picnoir
/
Systemd
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Systemd/src/core/main.c

2269 lines
93 KiB
C
Raw Normal View History

license: add GPLv2+ license blurbs everwhere
2010-02-03 13:03:47 +01:00
/***
This file is part of systemd.
Copyright 2010 Lennart Poettering
systemd is free software; you can redistribute it and/or modify it
relicense to LGPLv2.1 (with exceptions) We finally got the OK from all contributors with non-trivial commits to relicense systemd from GPL2+ to LGPL2.1+. Some udev bits continue to be GPL2+ for now, but we are looking into relicensing them too, to allow free copy/paste of all code within systemd. The bits that used to be MIT continue to be MIT. The big benefit of the relicensing is that closed source code may now link against libsystemd-login.so and friends.
2012-04-12 00:20:58 +02:00
under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 2.1 of the License, or
license: add GPLv2+ license blurbs everwhere
2010-02-03 13:03:47 +01:00
(at your option) any later version.
systemd is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
relicense to LGPLv2.1 (with exceptions) We finally got the OK from all contributors with non-trivial commits to relicense systemd from GPL2+ to LGPL2.1+. Some udev bits continue to be GPL2+ for now, but we are looking into relicensing them too, to allow free copy/paste of all code within systemd. The bits that used to be MIT continue to be MIT. The big benefit of the relicensing is that closed source code may now link against libsystemd-login.so and friends.
2012-04-12 00:20:58 +02:00
Lesser General Public License for more details.
license: add GPLv2+ license blurbs everwhere
2010-02-03 13:03:47 +01:00
relicense to LGPLv2.1 (with exceptions) We finally got the OK from all contributors with non-trivial commits to relicense systemd from GPL2+ to LGPL2.1+. Some udev bits continue to be GPL2+ for now, but we are looking into relicensing them too, to allow free copy/paste of all code within systemd. The bits that used to be MIT continue to be MIT. The big benefit of the relicensing is that closed source code may now link against libsystemd-login.so and friends.
2012-04-12 00:20:58 +02:00
You should have received a copy of the GNU Lesser General Public License
license: add GPLv2+ license blurbs everwhere
2010-02-03 13:03:47 +01:00
along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/
initial commit
2009-11-18 00:42:52 +01:00
#include <errno.h>
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include <fcntl.h>
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
#include <getopt.h>
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
#include <signal.h>
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include <stdio.h>
#include <string.h>
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
#include <sys/mount.h>
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include <sys/prctl.h>
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
#include <sys/reboot.h>
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include <sys/stat.h>
#include <unistd.h>
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_SECCOMP
execute: no need to include seccomp.h from execute.h
2014-03-03 17:12:56 +01:00
#include <seccomp.h>
#endif
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_VALGRIND_VALGRIND_H
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include <valgrind/valgrind.h>
#endif
valgrind: make running PID 1 in valgrind useful Since valgrind only generates useful output on exit() (rather than exec()) we need to explicitly exit when valgrind is detected.
2013-11-20 22:11:10 +01:00
core: convert PID 1 to libsystemd-bus This patch converts PID 1 to libsystemd-bus and thus drops the dependency on libdbus. The only remaining code using libdbus is a test case that validates our bus marshalling against libdbus' marshalling, and this dependency can be turned off. This patch also adds a couple of things to libsystem-bus, that are necessary to make the port work: - Synthesizing of "Disconnected" messages when bus connections are severed. - Support for attaching multiple vtables for the same interface on the same path. This patch also fixes the SetDefaultTarget() and GetDefaultTarget() bus calls which used an inappropriate signature. As a side effect we will now generate PropertiesChanged messages which carry property contents, rather than just invalidation information.
2013-11-19 21:12:59 +01:00
#include "sd-bus.h"
tree-wide: sort includes Sort the includes accoding to the new coding style.
2015-11-16 22:09:36 +01:00
#include "sd-daemon.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
util-lib: split out allocation calls into alloc-util.[ch]
2015-10-27 03:01:06 +01:00
#include "alloc-util.h"
hostnamectl: should the sanitized arch, not the native uname() one
2014-02-21 02:28:54 +01:00
#include "architecture.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "build.h"
#include "bus-error.h"
#include "bus-util.h"
src/basic: rename audit.[ch] → audit-util.[ch] and capability.[ch] → capability-util.[ch] The files are named too generically, so that they might conflict with the upstream project headers. Hence, let's add a "-util" suffix, to clarify that this are just our utility headers and not any official upstream headers.
2015-10-26 23:32:16 +01:00
#include "capability-util.h"
shared: rename hwclock.[ch] to clock-util.[ch]
2014-05-22 14:21:38 +02:00
#include "clock-util.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "conf-parser.h"
basic: split out cpu set specific APIs into cpu-set-util.[ch]
2015-09-30 21:50:22 +02:00
#include "cpu-set-util.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "dbus-manager.h"
#include "def.h"
core: minor coding style/wording fixes
2016-12-08 17:23:08 +01:00
#include "emergency-action.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "env-util.h"
util-lib: split out fd-related operations into fd-util.[ch] There are more than enough to deserve their own .c file, hence move them over.
2015-10-25 13:14:12 +01:00
#include "fd-util.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "fdset.h"
core: convert PID 1 to libsystemd-bus This patch converts PID 1 to libsystemd-bus and thus drops the dependency on libdbus. The only remaining code using libdbus is a test case that validates our bus marshalling against libdbus' marshalling, and this dependency can be turned off. This patch also adds a couple of things to libsystem-bus, that are necessary to make the port work: - Synthesizing of "Disconnected" messages when bus connections are severed. - Support for attaching multiple vtables for the same interface on the same path. This patch also fixes the SetDefaultTarget() and GetDefaultTarget() bus calls which used an inappropriate signature. As a side effect we will now generate PropertiesChanged messages which carry property contents, rather than just invalidation information.
2013-11-19 21:12:59 +01:00
#include "fileio.h"
Rename formats-util.h to format-util.h We don't have plural in the name of any other -util files and this inconsistency trips me up every time I try to type this file name from memory. "formats-util" is even hard to pronounce.
2016-11-07 16:14:59 +01:00
#include "format-util.h"
util-lib: move a number of fs operations into fs-util.[ch]
2015-10-26 21:16:26 +01:00
#include "fs-util.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "hostname-setup.h"
#include "ima-setup.h"
#include "killall.h"
#include "kmod-setup.h"
tmpfiles: add new line type 'v' for creating btrfs subvolumes
2014-12-27 18:46:36 +01:00
#include "load-fragment.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "log.h"
main: we want all setup functions to be in files called xxx-setup.[ch]
2012-04-12 14:28:43 +02:00
#include "loopback-setup.h"
#include "machine-id-setup.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "manager.h"
#include "missing.h"
#include "mount-setup.h"
#include "pager.h"
util-lib: split string parsing related calls from util.[ch] into parse-util.[ch]
2015-10-26 16:18:16 +01:00
#include "parse-util.h"
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
#include "path-util.h"
util-lib: move /proc/cmdline parsing code to proc-cmdline.[ch]
2015-10-27 00:06:29 +01:00
#include "proc-cmdline.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "process-util.h"
util-lib: Add sparc64 support for process creation (#3348) The current raw_clone function takes two arguments, the cloning flags and a pointer to the stack for the cloned child. The raw cloning without passing a "thread main" function does not make sense if a new stack is specified, as it returns in both the parent and the child, which will fail in the child as the stack is virgin. All uses of raw_clone indeed pass NULL for the stack pointer which indicates that both processes should share the stack address (so you better don't pass CLONE_VM). This commit refactors the code to not require the caller to pass the stack address, as NULL is the only sensible option. It also adds the magic code needed to make raw_clone work on sparc64, which does not return 0 in %o0 for the child, but indicates the child process by setting %o1 to non-zero. This refactoring is not plain aesthetic, because non-NULL stack addresses need to get mangled before being passed to the clone syscall (you have to apply STACK_BIAS), whereas NULL must not be mangled. Implementing the conditional mangling of the stack address would needlessly complicate the code. raw_clone is moved to a separete header, because the burden of including the assert machinery and sched.h shouldn't be applied to every user of missing_syscalls.h
2016-05-30 02:03:51 +02:00
#include "raw-clone.h"
util-lib: split out resource limits related calls into rlimit-util.[ch]
2015-10-26 19:40:43 +01:00
#include "rlimit-util.h"
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_SECCOMP
core: do not fail at step SECCOMP if there is no kernel support (#4004) Fixes #3882
2016-08-22 21:40:58 +02:00
#include "seccomp-util.h"
#endif
main: we want all setup functions to be in files called xxx-setup.[ch]
2012-04-12 14:28:43 +02:00
#include "selinux-setup.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "selinux-util.h"
#include "signal-util.h"
core: mount and initialize Smack SMACK is the Simple Mandatory Access Control Kernel, a minimal approach to Access Control implemented as a kernel LSM. The kernel exposes the smackfs filesystem API through which access rules can be loaded. At boot time, we want to load the access rules as early as possible to ensure all early boot steps are checked by Smack. This patch mounts smackfs at the new location at /sys/fs/smackfs for kernels 3.8 and above. The /smack mountpoint is not supported. After mounting smackfs, rules are loaded from the usual location. For more information about Smack see: http://www.kernel.org/doc/Documentation/security/Smack.txt
2013-03-07 20:06:58 +01:00
#include "smack-setup.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "special.h"
util-lib: split stat()/statfs()/stavfs() related calls into stat-util.[ch]
2015-10-26 22:01:44 +01:00
#include "stat-util.h"
util-lib: split out printf() helpers to stdio-util.h
2015-10-27 01:26:31 +01:00
#include "stdio-util.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "strv.h"
#include "switch-root.h"
#include "terminal-util.h"
core: be more paranoid when mixing umask and fopen() Let's be extra careful with the umask when we use simple fopen(), as this creates files with 0777 by default.
2016-04-07 16:15:26 +02:00
#include "umask-util.h"
util-lib: split out user/group/uid/gid calls into user-util.[ch]
2015-10-25 22:32:30 +01:00
#include "user-util.h"
core: order #includes in main.c
2015-09-23 01:04:46 +02:00
#include "virt.h"
#include "watchdog.h"
main: we want all setup functions to be in files called xxx-setup.[ch]
2012-04-12 14:28:43 +02:00
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
static enum {
ACTION_RUN,
main: implement --test
2010-04-07 00:10:17 +02:00
ACTION_HELP,
systemd: add --version option systemd --version mirrors systemctl --version: $ ./systemd --version systemd 186 other +PAM +LIBWRAP +AUDIT +SELINUX +IMA +SYSVINIT -LIBCRYPTSETUP This information can be retrieved by other means (systemctl, etc.), but it's easier for a newbie if 'systemd --version' says something useful. And 'systemd --help' is already there, so let's complement that with '--version'.
2012-07-17 07:31:47 +02:00
ACTION_VERSION,
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
ACTION_TEST,
main: get rid of ACTION_DONE (#3849) the ACTION_DONE was introduced in the 4288f61921 (dbus: automatically generate and install introspection files ) commit and was used in systemd --introspect command. Later 'introspect' command was removed in the ca2871d9b (bus: remove static introspection file export) commit and have no users anymore. So we can remove it.
2016-08-01 12:38:25 +02:00
ACTION_DUMP_CONFIGURATION_ITEMS
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
} arg_action = ACTION_RUN;
static char *arg_default_unit = NULL;
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
static bool arg_system = false;
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
static bool arg_dump_core = true;
static int arg_crash_chvt = -1;
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
static bool arg_crash_shell = false;
static bool arg_crash_reboot = false;
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
static char *arg_confirm_spawn = NULL;
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
static bool arg_switched_root = false;
tree-wide: merge pager_open_if_enabled() to the pager_open() Many subsystems define own pager_open_if_enabled() function which checks '--no-pager' command line argument and open pager depends on its value. All implementations of pager_open_if_enabled() are the same. Let's merger this function with pager_open() from the shared/pager.c and remove pager_open_if_enabled() from all subsytems to prevent code duplication.
2016-02-19 19:25:13 +01:00
static bool arg_no_pager = false;
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
static char ***arg_join_controllers = NULL;
journal: introduce log target 'journal' for executed processes
2012-01-05 23:54:45 +01:00
static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
Configurable Timeouts/Restarts default values https://bugs.freedesktop.org/show_bug.cgi?id=71132 Patch adds DefaultTimeoutStartSec, DefaultTimeoutStopSec, DefaultRestartSec configuration options to manager configuration file.
2013-11-04 17:47:43 +01:00
static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
manager: configurable StartLimit default values https://bugzilla.redhat.com/show_bug.cgi?id=821723
2013-11-08 16:01:22 +01:00
static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
static usec_t arg_runtime_watchdog = 0;
static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
manager: add DefaultEnvironment option This complements existing functionality of setting variables through 'systemctl set-environment', the kernel command line, and through normal environment variables for systemd in session mode.
2013-06-09 07:08:46 +02:00
static char **arg_default_environment = NULL;
missing: if RLIMIT_RTTIME is not defined by the libc, then we need a new define for the max number of rlimits, too
2014-03-05 02:29:58 +01:00
static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
capabilities: keep bounding set in non-inverted format. Change the capability bounding set parser and logic so that the bounding set is kept as a positive set internally. This means that the set reflects those capabilities that we want to keep instead of drop.
2016-01-07 23:00:04 +01:00
static uint64_t arg_capability_bounding_set = CAP_ALL;
time-util: add and use USEC/NSEC_INFINIY
2014-07-29 12:23:31 +02:00
static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
core: add a setting to globally control the default for timer unit accuracy
2014-03-24 16:22:34 +01:00
static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
static Set* arg_syscall_archs = NULL;
static FILE* arg_serialization = NULL;
core: add global settings for enabling CPUAccounting=, MemoryAccounting=, BlockIOAccounting= for all units at once
2014-02-24 23:50:10 +01:00
static bool arg_default_cpu_accounting = false;
core: add io controller support on the unified hierarchy On the unified hierarchy, blkio controller is renamed to io and the interface is changed significantly. * blkio.weight and blkio.weight_device are consolidated into io.weight which uses the standardized weight range [1, 10000] with 100 as the default value. * blkio.throttle.{read|write}_{bps|iops}_device are consolidated into io.max. Expansion of throttling features is being worked on to support work-conserving absolute limits (io.low and io.high). * All stats are consolidated into io.stats. This patchset adds support for the new interface. As the interface has been revamped and new features are expected to be added, it seems best to treat it as a separate controller rather than trying to expand the blkio settings although we might add automatic translation if only blkio settings are specified. * io.weight handling is mostly identical to blkio.weight[_device] handling except that the weight range is different. * Both read and write bandwidth settings are consolidated into CGroupIODeviceLimit which describes all limits applicable to the device. This makes it less painful to add new limits. * "max" can be used to specify the maximum limit which is equivalent to no config for max limits and treated as such. If a given CGroupIODeviceLimit doesn't contain any non-default configs, the config struct is discarded once the no limit config is applied to cgroup. * lookup_blkio_device() is renamed to lookup_block_device(). Signed-off-by: Tejun Heo <htejun@fb.com>
2016-05-05 22:42:55 +02:00
static bool arg_default_io_accounting = false;
manager: hook up IP accounting defaults
2017-09-01 16:04:50 +02:00
static bool arg_default_ip_accounting = false;
core: add global settings for enabling CPUAccounting=, MemoryAccounting=, BlockIOAccounting= for all units at once
2014-02-24 23:50:10 +01:00
static bool arg_default_blockio_accounting = false;
static bool arg_default_memory_accounting = false;
core: enable TasksMax= for all services by default, and set it to 512 Also, enable TasksAccounting= for all services by default, too. See: http://lists.freedesktop.org/archives/systemd-devel/2015-November/035006.html
2015-11-13 19:28:32 +01:00
static bool arg_default_tasks_accounting = true;
core: change TasksMax= default for system services to 15% As it turns out 512 is max number of tasks per service is hit by too many applications, hence let's bump it a bit, and make it relative to the system's maximum number of PIDs. With this change the new default is 15%. At the kernel's default pids_max value of 32768 this translates to 4915. At machined's default TasksMax= setting of 16384 this translates to 2457. Why 15%? Because it sounds like a round number and is close enough to 4096 which I was going for, i.e. an eight-fold increase over the old 512 Summary: | on the host | in a container old default | 512 | 512 new default | 4915 | 2457
2016-07-19 17:29:00 +02:00
static uint64_t arg_default_tasks_max = UINT64_MAX;
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
static sd_id128_t arg_machine_id = {};
core: use emergency_action for ctr+alt+del burst Fixes #4306
2016-10-18 12:16:32 +02:00
static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
noreturn static void freeze_or_reboot(void) {
if (arg_crash_reboot) {
log_notice("Rebooting in 10s...");
(void) sleep(10);
log_notice("Rebooting now...");
(void) reboot(RB_AUTOBOOT);
log_emergency_errno(errno, "Failed to reboot: %m");
}
log_emergency("Freezing execution.");
freeze();
}
_noreturn_ --> noreturn for C11 compat also define noreturn w/o <stdnoreturn.h>
2013-12-16 17:53:53 +01:00
noreturn static void crash(int sig) {
core: always let the kernel reap zombies when we're about to freeze Regardless of whether we're going to spawn a crash shell or not, let the kernel reap zombies. It's more consistent this way.
2015-09-30 17:08:04 +02:00
struct sigaction sa;
pid_t pid;
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() != 1)
core: don't do runaway fork()s if we hit a segfault from our segfault handler
2013-06-27 02:28:12 +02:00
/* Pass this on immediately, if this is not PID 1 */
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) raise(sig);
core: don't do runaway fork()s if we hit a segfault from our segfault handler
2013-06-27 02:28:12 +02:00
else if (!arg_dump_core)
manager: log some fatal errors at emergency level This adds a new log_emergency() function, which is equivalent to log_error() for non-PID-1, and logs at the highest priority for PID 1. Some messages which occur before freezing are converted to use it.
2014-11-06 06:04:06 +01:00
log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
else {
core: always let the kernel reap zombies when we're about to freeze Regardless of whether we're going to spawn a crash shell or not, let the kernel reap zombies. It's more consistent this way.
2015-09-30 17:08:04 +02:00
sa = (struct sigaction) {
util: unify implementation of NOP signal handler This is highly complex code after all, we really should make sure to only keep one implementation of this extremely difficult function around.
2015-09-23 01:32:44 +02:00
.sa_handler = nop_signal_handler,
Use initalization instead of explicit zeroing Before, we would initialize many fields twice: first by filling the structure with zeros, and then a second time with the real values. We can let the compiler do the job for us, avoiding one copy. A downside of this patch is that text gets slightly bigger. This is because all zero() calls are effectively inlined: $ size build/.libs/systemd text data bss dec hex filename before 897737 107300 2560 1007597 f5fed build/.libs/systemd after 897873 107300 2560 1007733 f6075 build/.libs/systemd … actually less than 1‰. A few asserts that the parameter is not null had to be removed. I don't think this changes much, because first, it is quite unlikely for the assert to fail, and second, an immediate SEGV is almost as good as an assert.
2013-03-25 00:59:00 +01:00
.sa_flags = SA_NOCLDSTOP|SA_RESTART,
};
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
main: run crash shell as subprocess, so that we can gdb pid 1
2010-04-13 20:26:54 +02:00
/* We want to wait for the core process, hence let's enable SIGCHLD */
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) sigaction(SIGCHLD, &sa, NULL);
main: run crash shell as subprocess, so that we can gdb pid 1
2010-04-13 20:26:54 +02:00
util-lib: Add sparc64 support for process creation (#3348) The current raw_clone function takes two arguments, the cloning flags and a pointer to the stack for the cloned child. The raw cloning without passing a "thread main" function does not make sense if a new stack is specified, as it returns in both the parent and the child, which will fail in the child as the stack is virgin. All uses of raw_clone indeed pass NULL for the stack pointer which indicates that both processes should share the stack address (so you better don't pass CLONE_VM). This commit refactors the code to not require the caller to pass the stack address, as NULL is the only sensible option. It also adds the magic code needed to make raw_clone work on sparc64, which does not return 0 in %o0 for the child, but indicates the child process by setting %o1 to non-zero. This refactoring is not plain aesthetic, because non-NULL stack addresses need to get mangled before being passed to the clone syscall (you have to apply STACK_BIAS), whereas NULL must not be mangled. Implementing the conditional mangling of the stack address would needlessly complicate the code. raw_clone is moved to a separete header, because the burden of including the assert machinery and sched.h shouldn't be applied to every user of missing_syscalls.h
2016-05-30 02:03:51 +02:00
pid = raw_clone(SIGCHLD);
Modernization Use _cleanup_ and wrap lines to ~80 chars and such.
2013-03-25 00:45:16 +01:00
if (pid < 0)
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
else if (pid == 0) {
/* Enable default signal handler for core dump */
core: set RLIMIT_CORE to unlimited by default The kernel sets RLIMIT_CORE to 0 by default. Let's bump this to unlimited by default (for systemd itself and all processes we fork off), so that the coredump hooks have an effect if they honour it. Bumping RLIMIT_CORE of course would have the effect that "core" files will end up on the system at various places, if no coredump hook is used. To avoid this, make sure PID1 sets the core pattern to the empty string by default, so that this logic is disabled. This change in defaults should be useful for all systems where coredump hooks are used, as it allows useful usage of RLIMIT_CORE from these hooks again. OTOH systems that expect that coredumps are placed under the name "core" in the current directory will break with this change. Given how questionnable this behaviour is, and given that no common distro makes use of this by default it shouldn't be too much of a loss. Also, the old behaviour may be restored by explicitly configuring a "core_pattern" of "core", and setting the default system RLIMIT_CORE to 0 again via system.conf.
2016-02-08 22:30:58 +01:00
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
sa = (struct sigaction) {
.sa_handler = SIG_DFL,
};
(void) sigaction(sig, &sa, NULL);
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
core: set RLIMIT_CORE to unlimited by default The kernel sets RLIMIT_CORE to 0 by default. Let's bump this to unlimited by default (for systemd itself and all processes we fork off), so that the coredump hooks have an effect if they honour it. Bumping RLIMIT_CORE of course would have the effect that "core" files will end up on the system at various places, if no coredump hook is used. To avoid this, make sure PID1 sets the core pattern to the empty string by default, so that this logic is disabled. This change in defaults should be useful for all systems where coredump hooks are used, as it allows useful usage of RLIMIT_CORE from these hooks again. OTOH systems that expect that coredumps are placed under the name "core" in the current directory will break with this change. Given how questionnable this behaviour is, and given that no common distro makes use of this by default it shouldn't be too much of a loss. Also, the old behaviour may be restored by explicitly configuring a "core_pattern" of "core", and setting the default system RLIMIT_CORE to 0 again via system.conf.
2016-02-08 22:30:58 +01:00
/* Don't limit the coredump size */
(void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
/* Just to be sure... */
Add (void) where we don't care about return value
2015-03-15 22:17:24 +01:00
(void) chdir("/");
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
/* Raise the signal again */
core: use raw_clone instead of fork in signal handler fork() is not async-signal-safe and calling it from the signal handler could result in a deadlock when at_fork() handlers are called. Using the raw clone() syscall sidesteps that problem. The tricky part is that raise() does not work, since getpid() does not work. Add raw_getpid() to get the real pid, and use kill() instead of raise(). https://bugs.freedesktop.org/show_bug.cgi?id=86604
2014-12-17 05:53:23 +01:00
pid = raw_getpid();
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) kill(pid, sig); /* raise() would kill the parent */
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
assert_not_reached("We shouldn't be here...");
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
_exit(EXIT_FAILURE);
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
} else {
util: use waitid() instead of waitpid() everywhere to avoid confusion due to SIGSTOP
2010-09-15 14:48:59 +02:00
siginfo_t status;
int r;
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
/* Order things nicely. */
Modernization Use _cleanup_ and wrap lines to ~80 chars and such.
2013-03-25 00:45:16 +01:00
r = wait_for_terminate(pid, &status);
if (r < 0)
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
util: use waitid() instead of waitpid() everywhere to avoid confusion due to SIGSTOP
2010-09-15 14:48:59 +02:00
else if (status.si_code != CLD_DUMPED)
core: use raw_clone instead of fork in signal handler fork() is not async-signal-safe and calling it from the signal handler could result in a deadlock when at_fork() handlers are called. Using the raw clone() syscall sidesteps that problem. The tricky part is that raise() does not work, since getpid() does not work. Add raw_getpid() to get the real pid, and use kill() instead of raise(). https://bugs.freedesktop.org/show_bug.cgi?id=86604
2014-12-17 05:53:23 +01:00
log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
signal_to_string(sig),
pid, sigchld_code_to_string(status.si_code),
status.si_status,
strna(status.si_code == CLD_EXITED
core: lower exit status "level" at one place When we print information about PID 1's crashdump subprocess failing. In this case we *know* that we do not generate LSB exit codes, as it's basically PID 1 itself that exited there.
2016-10-10 21:56:36 +02:00
? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
core: use raw_clone instead of fork in signal handler fork() is not async-signal-safe and calling it from the signal handler could result in a deadlock when at_fork() handlers are called. Using the raw clone() syscall sidesteps that problem. The tricky part is that raise() does not work, since getpid() does not work. Add raw_getpid() to get the real pid, and use kill() instead of raise(). https://bugs.freedesktop.org/show_bug.cgi?id=86604
2014-12-17 05:53:23 +01:00
: signal_to_string(status.si_status)));
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
else
manager: log some fatal errors at emergency level This adds a new log_emergency() function, which is equivalent to log_error() for non-PID-1, and logs at the highest priority for PID 1. Some messages which occur before freezing are converted to use it.
2014-11-06 06:04:06 +01:00
log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
}
}
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
if (arg_crash_chvt >= 0)
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) chvt(arg_crash_chvt);
main: switch to primary console vt on crash
2010-04-10 23:36:43 +02:00
core: always let the kernel reap zombies when we're about to freeze Regardless of whether we're going to spawn a crash shell or not, let the kernel reap zombies. It's more consistent this way.
2015-09-30 17:08:04 +02:00
sa = (struct sigaction) {
.sa_handler = SIG_IGN,
.sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
};
/* Let the kernel reap children for us */
(void) sigaction(SIGCHLD, &sa, NULL);
main: try to block signals before executing crash shell
2010-04-13 04:07:19 +02:00
core: always let the kernel reap zombies when we're about to freeze Regardless of whether we're going to spawn a crash shell or not, let the kernel reap zombies. It's more consistent this way.
2015-09-30 17:08:04 +02:00
if (arg_crash_shell) {
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
log_notice("Executing crash shell in 10s...");
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) sleep(10);
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
util-lib: Add sparc64 support for process creation (#3348) The current raw_clone function takes two arguments, the cloning flags and a pointer to the stack for the cloned child. The raw cloning without passing a "thread main" function does not make sense if a new stack is specified, as it returns in both the parent and the child, which will fail in the child as the stack is virgin. All uses of raw_clone indeed pass NULL for the stack pointer which indicates that both processes should share the stack address (so you better don't pass CLONE_VM). This commit refactors the code to not require the caller to pass the stack address, as NULL is the only sensible option. It also adds the magic code needed to make raw_clone work on sparc64, which does not return 0 in %o0 for the child, but indicates the child process by setting %o1 to non-zero. This refactoring is not plain aesthetic, because non-NULL stack addresses need to get mangled before being passed to the clone syscall (you have to apply STACK_BIAS), whereas NULL must not be mangled. Implementing the conditional mangling of the stack address would needlessly complicate the code. raw_clone is moved to a separete header, because the burden of including the assert machinery and sched.h shouldn't be applied to every user of missing_syscalls.h
2016-05-30 02:03:51 +02:00
pid = raw_clone(SIGCHLD);
switch-root: reopen /dev/console before we switch root
2012-07-10 19:19:59 +02:00
if (pid < 0)
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_emergency_errno(errno, "Failed to fork off crash shell: %m");
main: run crash shell as subprocess, so that we can gdb pid 1
2010-04-13 20:26:54 +02:00
else if (pid == 0) {
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
(void) setsid();
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) make_console_stdio();
(void) execle("/bin/sh", "/bin/sh", NULL, environ);
main: run crash shell as subprocess, so that we can gdb pid 1
2010-04-13 20:26:54 +02:00
core: use raw_clone instead of fork in signal handler fork() is not async-signal-safe and calling it from the signal handler could result in a deadlock when at_fork() handlers are called. Using the raw clone() syscall sidesteps that problem. The tricky part is that raise() does not work, since getpid() does not work. Add raw_getpid() to get the real pid, and use kill() instead of raise(). https://bugs.freedesktop.org/show_bug.cgi?id=86604
2014-12-17 05:53:23 +01:00
log_emergency_errno(errno, "execle() failed: %m");
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
_exit(EXIT_FAILURE);
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
} else {
log_info("Spawned crash shell as PID "PID_FMT".", pid);
core: change how crash_shell and crash_reboot interact Instead of freezing in PID1 and letting the forked child freeze or reboot when exec("/bin/sh") fails, just wait for the child's exit and then do the freeze_or_reboot in PID1 as usual. This means that when both crash_shell and crash_reboot are enabled, the system will reboot after the shell exits.
2015-09-30 15:12:19 +02:00
(void) wait_for_terminate(pid, NULL);
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
}
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
}
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
freeze_or_reboot();
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
}
static void install_crash_handler(void) {
core: explain why failing to set up the crash handler is not a real problem http://lists.freedesktop.org/archives/systemd-devel/2015-January/027428.html
2015-01-27 01:47:37 +01:00
static const struct sigaction sa = {
Use initalization instead of explicit zeroing Before, we would initialize many fields twice: first by filling the structure with zeros, and then a second time with the real values. We can let the compiler do the job for us, avoiding one copy. A downside of this patch is that text gets slightly bigger. This is because all zero() calls are effectively inlined: $ size build/.libs/systemd text data bss dec hex filename before 897737 107300 2560 1007597 f5fed build/.libs/systemd after 897873 107300 2560 1007733 f6075 build/.libs/systemd … actually less than 1‰. A few asserts that the parameter is not null had to be removed. I don't think this changes much, because first, it is quite unlikely for the assert to fail, and second, an immediate SEGV is almost as good as an assert.
2013-03-25 00:59:00 +01:00
.sa_handler = crash,
core: explain why failing to set up the crash handler is not a real problem http://lists.freedesktop.org/archives/systemd-devel/2015-January/027428.html
2015-01-27 01:47:37 +01:00
.sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
Use initalization instead of explicit zeroing Before, we would initialize many fields twice: first by filling the structure with zeros, and then a second time with the real values. We can let the compiler do the job for us, avoiding one copy. A downside of this patch is that text gets slightly bigger. This is because all zero() calls are effectively inlined: $ size build/.libs/systemd text data bss dec hex filename before 897737 107300 2560 1007597 f5fed build/.libs/systemd after 897873 107300 2560 1007733 f6075 build/.libs/systemd … actually less than 1‰. A few asserts that the parameter is not null had to be removed. I don't think this changes much, because first, it is quite unlikely for the assert to fail, and second, an immediate SEGV is almost as good as an assert.
2013-03-25 00:59:00 +01:00
};
core: explain why failing to set up the crash handler is not a real problem http://lists.freedesktop.org/archives/systemd-devel/2015-January/027428.html
2015-01-27 01:47:37 +01:00
int r;
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
core: explain why failing to set up the crash handler is not a real problem http://lists.freedesktop.org/archives/systemd-devel/2015-January/027428.html
2015-01-27 01:47:37 +01:00
/* We ignore the return value here, since, we don't mind if we
* cannot set up a crash handler */
r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
if (r < 0)
log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
main: minor code modernization for initializing the console
2014-08-15 18:01:52 +02:00
static int console_setup(void) {
_cleanup_close_ int tty_fd = -1;
int r;
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
main: don't force text mode in console_setup() When systemd starts, plymouth may be already displaying progress graphically. Do not switch the console to text mode at that time. All other users of reset_terminal_fd() do the switch as before. This avoids a graphical glitch with plymouth, especially visible with vesafb, but could be also seen as a sub-second blink with radeon. https://bugzilla.redhat.com/show_bug.cgi?id=785548
2012-01-29 21:55:51 +01:00
tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
treewide: more log_*_errno + return simplifications
2014-11-28 18:23:20 +01:00
if (tty_fd < 0)
return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
main: minor code modernization for initializing the console
2014-08-15 18:01:52 +02:00
/* We don't want to force text mode. plymouth may be showing
* pictures already from initrd. */
main: don't force text mode in console_setup() When systemd starts, plymouth may be already displaying progress graphically. Do not switch the console to text mode at that time. All other users of reset_terminal_fd() do the switch as before. This avoids a graphical glitch with plymouth, especially visible with vesafb, but could be also seen as a sub-second blink with radeon. https://bugzilla.redhat.com/show_bug.cgi?id=785548
2012-01-29 21:55:51 +01:00
r = reset_terminal_fd(tty_fd, false);
treewide: more log_*_errno + return simplifications
2014-11-28 18:23:20 +01:00
if (r < 0)
return log_error_errno(r, "Failed to reset /dev/console: %m");
log: rework logging logic so that we don't keep /dev/console open
2010-05-15 17:25:08 +02:00
main: minor code modernization for initializing the console
2014-08-15 18:01:52 +02:00
return 0;
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
}
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
static int parse_crash_chvt(const char *value) {
int b;
if (safe_atoi(value, &arg_crash_chvt) >= 0)
return 0;
b = parse_boolean(value);
if (b < 0)
return b;
if (b > 0)
arg_crash_chvt = 0; /* switch to where kmsg goes */
else
arg_crash_chvt = -1; /* turn off switching */
return 0;
}
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
static int parse_confirm_spawn(const char *value, char **console) {
char *s;
int r;
r = value ? parse_boolean(value) : 1;
if (r == 0) {
*console = NULL;
return 0;
}
if (r > 0) /* on with default tty */
s = strdup("/dev/console");
else if (is_path(value)) /* on with fully qualified path */
s = strdup(value);
else /* on with only a tty file name, not a fully qualified path */
s = strjoin("/dev/", value);
if (!s)
return -ENOMEM;
*console = s;
return 0;
}
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
static int set_machine_id(const char *m) {
main: make sure set_machine_id() doesn't clobber arg_machine_id on failure
2016-07-21 20:22:42 +02:00
sd_id128_t t;
core: introduce activation rate limiting for socket units This adds two new settings TriggerLimitIntervalSec= and TriggerLimitBurst= that define a rate limit for activation of socket units. When the limit is hit, the socket is is put into a failure mode. This is an alternative fix for #2467, since the original fix resulted in issue #2684. In a later commit the StartLimitInterval=/StartLimitBurst= rate limiter will be changed to be applied after any start conditions checks are made. This way, there are two separate rate limiters enforced: one at triggering time, before any jobs are queued with this patch, as well as the start limit that is moved again to be run immediately before the unit is activated. Condition checks are done in between the two, and thus no longer affect the start limit.
2016-04-26 20:26:15 +02:00
assert(m);
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
main: make sure set_machine_id() doesn't clobber arg_machine_id on failure
2016-07-21 20:22:42 +02:00
if (sd_id128_from_string(m, &t) < 0)
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
return -EINVAL;
main: make sure set_machine_id() doesn't clobber arg_machine_id on failure
2016-07-21 20:22:42 +02:00
if (sd_id128_is_null(t))
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
return -EINVAL;
main: make sure set_machine_id() doesn't clobber arg_machine_id on failure
2016-07-21 20:22:42 +02:00
arg_machine_id = t;
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
return 0;
}
tree-wide: allow state to be passed through to parse_proc_cmdline_item No functional change.
2016-10-22 20:24:52 +02:00
static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
int r;
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
assert(key);
auto-getty: rework auto console getty logic to work in conjunction with single user mode
2010-08-17 03:29:46 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
if (proc_cmdline_value_missing(key, value))
return 0;
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
else if (in_initrd() == !!startswith(key, "rd.")) {
if (free_and_strdup(&arg_default_unit, value) < 0)
return log_oom();
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
r = value ? parse_boolean(value) : true;
Some modernizations
2014-02-16 00:10:36 +01:00
if (r < 0)
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
log_warning("Failed to parse dump core switch %s. Ignoring.", value);
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
else
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
arg_dump_core = r;
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
if (!value)
arg_crash_chvt = 0; /* turn on */
else if (parse_crash_chvt(value) < 0)
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
r = value ? parse_boolean(value) : true;
Some modernizations
2014-02-16 00:10:36 +01:00
if (r < 0)
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
else
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
arg_crash_shell = r;
main: add kernel option to enable confirm_spawn
2010-04-13 19:10:01 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
main: add kernel option to enable confirm_spawn
2010-04-13 19:10:01 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
r = value ? parse_boolean(value) : true;
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
if (r < 0)
log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
main: add kernel option to enable confirm_spawn
2010-04-13 19:10:01 +02:00
else
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
arg_crash_reboot = r;
main: add kernel option to enable confirm_spawn
2010-04-13 19:10:01 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
char *s;
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
r = parse_confirm_spawn(value, &s);
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
if (r < 0)
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
else {
free(arg_confirm_spawn);
arg_confirm_spawn = s;
}
main: switch to primary console vt on crash
2010-04-10 23:36:43 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
if (value) {
r = parse_show_status(value, &arg_show_status);
if (r < 0)
log_warning("Failed to parse show status switch %s. Ignoring.", value);
} else
arg_show_status = SHOW_STATUS_YES;
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
if (proc_cmdline_value_missing(key, value))
return 0;
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
r = exec_output_from_string(value);
Some modernizations
2014-02-16 00:10:36 +01:00
if (r < 0)
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
else
arg_default_std_output = r;
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
if (proc_cmdline_value_missing(key, value))
return 0;
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
r = exec_output_from_string(value);
Some modernizations
2014-02-16 00:10:36 +01:00
if (r < 0)
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
else
arg_default_std_error = r;
systemd: Add systemd.setenv for /proc/cmdline parsing. Check for systemd.setenv when parsing /proc/cmdline. ex: systemd.setenv=PATH=/opt/bin
2012-02-07 21:31:20 +01:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (streq(key, "systemd.setenv")) {
if (proc_cmdline_value_missing(key, value))
return 0;
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
if (env_assignment_is_valid(value)) {
rework systemd's own process environment handling/passing Stop importing non-sensical kernel-exported variables. All parameters in the kernel command line are exported to the initial environment of PID1, but suppressed if they are recognized by kernel built-in code. The EFI booted kernel will add further kernel-internal things which do not belong into userspace. The passed original environ data of the process is not touched and preserved across re-execution, to allow external reading of /proc/self/environ for process properties like container*=.
2013-07-26 05:22:22 +02:00
char **env;
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
env = strv_env_set(arg_default_environment, value);
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
if (!env)
return log_oom();
arg_default_environment = env;
rework systemd's own process environment handling/passing Stop importing non-sensical kernel-exported variables. All parameters in the kernel command line are exported to the initial environment of PID1, but suppressed if they are recognized by kernel built-in code. The EFI booted kernel will add further kernel-internal things which do not belong into userspace. The passed original environ data of the process is not touched and preserved across re-execution, to allow external reading of /proc/self/environ for process properties like container*=.
2013-07-26 05:22:22 +02:00
} else
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
if (proc_cmdline_value_missing(key, value))
return 0;
r = set_machine_id(value);
if (r < 0)
log_warning("MachineID '%s' is not valid. Ignoring.", value);
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
} else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
if (proc_cmdline_value_missing(key, value))
return 0;
r = parse_sec(value, &arg_default_timeout_start_usec);
if (r < 0)
log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
if (arg_default_timeout_start_usec <= 0)
arg_default_timeout_start_usec = USEC_INFINITY;
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
} else if (streq(key, "quiet") && !value) {
log: don't downgrade log level in non-PID 1 if "quiet" is passed on kernel cmdline "debug" should apply to all tools, but "quiet" only to PID1.
2014-06-17 01:05:39 +02:00
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
if (arg_show_status == _SHOW_STATUS_UNSET)
manager: add systemd.show_status=auto mode When set to auto, status will shown when the first ephemeral message is shown (a job has been running for five seconds). Then until the boot or shutdown ends, status messages will be shown. No indication about the switch is done: I think it should be clear for the user that first the cylon eye and the ephemeral messages appear, and afterwards messages are displayed. The initial arming of the event source was still wrong, but now should really be fixed.
2014-01-28 04:27:07 +01:00
arg_show_status = SHOW_STATUS_AUTO;
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
} else if (streq(key, "debug") && !value) {
log: don't downgrade log level in non-PID 1 if "quiet" is passed on kernel cmdline "debug" should apply to all tools, but "quiet" only to PID1.
2014-06-17 01:05:39 +02:00
main,log: parse the log related kernel command line parameters at one place only, and for all tools Previously, we ended up parsing some of them three times: in main.c when processing the kernel cmdline, in main.c when processing the process cmdline (only for containers), and in log.c again. Let's streamline this, and only parse them in log.c In PID 1 also make sure we parse "quiet" first, and then override this with the more specific checks in log.c
2014-08-15 18:07:36 +02:00
/* Note that log_parse_environment() handles 'debug'
* too, and sets the log level to LOG_DEBUG. */
log: don't downgrade log level in non-PID 1 if "quiet" is passed on kernel cmdline "debug" should apply to all tools, but "quiet" only to PID1.
2014-06-17 01:05:39 +02:00
basic: rework virtualization detection API Introduce a proper enum, and don't pass around string ids anymore. This simplifies things quite a bit, and makes virtualization detection more similar to architecture detection.
2015-09-07 13:42:47 +02:00
if (detect_container() > 0)
reduce the amount of messages logged to /dev/kmsg when "debug" is specified
2014-04-05 19:59:01 +02:00
log_set_target(LOG_TARGET_CONSOLE);
util: move more intellegince into parse_proc_cmdline() Already split variable assignments before invoking the callback. And drop "rd." settings if we are not in an initrd.
2014-03-06 17:05:55 +01:00
core: parse `rd.rescue` and `rd.emergency` as initrd-specific shorthands (#3488) Typing `rd.rescue` is easier than `rd.systemd.unit=rescue.target`.
2016-06-13 16:28:42 +02:00
} else if (!value) {
core: use runlevel_to_target for /proc/cmdline parsing
2015-11-03 12:24:52 +01:00
const char *target;
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
/* SysV compatibility */
core: use runlevel_to_target for /proc/cmdline parsing
2015-11-03 12:24:52 +01:00
target = runlevel_to_target(key);
if (target)
return free_and_strdup(&arg_default_unit, target);
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
}
return 0;
}
Report about syntax errors with metadata The information about the unit for which files are being parsed is passed all the way down. This way messages land in the journal with proper UNIT=... or USER_UNIT=... attribution. 'systemctl status' and 'journalctl -u' not displaying those messages has been a source of confusion for users, since the journal entry for a misspelt setting was often logged quite a bit earlier than the failure to start a unit. Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
2013-04-16 04:25:58 +02:00
#define DEFINE_SETTER(name, func, descr) \
static int name(const char *unit, \
const char *filename, \
unsigned line, \
const char *section, \
conf-parser: distinguish between multiple sections with the same name Pass on the line on which a section was decleared to the parsers, so they can distinguish between multiple sections (if they chose to). Currently no parsers take advantage of this, but a follow-up patch will do that to distinguish [Address] Address=192.168.0.1/24 Label=one [Address] Address=192.168.0.2/24 Label=two from [Address] Address=192.168.0.1/24 Label=one Address=192.168.0.2/24 Label=two
2013-11-19 16:17:55 +01:00
unsigned section_line, \
Report about syntax errors with metadata The information about the unit for which files are being parsed is passed all the way down. This way messages land in the journal with proper UNIT=... or USER_UNIT=... attribution. 'systemctl status' and 'journalctl -u' not displaying those messages has been a source of confusion for users, since the journal entry for a misspelt setting was often logged quite a bit earlier than the failure to start a unit. Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
2013-04-16 04:25:58 +02:00
const char *lvalue, \
int ltype, \
const char *rvalue, \
void *data, \
void *userdata) { \
\
int r; \
\
assert(filename); \
assert(lvalue); \
assert(rvalue); \
\
r = func(rvalue); \
if (r < 0) \
core: use %m rather than strerror() where we can
2015-09-23 01:10:47 +02:00
log_syntax(unit, LOG_ERR, filename, line, r, \
"Invalid " descr "'%s': %m", \
rvalue); \
Report about syntax errors with metadata The information about the unit for which files are being parsed is passed all the way down. This way messages land in the journal with proper UNIT=... or USER_UNIT=... attribution. 'systemctl status' and 'journalctl -u' not displaying those messages has been a source of confusion for users, since the journal entry for a misspelt setting was often logged quite a bit earlier than the failure to start a unit. Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
2013-04-16 04:25:58 +02:00
\
return 0; \
}
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
Report about syntax errors with metadata The information about the unit for which files are being parsed is passed all the way down. This way messages land in the journal with proper UNIT=... or USER_UNIT=... attribution. 'systemctl status' and 'journalctl -u' not displaying those messages has been a source of confusion for users, since the journal entry for a misspelt setting was often logged quite a bit earlier than the failure to start a unit. Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
2013-04-16 04:25:58 +02:00
DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
core: move ShowStatus type into the core Let's make the scope of the show-status stuff a bit smaller, and make it private to the core, rather than shared API in shared/.
2014-03-03 21:23:12 +01:00
static int config_parse_cpu_affinity2(
const char *unit,
const char *filename,
unsigned line,
const char *section,
unsigned section_line,
const char *lvalue,
int ltype,
const char *rvalue,
void *data,
void *userdata) {
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
util: Declare a cleanup routine for a cpu_set_t Make use of it in config_parse_cpu_affinity2. Tested by tweaking the `CPUAffinity' setting in /etc/systemd/system.conf and reloading the daemon to confirm it is working as expected. No regressions observed in test cases.
2015-08-31 05:46:27 +02:00
_cleanup_cpu_free_ cpu_set_t *c = NULL;
util: refactor cpu_set parsing into its own function Use the new code in config_parse_cpu_affinity2. Tested by modifying CPUAffinity=... setting in /etc/systemd/system.conf and reloading the daemon, then checking ^Cpus_allowed in /proc/1/status to confirm the correct CPU mask is in place.
2015-09-25 02:58:49 +02:00
int ncpus;
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
util: rename parse_cpu_set() to parse_cpu_set_and_warn() It's pretty untypical for our parsing functions to log on their own. Clarify in the name that this one does.
2015-09-30 20:16:51 +02:00
ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
util: refactor cpu_set parsing into its own function Use the new code in config_parse_cpu_affinity2. Tested by modifying CPUAffinity=... setting in /etc/systemd/system.conf and reloading the daemon, then checking ^Cpus_allowed in /proc/1/status to confirm the correct CPU mask is in place.
2015-09-25 02:58:49 +02:00
if (ncpus < 0)
return ncpus;
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
util: refactor cpu_set parsing into its own function Use the new code in config_parse_cpu_affinity2. Tested by modifying CPUAffinity=... setting in /etc/systemd/system.conf and reloading the daemon, then checking ^Cpus_allowed in /proc/1/status to confirm the correct CPU mask is in place.
2015-09-25 02:58:49 +02:00
if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
tree-wide: when %m is used in log_*, always specify errno explicitly All those uses were correct, but I think it's better to be explicit. Using implicit errno is too error prone, and with this change we can require (in the sense of a style guideline) that the code is always specified. Helpful query: git grep -n -P 'log_[^s][a-z]+\(.*%m'
2017-05-13 17:26:55 +02:00
log_warning_errno(errno, "Failed to set CPU affinity: %m");
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
return 0;
}
core: move ShowStatus type into the core Let's make the scope of the show-status stuff a bit smaller, and make it private to the core, rather than shared API in shared/.
2014-03-03 21:23:12 +01:00
static int config_parse_show_status(
const char* unit,
const char *filename,
unsigned line,
const char *section,
unsigned section_line,
const char *lvalue,
int ltype,
const char *rvalue,
void *data,
void *userdata) {
int k;
ShowStatus *b = data;
assert(filename);
assert(lvalue);
assert(rvalue);
assert(data);
k = parse_show_status(rvalue, b);
if (k < 0) {
tree-wide: clean up log_syntax() usage - Rely everywhere that we use abs() on the error code passed in anyway, thus don't need to explicitly negate what we pass in - Never attach synthetic error number information to log messages. Only log about errors we *receive* with the error number we got there, don't log any synthetic error, that don#t even propagate, but just eat up. - Be more careful with attaching exactly the error we get, instead of errno or unrelated errors randomly. - Fix one occasion where the error number and line number got swapped. - Make sure we never tape over OOM issues, or inability to resolve specifiers
2015-09-30 18:22:42 +02:00
log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
core: move ShowStatus type into the core Let's make the scope of the show-status stuff a bit smaller, and make it private to the core, rather than shared API in shared/.
2014-03-03 21:23:12 +01:00
return 0;
}
return 0;
}
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
static int config_parse_crash_chvt(
const char* unit,
const char *filename,
unsigned line,
const char *section,
unsigned section_line,
const char *lvalue,
int ltype,
const char *rvalue,
void *data,
void *userdata) {
int r;
assert(filename);
assert(lvalue);
assert(rvalue);
r = parse_crash_chvt(rvalue);
if (r < 0) {
log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
return 0;
}
return 0;
}
Report about syntax errors with metadata The information about the unit for which files are being parsed is passed all the way down. This way messages land in the journal with proper UNIT=... or USER_UNIT=... attribution. 'systemctl status' and 'journalctl -u' not displaying those messages has been a source of confusion for users, since the journal entry for a misspelt setting was often logged quite a bit earlier than the failure to start a unit. Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
2013-04-16 04:25:58 +02:00
static int config_parse_join_controllers(const char *unit,
const char *filename,
unsigned line,
const char *section,
conf-parser: distinguish between multiple sections with the same name Pass on the line on which a section was decleared to the parsers, so they can distinguish between multiple sections (if they chose to). Currently no parsers take advantage of this, but a follow-up patch will do that to distinguish [Address] Address=192.168.0.1/24 Label=one [Address] Address=192.168.0.2/24 Label=two from [Address] Address=192.168.0.1/24 Label=one Address=192.168.0.2/24 Label=two
2013-11-19 16:17:55 +01:00
unsigned section_line,
Report about syntax errors with metadata The information about the unit for which files are being parsed is passed all the way down. This way messages land in the journal with proper UNIT=... or USER_UNIT=... attribution. 'systemctl status' and 'journalctl -u' not displaying those messages has been a source of confusion for users, since the journal entry for a misspelt setting was often logged quite a bit earlier than the failure to start a unit. Based-on-a-patch-by: Oleksii Shevchuk <alxchk@gmail.com>
2013-04-16 04:25:58 +02:00
const char *lvalue,
int ltype,
const char *rvalue,
void *data,
void *userdata) {
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
core: Log parse errors in config_parse_join_controllers
2015-09-01 17:30:26 +02:00
const char *whole_rvalue = rvalue;
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
unsigned n = 0;
assert(filename);
assert(lvalue);
assert(rvalue);
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
arg_join_controllers = strv_free_free(arg_join_controllers);
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
core: Use extract_first_word in config_parse_join_controllers Related to the TODO item to replace FOREACH_WORD_QUOTED with it. Tested by setting `JoinControllers=cpu,cpuacct,memory net_cls,blkio' in /etc/systemd/system.conf, rebooting the system with the patched binaries and checking that the desired setup was created by inspecting the entries under /sys/fs/cgroup. No regressions observed in test cases.
2015-08-31 06:00:30 +02:00
for (;;) {
_cleanup_free_ char *word = NULL;
char **l;
int r;
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
tree-wide: drop unneded WHITESPACE param to extract_first_word It's the default, and NULL is shorter.
2016-10-28 02:06:44 +02:00
r = extract_first_word(&rvalue, &word, NULL, EXTRACT_QUOTES);
core: Log parse errors in config_parse_join_controllers
2015-09-01 17:30:26 +02:00
if (r < 0) {
log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
core: Use extract_first_word in config_parse_join_controllers Related to the TODO item to replace FOREACH_WORD_QUOTED with it. Tested by setting `JoinControllers=cpu,cpuacct,memory net_cls,blkio' in /etc/systemd/system.conf, rebooting the system with the patched binaries and checking that the desired setup was created by inspecting the entries under /sys/fs/cgroup. No regressions observed in test cases.
2015-08-31 06:00:30 +02:00
return r;
core: Log parse errors in config_parse_join_controllers
2015-09-01 17:30:26 +02:00
}
core: Use extract_first_word in config_parse_join_controllers Related to the TODO item to replace FOREACH_WORD_QUOTED with it. Tested by setting `JoinControllers=cpu,cpuacct,memory net_cls,blkio' in /etc/systemd/system.conf, rebooting the system with the patched binaries and checking that the desired setup was created by inspecting the entries under /sys/fs/cgroup. No regressions observed in test cases.
2015-08-31 06:00:30 +02:00
if (r == 0)
break;
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
core: Use extract_first_word in config_parse_join_controllers Related to the TODO item to replace FOREACH_WORD_QUOTED with it. Tested by setting `JoinControllers=cpu,cpuacct,memory net_cls,blkio' in /etc/systemd/system.conf, rebooting the system with the patched binaries and checking that the desired setup was created by inspecting the entries under /sys/fs/cgroup. No regressions observed in test cases.
2015-08-31 06:00:30 +02:00
l = strv_split(word, ",");
core: add OOM check in config_parse_join_controllers
2015-09-01 17:31:34 +02:00
if (!l)
core: exit early if we hit OOM
2015-09-23 01:11:08 +02:00
return log_oom();
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
strv_uniq(l);
if (strv_length(l) <= 1) {
strv_free(l);
continue;
}
if (!arg_join_controllers) {
arg_join_controllers = new(char**, 2);
if (!arg_join_controllers) {
strv_free(l);
continue work with error messages, log_oom() Adds messages for formally silent errors: new "Failed on cmdline argument %s: %s". Removes some specific error messages for -ENOMEM in mount-setup.c. A few specific ones have been left in other binaries.
2012-08-04 02:22:09 +02:00
return log_oom();
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
}
arg_join_controllers[0] = l;
arg_join_controllers[1] = NULL;
n = 1;
} else {
char ***a;
char ***t;
t = new0(char**, n+2);
if (!t) {
strv_free(l);
continue work with error messages, log_oom() Adds messages for formally silent errors: new "Failed on cmdline argument %s: %s". Removes some specific error messages for -ENOMEM in mount-setup.c. A few specific ones have been left in other binaries.
2012-08-04 02:22:09 +02:00
return log_oom();
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
}
n = 0;
for (a = arg_join_controllers; *a; a++) {
if (strv_overlap(*a, l)) {
ask-password: add support for caching passwords in the kernel keyring This adds support for caching harddisk passwords in the kernel keyring if it is available, thus supporting caching without Plymouth being around. This is also useful for hooking up "gdm-auto-login" with the collected boot-time harddisk password, in order to support gnome keyring passphrase unlocking via the HDD password, if it is the same. Any passwords added to the kernel keyring this way have a timeout of 2.5min at which time they are purged from the kernel.
2015-10-07 11:26:10 +02:00
if (strv_extend_strv(&l, *a, false) < 0) {
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
strv_free(l);
strv_free_free(t);
continue work with error messages, log_oom() Adds messages for formally silent errors: new "Failed on cmdline argument %s: %s". Removes some specific error messages for -ENOMEM in mount-setup.c. A few specific ones have been left in other binaries.
2012-08-04 02:22:09 +02:00
return log_oom();
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
}
} else {
char **c;
c = strv_copy(*a);
if (!c) {
strv_free(l);
strv_free_free(t);
continue work with error messages, log_oom() Adds messages for formally silent errors: new "Failed on cmdline argument %s: %s". Removes some specific error messages for -ENOMEM in mount-setup.c. A few specific ones have been left in other binaries.
2012-08-04 02:22:09 +02:00
return log_oom();
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
}
t[n++] = c;
}
}
t[n++] = strv_uniq(l);
strv_free_free(arg_join_controllers);
arg_join_controllers = t;
}
}
core: Use extract_first_word in config_parse_join_controllers Related to the TODO item to replace FOREACH_WORD_QUOTED with it. Tested by setting `JoinControllers=cpu,cpuacct,memory net_cls,blkio' in /etc/systemd/system.conf, rebooting the system with the patched binaries and checking that the desired setup was created by inspecting the entries under /sys/fs/cgroup. No regressions observed in test cases.
2015-08-31 06:00:30 +02:00
if (!isempty(rvalue))
tree-wide: clean up log_syntax() usage - Rely everywhere that we use abs() on the error code passed in anyway, thus don't need to explicitly negate what we pass in - Never attach synthetic error number information to log messages. Only log about errors we *receive* with the error number we got there, don't log any synthetic error, that don#t even propagate, but just eat up. - Be more careful with attaching exactly the error we get, instead of errno or unrelated errors randomly. - Fix one occasion where the error number and line number got swapped. - Make sure we never tape over OOM issues, or inability to resolve specifiers
2015-09-30 18:22:42 +02:00
log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
return 0;
}
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
static int parse_config_file(void) {
load-fragment: speed up parsing by using a perfect hash table with configuration settings built by gperf
2011-08-01 00:43:05 +02:00
const ConfigTableItem items[] = {
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "LogLevel", config_parse_level2, 0, NULL },
{ "Manager", "LogTarget", config_parse_target, 0, NULL },
{ "Manager", "LogColor", config_parse_color, 0, NULL },
{ "Manager", "LogLocation", config_parse_location, 0, NULL },
{ "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
{ "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
{ "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
{ "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
{ "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
{ "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
{ "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
{ "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
capabilities: keep bounding set in non-inverted format. Change the capability bounding set parser and logic so that the bounding set is kept as a positive set internally. This means that the set reflects those capabilities that we want to keep instead of drop.
2016-01-07 23:00:04 +01:00
{ "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_SECCOMP
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
seccomp: fix build again if libseccomp is missing
2014-02-13 02:25:45 +01:00
#endif
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
core: add a setting to globally control the default for timer unit accuracy
2014-03-24 16:22:34 +01:00
{ "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
{ "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
{ "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
{ "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
{ "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
core: rename StartLimitInterval= to StartLimitIntervalSec= We generally follow the rule that for time settings we suffix the setting name with "Sec" to indicate the default unit if none is specified. The only exception was the rate limiting interval settings. Fix this, and keep the old names for compatibility. Do the same for journald's RateLimitInterval= setting
2016-04-26 20:46:20 +02:00
{ "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
{ "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{ "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
{ "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
core: move parsing of rlimits into rlimit-util.[ch] This way we can reuse it for parsing rlimit settings in "systemctl set-property" and related commands.
2016-02-01 21:07:09 +01:00
{ "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
{ "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
{ "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
{ "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
{ "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
{ "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
{ "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
{ "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
{ "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
{ "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
{ "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
{ "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
{ "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
{ "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
{ "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
{ "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
core: add global settings for enabling CPUAccounting=, MemoryAccounting=, BlockIOAccounting= for all units at once
2014-02-24 23:50:10 +01:00
{ "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
core: add io controller support on the unified hierarchy On the unified hierarchy, blkio controller is renamed to io and the interface is changed significantly. * blkio.weight and blkio.weight_device are consolidated into io.weight which uses the standardized weight range [1, 10000] with 100 as the default value. * blkio.throttle.{read|write}_{bps|iops}_device are consolidated into io.max. Expansion of throttling features is being worked on to support work-conserving absolute limits (io.low and io.high). * All stats are consolidated into io.stats. This patchset adds support for the new interface. As the interface has been revamped and new features are expected to be added, it seems best to treat it as a separate controller rather than trying to expand the blkio settings although we might add automatic translation if only blkio settings are specified. * io.weight handling is mostly identical to blkio.weight[_device] handling except that the weight range is different. * Both read and write bandwidth settings are consolidated into CGroupIODeviceLimit which describes all limits applicable to the device. This makes it less painful to add new limits. * "max" can be used to specify the maximum limit which is equivalent to no config for max limits and treated as such. If a given CGroupIODeviceLimit doesn't contain any non-default configs, the config struct is discarded once the no limit config is applied to cgroup. * lookup_blkio_device() is renamed to lookup_block_device(). Signed-off-by: Tejun Heo <htejun@fb.com>
2016-05-05 22:42:55 +02:00
{ "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
manager: hook up IP accounting defaults
2017-09-01 16:04:50 +02:00
{ "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
core: add global settings for enabling CPUAccounting=, MemoryAccounting=, BlockIOAccounting= for all units at once
2014-02-24 23:50:10 +01:00
{ "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
{ "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
core: add support for the "pids" cgroup controller This adds support for the new "pids" cgroup controller of 4.3 kernels. It allows accounting the number of tasks in a cgroup and enforcing limits on it. This adds two new setting TasksAccounting= and TasksMax= to each unit, as well as a gloabl option DefaultTasksAccounting=. This also updated "cgtop" to optionally make use of the new kernel-provided accounting. systemctl has been updated to show the number of tasks for each service if it is available. This patch also adds correct support for undoing memory limits for units using a MemoryLimit=infinity syntax. We do the same for TasksMax= now and hence keep things in sync here.
2015-09-10 12:32:16 +02:00
{ "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
core: add new DefaultTasksMax= setting for system.conf This allows initializing the TasksMax= setting of all units by default to some fixed value, instead of leaving it at infinity as before.
2015-11-13 17:13:55 +01:00
{ "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
core: use emergency_action for ctr+alt+del burst Fixes #4306
2016-10-18 12:16:32 +02:00
{ "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
{}
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
};
core: Support system.conf.d and user.conf.d directories in the usual search paths
2014-11-29 10:06:04 +01:00
const char *fn, *conf_dirs_nulstr;
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
fn = arg_system ?
defs: rework CONF_DIRS_NULSTR() macro The macro is generically useful for putting together search paths, hence let's make it truly generic, by dropping the implicit ".d" appending it does, and leave that to the caller. Also rename it from CONF_DIRS_NULSTR() to CONF_PATHS_NULSTR(), since it's not strictly about dirs that way, but any kind of file system path. Also, mark CONF_DIR_SPLIT_USR() as internal macro by renaming it to _CONF_PATHS_SPLIT_USR() so that the leading underscore indicates that it's internal.
2015-11-10 15:57:21 +01:00
PKGSYSCONFDIR "/system.conf" :
PKGSYSCONFDIR "/user.conf";
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
conf_dirs_nulstr = arg_system ?
defs: rework CONF_DIRS_NULSTR() macro The macro is generically useful for putting together search paths, hence let's make it truly generic, by dropping the implicit ".d" appending it does, and leave that to the caller. Also rename it from CONF_DIRS_NULSTR() to CONF_PATHS_NULSTR(), since it's not strictly about dirs that way, but any kind of file system path. Also, mark CONF_DIR_SPLIT_USR() as internal macro by renaming it to _CONF_PATHS_SPLIT_USR() so that the leading underscore indicates that it's internal.
2015-11-10 15:57:21 +01:00
CONF_PATHS_NULSTR("systemd/system.conf.d") :
CONF_PATHS_NULSTR("systemd/user.conf.d");
tree-wide: rename config_parse_many to …_nulstr In preparation for adding a version which takes a strv.
2016-09-10 12:02:40 +02:00
config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
core: rework unit timeout handling, and add new setting RuntimeMaxSec= This clean-ups timeout handling in PID 1. Specifically, instead of storing 0 in internal timeout variables as indication for a disabled timeout, use USEC_INFINITY which is in-line with how we do this in the rest of our code (following the logic that 0 means "no", and USEC_INFINITY means "never"). This also replace all usec_t additions with invocations to usec_add(), so that USEC_INFINITY is properly propagated, and sd-event considers it has indication for turning off the event source. This also alters the deserialization of the units to restart timeouts from the time they were originally started from. Before this patch timeouts would be restarted beginning with the time of the deserialization, which could lead to artificially prolonged timeouts if a daemon reload took place. Finally, a new RuntimeMaxSec= setting is introduced for service units, that specifies a maximum runtime after which a specific service is forcibly terminated. This is useful to put time limits on time-intensive processing jobs. This also simplifies the various xyz_spawn() calls of the various types in that explicit distruction of the timers is removed, as that is done anyway by the state change handlers, and a state change is always done when the xyz_spawn() calls fail. Fixes: #2249
2016-02-01 21:48:10 +01:00
/* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
* like everywhere else. */
if (arg_default_timeout_start_usec <= 0)
arg_default_timeout_start_usec = USEC_INFINITY;
if (arg_default_timeout_stop_usec <= 0)
arg_default_timeout_stop_usec = USEC_INFINITY;
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
return 0;
}
Reload manager defaults at daemon-reload "systemctl daemon-reload" should also update the manager defaults from /etc/systemd/system.conf. For details, see: http://lists.freedesktop.org/archives/systemd-devel/2015-June/033062.html Amended to use manager_set_defaults() as common function.
2015-06-29 11:26:27 +02:00
static void manager_set_defaults(Manager *m) {
assert(m);
m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
m->default_std_output = arg_default_std_output;
m->default_std_error = arg_default_std_error;
m->default_timeout_start_usec = arg_default_timeout_start_usec;
m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
m->default_restart_usec = arg_default_restart_usec;
m->default_start_limit_interval = arg_default_start_limit_interval;
m->default_start_limit_burst = arg_default_start_limit_burst;
m->default_cpu_accounting = arg_default_cpu_accounting;
core: add io controller support on the unified hierarchy On the unified hierarchy, blkio controller is renamed to io and the interface is changed significantly. * blkio.weight and blkio.weight_device are consolidated into io.weight which uses the standardized weight range [1, 10000] with 100 as the default value. * blkio.throttle.{read|write}_{bps|iops}_device are consolidated into io.max. Expansion of throttling features is being worked on to support work-conserving absolute limits (io.low and io.high). * All stats are consolidated into io.stats. This patchset adds support for the new interface. As the interface has been revamped and new features are expected to be added, it seems best to treat it as a separate controller rather than trying to expand the blkio settings although we might add automatic translation if only blkio settings are specified. * io.weight handling is mostly identical to blkio.weight[_device] handling except that the weight range is different. * Both read and write bandwidth settings are consolidated into CGroupIODeviceLimit which describes all limits applicable to the device. This makes it less painful to add new limits. * "max" can be used to specify the maximum limit which is equivalent to no config for max limits and treated as such. If a given CGroupIODeviceLimit doesn't contain any non-default configs, the config struct is discarded once the no limit config is applied to cgroup. * lookup_blkio_device() is renamed to lookup_block_device(). Signed-off-by: Tejun Heo <htejun@fb.com>
2016-05-05 22:42:55 +02:00
m->default_io_accounting = arg_default_io_accounting;
manager: hook up IP accounting defaults
2017-09-01 16:04:50 +02:00
m->default_ip_accounting = arg_default_ip_accounting;
Reload manager defaults at daemon-reload "systemctl daemon-reload" should also update the manager defaults from /etc/systemd/system.conf. For details, see: http://lists.freedesktop.org/archives/systemd-devel/2015-June/033062.html Amended to use manager_set_defaults() as common function.
2015-06-29 11:26:27 +02:00
m->default_blockio_accounting = arg_default_blockio_accounting;
m->default_memory_accounting = arg_default_memory_accounting;
core: add support for the "pids" cgroup controller This adds support for the new "pids" cgroup controller of 4.3 kernels. It allows accounting the number of tasks in a cgroup and enforcing limits on it. This adds two new setting TasksAccounting= and TasksMax= to each unit, as well as a gloabl option DefaultTasksAccounting=. This also updated "cgtop" to optionally make use of the new kernel-provided accounting. systemctl has been updated to show the number of tasks for each service if it is available. This patch also adds correct support for undoing memory limits for units using a MemoryLimit=infinity syntax. We do the same for TasksMax= now and hence keep things in sync here.
2015-09-10 12:32:16 +02:00
m->default_tasks_accounting = arg_default_tasks_accounting;
core: add new DefaultTasksMax= setting for system.conf This allows initializing the TasksMax= setting of all units by default to some fixed value, instead of leaving it at infinity as before.
2015-11-13 17:13:55 +01:00
m->default_tasks_max = arg_default_tasks_max;
Reload manager defaults at daemon-reload "systemctl daemon-reload" should also update the manager defaults from /etc/systemd/system.conf. For details, see: http://lists.freedesktop.org/archives/systemd-devel/2015-June/033062.html Amended to use manager_set_defaults() as common function.
2015-06-29 11:26:27 +02:00
manager_set_default_rlimits(m, arg_default_rlimit);
manager_environment_add(m, NULL, arg_default_environment);
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
static int parse_argv(int argc, char *argv[]) {
enum {
ARG_LOG_LEVEL = 0x100,
ARG_LOG_TARGET,
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
ARG_LOG_COLOR,
ARG_LOG_LOCATION,
main: rename systemd.default= to systemd.unit=
2010-06-09 15:38:46 +02:00
ARG_UNIT,
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
ARG_SYSTEM,
drop support for MANAGER_SESSION, introduce MANAGER_USER instead
2010-11-15 22:12:41 +01:00
ARG_USER,
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
ARG_TEST,
systemd: use pager for --test and --help
2014-01-07 06:00:05 +01:00
ARG_NO_PAGER,
systemd: add --version option systemd --version mirrors systemctl --version: $ ./systemd --version systemd 186 other +PAM +LIBWRAP +AUDIT +SELINUX +IMA +SYSVINIT -LIBCRYPTSETUP This information can be retrieved by other means (systemctl, etc.), but it's easier for a newbie if 'systemd --version' says something useful. And 'systemd --help' is already there, so let's complement that with '--version'.
2012-07-17 07:31:47 +02:00
ARG_VERSION,
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
ARG_DUMP_CONFIGURATION_ITEMS,
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
ARG_DUMP_CORE,
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
ARG_CRASH_CHVT,
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
ARG_CRASH_SHELL,
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
ARG_CRASH_REBOOT,
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
ARG_CONFIRM_SPAWN,
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
ARG_SHOW_STATUS,
dbus: automatically generate and install introspection files
2010-05-23 03:45:33 +02:00
ARG_DESERIALIZE,
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
ARG_SWITCHED_ROOT,
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
ARG_DEFAULT_STD_OUTPUT,
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
ARG_DEFAULT_STD_ERROR,
ARG_MACHINE_ID
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
};
static const struct option options[] = {
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
{ "log-level", required_argument, NULL, ARG_LOG_LEVEL },
{ "log-target", required_argument, NULL, ARG_LOG_TARGET },
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
{ "log-color", optional_argument, NULL, ARG_LOG_COLOR },
{ "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
main: rename systemd.default= to systemd.unit=
2010-06-09 15:38:46 +02:00
{ "unit", required_argument, NULL, ARG_UNIT },
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
{ "system", no_argument, NULL, ARG_SYSTEM },
drop support for MANAGER_SESSION, introduce MANAGER_USER instead
2010-11-15 22:12:41 +01:00
{ "user", no_argument, NULL, ARG_USER },
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
{ "test", no_argument, NULL, ARG_TEST },
systemd: use pager for --test and --help
2014-01-07 06:00:05 +01:00
{ "no-pager", no_argument, NULL, ARG_NO_PAGER },
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
{ "help", no_argument, NULL, 'h' },
systemd: add --version option systemd --version mirrors systemctl --version: $ ./systemd --version systemd 186 other +PAM +LIBWRAP +AUDIT +SELINUX +IMA +SYSVINIT -LIBCRYPTSETUP This information can be retrieved by other means (systemctl, etc.), but it's easier for a newbie if 'systemd --version' says something useful. And 'systemd --help' is already there, so let's complement that with '--version'.
2012-07-17 07:31:47 +02:00
{ "version", no_argument, NULL, ARG_VERSION },
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
{ "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
main: simplify+unify logic for parsing runtime booleans
2012-05-08 23:36:55 +02:00
{ "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
{ "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
main: simplify+unify logic for parsing runtime booleans
2012-05-08 23:36:55 +02:00
{ "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
{ "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
main: simplify+unify logic for parsing runtime booleans
2012-05-08 23:36:55 +02:00
{ "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
service: hide output of sysv scripts if quiet is passed on the kernel cmdline
2010-08-09 18:00:24 +02:00
{ "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
{ "deserialize", required_argument, NULL, ARG_DESERIALIZE },
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
{ "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
{ "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
{ "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
{ "machine-id", required_argument, NULL, ARG_MACHINE_ID },
Some modernizations
2014-02-16 00:10:36 +01:00
{}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
};
int c, r;
assert(argc >= 1);
assert(argv);
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1)
main: don't parse /proc/cmdline in containers
2011-03-14 04:08:12 +01:00
opterr = 0;
main: properly handle -b boot option
2011-03-10 23:01:42 +01:00
while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
switch (c) {
case ARG_LOG_LEVEL:
Some modernizations
2014-02-16 00:10:36 +01:00
r = log_set_max_level_from_string(optarg);
if (r < 0) {
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
log_error("Failed to parse log level %s.", optarg);
return r;
}
break;
case ARG_LOG_TARGET:
Some modernizations
2014-02-16 00:10:36 +01:00
r = log_set_target_from_string(optarg);
if (r < 0) {
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
log_error("Failed to parse log target %s.", optarg);
return r;
}
break;
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
case ARG_LOG_COLOR:
main: don't segfault when --log-color is passed without parameter
2010-06-18 23:13:15 +02:00
if (optarg) {
Some modernizations
2014-02-16 00:10:36 +01:00
r = log_show_color_from_string(optarg);
if (r < 0) {
main: don't segfault when --log-color is passed without parameter
2010-06-18 23:13:15 +02:00
log_error("Failed to parse log color setting %s.", optarg);
return r;
}
} else
log_show_color(true);
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
break;
case ARG_LOG_LOCATION:
main: don't segfault when --log-color is passed without parameter
2010-06-18 23:13:15 +02:00
if (optarg) {
Some modernizations
2014-02-16 00:10:36 +01:00
r = log_show_location_from_string(optarg);
if (r < 0) {
main: don't segfault when --log-color is passed without parameter
2010-06-18 23:13:15 +02:00
log_error("Failed to parse log location setting %s.", optarg);
return r;
}
} else
log_show_location(true);
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
break;
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
case ARG_DEFAULT_STD_OUTPUT:
Some modernizations
2014-02-16 00:10:36 +01:00
r = exec_output_from_string(optarg);
if (r < 0) {
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
log_error("Failed to parse default standard output setting %s.", optarg);
return r;
} else
arg_default_std_output = r;
break;
case ARG_DEFAULT_STD_ERROR:
Some modernizations
2014-02-16 00:10:36 +01:00
r = exec_output_from_string(optarg);
if (r < 0) {
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
log_error("Failed to parse default standard error output setting %s.", optarg);
return r;
} else
arg_default_std_error = r;
break;
main: rename systemd.default= to systemd.unit=
2010-06-09 15:38:46 +02:00
case ARG_UNIT:
core: remove set_default_unit() The new free_and_strdup() call does pretty much the same thing these days, no need to keep a private limited purpose version around.
2015-09-23 01:14:34 +02:00
r = free_and_strdup(&arg_default_unit, optarg);
treewide: more log_*_errno + return simplifications
2014-11-28 18:23:20 +01:00
if (r < 0)
return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
break;
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
case ARG_SYSTEM:
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
arg_system = true;
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
break;
manager: make running_as configurable
2010-04-06 23:55:42 +02:00
drop support for MANAGER_SESSION, introduce MANAGER_USER instead
2010-11-15 22:12:41 +01:00
case ARG_USER:
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
arg_system = false;
manager: make running_as configurable
2010-04-06 23:55:42 +02:00
break;
main: implement --test
2010-04-07 00:10:17 +02:00
case ARG_TEST:
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
arg_action = ACTION_TEST;
systemd: use pager for --test and --help
2014-01-07 06:00:05 +01:00
break;
case ARG_NO_PAGER:
arg_no_pager = true;
main: implement --test
2010-04-07 00:10:17 +02:00
break;
systemd: add --version option systemd --version mirrors systemctl --version: $ ./systemd --version systemd 186 other +PAM +LIBWRAP +AUDIT +SELINUX +IMA +SYSVINIT -LIBCRYPTSETUP This information can be retrieved by other means (systemctl, etc.), but it's easier for a newbie if 'systemd --version' says something useful. And 'systemd --help' is already there, so let's complement that with '--version'.
2012-07-17 07:31:47 +02:00
case ARG_VERSION:
arg_action = ACTION_VERSION;
break;
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
case ARG_DUMP_CONFIGURATION_ITEMS:
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
break;
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
case ARG_DUMP_CORE:
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
if (!optarg)
arg_dump_core = true;
else {
r = parse_boolean(optarg);
if (r < 0)
return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
arg_dump_core = r;
main: simplify+unify logic for parsing runtime booleans
2012-05-08 23:36:55 +02:00
}
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
break;
case ARG_CRASH_CHVT:
r = parse_crash_chvt(optarg);
if (r < 0)
return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
break;
case ARG_CRASH_SHELL:
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
if (!optarg)
arg_crash_shell = true;
else {
r = parse_boolean(optarg);
if (r < 0)
return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
arg_crash_shell = r;
}
break;
case ARG_CRASH_REBOOT:
if (!optarg)
arg_crash_reboot = true;
else {
r = parse_boolean(optarg);
if (r < 0)
return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
arg_crash_reboot = r;
main: simplify+unify logic for parsing runtime booleans
2012-05-08 23:36:55 +02:00
}
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
break;
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
case ARG_CONFIRM_SPAWN:
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
arg_confirm_spawn = mfree(arg_confirm_spawn);
r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
if (r < 0)
return log_error_errno(r, "Failed to parse confirm spawn option: %m");
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
break;
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
case ARG_SHOW_STATUS:
manager: add systemd.show_status=auto mode When set to auto, status will shown when the first ephemeral message is shown (a job has been running for five seconds). Then until the boot or shutdown ends, status messages will be shown. No indication about the switch is done: I think it should be clear for the user that first the cylon eye and the ephemeral messages appear, and afterwards messages are displayed. The initial arming of the event source was still wrong, but now should really be fixed.
2014-01-28 04:27:07 +01:00
if (optarg) {
r = parse_show_status(optarg, &arg_show_status);
if (r < 0) {
log_error("Failed to parse show status boolean %s.", optarg);
return r;
}
} else
arg_show_status = SHOW_STATUS_YES;
service: hide output of sysv scripts if quiet is passed on the kernel cmdline
2010-08-09 18:00:24 +02:00
break;
main: simplify+unify logic for parsing runtime booleans
2012-05-08 23:36:55 +02:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
case ARG_DESERIALIZE: {
int fd;
FILE *f;
socket: support socket activation of containers
2012-12-22 19:30:07 +01:00
r = safe_atoi(optarg, &fd);
if (r < 0 || fd < 0) {
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
log_error("Failed to parse deserialize option %s.", optarg);
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
return -EINVAL;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
}
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
(void) fd_cloexec(fd, true);
socket: support socket activation of containers
2012-12-22 19:30:07 +01:00
f = fdopen(fd, "r");
treewide: another round of simplifications Using the same scripts as in f647962d64e "treewide: yet more log_*_errno + return simplifications".
2014-11-28 19:57:32 +01:00
if (!f)
return log_error_errno(errno, "Failed to open serialization fd: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
util: introduce safe_fclose() and port everything over to it Adds a coccinelle script to port things over automatically.
2015-09-09 15:20:10 +02:00
safe_fclose(arg_serialization);
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
arg_serialization = f;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
break;
}
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
case ARG_SWITCHED_ROOT:
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
arg_switched_root = true;
core/main.c: add "--switchedroot" parameter If systemd serializes from a switch_root, it adds "--switchedroot" to the systemd in the real root. If "--switchedroot" is found, then we do not skip all the stuff, which is skipped for normal rexecs.
2012-05-16 14:22:42 +02:00
break;
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
case ARG_MACHINE_ID:
r = set_machine_id(optarg);
main: minor simplification
2016-08-26 17:25:08 +02:00
if (r < 0)
return log_error_errno(r, "MachineID '%s' is not valid.", optarg);
core: Add machine-id setting Allow for overriding all other machine-ids which may be present on the system using a kernel command line systemd.machine_id or --machine-id= option. This is especially useful for network booted systems where the machine-id needs to be static, or for containers where a specific machine-id is wanted.
2015-07-06 00:00:59 +02:00
break;
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
case 'h':
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
arg_action = ACTION_HELP;
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
break;
main: introduce -D as quick acess to debugging
2010-07-13 20:06:29 +02:00
case 'D':
log_set_max_level(LOG_DEBUG);
break;
main: properly handle -b boot option
2011-03-10 23:01:42 +01:00
case 'b':
case 's':
case 'z':
/* Just to eat away the sysvinit kernel
* cmdline args without getopt() error
* messages that we'll parse in
* parse_proc_cmdline_word() or ignore. */
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
main: properly handle -b boot option
2011-03-10 23:01:42 +01:00
case '?':
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() != 1)
main: properly handle -b boot option
2011-03-10 23:01:42 +01:00
return -EINVAL;
Unify parse_argv style getopt is usually good at printing out a nice error message when commandline options are invalid. It distinguishes between an unknown option and a known option with a missing arg. It is better to let it do its job and not use opterr=0 unless we actually want to suppress messages. So remove opterr=0 in the few places where it wasn't really useful. When an error in options is encountered, we should not print a lengthy help() and overwhelm the user, when we know precisely what is wrong with the commandline. In addition, since help() prints to stdout, it should not be used except when requested with -h or --help. Also, simplify things here and there.
2014-08-02 17:12:21 +02:00
else
return 0;
main: properly handle -b boot option
2011-03-10 23:01:42 +01:00
Unify parse_argv style getopt is usually good at printing out a nice error message when commandline options are invalid. It distinguishes between an unknown option and a known option with a missing arg. It is better to let it do its job and not use opterr=0 unless we actually want to suppress messages. So remove opterr=0 in the few places where it wasn't really useful. When an error in options is encountered, we should not print a lengthy help() and overwhelm the user, when we know precisely what is wrong with the commandline. In addition, since help() prints to stdout, it should not be used except when requested with -h or --help. Also, simplify things here and there.
2014-08-02 17:12:21 +02:00
default:
assert_not_reached("Unhandled option code.");
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
}
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (optind < argc && getpid_cached() != 1) {
main: interpret all argv[] arguments unconditionally when run in a container
2011-03-16 03:35:59 +01:00
/* Hmm, when we aren't run as init system
* let's complain about excess arguments */
log_error("Excess arguments.");
return -EINVAL;
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
return 0;
}
static int help(void) {
minor fixes to help texts
2010-06-24 00:08:42 +02:00
printf("%s [OPTIONS...]\n\n"
drop support for MANAGER_SESSION, introduce MANAGER_USER instead
2010-11-15 22:12:41 +01:00
"Starts up and maintains the system or user services.\n\n"
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
" -h --help Show this help\n"
add version argument to help function (#6377) Signed-off-by: Stefan Pietsch <mail.ipv4v6+gh@gmail.com>
2017-07-15 13:53:21 +02:00
" --version Show version\n"
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
" --test Determine startup sequence, dump it and exit\n"
systemd: use pager for --test and --help
2014-01-07 06:00:05 +01:00
" --no-pager Do not pipe output into a pager\n"
rework tty handling We now make sure to run all services in their own session, possibly with a controlling terminal. This also extends the service and socket state machines a little.
2010-04-13 02:06:27 +02:00
" --dump-configuration-items Dump understood unit configuration items\n"
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
" --unit=UNIT Set default unit\n"
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
" --system Run a system instance, even if PID != 1\n"
drop support for MANAGER_SESSION, introduce MANAGER_USER instead
2010-11-15 22:12:41 +01:00
" --user Run a user instance\n"
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
" --dump-core[=BOOL] Dump core on crash\n"
" --crash-vt=NR Change to specified VT on crash\n"
" --crash-reboot[=BOOL] Reboot on crash\n"
" --crash-shell[=BOOL] Run shell on crash\n"
" --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
" --show-status[=BOOL] Show status updates on the console during bootup\n"
log: never ever log to syslog from PID 1, log to the journal again We don't support journal-less systems anyway, so let's avoid the confusion.
2014-08-11 20:08:08 +02:00
" --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
" --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
" --log-color[=BOOL] Highlight important log messages\n"
" --log-location[=BOOL] Include code location in log messages\n"
exec: introduce global defaults for the standard output of services
2011-02-15 11:52:29 +01:00
" --default-standard-output= Set default standard output for services\n"
" --default-standard-error= Set default standard error output for services\n",
service: optionally call into PAM when dropping priviliges
2010-06-16 21:54:17 +02:00
program_invocation_short_name);
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
return 0;
}
Do not serialize environment, when switching root When switching root, i.e. LANG can be set to the locale of the initramfs or "C", if it was unset. When systemd deserializes LANG in the real root this would overwrite the setting previously gathered by locale_set(). To reproduce, boot with an initramfs without locale.conf or change /etc/locale.conf to a different language than the initramfs and check a daemon started by systemd: $ tr "$\000" '\n' </proc/$(pidof sshd)/environ | grep LANG LANG=C To prevent that, serialization of environment variables is skipped, when serializing for switching root. https://bugzilla.redhat.com/show_bug.cgi?id=949525
2013-04-08 14:05:24 +02:00
static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
_cleanup_fdset_free_ FDSet *fds = NULL;
_cleanup_fclose_ FILE *f = NULL;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
int r;
assert(m);
assert(_f);
assert(_fds);
unit: don't serialize job state, only unit state across switch-root
2012-07-18 01:46:52 +02:00
r = manager_open_serialization(m, &f);
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
if (r < 0)
return log_error_errno(r, "Failed to create serialization file: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
core: send out "Reloading" signal before and after doing a full reload/reexec of PID 1 Since we'll unload all units/job during a reload, and then readd them it is really useful for clients to be aware of this phase hence sent a signal out before and after. This signal is called "Reloading" (despite the fact that it is also sent out during reexecution, which we consider a special case in this context) and has one boolean parameter which is true for the signal sent before the reload, and false for the signal after the reload. The UnitRemoved/JobRremoved and UnitNew/JobNew due to the reloading are guranteed to be between the pair of Reloading messages.
2013-07-10 21:10:53 +02:00
/* Make sure nothing is really destructed when we shut down */
tree-wide: make ++/-- usage consistent WRT spacing Throughout the tree there's spurious use of spaces separating ++ and -- operators from their respective operands. Make ++ and -- operator consistent with the majority of existing uses; discard the spaces.
2016-02-23 05:32:04 +01:00
m->n_reloading++;
core: convert PID 1 to libsystemd-bus This patch converts PID 1 to libsystemd-bus and thus drops the dependency on libdbus. The only remaining code using libdbus is a test case that validates our bus marshalling against libdbus' marshalling, and this dependency can be turned off. This patch also adds a couple of things to libsystem-bus, that are necessary to make the port work: - Synthesizing of "Disconnected" messages when bus connections are severed. - Support for attaching multiple vtables for the same interface on the same path. This patch also fixes the SetDefaultTarget() and GetDefaultTarget() bus calls which used an inappropriate signature. As a side effect we will now generate PropertiesChanged messages which carry property contents, rather than just invalidation information.
2013-11-19 21:12:59 +01:00
bus_manager_send_reloading(m, true);
core: send out "Reloading" signal before and after doing a full reload/reexec of PID 1 Since we'll unload all units/job during a reload, and then readd them it is really useful for clients to be aware of this phase hence sent a signal out before and after. This signal is called "Reloading" (despite the fact that it is also sent out during reexecution, which we consider a special case in this context) and has one boolean parameter which is true for the signal sent before the reload, and false for the signal after the reload. The UnitRemoved/JobRremoved and UnitNew/JobNew due to the reloading are guranteed to be between the pair of Reloading messages.
2013-07-10 21:10:53 +02:00
unit: don't serialize job state, only unit state across switch-root
2012-07-18 01:46:52 +02:00
fds = fdset_new();
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
if (!fds)
return log_oom();
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
Do not serialize environment, when switching root When switching root, i.e. LANG can be set to the locale of the initramfs or "C", if it was unset. When systemd deserializes LANG in the real root this would overwrite the setting previously gathered by locale_set(). To reproduce, boot with an initramfs without locale.conf or change /etc/locale.conf to a different language than the initramfs and check a daemon started by systemd: $ tr "$\000" '\n' </proc/$(pidof sshd)/environ | grep LANG LANG=C To prevent that, serialization of environment variables is skipped, when serializing for switching root. https://bugzilla.redhat.com/show_bug.cgi?id=949525
2013-04-08 14:05:24 +02:00
r = manager_serialize(m, f, fds, switching_root);
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
if (r < 0)
return log_error_errno(r, "Failed to serialize state: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
return log_error_errno(errno, "Failed to rewind serialization fd: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
unit: don't serialize job state, only unit state across switch-root
2012-07-18 01:46:52 +02:00
r = fd_cloexec(fileno(f), false);
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
if (r < 0)
return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
unit: don't serialize job state, only unit state across switch-root
2012-07-18 01:46:52 +02:00
r = fdset_cloexec(fds, false);
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
if (r < 0)
return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
*_f = f;
*_fds = fds;
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
f = NULL;
fds = NULL;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
core: modernize prepare_reexecute()
2015-09-23 01:11:30 +02:00
return 0;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
}
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
struct rlimit nl;
int r;
main: improve RLIMIT_NOFILE handling (#5795) This has systemd look at /proc/sys/fs/nr_open to find the current maximum of open files compiled into the kernel and tries to set the RLIMIT_NOFILE max to it. This has the advantage the value chosen as limit is less arbitrary and also improves the behavior of systemd in containers that have an rlimit set: When systemd currently starts in a container that has RLIMIT_NOFILE set to e.g. 100000 systemd will lower it to 65536. With this patch systemd will try to set the nofile limit to the allowed kernel maximum. If this fails, it will compute the minimum of the current set value (the limit that is set on the container) and the maximum value as soft limit and the currently set maximum value as the maximum value. This way it retains the limit set on the container.
2017-04-26 06:18:10 +02:00
int min_max;
_cleanup_free_ char *nr_open = NULL;
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
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. */
treewide: another round of simplifications Using the same scripts as in f647962d64e "treewide: yet more log_*_errno + return simplifications".
2014-11-28 19:57:32 +01:00
if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
pid1: downgrade some rlimit warnings Since we ignore the result anyway, downgrade errors to warning. log_oom() will still emit an error, but that's mostly theoretical, so it is not worth complicating the code to avoid the small inconsistency
2016-10-18 19:38:41 +02:00
return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
/* 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;
}
main: improve RLIMIT_NOFILE handling (#5795) This has systemd look at /proc/sys/fs/nr_open to find the current maximum of open files compiled into the kernel and tries to set the RLIMIT_NOFILE max to it. This has the advantage the value chosen as limit is less arbitrary and also improves the behavior of systemd in containers that have an rlimit set: When systemd currently starts in a container that has RLIMIT_NOFILE set to e.g. 100000 systemd will lower it to 65536. With this patch systemd will try to set the nofile limit to the allowed kernel maximum. If this fails, it will compute the minimum of the current set value (the limit that is set on the container) and the maximum value as soft limit and the currently set maximum value as the maximum value. This way it retains the limit set on the container.
2017-04-26 06:18:10 +02:00
/* Get current RLIMIT_NOFILE maximum compiled into the kernel. */
r = read_one_line_file("/proc/sys/fs/nr_open", &nr_open);
if (r == 0)
r = safe_atoi(nr_open, &min_max);
/* If we fail, fallback to the hard-coded kernel limit of 1024 * 1024. */
if (r < 0)
min_max = 1024 * 1024;
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
/* Bump up the resource limit for ourselves substantially */
main: improve RLIMIT_NOFILE handling (#5795) This has systemd look at /proc/sys/fs/nr_open to find the current maximum of open files compiled into the kernel and tries to set the RLIMIT_NOFILE max to it. This has the advantage the value chosen as limit is less arbitrary and also improves the behavior of systemd in containers that have an rlimit set: When systemd currently starts in a container that has RLIMIT_NOFILE set to e.g. 100000 systemd will lower it to 65536. With this patch systemd will try to set the nofile limit to the allowed kernel maximum. If this fails, it will compute the minimum of the current set value (the limit that is set on the container) and the maximum value as soft limit and the currently set maximum value as the maximum value. This way it retains the limit set on the container.
2017-04-26 06:18:10 +02:00
nl.rlim_cur = nl.rlim_max = min_max;
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
r = setrlimit_closest(RLIMIT_NOFILE, &nl);
treewide: more log_*_errno + return simplifications
2014-11-28 18:23:20 +01:00
if (r < 0)
pid1: downgrade some rlimit warnings Since we ignore the result anyway, downgrade errors to warning. log_oom() will still emit an error, but that's mostly theoretical, so it is not worth complicating the code to avoid the small inconsistency
2016-10-18 19:38:41 +02:00
return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
return 0;
}
main: bump RLIMIT_NOFILE for the root user substantially On current kernels BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK even for privileged users that have CAP_IPC_LOCK. Given that mlock() generally ignores RLIMIT_MEMLOCK if CAP_IPC_LOCK is set this appears to be an oversight in the kernel. Either way, until that's fixed, let's just bump RLIMIT_MEMLOCK for the root user considerably, as the default is quite limiting, and doesn't permit us to create more than a few TRIE maps.
2017-09-21 19:43:07 +02:00
static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
int r;
assert(saved_rlimit);
assert(getuid() == 0);
/* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even though we have CAP_IPC_LOCK which
* should normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's
* bump the value high enough for the root user. */
if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*16ULL));
if (r < 0)
return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
return 0;
}
main: print warning if /usr is on a seperate partition
2011-02-23 01:10:20 +01:00
static void test_usr(void) {
s/seperate/separate/
2011-03-04 03:58:52 +01:00
/* Check that /usr is not a separate fs */
main: print warning if /usr is on a seperate partition
2011-02-23 01:10:20 +01:00
taint: add missing cgroups taint flag
2011-03-30 02:12:46 +02:00
if (dir_is_empty("/usr") <= 0)
return;
core: fix typo in log message
2015-01-05 14:12:47 +01:00
log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
taint: add missing cgroups taint flag
2011-03-30 02:12:46 +02:00
"Some things will probably break (sometimes even silently) in mysterious ways. "
"Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
}
main: don't try to mout cpuset with cpu+cpuacct anymore Turns out cpuset needs explicit initialization before we could make use of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to just create a group in "cpu" and start it.
2012-09-24 11:35:51 +02:00
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
tree-wide: spelling fixes https://github.com/vlajos/misspell_fixer https://github.com/torstehu/systemd/commit/b6fdeb618cf2f3ce1645b3315f15f482710c7ffa Thanks to Torstein Husebo <torstein@huseboe.net>.
2014-12-29 10:45:58 +01:00
* "cpuset" doesn't really work for groups with no initialized
main: don't try to mout cpuset with cpu+cpuacct anymore Turns out cpuset needs explicit initialization before we could make use of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to just create a group in "cpu" and start it.
2012-09-24 11:35:51 +02:00
* attributes. */
arg_join_controllers = new(char**, 3);
if (!arg_join_controllers)
return -ENOMEM;
arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
if (!arg_join_controllers[0])
goto oom;
Small cleanup
2013-04-23 14:28:10 +02:00
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
if (!arg_join_controllers[1])
goto oom;
main: don't try to mout cpuset with cpu+cpuacct anymore Turns out cpuset needs explicit initialization before we could make use of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to just create a group in "cpu" and start it.
2012-09-24 11:35:51 +02:00
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
arg_join_controllers[2] = NULL;
main: don't try to mout cpuset with cpu+cpuacct anymore Turns out cpuset needs explicit initialization before we could make use of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to just create a group in "cpu" and start it.
2012-09-24 11:35:51 +02:00
return 0;
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
oom:
arg_join_controllers = strv_free_free(arg_join_controllers);
return -ENOMEM;
main: don't try to mout cpuset with cpu+cpuacct anymore Turns out cpuset needs explicit initialization before we could make use of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to just create a group in "cpu" and start it.
2012-09-24 11:35:51 +02:00
}
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
static int enforce_syscall_archs(Set *archs) {
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_SECCOMP
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
int r;
core: do not fail at step SECCOMP if there is no kernel support (#4004) Fixes #3882
2016-08-22 21:40:58 +02:00
if (!is_seccomp_available())
return 0;
seccomp: rework seccomp code, to improve compat with some archs This substantially reworks the seccomp code, to ensure better compatibility with some architectures, including i386. So far we relied on libseccomp's internal handling of the multiple syscall ABIs supported on Linux. This is problematic however, as it does not define clear semantics if an ABI is not able to support specific seccomp rules we install. This rework hence changes a couple of things: - We no longer use seccomp_rule_add(), but only seccomp_rule_add_exact(), and fail the installation of a filter if the architecture doesn't support it. - We no longer rely on adding multiple syscall architectures to a single filter, but instead install a separate filter for each syscall architecture supported. This way, we can install a strict filter for x86-64, while permitting a less strict filter for i386. - All high-level filter additions are now moved from execute.c to seccomp-util.c, so that we can test them independently of the service execution logic. - Tests have been added for all types of our seccomp filters. - SystemCallFilters= and SystemCallArchitectures= are now implemented in independent filters and installation logic, as they semantically are very much independent of each other. Fixes: #4575
2016-12-27 15:28:25 +01:00
r = seccomp_restrict_archs(arg_syscall_archs);
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
if (r < 0)
seccomp: rework seccomp code, to improve compat with some archs This substantially reworks the seccomp code, to ensure better compatibility with some architectures, including i386. So far we relied on libseccomp's internal handling of the multiple syscall ABIs supported on Linux. This is problematic however, as it does not define clear semantics if an ABI is not able to support specific seccomp rules we install. This rework hence changes a couple of things: - We no longer use seccomp_rule_add(), but only seccomp_rule_add_exact(), and fail the installation of a filter if the architecture doesn't support it. - We no longer rely on adding multiple syscall architectures to a single filter, but instead install a separate filter for each syscall architecture supported. This way, we can install a strict filter for x86-64, while permitting a less strict filter for i386. - All high-level filter additions are now moved from execute.c to seccomp-util.c, so that we can test them independently of the service execution logic. - Tests have been added for all types of our seccomp filters. - SystemCallFilters= and SystemCallArchitectures= are now implemented in independent filters and installation logic, as they semantically are very much independent of each other. Fixes: #4575
2016-12-27 15:28:25 +01:00
return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
#endif
seccomp: rework seccomp code, to improve compat with some archs This substantially reworks the seccomp code, to ensure better compatibility with some architectures, including i386. So far we relied on libseccomp's internal handling of the multiple syscall ABIs supported on Linux. This is problematic however, as it does not define clear semantics if an ABI is not able to support specific seccomp rules we install. This rework hence changes a couple of things: - We no longer use seccomp_rule_add(), but only seccomp_rule_add_exact(), and fail the installation of a filter if the architecture doesn't support it. - We no longer rely on adding multiple syscall architectures to a single filter, but instead install a separate filter for each syscall architecture supported. This way, we can install a strict filter for x86-64, while permitting a less strict filter for i386. - All high-level filter additions are now moved from execute.c to seccomp-util.c, so that we can test them independently of the service execution logic. - Tests have been added for all types of our seccomp filters. - SystemCallFilters= and SystemCallArchitectures= are now implemented in independent filters and installation logic, as they semantically are very much independent of each other. Fixes: #4575
2016-12-27 15:28:25 +01:00
return 0;
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
}
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
static int status_welcome(void) {
_cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
int r;
r = parse_env_file("/etc/os-release", NEWLINE,
"PRETTY_NAME", &pretty_name,
"ANSI_COLOR", &ansi_color,
NULL);
tree-wide: drop {} from one-line if blocks Patch via coccinelle.
2015-09-08 23:03:38 +02:00
if (r == -ENOENT)
os-release: define /usr/lib/os-release as fallback for /etc/os-release The file should have been in /usr/lib/ in the first place, since it describes the OS container in /usr (and not the configuration in /etc), hence, let's support os-release files in /usr/lib as fallback if no version in /etc exists, following the usual override logic. A prior commit already enabled tmpfiles to create /etc/os-release as a symlink to /usr/lib/os-release should it be missing, thus providing nice compatibility with applications only checking in /etc. While it's probably a good idea if all apps check both locations via a fallback logic, it is only necessary in the early boot process, as long as the /etc/os-release symlink has not been restored, in case we boot with an empty /etc.
2014-06-13 19:45:52 +02:00
r = parse_env_file("/usr/lib/os-release", NEWLINE,
"PRETTY_NAME", &pretty_name,
"ANSI_COLOR", &ansi_color,
NULL);
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
if (r < 0 && r != -ENOENT)
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_warning_errno(r, "Failed to read os-release file: %m");
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
systemd: obey systemd.log_color config Fixes #2845.
2016-03-16 14:27:37 +01:00
if (log_get_show_color())
return status_printf(NULL, false, false,
"\nWelcome to \x1B[%sm%s\x1B[0m!\n",
isempty(ansi_color) ? "1" : ansi_color,
isempty(pretty_name) ? "Linux" : pretty_name);
else
return status_printf(NULL, false, false,
"\nWelcome to %s!\n",
isempty(pretty_name) ? "Linux" : pretty_name);
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
}
virt: rework container detection logic Instead of accessing /proc/1/environ directly, trying to read the $container variable from it, let's make PID 1 save the contents of that variable to /run/systemd/container. This allows us to detect containers without the need for CAP_SYS_PTRACE, which allows us to drop it from a number of daemons and from the file capabilities of systemd-detect-virt. Also, don't consider chroot a container technology anymore. After all, we don't consider file system namespaces container technology anymore, and hence chroot() should be considered a container even less.
2014-05-28 12:37:11 +02:00
static int write_container_id(void) {
const char *c;
core: bump net.unix.max_dgram_qlen really early during boot Only that way it actually has an effect on all our sockets, including $NOTIFY_SOCKET.
2015-11-02 09:34:05 +01:00
int r;
virt: rework container detection logic Instead of accessing /proc/1/environ directly, trying to read the $container variable from it, let's make PID 1 save the contents of that variable to /run/systemd/container. This allows us to detect containers without the need for CAP_SYS_PTRACE, which allows us to drop it from a number of daemons and from the file capabilities of systemd-detect-virt. Also, don't consider chroot a container technology anymore. After all, we don't consider file system namespaces container technology anymore, and hence chroot() should be considered a container even less.
2014-05-28 12:37:11 +02:00
c = getenv("container");
if (isempty(c))
return 0;
core: be more paranoid when mixing umask and fopen() Let's be extra careful with the umask when we use simple fopen(), as this creates files with 0777 by default.
2016-04-07 16:15:26 +02:00
RUN_WITH_UMASK(0022)
r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
core: bump net.unix.max_dgram_qlen really early during boot Only that way it actually has an effect on all our sockets, including $NOTIFY_SOCKET.
2015-11-02 09:34:05 +01:00
if (r < 0)
core: fix typo
2015-11-04 13:18:59 +01:00
return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
core: bump net.unix.max_dgram_qlen really early during boot Only that way it actually has an effect on all our sockets, including $NOTIFY_SOCKET.
2015-11-02 09:34:05 +01:00
return 1;
}
static int bump_unix_max_dgram_qlen(void) {
_cleanup_free_ char *qlen = NULL;
unsigned long v;
int r;
/* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
* default of 16 is simply too low. We set the value really
* really early during boot, so that it is actually applied to
* all our sockets, including the $NOTIFY_SOCKET one. */
r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
if (r < 0)
return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
r = safe_atolu(qlen, &v);
if (r < 0)
return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
return 0;
qlen = mfree(qlen);
if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
return log_oom();
r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
if (r < 0)
return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
"Failed to bump AF_UNIX datagram queue length, ignoring: %m");
return 1;
virt: rework container detection logic Instead of accessing /proc/1/environ directly, trying to read the $container variable from it, let's make PID 1 save the contents of that variable to /run/systemd/container. This allows us to detect containers without the need for CAP_SYS_PTRACE, which allows us to drop it from a number of daemons and from the file capabilities of systemd-detect-virt. Also, don't consider chroot a container technology anymore. After all, we don't consider file system namespaces container technology anymore, and hence chroot() should be considered a container even less.
2014-05-28 12:37:11 +02:00
}
pid1: initialize TERM environment variable correctly When systemd is started by the kernel, the kernel set the TERM environment variable unconditionnally to "linux" no matter the console device used. This might be an issue for dumb devices with no colors support. This patch uses default_term_for_tty() for getting a more accurate value. But it makes sure to keep the user preferences (if any) which might be passed via the kernel command line. For that purpose /proc should be mounted.
2016-06-20 18:54:21 +02:00
static int fixup_environment(void) {
_cleanup_free_ char *term = NULL;
int r;
main: simplify things a bit by moving container check into fixup_environment()
2016-07-19 17:00:36 +02:00
/* We expect the environment to be set correctly
* if run inside a container. */
if (detect_container() > 0)
return 0;
pid1: initialize TERM environment variable correctly When systemd is started by the kernel, the kernel set the TERM environment variable unconditionnally to "linux" no matter the console device used. This might be an issue for dumb devices with no colors support. This patch uses default_term_for_tty() for getting a more accurate value. But it makes sure to keep the user preferences (if any) which might be passed via the kernel command line. For that purpose /proc should be mounted.
2016-06-20 18:54:21 +02:00
/* When started as PID1, the kernel uses /dev/console
* for our stdios and uses TERM=linux whatever the
* backend device used by the console. We try to make
* a better guess here since some consoles might not
* have support for color mode for example.
*
* However if TERM was configured through the kernel
* command line then leave it alone. */
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
r = proc_cmdline_get_key("TERM", 0, &term);
pid1: initialize TERM environment variable correctly When systemd is started by the kernel, the kernel set the TERM environment variable unconditionnally to "linux" no matter the console device used. This might be an issue for dumb devices with no colors support. This patch uses default_term_for_tty() for getting a more accurate value. But it makes sure to keep the user preferences (if any) which might be passed via the kernel command line. For that purpose /proc should be mounted.
2016-06-20 18:54:21 +02:00
if (r < 0)
return r;
if (r == 0) {
core: inherit TERM from PID 1 for all services started on /dev/console This way, invoking nspawn from a shell in the best case inherits the TERM setting all the way down into the login shell spawned in the container. Fixes: #3697
2016-07-27 15:25:55 +02:00
term = strdup(default_term_for_tty("/dev/console"));
pid1: initialize TERM environment variable correctly When systemd is started by the kernel, the kernel set the TERM environment variable unconditionnally to "linux" no matter the console device used. This might be an issue for dumb devices with no colors support. This patch uses default_term_for_tty() for getting a more accurate value. But it makes sure to keep the user preferences (if any) which might be passed via the kernel command line. For that purpose /proc should be mounted.
2016-06-20 18:54:21 +02:00
if (!term)
main: simplify things a bit by moving container check into fixup_environment()
2016-07-19 17:00:36 +02:00
return -ENOMEM;
pid1: initialize TERM environment variable correctly When systemd is started by the kernel, the kernel set the TERM environment variable unconditionnally to "linux" no matter the console device used. This might be an issue for dumb devices with no colors support. This patch uses default_term_for_tty() for getting a more accurate value. But it makes sure to keep the user preferences (if any) which might be passed via the kernel command line. For that purpose /proc should be mounted.
2016-06-20 18:54:21 +02:00
}
if (setenv("TERM", term, 1) < 0)
return -errno;
return 0;
}
initial commit
2009-11-18 00:42:52 +01:00
int main(int argc, char *argv[]) {
Manager *m = NULL;
systemctl: rework exit codes for all utility programs to follow LSB or other standards
2010-08-31 21:05:54 +02:00
int r, retval = EXIT_FAILURE;
main: profile unit file loading
2011-07-28 23:41:57 +02:00
usec_t before_startup, after_startup;
char timespan[FORMAT_TIMESPAN_MAX];
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
FDSet *fds = NULL;
bool reexecute = false;
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
const char *shutdown_verb = NULL;
core: use some nice macros where appropriate
2015-02-03 18:16:35 +01:00
dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
exec: include path name of binary we are about to execute when renaming forked off processes Immediately after forking off a process change the comm name and argv[0] to "(foobar)" where "foobar" is the basename of the path we are about to execute. This should be useful when charting boot progress.
2012-02-01 22:33:15 +01:00
static char systemd[] = "systemd";
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
bool skip_setup = false;
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
unsigned j;
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
bool loaded_policy = false;
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
bool arm_reboot_watchdog = false;
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
bool queue_default_job = false;
core: populate unit file set with preset data if we boot with empty /etc
2014-06-17 03:25:34 +02:00
bool empty_etc = false;
switch-root: move switch_root() call into its own .c file
2012-05-21 19:33:39 +02:00
char *switch_root_dir = NULL, *switch_root_init = NULL;
main: bump RLIMIT_NOFILE for the root user substantially On current kernels BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK even for privileged users that have CAP_IPC_LOCK. Given that mlock() generally ignores RLIMIT_MEMLOCK if CAP_IPC_LOCK is set this appears to be an oversight in the kernel. Either way, until that's fixed, let's just bump RLIMIT_MEMLOCK for the root user considerably, as the default is quite limiting, and doesn't permit us to create more than a few TRIE maps.
2017-09-21 19:43:07 +02:00
struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
const char *error_message = NULL;
main: initialize default boot target from argv[1]
2010-04-04 22:49:26 +02:00
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_SYSV_COMPAT
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() != 1 && strstr(program_invocation_short_name, "init")) {
Spelling Corrections Just some lame spelling corrections with no functionality.
2011-02-21 15:32:17 +01:00
/* This is compatibility support for SysV, where
init: call telinit in case we are run as init and not pid1
2010-06-18 20:01:01 +02:00
* calling init as a user is identical to telinit. */
execv(SYSTEMCTL_BINARY_PATH, argv);
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
init: call telinit in case we are run as init and not pid1
2010-06-18 20:01:01 +02:00
return 1;
}
main: redirection to telinit is unnecessary if sysv compat is disabled
2011-07-28 23:42:47 +02:00
#endif
init: call telinit in case we are run as init and not pid1
2010-06-18 20:01:01 +02:00
systemd: record the timestamps as early as possible The time for systemd initialization and selinux policy loading is accounted to the initrd or the kernel, which is wrong. Instead of: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 975ms (initrd) + 1.410s (userspace) = 8.647s the more correct output is: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 475ms (initrd) + 1.910s (userspace) = 8.647s
2013-04-24 17:15:47 +02:00
dual_timestamp_from_monotonic(&kernel_timestamp, 0);
dual_timestamp_get(&userspace_timestamp);
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
/* 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. */
main: use strv_find() where we can
2013-03-22 04:42:26 +01:00
if (strv_find(argv+1, "--deserialize"))
skip_setup = true;
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
/* If we have switched root, do all the special setup
* things */
main: use strv_find() where we can
2013-03-22 04:42:26 +01:00
if (strv_find(argv+1, "--switched-root"))
skip_setup = false;
core/main.c: add "--switchedroot" parameter If systemd serializes from a switch_root, it adds "--switchedroot" to the systemd in the real root. If "--switchedroot" is found, then we do not skip all the stuff, which is skipped for normal rexecs.
2012-05-16 14:22:42 +02:00
main: rename process on startup to 'systemd' to avoid confusion
2010-11-10 22:35:05 +01:00
/* If we get started via the /sbin/init symlink then we are
called 'init'. After a subsequent reexecution we are then
called 'systemd'. That is confusing, hence let's call us
systemd right-away. */
program_invocation_short_name = systemd;
core: minor coding style/wording fixes
2016-12-08 17:23:08 +01:00
(void) prctl(PR_SET_NAME, systemd);
exec: include path name of binary we are about to execute when renaming forked off processes Immediately after forking off a process change the comm name and argv[0] to "(foobar)" where "foobar" is the basename of the path we are about to execute. This should be useful when charting boot progress.
2012-02-01 22:33:15 +01:00
util: try harder to rename the process
2011-06-30 04:16:10 +02:00
saved_argv = argv;
saved_argc = argc;
main: rename process on startup to 'systemd' to avoid confusion
2010-11-10 22:35:05 +01:00
log: never ever log to syslog from PID 1, log to the journal again We don't support journal-less systems anyway, so let's avoid the confusion.
2014-08-11 20:08:08 +02:00
log_set_upgrade_syslog_to_journal(true);
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1) {
log: never log into foreign fd #2 in PID 1 or its pre-execve() children Fixes: #5401
2017-02-21 17:57:55 +01:00
/* Disable the umask logic */
main: set umask before creating any files This avoids a problem when we inherit a non-zero umask from the initramfs. This would cause /run/systemd to be created with the wrong mode.
2013-09-26 20:39:41 +02:00
umask(0);
log: never log into foreign fd #2 in PID 1 or its pre-execve() children Fixes: #5401
2017-02-21 17:57:55 +01:00
/* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
* important so that we never end up logging to any foreign stderr, for example if we have to log in a
* child process right before execve()'ing the actual binary, at a point in time where socket
* activation stderr/stdout area already set up. */
log_set_always_reopen_console(true);
}
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1 && detect_container() <= 0) {
systemd: record efi timestamps after /sys is mounted This partially reverts commit c3a170f3, which moved efi_get_boot_timestamps too early in main(), before /sys is assured to be mounted Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=64371 [tomegun: in particular /sys/firmware/efi/efivars needs to be mounted, which is not a problem if a systemd-initramfs containing the correct module is being used. But not everyone uses an initramfs...]
2013-05-14 00:00:37 +02:00
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
/* Running outside of a container as PID 1 */
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
arg_system = true;
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
log_set_target(LOG_TARGET_KMSG);
log_open();
remove left-over initrd time stamp handling
2013-07-23 12:25:32 +02:00
if (in_initrd())
systemd: record the timestamps as early as possible The time for systemd initialization and selinux policy loading is accounted to the initrd or the kernel, which is wrong. Instead of: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 975ms (initrd) + 1.410s (userspace) = 8.647s the more correct output is: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 475ms (initrd) + 1.910s (userspace) = 8.647s
2013-04-24 17:15:47 +02:00
initrd_timestamp = userspace_timestamp;
core/main.c: handle the initrd timestamp differently, if in the initrd If systemd is in the initrd/initramfs, set the initrd timestamp and do not try to read it from the RD_TIMESTAMP environment variable.
2012-05-16 14:22:44 +02:00
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
if (!skip_setup) {
manager: print fatal error if early mount failed The mount_setup_early() can fail and if it will occur, there is no sense to make selinux setup and etc.
2016-02-02 20:36:33 +01:00
r = mount_setup_early();
if (r < 0) {
core: minor coding style/wording fixes
2016-12-08 17:23:08 +01:00
error_message = "Failed to mount early API filesystems";
manager: print fatal error if early mount failed The mount_setup_early() can fail and if it will occur, there is no sense to make selinux setup and etc.
2016-02-02 20:36:33 +01:00
goto finish;
}
core: minor coding style/wording fixes
2016-12-08 17:23:08 +01:00
analyze: plot the time spent setting up security modules
2013-11-10 23:21:15 +01:00
dual_timestamp_get(&security_start_timestamp);
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (mac_selinux_setup(&loaded_policy) < 0) {
error_message = "Failed to load SELinux policy";
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
} else if (mac_smack_setup(&loaded_policy) < 0) {
error_message = "Failed to load SMACK policy";
core: mount and initialize Smack SMACK is the Simple Mandatory Access Control Kernel, a minimal approach to Access Control implemented as a kernel LSM. The kernel exposes the smackfs filesystem API through which access rules can be loaded. At boot time, we want to load the access rules as early as possible to ensure all early boot steps are checked by Smack. This patch mounts smackfs at the new location at /sys/fs/smackfs for kernels 3.8 and above. The /smack mountpoint is not supported. After mounting smackfs, rules are loaded from the usual location. For more information about Smack see: http://www.kernel.org/doc/Documentation/security/Smack.txt
2013-03-07 20:06:58 +01:00
goto finish;
main: load Smack policy before IMA policy (#3859) IMA wiki says: "If the IMA policy contains LSM labels, then the LSM policy must be loaded prior to the IMA policy." Right now, in case of Smack, the IMA policy is loaded before the Smack policy. Move the order around to allow Smack labels to be used in IMA policy.
2016-08-02 14:58:30 +02:00
} else if (ima_setup() < 0) {
error_message = "Failed to load IMA policy";
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
analyze: plot the time spent setting up security modules
2013-11-10 23:21:15 +01:00
dual_timestamp_get(&security_finish_timestamp);
main: added support for loading IMA custom policies This is an S/MIME signed message The new function ima_setup() loads an IMA custom policy from a file in the default location '/etc/ima/ima-policy', if present, and writes it to the path 'ima/policy' in the security filesystem. This function is executed at early stage in order to avoid that some file operations are not measured by IMA and it is placed after the initialization of SELinux because IMA needs the latter (or other security modules) to understand LSM-specific rules. This feature is enabled by default and can be disabled by providing the option '--disable-ima' to the configure script. Signed-off-by: Roberto Sassu <roberto.sassu@polito.it> Acked-by: Gianluca Ramunno <ramunno@polito.it>
2012-03-15 19:06:11 +01:00
}
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
selinux: always try to load the full selinux db https://github.com/systemd/systemd/pull/2508#issuecomment-190901170 Maybe fixes https://bugzilla.redhat.com/show_bug.cgi?id=1308771.
2016-03-02 02:35:55 +01:00
if (mac_selinux_init() < 0) {
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to initialize SELinux policy";
main: automatically spawn a getty on the kernel configured serial console
2010-08-06 00:42:24 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
rtc in localtime: use settimeofday(NULL, tz) instead of hwclock(8) We check for LOCAL in /etc/adjtime and if needed, ask the kernel to apply the timezone delta to the system clock. The very first call of settimeofday() without a time, but a timezone warps the system clock, so that it properly runs in UTC.
2011-05-24 20:23:07 +02:00
hwclock: always set the kernel's timezone Properly tell the kernel at bootup, and any later time zone changes, the actual system time zone. Things like the kernel's FAT filesystem driver needs the actual time zone to calculate the proper local time to use for the on-disk time stamps. https://bugzilla.redhat.com/show_bug.cgi?id=802198
2012-09-17 16:41:13 +02:00
if (!skip_setup) {
clock-util: make clock_is_localtime() testable and add initial tests Add path argument to clock_is_localtime() and default to "/etc/adjtime" if it's NULL. This makes the function testable. Add test-clock: initial test cases for some scenarios, using a temporary file. This also checks the behaviour with a NULL (i. e. the system's /etc/adjtime) file.
2016-02-26 11:25:22 +01:00
if (clock_is_localtime(NULL) > 0) {
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
int min;
rtc in localtime: use settimeofday(NULL, tz) instead of hwclock(8) We check for LOCAL in /etc/adjtime and if needed, ask the kernel to apply the timezone delta to the system clock. The very first call of settimeofday() without a time, but a timezone warps the system clock, so that it properly runs in UTC.
2011-05-24 20:23:07 +02:00
core: only set the kernel's timezone when the RTC runs in local time We can not reliably manage any notion of local time. Every daylight saving time change or time zone change by traveling will make the time jump, and the local time might jump backwards which creates unsolvable problems with file timestamps. We will no longer tell the kernel our local time zone and leave everything set to UTC. This will effectively turn FAT timestamps into UTC timestamps. If and only if the machine is configured to read the RTC in local time mode, the kernel's time zone will be configured, but systemd-timesysnc will disable the kernel's system time to RTC syncing. In this mode, the RTC will not be managed, and external tools like Windows bootups are expected to manage the RTC's time. https://bugs.freedesktop.org/show_bug.cgi?id=81538
2014-07-27 00:11:08 +02:00
/*
* The very first call of settimeofday() also does a time warp in the kernel.
*
* In the rtc-in-local time mode, we set the kernel's timezone, and rely on
* external tools to take care of maintaining the RTC and do all adjustments.
* This matches the behavior of Windows, which leaves the RTC alone if the
* registry tells that the RTC runs in UTC.
*/
shared: rename hwclock.[ch] to clock-util.[ch]
2014-05-22 14:21:38 +02:00
r = clock_set_timezone(&min);
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
if (r < 0)
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
selinux: use setcon() instead of reexec to apply selinux policy
2011-07-28 23:52:23 +02:00
else
log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
hwclock: do not seal the kernel's time-warp call from inside the initrd
2012-10-27 16:23:32 +02:00
} else if (!in_initrd()) {
/*
core: only set the kernel's timezone when the RTC runs in local time We can not reliably manage any notion of local time. Every daylight saving time change or time zone change by traveling will make the time jump, and the local time might jump backwards which creates unsolvable problems with file timestamps. We will no longer tell the kernel our local time zone and leave everything set to UTC. This will effectively turn FAT timestamps into UTC timestamps. If and only if the machine is configured to read the RTC in local time mode, the kernel's time zone will be configured, but systemd-timesysnc will disable the kernel's system time to RTC syncing. In this mode, the RTC will not be managed, and external tools like Windows bootups are expected to manage the RTC's time. https://bugs.freedesktop.org/show_bug.cgi?id=81538
2014-07-27 00:11:08 +02:00
* Do a dummy very first call to seal the kernel's time warp magic.
hwclock: do not seal the kernel's time-warp call from inside the initrd
2012-10-27 16:23:32 +02:00
*
treewide: fix typos and remove accidental repetition of words
2016-07-10 14:48:23 +02:00
* Do not call this from inside the initrd. The initrd might not
hwclock: do not seal the kernel's time-warp call from inside the initrd
2012-10-27 16:23:32 +02:00
* 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.
core: only set the kernel's timezone when the RTC runs in local time We can not reliably manage any notion of local time. Every daylight saving time change or time zone change by traveling will make the time jump, and the local time might jump backwards which creates unsolvable problems with file timestamps. We will no longer tell the kernel our local time zone and leave everything set to UTC. This will effectively turn FAT timestamps into UTC timestamps. If and only if the machine is configured to read the RTC in local time mode, the kernel's time zone will be configured, but systemd-timesysnc will disable the kernel's system time to RTC syncing. In this mode, the RTC will not be managed, and external tools like Windows bootups are expected to manage the RTC's time. https://bugs.freedesktop.org/show_bug.cgi?id=81538
2014-07-27 00:11:08 +02:00
*
* Do no set the kernel's timezone. The concept of local time cannot
* be supported reliably, the time will jump or be incorrect at every daylight
* saving time change. All kernel local time concepts will be treated
* as UTC that way.
hwclock: do not seal the kernel's time-warp call from inside the initrd
2012-10-27 16:23:32 +02:00
*/
resolved: apply epoch to system time from PID 1 For use in timesyncd we already defined a compile-time "epoch" value, which is based on the mtime of the NEWS file, and specifies a point in time we know lies in the past at runtime. timesyncd uses this to filter out nonsensical timestamp file data, and bump the system clock to a time that is after the build time of systemd. This patch adds similar bumping code to earliest PID 1 initialization, so that the system never continues operation with a clock that is in the 1970ies or even 1930s.
2016-02-03 18:28:40 +01:00
(void) clock_reset_timewarp();
hwclock: always set the kernel's timezone Properly tell the kernel at bootup, and any later time zone changes, the actual system time zone. Things like the kernel's FAT filesystem driver needs the actual time zone to calculate the proper local time to use for the on-disk time stamps. https://bugzilla.redhat.com/show_bug.cgi?id=802198
2012-09-17 16:41:13 +02:00
}
resolved: apply epoch to system time from PID 1 For use in timesyncd we already defined a compile-time "epoch" value, which is based on the mtime of the NEWS file, and specifies a point in time we know lies in the past at runtime. timesyncd uses this to filter out nonsensical timestamp file data, and bump the system clock to a time that is after the build time of systemd. This patch adds similar bumping code to earliest PID 1 initialization, so that the system never continues operation with a clock that is in the 1970ies or even 1930s.
2016-02-03 18:28:40 +01:00
r = clock_apply_epoch();
if (r < 0)
log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
else if (r > 0)
log_info("System time before build time, advancing clock.");
hwclock: always set the kernel's timezone Properly tell the kernel at bootup, and any later time zone changes, the actual system time zone. Things like the kernel's FAT filesystem driver needs the actual time zone to calculate the proper local time to use for the on-disk time stamps. https://bugzilla.redhat.com/show_bug.cgi?id=802198
2012-09-17 16:41:13 +02:00
}
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
/* 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);
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
} else if (getpid_cached() == 1) {
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
/* Running inside a container, as PID 1 */
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
arg_system = true;
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
log_set_target(LOG_TARGET_CONSOLE);
log: don't reopen /dev/console each time we call log_open() Instead, force reopen it only if we really really have to.
2013-12-18 05:07:34 +01:00
log_close_console(); /* force reopen of /dev/console */
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
log_open();
core: minor coding style/wording fixes
2016-12-08 17:23:08 +01:00
/* For later on, see above... */
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
log_set_target(LOG_TARGET_JOURNAL);
systemd: record the timestamps as early as possible The time for systemd initialization and selinux policy loading is accounted to the initrd or the kernel, which is wrong. Instead of: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 975ms (initrd) + 1.410s (userspace) = 8.647s the more correct output is: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 475ms (initrd) + 1.910s (userspace) = 8.647s
2013-04-24 17:15:47 +02:00
/* clear the kernel timestamp,
* because we are in a container */
core: use DUAL_TIMESTAMP_NULL to reset kernel_timestamp instead of direct reset of kernel_timestamp fields.
2016-02-29 13:56:57 +01:00
kernel_timestamp = DUAL_TIMESTAMP_NULL;
systemd: record the timestamps as early as possible The time for systemd initialization and selinux policy loading is accounted to the initrd or the kernel, which is wrong. Instead of: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 975ms (initrd) + 1.410s (userspace) = 8.647s the more correct output is: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 475ms (initrd) + 1.910s (userspace) = 8.647s
2013-04-24 17:15:47 +02:00
} else {
main: when transitioning from initrd to the main system log to kmsg When the new PID is invoked the journal socket from the initrd might still be around. Due to the default log target being journal we'd log to that initially when the new main systemd initializes even if the kernel command line included a directive to redirect systemd's logging elsewhere. With this fix we initially always log to kmsg now, if we are PID1, and only after parsing the kernel cmdline try to open the journal if that's desired. (The effective benefit of this is that SELinux performance data is now logged again to kmsg like it used to be.)
2012-09-17 17:42:13 +02:00
/* Running as user instance */
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
arg_system = false;
log: log to syslog unless connected to a tty in user mode
2011-07-01 22:35:34 +02:00
log_set_target(LOG_TARGET_AUTO);
selinux: log how much time it takes to load the SELinux policy and database
2011-07-25 21:22:57 +02:00
log_open();
systemd: record the timestamps as early as possible The time for systemd initialization and selinux policy loading is accounted to the initrd or the kernel, which is wrong. Instead of: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 975ms (initrd) + 1.410s (userspace) = 8.647s the more correct output is: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 475ms (initrd) + 1.910s (userspace) = 8.647s
2013-04-24 17:15:47 +02:00
/* clear the kernel timestamp,
* because we are not PID 1 */
core: use DUAL_TIMESTAMP_NULL where we can
2015-08-28 17:11:37 +02:00
kernel_timestamp = DUAL_TIMESTAMP_NULL;
log: make color/location logging optional
2010-06-17 22:52:55 +02:00
}
manager: make running_as configurable
2010-04-06 23:55:42 +02:00
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1) {
core: set RLIMIT_CORE to unlimited by default The kernel sets RLIMIT_CORE to 0 by default. Let's bump this to unlimited by default (for systemd itself and all processes we fork off), so that the coredump hooks have an effect if they honour it. Bumping RLIMIT_CORE of course would have the effect that "core" files will end up on the system at various places, if no coredump hook is used. To avoid this, make sure PID1 sets the core pattern to the empty string by default, so that this logic is disabled. This change in defaults should be useful for all systems where coredump hooks are used, as it allows useful usage of RLIMIT_CORE from these hooks again. OTOH systems that expect that coredumps are placed under the name "core" in the current directory will break with this change. Given how questionnable this behaviour is, and given that no common distro makes use of this by default it shouldn't be too much of a loss. Also, the old behaviour may be restored by explicitly configuring a "core_pattern" of "core", and setting the default system RLIMIT_CORE to 0 again via system.conf.
2016-02-08 22:30:58 +01:00
/* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
* will process core dumps for system services by default. */
systemd: warn when setrlimit fails This should make it easier to figure things out.
2016-08-17 18:06:07 +02:00
if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
core: set RLIMIT_CORE to unlimited by default The kernel sets RLIMIT_CORE to 0 by default. Let's bump this to unlimited by default (for systemd itself and all processes we fork off), so that the coredump hooks have an effect if they honour it. Bumping RLIMIT_CORE of course would have the effect that "core" files will end up on the system at various places, if no coredump hook is used. To avoid this, make sure PID1 sets the core pattern to the empty string by default, so that this logic is disabled. This change in defaults should be useful for all systems where coredump hooks are used, as it allows useful usage of RLIMIT_CORE from these hooks again. OTOH systems that expect that coredumps are placed under the name "core" in the current directory will break with this change. Given how questionnable this behaviour is, and given that no common distro makes use of this by default it shouldn't be too much of a loss. Also, the old behaviour may be restored by explicitly configuring a "core_pattern" of "core", and setting the default system RLIMIT_CORE to 0 again via system.conf.
2016-02-08 22:30:58 +01:00
/* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
* until the systemd-coredump tool is enabled via sysctl. */
if (!skip_setup)
(void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
}
pid1: restore console color support for containers (#3595) Commit 3a18b60489504056f9b0b1a139439cbfa60a87e1 introduced a regression that disabled the color mode for container. This patch fixes this.
2016-06-24 16:08:43 +02:00
if (arg_system) {
main: simplify things a bit by moving container check into fixup_environment()
2016-07-19 17:00:36 +02:00
if (fixup_environment() < 0) {
error_message = "Failed to fix up PID1 environment";
goto finish;
}
pid1: initialize TERM environment variable correctly When systemd is started by the kernel, the kernel set the TERM environment variable unconditionnally to "linux" no matter the console device used. This might be an issue for dumb devices with no colors support. This patch uses default_term_for_tty() for getting a more accurate value. But it makes sure to keep the user preferences (if any) which might be passed via the kernel command line. For that purpose /proc should be mounted.
2016-06-20 18:54:21 +02:00
pid1: initialize status color mode after setting up TERM Also we had to connect PID's stdio to null later since colors_enabled() assume that stdout is connected to the console.
2016-06-20 21:45:28 +02:00
/* Try to figure out if we can use colors with the console. No
* need to do that for user instances since they never log
* into the console. */
log_show_color(colors_enabled());
tree-wide: pass return value of make_null_stdio() to warning instead of errno (#4328) as @poettering suggested in the #4320
2016-10-10 19:51:33 +02:00
r = make_null_stdio();
if (r < 0)
log_warning_errno(r, "Failed to redirect standard streams to /dev/null: %m");
pid1: initialize status color mode after setting up TERM Also we had to connect PID's stdio to null later since colors_enabled() assume that stdout is connected to the console.
2016-06-20 21:45:28 +02:00
}
main: don't try to mout cpuset with cpu+cpuacct anymore Turns out cpuset needs explicit initialization before we could make use of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to just create a group in "cpu" and start it.
2012-09-24 11:35:51 +02:00
r = initialize_join_controllers();
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (r < 0) {
treewide: Correct typos and spell plural of bus consistent
2015-05-11 13:49:29 +02:00
error_message = "Failed to initialize cgroup controllers";
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
/* Mount /proc, /sys and friends, so that /proc/cmdline and
* /proc/$PID/fd is available. */
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1) {
kmod: move #ifdef checks for kmod-setup out of main.c into kmod-setup.c
2014-11-14 17:58:32 +01:00
Drop kdbus bits Some kdbus_flag and memfd related parts are left behind, because they are entangled with the "legacy" dbus support. test-bus-benchmark is switched to "manual". It was already broken before (in the non-kdbus mode) but apparently nobody noticed. Hopefully it can be fixed later.
2017-07-23 17:45:57 +02:00
/* Load the kernel modules early. */
mount-setup: remove mount_setup_late() Turns out we can just do kmod_setup() earlier, before we do mount_setup(), so there's no need for mount_setup_late() anymore. Instead, put kdbusfs in mount_table[].
2014-11-14 15:18:56 +01:00
if (!skip_setup)
kmod_setup();
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
r = mount_setup(loaded_policy);
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (r < 0) {
error_message = "Failed to mount API filesystems";
main: don't try to mount api dirs if we are not root
2010-05-24 18:59:13 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
}
main: move basic setup into main.c
2010-04-06 21:59:25 +02:00
/* Reset all signal handlers. */
tree-wide: whenever we fork off a foreign child process reset signal mask/handlers Also, when the child is potentially long-running make sure to set a death signal. Also, ignore the result of the reset operations explicitly by casting them to (void).
2015-05-31 23:55:55 +02:00
(void) reset_all_signal_handlers();
(void) ignore_signals(SIGNALS_IGNORE, -1);
main: ignore SIGKILL and SIGPIPE so that nothing intereferes with us
2010-04-13 02:01:28 +02:00
logind: update empty and "infinity" handling for [User]TasksMax (#3835) The parsing functions for [User]TasksMax were inconsistent. Empty string and "infinity" were interpreted as no limit for TasksMax but not accepted for UserTasksMax. Update them so that they're consistent with other knobs. * Empty string indicates the default value. * "infinity" indicates no limit. While at it, replace opencoded (uint64_t) -1 with CGROUP_LIMIT_MAX in TasksMax handling. v2: Update empty string to indicate the default value as suggested by Zbigniew Jędrzejewski-Szmek. v3: Fixed empty UserTasksMax handling.
2016-08-19 04:57:53 +02:00
arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
core: change TasksMax= default for system services to 15% As it turns out 512 is max number of tasks per service is hit by too many applications, hence let's bump it a bit, and make it relative to the system's maximum number of PIDs. With this change the new default is 15%. At the kernel's default pids_max value of 32768 this translates to 4915. At machined's default TasksMax= setting of 16384 this translates to 2457. Why 15%? Because it sounds like a round number and is close enough to 4096 which I was going for, i.e. an eight-fold increase over the old 512 Summary: | on the host | in a container old default | 512 | 512 new default | 4915 | 2457
2016-07-19 17:29:00 +02:00
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (parse_config_file() < 0) {
error_message = "Failed to parse config file";
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
main: implement manager configuration file
2010-07-07 01:10:27 +02:00
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (arg_system) {
util-lib: various improvements to kernel command line parsing This improves kernel command line parsing in a number of ways: a) An kernel option "foo_bar=xyz" is now considered equivalent to "foo-bar-xyz", i.e. when comparing kernel command line option names "-" and "_" are now considered equivalent (this only applies to the option names though, not the option values!). Most of our kernel options used "-" as word separator in kernel command line options so far, but some used "_". With this change, which was a source of confusion for users (well, at least of one user: myself, I just couldn't remember that it's systemd.debug-shell, not systemd.debug_shell). Considering both as equivalent is inspired how modern kernel module loading normalizes all kernel module names to use underscores now too. b) All options previously using a dash for separating words in kernel command line options now use an underscore instead, in all documentation and in code. Since a) has been implemented this should not create any compatibility problems, but normalizes our documentation and our code. c) All kernel command line options which take booleans (or are boolean-like) have been reworked so that "foobar" (without argument) is now equivalent to "foobar=1" (but not "foobar=0"), thus normalizing the handling of our boolean arguments. Specifically this means systemd.debug-shell and systemd_debug_shell=1 are now entirely equivalent. d) All kernel command line options which take an argument, and where no argument is specified will now result in a log message. e.g. passing just "systemd.unit" will no result in a complain that it needs an argument. This is implemented in the proc_cmdline_missing_value() function. e) There's now a call proc_cmdline_get_bool() similar to proc_cmdline_get_key() that parses booleans (following the logic explained in c). f) The proc_cmdline_parse() call's boolean argument has been replaced by a new flags argument that takes a common set of bits with proc_cmdline_get_key(). g) All kernel command line APIs now begin with the same "proc_cmdline_" prefix. h) There are now tests for much of this. Yay!
2016-12-12 18:29:15 +01:00
r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
util: simplify proc_cmdline() to reuse get_process_cmdline() Also, make all parsing of the kernel cmdline non-fatal.
2014-11-06 21:53:34 +01:00
if (r < 0)
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
util: simplify proc_cmdline() to reuse get_process_cmdline() Also, make all parsing of the kernel cmdline non-fatal.
2014-11-06 21:53:34 +01:00
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
main,log: parse the log related kernel command line parameters at one place only, and for all tools Previously, we ended up parsing some of them three times: in main.c when processing the kernel cmdline, in main.c when processing the process cmdline (only for containers), and in log.c again. Let's streamline this, and only parse them in log.c In PID 1 also make sure we parse "quiet" first, and then override this with the more specific checks in log.c
2014-08-15 18:07:36 +02:00
/* Note that this also parses bits from the kernel command
* line, including "debug". */
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
log_parse_environment();
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (parse_argv(argc, argv) < 0) {
error_message = "Failed to parse commandline arguments";
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
main: initialize default unit little later (#4321) systemd fills arg_default_unit during startup with default.target value. But arg_default_unit may be overwritten in parse_argv() or parse_proc_cmdline_item(). Let's check value of arg_default_unit after calls of parse_argv() and parse_proc_cmdline_item() and fill it with default.target if it wasn't filled before. In this way we will not spend unnecessary time to for filling arg_default_unit with default.target.
2016-10-10 04:57:03 +02:00
/* Initialize default unit */
if (!arg_default_unit) {
main: use strdup instead of free_and_strdup to initialize default unit (#4335) Previously we've used free_and_strdup() to fill arg_default_unit with unit name, If we didn't pass default unit name through a kernel command line or command line arguments. But we can use just strdup() instead of free_and_strdup() for this, because we will start fill arg_default_unit only if it wasn't set before.
2016-10-10 22:11:36 +02:00
arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
if (!arg_default_unit) {
r = log_oom();
main: initialize default unit little later (#4321) systemd fills arg_default_unit during startup with default.target value. But arg_default_unit may be overwritten in parse_argv() or parse_proc_cmdline_item(). Let's check value of arg_default_unit after calls of parse_argv() and parse_proc_cmdline_item() and fill it with default.target if it wasn't filled before. In this way we will not spend unnecessary time to for filling arg_default_unit with default.target.
2016-10-10 04:57:03 +02:00
error_message = "Failed to set default unit";
goto finish;
}
}
core: Refuse to run a user instance when the system hasn't been booted with systemd. Running as a user instance won't work at all if systemd isn't running as system manager, so refuse to start in that case.
2012-10-06 01:11:52 +02:00
if (arg_action == ACTION_TEST &&
geteuid() == 0) {
main: refuse --test as root
2011-02-19 14:20:00 +01:00
log_error("Don't run test mode as root.");
goto finish;
}
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (!arg_system &&
core: Refuse to run a user instance when the system hasn't been booted with systemd. Running as a user instance won't work at all if systemd isn't running as system manager, so refuse to start in that case.
2012-10-06 01:11:52 +02:00
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;
}
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (arg_system &&
main: refuse system to be started in a chroot
2011-03-09 23:59:27 +01:00
arg_action == ACTION_RUN &&
running_in_chroot() > 0) {
log_error("Cannot be run in a chroot() environment.");
goto finish;
}
core/main: get rid from excess check of ACTION_TEST (#4350) If `--test` command line option was passed, the systemd set skip_setup to true during bootup. But after this we check again that arg_action is test or help and opens pager depends on result. We should skip setup in a case when `--test` is passed, but it is also safe to set skip_setup in a case of `--help`. So let's remove first check and move skip_setup = true to the second check.
2016-10-11 23:30:04 +02:00
if (arg_action == ACTION_TEST || arg_action == ACTION_HELP) {
tree-wide: merge pager_open_if_enabled() to the pager_open() Many subsystems define own pager_open_if_enabled() function which checks '--no-pager' command line argument and open pager depends on its value. All implementations of pager_open_if_enabled() are the same. Let's merger this function with pager_open() from the shared/pager.c and remove pager_open_if_enabled() from all subsytems to prevent code duplication.
2016-02-19 19:25:13 +01:00
pager_open(arg_no_pager, false);
core/main: get rid from excess check of ACTION_TEST (#4350) If `--test` command line option was passed, the systemd set skip_setup to true during bootup. But after this we check again that arg_action is test or help and opens pager depends on result. We should skip setup in a case when `--test` is passed, but it is also safe to set skip_setup in a case of `--help`. So let's remove first check and move skip_setup = true to the second check.
2016-10-11 23:30:04 +02:00
skip_setup = true;
}
systemd: use pager for --test and --help
2014-01-07 06:00:05 +01:00
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
if (arg_action == ACTION_HELP) {
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
retval = help();
goto finish;
systemd: add --version option systemd --version mirrors systemctl --version: $ ./systemd --version systemd 186 other +PAM +LIBWRAP +AUDIT +SELINUX +IMA +SYSVINIT -LIBCRYPTSETUP This information can be retrieved by other means (systemctl, etc.), but it's easier for a newbie if 'systemd --version' says something useful. And 'systemd --help' is already there, so let's complement that with '--version'.
2012-07-17 07:31:47 +02:00
} else if (arg_action == ACTION_VERSION) {
retval = version();
goto finish;
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
} else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
main: use pager for --dump-configuration-items (#3894)
2016-08-04 22:52:24 +02:00
pager_open(arg_no_pager, false);
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
unit_dump_config_items(stdout);
systemctl: rework exit codes for all utility programs to follow LSB or other standards
2010-08-31 21:05:54 +02:00
retval = EXIT_SUCCESS;
mount: implement mounting properly This also includes code that writes utmp/wtmp records when applicable, making use the mount infrastructure to detct when those files are accessible. Finally, this also introduces a --dump-configuration-items switch.
2010-04-10 17:53:17 +02:00
goto finish;
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
}
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (!arg_system &&
core: fix require $XDG_RUNTIME_DIR $ sudo dracut -f Trying to run as user instance, but $XDG_RUNTIME_DIR is not set. /usr/lib/dracut/modules.d/98systemd/module-setup.sh: line 10: ((: >= 198 : syntax error: operand expected (error token is ">= 198 ")
2013-11-07 14:14:22 +01:00
!getenv("XDG_RUNTIME_DIR")) {
log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
goto finish;
}
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
selinux: log how much time it takes to load the SELinux policy and database
2011-07-25 21:22:57 +02:00
/* Close logging fds, in order not to confuse fdset below */
log_close();
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
/* Remember open file descriptors for later deserialization */
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_action == ACTION_RUN) {
r = fdset_new_fill(&fds);
if (r < 0) {
log_emergency_errno(r, "Failed to allocate fd set: %m");
error_message = "Failed to allocate fd set";
goto finish;
} else
fdset_cloexec(fds, true);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_serialization)
assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_system)
/* Become a session leader if we aren't one yet. */
setsid();
}
main: move basic setup into main.c
2010-04-06 21:59:25 +02:00
core: rearrange if blocks a bit
2013-07-11 00:48:52 +02:00
/* Move out of the way, so that we won't block unmounts */
fix extraneous space in equality check
2015-05-27 21:02:24 +02:00
assert_se(chdir("/") == 0);
core: rearrange if blocks a bit
2013-07-11 00:48:52 +02:00
main: never reset console unless pid=1, to make sure that we don't kill X when somebody passes --test
2010-04-23 04:06:23 +02:00
/* Reset the console, but only if this is really init and we
* are freshly booted */
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (arg_system && arg_action == ACTION_RUN) {
main: minor code modernization for initializing the console
2014-08-15 18:01:52 +02:00
/* If we are init, we connect stdin/stdout/stderr to
* /dev/null and make sure we don't have a controlling
* tty. */
release_terminal();
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1 && !skip_setup)
main: minor code modernization for initializing the console
2014-08-15 18:01:52 +02:00
console_setup();
}
main: move basic setup into main.c
2010-04-06 21:59:25 +02:00
main: remove testing assert
2010-04-10 21:42:55 +02:00
/* Open the logging devices, if possible and necessary */
log: rework logging logic so that we don't keep /dev/console open
2010-05-15 17:25:08 +02:00
log_open();
main: move basic setup into main.c
2010-04-06 21:59:25 +02:00
core: fixate show_status earlier, so that we actually print the welcome message Previously, we'd fixed show_state only after printing the welcome message which had the effect that the welcome message was almost always suppressed.
2014-02-17 16:17:08 +01:00
if (arg_show_status == _SHOW_STATUS_UNSET)
arg_show_status = SHOW_STATUS_YES;
main: install crash handler in all signals that usually dump core
2010-04-13 02:00:30 +02:00
/* Make sure we leave a core dump without panicing the
* kernel. */
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1) {
main: introduce configurable crash shell
2010-04-10 22:35:37 +02:00
install_crash_handler();
main: install crash handler for creating core dumps
2010-04-10 21:40:40 +02:00
cgroup: optionally mount a specific cgroup controllers together, and add cpu+cpuacct to the default
2011-08-23 00:37:35 +02:00
r = mount_cgroup_controllers(arg_join_controllers);
if (r < 0)
goto finish;
}
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (arg_system) {
basic: rework virtualization detection API Introduce a proper enum, and don't pass around string ids anymore. This simplifies things quite a bit, and makes virtualization detection more similar to architecture detection.
2015-09-07 13:42:47 +02:00
int v;
main: print detected virtualiaztion at boot
2012-06-28 13:46:45 +02:00
test-engine: fix access to unit load path Also add a bit of debugging output to help diagnose problems, add missing units, and simplify cppflags. Move test-engine to normal tests from manual tests, it should now work without destroying the system.
2014-01-07 14:41:24 +01:00
log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
arg_action == ACTION_TEST ? "test " : "" );
main: print detected virtualiaztion at boot
2012-06-28 13:46:45 +02:00
basic: rework virtualization detection API Introduce a proper enum, and don't pass around string ids anymore. This simplifies things quite a bit, and makes virtualization detection more similar to architecture detection.
2015-09-07 13:42:47 +02:00
v = detect_virtualization();
if (v > 0)
log_info("Detected virtualization %s.", virtualization_to_string(v));
main: print detected virtualiaztion at boot
2012-06-28 13:46:45 +02:00
virt: rework container detection logic Instead of accessing /proc/1/environ directly, trying to read the $container variable from it, let's make PID 1 save the contents of that variable to /run/systemd/container. This allows us to detect containers without the need for CAP_SYS_PTRACE, which allows us to drop it from a number of daemons and from the file capabilities of systemd-detect-virt. Also, don't consider chroot a container technology anymore. After all, we don't consider file system namespaces container technology anymore, and hence chroot() should be considered a container even less.
2014-05-28 12:37:11 +02:00
write_container_id();
core: do not use quotes around virt and arch Quotes are useful when the string can contain spaces or be otherwise confusing. Not possible with those two.
2015-03-15 02:49:10 +01:00
log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
hostnamectl: should the sanitized arch, not the native uname() one
2014-02-21 02:28:54 +01:00
core: mention if we are run from an initial RAM disk
2012-07-02 11:38:32 +02:00
if (in_initrd())
log_info("Running in initial RAM disk.");
main: explain our /etc empty check a bit in a comment
2014-07-07 11:47:46 +02:00
/* Let's check whether /etc is already populated. We
* don't actually really check for that, but use
* /etc/machine-id as flag file. This allows container
* managers and installers to provision a couple of
* files already. If the container manager wants to
* provision the machine ID itself it should pass
core: correct spacing near eol in code comments
2014-12-10 20:00:08 +01:00
* $container_uuid to PID 1. */
main: explain our /etc empty check a bit in a comment
2014-07-07 11:47:46 +02:00
main: change check whether /etc is unpopulated to look for /etc/machine-id Previously, we checked whether /etc was completely empty. This makes it difficult though for container managers such as nspawn to install a small number of files (such as /etc/timezone), and have the system otherwise populate its own tree. Hence, change this by looking for /etc/machine-id, which should be a good sign whether /etc is populated or not.
2014-07-04 03:13:05 +02:00
empty_etc = access("/etc/machine-id", F_OK) < 0;
core: populate unit file set with preset data if we boot with empty /etc
2014-06-17 03:25:34 +02:00
if (empty_etc)
log_info("Running with unpopulated /etc.");
Improve messages about user mode a bit
2014-01-08 23:20:45 +01:00
} else {
main: uid_to_name() might fail due to OOM, protect against that
2014-06-27 18:34:37 +02:00
_cleanup_free_ char *t;
t = uid_to_name(getuid());
test-engine: fix access to unit load path Also add a bit of debugging output to help diagnose problems, add missing units, and simplify cppflags. Move test-engine to normal tests from manual tests, it should now work without destroying the system.
2014-01-07 14:41:24 +01:00
log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
arg_action == ACTION_TEST ? " test" : "", getuid(), t);
Improve messages about user mode a bit
2014-01-08 23:20:45 +01:00
}
manager: make running_as configurable
2010-04-06 23:55:42 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_action == ACTION_RUN) {
if (arg_system && !skip_setup) {
if (arg_show_status > 0)
status_welcome();
nss: don't disable nscd anymore, since it doesn't make sense to socket-activate nscd anyway
2010-08-17 18:00:48 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
hostname_setup();
machine_id_setup(NULL, arg_machine_id, NULL);
loopback_setup();
bump_unix_max_dgram_qlen();
ask-password: add minimal framework to allow services query SSL/harddisk passphrases from the user
2010-09-17 01:26:29 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
test_usr();
}
hostname: set hostname early during boottime
2010-04-10 05:00:38 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
watchdog_set_timeout(&arg_runtime_watchdog);
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_timer_slack_nsec != NSEC_INFINITY)
if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
log_error_errno(errno, "Failed to adjust timer slack: %m");
main: allow setting of timer slack for PID 1
2012-05-31 04:36:08 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
if (r < 0) {
log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
error_message = "Failed to drop capability bounding set of usermode helpers";
goto finish;
}
r = capability_bounding_set_drop(arg_capability_bounding_set, true);
if (r < 0) {
log_emergency_errno(r, "Failed to drop capability bounding set: %m");
error_message = "Failed to drop capability bounding set";
goto finish;
}
capabilities: when dropping capabilities system-wide also drop them from usermode helpers This hooks things up with /proc/sys/kernel/usermodehelper/bset and /proc/sys/kernel/usermodehelper/inheritable.
2012-05-29 23:33:38 +02:00
}
main: add configuration option to alter capability bounding set for PID 1 This also ensures that caps dropped from the bounding set are also dropped from the inheritable set, to be extra-secure. Usually that should change very little though as the inheritable set is empty for all our uses anyway.
2012-05-24 04:00:56 +02:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (arg_syscall_archs) {
r = enforce_syscall_archs(arg_syscall_archs);
if (r < 0) {
error_message = "Failed to set syscall architectures";
goto finish;
}
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
if (!arg_system)
/* Become reaper of our children */
if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
log_warning_errno(errno, "Failed to make us a subreaper: %m");
main: set PR_SET_CHILD_REAPER for MANAGER_USER Become the reaper for all children part of the user session. Tested with several forking services.
2012-07-25 00:13:11 +02:00
main: bump RLIMIT_NOFILE for the root user substantially On current kernels BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK even for privileged users that have CAP_IPC_LOCK. Given that mlock() generally ignores RLIMIT_MEMLOCK if CAP_IPC_LOCK is set this appears to be an oversight in the kernel. Either way, until that's fixed, let's just bump RLIMIT_MEMLOCK for the root user considerably, as the default is quite limiting, and doesn't permit us to create more than a few TRIE maps.
2017-09-21 19:43:07 +02:00
if (arg_system) {
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
/* Bump up RLIMIT_NOFILE for systemd itself */
(void) bump_rlimit_nofile(&saved_rlimit_nofile);
main: bump RLIMIT_NOFILE for the root user substantially On current kernels BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK even for privileged users that have CAP_IPC_LOCK. Given that mlock() generally ignores RLIMIT_MEMLOCK if CAP_IPC_LOCK is set this appears to be an oversight in the kernel. Either way, until that's fixed, let's just bump RLIMIT_MEMLOCK for the root user considerably, as the default is quite limiting, and doesn't permit us to create more than a few TRIE maps.
2017-09-21 19:43:07 +02:00
(void) bump_rlimit_memlock(&saved_rlimit_memlock);
}
main: skip many initialization steps when running in --test mode Most importantly, don't collect open socket activation fds when in --test mode. This specifically created a problem because we invoke pager_open() beforehand (which these days makes copies of the original stdout/stderr in order to be able to restore them when the pager goes away) and we might mistakenly the fd copies it creates as socket activation fds. Fixes: #6383
2017-09-13 10:31:40 +02:00
}
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
Make test_run into a flags field and disable generators again Now generators are only run in systemd --test mode, where this makes most sense (how are you going to test what would happen otherwise?). Fixes #6842. v2: - rename test_run to test_run_flags
2017-09-16 11:19:43 +02:00
r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
&m);
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
if (r < 0) {
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_emergency_errno(r, "Failed to allocate manager object: %m");
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to allocate manager object";
initial commit
2009-11-18 00:42:52 +01:00
goto finish;
}
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
m->confirm_spawn = arg_confirm_spawn;
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
m->runtime_watchdog = arg_runtime_watchdog;
m->shutdown_watchdog = arg_shutdown_watchdog;
systemd: record the timestamps as early as possible The time for systemd initialization and selinux policy loading is accounted to the initrd or the kernel, which is wrong. Instead of: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 975ms (initrd) + 1.410s (userspace) = 8.647s the more correct output is: Startup finished in 5.559s (firmware) + 36ms (loader) + 665ms (kernel) + 475ms (initrd) + 1.910s (userspace) = 8.647s
2013-04-24 17:15:47 +02:00
m->userspace_timestamp = userspace_timestamp;
m->kernel_timestamp = kernel_timestamp;
m->initrd_timestamp = initrd_timestamp;
analyze: plot the time spent setting up security modules
2013-11-10 23:21:15 +01:00
m->security_start_timestamp = security_start_timestamp;
m->security_finish_timestamp = security_finish_timestamp;
core: add possibility to set action for ctrl-alt-del burst (#4105) For some certification, it should not be possible to reboot the machine through ctrl-alt-delete. Currently we suggest our customers to mask the ctrl-alt-delete target, but that is obviously not enough. Patching the keymaps to disable that is really not a way to go for them, because the settings need to be easily checked by some SCAP tools.
2016-10-07 03:08:21 +02:00
m->cad_burst_action = arg_cad_burst_action;
manager: optionally print status updates to console on boot
2010-07-07 00:00:59 +02:00
Reload manager defaults at daemon-reload "systemctl daemon-reload" should also update the manager defaults from /etc/systemd/system.conf. For details, see: http://lists.freedesktop.org/archives/systemd-devel/2015-June/033062.html Amended to use manager_set_defaults() as common function.
2015-06-29 11:26:27 +02:00
manager_set_defaults(m);
fsck: show progress while fscking at boot
2011-09-01 21:05:06 +02:00
manager_set_show_status(m, arg_show_status);
firstboot: get rid of firstboot generator again, introduce ConditionFirstBoot= instead As Zbigniew pointed out a new ConditionFirstBoot= appears like the nicer way to hook in systemd-firstboot.service on first boots (those with /etc unpopulated), so let's do this, and get rid of the generator again.
2014-07-07 19:25:31 +02:00
manager_set_first_boot(m, empty_etc);
fsck: show progress while fscking at boot
2011-09-01 21:05:06 +02:00
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
/* Remember whether we should queue the default job */
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
queue_default_job = !arg_serialization || arg_switched_root;
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
main: profile unit file loading
2011-07-28 23:41:57 +02:00
before_startup = now(CLOCK_MONOTONIC);
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
r = manager_startup(m, arg_serialization, fds);
manager: fix handling of failure in initialization We would warn and continue after failure in manager_startup, but there's no way we can continue. We must fail.
2017-01-22 07:35:33 +01:00
if (r < 0) {
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_error_errno(r, "Failed to fully start up daemon: %m");
manager: fix handling of failure in initialization We would warn and continue after failure in manager_startup, but there's no way we can continue. We must fail.
2017-01-22 07:35:33 +01:00
goto finish;
}
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
/* This will close all file descriptors that were opened, but
* not claimed by any unit. */
tree-wide: take benefit of the fact that fdset_free() returns NULL
2015-09-23 01:05:55 +02:00
fds = fdset_free(fds);
implement coldpluggin
2010-01-29 03:18:09 +01:00
util: introduce safe_fclose() and port everything over to it Adds a coccinelle script to port things over automatically.
2015-09-09 15:20:10 +02:00
arg_serialization = safe_fclose(arg_serialization);
main: properly queue default.target after switched root This also introduces rd.systemd.unit= to specify the unit to boot into in the initrd.
2012-05-22 02:35:22 +02:00
if (queue_default_job) {
tree-wide: expose "p"-suffix unref calls in public APIs to make gcc cleanup easy GLIB has recently started to officially support the gcc cleanup attribute in its public API, hence let's do the same for our APIs. With this patch we'll define an xyz_unrefp() call for each public xyz_unref() call, to make it easy to use inside a __attribute__((cleanup())) expression. Then, all code is ported over to make use of this. The new calls are also documented in the man pages, with examples how to use them (well, I only added docs where the _unref() call itself already had docs, and the examples, only cover sd_bus_unrefp() and sd_event_unrefp()). This also renames sd_lldp_free() to sd_lldp_unref(), since that's how we tend to call our destructors these days. Note that this defines no public macro that wraps gcc's attribute and makes it easier to use. While I think it's our duty in the library to make our stuff easy to use, I figure it's not our duty to make gcc's own features easy to use on its own. Most likely, client code which wants to make use of this should define its own: #define _cleanup_(function) __attribute__((cleanup(function))) Or similar, to make the gcc feature easier to use. Making this logic public has the benefit that we can remove three header files whose only purpose was to define these functions internally. See #2008.
2015-11-27 19:13:45 +01:00
_cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
main: properly fallback to rescue.target if default.target is hosed
2010-08-17 20:42:53 +02:00
Unit *target = NULL;
manager: tell correctly if the manager is booting The assumption that the initial job is the job with id==1 is incorrect. Some jobs may be enqueued before the job that starts the default unit as in this example: -.mount changed dead -> mounted Trying to enqueue job quotacheck.service/start/fail Installed new job quotacheck.service/start as 1 Installed new job systemd-stdout-syslog-bridge.socket/start as 2 Enqueued job quotacheck.service/start as 1 Trying to enqueue job quotaon.service/start/fail Installed new job quotaon.service/start as 5 Enqueued job quotaon.service/start as 5 Activating default unit: default.target Trying to enqueue job graphical.target/start/replace This fixes a bug where displaying of boot status messages was turned off too early.
2012-02-02 12:39:33 +01:00
Job *default_unit_job;
dbus: make errors reported via D-Bus more useful
2010-07-08 02:43:18 +02:00
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
log_debug("Activating default unit: %s", arg_default_unit);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
core: convert PID 1 to libsystemd-bus This patch converts PID 1 to libsystemd-bus and thus drops the dependency on libdbus. The only remaining code using libdbus is a test case that validates our bus marshalling against libdbus' marshalling, and this dependency can be turned off. This patch also adds a couple of things to libsystem-bus, that are necessary to make the port work: - Synthesizing of "Disconnected" messages when bus connections are severed. - Support for attaching multiple vtables for the same interface on the same path. This patch also fixes the SetDefaultTarget() and GetDefaultTarget() bus calls which used an inappropriate signature. As a side effect we will now generate PropertiesChanged messages which carry property contents, rather than just invalidation information.
2013-11-19 21:12:59 +01:00
if (r < 0)
log_error("Failed to load default target: %s", bus_error_message(&error, r));
else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
treewide: more log_*_errno() conversions, multiline calls Basically: find . -name '*.[ch]' | while read f; do perl -i.mmm -e \ 'local $/; local $_=<>; s/log_(debug|info|notice|warning|error|emergency)\("([^"]*)%s"([^;]*),\s*strerror\(-?([->a-zA-Z_]+)\)\);/log_\1_errno(\4, "\2%m"\3);/gms;print;' \ $f; done Plus manual indentation fixups.
2014-11-28 17:09:20 +01:00
log_error_errno(target->load_error, "Failed to load default target: %m");
unit: remove union Unit Now that objects of all unit types are allocated the exact amount of memory they need, the Unit union has lost its purpose. Remove it. "Unit" is a more natural name for the base unit class than "Meta", so rename Meta to Unit. Access to members of the base class gets simplified.
2012-01-15 12:04:08 +01:00
else if (target->load_state == UNIT_MASKED)
unit: rename 'banned' load state to 'masked'
2010-10-08 18:21:52 +02:00
log_error("Default target masked.");
main: initialize default boot target from argv[1]
2010-04-04 22:49:26 +02:00
unit: remove union Unit Now that objects of all unit types are allocated the exact amount of memory they need, the Unit union has lost its purpose. Remove it. "Unit" is a more natural name for the base unit class than "Meta", so rename Meta to Unit. Access to members of the base class gets simplified.
2012-01-15 12:04:08 +01:00
if (!target || target->load_state != UNIT_LOADED) {
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
log_info("Trying to load rescue target...");
main: properly fallback to rescue.target if default.target is hosed
2010-08-17 20:42:53 +02:00
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
if (r < 0) {
manager: log some fatal errors at emergency level This adds a new log_emergency() function, which is equivalent to log_error() for non-PID-1, and logs at the highest priority for PID 1. Some messages which occur before freezing are converted to use it.
2014-11-06 06:04:06 +01:00
log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to load rescue target";
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
goto finish;
main: fix loading of default target
2013-06-28 18:37:15 +02:00
} else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
treewide: more log_*_errno() conversions, multiline calls Basically: find . -name '*.[ch]' | while read f; do perl -i.mmm -e \ 'local $/; local $_=<>; s/log_(debug|info|notice|warning|error|emergency)\("([^"]*)%s"([^;]*),\s*strerror\(-?([->a-zA-Z_]+)\)\);/log_\1_errno(\4, "\2%m"\3);/gms;print;' \ $f; done Plus manual indentation fixups.
2014-11-28 17:09:20 +01:00
log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to load rescue target";
main: properly fallback to rescue.target if default.target is hosed
2010-08-17 20:42:53 +02:00
goto finish;
unit: remove union Unit Now that objects of all unit types are allocated the exact amount of memory they need, the Unit union has lost its purpose. Remove it. "Unit" is a more natural name for the base unit class than "Meta", so rename Meta to Unit. Access to members of the base class gets simplified.
2012-01-15 12:04:08 +01:00
} else if (target->load_state == UNIT_MASKED) {
manager: log some fatal errors at emergency level This adds a new log_emergency() function, which is equivalent to log_error() for non-PID-1, and logs at the highest priority for PID 1. Some messages which occur before freezing are converted to use it.
2014-11-06 06:04:06 +01:00
log_emergency("Rescue target masked.");
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Rescue target masked";
unit: introduce 'banned' load state for units symlinked to /dev/null
2010-10-08 02:31:36 +02:00
goto finish;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
}
}
main: fall back to rescue target when default target cannot be loaded
2010-04-08 02:00:40 +02:00
unit: remove union Unit Now that objects of all unit types are allocated the exact amount of memory they need, the Unit union has lost its purpose. Remove it. "Unit" is a more natural name for the base unit class than "Meta", so rename Meta to Unit. Access to members of the base class gets simplified.
2012-01-15 12:04:08 +01:00
assert(target->load_state == UNIT_LOADED);
unit: introduce 'banned' load state for units symlinked to /dev/null
2010-10-08 02:31:36 +02:00
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
if (arg_action == ACTION_TEST) {
get rid of unicode arrows, since the linux console cannot show them, not even in unicode mode
2010-04-23 20:25:55 +02:00
printf("-> By units:\n");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
manager_dump_units(m, stdout, "\t");
}
core: drop "override" flag when building transactions Now that we don't have RequiresOverridable= and RequisiteOverridable= dependencies anymore, we can get rid of tracking the "override" boolean for jobs in the job engine, as it serves no purpose anymore. While we are at it, fix some error messages we print when invoking functions that take the override parameter.
2015-11-12 19:52:31 +01:00
r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
core: if we cannot JOB_ISOLATE the default target JOB_REPLACE it instead In order to maintain compatibility with older initrds which do not have AllowIsolate=yes set for their target units, fallback to JOB_REPLACE if JOB_ISOLATE doesn't work, but complain about it.
2013-03-07 22:18:34 +01:00
if (r == -EPERM) {
core: convert PID 1 to libsystemd-bus This patch converts PID 1 to libsystemd-bus and thus drops the dependency on libdbus. The only remaining code using libdbus is a test case that validates our bus marshalling against libdbus' marshalling, and this dependency can be turned off. This patch also adds a couple of things to libsystem-bus, that are necessary to make the port work: - Synthesizing of "Disconnected" messages when bus connections are severed. - Support for attaching multiple vtables for the same interface on the same path. This patch also fixes the SetDefaultTarget() and GetDefaultTarget() bus calls which used an inappropriate signature. As a side effect we will now generate PropertiesChanged messages which carry property contents, rather than just invalidation information.
2013-11-19 21:12:59 +01:00
log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
core: if we cannot JOB_ISOLATE the default target JOB_REPLACE it instead In order to maintain compatibility with older initrds which do not have AllowIsolate=yes set for their target units, fallback to JOB_REPLACE if JOB_ISOLATE doesn't work, but complain about it.
2013-03-07 22:18:34 +01:00
core: make sure to reset the bus error struct before reusing it Otherwise the call might fail, because the error structure is already initialized.
2015-11-12 20:14:33 +01:00
sd_bus_error_free(&error);
core: drop "override" flag when building transactions Now that we don't have RequiresOverridable= and RequisiteOverridable= dependencies anymore, we can get rid of tracking the "override" boolean for jobs in the job engine, as it serves no purpose anymore. While we are at it, fix some error messages we print when invoking functions that take the override parameter.
2015-11-12 19:52:31 +01:00
r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
core: if we cannot JOB_ISOLATE the default target JOB_REPLACE it instead In order to maintain compatibility with older initrds which do not have AllowIsolate=yes set for their target units, fallback to JOB_REPLACE if JOB_ISOLATE doesn't work, but complain about it.
2013-03-07 22:18:34 +01:00
if (r < 0) {
manager: log some fatal errors at emergency level This adds a new log_emergency() function, which is equivalent to log_error() for non-PID-1, and logs at the highest priority for PID 1. Some messages which occur before freezing are converted to use it.
2014-11-06 06:04:06 +01:00
log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to start default target";
core: if we cannot JOB_ISOLATE the default target JOB_REPLACE it instead In order to maintain compatibility with older initrds which do not have AllowIsolate=yes set for their target units, fallback to JOB_REPLACE if JOB_ISOLATE doesn't work, but complain about it.
2013-03-07 22:18:34 +01:00
goto finish;
}
} else if (r < 0) {
manager: log some fatal errors at emergency level This adds a new log_emergency() function, which is equivalent to log_error() for non-PID-1, and logs at the highest priority for PID 1. Some messages which occur before freezing are converted to use it.
2014-11-06 06:04:06 +01:00
log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to isolate default target";
main: fall back to rescue target when default target cannot be loaded
2010-04-08 02:00:40 +02:00
goto finish;
}
core: if we cannot JOB_ISOLATE the default target JOB_REPLACE it instead In order to maintain compatibility with older initrds which do not have AllowIsolate=yes set for their target units, fallback to JOB_REPLACE if JOB_ISOLATE doesn't work, but complain about it.
2013-03-07 22:18:34 +01:00
manager: tell correctly if the manager is booting The assumption that the initial job is the job with id==1 is incorrect. Some jobs may be enqueued before the job that starts the default unit as in this example: -.mount changed dead -> mounted Trying to enqueue job quotacheck.service/start/fail Installed new job quotacheck.service/start as 1 Installed new job systemd-stdout-syslog-bridge.socket/start as 2 Enqueued job quotacheck.service/start as 1 Trying to enqueue job quotaon.service/start/fail Installed new job quotaon.service/start as 5 Enqueued job quotaon.service/start as 5 Activating default unit: default.target Trying to enqueue job graphical.target/start/replace This fixes a bug where displaying of boot status messages was turned off too early.
2012-02-02 12:39:33 +01:00
m->default_unit_job_id = default_unit_job->id;
initial commit
2009-11-18 00:42:52 +01:00
main: show load profiling in test mode, too
2011-07-31 18:13:59 +02:00
after_startup = now(CLOCK_MONOTONIC);
log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
"Loaded units and determined initial transaction in %s.",
core: limit timestamp to sane precision Anything below .1 s is meaningless anyway.
2014-10-01 14:34:05 +02:00
format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
main: show load profiling in test mode, too
2011-07-31 18:13:59 +02:00
main: add a native implementation of the 'nomodules' kernel option understood by fedora init scripts
2010-07-06 20:21:08 +02:00
if (arg_action == ACTION_TEST) {
get rid of unicode arrows, since the linux console cannot show them, not even in unicode mode
2010-04-23 20:25:55 +02:00
printf("-> By jobs:\n");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
manager_dump_jobs(m, stdout, "\t");
systemctl: rework exit codes for all utility programs to follow LSB or other standards
2010-08-31 21:05:54 +02:00
retval = EXIT_SUCCESS;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
goto finish;
}
main: implement --test
2010-04-07 00:10:17 +02:00
}
rework config file load logic
2010-01-28 02:44:47 +01:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
for (;;) {
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
r = manager_loop(m);
if (r < 0) {
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_emergency_errno(r, "Failed to run main loop: %m");
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
error_message = "Failed to run main loop";
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
goto finish;
}
first try at implementing job creation
2010-01-19 04:15:20 +01:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
switch (m->exit_code) {
main: implement --test
2010-04-07 00:10:17 +02:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
case MANAGER_RELOAD:
main: always log when we reexecute or reload
2010-07-07 17:44:11 +02:00
log_info("Reloading.");
Reload manager defaults at daemon-reload "systemctl daemon-reload" should also update the manager defaults from /etc/systemd/system.conf. For details, see: http://lists.freedesktop.org/archives/systemd-devel/2015-June/033062.html Amended to use manager_set_defaults() as common function.
2015-06-29 11:26:27 +02:00
r = parse_config_file();
if (r < 0)
log_error("Failed to parse config file.");
manager_set_defaults(m);
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
r = manager_reload(m);
if (r < 0)
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_error_errno(r, "Failed to reload: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
break;
extend test a little
2010-01-20 04:02:39 +01:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
case MANAGER_REEXECUTE:
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
core: s/reexection/reexecution/ typo fix Signed-off-by: Jan Pokorný <jpokorny@redhat.com>
2015-08-07 17:13:15 +02:00
error_message = "Failed to prepare for reexecution";
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
initial commit
2009-11-18 00:42:52 +01:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
reexecute = true;
main: always log when we reexecute or reload
2010-07-07 17:44:11 +02:00
log_notice("Reexecuting.");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
goto finish;
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
case MANAGER_SWITCH_ROOT:
/* Steal the switch root parameters */
switch-root: move switch_root() call into its own .c file
2012-05-21 19:33:39 +02:00
switch_root_dir = m->switch_root;
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
switch_root_init = m->switch_root_init;
m->switch_root = m->switch_root_init = NULL;
if (!switch_root_init)
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
core: s/reexection/reexecution/ typo fix Signed-off-by: Jan Pokorný <jpokorny@redhat.com>
2015-08-07 17:13:15 +02:00
error_message = "Failed to prepare for reexecution";
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
goto finish;
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
reexecute = true;
log_notice("Switching root.");
goto finish;
containers: systemd exits with non-zero code When a systemd service running in a container exits with a non-zero code, it can be useful to terminate the container immediately and get the exit code back to the host, when systemd-nspawn returns. This was not possible to do. This patch adds the following to make it possible: - Add a read-only "ExitCode" property on PID 1's "Manager" bus object. By default, it is 0 so the behaviour stays the same as previously. - Add a method "SetExitCode" on the same object. The method fails when called on baremetal: it is only allowed in containers or in user session. - Add support in systemctl to call "systemctl exit 42". It reuses the existing code for user session. - Add exit.target and systemd-exit.service to the system instance. - Change main() to actually call systemd-shutdown to exit() with the correct value. - Add verb 'exit' in systemd-shutdown with parameter --exit-code - Update systemctl manpage. I used the following to test it: | $ sudo rkt --debug --insecure-skip-verify run \ | --mds-register=false --local docker://busybox \ | --exec=/bin/chroot -- /proc/1/root \ | systemctl --force exit 42 | ... | Container rkt-895a0cba-5c66-4fa5-831c-e3f8ddc5810d failed with error code 42. | $ echo $? | 42 Fixes https://github.com/systemd/systemd/issues/1290
2015-09-18 13:37:34 +02:00
case MANAGER_EXIT:
core: properly use the ExitCode bus property when exiting is --user instance
2015-09-23 19:50:10 +02:00
retval = m->return_value;
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
if (MANAGER_IS_USER(m)) {
core: fix shutdown of --user Shutting down a user session currently fails with: Sep 22 22:35:38 david-t2 systemd[640]: Reached target Shutdown. Sep 22 22:35:38 david-t2 systemd[640]: Starting Exit the Session... Sep 22 22:35:38 david-t2 systemd[640]: Received SIGRTMIN+24 from PID 659 (kill). Sep 22 22:35:38 david-t2 systemd[640]: Shutting down. Sep 22 22:35:38 david-t2 systemd[640]: Not executed by init (PID 1). Sep 22 22:35:38 david-t2 systemd[640]: Critical error while doing system shutdown: Operation not permitted This is a regression from: commit 287419c119ef961db487a281162ab037eba70c61 Author: Alban Crequy <alban.crequy@gmail.com> Date: Fri Sep 18 13:37:34 2015 +0200 containers: systemd exits with non-zero code Make sure we never ever execute systemd-shutdown from within a user-manager. Restore the previous behavior by partially reverting given commit.
2015-09-22 22:56:01 +02:00
log_debug("Exit.");
goto finish;
}
/* fallthrough */
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
case MANAGER_REBOOT:
case MANAGER_POWEROFF:
case MANAGER_HALT:
case MANAGER_KEXEC: {
static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
containers: systemd exits with non-zero code When a systemd service running in a container exits with a non-zero code, it can be useful to terminate the container immediately and get the exit code back to the host, when systemd-nspawn returns. This was not possible to do. This patch adds the following to make it possible: - Add a read-only "ExitCode" property on PID 1's "Manager" bus object. By default, it is 0 so the behaviour stays the same as previously. - Add a method "SetExitCode" on the same object. The method fails when called on baremetal: it is only allowed in containers or in user session. - Add support in systemctl to call "systemctl exit 42". It reuses the existing code for user session. - Add exit.target and systemd-exit.service to the system instance. - Change main() to actually call systemd-shutdown to exit() with the correct value. - Add verb 'exit' in systemd-shutdown with parameter --exit-code - Update systemctl manpage. I used the following to test it: | $ sudo rkt --debug --insecure-skip-verify run \ | --mds-register=false --local docker://busybox \ | --exec=/bin/chroot -- /proc/1/root \ | systemctl --force exit 42 | ... | Container rkt-895a0cba-5c66-4fa5-831c-e3f8ddc5810d failed with error code 42. | $ echo $? | 42 Fixes https://github.com/systemd/systemd/issues/1290
2015-09-18 13:37:34 +02:00
[MANAGER_EXIT] = "exit",
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
[MANAGER_REBOOT] = "reboot",
[MANAGER_POWEROFF] = "poweroff",
[MANAGER_HALT] = "halt",
[MANAGER_KEXEC] = "kexec"
};
assert_se(shutdown_verb = table[m->exit_code]);
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
log_notice("Shutting down.");
goto finish;
}
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
default:
assert_not_reached("Unknown exit code.");
}
}
main: parse our own command line and the kernel command line
2010-04-06 23:40:24 +02:00
initial commit
2009-11-18 00:42:52 +01:00
finish:
systemd: use pager for --test and --help
2014-01-07 06:00:05 +01:00
pager_close();
core: properly use the ExitCode bus property when exiting is --user instance
2015-09-23 19:50:10 +02:00
if (m)
core: make setting the shutdown watchdog configuration via dbus work https://bugs.freedesktop.org/show_bug.cgi?id=88284
2015-01-29 01:23:07 +01:00
arg_shutdown_watchdog = m->shutdown_watchdog;
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
manager: let manager_free() handle NULLs This makes the calling code a bit simpler.
2014-11-08 16:06:12 +01:00
m = manager_free(m);
initial commit
2009-11-18 00:42:52 +01:00
tree-wide: update empty-if coccinelle script to cover empty-while and more Let's also clean up single-line while and for blocks.
2015-09-09 14:23:02 +02:00
for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
tree-wide: use coccinelle to patch a lot of code to use mfree() This replaces this: free(p); p = NULL; by this: p = mfree(p); Change generated using coccinelle. Semantic patch is added to the sources.
2015-09-08 18:43:11 +02:00
arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
main: allow system wide limits for services
2012-03-21 18:03:40 +01:00
tree-wide: introduce mfree() Pretty trivial helper which wraps free() but returns NULL, so we can simplify this: free(foobar); foobar = NULL; to this: foobar = mfree(foobar);
2015-07-31 19:56:38 +02:00
arg_default_unit = mfree(arg_default_unit);
core: allow to redirect confirmation messages to a different console It's rather hard to parse the confirmation messages (enabled with systemd.confirm_spawn=true) amongst the status messages and the kernel ones (if enabled). This patch gives the possibility to the user to redirect the confirmation message to a different virtual console, either by giving its name or its path, so those messages are separated from the other ones and easier to read.
2016-11-02 10:38:22 +01:00
arg_confirm_spawn = mfree(arg_confirm_spawn);
strv: add strv_free_free() to strv.c and make use of it Let's teach it a new trick, and make it return NULL.
2015-09-23 01:01:26 +02:00
arg_join_controllers = strv_free_free(arg_join_controllers);
tree-wide: introduce mfree() Pretty trivial helper which wraps free() but returns NULL, so we can simplify this: free(foobar); foobar = NULL; to this: foobar = mfree(foobar);
2015-07-31 19:56:38 +02:00
arg_default_environment = strv_free(arg_default_environment);
tree-wide: take benefit of the fact that hashmap_free() returns NULL And set_free() too. Another Coccinelle patch.
2015-09-09 23:12:07 +02:00
arg_syscall_archs = set_free(arg_syscall_archs);
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
mac: rename apis with mac_{selinux/smack}_ prefix
2014-10-23 10:23:46 +02:00
mac_selinux_finish();
console: rework automatic getty on kernel console logic again It is essential that the gettys are proper dependencies from getty.target so that they aren't killed and immediately restarted on runlevel changes. Hence rework the logic to implicitly add console gettys to getty.target as dependencies. This also adds an automatic hvc console for virtualizers. https://bugzilla.redhat.com/show_bug.cgi?id=501720
2010-08-30 21:31:40 +02:00
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
if (reexecute) {
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
const char **args;
main: fix assertion failure due to use of ELEMENTSOF on a non-array
2012-05-09 11:13:26 +02:00
unsigned i, args_size;
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
/* 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);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
/* 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)
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
main: bump RLIMIT_NOFILE for the root user substantially On current kernels BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK even for privileged users that have CAP_IPC_LOCK. Given that mlock() generally ignores RLIMIT_MEMLOCK if CAP_IPC_LOCK is set this appears to be an oversight in the kernel. Either way, until that's fixed, let's just bump RLIMIT_MEMLOCK for the root user considerably, as the default is quite limiting, and doesn't permit us to create more than a few TRIE maps.
2017-09-21 19:43:07 +02:00
if (saved_rlimit_memlock.rlim_cur != (rlim_t) -1)
(void) setrlimit(RLIMIT_MEMLOCK, &saved_rlimit_memlock);
main: bump up RLIMIT_NOFILE for systemd itself For setups with many listening sockets the default kernel resource limit of 1024 fds is not enough. Bump this up to 64K to avoid any limitations in this regard. We are careful to pass on the kernel default to daemons however, since normally resource limits are a good to enforce, especially since select() can't handle fds > 1023.
2012-09-17 16:35:59 +02:00
switch-root: move switch_root() call into its own .c file
2012-05-21 19:33:39 +02:00
if (switch_root_dir) {
switch-root: don't wait for processes When we transition from the initrd to the main system, don't reap processes, so that they can be handled normally after deserialization.
2012-07-17 17:44:26 +02:00
/* Kill all remaining processes from the
* initrd, but don't wait for them, so that we
* can handle the SIGCHLD for them after
* deserializing. */
shutdown: during final killing spree also send SIGHUP in addition to SIGTERM to deal with shells This makes shutdown a bit faster if debug-shell.service is enabled.
2013-11-25 18:08:02 +01:00
broadcast_signal(SIGTERM, false, true);
core: go on a killing spree when transitioning from initrd to main system
2012-07-13 14:41:57 +02:00
use the switch_root function in shutdown removes code duplication also move switch-root to shared
2014-08-21 16:21:26 +02:00
/* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
switch-root: move switch_root() call into its own .c file
2012-05-21 19:33:39 +02:00
if (r < 0)
treewide: no need to negate errno for log_*_errno() It corrrectly handles both positive and negative errno values.
2014-11-28 13:19:16 +01:00
log_error_errno(r, "Failed to switch root, trying to continue: %m");
switch-root: move switch_root() call into its own .c file
2012-05-21 19:33:39 +02:00
}
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
core/main.c: add "--switchedroot" parameter If systemd serializes from a switch_root, it adds "--switchedroot" to the systemd in the real root. If "--switchedroot" is found, then we do not skip all the stuff, which is skipped for normal rexecs.
2012-05-16 14:22:42 +02:00
args_size = MAX(6, argc+1);
main: fix assertion failure due to use of ELEMENTSOF on a non-array
2012-05-09 11:13:26 +02:00
args = newa(const char*, args_size);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
if (!switch_root_init) {
Add a snprinf wrapper which checks that the buffer was big enough If we scale our buffer to be wide enough for the format string, we should expect that the calculation was correct. char_array_0() invocations are removed, since snprintf nul-terminates the output in any case. A similar wrapper is used for strftime calls, but only in timedatectl.c.
2015-01-27 14:00:11 +01:00
char sfd[DECIMAL_STR_MAX(int) + 1];
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
/* 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. */
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
assert(arg_serialization);
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
assert(fds);
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
Add a snprinf wrapper which checks that the buffer was big enough If we scale our buffer to be wide enough for the format string, we should expect that the calculation was correct. char_array_0() invocations are removed, since snprintf nul-terminates the output in any case. A similar wrapper is used for strftime calls, but only in timedatectl.c.
2015-01-27 14:00:11 +01:00
xsprintf(sfd, "%i", fileno(arg_serialization));
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
i = 0;
args[i++] = SYSTEMD_BINARY_PATH;
switch-root: move switch_root() call into its own .c file
2012-05-21 19:33:39 +02:00
if (switch_root_dir)
main: rename a few fix to follow general naming style
2012-05-21 19:48:04 +02:00
args[i++] = "--switched-root";
core: remove ManagerRunningAs enum Previously, we had two enums ManagerRunningAs and UnitFileScope, that were mostly identical and converted from one to the other all the time. The latter had one more value UNIT_FILE_GLOBAL however. Let's simplify things, and remove ManagerRunningAs and replace it by UnitFileScope everywhere, thus making the translation unnecessary. Introduce two new macros MANAGER_IS_SYSTEM() and MANAGER_IS_USER() to simplify checking if we are running in one or the user context.
2016-02-24 21:24:23 +01:00
args[i++] = arg_system ? "--system" : "--user";
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
args[i++] = "--deserialize";
args[i++] = sfd;
args[i++] = NULL;
main: replace --running-as= by --session and --system do mimic related tools and D-Bus
2010-07-13 18:57:58 +02:00
main: fix assertion failure due to use of ELEMENTSOF on a non-array
2012-05-09 11:13:26 +02:00
assert(i <= args_size);
core: add valgrind helper for daemon-reexec Inspired by https://github.com/systemd/systemd/issues/2187#issuecomment-165587140
2016-01-19 16:48:45 +01:00
/*
* We want valgrind to print its memory usage summary before reexecution.
* Valgrind won't do this is on its own on exec(), but it will do it on exit().
* Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
* so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
*/
valgrind_summary_hack();
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) execv(args[0], (char* const*) args);
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
}
service: hide output of sysv scripts if quiet is passed on the kernel cmdline
2010-08-09 18:00:24 +02:00
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
/* 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.) */
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
util: introduce safe_fclose() and port everything over to it Adds a coccinelle script to port things over automatically.
2015-09-09 15:20:10 +02:00
arg_serialization = safe_fclose(arg_serialization);
tree-wide: take benefit of the fact that fdset_free() returns NULL
2015-09-23 01:05:55 +02:00
fds = fdset_free(fds);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
Revert "pid1: reconnect to the console before being re-executed" This reverts commit affd7ed1a923b0df8479cff1bd9eafb625fdaa66. > So it looks like make_console_stdio() has bad side effect. More specifically it > does a TIOCSCTTY ioctl (via acquire_terminal()) which sees to disturb the > process which was using/owning the console. Fixes #3842. https://bugs.debian.org/834367 https://bugzilla.redhat.com/show_bug.cgi?id=1367766
2016-08-19 03:39:39 +02:00
/* Reopen the console */
(void) make_console_stdio();
core: add a system-wide SystemCallArchitectures= setting This is useful to prohibit execution of non-native processes on systems, for example 32bit binaries on 64bit systems, this lowering the attack service on incorrect syscall and ioctl 32→64bit mappings.
2014-02-13 01:35:27 +01:00
for (j = 1, i = 1; j < (unsigned) argc; j++)
manager: introduce SwitchRoot bus call for initrd/main transition
2012-05-09 01:24:50 +02:00
args[i++] = argv[j];
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
args[i++] = NULL;
main: fix assertion failure due to use of ELEMENTSOF on a non-array
2012-05-09 11:13:26 +02:00
assert(i <= args_size);
main: try a bit harder to find an init process to execute after reexec/switch-root
2012-05-21 17:26:19 +02:00
util: make use of newly added reset_signal_mask() call wherever appropriate
2014-08-26 21:11:35 +02:00
/* Reenable any blocked signals, especially important
Reset signal-mask on re-exec to init=.. Process 1 (aka init) needs to be started with an empty signal mask. That includes the process 1 that's started after the initrd is finished. When the initrd is using systemd (as it does with dracut based initrds) then it is systemd that calls the real init. Normally this is systemd again, except when the user uses for instance "init=/bin/bash" on the kernel command line.
2014-06-13 16:41:06 +02:00
* if we switch from initial ramdisk to init=... */
tree-wide: whenever we fork off a foreign child process reset signal mask/handlers Also, when the child is potentially long-running make sure to set a death signal. Also, ignore the result of the reset operations explicitly by casting them to (void).
2015-05-31 23:55:55 +02:00
(void) reset_all_signal_handlers();
(void) reset_signal_mask();
Reset signal-mask on re-exec to init=.. Process 1 (aka init) needs to be started with an empty signal mask. That includes the process 1 that's started after the initrd is finished. When the initrd is using systemd (as it does with dracut based initrds) then it is systemd that calls the real init. Normally this is systemd again, except when the user uses for instance "init=/bin/bash" on the kernel command line.
2014-06-13 16:41:06 +02:00
main: try a bit harder to find an init process to execute after reexec/switch-root
2012-05-21 17:26:19 +02:00
if (switch_root_init) {
args[0] = switch_root_init;
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) execv(args[0], (char* const*) args);
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
main: try a bit harder to find an init process to execute after reexec/switch-root
2012-05-21 17:26:19 +02:00
}
args[0] = "/sbin/init";
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) execv(args[0], (char* const*) args);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
main: only fall back to /bin/sh in case /sbin/init does not exist
2012-05-22 19:11:10 +02:00
if (errno == ENOENT) {
log_warning("No /sbin/init, trying fallback");
main: try a bit harder to find an init process to execute after reexec/switch-root
2012-05-21 17:26:19 +02:00
main: only fall back to /bin/sh in case /sbin/init does not exist
2012-05-22 19:11:10 +02:00
args[0] = "/bin/sh";
args[1] = NULL;
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) execv(args[0], (char* const*) args);
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
main: only fall back to /bin/sh in case /sbin/init does not exist
2012-05-22 19:11:10 +02:00
} else
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
}
util: introduce safe_fclose() and port everything over to it Adds a coccinelle script to port things over automatically.
2015-09-09 15:20:10 +02:00
arg_serialization = safe_fclose(arg_serialization);
tree-wide: take benefit of the fact that fdset_free() returns NULL
2015-09-23 01:05:55 +02:00
fds = fdset_free(fds);
reload: implement reload/reexec logic
2010-04-21 03:27:44 +02:00
build-sys: use #if Y instead of #ifdef Y everywhere The advantage is that is the name is mispellt, cpp will warn us. $ git grep -Ee "conf.set\('(HAVE|ENABLE)_" -l|xargs sed -r -i "s/conf.set\('(HAVE|ENABLE)_/conf.set10('\1_/" $ git grep -Ee '#ifn?def (HAVE|ENABLE)' -l|xargs sed -r -i 's/#ifdef (HAVE|ENABLE)/#if \1/; s/#ifndef (HAVE|ENABLE)/#if ! \1/;' $ git grep -Ee 'if.*defined\(HAVE' -l|xargs sed -i -r 's/defined\((HAVE_[A-Z0-9_]*)\)/\1/g' $ git grep -Ee 'if.*defined\(ENABLE' -l|xargs sed -i -r 's/defined\((ENABLE_[A-Z0-9_]*)\)/\1/g' + manual changes to meson.build squash! build-sys: use #if Y instead of #ifdef Y everywhere v2: - fix incorrect setting of HAVE_LIBIDN2
2017-10-03 10:41:51 +02:00
#if HAVE_VALGRIND_VALGRIND_H
valgrind: make running PID 1 in valgrind useful Since valgrind only generates useful output on exit() (rather than exec()) we need to explicitly exit when valgrind is detected.
2013-11-20 22:11:10 +01:00
/* If we are PID 1 and running under valgrind, then let's exit
* here explicitly. valgrind will only generate nice output on
* exit(), not on exec(), hence let's do the former not the
* latter here. */
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1 && RUNNING_ON_VALGRIND)
valgrind: make running PID 1 in valgrind useful Since valgrind only generates useful output on exit() (rather than exec()) we need to explicitly exit when valgrind is detected.
2013-11-20 22:11:10 +01:00
return 0;
#endif
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
if (shutdown_verb) {
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
char log_level[DECIMAL_STR_MAX(int) + 1];
containers: systemd exits with non-zero code When a systemd service running in a container exits with a non-zero code, it can be useful to terminate the container immediately and get the exit code back to the host, when systemd-nspawn returns. This was not possible to do. This patch adds the following to make it possible: - Add a read-only "ExitCode" property on PID 1's "Manager" bus object. By default, it is 0 so the behaviour stays the same as previously. - Add a method "SetExitCode" on the same object. The method fails when called on baremetal: it is only allowed in containers or in user session. - Add support in systemctl to call "systemctl exit 42". It reuses the existing code for user session. - Add exit.target and systemd-exit.service to the system instance. - Change main() to actually call systemd-shutdown to exit() with the correct value. - Add verb 'exit' in systemd-shutdown with parameter --exit-code - Update systemctl manpage. I used the following to test it: | $ sudo rkt --debug --insecure-skip-verify run \ | --mds-register=false --local docker://busybox \ | --exec=/bin/chroot -- /proc/1/root \ | systemctl --force exit 42 | ... | Container rkt-895a0cba-5c66-4fa5-831c-e3f8ddc5810d failed with error code 42. | $ echo $? | 42 Fixes https://github.com/systemd/systemd/issues/1290
2015-09-18 13:37:34 +02:00
char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
const char* command_line[11] = {
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
SYSTEMD_SHUTDOWN_BINARY_PATH,
shutdown_verb,
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
"--log-level", log_level,
"--log-target",
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
};
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
unsigned pos = 5;
strv: multiple cleanups - turn strv_merge into strv_extend_strv. appending strv b to the end of strv a instead of creating a new strv - strv_append: remove in favor of strv_extend and strv_push. - strv_remove: write slightly more elegant - strv_remove_prefix: remove unused function - strv_overlap: use strv_contains - strv_printf: STRV_FOREACH handles NULL correctly
2014-01-04 02:35:27 +01:00
_cleanup_strv_free_ char **env_block = NULL;
main: make gcc shut up
2014-02-17 16:18:17 +01:00
assert(command_line[pos] == NULL);
strv: multiple cleanups - turn strv_merge into strv_extend_strv. appending strv b to the end of strv a instead of creating a new strv - strv_append: remove in favor of strv_extend and strv_push. - strv_remove: write slightly more elegant - strv_remove_prefix: remove unused function - strv_overlap: use strv_contains - strv_printf: STRV_FOREACH handles NULL correctly
2014-01-04 02:35:27 +01:00
env_block = strv_copy(environ);
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
Add a snprinf wrapper which checks that the buffer was big enough If we scale our buffer to be wide enough for the format string, we should expect that the calculation was correct. char_array_0() invocations are removed, since snprintf nul-terminates the output in any case. A similar wrapper is used for strftime calls, but only in timedatectl.c.
2015-01-27 14:00:11 +01:00
xsprintf(log_level, "%d", log_get_max_level());
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
switch (log_get_target()) {
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
case LOG_TARGET_KMSG:
case LOG_TARGET_JOURNAL_OR_KMSG:
case LOG_TARGET_SYSLOG_OR_KMSG:
command_line[pos++] = "kmsg";
break;
core: handle --log-target=null when calling systemd-shutdown When shutting down, if systemd was started with --log-target=null, systemd-shutdown was being called with --log-target=console.
2015-06-30 15:08:49 +02:00
case LOG_TARGET_NULL:
command_line[pos++] = "null";
break;
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
case LOG_TARGET_CONSOLE:
default:
command_line[pos++] = "console";
break;
};
if (log_get_show_color())
command_line[pos++] = "--log-color";
if (log_get_show_location())
command_line[pos++] = "--log-location";
containers: systemd exits with non-zero code When a systemd service running in a container exits with a non-zero code, it can be useful to terminate the container immediately and get the exit code back to the host, when systemd-nspawn returns. This was not possible to do. This patch adds the following to make it possible: - Add a read-only "ExitCode" property on PID 1's "Manager" bus object. By default, it is 0 so the behaviour stays the same as previously. - Add a method "SetExitCode" on the same object. The method fails when called on baremetal: it is only allowed in containers or in user session. - Add support in systemctl to call "systemctl exit 42". It reuses the existing code for user session. - Add exit.target and systemd-exit.service to the system instance. - Change main() to actually call systemd-shutdown to exit() with the correct value. - Add verb 'exit' in systemd-shutdown with parameter --exit-code - Update systemctl manpage. I used the following to test it: | $ sudo rkt --debug --insecure-skip-verify run \ | --mds-register=false --local docker://busybox \ | --exec=/bin/chroot -- /proc/1/root \ | systemctl --force exit 42 | ... | Container rkt-895a0cba-5c66-4fa5-831c-e3f8ddc5810d failed with error code 42. | $ echo $? | 42 Fixes https://github.com/systemd/systemd/issues/1290
2015-09-18 13:37:34 +02:00
if (streq(shutdown_verb, "exit")) {
command_line[pos++] = "--exit-code";
command_line[pos++] = exit_code;
core: properly use the ExitCode bus property when exiting is --user instance
2015-09-23 19:50:10 +02:00
xsprintf(exit_code, "%d", retval);
containers: systemd exits with non-zero code When a systemd service running in a container exits with a non-zero code, it can be useful to terminate the container immediately and get the exit code back to the host, when systemd-nspawn returns. This was not possible to do. This patch adds the following to make it possible: - Add a read-only "ExitCode" property on PID 1's "Manager" bus object. By default, it is 0 so the behaviour stays the same as previously. - Add a method "SetExitCode" on the same object. The method fails when called on baremetal: it is only allowed in containers or in user session. - Add support in systemctl to call "systemctl exit 42". It reuses the existing code for user session. - Add exit.target and systemd-exit.service to the system instance. - Change main() to actually call systemd-shutdown to exit() with the correct value. - Add verb 'exit' in systemd-shutdown with parameter --exit-code - Update systemctl manpage. I used the following to test it: | $ sudo rkt --debug --insecure-skip-verify run \ | --mds-register=false --local docker://busybox \ | --exec=/bin/chroot -- /proc/1/root \ | systemctl --force exit 42 | ... | Container rkt-895a0cba-5c66-4fa5-831c-e3f8ddc5810d failed with error code 42. | $ echo $? | 42 Fixes https://github.com/systemd/systemd/issues/1290
2015-09-18 13:37:34 +02:00
}
fix off by one error in array index assertion Since the index is already post-incremented when the array is appended to, this assertion can be wrongly reached when the array is at capacity (with the NULL terminator). The bug is reproducible on shutdown with the following settings in /etc/systemd/system.conf: LogTarget=journal-or-kmsg LogColor=yes LogLocation=yes Reported by Thermi on IRC.
2014-03-08 23:32:53 +01:00
assert(pos < ELEMENTSOF(command_line));
Pass log config from systemd to systemd-shutdown If PID 1 debug logging is enabled, it is nice to keep those settings when switching to systemd-shutdown binary, independently of whether this was done through /proc/cmdline options, or through runtime manipulations.
2014-02-16 00:13:46 +01:00
core: ShutdownWatchdogSec=infinity disables the watchdog logic too This is a follow-up for 947292eef
2016-03-29 13:04:04 +02:00
if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) {
strv: multiple cleanups - turn strv_merge into strv_extend_strv. appending strv b to the end of strv a instead of creating a new strv - strv_append: remove in favor of strv_extend and strv_push. - strv_remove: write slightly more elegant - strv_remove_prefix: remove unused function - strv_overlap: use strv_contains - strv_printf: STRV_FOREACH handles NULL correctly
2014-01-04 02:35:27 +01:00
char *e;
main: pass original environment block to shutdown binary so that it can detect container environments
2012-04-12 03:19:28 +02:00
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
/* If we reboot let's set the shutdown
* watchdog and tell the shutdown binary to
* repeatedly ping it */
core: disarm shutdown watchdog if we fail to set timeout Better safe than sorry, if drivers are stupid, and reset immediately on device closing if the timeout could not be initialized. https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=777735
2015-02-12 12:28:48 +01:00
r = watchdog_set_timeout(&arg_shutdown_watchdog);
watchdog_close(r < 0);
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
strv: multiple cleanups - turn strv_merge into strv_extend_strv. appending strv b to the end of strv a instead of creating a new strv - strv_append: remove in favor of strv_extend and strv_push. - strv_remove: write slightly more elegant - strv_remove_prefix: remove unused function - strv_overlap: use strv_contains - strv_printf: STRV_FOREACH handles NULL correctly
2014-01-04 02:35:27 +01:00
/* Tell the binary how often to ping, ignore failure */
if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
main: minor clean-ups Add (void) casting for a couple of functions where we knowingly ignore the returning error code. Use EXIT_FAILURE where appropriate. Try to initialize structures at declaration time, or at once.
2015-09-23 01:06:56 +02:00
(void) strv_push(&env_block, e);
strv: multiple cleanups - turn strv_merge into strv_extend_strv. appending strv b to the end of strv a instead of creating a new strv - strv_append: remove in favor of strv_extend and strv_push. - strv_remove: write slightly more elegant - strv_remove_prefix: remove unused function - strv_overlap: use strv_contains - strv_printf: STRV_FOREACH handles NULL correctly
2014-01-04 02:35:27 +01:00
} else
systemd: add hardware watchdog support This adds minimal hardware watchdog support to PID 1. The idea is that PID 1 supervises and watchdogs system services, while the hardware watchdog is used to supervise PID 1. This adds two hardware watchdog configuration options, for the runtime watchdog and for a shutdown watchdog. The former is active during normal operation, the latter only at reboots to ensure that if a clean reboot times out we reboot nonetheless. If the runtime watchdog is enabled PID 1 will automatically wake up at half the configured interval and write to the watchdog daemon. By default we enable the shutdown watchdog, but leave the runtime watchdog disabled in order not to break independent hardware watchdog daemons people might be using. This is only the most basic hookup. If necessary we can later on hook up the watchdog ping more closely with services deemed crucial.
2012-04-05 22:08:10 +02:00
watchdog_close(true);
core: uninstall cgroup agent only if we are running outside of a container Since the cgroupfs is currently not virtualized for containers we shouldn't reset the hosts agent from the container.
2013-07-10 23:50:28 +02:00
/* Avoid the creation of new processes forked by the
* kernel; at this point, we will not listen to the
* signals anyway */
basic: rework virtualization detection API Introduce a proper enum, and don't pass around string ids anymore. This simplifies things quite a bit, and makes virtualization detection more similar to architecture detection.
2015-09-07 13:42:47 +02:00
if (detect_container() <= 0)
cgroup: small cleanups and coding style fixes A number of simplications and adjustments to brings things closer to our coding style.
2015-09-01 18:02:43 +02:00
(void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
disable the cgroups release agent when shutting down During shutdown, when we try to clean up all remaining processes, the kernel will fork new agents every time a cgroup runs empty. These new processes cause delays in the final SIGTERM, SIGKILL logic. Apart from that, this should also avoid that the kernel-forked binaries cause unpredictably timed access to the filesystem which we might need to unmount.
2013-07-04 20:31:18 +02:00
main: pass original environment block to shutdown binary so that it can detect container environments
2012-04-12 03:19:28 +02:00
execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
treewide: use log_*_errno whenever %m is in the format string If the format string contains %m, clearly errno must have a meaningful value, so we might as well use log_*_errno to have ERRNO= logged. Using: find . -name '*.[ch]' | xargs sed -r -i -e \ 's/log_(debug|info|notice|warning|error|emergency)\((".*%m.*")/log_\1_errno(errno, \2/' Plus some whitespace, linewrap, and indent adjustments.
2014-11-28 19:29:59 +01:00
log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
getpid_cached() == 1 ? "freezing" : "quitting");
manager: hookup execution of systemd-shutdown helper (Modified by Lennart Poettering)
2010-10-14 00:52:26 +02:00
}
tree-wide: make use of getpid_cached() wherever we can This moves pretty much all uses of getpid() over to getpid_raw(). I didn't specifically check whether the optimization is worth it for each replacement, but in order to keep things simple and systematic I switched over everything at once.
2017-07-20 16:19:18 +02:00
if (getpid_cached() == 1) {
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
if (error_message)
manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
cgtop: underline table header Let's underline the header line of the table shown by cgtop, how it is customary for tables. In order to do this, let's introduce new ANSI underline macros, and clean up the existing ones as side effect.
2015-09-19 00:45:05 +02:00
ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
"%s, freezing.", error_message);
core: rework crash handling This introduces a new systemd.crash_reboot=1 kernel command line option that triggers a reboot after crashing. This also cleans up crash VT handling. Specifically, it cleans up the configuration setting, to be between 1..63 or a boolean. This is to replace the previous logic where "-1" meant disabled. We continue to accept that setting, but only document the boolean syntax instead. This also brings the documentation of the default settings in sync with what actually happens. The CrashChVT= configuration file setting is renamed to CrashChangeVT=, following our usual logic of not abbreviating unnecessarily. The old setting stays support for compat reasons. Fixes #1300
2015-09-23 23:13:06 +02:00
freeze_or_reboot();
manager: print fatal errors on the console too When booting in quiet mode, fatal messages would not be shown at all to the user. https://bugzilla.redhat.com/show_bug.cgi?id=1155468
2014-11-06 06:05:38 +01:00
}
main: freeze instead of exiting when run as init
2010-04-13 04:19:02 +02:00
initial commit
2009-11-18 00:42:52 +01:00
return retval;
}