwclock_apply_localtime_delta(&min); + /* The first-time call to settimeofday() does a time warp in the kernel */ + r = hwclock_set_timezone(&min); if (r < 0) log_error("Failed to apply local time delta, ignoring: %s", strerror(-r)); else log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min); + } else if (!in_initrd()) { + /* + * Do dummy first-time call to seal the kernel's time warp magic + * + * Do not call this this from inside the initrd. The initrd might not + * carry /etc/adjtime with LOCAL, but the real system could be set up + * that way. In such case, we need to delay the time-warp or the sealing + * until we reach the real system. + */ + hwclock_reset_timezone(); + + /* Tell the kernel our time zone */ + r = hwclock_set_timezone(NULL); + if (r < 0) + log_error("Failed to set the kernel's time zone, ignoring: %s", strerror(-r)); } + } + + /* Set the default for later on, but don't actually + * open the logs like this for now. Note that if we + * are transitioning from the initrd there might still + * be journal fd open, and we shouldn't attempt + * opening that before we parsed /proc/cmdline which + * might redirect output elsewhere. */ + log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); + + } else if (getpid() == 1) { + + /* Running inside a container, as PID 1 */ + arg_running_as = SYSTEMD_SYSTEM; + log_set_target(LOG_TARGET_CONSOLE); + log_open(); + + /* For the later on, see above... */ + log_set_target(LOG_TARGET_JOURNAL); } else { - arg_running_as = MANAGER_USER; + + /* Running as user instance */ + arg_running_as = SYSTEMD_USER; log_set_target(LOG_TARGET_AUTO); log_open(); } /* Initialize default unit */ - if (set_default_unit(SPECIAL_DEFAULT_TARGET) < 0) - goto finish; - - /* By default, mount "cpu" and "cpuacct" together */ - arg_join_controllers = new(char**, 2); - if (!arg_join_controllers) + r = set_default_unit(SPECIAL_DEFAULT_TARGET); + if (r < 0) { + log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r)); goto finish; + } - arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL); - arg_join_controllers[1] = NULL; - - if (!arg_join_controllers[0]) + r = initialize_join_controllers(); + if (r < 0) goto finish; /* Mount /proc, /sys and friends, so that /proc/cmdline and @@ -1303,7 +1422,7 @@ int main(int argc, char *argv[]) { if (parse_config_file() < 0) goto finish; - if (arg_running_as == MANAGER_SYSTEM) + if (arg_running_as == SYSTEMD_SYSTEM) if (parse_proc_cmdline() < 0) goto finish; @@ -1312,12 +1431,20 @@ int main(int argc, char *argv[]) { if (parse_argv(argc, argv) < 0) goto finish; - if (arg_action == ACTION_TEST && geteuid() == 0) { + if (arg_action == ACTION_TEST && + geteuid() == 0) { log_error("Don't run test mode as root."); goto finish; } - if (arg_running_as == MANAGER_SYSTEM && + if (arg_running_as == SYSTEMD_USER && + arg_action == ACTION_RUN && + sd_booted() <= 0) { + log_error("Trying to run as user instance, but the system has not been booted with systemd."); + goto finish; + } + + if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN && running_in_chroot() > 0) { log_error("Cannot be run in a chroot() environment."); @@ -1327,6 +1454,9 @@ int main(int argc, char *argv[]) { if (arg_action == ACTION_HELP) { retval = help(); goto finish; + } else if (arg_action == ACTION_VERSION) { + retval = version(); + goto finish; } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) { unit_dump_config_items(stdout); retval = EXIT_SUCCESS; @@ -1360,9 +1490,9 @@ int main(int argc, char *argv[]) { #else "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin", #endif - arg_running_as == MANAGER_SYSTEM); + arg_running_as == SYSTEMD_SYSTEM); - if (arg_running_as == MANAGER_SYSTEM) { + if (arg_running_as == SYSTEMD_SYSTEM) { /* Parse the data passed to us. We leave this * variables set, but the manager later on will not * pass them on to our children. */ @@ -1393,7 +1523,7 @@ int main(int argc, char *argv[]) { /* Move out of the way, so that we won't block unmounts */ assert_se(chdir("/") == 0); - if (arg_running_as == MANAGER_SYSTEM) { + if (arg_running_as == SYSTEMD_SYSTEM) { /* Become a session leader if we aren't one yet. */ setsid(); @@ -1406,10 +1536,8 @@ int main(int argc, char *argv[]) { /* 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) { + if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN) console_setup(getpid() == 1 && !skip_setup); - make_null_stdio(); - } /* Open the logging devices, if possible and necessary */ log_open(); @@ -1425,7 +1553,7 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_running_as == MANAGER_SYSTEM) { + if (arg_running_as == SYSTEMD_SYSTEM) { const char *virtualization = NULL; log_info(PACKAGE_STRING " running in system mode. (" SYSTEMD_FEATURES "; " DISTRIBUTION ")"); @@ -1440,7 +1568,7 @@ int main(int argc, char *argv[]) { } else log_debug(PACKAGE_STRING " running in user mode. (" SYSTEMD_FEATURES "; " DISTRIBUTION ")"); - if (arg_running_as == MANAGER_SYSTEM && !skip_setup) { + if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) { locale_setup(); if (arg_show_status || plymouth_running()) @@ -1456,7 +1584,7 @@ int main(int argc, char *argv[]) { test_cgroups(); } - if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0) + if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0) watchdog_set_timeout(&arg_runtime_watchdog); if (arg_timer_slack_nsec != (nsec_t) -1) @@ -1476,6 +1604,18 @@ int main(int argc, char *argv[]) { } } + if (arg_running_as == SYSTEMD_USER) { + /* Become reaper of our children */ + if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) { + log_warning("Failed to make us a subreaper: %m"); + if (errno == EINVAL) + log_info("Perhaps the kernel version is too old (< 3.4?)"); + } + } + + if (arg_running_as == SYSTEMD_SYSTEM) + bump_rlimit_nofile(&saved_rlimit_nofile); + r = manager_new(arg_running_as, &m); if (r < 0) { log_error("Failed to allocate manager object: %s", strerror(-r)); @@ -1605,7 +1745,7 @@ int main(int argc, char *argv[]) { case MANAGER_REEXECUTE: - if (prepare_reexecute(m, &serialization, &fds) < 0) + if (prepare_reexecute(m, &serialization, &fds, true) < 0) goto finish; reexecute = true; @@ -1619,7 +1759,7 @@ int main(int argc, char *argv[]) { m->switch_root = m->switch_root_init = NULL; if (!switch_root_init) - if (prepare_reexecute(m, &serialization, &fds) < 0) + if (prepare_reexecute(m, &serialization, &fds, false) < 0) goto finish; reexecute = true; @@ -1654,7 +1794,7 @@ finish: manager_free(m); for (j = 0; j < RLIMIT_NLIMITS; j++) - free (arg_default_rlimit[j]); + free(arg_default_rlimit[j]); free(arg_default_unit); strv_free(arg_default_controllers); @@ -1672,10 +1812,20 @@ finish: * rebooted while we do that */ watchdog_close(true); - /* Reopen the console */ - make_console_stdio(); + /* Reset the RLIMIT_NOFILE to the kernel default, so + * that the new systemd can pass the kernel default to + * its child processes */ + if (saved_rlimit_nofile.rlim_cur > 0) + setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile); if (switch_root_dir) { + /* Kill all remaining processes from the + * initrd, but don't wait for them, so that we + * can handle the SIGCHLD for them after + * deserializing. */ + broadcast_signal(SIGTERM, false); + + /* And switch root */ r = switch_root(switch_root_dir); if (r < 0) log_error("Failed to switch root, ignoring: %s", strerror(-r)); @@ -1702,7 +1852,7 @@ finish: args[i++] = SYSTEMD_BINARY_PATH; if (switch_root_dir) args[i++] = "--switched-root"; - args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user"; + args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user"; args[i++] = "--deserialize"; args[i++] = sfd; args[i++] = NULL; @@ -1727,6 +1877,9 @@ finish: fds = NULL; } + /* Reopen the console */ + make_console_stdio(); + for (j = 1, i = 1; j < argc; j++) args[i++] = argv[j]; args[i++] = NULL;