diff options
Diffstat (limited to 'src/grp-system/systemctl')
-rw-r--r-- | src/grp-system/systemctl/.gitignore | 2 | ||||
-rw-r--r-- | src/grp-system/systemctl/halt.xml | 176 | ||||
-rw-r--r-- | src/grp-system/systemctl/runlevel.xml | 192 | ||||
-rw-r--r-- | src/grp-system/systemctl/shutdown.xml | 175 | ||||
-rw-r--r-- | src/grp-system/systemctl/systemctl.c | 7815 | ||||
-rw-r--r-- | src/grp-system/systemctl/systemctl.completion.bash.in | 284 | ||||
-rw-r--r-- | src/grp-system/systemctl/systemctl.completion.zsh.in | 391 | ||||
-rw-r--r-- | src/grp-system/systemctl/systemctl.xml | 1838 | ||||
-rwxr-xr-x | src/grp-system/systemctl/systemd-sysv-install.SKELETON | 47 | ||||
-rw-r--r-- | src/grp-system/systemctl/telinit.xml | 179 |
10 files changed, 11099 insertions, 0 deletions
diff --git a/src/grp-system/systemctl/.gitignore b/src/grp-system/systemctl/.gitignore new file mode 100644 index 0000000000..ebd59d3c9e --- /dev/null +++ b/src/grp-system/systemctl/.gitignore @@ -0,0 +1,2 @@ +/systemctl +/_systemctl diff --git a/src/grp-system/systemctl/halt.xml b/src/grp-system/systemctl/halt.xml new file mode 100644 index 0000000000..e3fa60a915 --- /dev/null +++ b/src/grp-system/systemctl/halt.xml @@ -0,0 +1,176 @@ +<?xml version='1.0'?> <!--*-nxml-*--> +<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" + "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd"> + +<!-- + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + 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 + (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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +--> + +<refentry id="halt" + xmlns:xi="http://www.w3.org/2001/XInclude"> + + <refentryinfo> + <title>halt</title> + <productname>systemd</productname> + + <authorgroup> + <author> + <contrib>Developer</contrib> + <firstname>Lennart</firstname> + <surname>Poettering</surname> + <email>lennart@poettering.net</email> + </author> + </authorgroup> + </refentryinfo> + + <refmeta> + <refentrytitle>halt</refentrytitle> + <manvolnum>8</manvolnum> + </refmeta> + + <refnamediv> + <refname>halt</refname> + <refname>poweroff</refname> + <refname>reboot</refname> + <refpurpose>Halt, power-off or reboot the machine</refpurpose> + </refnamediv> + + <refsynopsisdiv> + <cmdsynopsis> + <command>halt</command> + <arg choice="opt" rep="repeat">OPTIONS</arg> + </cmdsynopsis> + <cmdsynopsis> + <command>poweroff</command> + <arg choice="opt" rep="repeat">OPTIONS</arg> + </cmdsynopsis> + <cmdsynopsis> + <command>reboot</command> + <arg choice="opt" rep="repeat">OPTIONS</arg> + </cmdsynopsis> + </refsynopsisdiv> + + <refsect1> + <title>Description</title> + + <para><command>halt</command>, <command>poweroff</command>, + <command>reboot</command> may be used to halt, power-off or reboot + the machine.</para> + + </refsect1> + + <refsect1> + <title>Options</title> + + <para>The following options are understood:</para> + + <variablelist> + <varlistentry> + <term><option>--help</option></term> + + <xi:include href="standard-options.xml" xpointer="help-text" /> + </varlistentry> + + <varlistentry> + <term><option>--halt</option></term> + + <listitem><para>Halt the machine, regardless of which one of + the three commands is invoked.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-p</option></term> + <term><option>--poweroff</option></term> + + <listitem><para>Power-off the machine, regardless of which one + of the three commands is invoked.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>--reboot</option></term> + + <listitem><para>Reboot the machine, regardless of which one of + the three commands is invoked.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-f</option></term> + <term><option>--force</option></term> + + <listitem><para>Force immediate halt, power-off, reboot. Do + not contact the init system.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-w</option></term> + <term><option>--wtmp-only</option></term> + + <listitem><para>Only write wtmp shutdown entry, do not + actually halt, power-off, reboot.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-d</option></term> + <term><option>--no-wtmp</option></term> + + <listitem><para>Do not write wtmp shutdown + entry.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-n</option></term> + <term><option>--no-sync</option></term> + + <listitem><para>Don't sync hard disks/storage media before + halt, power-off, reboot.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>--no-wall</option></term> + + <listitem><para>Do not send wall message before halt, + power-off, reboot.</para></listitem> + </varlistentry> + </variablelist> + </refsect1> + + <refsect1> + <title>Exit status</title> + + <para>On success, 0 is returned, a non-zero failure code + otherwise.</para> + </refsect1> + + <refsect1> + <title>Notes</title> + + <para>These are legacy commands available for compatibility + only.</para> + </refsect1> + + <refsect1> + <title>See Also</title> + <para> + <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemctl</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>shutdown</refentrytitle><manvolnum>8</manvolnum></citerefentry>, + <citerefentry project='man-pages'><refentrytitle>wall</refentrytitle><manvolnum>1</manvolnum></citerefentry> + </para> + </refsect1> + +</refentry> diff --git a/src/grp-system/systemctl/runlevel.xml b/src/grp-system/systemctl/runlevel.xml new file mode 100644 index 0000000000..ca29c7c22c --- /dev/null +++ b/src/grp-system/systemctl/runlevel.xml @@ -0,0 +1,192 @@ +<?xml version='1.0'?> <!--*-nxml-*--> +<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" + "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd"> + +<!-- + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + 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 + (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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +--> + +<refentry id="runlevel" + xmlns:xi="http://www.w3.org/2001/XInclude" + conditional="HAVE_UTMP"> + + <refentryinfo> + <title>runlevel</title> + <productname>systemd</productname> + + <authorgroup> + <author> + <contrib>Developer</contrib> + <firstname>Lennart</firstname> + <surname>Poettering</surname> + <email>lennart@poettering.net</email> + </author> + </authorgroup> + </refentryinfo> + + <refmeta> + <refentrytitle>runlevel</refentrytitle> + <manvolnum>8</manvolnum> + </refmeta> + + <refnamediv> + <refname>runlevel</refname> + <refpurpose>Print previous and current SysV runlevel</refpurpose> + </refnamediv> + + <refsynopsisdiv> + <cmdsynopsis> + <command>runlevel</command> + <arg choice="opt" rep="repeat">options</arg> + </cmdsynopsis> + </refsynopsisdiv> + + <refsect1> + <title>Overview</title> + + <para>"Runlevels" are an obsolete way to start and stop groups of + services used in SysV init. systemd provides a compatibility layer + that maps runlevels to targets, and associated binaries like + <command>runlevel</command>. Nevertheless, only one runlevel can + be "active" at a given time, while systemd can activate multiple + targets concurrently, so the mapping to runlevels is confusing + and only approximate. Runlevels should not be used in new code, + and are mostly useful as a shorthand way to refer the matching + systemd targets in kernel boot parameters.</para> + + <table> + <title>Mapping between runlevels and systemd targets</title> + <tgroup cols='2' align='left' colsep='1' rowsep='1'> + <colspec colname="runlevel" /> + <colspec colname="target" /> + <thead> + <row> + <entry>Runlevel</entry> + <entry>Target</entry> + </row> + </thead> + <tbody> + <row> + <entry>0</entry> + <entry><filename>poweroff.target</filename></entry> + </row> + <row> + <entry>1</entry> + <entry><filename>rescue.target</filename></entry> + </row> + <row> + <entry>2, 3, 4</entry> + <entry><filename>multi-user.target</filename></entry> + </row> + <row> + <entry>5</entry> + <entry><filename>graphical.target</filename></entry> + </row> + <row> + <entry>6</entry> + <entry><filename>reboot.target</filename></entry> + </row> + </tbody> + </tgroup> + </table> + </refsect1> + + <refsect1> + <title>Description</title> + + <para><command>runlevel</command> prints the previous and current + SysV runlevel if they are known.</para> + + <para>The two runlevel characters are separated by a single space + character. If a runlevel cannot be determined, N is printed + instead. If neither can be determined, the word "unknown" is + printed.</para> + + <para>Unless overridden in the environment, this will check the + utmp database for recent runlevel changes.</para> + </refsect1> + + <refsect1> + <title>Options</title> + + <para>The following option is understood:</para> + + <variablelist> + <varlistentry> + <term><option>--help</option></term> + + <xi:include href="standard-options.xml" xpointer="help-text" /> + </varlistentry> + </variablelist> + + </refsect1> + + <refsect1> + <title>Exit status</title> + + <para>If one or both runlevels could be determined, 0 is returned, + a non-zero failure code otherwise.</para> + + </refsect1> + + <refsect1> + <title>Environment</title> + + <variablelist class='environment-variables'> + <varlistentry> + <term><varname>$RUNLEVEL</varname></term> + + <listitem><para>If <varname>$RUNLEVEL</varname> is set, + <command>runlevel</command> will print this value as current + runlevel and ignore utmp.</para></listitem> + </varlistentry> + + <varlistentry> + <term><varname>$PREVLEVEL</varname></term> + + <listitem><para>If <varname>$PREVLEVEL</varname> is set, + <command>runlevel</command> will print this value as previous + runlevel and ignore utmp.</para></listitem> + </varlistentry> + </variablelist> + </refsect1> + + <refsect1> + <title>Files</title> + + <variablelist> + <varlistentry> + <term><filename>/var/run/utmp</filename></term> + + <listitem><para>The utmp database <command>runlevel</command> + reads the previous and current runlevel + from.</para></listitem> + </varlistentry> + </variablelist> + </refsect1> + + <refsect1> + <title>See Also</title> + <para> + <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.target</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemctl</refentrytitle><manvolnum>1</manvolnum></citerefentry> + </para> + </refsect1> + +</refentry> diff --git a/src/grp-system/systemctl/shutdown.xml b/src/grp-system/systemctl/shutdown.xml new file mode 100644 index 0000000000..a8af387c67 --- /dev/null +++ b/src/grp-system/systemctl/shutdown.xml @@ -0,0 +1,175 @@ +<?xml version='1.0'?> <!--*-nxml-*--> +<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" + "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd"> + +<!-- + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + 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 + (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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +--> + +<refentry id="shutdown" + xmlns:xi="http://www.w3.org/2001/XInclude"> + + <refentryinfo> + <title>shutdown</title> + <productname>systemd</productname> + + <authorgroup> + <author> + <contrib>Developer</contrib> + <firstname>Lennart</firstname> + <surname>Poettering</surname> + <email>lennart@poettering.net</email> + </author> + </authorgroup> + </refentryinfo> + + <refmeta> + <refentrytitle>shutdown</refentrytitle> + <manvolnum>8</manvolnum> + </refmeta> + + <refnamediv> + <refname>shutdown</refname> + <refpurpose>Halt, power-off or reboot the machine</refpurpose> + </refnamediv> + + <refsynopsisdiv> + <cmdsynopsis> + <command>shutdown</command> + <arg choice="opt" rep="repeat">OPTIONS</arg> + <arg choice="opt">TIME</arg> + <arg choice="opt" rep="repeat">WALL</arg> + </cmdsynopsis> + </refsynopsisdiv> + + <refsect1> + <title>Description</title> + + <para><command>shutdown</command> may be used to halt, power-off + or reboot the machine.</para> + + <para>The first argument may be a time string (which is usually + <literal>now</literal>). Optionally, this may be followed by a + wall message to be sent to all logged-in users before going + down.</para> + + <para>The time string may either be in the format + <literal>hh:mm</literal> for hour/minutes specifying the time to + execute the shutdown at, specified in 24h clock format. + Alternatively it may be in the syntax <literal>+m</literal> + referring to the specified number of minutes m from now. + <literal>now</literal> is an alias for <literal>+0</literal>, i.e. + for triggering an immediate shutdown. If no time argument is + specified, <literal>+1</literal> is implied.</para> + + <para>Note that to specify a wall message you must specify a time + argument, too.</para> + + <para>If the time argument is used, 5 minutes before the system + goes down the <filename>/run/nologin</filename> file is created to + ensure that further logins shall not be allowed.</para> + </refsect1> + + <refsect1> + <title>Options</title> + + <para>The following options are understood:</para> + + <variablelist> + <varlistentry> + <term><option>--help</option></term> + + <xi:include href="standard-options.xml" xpointer="help-text" /> + </varlistentry> + + <varlistentry> + <term><option>-H</option></term> + <term><option>--halt</option></term> + + <listitem><para>Halt the machine.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-P</option></term> + <term><option>--poweroff</option></term> + + <listitem><para>Power-off the machine (the + default).</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-r</option></term> + <term><option>--reboot</option></term> + + <listitem><para>Reboot the + machine.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-h</option></term> + + <listitem><para>Equivalent to <option>--poweroff</option>, + unless <option>--halt</option> is specified.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-k</option></term> + + <listitem><para>Do not halt, power-off, reboot, just write + wall message.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>--no-wall</option></term> + + <listitem><para>Do not send wall + message before + halt, power-off, reboot.</para></listitem> + </varlistentry> + + <varlistentry> + <term><option>-c</option></term> + + <listitem><para>Cancel a pending shutdown. This may be used + cancel the effect of an invocation of + <command>shutdown</command> with a time argument that is not + <literal>+0</literal> or + <literal>now</literal>.</para></listitem> + </varlistentry> + + </variablelist> + </refsect1> + + <refsect1> + <title>Exit status</title> + + <para>On success, 0 is returned, a non-zero failure code + otherwise.</para> + </refsect1> + + <refsect1> + <title>See Also</title> + <para> + <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemctl</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>halt</refentrytitle><manvolnum>8</manvolnum></citerefentry>, + <citerefentry project='man-pages'><refentrytitle>wall</refentrytitle><manvolnum>1</manvolnum></citerefentry> + </para> + </refsect1> + +</refentry> diff --git a/src/grp-system/systemctl/systemctl.c b/src/grp-system/systemctl/systemctl.c new file mode 100644 index 0000000000..c7c00bbbc3 --- /dev/null +++ b/src/grp-system/systemctl/systemctl.c @@ -0,0 +1,7815 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2013 Marc-Antoine Perennou + + systemd is free software; you can redistribute it and/or modify it + 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 + (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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <getopt.h> +#include <linux/reboot.h> +#include <locale.h> +#include <stdbool.h> +#include <stddef.h> +#include <stdio.h> +#include <string.h> +#include <sys/reboot.h> +#include <sys/socket.h> +#include <unistd.h> + +#include <systemd/sd-bus.h> +#include <systemd/sd-daemon.h> +#include <systemd/sd-login.h> + +#include "alloc-util.h" +#include "bus-common-errors.h" +#include "bus-error.h" +#include "bus-message.h" +#include "bus-unit-util.h" +#include "bus-util.h" +#include "cgroup-show.h" +#include "cgroup-util.h" +#include "copy.h" +#include "dropin.h" +#include "efivars.h" +#include "env-util.h" +#include "exit-status.h" +#include "fd-util.h" +#include "fileio.h" +#include "formats-util.h" +#include "fs-util.h" +#include "glob-util.h" +#include "hostname-util.h" +#include "initreq.h" +#include "install.h" +#include "io-util.h" +#include "list.h" +#include "locale-util.h" +#include "log.h" +#include "logs-show.h" +#include "macro.h" +#include "mkdir.h" +#include "pager.h" +#include "parse-util.h" +#include "path-lookup.h" +#include "path-util.h" +#include "process-util.h" +#include "rlimit-util.h" +#include "set.h" +#include "signal-util.h" +#include "socket-util.h" +#include "spawn-ask-password-agent.h" +#include "spawn-polkit-agent.h" +#include "special.h" +#include "stat-util.h" +#include "strv.h" +#include "terminal-util.h" +#include "unit-name.h" +#include "user-util.h" +#include "util.h" +#include "utmp-wtmp.h" +#include "verbs.h" +#include "virt.h" + +static char **arg_types = NULL; +static char **arg_states = NULL; +static char **arg_properties = NULL; +static bool arg_all = false; +static enum dependency { + DEPENDENCY_FORWARD, + DEPENDENCY_REVERSE, + DEPENDENCY_AFTER, + DEPENDENCY_BEFORE, + _DEPENDENCY_MAX +} arg_dependency = DEPENDENCY_FORWARD; +static const char *arg_job_mode = "replace"; +static UnitFileScope arg_scope = UNIT_FILE_SYSTEM; +static bool arg_no_block = false; +static bool arg_no_legend = false; +static bool arg_no_pager = false; +static bool arg_no_wtmp = false; +static bool arg_no_sync = false; +static bool arg_no_wall = false; +static bool arg_no_reload = false; +static bool arg_value = false; +static bool arg_show_types = false; +static bool arg_ignore_inhibitors = false; +static bool arg_dry = false; +static bool arg_quiet = false; +static bool arg_full = false; +static bool arg_recursive = false; +static int arg_force = 0; +static bool arg_ask_password = false; +static bool arg_runtime = false; +static UnitFilePresetMode arg_preset_mode = UNIT_FILE_PRESET_FULL; +static char **arg_wall = NULL; +static const char *arg_kill_who = NULL; +static int arg_signal = SIGTERM; +static char *arg_root = NULL; +static usec_t arg_when = 0; +static enum action { + _ACTION_INVALID, + ACTION_SYSTEMCTL, + ACTION_HALT, + ACTION_POWEROFF, + ACTION_REBOOT, + ACTION_KEXEC, + ACTION_EXIT, + ACTION_SUSPEND, + ACTION_HIBERNATE, + ACTION_HYBRID_SLEEP, + ACTION_RUNLEVEL2, + ACTION_RUNLEVEL3, + ACTION_RUNLEVEL4, + ACTION_RUNLEVEL5, + ACTION_RESCUE, + ACTION_EMERGENCY, + ACTION_DEFAULT, + ACTION_RELOAD, + ACTION_REEXEC, + ACTION_RUNLEVEL, + ACTION_CANCEL_SHUTDOWN, + _ACTION_MAX +} arg_action = ACTION_SYSTEMCTL; +static BusTransport arg_transport = BUS_TRANSPORT_LOCAL; +static const char *arg_host = NULL; +static unsigned arg_lines = 10; +static OutputMode arg_output = OUTPUT_SHORT; +static bool arg_plain = false; +static bool arg_firmware_setup = false; +static bool arg_now = false; + +static int daemon_reload(int argc, char *argv[], void* userdata); +static int halt_now(enum action a); +static int get_state_one_unit(sd_bus *bus, const char *name, UnitActiveState *active_state); + +static bool original_stdout_is_tty; + +typedef enum BusFocus { + BUS_FULL, /* The full bus indicated via --system or --user */ + BUS_MANAGER, /* The manager itself, possibly directly, possibly via the bus */ + _BUS_FOCUS_MAX +} BusFocus; + +static sd_bus *busses[_BUS_FOCUS_MAX] = {}; + +static int acquire_bus(BusFocus focus, sd_bus **ret) { + int r; + + assert(focus < _BUS_FOCUS_MAX); + assert(ret); + + /* We only go directly to the manager, if we are using a local transport */ + if (arg_transport != BUS_TRANSPORT_LOCAL) + focus = BUS_FULL; + + if (!busses[focus]) { + bool user; + + user = arg_scope != UNIT_FILE_SYSTEM; + + if (focus == BUS_MANAGER) + r = bus_connect_transport_systemd(arg_transport, arg_host, user, &busses[focus]); + else + r = bus_connect_transport(arg_transport, arg_host, user, &busses[focus]); + if (r < 0) + return log_error_errno(r, "Failed to connect to bus: %m"); + + (void) sd_bus_set_allow_interactive_authorization(busses[focus], arg_ask_password); + } + + *ret = busses[focus]; + return 0; +} + +static void release_busses(void) { + BusFocus w; + + for (w = 0; w < _BUS_FOCUS_MAX; w++) + busses[w] = sd_bus_flush_close_unref(busses[w]); +} + +static void ask_password_agent_open_if_enabled(void) { + + /* Open the password agent as a child process if necessary */ + + if (!arg_ask_password) + return; + + if (arg_scope != UNIT_FILE_SYSTEM) + return; + + if (arg_transport != BUS_TRANSPORT_LOCAL) + return; + + ask_password_agent_open(); +} + +static void polkit_agent_open_if_enabled(void) { + + /* Open the polkit agent as a child process if necessary */ + + if (!arg_ask_password) + return; + + if (arg_scope != UNIT_FILE_SYSTEM) + return; + + if (arg_transport != BUS_TRANSPORT_LOCAL) + return; + + polkit_agent_open(); +} + +static OutputFlags get_output_flags(void) { + return + arg_all * OUTPUT_SHOW_ALL | + arg_full * OUTPUT_FULL_WIDTH | + (!on_tty() || pager_have()) * OUTPUT_FULL_WIDTH | + colors_enabled() * OUTPUT_COLOR | + !arg_quiet * OUTPUT_WARN_CUTOFF; +} + +static int translate_bus_error_to_exit_status(int r, const sd_bus_error *error) { + assert(error); + + if (!sd_bus_error_is_set(error)) + return r; + + if (sd_bus_error_has_name(error, SD_BUS_ERROR_ACCESS_DENIED) || + sd_bus_error_has_name(error, BUS_ERROR_ONLY_BY_DEPENDENCY) || + sd_bus_error_has_name(error, BUS_ERROR_NO_ISOLATION) || + sd_bus_error_has_name(error, BUS_ERROR_TRANSACTION_IS_DESTRUCTIVE)) + return EXIT_NOPERMISSION; + + if (sd_bus_error_has_name(error, BUS_ERROR_NO_SUCH_UNIT)) + return EXIT_NOTINSTALLED; + + if (sd_bus_error_has_name(error, BUS_ERROR_JOB_TYPE_NOT_APPLICABLE) || + sd_bus_error_has_name(error, SD_BUS_ERROR_NOT_SUPPORTED)) + return EXIT_NOTIMPLEMENTED; + + if (sd_bus_error_has_name(error, BUS_ERROR_LOAD_FAILED)) + return EXIT_NOTCONFIGURED; + + if (r != 0) + return r; + + return EXIT_FAILURE; +} + +static bool install_client_side(void) { + + /* Decides when to execute enable/disable/... operations + * client-side rather than server-side. */ + + if (running_in_chroot() > 0) + return true; + + if (sd_booted() <= 0) + return true; + + if (!isempty(arg_root)) + return true; + + if (arg_scope == UNIT_FILE_GLOBAL) + return true; + + /* Unsupported environment variable, mostly for debugging purposes */ + if (getenv_bool("SYSTEMCTL_INSTALL_CLIENT_SIDE") > 0) + return true; + + return false; +} + +static int compare_unit_info(const void *a, const void *b) { + const UnitInfo *u = a, *v = b; + const char *d1, *d2; + int r; + + /* First, order by machine */ + if (!u->machine && v->machine) + return -1; + if (u->machine && !v->machine) + return 1; + if (u->machine && v->machine) { + r = strcasecmp(u->machine, v->machine); + if (r != 0) + return r; + } + + /* Second, order by unit type */ + d1 = strrchr(u->id, '.'); + d2 = strrchr(v->id, '.'); + if (d1 && d2) { + r = strcasecmp(d1, d2); + if (r != 0) + return r; + } + + /* Third, order by name */ + return strcasecmp(u->id, v->id); +} + +static bool output_show_unit(const UnitInfo *u, char **patterns) { + assert(u); + + if (!strv_fnmatch_or_empty(patterns, u->id, FNM_NOESCAPE)) + return false; + + if (arg_types) { + const char *dot; + + dot = strrchr(u->id, '.'); + if (!dot) + return false; + + if (!strv_find(arg_types, dot+1)) + return false; + } + + if (arg_all) + return true; + + /* Note that '--all' is not purely a state filter, but also a + * filter that hides units that "follow" other units (which is + * used for device units that appear under different names). */ + if (!isempty(u->following)) + return false; + + if (!strv_isempty(arg_states)) + return true; + + /* By default show all units except the ones in inactive + * state and with no pending job */ + if (u->job_id > 0) + return true; + + if (streq(u->active_state, "inactive")) + return false; + + return true; +} + +static int output_units_list(const UnitInfo *unit_infos, unsigned c) { + unsigned circle_len = 0, id_len, max_id_len, load_len, active_len, sub_len, job_len, desc_len; + const UnitInfo *u; + unsigned n_shown = 0; + int job_count = 0; + + max_id_len = strlen("UNIT"); + load_len = strlen("LOAD"); + active_len = strlen("ACTIVE"); + sub_len = strlen("SUB"); + job_len = strlen("JOB"); + desc_len = 0; + + for (u = unit_infos; u < unit_infos + c; u++) { + max_id_len = MAX(max_id_len, strlen(u->id) + (u->machine ? strlen(u->machine)+1 : 0)); + load_len = MAX(load_len, strlen(u->load_state)); + active_len = MAX(active_len, strlen(u->active_state)); + sub_len = MAX(sub_len, strlen(u->sub_state)); + + if (u->job_id != 0) { + job_len = MAX(job_len, strlen(u->job_type)); + job_count++; + } + + if (!arg_no_legend && + (streq(u->active_state, "failed") || + STR_IN_SET(u->load_state, "error", "not-found", "masked"))) + circle_len = 2; + } + + if (!arg_full && original_stdout_is_tty) { + unsigned basic_len; + + id_len = MIN(max_id_len, 25u); + basic_len = circle_len + 5 + id_len + 5 + active_len + sub_len; + + if (job_count) + basic_len += job_len + 1; + + if (basic_len < (unsigned) columns()) { + unsigned extra_len, incr; + extra_len = columns() - basic_len; + + /* Either UNIT already got 25, or is fully satisfied. + * Grant up to 25 to DESC now. */ + incr = MIN(extra_len, 25u); + desc_len += incr; + extra_len -= incr; + + /* split the remaining space between UNIT and DESC, + * but do not give UNIT more than it needs. */ + if (extra_len > 0) { + incr = MIN(extra_len / 2, max_id_len - id_len); + id_len += incr; + desc_len += extra_len - incr; + } + } + } else + id_len = max_id_len; + + for (u = unit_infos; u < unit_infos + c; u++) { + _cleanup_free_ char *e = NULL, *j = NULL; + const char *on_loaded = "", *off_loaded = ""; + const char *on_active = "", *off_active = ""; + const char *on_circle = "", *off_circle = ""; + const char *id; + bool circle = false; + + if (!n_shown && !arg_no_legend) { + + if (circle_len > 0) + fputs(" ", stdout); + + printf("%-*s %-*s %-*s %-*s ", + id_len, "UNIT", + load_len, "LOAD", + active_len, "ACTIVE", + sub_len, "SUB"); + + if (job_count) + printf("%-*s ", job_len, "JOB"); + + if (!arg_full && arg_no_pager) + printf("%.*s\n", desc_len, "DESCRIPTION"); + else + printf("%s\n", "DESCRIPTION"); + } + + n_shown++; + + if (STR_IN_SET(u->load_state, "error", "not-found", "masked") && !arg_plain) { + on_loaded = ansi_highlight_red(); + on_circle = ansi_highlight_yellow(); + off_loaded = off_circle = ansi_normal(); + circle = true; + } else if (streq(u->active_state, "failed") && !arg_plain) { + on_circle = on_active = ansi_highlight_red(); + off_circle = off_active = ansi_normal(); + circle = true; + } + + if (u->machine) { + j = strjoin(u->machine, ":", u->id, NULL); + if (!j) + return log_oom(); + + id = j; + } else + id = u->id; + + if (arg_full) { + e = ellipsize(id, id_len, 33); + if (!e) + return log_oom(); + + id = e; + } + + if (circle_len > 0) + printf("%s%s%s ", on_circle, circle ? special_glyph(BLACK_CIRCLE) : " ", off_circle); + + printf("%s%-*s%s %s%-*s%s %s%-*s %-*s%s %-*s", + on_active, id_len, id, off_active, + on_loaded, load_len, u->load_state, off_loaded, + on_active, active_len, u->active_state, + sub_len, u->sub_state, off_active, + job_count ? job_len + 1 : 0, u->job_id ? u->job_type : ""); + + if (desc_len > 0) + printf("%.*s\n", desc_len, u->description); + else + printf("%s\n", u->description); + } + + if (!arg_no_legend) { + const char *on, *off; + + if (n_shown) { + puts("\n" + "LOAD = Reflects whether the unit definition was properly loaded.\n" + "ACTIVE = The high-level unit activation state, i.e. generalization of SUB.\n" + "SUB = The low-level unit activation state, values depend on unit type."); + puts(job_count ? "JOB = Pending job for the unit.\n" : ""); + on = ansi_highlight(); + off = ansi_normal(); + } else { + on = ansi_highlight_red(); + off = ansi_normal(); + } + + if (arg_all) + printf("%s%u loaded units listed.%s\n" + "To show all installed unit files use 'systemctl list-unit-files'.\n", + on, n_shown, off); + else + printf("%s%u loaded units listed.%s Pass --all to see loaded but inactive units, too.\n" + "To show all installed unit files use 'systemctl list-unit-files'.\n", + on, n_shown, off); + } + + return 0; +} + +static int get_unit_list( + sd_bus *bus, + const char *machine, + char **patterns, + UnitInfo **unit_infos, + int c, + sd_bus_message **_reply) { + + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + size_t size = c; + int r; + UnitInfo u; + bool fallback = false; + + assert(bus); + assert(unit_infos); + assert(_reply); + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "ListUnitsByPatterns"); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, arg_states); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, patterns); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, &reply); + if (r < 0 && sd_bus_error_has_name(&error, SD_BUS_ERROR_UNKNOWN_METHOD)) { + /* Fallback to legacy ListUnitsFiltered method */ + fallback = true; + log_debug_errno(r, "Failed to list units: %s Falling back to ListUnitsFiltered method.", bus_error_message(&error, r)); + m = sd_bus_message_unref(m); + sd_bus_error_free(&error); + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "ListUnitsFiltered"); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, arg_states); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, &reply); + } + if (r < 0) + return log_error_errno(r, "Failed to list units: %s", bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "(ssssssouso)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = bus_parse_unit_info(reply, &u)) > 0) { + u.machine = machine; + + if (!output_show_unit(&u, fallback ? patterns : NULL)) + continue; + + if (!GREEDY_REALLOC(*unit_infos, size, c+1)) + return log_oom(); + + (*unit_infos)[c++] = u; + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + *_reply = reply; + reply = NULL; + + return c; +} + +static void message_set_freep(Set **set) { + sd_bus_message *m; + + while ((m = set_steal_first(*set))) + sd_bus_message_unref(m); + + set_free(*set); +} + +static int get_unit_list_recursive( + sd_bus *bus, + char **patterns, + UnitInfo **_unit_infos, + Set **_replies, + char ***_machines) { + + _cleanup_free_ UnitInfo *unit_infos = NULL; + _cleanup_(message_set_freep) Set *replies; + sd_bus_message *reply; + int c, r; + + assert(bus); + assert(_replies); + assert(_unit_infos); + assert(_machines); + + replies = set_new(NULL); + if (!replies) + return log_oom(); + + c = get_unit_list(bus, NULL, patterns, &unit_infos, 0, &reply); + if (c < 0) + return c; + + r = set_put(replies, reply); + if (r < 0) { + sd_bus_message_unref(reply); + return log_oom(); + } + + if (arg_recursive) { + _cleanup_strv_free_ char **machines = NULL; + char **i; + + r = sd_get_machine_names(&machines); + if (r < 0) + return log_error_errno(r, "Failed to get machine names: %m"); + + STRV_FOREACH(i, machines) { + _cleanup_(sd_bus_flush_close_unrefp) sd_bus *container = NULL; + int k; + + r = sd_bus_open_system_machine(&container, *i); + if (r < 0) { + log_warning_errno(r, "Failed to connect to container %s, ignoring: %m", *i); + continue; + } + + k = get_unit_list(container, *i, patterns, &unit_infos, c, &reply); + if (k < 0) + return k; + + c = k; + + r = set_put(replies, reply); + if (r < 0) { + sd_bus_message_unref(reply); + return log_oom(); + } + } + + *_machines = machines; + machines = NULL; + } else + *_machines = NULL; + + *_unit_infos = unit_infos; + unit_infos = NULL; + + *_replies = replies; + replies = NULL; + + return c; +} + +static int list_units(int argc, char *argv[], void *userdata) { + _cleanup_free_ UnitInfo *unit_infos = NULL; + _cleanup_(message_set_freep) Set *replies = NULL; + _cleanup_strv_free_ char **machines = NULL; + sd_bus *bus; + int r; + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = get_unit_list_recursive(bus, strv_skip(argv, 1), &unit_infos, &replies, &machines); + if (r < 0) + return r; + + qsort_safe(unit_infos, r, sizeof(UnitInfo), compare_unit_info); + return output_units_list(unit_infos, r); +} + +static int get_triggered_units( + sd_bus *bus, + const char* path, + char*** ret) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int r; + + assert(bus); + assert(path); + assert(ret); + + r = sd_bus_get_property_strv( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Unit", + "Triggers", + &error, + ret); + if (r < 0) + return log_error_errno(r, "Failed to determine triggers: %s", bus_error_message(&error, r)); + + return 0; +} + +static int get_listening( + sd_bus *bus, + const char* unit_path, + char*** listening) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + const char *type, *path; + int r, n = 0; + + r = sd_bus_get_property( + bus, + "org.freedesktop.systemd1", + unit_path, + "org.freedesktop.systemd1.Socket", + "Listen", + &error, + &reply, + "a(ss)"); + if (r < 0) + return log_error_errno(r, "Failed to get list of listening sockets: %s", bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "(ss)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(reply, "(ss)", &type, &path)) > 0) { + + r = strv_extend(listening, type); + if (r < 0) + return log_oom(); + + r = strv_extend(listening, path); + if (r < 0) + return log_oom(); + + n++; + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + return n; +} + +struct socket_info { + const char *machine; + const char* id; + + char* type; + char* path; + + /* Note: triggered is a list here, although it almost certainly + * will always be one unit. Nevertheless, dbus API allows for multiple + * values, so let's follow that. */ + char** triggered; + + /* The strv above is shared. free is set only in the first one. */ + bool own_triggered; +}; + +static int socket_info_compare(const struct socket_info *a, const struct socket_info *b) { + int o; + + assert(a); + assert(b); + + if (!a->machine && b->machine) + return -1; + if (a->machine && !b->machine) + return 1; + if (a->machine && b->machine) { + o = strcasecmp(a->machine, b->machine); + if (o != 0) + return o; + } + + o = strcmp(a->path, b->path); + if (o == 0) + o = strcmp(a->type, b->type); + + return o; +} + +static int output_sockets_list(struct socket_info *socket_infos, unsigned cs) { + struct socket_info *s; + unsigned pathlen = strlen("LISTEN"), + typelen = strlen("TYPE") * arg_show_types, + socklen = strlen("UNIT"), + servlen = strlen("ACTIVATES"); + const char *on, *off; + + for (s = socket_infos; s < socket_infos + cs; s++) { + unsigned tmp = 0; + char **a; + + socklen = MAX(socklen, strlen(s->id)); + if (arg_show_types) + typelen = MAX(typelen, strlen(s->type)); + pathlen = MAX(pathlen, strlen(s->path) + (s->machine ? strlen(s->machine)+1 : 0)); + + STRV_FOREACH(a, s->triggered) + tmp += strlen(*a) + 2*(a != s->triggered); + servlen = MAX(servlen, tmp); + } + + if (cs) { + if (!arg_no_legend) + printf("%-*s %-*.*s%-*s %s\n", + pathlen, "LISTEN", + typelen + arg_show_types, typelen + arg_show_types, "TYPE ", + socklen, "UNIT", + "ACTIVATES"); + + for (s = socket_infos; s < socket_infos + cs; s++) { + _cleanup_free_ char *j = NULL; + const char *path; + char **a; + + if (s->machine) { + j = strjoin(s->machine, ":", s->path, NULL); + if (!j) + return log_oom(); + path = j; + } else + path = s->path; + + if (arg_show_types) + printf("%-*s %-*s %-*s", + pathlen, path, typelen, s->type, socklen, s->id); + else + printf("%-*s %-*s", + pathlen, path, socklen, s->id); + STRV_FOREACH(a, s->triggered) + printf("%s %s", + a == s->triggered ? "" : ",", *a); + printf("\n"); + } + + on = ansi_highlight(); + off = ansi_normal(); + if (!arg_no_legend) + printf("\n"); + } else { + on = ansi_highlight_red(); + off = ansi_normal(); + } + + if (!arg_no_legend) { + printf("%s%u sockets listed.%s\n", on, cs, off); + if (!arg_all) + printf("Pass --all to see loaded but inactive sockets, too.\n"); + } + + return 0; +} + +static int list_sockets(int argc, char *argv[], void *userdata) { + _cleanup_(message_set_freep) Set *replies = NULL; + _cleanup_strv_free_ char **machines = NULL; + _cleanup_free_ UnitInfo *unit_infos = NULL; + _cleanup_free_ struct socket_info *socket_infos = NULL; + const UnitInfo *u; + struct socket_info *s; + unsigned cs = 0; + size_t size = 0; + int r = 0, n; + sd_bus *bus; + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + n = get_unit_list_recursive(bus, strv_skip(argv, 1), &unit_infos, &replies, &machines); + if (n < 0) + return n; + + for (u = unit_infos; u < unit_infos + n; u++) { + _cleanup_strv_free_ char **listening = NULL, **triggered = NULL; + int i, c; + + if (!endswith(u->id, ".socket")) + continue; + + r = get_triggered_units(bus, u->unit_path, &triggered); + if (r < 0) + goto cleanup; + + c = get_listening(bus, u->unit_path, &listening); + if (c < 0) { + r = c; + goto cleanup; + } + + if (!GREEDY_REALLOC(socket_infos, size, cs + c)) { + r = log_oom(); + goto cleanup; + } + + for (i = 0; i < c; i++) + socket_infos[cs + i] = (struct socket_info) { + .machine = u->machine, + .id = u->id, + .type = listening[i*2], + .path = listening[i*2 + 1], + .triggered = triggered, + .own_triggered = i==0, + }; + + /* from this point on we will cleanup those socket_infos */ + cs += c; + free(listening); + listening = triggered = NULL; /* avoid cleanup */ + } + + qsort_safe(socket_infos, cs, sizeof(struct socket_info), + (__compar_fn_t) socket_info_compare); + + output_sockets_list(socket_infos, cs); + + cleanup: + assert(cs == 0 || socket_infos); + for (s = socket_infos; s < socket_infos + cs; s++) { + free(s->type); + free(s->path); + if (s->own_triggered) + strv_free(s->triggered); + } + + return r; +} + +static int get_next_elapse( + sd_bus *bus, + const char *path, + dual_timestamp *next) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + dual_timestamp t; + int r; + + assert(bus); + assert(path); + assert(next); + + r = sd_bus_get_property_trivial( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Timer", + "NextElapseUSecMonotonic", + &error, + 't', + &t.monotonic); + if (r < 0) + return log_error_errno(r, "Failed to get next elapsation time: %s", bus_error_message(&error, r)); + + r = sd_bus_get_property_trivial( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Timer", + "NextElapseUSecRealtime", + &error, + 't', + &t.realtime); + if (r < 0) + return log_error_errno(r, "Failed to get next elapsation time: %s", bus_error_message(&error, r)); + + *next = t; + return 0; +} + +static int get_last_trigger( + sd_bus *bus, + const char *path, + usec_t *last) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int r; + + assert(bus); + assert(path); + assert(last); + + r = sd_bus_get_property_trivial( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Timer", + "LastTriggerUSec", + &error, + 't', + last); + if (r < 0) + return log_error_errno(r, "Failed to get last trigger time: %s", bus_error_message(&error, r)); + + return 0; +} + +struct timer_info { + const char* machine; + const char* id; + usec_t next_elapse; + usec_t last_trigger; + char** triggered; +}; + +static int timer_info_compare(const struct timer_info *a, const struct timer_info *b) { + int o; + + assert(a); + assert(b); + + if (!a->machine && b->machine) + return -1; + if (a->machine && !b->machine) + return 1; + if (a->machine && b->machine) { + o = strcasecmp(a->machine, b->machine); + if (o != 0) + return o; + } + + if (a->next_elapse < b->next_elapse) + return -1; + if (a->next_elapse > b->next_elapse) + return 1; + + return strcmp(a->id, b->id); +} + +static int output_timers_list(struct timer_info *timer_infos, unsigned n) { + struct timer_info *t; + unsigned + nextlen = strlen("NEXT"), + leftlen = strlen("LEFT"), + lastlen = strlen("LAST"), + passedlen = strlen("PASSED"), + unitlen = strlen("UNIT"), + activatelen = strlen("ACTIVATES"); + + const char *on, *off; + + assert(timer_infos || n == 0); + + for (t = timer_infos; t < timer_infos + n; t++) { + unsigned ul = 0; + char **a; + + if (t->next_elapse > 0) { + char tstamp[FORMAT_TIMESTAMP_MAX] = "", trel[FORMAT_TIMESTAMP_RELATIVE_MAX] = ""; + + format_timestamp(tstamp, sizeof(tstamp), t->next_elapse); + nextlen = MAX(nextlen, strlen(tstamp) + 1); + + format_timestamp_relative(trel, sizeof(trel), t->next_elapse); + leftlen = MAX(leftlen, strlen(trel)); + } + + if (t->last_trigger > 0) { + char tstamp[FORMAT_TIMESTAMP_MAX] = "", trel[FORMAT_TIMESTAMP_RELATIVE_MAX] = ""; + + format_timestamp(tstamp, sizeof(tstamp), t->last_trigger); + lastlen = MAX(lastlen, strlen(tstamp) + 1); + + format_timestamp_relative(trel, sizeof(trel), t->last_trigger); + passedlen = MAX(passedlen, strlen(trel)); + } + + unitlen = MAX(unitlen, strlen(t->id) + (t->machine ? strlen(t->machine)+1 : 0)); + + STRV_FOREACH(a, t->triggered) + ul += strlen(*a) + 2*(a != t->triggered); + + activatelen = MAX(activatelen, ul); + } + + if (n > 0) { + if (!arg_no_legend) + printf("%-*s %-*s %-*s %-*s %-*s %s\n", + nextlen, "NEXT", + leftlen, "LEFT", + lastlen, "LAST", + passedlen, "PASSED", + unitlen, "UNIT", + "ACTIVATES"); + + for (t = timer_infos; t < timer_infos + n; t++) { + _cleanup_free_ char *j = NULL; + const char *unit; + char tstamp1[FORMAT_TIMESTAMP_MAX] = "n/a", trel1[FORMAT_TIMESTAMP_RELATIVE_MAX] = "n/a"; + char tstamp2[FORMAT_TIMESTAMP_MAX] = "n/a", trel2[FORMAT_TIMESTAMP_RELATIVE_MAX] = "n/a"; + char **a; + + format_timestamp(tstamp1, sizeof(tstamp1), t->next_elapse); + format_timestamp_relative(trel1, sizeof(trel1), t->next_elapse); + + format_timestamp(tstamp2, sizeof(tstamp2), t->last_trigger); + format_timestamp_relative(trel2, sizeof(trel2), t->last_trigger); + + if (t->machine) { + j = strjoin(t->machine, ":", t->id, NULL); + if (!j) + return log_oom(); + unit = j; + } else + unit = t->id; + + printf("%-*s %-*s %-*s %-*s %-*s", + nextlen, tstamp1, leftlen, trel1, lastlen, tstamp2, passedlen, trel2, unitlen, unit); + + STRV_FOREACH(a, t->triggered) + printf("%s %s", + a == t->triggered ? "" : ",", *a); + printf("\n"); + } + + on = ansi_highlight(); + off = ansi_normal(); + if (!arg_no_legend) + printf("\n"); + } else { + on = ansi_highlight_red(); + off = ansi_normal(); + } + + if (!arg_no_legend) { + printf("%s%u timers listed.%s\n", on, n, off); + if (!arg_all) + printf("Pass --all to see loaded but inactive timers, too.\n"); + } + + return 0; +} + +static usec_t calc_next_elapse(dual_timestamp *nw, dual_timestamp *next) { + usec_t next_elapse; + + assert(nw); + assert(next); + + if (next->monotonic != USEC_INFINITY && next->monotonic > 0) { + usec_t converted; + + if (next->monotonic > nw->monotonic) + converted = nw->realtime + (next->monotonic - nw->monotonic); + else + converted = nw->realtime - (nw->monotonic - next->monotonic); + + if (next->realtime != USEC_INFINITY && next->realtime > 0) + next_elapse = MIN(converted, next->realtime); + else + next_elapse = converted; + + } else + next_elapse = next->realtime; + + return next_elapse; +} + +static int list_timers(int argc, char *argv[], void *userdata) { + _cleanup_(message_set_freep) Set *replies = NULL; + _cleanup_strv_free_ char **machines = NULL; + _cleanup_free_ struct timer_info *timer_infos = NULL; + _cleanup_free_ UnitInfo *unit_infos = NULL; + struct timer_info *t; + const UnitInfo *u; + size_t size = 0; + int n, c = 0; + dual_timestamp nw; + sd_bus *bus; + int r = 0; + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + n = get_unit_list_recursive(bus, strv_skip(argv, 1), &unit_infos, &replies, &machines); + if (n < 0) + return n; + + dual_timestamp_get(&nw); + + for (u = unit_infos; u < unit_infos + n; u++) { + _cleanup_strv_free_ char **triggered = NULL; + dual_timestamp next = DUAL_TIMESTAMP_NULL; + usec_t m, last = 0; + + if (!endswith(u->id, ".timer")) + continue; + + r = get_triggered_units(bus, u->unit_path, &triggered); + if (r < 0) + goto cleanup; + + r = get_next_elapse(bus, u->unit_path, &next); + if (r < 0) + goto cleanup; + + get_last_trigger(bus, u->unit_path, &last); + + if (!GREEDY_REALLOC(timer_infos, size, c+1)) { + r = log_oom(); + goto cleanup; + } + + m = calc_next_elapse(&nw, &next); + + timer_infos[c++] = (struct timer_info) { + .machine = u->machine, + .id = u->id, + .next_elapse = m, + .last_trigger = last, + .triggered = triggered, + }; + + triggered = NULL; /* avoid cleanup */ + } + + qsort_safe(timer_infos, c, sizeof(struct timer_info), + (__compar_fn_t) timer_info_compare); + + output_timers_list(timer_infos, c); + + cleanup: + for (t = timer_infos; t < timer_infos + c; t++) + strv_free(t->triggered); + + return r; +} + +static int compare_unit_file_list(const void *a, const void *b) { + const char *d1, *d2; + const UnitFileList *u = a, *v = b; + + d1 = strrchr(u->path, '.'); + d2 = strrchr(v->path, '.'); + + if (d1 && d2) { + int r; + + r = strcasecmp(d1, d2); + if (r != 0) + return r; + } + + return strcasecmp(basename(u->path), basename(v->path)); +} + +static bool output_show_unit_file(const UnitFileList *u, char **states, char **patterns) { + assert(u); + + if (!strv_fnmatch_or_empty(patterns, basename(u->path), FNM_NOESCAPE)) + return false; + + if (!strv_isempty(arg_types)) { + const char *dot; + + dot = strrchr(u->path, '.'); + if (!dot) + return false; + + if (!strv_find(arg_types, dot+1)) + return false; + } + + if (!strv_isempty(states) && + !strv_find(states, unit_file_state_to_string(u->state))) + return false; + + return true; +} + +static void output_unit_file_list(const UnitFileList *units, unsigned c) { + unsigned max_id_len, id_cols, state_cols; + const UnitFileList *u; + + max_id_len = strlen("UNIT FILE"); + state_cols = strlen("STATE"); + + for (u = units; u < units + c; u++) { + max_id_len = MAX(max_id_len, strlen(basename(u->path))); + state_cols = MAX(state_cols, strlen(unit_file_state_to_string(u->state))); + } + + if (!arg_full) { + unsigned basic_cols; + + id_cols = MIN(max_id_len, 25u); + basic_cols = 1 + id_cols + state_cols; + if (basic_cols < (unsigned) columns()) + id_cols += MIN(columns() - basic_cols, max_id_len - id_cols); + } else + id_cols = max_id_len; + + if (!arg_no_legend && c > 0) + printf("%-*s %-*s\n", + id_cols, "UNIT FILE", + state_cols, "STATE"); + + for (u = units; u < units + c; u++) { + _cleanup_free_ char *e = NULL; + const char *on, *off; + const char *id; + + if (IN_SET(u->state, + UNIT_FILE_MASKED, + UNIT_FILE_MASKED_RUNTIME, + UNIT_FILE_DISABLED, + UNIT_FILE_BAD)) { + on = ansi_highlight_red(); + off = ansi_normal(); + } else if (u->state == UNIT_FILE_ENABLED) { + on = ansi_highlight_green(); + off = ansi_normal(); + } else + on = off = ""; + + id = basename(u->path); + + e = arg_full ? NULL : ellipsize(id, id_cols, 33); + + printf("%-*s %s%-*s%s\n", + id_cols, e ? e : id, + on, state_cols, unit_file_state_to_string(u->state), off); + } + + if (!arg_no_legend) + printf("\n%u unit files listed.\n", c); +} + +static int list_unit_files(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_free_ UnitFileList *units = NULL; + UnitFileList *unit; + size_t size = 0; + unsigned c = 0; + const char *state; + char *path; + int r; + bool fallback = false; + + pager_open(arg_no_pager, false); + + if (install_client_side()) { + Hashmap *h; + UnitFileList *u; + Iterator i; + unsigned n_units; + + h = hashmap_new(&string_hash_ops); + if (!h) + return log_oom(); + + r = unit_file_get_list(arg_scope, arg_root, h, arg_states, strv_skip(argv, 1)); + if (r < 0) { + unit_file_list_free(h); + return log_error_errno(r, "Failed to get unit file list: %m"); + } + + n_units = hashmap_size(h); + + units = new(UnitFileList, n_units ?: 1); /* avoid malloc(0) */ + if (!units) { + unit_file_list_free(h); + return log_oom(); + } + + HASHMAP_FOREACH(u, h, i) { + if (!output_show_unit_file(u, NULL, NULL)) + continue; + + units[c++] = *u; + free(u); + } + + assert(c <= n_units); + hashmap_free(h); + + r = 0; + } else { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "ListUnitFilesByPatterns"); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, arg_states); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, strv_skip(argv, 1)); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, &reply); + if (r < 0 && sd_bus_error_has_name(&error, SD_BUS_ERROR_UNKNOWN_METHOD)) { + /* Fallback to legacy ListUnitFiles method */ + fallback = true; + log_debug_errno(r, "Failed to list unit files: %s Falling back to ListUnitsFiles method.", bus_error_message(&error, r)); + m = sd_bus_message_unref(m); + sd_bus_error_free(&error); + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "ListUnitFiles"); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, &reply); + } + if (r < 0) + return log_error_errno(r, "Failed to list unit files: %s", bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "(ss)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(reply, "(ss)", &path, &state)) > 0) { + + if (!GREEDY_REALLOC(units, size, c + 1)) + return log_oom(); + + units[c] = (struct UnitFileList) { + path, + unit_file_state_from_string(state) + }; + + if (output_show_unit_file(&units[c], + fallback ? arg_states : NULL, + fallback ? strv_skip(argv, 1) : NULL)) + c++; + + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + } + + qsort_safe(units, c, sizeof(UnitFileList), compare_unit_file_list); + output_unit_file_list(units, c); + + if (install_client_side()) + for (unit = units; unit < units + c; unit++) + free(unit->path); + + return 0; +} + +static int list_dependencies_print(const char *name, int level, unsigned int branches, bool last) { + _cleanup_free_ char *n = NULL; + size_t max_len = MAX(columns(),20u); + size_t len = 0; + int i; + + if (!arg_plain) { + + for (i = level - 1; i >= 0; i--) { + len += 2; + if (len > max_len - 3 && !arg_full) { + printf("%s...\n",max_len % 2 ? "" : " "); + return 0; + } + printf("%s", special_glyph(branches & (1 << i) ? TREE_VERTICAL : TREE_SPACE)); + } + len += 2; + + if (len > max_len - 3 && !arg_full) { + printf("%s...\n",max_len % 2 ? "" : " "); + return 0; + } + + printf("%s", special_glyph(last ? TREE_RIGHT : TREE_BRANCH)); + } + + if (arg_full) { + printf("%s\n", name); + return 0; + } + + n = ellipsize(name, max_len-len, 100); + if (!n) + return log_oom(); + + printf("%s\n", n); + return 0; +} + +static int list_dependencies_get_dependencies(sd_bus *bus, const char *name, char ***deps) { + + static const char *dependencies[_DEPENDENCY_MAX] = { + [DEPENDENCY_FORWARD] = "Requires\0" + "Requisite\0" + "Wants\0" + "ConsistsOf\0" + "BindsTo\0", + [DEPENDENCY_REVERSE] = "RequiredBy\0" + "RequisiteOf\0" + "WantedBy\0" + "PartOf\0" + "BoundBy\0", + [DEPENDENCY_AFTER] = "After\0", + [DEPENDENCY_BEFORE] = "Before\0", + }; + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_strv_free_ char **ret = NULL; + _cleanup_free_ char *path = NULL; + int r; + + assert(bus); + assert(name); + assert(deps); + assert_cc(ELEMENTSOF(dependencies) == _DEPENDENCY_MAX); + + path = unit_dbus_path_from_name(name); + if (!path) + return log_oom(); + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.DBus.Properties", + "GetAll", + &error, + &reply, + "s", "org.freedesktop.systemd1.Unit"); + if (r < 0) + return log_error_errno(r, "Failed to get properties of %s: %s", name, bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "{sv}"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_DICT_ENTRY, "sv")) > 0) { + const char *prop; + + r = sd_bus_message_read(reply, "s", &prop); + if (r < 0) + return bus_log_parse_error(r); + + if (!nulstr_contains(dependencies[arg_dependency], prop)) { + r = sd_bus_message_skip(reply, "v"); + if (r < 0) + return bus_log_parse_error(r); + } else { + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_VARIANT, "as"); + if (r < 0) + return bus_log_parse_error(r); + + r = bus_message_read_strv_extend(reply, &ret); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + } + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + *deps = ret; + ret = NULL; + + return 0; +} + +static int list_dependencies_compare(const void *_a, const void *_b) { + const char **a = (const char**) _a, **b = (const char**) _b; + + if (unit_name_to_type(*a) == UNIT_TARGET && unit_name_to_type(*b) != UNIT_TARGET) + return 1; + if (unit_name_to_type(*a) != UNIT_TARGET && unit_name_to_type(*b) == UNIT_TARGET) + return -1; + + return strcasecmp(*a, *b); +} + +static int list_dependencies_one( + sd_bus *bus, + const char *name, + int level, + char ***units, + unsigned int branches) { + + _cleanup_strv_free_ char **deps = NULL; + char **c; + int r = 0; + + assert(bus); + assert(name); + assert(units); + + r = strv_extend(units, name); + if (r < 0) + return log_oom(); + + r = list_dependencies_get_dependencies(bus, name, &deps); + if (r < 0) + return r; + + qsort_safe(deps, strv_length(deps), sizeof (char*), list_dependencies_compare); + + STRV_FOREACH(c, deps) { + if (strv_contains(*units, *c)) { + if (!arg_plain) { + r = list_dependencies_print("...", level + 1, (branches << 1) | (c[1] == NULL ? 0 : 1), 1); + if (r < 0) + return r; + } + continue; + } + + if (arg_plain) + printf(" "); + else { + UnitActiveState active_state = _UNIT_ACTIVE_STATE_INVALID; + const char *on; + + (void) get_state_one_unit(bus, *c, &active_state); + + switch (active_state) { + case UNIT_ACTIVE: + case UNIT_RELOADING: + case UNIT_ACTIVATING: + on = ansi_highlight_green(); + break; + + case UNIT_INACTIVE: + case UNIT_DEACTIVATING: + on = ansi_normal(); + break; + + default: + on = ansi_highlight_red(); + break; + } + + printf("%s%s%s ", on, special_glyph(BLACK_CIRCLE), ansi_normal()); + } + + r = list_dependencies_print(*c, level, branches, c[1] == NULL); + if (r < 0) + return r; + + if (arg_all || unit_name_to_type(*c) == UNIT_TARGET) { + r = list_dependencies_one(bus, *c, level + 1, units, (branches << 1) | (c[1] == NULL ? 0 : 1)); + if (r < 0) + return r; + } + } + + if (!arg_plain) + strv_remove(*units, name); + + return 0; +} + +static int list_dependencies(int argc, char *argv[], void *userdata) { + _cleanup_strv_free_ char **units = NULL; + _cleanup_free_ char *unit = NULL; + const char *u; + sd_bus *bus; + int r; + + if (argv[1]) { + r = unit_name_mangle(argv[1], UNIT_NAME_NOGLOB, &unit); + if (r < 0) + return log_error_errno(r, "Failed to mangle unit name: %m"); + + u = unit; + } else + u = SPECIAL_DEFAULT_TARGET; + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + puts(u); + + return list_dependencies_one(bus, u, 0, &units, 0); +} + +struct machine_info { + bool is_host; + char *name; + char *state; + char *control_group; + uint32_t n_failed_units; + uint32_t n_jobs; + usec_t timestamp; +}; + +static const struct bus_properties_map machine_info_property_map[] = { + { "SystemState", "s", NULL, offsetof(struct machine_info, state) }, + { "NJobs", "u", NULL, offsetof(struct machine_info, n_jobs) }, + { "NFailedUnits", "u", NULL, offsetof(struct machine_info, n_failed_units) }, + { "ControlGroup", "s", NULL, offsetof(struct machine_info, control_group) }, + { "UserspaceTimestamp", "t", NULL, offsetof(struct machine_info, timestamp) }, + {} +}; + +static void machine_info_clear(struct machine_info *info) { + if (info) { + free(info->name); + free(info->state); + free(info->control_group); + zero(*info); + } +} + +static void free_machines_list(struct machine_info *machine_infos, int n) { + int i; + + if (!machine_infos) + return; + + for (i = 0; i < n; i++) + machine_info_clear(&machine_infos[i]); + + free(machine_infos); +} + +static int compare_machine_info(const void *a, const void *b) { + const struct machine_info *u = a, *v = b; + + if (u->is_host != v->is_host) + return u->is_host > v->is_host ? -1 : 1; + + return strcasecmp(u->name, v->name); +} + +static int get_machine_properties(sd_bus *bus, struct machine_info *mi) { + _cleanup_(sd_bus_flush_close_unrefp) sd_bus *container = NULL; + int r; + + assert(mi); + + if (!bus) { + r = sd_bus_open_system_machine(&container, mi->name); + if (r < 0) + return r; + + bus = container; + } + + r = bus_map_all_properties(bus, "org.freedesktop.systemd1", "/org/freedesktop/systemd1", machine_info_property_map, mi); + if (r < 0) + return r; + + return 0; +} + +static bool output_show_machine(const char *name, char **patterns) { + return strv_fnmatch_or_empty(patterns, name, FNM_NOESCAPE); +} + +static int get_machine_list( + sd_bus *bus, + struct machine_info **_machine_infos, + char **patterns) { + + struct machine_info *machine_infos = NULL; + _cleanup_strv_free_ char **m = NULL; + _cleanup_free_ char *hn = NULL; + size_t sz = 0; + char **i; + int c = 0, r; + + hn = gethostname_malloc(); + if (!hn) + return log_oom(); + + if (output_show_machine(hn, patterns)) { + if (!GREEDY_REALLOC0(machine_infos, sz, c+1)) + return log_oom(); + + machine_infos[c].is_host = true; + machine_infos[c].name = hn; + hn = NULL; + + get_machine_properties(bus, &machine_infos[c]); + c++; + } + + r = sd_get_machine_names(&m); + if (r < 0) + return log_error_errno(r, "Failed to get machine list: %m"); + + STRV_FOREACH(i, m) { + _cleanup_free_ char *class = NULL; + + if (!output_show_machine(*i, patterns)) + continue; + + sd_machine_get_class(*i, &class); + if (!streq_ptr(class, "container")) + continue; + + if (!GREEDY_REALLOC0(machine_infos, sz, c+1)) { + free_machines_list(machine_infos, c); + return log_oom(); + } + + machine_infos[c].is_host = false; + machine_infos[c].name = strdup(*i); + if (!machine_infos[c].name) { + free_machines_list(machine_infos, c); + return log_oom(); + } + + get_machine_properties(NULL, &machine_infos[c]); + c++; + } + + *_machine_infos = machine_infos; + return c; +} + +static void output_machines_list(struct machine_info *machine_infos, unsigned n) { + struct machine_info *m; + unsigned + circle_len = 0, + namelen = sizeof("NAME") - 1, + statelen = sizeof("STATE") - 1, + failedlen = sizeof("FAILED") - 1, + jobslen = sizeof("JOBS") - 1; + + assert(machine_infos || n == 0); + + for (m = machine_infos; m < machine_infos + n; m++) { + namelen = MAX(namelen, strlen(m->name) + (m->is_host ? sizeof(" (host)") - 1 : 0)); + statelen = MAX(statelen, m->state ? strlen(m->state) : 0); + failedlen = MAX(failedlen, DECIMAL_STR_WIDTH(m->n_failed_units)); + jobslen = MAX(jobslen, DECIMAL_STR_WIDTH(m->n_jobs)); + + if (!arg_plain && !streq_ptr(m->state, "running")) + circle_len = 2; + } + + if (!arg_no_legend) { + if (circle_len > 0) + fputs(" ", stdout); + + printf("%-*s %-*s %-*s %-*s\n", + namelen, "NAME", + statelen, "STATE", + failedlen, "FAILED", + jobslen, "JOBS"); + } + + for (m = machine_infos; m < machine_infos + n; m++) { + const char *on_state = "", *off_state = ""; + const char *on_failed = "", *off_failed = ""; + bool circle = false; + + if (streq_ptr(m->state, "degraded")) { + on_state = ansi_highlight_red(); + off_state = ansi_normal(); + circle = true; + } else if (!streq_ptr(m->state, "running")) { + on_state = ansi_highlight_yellow(); + off_state = ansi_normal(); + circle = true; + } + + if (m->n_failed_units > 0) { + on_failed = ansi_highlight_red(); + off_failed = ansi_normal(); + } else + on_failed = off_failed = ""; + + if (circle_len > 0) + printf("%s%s%s ", on_state, circle ? special_glyph(BLACK_CIRCLE) : " ", off_state); + + if (m->is_host) + printf("%-*s (host) %s%-*s%s %s%*" PRIu32 "%s %*" PRIu32 "\n", + (int) (namelen - (sizeof(" (host)")-1)), strna(m->name), + on_state, statelen, strna(m->state), off_state, + on_failed, failedlen, m->n_failed_units, off_failed, + jobslen, m->n_jobs); + else + printf("%-*s %s%-*s%s %s%*" PRIu32 "%s %*" PRIu32 "\n", + namelen, strna(m->name), + on_state, statelen, strna(m->state), off_state, + on_failed, failedlen, m->n_failed_units, off_failed, + jobslen, m->n_jobs); + } + + if (!arg_no_legend) + printf("\n%u machines listed.\n", n); +} + +static int list_machines(int argc, char *argv[], void *userdata) { + struct machine_info *machine_infos = NULL; + sd_bus *bus; + int r; + + if (geteuid() != 0) { + log_error("Must be root."); + return -EPERM; + } + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = get_machine_list(bus, &machine_infos, strv_skip(argv, 1)); + if (r < 0) + return r; + + qsort_safe(machine_infos, r, sizeof(struct machine_info), compare_machine_info); + output_machines_list(machine_infos, r); + free_machines_list(machine_infos, r); + + return 0; +} + +static int get_default(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_free_ char *_path = NULL; + const char *path; + int r; + + if (install_client_side()) { + r = unit_file_get_default(arg_scope, arg_root, &_path); + if (r < 0) + return log_error_errno(r, "Failed to get default target: %m"); + path = _path; + + r = 0; + } else { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "GetDefaultTarget", + &error, + &reply, + NULL); + if (r < 0) + return log_error_errno(r, "Failed to get default target: %s", bus_error_message(&error, r)); + + r = sd_bus_message_read(reply, "s", &path); + if (r < 0) + return bus_log_parse_error(r); + } + + if (path) + printf("%s\n", path); + + return 0; +} + +static int set_default(int argc, char *argv[], void *userdata) { + _cleanup_free_ char *unit = NULL; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + int r; + + assert(argc >= 2); + assert(argv); + + r = unit_name_mangle_with_suffix(argv[1], UNIT_NAME_NOGLOB, ".target", &unit); + if (r < 0) + return log_error_errno(r, "Failed to mangle unit name: %m"); + + if (install_client_side()) { + r = unit_file_set_default(arg_scope, arg_root, unit, true, &changes, &n_changes); + unit_file_dump_changes(r, "set default", changes, n_changes, arg_quiet); + + if (r > 0) + r = 0; + } else { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + sd_bus *bus; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "SetDefaultTarget", + &error, + &reply, + "sb", unit, 1); + if (r < 0) + return log_error_errno(r, "Failed to set default target: %s", bus_error_message(&error, r)); + + r = bus_deserialize_and_dump_unit_file_changes(reply, arg_quiet, &changes, &n_changes); + if (r < 0) + goto finish; + + /* Try to reload if enabled */ + if (!arg_no_reload) + r = daemon_reload(argc, argv, userdata); + else + r = 0; + } + +finish: + unit_file_changes_free(changes, n_changes); + + return r; +} + +struct job_info { + uint32_t id; + const char *name, *type, *state; +}; + +static void output_jobs_list(const struct job_info* jobs, unsigned n, bool skipped) { + unsigned id_len, unit_len, type_len, state_len; + const struct job_info *j; + const char *on, *off; + bool shorten = false; + + assert(n == 0 || jobs); + + if (n == 0) { + if (!arg_no_legend) { + on = ansi_highlight_green(); + off = ansi_normal(); + + printf("%sNo jobs %s.%s\n", on, skipped ? "listed" : "running", off); + } + return; + } + + pager_open(arg_no_pager, false); + + id_len = strlen("JOB"); + unit_len = strlen("UNIT"); + type_len = strlen("TYPE"); + state_len = strlen("STATE"); + + for (j = jobs; j < jobs + n; j++) { + uint32_t id = j->id; + assert(j->name && j->type && j->state); + + id_len = MAX(id_len, DECIMAL_STR_WIDTH(id)); + unit_len = MAX(unit_len, strlen(j->name)); + type_len = MAX(type_len, strlen(j->type)); + state_len = MAX(state_len, strlen(j->state)); + } + + if (!arg_full && id_len + 1 + unit_len + type_len + 1 + state_len > columns()) { + unit_len = MAX(33u, columns() - id_len - type_len - state_len - 3); + shorten = true; + } + + if (!arg_no_legend) + printf("%*s %-*s %-*s %-*s\n", + id_len, "JOB", + unit_len, "UNIT", + type_len, "TYPE", + state_len, "STATE"); + + for (j = jobs; j < jobs + n; j++) { + _cleanup_free_ char *e = NULL; + + if (streq(j->state, "running")) { + on = ansi_highlight(); + off = ansi_normal(); + } else + on = off = ""; + + e = shorten ? ellipsize(j->name, unit_len, 33) : NULL; + printf("%*u %s%-*s%s %-*s %s%-*s%s\n", + id_len, j->id, + on, unit_len, e ? e : j->name, off, + type_len, j->type, + on, state_len, j->state, off); + } + + if (!arg_no_legend) { + on = ansi_highlight(); + off = ansi_normal(); + + printf("\n%s%u jobs listed%s.\n", on, n, off); + } +} + +static bool output_show_job(struct job_info *job, char **patterns) { + return strv_fnmatch_or_empty(patterns, job->name, FNM_NOESCAPE); +} + +static int list_jobs(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + const char *name, *type, *state, *job_path, *unit_path; + _cleanup_free_ struct job_info *jobs = NULL; + size_t size = 0; + unsigned c = 0; + sd_bus *bus; + uint32_t id; + int r; + bool skipped = false; + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "ListJobs", + &error, + &reply, + NULL); + if (r < 0) + return log_error_errno(r, "Failed to list jobs: %s", bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, 'a', "(usssoo)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(reply, "(usssoo)", &id, &name, &type, &state, &job_path, &unit_path)) > 0) { + struct job_info job = { id, name, type, state }; + + if (!output_show_job(&job, strv_skip(argv, 1))) { + skipped = true; + continue; + } + + if (!GREEDY_REALLOC(jobs, size, c + 1)) + return log_oom(); + + jobs[c++] = job; + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + output_jobs_list(jobs, c, skipped); + return 0; +} + +static int cancel_job(int argc, char *argv[], void *userdata) { + sd_bus *bus; + char **name; + int r = 0; + + if (argc <= 1) + return daemon_reload(argc, argv, userdata); + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + STRV_FOREACH(name, strv_skip(argv, 1)) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + uint32_t id; + int q; + + q = safe_atou32(*name, &id); + if (q < 0) + return log_error_errno(q, "Failed to parse job id \"%s\": %m", *name); + + q = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "CancelJob", + &error, + NULL, + "u", id); + if (q < 0) { + log_error_errno(q, "Failed to cancel job %"PRIu32": %s", id, bus_error_message(&error, q)); + if (r == 0) + r = q; + } + } + + return r; +} + +static int need_daemon_reload(sd_bus *bus, const char *unit) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + const char *path; + int b, r; + + /* We ignore all errors here, since this is used to show a + * warning only */ + + /* We don't use unit_dbus_path_from_name() directly since we + * don't want to load the unit if it isn't loaded. */ + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "GetUnit", + NULL, + &reply, + "s", unit); + if (r < 0) + return r; + + r = sd_bus_message_read(reply, "o", &path); + if (r < 0) + return r; + + r = sd_bus_get_property_trivial( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Unit", + "NeedDaemonReload", + NULL, + 'b', &b); + if (r < 0) + return r; + + return b; +} + +static void warn_unit_file_changed(const char *name) { + assert(name); + + log_warning("%sWarning:%s %s changed on disk. Run 'systemctl%s daemon-reload' to reload units.", + ansi_highlight_red(), + ansi_normal(), + name, + arg_scope == UNIT_FILE_SYSTEM ? "" : " --user"); +} + +static int unit_file_find_path(LookupPaths *lp, const char *unit_name, char **unit_path) { + char **p; + + assert(lp); + assert(unit_name); + assert(unit_path); + + STRV_FOREACH(p, lp->search_path) { + _cleanup_free_ char *path; + + path = path_join(arg_root, *p, unit_name); + if (!path) + return log_oom(); + + if (access(path, F_OK) == 0) { + *unit_path = path; + path = NULL; + return 1; + } + } + + return 0; +} + +static int unit_find_paths( + sd_bus *bus, + const char *unit_name, + LookupPaths *lp, + char **fragment_path, + char ***dropin_paths) { + + _cleanup_free_ char *path = NULL; + _cleanup_strv_free_ char **dropins = NULL; + int r; + + /** + * Finds where the unit is defined on disk. Returns 0 if the unit + * is not found. Returns 1 if it is found, and sets + * - the path to the unit in *path, if it exists on disk, + * - and a strv of existing drop-ins in *dropins, + * if the arg is not NULL and any dropins were found. + */ + + assert(unit_name); + assert(fragment_path); + assert(lp); + + if (!install_client_side() && !unit_name_is_valid(unit_name, UNIT_NAME_TEMPLATE)) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_free_ char *unit = NULL; + + unit = unit_dbus_path_from_name(unit_name); + if (!unit) + return log_oom(); + + r = sd_bus_get_property_string( + bus, + "org.freedesktop.systemd1", + unit, + "org.freedesktop.systemd1.Unit", + "FragmentPath", + &error, + &path); + if (r < 0) + return log_error_errno(r, "Failed to get FragmentPath: %s", bus_error_message(&error, r)); + + if (dropin_paths) { + r = sd_bus_get_property_strv( + bus, + "org.freedesktop.systemd1", + unit, + "org.freedesktop.systemd1.Unit", + "DropInPaths", + &error, + &dropins); + if (r < 0) + return log_error_errno(r, "Failed to get DropInPaths: %s", bus_error_message(&error, r)); + } + } else { + _cleanup_set_free_ Set *names; + + names = set_new(NULL); + if (!names) + return log_oom(); + + r = set_put(names, unit_name); + if (r < 0) + return log_error_errno(r, "Failed to add unit name: %m"); + + r = unit_file_find_path(lp, unit_name, &path); + if (r < 0) + return r; + + if (r == 0) { + _cleanup_free_ char *template = NULL; + + r = unit_name_template(unit_name, &template); + if (r < 0 && r != -EINVAL) + return log_error_errno(r, "Failed to determine template name: %m"); + if (r >= 0) { + r = unit_file_find_path(lp, template, &path); + if (r < 0) + return r; + } + } + + if (dropin_paths) { + r = unit_file_find_dropin_paths(lp->search_path, NULL, names, &dropins); + if (r < 0) + return r; + } + } + + r = 0; + + if (!isempty(path)) { + *fragment_path = path; + path = NULL; + r = 1; + } + + if (dropin_paths && !strv_isempty(dropins)) { + *dropin_paths = dropins; + dropins = NULL; + r = 1; + } + + if (r == 0) + log_error("No files found for %s.", unit_name); + + return r; +} + +static int get_state_one_unit(sd_bus *bus, const char *name, UnitActiveState *active_state) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_free_ char *buf = NULL; + UnitActiveState state; + const char *path; + int r; + + assert(name); + assert(active_state); + + /* We don't use unit_dbus_path_from_name() directly since we don't want to load the unit unnecessarily, if it + * isn't loaded. */ + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "GetUnit", + &error, + &reply, + "s", name); + if (r < 0) { + if (!sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_UNIT)) + return log_error_errno(r, "Failed to retrieve unit: %s", bus_error_message(&error, r)); + + /* The unit is currently not loaded, hence say it's "inactive", since all units that aren't loaded are + * considered inactive. */ + state = UNIT_INACTIVE; + + } else { + r = sd_bus_message_read(reply, "o", &path); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_get_property_string( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Unit", + "ActiveState", + &error, + &buf); + if (r < 0) + return log_error_errno(r, "Failed to retrieve unit state: %s", bus_error_message(&error, r)); + + state = unit_active_state_from_string(buf); + if (state == _UNIT_ACTIVE_STATE_INVALID) { + log_error("Invalid unit state '%s' for: %s", buf, name); + return -EINVAL; + } + } + + *active_state = state; + return 0; +} + +static int check_triggering_units( + sd_bus *bus, + const char *name) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_free_ char *path = NULL, *n = NULL, *load_state = NULL; + _cleanup_strv_free_ char **triggered_by = NULL; + bool print_warning_label = true; + UnitActiveState active_state; + char **i; + int r; + + r = unit_name_mangle(name, UNIT_NAME_NOGLOB, &n); + if (r < 0) + return log_error_errno(r, "Failed to mangle unit name: %m"); + + path = unit_dbus_path_from_name(n); + if (!path) + return log_oom(); + + r = sd_bus_get_property_string( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Unit", + "LoadState", + &error, + &load_state); + if (r < 0) + return log_error_errno(r, "Failed to get load state of %s: %s", n, bus_error_message(&error, r)); + + if (streq(load_state, "masked")) + return 0; + + r = sd_bus_get_property_strv( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.systemd1.Unit", + "TriggeredBy", + &error, + &triggered_by); + if (r < 0) + return log_error_errno(r, "Failed to get triggered by array of %s: %s", n, bus_error_message(&error, r)); + + STRV_FOREACH(i, triggered_by) { + r = get_state_one_unit(bus, *i, &active_state); + if (r < 0) + return r; + + if (!IN_SET(active_state, UNIT_ACTIVE, UNIT_RELOADING)) + continue; + + if (print_warning_label) { + log_warning("Warning: Stopping %s, but it can still be activated by:", n); + print_warning_label = false; + } + + log_warning(" %s", *i); + } + + return 0; +} + +static const struct { + const char *verb; + const char *method; +} unit_actions[] = { + { "start", "StartUnit" }, + { "stop", "StopUnit" }, + { "condstop", "StopUnit" }, + { "reload", "ReloadUnit" }, + { "restart", "RestartUnit" }, + { "try-restart", "TryRestartUnit" }, + { "condrestart", "TryRestartUnit" }, + { "reload-or-restart", "ReloadOrRestartUnit" }, + { "try-reload-or-restart", "ReloadOrTryRestartUnit" }, + { "reload-or-try-restart", "ReloadOrTryRestartUnit" }, + { "condreload", "ReloadOrTryRestartUnit" }, + { "force-reload", "ReloadOrTryRestartUnit" } +}; + +static const char *verb_to_method(const char *verb) { + uint i; + + for (i = 0; i < ELEMENTSOF(unit_actions); i++) + if (streq_ptr(unit_actions[i].verb, verb)) + return unit_actions[i].method; + + return "StartUnit"; +} + +static const char *method_to_verb(const char *method) { + uint i; + + for (i = 0; i < ELEMENTSOF(unit_actions); i++) + if (streq_ptr(unit_actions[i].method, method)) + return unit_actions[i].verb; + + return "n/a"; +} + +static int start_unit_one( + sd_bus *bus, + const char *method, + const char *name, + const char *mode, + sd_bus_error *error, + BusWaitForJobs *w) { + + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + const char *path; + int r; + + assert(method); + assert(name); + assert(mode); + assert(error); + + log_debug("Calling manager for %s on %s, %s", method, name, mode); + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + method, + error, + &reply, + "ss", name, mode); + if (r < 0) { + const char *verb; + + if (r == -ENOENT && arg_action != ACTION_SYSTEMCTL) + /* There's always a fallback possible for + * legacy actions. */ + return -EADDRNOTAVAIL; + + verb = method_to_verb(method); + + log_error("Failed to %s %s: %s", verb, name, bus_error_message(error, r)); + + if (!sd_bus_error_has_name(error, BUS_ERROR_NO_SUCH_UNIT) && + !sd_bus_error_has_name(error, BUS_ERROR_UNIT_MASKED)) + log_error("See %s logs and 'systemctl%s status %s' for details.", + arg_scope == UNIT_FILE_SYSTEM ? "system" : "user", + arg_scope == UNIT_FILE_SYSTEM ? "" : " --user", + name); + + return r; + } + + r = sd_bus_message_read(reply, "o", &path); + if (r < 0) + return bus_log_parse_error(r); + + if (need_daemon_reload(bus, name) > 0) + warn_unit_file_changed(name); + + if (w) { + log_debug("Adding %s to the set", path); + r = bus_wait_for_jobs_add(w, path); + if (r < 0) + return log_oom(); + } + + return 0; +} + +static int expand_names(sd_bus *bus, char **names, const char* suffix, char ***ret) { + _cleanup_strv_free_ char **mangled = NULL, **globs = NULL; + char **name; + int r, i; + + assert(bus); + assert(ret); + + STRV_FOREACH(name, names) { + char *t; + + if (suffix) + r = unit_name_mangle_with_suffix(*name, UNIT_NAME_GLOB, suffix, &t); + else + r = unit_name_mangle(*name, UNIT_NAME_GLOB, &t); + if (r < 0) + return log_error_errno(r, "Failed to mangle name: %m"); + + if (string_is_glob(t)) + r = strv_consume(&globs, t); + else + r = strv_consume(&mangled, t); + if (r < 0) + return log_oom(); + } + + /* Query the manager only if any of the names are a glob, since + * this is fairly expensive */ + if (!strv_isempty(globs)) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_free_ UnitInfo *unit_infos = NULL; + size_t allocated, n; + + r = get_unit_list(bus, NULL, globs, &unit_infos, 0, &reply); + if (r < 0) + return r; + + n = strv_length(mangled); + allocated = n + 1; + + for (i = 0; i < r; i++) { + if (!GREEDY_REALLOC(mangled, allocated, n+2)) + return log_oom(); + + mangled[n] = strdup(unit_infos[i].id); + if (!mangled[n]) + return log_oom(); + + mangled[++n] = NULL; + } + } + + *ret = mangled; + mangled = NULL; /* do not free */ + + return 0; +} + +static const struct { + const char *target; + const char *verb; + const char *mode; +} action_table[_ACTION_MAX] = { + [ACTION_HALT] = { SPECIAL_HALT_TARGET, "halt", "replace-irreversibly" }, + [ACTION_POWEROFF] = { SPECIAL_POWEROFF_TARGET, "poweroff", "replace-irreversibly" }, + [ACTION_REBOOT] = { SPECIAL_REBOOT_TARGET, "reboot", "replace-irreversibly" }, + [ACTION_KEXEC] = { SPECIAL_KEXEC_TARGET, "kexec", "replace-irreversibly" }, + [ACTION_RUNLEVEL2] = { SPECIAL_MULTI_USER_TARGET, NULL, "isolate" }, + [ACTION_RUNLEVEL3] = { SPECIAL_MULTI_USER_TARGET, NULL, "isolate" }, + [ACTION_RUNLEVEL4] = { SPECIAL_MULTI_USER_TARGET, NULL, "isolate" }, + [ACTION_RUNLEVEL5] = { SPECIAL_GRAPHICAL_TARGET, NULL, "isolate" }, + [ACTION_RESCUE] = { SPECIAL_RESCUE_TARGET, "rescue", "isolate" }, + [ACTION_EMERGENCY] = { SPECIAL_EMERGENCY_TARGET, "emergency", "isolate" }, + [ACTION_DEFAULT] = { SPECIAL_DEFAULT_TARGET, "default", "isolate" }, + [ACTION_EXIT] = { SPECIAL_EXIT_TARGET, "exit", "replace-irreversibly" }, + [ACTION_SUSPEND] = { SPECIAL_SUSPEND_TARGET, "suspend", "replace-irreversibly" }, + [ACTION_HIBERNATE] = { SPECIAL_HIBERNATE_TARGET, "hibernate", "replace-irreversibly" }, + [ACTION_HYBRID_SLEEP] = { SPECIAL_HYBRID_SLEEP_TARGET, "hybrid-sleep", "replace-irreversibly" }, +}; + +static enum action verb_to_action(const char *verb) { + enum action i; + + for (i = _ACTION_INVALID; i < _ACTION_MAX; i++) + if (streq_ptr(action_table[i].verb, verb)) + return i; + + return _ACTION_INVALID; +} + +static int start_unit(int argc, char *argv[], void *userdata) { + _cleanup_(bus_wait_for_jobs_freep) BusWaitForJobs *w = NULL; + const char *method, *mode, *one_name, *suffix = NULL; + _cleanup_strv_free_ char **names = NULL; + sd_bus *bus; + char **name; + int r = 0; + + ask_password_agent_open_if_enabled(); + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + if (arg_action == ACTION_SYSTEMCTL) { + enum action action; + + method = verb_to_method(argv[0]); + action = verb_to_action(argv[0]); + + if (streq(argv[0], "isolate")) { + mode = "isolate"; + suffix = ".target"; + } else + mode = action_table[action].mode ?: arg_job_mode; + + one_name = action_table[action].target; + } else { + assert(arg_action < ELEMENTSOF(action_table)); + assert(action_table[arg_action].target); + + method = "StartUnit"; + + mode = action_table[arg_action].mode; + one_name = action_table[arg_action].target; + } + + if (one_name) + names = strv_new(one_name, NULL); + else { + r = expand_names(bus, strv_skip(argv, 1), suffix, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + } + + if (!arg_no_block) { + r = bus_wait_for_jobs_new(bus, &w); + if (r < 0) + return log_error_errno(r, "Could not watch jobs: %m"); + } + + STRV_FOREACH(name, names) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int q; + + q = start_unit_one(bus, method, *name, mode, &error, w); + if (r >= 0 && q < 0) + r = translate_bus_error_to_exit_status(q, &error); + } + + if (!arg_no_block) { + int q, arg_count = 0; + const char* extra_args[4] = {}; + + if (arg_scope != UNIT_FILE_SYSTEM) + extra_args[arg_count++] = "--user"; + + assert(IN_SET(arg_transport, BUS_TRANSPORT_LOCAL, BUS_TRANSPORT_REMOTE, BUS_TRANSPORT_MACHINE)); + if (arg_transport == BUS_TRANSPORT_REMOTE) { + extra_args[arg_count++] = "-H"; + extra_args[arg_count++] = arg_host; + } else if (arg_transport == BUS_TRANSPORT_MACHINE) { + extra_args[arg_count++] = "-M"; + extra_args[arg_count++] = arg_host; + } + + q = bus_wait_for_jobs(w, arg_quiet, extra_args); + if (q < 0) + return q; + + /* When stopping units, warn if they can still be triggered by + * another active unit (socket, path, timer) */ + if (!arg_quiet && streq(method, "StopUnit")) + STRV_FOREACH(name, names) + check_triggering_units(bus, *name); + } + + return r; +} + +static int logind_set_wall_message(void) { +#ifdef HAVE_LOGIND + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + _cleanup_free_ char *m = NULL; + int r; + + r = acquire_bus(BUS_FULL, &bus); + if (r < 0) + return r; + + m = strv_join(arg_wall, " "); + if (!m) + return log_oom(); + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "SetWallMessage", + &error, + NULL, + "sb", + m, + !arg_no_wall); + + if (r < 0) + return log_warning_errno(r, "Failed to set wall message, ignoring: %s", bus_error_message(&error, r)); + +#endif + return 0; +} + +/* Ask systemd-logind, which might grant access to unprivileged users + * through PolicyKit */ +static int logind_reboot(enum action a) { +#ifdef HAVE_LOGIND + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + const char *method, *description; + sd_bus *bus; + int r; + + polkit_agent_open_if_enabled(); + (void) logind_set_wall_message(); + + r = acquire_bus(BUS_FULL, &bus); + if (r < 0) + return r; + + switch (a) { + + case ACTION_REBOOT: + method = "Reboot"; + description = "reboot system"; + break; + + case ACTION_POWEROFF: + method = "PowerOff"; + description = "power off system"; + break; + + case ACTION_SUSPEND: + method = "Suspend"; + description = "suspend system"; + break; + + case ACTION_HIBERNATE: + method = "Hibernate"; + description = "hibernate system"; + break; + + case ACTION_HYBRID_SLEEP: + method = "HybridSleep"; + description = "put system into hybrid sleep"; + break; + + default: + return -EINVAL; + } + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + method, + &error, + NULL, + "b", arg_ask_password); + if (r < 0) + return log_error_errno(r, "Failed to %s via logind: %s", description, bus_error_message(&error, r)); + + return 0; +#else + return -ENOSYS; +#endif +} + +static int logind_check_inhibitors(enum action a) { +#ifdef HAVE_LOGIND + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_strv_free_ char **sessions = NULL; + const char *what, *who, *why, *mode; + uint32_t uid, pid; + sd_bus *bus; + unsigned c = 0; + char **s; + int r; + + if (arg_ignore_inhibitors || arg_force > 0) + return 0; + + if (arg_when > 0) + return 0; + + if (geteuid() == 0) + return 0; + + if (!on_tty()) + return 0; + + r = acquire_bus(BUS_FULL, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "ListInhibitors", + NULL, + &reply, + NULL); + if (r < 0) + /* If logind is not around, then there are no inhibitors... */ + return 0; + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "(ssssuu)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(reply, "(ssssuu)", &what, &who, &why, &mode, &uid, &pid)) > 0) { + _cleanup_free_ char *comm = NULL, *user = NULL; + _cleanup_strv_free_ char **sv = NULL; + + if (!streq(mode, "block")) + continue; + + sv = strv_split(what, ":"); + if (!sv) + return log_oom(); + + if ((pid_t) pid < 0) + return log_error_errno(ERANGE, "Bad PID %"PRIu32": %m", pid); + + if (!strv_contains(sv, + IN_SET(a, + ACTION_HALT, + ACTION_POWEROFF, + ACTION_REBOOT, + ACTION_KEXEC) ? "shutdown" : "sleep")) + continue; + + get_process_comm(pid, &comm); + user = uid_to_name(uid); + + log_warning("Operation inhibited by \"%s\" (PID "PID_FMT" \"%s\", user %s), reason is \"%s\".", + who, (pid_t) pid, strna(comm), strna(user), why); + + c++; + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + /* Check for current sessions */ + sd_get_sessions(&sessions); + STRV_FOREACH(s, sessions) { + _cleanup_free_ char *type = NULL, *tty = NULL, *seat = NULL, *user = NULL, *service = NULL, *class = NULL; + + if (sd_session_get_uid(*s, &uid) < 0 || uid == getuid()) + continue; + + if (sd_session_get_class(*s, &class) < 0 || !streq(class, "user")) + continue; + + if (sd_session_get_type(*s, &type) < 0 || (!streq(type, "x11") && !streq(type, "tty"))) + continue; + + sd_session_get_tty(*s, &tty); + sd_session_get_seat(*s, &seat); + sd_session_get_service(*s, &service); + user = uid_to_name(uid); + + log_warning("User %s is logged in on %s.", strna(user), isempty(tty) ? (isempty(seat) ? strna(service) : seat) : tty); + c++; + } + + if (c <= 0) + return 0; + + log_error("Please retry operation after closing inhibitors and logging out other users.\nAlternatively, ignore inhibitors and users with 'systemctl %s -i'.", + action_table[a].verb); + + return -EPERM; +#else + return 0; +#endif +} + +static int logind_prepare_firmware_setup(void) { +#ifdef HAVE_LOGIND + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + int r; + + r = acquire_bus(BUS_FULL, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "SetRebootToFirmwareSetup", + &error, + NULL, + "b", true); + if (r < 0) + return log_error_errno(r, "Cannot indicate to EFI to boot into setup mode: %s", bus_error_message(&error, r)); + + return 0; +#else + log_error("Cannot remotely indicate to EFI to boot into setup mode."); + return -ENOSYS; +#endif +} + +static int prepare_firmware_setup(void) { + int r; + + if (!arg_firmware_setup) + return 0; + + if (arg_transport == BUS_TRANSPORT_LOCAL) { + + r = efi_set_reboot_to_firmware(true); + if (r < 0) + log_debug_errno(r, "Cannot indicate to EFI to boot into setup mode, will retry via logind: %m"); + else + return r; + } + + return logind_prepare_firmware_setup(); +} + +static int set_exit_code(uint8_t code) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + int r; + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "SetExitCode", + &error, + NULL, + "y", code); + if (r < 0) + return log_error_errno(r, "Failed to set exit code: %s", bus_error_message(&error, r)); + + return 0; +} + +static int start_special(int argc, char *argv[], void *userdata) { + enum action a; + int r; + + assert(argv); + + a = verb_to_action(argv[0]); + + r = logind_check_inhibitors(a); + if (r < 0) + return r; + + if (arg_force >= 2 && geteuid() != 0) { + log_error("Must be root."); + return -EPERM; + } + + r = prepare_firmware_setup(); + if (r < 0) + return r; + + if (a == ACTION_REBOOT && argc > 1) { + r = update_reboot_parameter_and_warn(argv[1]); + if (r < 0) + return r; + + } else if (a == ACTION_EXIT && argc > 1) { + uint8_t code; + + /* If the exit code is not given on the command line, + * don't reset it to zero: just keep it as it might + * have been set previously. */ + + r = safe_atou8(argv[1], &code); + if (r < 0) + return log_error_errno(r, "Invalid exit code."); + + r = set_exit_code(code); + if (r < 0) + return r; + } + + if (arg_force >= 2 && + IN_SET(a, + ACTION_HALT, + ACTION_POWEROFF, + ACTION_REBOOT)) + return halt_now(a); + + if (arg_force >= 1 && + IN_SET(a, + ACTION_HALT, + ACTION_POWEROFF, + ACTION_REBOOT, + ACTION_KEXEC, + ACTION_EXIT)) + return daemon_reload(argc, argv, userdata); + + /* First try logind, to allow authentication with polkit */ + if (IN_SET(a, + ACTION_POWEROFF, + ACTION_REBOOT, + ACTION_SUSPEND, + ACTION_HIBERNATE, + ACTION_HYBRID_SLEEP)) { + r = logind_reboot(a); + if (r >= 0) + return r; + if (IN_SET(r, -EOPNOTSUPP, -EINPROGRESS)) + /* requested operation is not supported or already in progress */ + return r; + + /* On all other errors, try low-level operation */ + } + + return start_unit(argc, argv, userdata); +} + +static int check_unit_generic(int code, const UnitActiveState good_states[], int nb_states, char **args) { + _cleanup_strv_free_ char **names = NULL; + UnitActiveState active_state; + sd_bus *bus; + char **name; + int r, i; + bool found = false; + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = expand_names(bus, args, NULL, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + + STRV_FOREACH(name, names) { + r = get_state_one_unit(bus, *name, &active_state); + if (r < 0) + return r; + + if (!arg_quiet) + puts(unit_active_state_to_string(active_state)); + + for (i = 0; i < nb_states; ++i) + if (good_states[i] == active_state) + found = true; + } + + /* use the given return code for the case that we won't find + * any unit which matches the list */ + return found ? 0 : code; +} + +static int check_unit_active(int argc, char *argv[], void *userdata) { + const UnitActiveState states[] = { UNIT_ACTIVE, UNIT_RELOADING }; + /* According to LSB: 3, "program is not running" */ + return check_unit_generic(3, states, ELEMENTSOF(states), strv_skip(argv, 1)); +} + +static int check_unit_failed(int argc, char *argv[], void *userdata) { + const UnitActiveState states[] = { UNIT_FAILED }; + return check_unit_generic(1, states, ELEMENTSOF(states), strv_skip(argv, 1)); +} + +static int kill_unit(int argc, char *argv[], void *userdata) { + _cleanup_strv_free_ char **names = NULL; + char *kill_who = NULL, **name; + sd_bus *bus; + int r, q; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + if (!arg_kill_who) + arg_kill_who = "all"; + + /* --fail was specified */ + if (streq(arg_job_mode, "fail")) + kill_who = strjoina(arg_kill_who, "-fail"); + + r = expand_names(bus, strv_skip(argv, 1), NULL, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + + STRV_FOREACH(name, names) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + + q = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "KillUnit", + &error, + NULL, + "ssi", *names, kill_who ? kill_who : arg_kill_who, arg_signal); + if (q < 0) { + log_error_errno(q, "Failed to kill unit %s: %s", *names, bus_error_message(&error, q)); + if (r == 0) + r = q; + } + } + + return r; +} + +typedef struct ExecStatusInfo { + char *name; + + char *path; + char **argv; + + bool ignore; + + usec_t start_timestamp; + usec_t exit_timestamp; + pid_t pid; + int code; + int status; + + LIST_FIELDS(struct ExecStatusInfo, exec); +} ExecStatusInfo; + +static void exec_status_info_free(ExecStatusInfo *i) { + assert(i); + + free(i->name); + free(i->path); + strv_free(i->argv); + free(i); +} + +static int exec_status_info_deserialize(sd_bus_message *m, ExecStatusInfo *i) { + uint64_t start_timestamp, exit_timestamp, start_timestamp_monotonic, exit_timestamp_monotonic; + const char *path; + uint32_t pid; + int32_t code, status; + int ignore, r; + + assert(m); + assert(i); + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_STRUCT, "sasbttttuii"); + if (r < 0) + return bus_log_parse_error(r); + else if (r == 0) + return 0; + + r = sd_bus_message_read(m, "s", &path); + if (r < 0) + return bus_log_parse_error(r); + + i->path = strdup(path); + if (!i->path) + return log_oom(); + + r = sd_bus_message_read_strv(m, &i->argv); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_read(m, + "bttttuii", + &ignore, + &start_timestamp, &start_timestamp_monotonic, + &exit_timestamp, &exit_timestamp_monotonic, + &pid, + &code, &status); + if (r < 0) + return bus_log_parse_error(r); + + i->ignore = ignore; + i->start_timestamp = (usec_t) start_timestamp; + i->exit_timestamp = (usec_t) exit_timestamp; + i->pid = (pid_t) pid; + i->code = code; + i->status = status; + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 1; +} + +typedef struct UnitStatusInfo { + const char *id; + const char *load_state; + const char *active_state; + const char *sub_state; + const char *unit_file_state; + const char *unit_file_preset; + + const char *description; + const char *following; + + char **documentation; + + const char *fragment_path; + const char *source_path; + const char *control_group; + + char **dropin_paths; + + const char *load_error; + const char *result; + + usec_t inactive_exit_timestamp; + usec_t inactive_exit_timestamp_monotonic; + usec_t active_enter_timestamp; + usec_t active_exit_timestamp; + usec_t inactive_enter_timestamp; + + bool need_daemon_reload; + bool transient; + + /* Service */ + pid_t main_pid; + pid_t control_pid; + const char *status_text; + const char *pid_file; + bool running:1; + int status_errno; + + usec_t start_timestamp; + usec_t exit_timestamp; + + int exit_code, exit_status; + + usec_t condition_timestamp; + bool condition_result; + bool failed_condition_trigger; + bool failed_condition_negate; + const char *failed_condition; + const char *failed_condition_parameter; + + usec_t assert_timestamp; + bool assert_result; + bool failed_assert_trigger; + bool failed_assert_negate; + const char *failed_assert; + const char *failed_assert_parameter; + + /* Socket */ + unsigned n_accepted; + unsigned n_connections; + bool accept; + + /* Pairs of type, path */ + char **listen; + + /* Device */ + const char *sysfs_path; + + /* Mount, Automount */ + const char *where; + + /* Swap */ + const char *what; + + /* CGroup */ + uint64_t memory_current; + uint64_t memory_limit; + uint64_t cpu_usage_nsec; + uint64_t tasks_current; + uint64_t tasks_max; + + LIST_HEAD(ExecStatusInfo, exec); +} UnitStatusInfo; + +static void print_status_info( + sd_bus *bus, + UnitStatusInfo *i, + bool *ellipsized) { + + ExecStatusInfo *p; + const char *active_on, *active_off, *on, *off, *ss; + usec_t timestamp; + char since1[FORMAT_TIMESTAMP_RELATIVE_MAX], *s1; + char since2[FORMAT_TIMESTAMP_MAX], *s2; + const char *path; + char **t, **t2; + int r; + + assert(i); + + /* This shows pretty information about a unit. See + * print_property() for a low-level property printer */ + + if (streq_ptr(i->active_state, "failed")) { + active_on = ansi_highlight_red(); + active_off = ansi_normal(); + } else if (streq_ptr(i->active_state, "active") || streq_ptr(i->active_state, "reloading")) { + active_on = ansi_highlight_green(); + active_off = ansi_normal(); + } else + active_on = active_off = ""; + + printf("%s%s%s %s", active_on, special_glyph(BLACK_CIRCLE), active_off, strna(i->id)); + + if (i->description && !streq_ptr(i->id, i->description)) + printf(" - %s", i->description); + + printf("\n"); + + if (i->following) + printf(" Follow: unit currently follows state of %s\n", i->following); + + if (streq_ptr(i->load_state, "error")) { + on = ansi_highlight_red(); + off = ansi_normal(); + } else + on = off = ""; + + path = i->source_path ? i->source_path : i->fragment_path; + + if (i->load_error != 0) + printf(" Loaded: %s%s%s (Reason: %s)\n", + on, strna(i->load_state), off, i->load_error); + else if (path && !isempty(i->unit_file_state) && !isempty(i->unit_file_preset)) + printf(" Loaded: %s%s%s (%s; %s; vendor preset: %s)\n", + on, strna(i->load_state), off, path, i->unit_file_state, i->unit_file_preset); + else if (path && !isempty(i->unit_file_state)) + printf(" Loaded: %s%s%s (%s; %s)\n", + on, strna(i->load_state), off, path, i->unit_file_state); + else if (path) + printf(" Loaded: %s%s%s (%s)\n", + on, strna(i->load_state), off, path); + else + printf(" Loaded: %s%s%s\n", + on, strna(i->load_state), off); + + if (i->transient) + printf("Transient: yes\n"); + + if (!strv_isempty(i->dropin_paths)) { + _cleanup_free_ char *dir = NULL; + bool last = false; + char ** dropin; + + STRV_FOREACH(dropin, i->dropin_paths) { + if (! dir || last) { + printf(dir ? " " : " Drop-In: "); + + dir = mfree(dir); + + dir = dirname_malloc(*dropin); + if (!dir) { + log_oom(); + return; + } + + printf("%s\n %s", dir, + special_glyph(TREE_RIGHT)); + } + + last = ! (*(dropin + 1) && startswith(*(dropin + 1), dir)); + + printf("%s%s", basename(*dropin), last ? "\n" : ", "); + } + } + + ss = streq_ptr(i->active_state, i->sub_state) ? NULL : i->sub_state; + if (ss) + printf(" Active: %s%s (%s)%s", + active_on, strna(i->active_state), ss, active_off); + else + printf(" Active: %s%s%s", + active_on, strna(i->active_state), active_off); + + if (!isempty(i->result) && !streq(i->result, "success")) + printf(" (Result: %s)", i->result); + + timestamp = (streq_ptr(i->active_state, "active") || + streq_ptr(i->active_state, "reloading")) ? i->active_enter_timestamp : + (streq_ptr(i->active_state, "inactive") || + streq_ptr(i->active_state, "failed")) ? i->inactive_enter_timestamp : + streq_ptr(i->active_state, "activating") ? i->inactive_exit_timestamp : + i->active_exit_timestamp; + + s1 = format_timestamp_relative(since1, sizeof(since1), timestamp); + s2 = format_timestamp(since2, sizeof(since2), timestamp); + + if (s1) + printf(" since %s; %s\n", s2, s1); + else if (s2) + printf(" since %s\n", s2); + else + printf("\n"); + + if (!i->condition_result && i->condition_timestamp > 0) { + s1 = format_timestamp_relative(since1, sizeof(since1), i->condition_timestamp); + s2 = format_timestamp(since2, sizeof(since2), i->condition_timestamp); + + printf("Condition: start %scondition failed%s at %s%s%s\n", + ansi_highlight_yellow(), ansi_normal(), + s2, s1 ? "; " : "", strempty(s1)); + if (i->failed_condition_trigger) + printf(" none of the trigger conditions were met\n"); + else if (i->failed_condition) + printf(" %s=%s%s was not met\n", + i->failed_condition, + i->failed_condition_negate ? "!" : "", + i->failed_condition_parameter); + } + + if (!i->assert_result && i->assert_timestamp > 0) { + s1 = format_timestamp_relative(since1, sizeof(since1), i->assert_timestamp); + s2 = format_timestamp(since2, sizeof(since2), i->assert_timestamp); + + printf(" Assert: start %sassertion failed%s at %s%s%s\n", + ansi_highlight_red(), ansi_normal(), + s2, s1 ? "; " : "", strempty(s1)); + if (i->failed_assert_trigger) + printf(" none of the trigger assertions were met\n"); + else if (i->failed_assert) + printf(" %s=%s%s was not met\n", + i->failed_assert, + i->failed_assert_negate ? "!" : "", + i->failed_assert_parameter); + } + + if (i->sysfs_path) + printf(" Device: %s\n", i->sysfs_path); + if (i->where) + printf(" Where: %s\n", i->where); + if (i->what) + printf(" What: %s\n", i->what); + + STRV_FOREACH(t, i->documentation) + printf(" %*s %s\n", 9, t == i->documentation ? "Docs:" : "", *t); + + STRV_FOREACH_PAIR(t, t2, i->listen) + printf(" %*s %s (%s)\n", 9, t == i->listen ? "Listen:" : "", *t2, *t); + + if (i->accept) + printf(" Accepted: %u; Connected: %u\n", i->n_accepted, i->n_connections); + + LIST_FOREACH(exec, p, i->exec) { + _cleanup_free_ char *argv = NULL; + bool good; + + /* Only show exited processes here */ + if (p->code == 0) + continue; + + argv = strv_join(p->argv, " "); + printf(" Process: "PID_FMT" %s=%s ", p->pid, p->name, strna(argv)); + + good = is_clean_exit_lsb(p->code, p->status, NULL); + if (!good) { + on = ansi_highlight_red(); + off = ansi_normal(); + } else + on = off = ""; + + printf("%s(code=%s, ", on, sigchld_code_to_string(p->code)); + + if (p->code == CLD_EXITED) { + const char *c; + + printf("status=%i", p->status); + + c = exit_status_to_string(p->status, EXIT_STATUS_SYSTEMD); + if (c) + printf("/%s", c); + + } else + printf("signal=%s", signal_to_string(p->status)); + + printf(")%s\n", off); + + if (i->main_pid == p->pid && + i->start_timestamp == p->start_timestamp && + i->exit_timestamp == p->start_timestamp) + /* Let's not show this twice */ + i->main_pid = 0; + + if (p->pid == i->control_pid) + i->control_pid = 0; + } + + if (i->main_pid > 0 || i->control_pid > 0) { + if (i->main_pid > 0) { + printf(" Main PID: "PID_FMT, i->main_pid); + + if (i->running) { + _cleanup_free_ char *comm = NULL; + get_process_comm(i->main_pid, &comm); + if (comm) + printf(" (%s)", comm); + } else if (i->exit_code > 0) { + printf(" (code=%s, ", sigchld_code_to_string(i->exit_code)); + + if (i->exit_code == CLD_EXITED) { + const char *c; + + printf("status=%i", i->exit_status); + + c = exit_status_to_string(i->exit_status, EXIT_STATUS_SYSTEMD); + if (c) + printf("/%s", c); + + } else + printf("signal=%s", signal_to_string(i->exit_status)); + printf(")"); + } + + if (i->control_pid > 0) + printf(";"); + } + + if (i->control_pid > 0) { + _cleanup_free_ char *c = NULL; + + printf(" %8s: "PID_FMT, i->main_pid ? "" : " Control", i->control_pid); + + get_process_comm(i->control_pid, &c); + if (c) + printf(" (%s)", c); + } + + printf("\n"); + } + + if (i->status_text) + printf(" Status: \"%s\"\n", i->status_text); + if (i->status_errno > 0) + printf(" Error: %i (%s)\n", i->status_errno, strerror(i->status_errno)); + + if (i->tasks_current != (uint64_t) -1) { + printf(" Tasks: %" PRIu64, i->tasks_current); + + if (i->tasks_max != (uint64_t) -1) + printf(" (limit: %" PRIi64 ")\n", i->tasks_max); + else + printf("\n"); + } + + if (i->memory_current != (uint64_t) -1) { + char buf[FORMAT_BYTES_MAX]; + + printf(" Memory: %s", format_bytes(buf, sizeof(buf), i->memory_current)); + + if (i->memory_limit != (uint64_t) -1) + printf(" (limit: %s)\n", format_bytes(buf, sizeof(buf), i->memory_limit)); + else + printf("\n"); + } + + if (i->cpu_usage_nsec != (uint64_t) -1) { + char buf[FORMAT_TIMESPAN_MAX]; + printf(" CPU: %s\n", format_timespan(buf, sizeof(buf), i->cpu_usage_nsec / NSEC_PER_USEC, USEC_PER_MSEC)); + } + + if (i->control_group) + printf(" CGroup: %s\n", i->control_group); + + { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + static const char prefix[] = " "; + unsigned c; + + c = columns(); + if (c > sizeof(prefix) - 1) + c -= sizeof(prefix) - 1; + else + c = 0; + + r = unit_show_processes(bus, i->id, i->control_group, prefix, c, get_output_flags(), &error); + if (r == -EBADR) { + unsigned k = 0; + pid_t extra[2]; + + /* Fallback for older systemd versions where the GetUnitProcesses() call is not yet available */ + + if (i->main_pid > 0) + extra[k++] = i->main_pid; + + if (i->control_pid > 0) + extra[k++] = i->control_pid; + + show_cgroup_and_extra(SYSTEMD_CGROUP_CONTROLLER, i->control_group, prefix, c, extra, k, get_output_flags()); + } else if (r < 0) + log_warning_errno(r, "Failed to dump process list, ignoring: %s", bus_error_message(&error, r)); + } + + if (i->id && arg_transport == BUS_TRANSPORT_LOCAL) + show_journal_by_unit( + stdout, + i->id, + arg_output, + 0, + i->inactive_exit_timestamp_monotonic, + arg_lines, + getuid(), + get_output_flags() | OUTPUT_BEGIN_NEWLINE, + SD_JOURNAL_LOCAL_ONLY, + arg_scope == UNIT_FILE_SYSTEM, + ellipsized); + + if (i->need_daemon_reload) + warn_unit_file_changed(i->id); +} + +static void show_unit_help(UnitStatusInfo *i) { + char **p; + + assert(i); + + if (!i->documentation) { + log_info("Documentation for %s not known.", i->id); + return; + } + + STRV_FOREACH(p, i->documentation) + if (startswith(*p, "man:")) + show_man_page(*p + 4, false); + else + log_info("Can't show: %s", *p); +} + +static int status_property(const char *name, sd_bus_message *m, UnitStatusInfo *i, const char *contents) { + int r; + + assert(name); + assert(m); + assert(i); + + switch (contents[0]) { + + case SD_BUS_TYPE_STRING: { + const char *s; + + r = sd_bus_message_read(m, "s", &s); + if (r < 0) + return bus_log_parse_error(r); + + if (!isempty(s)) { + if (streq(name, "Id")) + i->id = s; + else if (streq(name, "LoadState")) + i->load_state = s; + else if (streq(name, "ActiveState")) + i->active_state = s; + else if (streq(name, "SubState")) + i->sub_state = s; + else if (streq(name, "Description")) + i->description = s; + else if (streq(name, "FragmentPath")) + i->fragment_path = s; + else if (streq(name, "SourcePath")) + i->source_path = s; +#ifndef NOLEGACY + else if (streq(name, "DefaultControlGroup")) { + const char *e; + e = startswith(s, SYSTEMD_CGROUP_CONTROLLER ":"); + if (e) + i->control_group = e; + } +#endif + else if (streq(name, "ControlGroup")) + i->control_group = s; + else if (streq(name, "StatusText")) + i->status_text = s; + else if (streq(name, "PIDFile")) + i->pid_file = s; + else if (streq(name, "SysFSPath")) + i->sysfs_path = s; + else if (streq(name, "Where")) + i->where = s; + else if (streq(name, "What")) + i->what = s; + else if (streq(name, "Following")) + i->following = s; + else if (streq(name, "UnitFileState")) + i->unit_file_state = s; + else if (streq(name, "UnitFilePreset")) + i->unit_file_preset = s; + else if (streq(name, "Result")) + i->result = s; + } + + break; + } + + case SD_BUS_TYPE_BOOLEAN: { + int b; + + r = sd_bus_message_read(m, "b", &b); + if (r < 0) + return bus_log_parse_error(r); + + if (streq(name, "Accept")) + i->accept = b; + else if (streq(name, "NeedDaemonReload")) + i->need_daemon_reload = b; + else if (streq(name, "ConditionResult")) + i->condition_result = b; + else if (streq(name, "AssertResult")) + i->assert_result = b; + else if (streq(name, "Transient")) + i->transient = b; + + break; + } + + case SD_BUS_TYPE_UINT32: { + uint32_t u; + + r = sd_bus_message_read(m, "u", &u); + if (r < 0) + return bus_log_parse_error(r); + + if (streq(name, "MainPID")) { + if (u > 0) { + i->main_pid = (pid_t) u; + i->running = true; + } + } else if (streq(name, "ControlPID")) + i->control_pid = (pid_t) u; + else if (streq(name, "ExecMainPID")) { + if (u > 0) + i->main_pid = (pid_t) u; + } else if (streq(name, "NAccepted")) + i->n_accepted = u; + else if (streq(name, "NConnections")) + i->n_connections = u; + + break; + } + + case SD_BUS_TYPE_INT32: { + int32_t j; + + r = sd_bus_message_read(m, "i", &j); + if (r < 0) + return bus_log_parse_error(r); + + if (streq(name, "ExecMainCode")) + i->exit_code = (int) j; + else if (streq(name, "ExecMainStatus")) + i->exit_status = (int) j; + else if (streq(name, "StatusErrno")) + i->status_errno = (int) j; + + break; + } + + case SD_BUS_TYPE_UINT64: { + uint64_t u; + + r = sd_bus_message_read(m, "t", &u); + if (r < 0) + return bus_log_parse_error(r); + + if (streq(name, "ExecMainStartTimestamp")) + i->start_timestamp = (usec_t) u; + else if (streq(name, "ExecMainExitTimestamp")) + i->exit_timestamp = (usec_t) u; + else if (streq(name, "ActiveEnterTimestamp")) + i->active_enter_timestamp = (usec_t) u; + else if (streq(name, "InactiveEnterTimestamp")) + i->inactive_enter_timestamp = (usec_t) u; + else if (streq(name, "InactiveExitTimestamp")) + i->inactive_exit_timestamp = (usec_t) u; + else if (streq(name, "InactiveExitTimestampMonotonic")) + i->inactive_exit_timestamp_monotonic = (usec_t) u; + else if (streq(name, "ActiveExitTimestamp")) + i->active_exit_timestamp = (usec_t) u; + else if (streq(name, "ConditionTimestamp")) + i->condition_timestamp = (usec_t) u; + else if (streq(name, "AssertTimestamp")) + i->assert_timestamp = (usec_t) u; + else if (streq(name, "MemoryCurrent")) + i->memory_current = u; + else if (streq(name, "MemoryLimit")) + i->memory_limit = u; + else if (streq(name, "TasksCurrent")) + i->tasks_current = u; + else if (streq(name, "TasksMax")) + i->tasks_max = u; + else if (streq(name, "CPUUsageNSec")) + i->cpu_usage_nsec = u; + + break; + } + + case SD_BUS_TYPE_ARRAY: + + if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && startswith(name, "Exec")) { + _cleanup_free_ ExecStatusInfo *info = NULL; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(sasbttttuii)"); + if (r < 0) + return bus_log_parse_error(r); + + info = new0(ExecStatusInfo, 1); + if (!info) + return log_oom(); + + while ((r = exec_status_info_deserialize(m, info)) > 0) { + + info->name = strdup(name); + if (!info->name) + return log_oom(); + + LIST_PREPEND(exec, i->exec, info); + + info = new0(ExecStatusInfo, 1); + if (!info) + return log_oom(); + } + + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "Listen")) { + const char *type, *path; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(ss)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(ss)", &type, &path)) > 0) { + + r = strv_extend(&i->listen, type); + if (r < 0) + return r; + + r = strv_extend(&i->listen, path); + if (r < 0) + return r; + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRING && streq(name, "DropInPaths")) { + + r = sd_bus_message_read_strv(m, &i->dropin_paths); + if (r < 0) + return bus_log_parse_error(r); + + } else if (contents[1] == SD_BUS_TYPE_STRING && streq(name, "Documentation")) { + + r = sd_bus_message_read_strv(m, &i->documentation); + if (r < 0) + return bus_log_parse_error(r); + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "Conditions")) { + const char *cond, *param; + int trigger, negate; + int32_t state; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(sbbsi)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(sbbsi)", &cond, &trigger, &negate, ¶m, &state)) > 0) { + log_debug("%s %d %d %s %d", cond, trigger, negate, param, state); + if (state < 0 && (!trigger || !i->failed_condition)) { + i->failed_condition = cond; + i->failed_condition_trigger = trigger; + i->failed_condition_negate = negate; + i->failed_condition_parameter = param; + } + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "Asserts")) { + const char *cond, *param; + int trigger, negate; + int32_t state; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(sbbsi)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(sbbsi)", &cond, &trigger, &negate, ¶m, &state)) > 0) { + log_debug("%s %d %d %s %d", cond, trigger, negate, param, state); + if (state < 0 && (!trigger || !i->failed_assert)) { + i->failed_assert = cond; + i->failed_assert_trigger = trigger; + i->failed_assert_negate = negate; + i->failed_assert_parameter = param; + } + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + } else + goto skip; + + break; + + case SD_BUS_TYPE_STRUCT_BEGIN: + + if (streq(name, "LoadError")) { + const char *n, *message; + + r = sd_bus_message_read(m, "(ss)", &n, &message); + if (r < 0) + return bus_log_parse_error(r); + + if (!isempty(message)) + i->load_error = message; + } else + goto skip; + + break; + + default: + goto skip; + } + + return 0; + +skip: + r = sd_bus_message_skip(m, contents); + if (r < 0) + return bus_log_parse_error(r); + + return 0; +} + +#define print_prop(name, fmt, ...) \ + do { \ + if (arg_value) \ + printf(fmt "\n", __VA_ARGS__); \ + else \ + printf("%s=" fmt "\n", name, __VA_ARGS__); \ + } while(0) + +static int print_property(const char *name, sd_bus_message *m, const char *contents) { + int r; + + assert(name); + assert(m); + + /* This is a low-level property printer, see + * print_status_info() for the nicer output */ + + if (arg_properties && !strv_find(arg_properties, name)) { + /* skip what we didn't read */ + r = sd_bus_message_skip(m, contents); + return r; + } + + switch (contents[0]) { + + case SD_BUS_TYPE_STRUCT_BEGIN: + + if (contents[1] == SD_BUS_TYPE_UINT32 && streq(name, "Job")) { + uint32_t u; + + r = sd_bus_message_read(m, "(uo)", &u, NULL); + if (r < 0) + return bus_log_parse_error(r); + + if (u > 0) + print_prop(name, "%"PRIu32, u); + else if (arg_all) + print_prop(name, "%s", ""); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRING && streq(name, "Unit")) { + const char *s; + + r = sd_bus_message_read(m, "(so)", &s, NULL); + if (r < 0) + return bus_log_parse_error(r); + + if (arg_all || !isempty(s)) + print_prop(name, "%s", s); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRING && streq(name, "LoadError")) { + const char *a = NULL, *b = NULL; + + r = sd_bus_message_read(m, "(ss)", &a, &b); + if (r < 0) + return bus_log_parse_error(r); + + if (arg_all || !isempty(a) || !isempty(b)) + print_prop(name, "%s \"%s\"", strempty(a), strempty(b)); + + return 0; + } else if (streq_ptr(name, "SystemCallFilter")) { + _cleanup_strv_free_ char **l = NULL; + int whitelist; + + r = sd_bus_message_enter_container(m, 'r', "bas"); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_read(m, "b", &whitelist); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_read_strv(m, &l); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + if (arg_all || whitelist || !strv_isempty(l)) { + bool first = true; + char **i; + + if (!arg_value) { + fputs(name, stdout); + fputc('=', stdout); + } + + if (!whitelist) + fputc('~', stdout); + + STRV_FOREACH(i, l) { + if (first) + first = false; + else + fputc(' ', stdout); + + fputs(*i, stdout); + } + fputc('\n', stdout); + } + + return 0; + } + + break; + + case SD_BUS_TYPE_ARRAY: + + if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "EnvironmentFiles")) { + const char *path; + int ignore; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(sb)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(sb)", &path, &ignore)) > 0) + print_prop("EnvironmentFile", "%s (ignore_errors=%s)\n", path, yes_no(ignore)); + + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "Paths")) { + const char *type, *path; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(ss)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(ss)", &type, &path)) > 0) + print_prop(type, "%s", path); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "Listen")) { + const char *type, *path; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(ss)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(ss)", &type, &path)) > 0) + if (arg_value) + puts(path); + else + printf("Listen%s=%s\n", type, path); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "Timers")) { + const char *base; + uint64_t value, next_elapse; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(stt)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(stt)", &base, &value, &next_elapse)) > 0) { + char timespan1[FORMAT_TIMESPAN_MAX], timespan2[FORMAT_TIMESPAN_MAX]; + + print_prop(base, "{ value=%s ; next_elapse=%s }", + format_timespan(timespan1, sizeof(timespan1), value, 0), + format_timespan(timespan2, sizeof(timespan2), next_elapse, 0)); + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && startswith(name, "Exec")) { + ExecStatusInfo info = {}; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(sasbttttuii)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = exec_status_info_deserialize(m, &info)) > 0) { + char timestamp1[FORMAT_TIMESTAMP_MAX], timestamp2[FORMAT_TIMESTAMP_MAX]; + _cleanup_free_ char *tt; + + tt = strv_join(info.argv, " "); + + print_prop(name, + "{ path=%s ; argv[]=%s ; ignore_errors=%s ; start_time=[%s] ; stop_time=[%s] ; pid="PID_FMT" ; code=%s ; status=%i%s%s }", + strna(info.path), + strna(tt), + yes_no(info.ignore), + strna(format_timestamp(timestamp1, sizeof(timestamp1), info.start_timestamp)), + strna(format_timestamp(timestamp2, sizeof(timestamp2), info.exit_timestamp)), + info.pid, + sigchld_code_to_string(info.code), + info.status, + info.code == CLD_EXITED ? "" : "/", + strempty(info.code == CLD_EXITED ? NULL : signal_to_string(info.status))); + + free(info.path); + strv_free(info.argv); + zero(info); + } + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && streq(name, "DeviceAllow")) { + const char *path, *rwm; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(ss)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(ss)", &path, &rwm)) > 0) + print_prop(name, "%s %s", strna(path), strna(rwm)); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && (streq(name, "IODeviceWeight") || streq(name, "BlockIODeviceWeight"))) { + const char *path; + uint64_t weight; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(st)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(st)", &path, &weight)) > 0) + print_prop(name, "%s %"PRIu64, strna(path), weight); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + + } else if (contents[1] == SD_BUS_TYPE_STRUCT_BEGIN && (cgroup_io_limit_type_from_string(name) >= 0 || + streq(name, "BlockIOReadBandwidth") || streq(name, "BlockIOWriteBandwidth"))) { + const char *path; + uint64_t bandwidth; + + r = sd_bus_message_enter_container(m, SD_BUS_TYPE_ARRAY, "(st)"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read(m, "(st)", &path, &bandwidth)) > 0) + print_prop(name, "%s %"PRIu64, strna(path), bandwidth); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(m); + if (r < 0) + return bus_log_parse_error(r); + + return 0; + } + + break; + } + + r = bus_print_property(name, m, arg_value, arg_all); + if (r < 0) + return bus_log_parse_error(r); + + if (r == 0) { + r = sd_bus_message_skip(m, contents); + if (r < 0) + return bus_log_parse_error(r); + + if (arg_all) + printf("%s=[unprintable]\n", name); + } + + return 0; +} + +static int show_one( + const char *verb, + sd_bus *bus, + const char *path, + bool show_properties, + bool *new_line, + bool *ellipsized) { + + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + UnitStatusInfo info = { + .memory_current = (uint64_t) -1, + .memory_limit = (uint64_t) -1, + .cpu_usage_nsec = (uint64_t) -1, + .tasks_current = (uint64_t) -1, + .tasks_max = (uint64_t) -1, + }; + ExecStatusInfo *p; + int r; + + assert(path); + assert(new_line); + + log_debug("Showing one %s", path); + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + path, + "org.freedesktop.DBus.Properties", + "GetAll", + &error, + &reply, + "s", ""); + if (r < 0) + return log_error_errno(r, "Failed to get properties: %s", bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "{sv}"); + if (r < 0) + return bus_log_parse_error(r); + + if (*new_line) + printf("\n"); + + *new_line = true; + + while ((r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_DICT_ENTRY, "sv")) > 0) { + const char *name, *contents; + + r = sd_bus_message_read(reply, "s", &name); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_peek_type(reply, NULL, &contents); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_VARIANT, contents); + if (r < 0) + return bus_log_parse_error(r); + + if (show_properties) + r = print_property(name, reply, contents); + else + r = status_property(name, reply, &info, contents); + if (r < 0) + return r; + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + } + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + r = 0; + + if (!show_properties) { + if (streq(verb, "help")) + show_unit_help(&info); + else + print_status_info(bus, &info, ellipsized); + } + + strv_free(info.documentation); + strv_free(info.dropin_paths); + strv_free(info.listen); + + if (!streq_ptr(info.active_state, "active") && + !streq_ptr(info.active_state, "reloading") && + streq(verb, "status")) { + /* According to LSB: "program not running" */ + /* 0: program is running or service is OK + * 1: program is dead and /run PID file exists + * 2: program is dead and /run/lock lock file exists + * 3: program is not running + * 4: program or service status is unknown + */ + if (info.pid_file && access(info.pid_file, F_OK) == 0) + r = 1; + else + r = 3; + } + + while ((p = info.exec)) { + LIST_REMOVE(exec, info.exec, p); + exec_status_info_free(p); + } + + return r; +} + +static int get_unit_dbus_path_by_pid( + sd_bus *bus, + uint32_t pid, + char **unit) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + char *u; + int r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "GetUnitByPID", + &error, + &reply, + "u", pid); + if (r < 0) + return log_error_errno(r, "Failed to get unit for PID %"PRIu32": %s", pid, bus_error_message(&error, r)); + + r = sd_bus_message_read(reply, "o", &u); + if (r < 0) + return bus_log_parse_error(r); + + u = strdup(u); + if (!u) + return log_oom(); + + *unit = u; + return 0; +} + +static int show_all( + const char* verb, + sd_bus *bus, + bool show_properties, + bool *new_line, + bool *ellipsized) { + + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_free_ UnitInfo *unit_infos = NULL; + const UnitInfo *u; + unsigned c; + int r, ret = 0; + + r = get_unit_list(bus, NULL, NULL, &unit_infos, 0, &reply); + if (r < 0) + return r; + + pager_open(arg_no_pager, false); + + c = (unsigned) r; + + qsort_safe(unit_infos, c, sizeof(UnitInfo), compare_unit_info); + + for (u = unit_infos; u < unit_infos + c; u++) { + _cleanup_free_ char *p = NULL; + + p = unit_dbus_path_from_name(u->id); + if (!p) + return log_oom(); + + r = show_one(verb, bus, p, show_properties, new_line, ellipsized); + if (r < 0) + return r; + else if (r > 0 && ret == 0) + ret = r; + } + + return ret; +} + +static int show_system_status(sd_bus *bus) { + char since1[FORMAT_TIMESTAMP_RELATIVE_MAX], since2[FORMAT_TIMESTAMP_MAX]; + _cleanup_free_ char *hn = NULL; + _cleanup_(machine_info_clear) struct machine_info mi = {}; + const char *on, *off; + int r; + + hn = gethostname_malloc(); + if (!hn) + return log_oom(); + + r = bus_map_all_properties(bus, "org.freedesktop.systemd1", "/org/freedesktop/systemd1", machine_info_property_map, &mi); + if (r < 0) + return log_error_errno(r, "Failed to read server status: %m"); + + if (streq_ptr(mi.state, "degraded")) { + on = ansi_highlight_red(); + off = ansi_normal(); + } else if (!streq_ptr(mi.state, "running")) { + on = ansi_highlight_yellow(); + off = ansi_normal(); + } else + on = off = ""; + + printf("%s%s%s %s\n", on, special_glyph(BLACK_CIRCLE), off, arg_host ? arg_host : hn); + + printf(" State: %s%s%s\n", + on, strna(mi.state), off); + + printf(" Jobs: %" PRIu32 " queued\n", mi.n_jobs); + printf(" Failed: %" PRIu32 " units\n", mi.n_failed_units); + + printf(" Since: %s; %s\n", + format_timestamp(since2, sizeof(since2), mi.timestamp), + format_timestamp_relative(since1, sizeof(since1), mi.timestamp)); + + printf(" CGroup: %s\n", mi.control_group ?: "/"); + if (IN_SET(arg_transport, + BUS_TRANSPORT_LOCAL, + BUS_TRANSPORT_MACHINE)) { + static const char prefix[] = " "; + unsigned c; + + c = columns(); + if (c > sizeof(prefix) - 1) + c -= sizeof(prefix) - 1; + else + c = 0; + + show_cgroup(SYSTEMD_CGROUP_CONTROLLER, strempty(mi.control_group), prefix, c, get_output_flags()); + } + + return 0; +} + +static int show(int argc, char *argv[], void *userdata) { + bool show_properties, show_status, show_help, new_line = false; + bool ellipsized = false; + int r, ret = 0; + sd_bus *bus; + + assert(argv); + + show_properties = streq(argv[0], "show"); + show_status = streq(argv[0], "status"); + show_help = streq(argv[0], "help"); + + if (show_help && argc <= 1) { + log_error("This command expects one or more unit names. Did you mean --help?"); + return -EINVAL; + } + + pager_open(arg_no_pager, false); + + if (show_status) + /* Increase max number of open files to 16K if we can, we + * might needs this when browsing journal files, which might + * be split up into many files. */ + setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(16384)); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + /* If no argument is specified inspect the manager itself */ + if (show_properties && argc <= 1) + return show_one(argv[0], bus, "/org/freedesktop/systemd1", show_properties, &new_line, &ellipsized); + + if (show_status && argc <= 1) { + + pager_open(arg_no_pager, false); + show_system_status(bus); + new_line = true; + + if (arg_all) + ret = show_all(argv[0], bus, false, &new_line, &ellipsized); + } else { + _cleanup_free_ char **patterns = NULL; + char **name; + + STRV_FOREACH(name, strv_skip(argv, 1)) { + _cleanup_free_ char *unit = NULL; + uint32_t id; + + if (safe_atou32(*name, &id) < 0) { + if (strv_push(&patterns, *name) < 0) + return log_oom(); + + continue; + } else if (show_properties) { + /* Interpret as job id */ + if (asprintf(&unit, "/org/freedesktop/systemd1/job/%u", id) < 0) + return log_oom(); + + } else { + /* Interpret as PID */ + r = get_unit_dbus_path_by_pid(bus, id, &unit); + if (r < 0) { + ret = r; + continue; + } + } + + r = show_one(argv[0], bus, unit, show_properties, &new_line, &ellipsized); + if (r < 0) + return r; + else if (r > 0 && ret == 0) + ret = r; + } + + if (!strv_isempty(patterns)) { + _cleanup_strv_free_ char **names = NULL; + + r = expand_names(bus, patterns, NULL, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + + STRV_FOREACH(name, names) { + _cleanup_free_ char *unit; + + unit = unit_dbus_path_from_name(*name); + if (!unit) + return log_oom(); + + r = show_one(argv[0], bus, unit, show_properties, &new_line, &ellipsized); + if (r < 0) + return r; + else if (r > 0 && ret == 0) + ret = r; + } + } + } + + if (ellipsized && !arg_quiet) + printf("Hint: Some lines were ellipsized, use -l to show in full.\n"); + + return ret; +} + +static int cat_file(const char *filename, bool newline) { + _cleanup_close_ int fd; + + fd = open(filename, O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (fd < 0) + return -errno; + + printf("%s%s# %s%s\n", + newline ? "\n" : "", + ansi_highlight_blue(), + filename, + ansi_normal()); + fflush(stdout); + + return copy_bytes(fd, STDOUT_FILENO, (uint64_t) -1, false); +} + +static int cat(int argc, char *argv[], void *userdata) { + _cleanup_lookup_paths_free_ LookupPaths lp = {}; + _cleanup_strv_free_ char **names = NULL; + char **name; + sd_bus *bus; + bool first = true; + int r; + + if (arg_transport != BUS_TRANSPORT_LOCAL) { + log_error("Cannot remotely cat units."); + return -EINVAL; + } + + r = lookup_paths_init(&lp, arg_scope, 0, arg_root); + if (r < 0) + return log_error_errno(r, "Failed to determine unit paths: %m"); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = expand_names(bus, strv_skip(argv, 1), NULL, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + + pager_open(arg_no_pager, false); + + STRV_FOREACH(name, names) { + _cleanup_free_ char *fragment_path = NULL; + _cleanup_strv_free_ char **dropin_paths = NULL; + char **path; + + r = unit_find_paths(bus, *name, &lp, &fragment_path, &dropin_paths); + if (r < 0) + return r; + else if (r == 0) + return -ENOENT; + + if (first) + first = false; + else + puts(""); + + if (fragment_path) { + r = cat_file(fragment_path, false); + if (r < 0) + return log_warning_errno(r, "Failed to cat %s: %m", fragment_path); + } + + STRV_FOREACH(path, dropin_paths) { + r = cat_file(*path, path == dropin_paths); + if (r < 0) + return log_warning_errno(r, "Failed to cat %s: %m", *path); + } + } + + return 0; +} + +static int set_property(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_free_ char *n = NULL; + sd_bus *bus; + char **i; + int r; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "SetUnitProperties"); + if (r < 0) + return bus_log_create_error(r); + + r = unit_name_mangle(argv[1], UNIT_NAME_NOGLOB, &n); + if (r < 0) + return log_error_errno(r, "Failed to mangle unit name: %m"); + + r = sd_bus_message_append(m, "sb", n, arg_runtime); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_open_container(m, SD_BUS_TYPE_ARRAY, "(sv)"); + if (r < 0) + return bus_log_create_error(r); + + STRV_FOREACH(i, strv_skip(argv, 2)) { + r = bus_append_unit_property_assignment(m, *i); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(m); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, NULL); + if (r < 0) + return log_error_errno(r, "Failed to set unit properties on %s: %s", n, bus_error_message(&error, r)); + + return 0; +} + +static int daemon_reload(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + const char *method; + sd_bus *bus; + int r; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + if (arg_action == ACTION_RELOAD) + method = "Reload"; + else if (arg_action == ACTION_REEXEC) + method = "Reexecute"; + else { + assert(arg_action == ACTION_SYSTEMCTL); + + method = + streq(argv[0], "clear-jobs") || + streq(argv[0], "cancel") ? "ClearJobs" : + streq(argv[0], "daemon-reexec") ? "Reexecute" : + streq(argv[0], "reset-failed") ? "ResetFailed" : + streq(argv[0], "halt") ? "Halt" : + streq(argv[0], "poweroff") ? "PowerOff" : + streq(argv[0], "reboot") ? "Reboot" : + streq(argv[0], "kexec") ? "KExec" : + streq(argv[0], "exit") ? "Exit" : + /* "daemon-reload" */ "Reload"; + } + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + method, + &error, + NULL, + NULL); + if (r == -ENOENT && arg_action != ACTION_SYSTEMCTL) + /* There's always a fallback possible for + * legacy actions. */ + r = -EADDRNOTAVAIL; + else if ((r == -ETIMEDOUT || r == -ECONNRESET) && streq(method, "Reexecute")) + /* On reexecution, we expect a disconnect, not a + * reply */ + r = 0; + else if (r < 0) + return log_error_errno(r, "Failed to reload daemon: %s", bus_error_message(&error, r)); + + return r < 0 ? r : 0; +} + +static int reset_failed(int argc, char *argv[], void *userdata) { + _cleanup_strv_free_ char **names = NULL; + sd_bus *bus; + char **name; + int r, q; + + if (argc <= 1) + return daemon_reload(argc, argv, userdata); + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = expand_names(bus, strv_skip(argv, 1), NULL, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + + STRV_FOREACH(name, names) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + + q = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "ResetFailedUnit", + &error, + NULL, + "s", *name); + if (q < 0) { + log_error_errno(q, "Failed to reset failed state of unit %s: %s", *name, bus_error_message(&error, q)); + if (r == 0) + r = q; + } + } + + return r; +} + +static int show_environment(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + const char *text; + sd_bus *bus; + int r; + + pager_open(arg_no_pager, false); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_get_property( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "Environment", + &error, + &reply, + "as"); + if (r < 0) + return log_error_errno(r, "Failed to get environment: %s", bus_error_message(&error, r)); + + r = sd_bus_message_enter_container(reply, SD_BUS_TYPE_ARRAY, "s"); + if (r < 0) + return bus_log_parse_error(r); + + while ((r = sd_bus_message_read_basic(reply, SD_BUS_TYPE_STRING, &text)) > 0) + puts(text); + if (r < 0) + return bus_log_parse_error(r); + + r = sd_bus_message_exit_container(reply); + if (r < 0) + return bus_log_parse_error(r); + + return 0; +} + +static int switch_root(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_free_ char *cmdline_init = NULL; + const char *root, *init; + sd_bus *bus; + int r; + + if (arg_transport != BUS_TRANSPORT_LOCAL) { + log_error("Cannot switch root remotely."); + return -EINVAL; + } + + if (argc < 2 || argc > 3) { + log_error("Wrong number of arguments."); + return -EINVAL; + } + + root = argv[1]; + + if (argc >= 3) + init = argv[2]; + else { + r = parse_env_file("/proc/cmdline", WHITESPACE, + "init", &cmdline_init, + NULL); + if (r < 0) + log_debug_errno(r, "Failed to parse /proc/cmdline: %m"); + + init = cmdline_init; + } + + if (isempty(init)) + init = NULL; + + if (init) { + const char *root_systemd_path = NULL, *root_init_path = NULL; + + root_systemd_path = strjoina(root, "/" SYSTEMD_BINARY_PATH); + root_init_path = strjoina(root, "/", init); + + /* If the passed init is actually the same as the + * systemd binary, then let's suppress it. */ + if (files_same(root_init_path, root_systemd_path) > 0) + init = NULL; + } + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + log_debug("Switching root - root: %s; init: %s", root, strna(init)); + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "SwitchRoot", + &error, + NULL, + "ss", root, init); + if (r < 0) + return log_error_errno(r, "Failed to switch root: %s", bus_error_message(&error, r)); + + return 0; +} + +static int set_environment(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + const char *method; + sd_bus *bus; + int r; + + assert(argc > 1); + assert(argv); + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + method = streq(argv[0], "set-environment") + ? "SetEnvironment" + : "UnsetEnvironment"; + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + method); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, strv_skip(argv, 1)); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, NULL); + if (r < 0) + return log_error_errno(r, "Failed to set environment: %s", bus_error_message(&error, r)); + + return 0; +} + +static int import_environment(int argc, char *argv[], void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + sd_bus *bus; + int r; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "SetEnvironment"); + if (r < 0) + return bus_log_create_error(r); + + if (argc < 2) + r = sd_bus_message_append_strv(m, environ); + else { + char **a, **b; + + r = sd_bus_message_open_container(m, 'a', "s"); + if (r < 0) + return bus_log_create_error(r); + + STRV_FOREACH(a, strv_skip(argv, 1)) { + + if (!env_name_is_valid(*a)) { + log_error("Not a valid environment variable name: %s", *a); + return -EINVAL; + } + + STRV_FOREACH(b, environ) { + const char *eq; + + eq = startswith(*b, *a); + if (eq && *eq == '=') { + + r = sd_bus_message_append(m, "s", *b); + if (r < 0) + return bus_log_create_error(r); + + break; + } + } + } + + r = sd_bus_message_close_container(m); + } + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, NULL); + if (r < 0) + return log_error_errno(r, "Failed to import environment: %s", bus_error_message(&error, r)); + + return 0; +} + +static int enable_sysv_units(const char *verb, char **args) { + int r = 0; + +#if defined(HAVE_SYSV_COMPAT) + _cleanup_lookup_paths_free_ LookupPaths paths = {}; + unsigned f = 0; + + /* Processes all SysV units, and reshuffles the array so that afterwards only the native units remain */ + + if (arg_scope != UNIT_FILE_SYSTEM) + return 0; + + if (getenv_bool("SYSTEMCTL_SKIP_SYSV") > 0) + return 0; + + if (!STR_IN_SET(verb, + "enable", + "disable", + "is-enabled")) + return 0; + + r = lookup_paths_init(&paths, arg_scope, LOOKUP_PATHS_EXCLUDE_GENERATED, arg_root); + if (r < 0) + return r; + + r = 0; + while (args[f]) { + + const char *argv[] = { + ROOTLIBEXECDIR "/systemd-sysv-install", + NULL, + NULL, + NULL, + NULL, + }; + + _cleanup_free_ char *p = NULL, *q = NULL, *l = NULL; + bool found_native = false, found_sysv; + siginfo_t status; + const char *name; + unsigned c = 1; + pid_t pid; + int j; + + name = args[f++]; + + if (!endswith(name, ".service")) + continue; + + if (path_is_absolute(name)) + continue; + + j = unit_file_exists(arg_scope, &paths, name); + if (j < 0 && !IN_SET(j, -ELOOP, -ERFKILL, -EADDRNOTAVAIL)) + return log_error_errno(j, "Failed to lookup unit file state: %m"); + found_native = j != 0; + + /* If we have both a native unit and a SysV script, enable/disable them both (below); for is-enabled, + * prefer the native unit */ + if (found_native && streq(verb, "is-enabled")) + continue; + + p = path_join(arg_root, SYSTEM_SYSVINIT_PATH, name); + if (!p) + return log_oom(); + + p[strlen(p) - strlen(".service")] = 0; + found_sysv = access(p, F_OK) >= 0; + if (!found_sysv) + continue; + + if (found_native) + log_info("Synchronizing state of %s with SysV service script with %s.", name, argv[0]); + else + log_info("%s is not a native service, redirecting to systemd-sysv-install.", name); + + if (!isempty(arg_root)) + argv[c++] = q = strappend("--root=", arg_root); + + argv[c++] = verb; + argv[c++] = basename(p); + argv[c] = NULL; + + l = strv_join((char**)argv, " "); + if (!l) + return log_oom(); + + log_info("Executing: %s", l); + + pid = fork(); + if (pid < 0) + return log_error_errno(errno, "Failed to fork: %m"); + else if (pid == 0) { + /* Child */ + + (void) reset_all_signal_handlers(); + (void) reset_signal_mask(); + + execv(argv[0], (char**) argv); + log_error_errno(errno, "Failed to execute %s: %m", argv[0]); + _exit(EXIT_FAILURE); + } + + j = wait_for_terminate(pid, &status); + if (j < 0) { + log_error_errno(j, "Failed to wait for child: %m"); + return j; + } + + if (status.si_code == CLD_EXITED) { + if (streq(verb, "is-enabled")) { + if (status.si_status == 0) { + if (!arg_quiet) + puts("enabled"); + r = 1; + } else { + if (!arg_quiet) + puts("disabled"); + } + + } else if (status.si_status != 0) + return -EBADE; /* We don't warn here, under the assumption the script already showed an explanation */ + } else { + log_error("Unexpected waitid() result."); + return -EPROTO; + } + + if (found_native) + continue; + + /* Remove this entry, so that we don't try enabling it as native unit */ + assert(f > 0); + f--; + assert(args[f] == name); + strv_remove(args, name); + } + +#endif + return r; +} + +static int mangle_names(char **original_names, char ***mangled_names) { + char **i, **l, **name; + int r; + + l = i = new(char*, strv_length(original_names) + 1); + if (!l) + return log_oom(); + + STRV_FOREACH(name, original_names) { + + /* When enabling units qualified path names are OK, + * too, hence allow them explicitly. */ + + if (is_path(*name)) { + *i = strdup(*name); + if (!*i) { + strv_free(l); + return log_oom(); + } + } else { + r = unit_name_mangle(*name, UNIT_NAME_NOGLOB, i); + if (r < 0) { + strv_free(l); + return log_error_errno(r, "Failed to mangle unit name: %m"); + } + } + + i++; + } + + *i = NULL; + *mangled_names = l; + + return 0; +} + +static int enable_unit(int argc, char *argv[], void *userdata) { + _cleanup_strv_free_ char **names = NULL; + const char *verb = argv[0]; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + int carries_install_info = -1; + bool ignore_carries_install_info = arg_quiet; + int r; + + if (!argv[1]) + return 0; + + r = mangle_names(strv_skip(argv, 1), &names); + if (r < 0) + return r; + + r = enable_sysv_units(verb, names); + if (r < 0) + return r; + + /* If the operation was fully executed by the SysV compat, let's finish early */ + if (strv_isempty(names)) { + if (arg_no_reload || install_client_side()) + return 0; + return daemon_reload(argc, argv, userdata); + } + + if (install_client_side()) { + if (streq(verb, "enable")) { + r = unit_file_enable(arg_scope, arg_runtime, arg_root, names, arg_force, &changes, &n_changes); + carries_install_info = r; + } else if (streq(verb, "disable")) + r = unit_file_disable(arg_scope, arg_runtime, arg_root, names, &changes, &n_changes); + else if (streq(verb, "reenable")) { + r = unit_file_reenable(arg_scope, arg_runtime, arg_root, names, arg_force, &changes, &n_changes); + carries_install_info = r; + } else if (streq(verb, "link")) + r = unit_file_link(arg_scope, arg_runtime, arg_root, names, arg_force, &changes, &n_changes); + else if (streq(verb, "preset")) { + r = unit_file_preset(arg_scope, arg_runtime, arg_root, names, arg_preset_mode, arg_force, &changes, &n_changes); + } else if (streq(verb, "mask")) + r = unit_file_mask(arg_scope, arg_runtime, arg_root, names, arg_force, &changes, &n_changes); + else if (streq(verb, "unmask")) + r = unit_file_unmask(arg_scope, arg_runtime, arg_root, names, &changes, &n_changes); + else if (streq(verb, "revert")) + r = unit_file_revert(arg_scope, arg_root, names, &changes, &n_changes); + else + assert_not_reached("Unknown verb"); + + unit_file_dump_changes(r, verb, changes, n_changes, arg_quiet); + if (r < 0) + return r; + r = 0; + } else { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL, *m = NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + bool expect_carries_install_info = false; + bool send_runtime = true, send_force = true, send_preset_mode = false; + const char *method; + sd_bus *bus; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + if (streq(verb, "enable")) { + method = "EnableUnitFiles"; + expect_carries_install_info = true; + } else if (streq(verb, "disable")) { + method = "DisableUnitFiles"; + send_force = false; + } else if (streq(verb, "reenable")) { + method = "ReenableUnitFiles"; + expect_carries_install_info = true; + } else if (streq(verb, "link")) + method = "LinkUnitFiles"; + else if (streq(verb, "preset")) { + + if (arg_preset_mode != UNIT_FILE_PRESET_FULL) { + method = "PresetUnitFilesWithMode"; + send_preset_mode = true; + } else + method = "PresetUnitFiles"; + + expect_carries_install_info = true; + ignore_carries_install_info = true; + } else if (streq(verb, "mask")) + method = "MaskUnitFiles"; + else if (streq(verb, "unmask")) { + method = "UnmaskUnitFiles"; + send_force = false; + } else if (streq(verb, "revert")) { + method = "RevertUnitFiles"; + send_runtime = send_force = false; + } else + assert_not_reached("Unknown verb"); + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + method); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, names); + if (r < 0) + return bus_log_create_error(r); + + if (send_preset_mode) { + r = sd_bus_message_append(m, "s", unit_file_preset_mode_to_string(arg_preset_mode)); + if (r < 0) + return bus_log_create_error(r); + } + + if (send_runtime) { + r = sd_bus_message_append(m, "b", arg_runtime); + if (r < 0) + return bus_log_create_error(r); + } + + if (send_force) { + r = sd_bus_message_append(m, "b", arg_force); + if (r < 0) + return bus_log_create_error(r); + } + + r = sd_bus_call(bus, m, 0, &error, &reply); + if (r < 0) + return log_error_errno(r, "Failed to %s unit: %s", verb, bus_error_message(&error, r)); + + if (expect_carries_install_info) { + r = sd_bus_message_read(reply, "b", &carries_install_info); + if (r < 0) + return bus_log_parse_error(r); + } + + r = bus_deserialize_and_dump_unit_file_changes(reply, arg_quiet, &changes, &n_changes); + if (r < 0) + return r; + + /* Try to reload if enabled */ + if (!arg_no_reload) + r = daemon_reload(argc, argv, userdata); + else + r = 0; + } + + if (carries_install_info == 0 && !ignore_carries_install_info) + log_warning("The unit files have no installation config (WantedBy, RequiredBy, Also, Alias\n" + "settings in the [Install] section, and DefaultInstance for template units).\n" + "This means they are not meant to be enabled using systemctl.\n" + "Possible reasons for having this kind of units are:\n" + "1) A unit may be statically enabled by being symlinked from another unit's\n" + " .wants/ or .requires/ directory.\n" + "2) A unit's purpose may be to act as a helper for some other unit which has\n" + " a requirement dependency on it.\n" + "3) A unit may be started when needed via activation (socket, path, timer,\n" + " D-Bus, udev, scripted systemctl call, ...).\n" + "4) In case of template units, the unit is meant to be enabled with some\n" + " instance name specified."); + + if (arg_now && n_changes > 0 && STR_IN_SET(argv[0], "enable", "disable", "mask")) { + char *new_args[n_changes + 2]; + sd_bus *bus; + unsigned i; + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + goto finish; + + new_args[0] = (char*) (streq(argv[0], "enable") ? "start" : "stop"); + for (i = 0; i < n_changes; i++) + new_args[i + 1] = basename(changes[i].path); + new_args[i + 1] = NULL; + + r = start_unit(strv_length(new_args), new_args, userdata); + } + +finish: + unit_file_changes_free(changes, n_changes); + + return r; +} + +static int add_dependency(int argc, char *argv[], void *userdata) { + _cleanup_strv_free_ char **names = NULL; + _cleanup_free_ char *target = NULL; + const char *verb = argv[0]; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + UnitDependency dep; + int r = 0; + + if (!argv[1]) + return 0; + + r = unit_name_mangle_with_suffix(argv[1], UNIT_NAME_NOGLOB, ".target", &target); + if (r < 0) + return log_error_errno(r, "Failed to mangle unit name: %m"); + + r = mangle_names(strv_skip(argv, 2), &names); + if (r < 0) + return r; + + if (streq(verb, "add-wants")) + dep = UNIT_WANTS; + else if (streq(verb, "add-requires")) + dep = UNIT_REQUIRES; + else + assert_not_reached("Unknown verb"); + + if (install_client_side()) { + r = unit_file_add_dependency(arg_scope, arg_runtime, arg_root, names, target, dep, arg_force, &changes, &n_changes); + unit_file_dump_changes(r, "add dependency on", changes, n_changes, arg_quiet); + + if (r > 0) + r = 0; + } else { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL, *m = NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_message_new_method_call( + bus, + &m, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "AddDependencyUnitFiles"); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append_strv(m, names); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_message_append(m, "ssbb", target, unit_dependency_to_string(dep), arg_runtime, arg_force); + if (r < 0) + return bus_log_create_error(r); + + r = sd_bus_call(bus, m, 0, &error, &reply); + if (r < 0) + return log_error_errno(r, "Failed to add dependency: %s", bus_error_message(&error, r)); + + r = bus_deserialize_and_dump_unit_file_changes(reply, arg_quiet, &changes, &n_changes); + if (r < 0) + goto finish; + + if (arg_no_reload) { + r = 0; + goto finish; + } + + r = daemon_reload(argc, argv, userdata); + } + +finish: + unit_file_changes_free(changes, n_changes); + + return r; +} + +static int preset_all(int argc, char *argv[], void *userdata) { + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + int r; + + if (install_client_side()) { + r = unit_file_preset_all(arg_scope, arg_runtime, arg_root, arg_preset_mode, arg_force, &changes, &n_changes); + unit_file_dump_changes(r, "preset", changes, n_changes, arg_quiet); + + if (r > 0) + r = 0; + } else { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + sd_bus *bus; + + polkit_agent_open_if_enabled(); + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "PresetAllUnitFiles", + &error, + &reply, + "sbb", + unit_file_preset_mode_to_string(arg_preset_mode), + arg_runtime, + arg_force); + if (r < 0) + return log_error_errno(r, "Failed to preset all units: %s", bus_error_message(&error, r)); + + r = bus_deserialize_and_dump_unit_file_changes(reply, arg_quiet, &changes, &n_changes); + if (r < 0) + goto finish; + + if (arg_no_reload) { + r = 0; + goto finish; + } + + r = daemon_reload(argc, argv, userdata); + } + +finish: + unit_file_changes_free(changes, n_changes); + + return r; +} + +static int unit_is_enabled(int argc, char *argv[], void *userdata) { + + _cleanup_strv_free_ char **names = NULL; + bool enabled; + char **name; + int r; + + r = mangle_names(strv_skip(argv, 1), &names); + if (r < 0) + return r; + + r = enable_sysv_units(argv[0], names); + if (r < 0) + return r; + + enabled = r > 0; + + if (install_client_side()) { + + STRV_FOREACH(name, names) { + UnitFileState state; + + r = unit_file_get_state(arg_scope, arg_root, *name, &state); + if (r < 0) + return log_error_errno(state, "Failed to get unit file state for %s: %m", *name); + + if (IN_SET(state, + UNIT_FILE_ENABLED, + UNIT_FILE_ENABLED_RUNTIME, + UNIT_FILE_STATIC, + UNIT_FILE_INDIRECT, + UNIT_FILE_GENERATED)) + enabled = true; + + if (!arg_quiet) + puts(unit_file_state_to_string(state)); + } + + r = 0; + } else { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + STRV_FOREACH(name, names) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + const char *s; + + r = sd_bus_call_method( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "GetUnitFileState", + &error, + &reply, + "s", *name); + if (r < 0) + return log_error_errno(r, "Failed to get unit file state for %s: %s", *name, bus_error_message(&error, r)); + + r = sd_bus_message_read(reply, "s", &s); + if (r < 0) + return bus_log_parse_error(r); + + if (STR_IN_SET(s, "enabled", "enabled-runtime", "static", "indirect", "generated")) + enabled = true; + + if (!arg_quiet) + puts(s); + } + } + + return enabled ? EXIT_SUCCESS : EXIT_FAILURE; +} + +static int is_system_running(int argc, char *argv[], void *userdata) { + _cleanup_free_ char *state = NULL; + sd_bus *bus; + int r; + + if (running_in_chroot() > 0 || (arg_transport == BUS_TRANSPORT_LOCAL && !sd_booted())) { + if (!arg_quiet) + puts("offline"); + return EXIT_FAILURE; + } + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = sd_bus_get_property_string( + bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "SystemState", + NULL, + &state); + if (r < 0) { + if (!arg_quiet) + puts("unknown"); + return 0; + } + + if (!arg_quiet) + puts(state); + + return streq(state, "running") ? EXIT_SUCCESS : EXIT_FAILURE; +} + +static int create_edit_temp_file(const char *new_path, const char *original_path, char **ret_tmp_fn) { + _cleanup_free_ char *t = NULL; + int r; + + assert(new_path); + assert(original_path); + assert(ret_tmp_fn); + + r = tempfn_random(new_path, NULL, &t); + if (r < 0) + return log_error_errno(r, "Failed to determine temporary filename for \"%s\": %m", new_path); + + r = mkdir_parents(new_path, 0755); + if (r < 0) + return log_error_errno(r, "Failed to create directories for \"%s\": %m", new_path); + + r = copy_file(original_path, t, 0, 0644, 0); + if (r == -ENOENT) { + + r = touch(t); + if (r < 0) + return log_error_errno(r, "Failed to create temporary file \"%s\": %m", t); + + } else if (r < 0) + return log_error_errno(r, "Failed to copy \"%s\" to \"%s\": %m", original_path, t); + + *ret_tmp_fn = t; + t = NULL; + + return 0; +} + +static int get_file_to_edit( + const LookupPaths *paths, + const char *name, + char **ret_path) { + + _cleanup_free_ char *path = NULL, *run = NULL; + + assert(name); + assert(ret_path); + + path = strjoin(paths->persistent_config, "/", name, NULL); + if (!path) + return log_oom(); + + if (arg_runtime) { + run = strjoin(paths->runtime_config, name, NULL); + if (!run) + return log_oom(); + } + + if (arg_runtime) { + if (access(path, F_OK) >= 0) { + log_error("Refusing to create \"%s\" because it would be overridden by \"%s\" anyway.", run, path); + return -EEXIST; + } + + *ret_path = run; + run = NULL; + } else { + *ret_path = path; + path = NULL; + } + + return 0; +} + +static int unit_file_create_dropin( + const LookupPaths *paths, + const char *unit_name, + char **ret_new_path, + char **ret_tmp_path) { + + char *tmp_new_path, *tmp_tmp_path, *ending; + int r; + + assert(unit_name); + assert(ret_new_path); + assert(ret_tmp_path); + + ending = strjoina(unit_name, ".d/override.conf"); + r = get_file_to_edit(paths, ending, &tmp_new_path); + if (r < 0) + return r; + + r = create_edit_temp_file(tmp_new_path, tmp_new_path, &tmp_tmp_path); + if (r < 0) { + free(tmp_new_path); + return r; + } + + *ret_new_path = tmp_new_path; + *ret_tmp_path = tmp_tmp_path; + + return 0; +} + +static int unit_file_create_copy( + const LookupPaths *paths, + const char *unit_name, + const char *fragment_path, + char **ret_new_path, + char **ret_tmp_path) { + + char *tmp_new_path, *tmp_tmp_path; + int r; + + assert(fragment_path); + assert(unit_name); + assert(ret_new_path); + assert(ret_tmp_path); + + r = get_file_to_edit(paths, unit_name, &tmp_new_path); + if (r < 0) + return r; + + if (!path_equal(fragment_path, tmp_new_path) && access(tmp_new_path, F_OK) == 0) { + char response; + + r = ask_char(&response, "yn", "\"%s\" already exists. Overwrite with \"%s\"? [(y)es, (n)o] ", tmp_new_path, fragment_path); + if (r < 0) { + free(tmp_new_path); + return r; + } + if (response != 'y') { + log_warning("%s ignored", unit_name); + free(tmp_new_path); + return -1; + } + } + + r = create_edit_temp_file(tmp_new_path, fragment_path, &tmp_tmp_path); + if (r < 0) { + log_error_errno(r, "Failed to create temporary file for \"%s\": %m", tmp_new_path); + free(tmp_new_path); + return r; + } + + *ret_new_path = tmp_new_path; + *ret_tmp_path = tmp_tmp_path; + + return 0; +} + +static int run_editor(char **paths) { + pid_t pid; + int r; + + assert(paths); + + pid = fork(); + if (pid < 0) + return log_error_errno(errno, "Failed to fork: %m"); + + if (pid == 0) { + const char **args; + char *editor, **editor_args = NULL; + char **tmp_path, **original_path, *p; + unsigned n_editor_args = 0, i = 1; + size_t argc; + + (void) reset_all_signal_handlers(); + (void) reset_signal_mask(); + + argc = strv_length(paths)/2 + 1; + + /* SYSTEMD_EDITOR takes precedence over EDITOR which takes precedence over VISUAL + * If neither SYSTEMD_EDITOR nor EDITOR nor VISUAL are present, + * we try to execute well known editors + */ + editor = getenv("SYSTEMD_EDITOR"); + if (!editor) + editor = getenv("EDITOR"); + if (!editor) + editor = getenv("VISUAL"); + + if (!isempty(editor)) { + editor_args = strv_split(editor, WHITESPACE); + if (!editor_args) { + (void) log_oom(); + _exit(EXIT_FAILURE); + } + n_editor_args = strv_length(editor_args); + argc += n_editor_args - 1; + } + args = newa(const char*, argc + 1); + + if (n_editor_args > 0) { + args[0] = editor_args[0]; + for (; i < n_editor_args; i++) + args[i] = editor_args[i]; + } + + STRV_FOREACH_PAIR(original_path, tmp_path, paths) { + args[i] = *tmp_path; + i++; + } + args[i] = NULL; + + if (n_editor_args > 0) + execvp(args[0], (char* const*) args); + + FOREACH_STRING(p, "editor", "nano", "vim", "vi") { + args[0] = p; + execvp(p, (char* const*) args); + /* We do not fail if the editor doesn't exist + * because we want to try each one of them before + * failing. + */ + if (errno != ENOENT) { + log_error_errno(errno, "Failed to execute %s: %m", editor); + _exit(EXIT_FAILURE); + } + } + + log_error("Cannot edit unit(s), no editor available. Please set either $SYSTEMD_EDITOR, $EDITOR or $VISUAL."); + _exit(EXIT_FAILURE); + } + + r = wait_for_terminate_and_warn("editor", pid, true); + if (r < 0) + return log_error_errno(r, "Failed to wait for child: %m"); + + return 0; +} + +static int find_paths_to_edit(sd_bus *bus, char **names, char ***paths) { + _cleanup_lookup_paths_free_ LookupPaths lp = {}; + char **name; + int r; + + assert(names); + assert(paths); + + r = lookup_paths_init(&lp, arg_scope, 0, arg_root); + if (r < 0) + return r; + + STRV_FOREACH(name, names) { + _cleanup_free_ char *path = NULL, *new_path = NULL, *tmp_path = NULL; + + r = unit_find_paths(bus, *name, &lp, &path, NULL); + if (r < 0) + return r; + else if (r == 0) + return -ENOENT; + else if (!path) { + // FIXME: support units with path==NULL (no FragmentPath) + log_error("No fragment exists for %s.", *name); + return -ENOENT; + } + + if (arg_full) + r = unit_file_create_copy(&lp, *name, path, &new_path, &tmp_path); + else + r = unit_file_create_dropin(&lp, *name, &new_path, &tmp_path); + if (r < 0) + return r; + + r = strv_push_pair(paths, new_path, tmp_path); + if (r < 0) + return log_oom(); + new_path = tmp_path = NULL; + } + + return 0; +} + +static int edit(int argc, char *argv[], void *userdata) { + _cleanup_strv_free_ char **names = NULL; + _cleanup_strv_free_ char **paths = NULL; + char **original, **tmp; + sd_bus *bus; + int r; + + if (!on_tty()) { + log_error("Cannot edit units if not on a tty."); + return -EINVAL; + } + + if (arg_transport != BUS_TRANSPORT_LOCAL) { + log_error("Cannot edit units remotely."); + return -EINVAL; + } + + r = acquire_bus(BUS_MANAGER, &bus); + if (r < 0) + return r; + + r = expand_names(bus, strv_skip(argv, 1), NULL, &names); + if (r < 0) + return log_error_errno(r, "Failed to expand names: %m"); + + r = find_paths_to_edit(bus, names, &paths); + if (r < 0) + return r; + + if (strv_isempty(paths)) + return -ENOENT; + + r = run_editor(paths); + if (r < 0) + goto end; + + STRV_FOREACH_PAIR(original, tmp, paths) { + /* If the temporary file is empty we ignore it. It's + * useful if the user wants to cancel its modification + */ + if (null_or_empty_path(*tmp)) { + log_warning("Editing \"%s\" canceled: temporary file is empty.", *original); + continue; + } + + r = rename(*tmp, *original); + if (r < 0) { + r = log_error_errno(errno, "Failed to rename \"%s\" to \"%s\": %m", *tmp, *original); + goto end; + } + } + + r = 0; + + if (!arg_no_reload && !install_client_side()) + r = daemon_reload(argc, argv, userdata); + +end: + STRV_FOREACH_PAIR(original, tmp, paths) { + (void) unlink(*tmp); + + /* Removing empty dropin dirs */ + if (!arg_full) { + _cleanup_free_ char *dir; + + dir = dirname_malloc(*original); + if (!dir) + return log_oom(); + + /* no need to check if the dir is empty, rmdir + * does nothing if it is not the case. + */ + (void) rmdir(dir); + } + } + + return r; +} + +static void systemctl_help(void) { + + pager_open(arg_no_pager, false); + + printf("%s [OPTIONS...] {COMMAND} ...\n\n" + "Query or send control commands to the systemd manager.\n\n" + " -h --help Show this help\n" + " --version Show package version\n" + " --system Connect to system manager\n" + " --user Connect to user service manager\n" + " -H --host=[USER@]HOST\n" + " Operate on remote host\n" + " -M --machine=CONTAINER\n" + " Operate on local container\n" + " -t --type=TYPE List units of a particular type\n" + " --state=STATE List units with particular LOAD or SUB or ACTIVE state\n" + " -p --property=NAME Show only properties by this name\n" + " -a --all Show all loaded units/properties, including dead/empty\n" + " ones. To list all units installed on the system, use\n" + " the 'list-unit-files' command instead.\n" + " -l --full Don't ellipsize unit names on output\n" + " -r --recursive Show unit list of host and local containers\n" + " --reverse Show reverse dependencies with 'list-dependencies'\n" + " --job-mode=MODE Specify how to deal with already queued jobs, when\n" + " queueing a new job\n" + " --show-types When showing sockets, explicitly show their type\n" + " --value When showing properties, only print the value\n" + " -i --ignore-inhibitors\n" + " When shutting down or sleeping, ignore inhibitors\n" + " --kill-who=WHO Who to send signal to\n" + " -s --signal=SIGNAL Which signal to send\n" + " --now Start or stop unit in addition to enabling or disabling it\n" + " -q --quiet Suppress output\n" + " --no-block Do not wait until operation finished\n" + " --no-wall Don't send wall message before halt/power-off/reboot\n" + " --no-reload Don't reload daemon after en-/dis-abling unit files\n" + " --no-legend Do not print a legend (column headers and hints)\n" + " --no-pager Do not pipe output into a pager\n" + " --no-ask-password\n" + " Do not ask for system passwords\n" + " --global Enable/disable unit files globally\n" + " --runtime Enable unit files only temporarily until next reboot\n" + " -f --force When enabling unit files, override existing symlinks\n" + " When shutting down, execute action immediately\n" + " --preset-mode= Apply only enable, only disable, or all presets\n" + " --root=PATH Enable unit files in the specified root directory\n" + " -n --lines=INTEGER Number of journal entries to show\n" + " -o --output=STRING Change journal output mode (short, short-iso,\n" + " short-precise, short-monotonic, verbose,\n" + " export, json, json-pretty, json-sse, cat)\n" + " --firmware-setup Tell the firmware to show the setup menu on next boot\n" + " --plain Print unit dependencies as a list instead of a tree\n\n" + "Unit Commands:\n" + " list-units [PATTERN...] List loaded units\n" + " list-sockets [PATTERN...] List loaded sockets ordered by address\n" + " list-timers [PATTERN...] List loaded timers ordered by next elapse\n" + " start NAME... Start (activate) one or more units\n" + " stop NAME... Stop (deactivate) one or more units\n" + " reload NAME... Reload one or more units\n" + " restart NAME... Start or restart one or more units\n" + " try-restart NAME... Restart one or more units if active\n" + " reload-or-restart NAME... Reload one or more units if possible,\n" + " otherwise start or restart\n" + " try-reload-or-restart NAME... If active, reload one or more units,\n" + " if supported, otherwise restart\n" + " isolate NAME Start one unit and stop all others\n" + " kill NAME... Send signal to processes of a unit\n" + " is-active PATTERN... Check whether units are active\n" + " is-failed PATTERN... Check whether units are failed\n" + " status [PATTERN...|PID...] Show runtime status of one or more units\n" + " show [PATTERN...|JOB...] Show properties of one or more\n" + " units/jobs or the manager\n" + " cat PATTERN... Show files and drop-ins of one or more units\n" + " set-property NAME ASSIGNMENT... Sets one or more properties of a unit\n" + " help PATTERN...|PID... Show manual for one or more units\n" + " reset-failed [PATTERN...] Reset failed state for all, one, or more\n" + " units\n" + " list-dependencies [NAME] Recursively show units which are required\n" + " or wanted by this unit or by which this\n" + " unit is required or wanted\n\n" + "Unit File Commands:\n" + " list-unit-files [PATTERN...] List installed unit files\n" + " enable NAME... Enable one or more unit files\n" + " disable NAME... Disable one or more unit files\n" + " reenable NAME... Reenable one or more unit files\n" + " preset NAME... Enable/disable one or more unit files\n" + " based on preset configuration\n" + " preset-all Enable/disable all unit files based on\n" + " preset configuration\n" + " is-enabled NAME... Check whether unit files are enabled\n" + " mask NAME... Mask one or more units\n" + " unmask NAME... Unmask one or more units\n" + " link PATH... Link one or more units files into\n" + " the search path\n" + " revert NAME... Revert one or more unit files to vendor\n" + " version\n" + " add-wants TARGET NAME... Add 'Wants' dependency for the target\n" + " on specified one or more units\n" + " add-requires TARGET NAME... Add 'Requires' dependency for the target\n" + " on specified one or more units\n" + " edit NAME... Edit one or more unit files\n" + " get-default Get the name of the default target\n" + " set-default NAME Set the default target\n\n" + "Machine Commands:\n" + " list-machines [PATTERN...] List local containers and host\n\n" + "Job Commands:\n" + " list-jobs [PATTERN...] List jobs\n" + " cancel [JOB...] Cancel all, one, or more jobs\n\n" + "Environment Commands:\n" + " show-environment Dump environment\n" + " set-environment NAME=VALUE... Set one or more environment variables\n" + " unset-environment NAME... Unset one or more environment variables\n" + " import-environment [NAME...] Import all or some environment variables\n\n" + "Manager Lifecycle Commands:\n" + " daemon-reload Reload systemd manager configuration\n" + " daemon-reexec Reexecute systemd manager\n\n" + "System Commands:\n" + " is-system-running Check whether system is fully running\n" + " default Enter system default mode\n" + " rescue Enter system rescue mode\n" + " emergency Enter system emergency mode\n" + " halt Shut down and halt the system\n" + " poweroff Shut down and power-off the system\n" + " reboot [ARG] Shut down and reboot the system\n" + " kexec Shut down and reboot the system with kexec\n" + " exit [EXIT_CODE] Request user instance or container exit\n" + " switch-root ROOT [INIT] Change to a different root file system\n" + " suspend Suspend the system\n" + " hibernate Hibernate the system\n" + " hybrid-sleep Hibernate and suspend the system\n", + program_invocation_short_name); +} + +static void halt_help(void) { + printf("%s [OPTIONS...]%s\n\n" + "%s the system.\n\n" + " --help Show this help\n" + " --halt Halt the machine\n" + " -p --poweroff Switch off the machine\n" + " --reboot Reboot the machine\n" + " -f --force Force immediate halt/power-off/reboot\n" + " -w --wtmp-only Don't halt/power-off/reboot, just write wtmp record\n" + " -d --no-wtmp Don't write wtmp record\n" + " --no-wall Don't send wall message before halt/power-off/reboot\n", + program_invocation_short_name, + arg_action == ACTION_REBOOT ? " [ARG]" : "", + arg_action == ACTION_REBOOT ? "Reboot" : + arg_action == ACTION_POWEROFF ? "Power off" : + "Halt"); +} + +static void shutdown_help(void) { + printf("%s [OPTIONS...] [TIME] [WALL...]\n\n" + "Shut down the system.\n\n" + " --help Show this help\n" + " -H --halt Halt the machine\n" + " -P --poweroff Power-off the machine\n" + " -r --reboot Reboot the machine\n" + " -h Equivalent to --poweroff, overridden by --halt\n" + " -k Don't halt/power-off/reboot, just send warnings\n" + " --no-wall Don't send wall message before halt/power-off/reboot\n" + " -c Cancel a pending shutdown\n", + program_invocation_short_name); +} + +static void telinit_help(void) { + printf("%s [OPTIONS...] {COMMAND}\n\n" + "Send control commands to the init daemon.\n\n" + " --help Show this help\n" + " --no-wall Don't send wall message before halt/power-off/reboot\n\n" + "Commands:\n" + " 0 Power-off the machine\n" + " 6 Reboot the machine\n" + " 2, 3, 4, 5 Start runlevelX.target unit\n" + " 1, s, S Enter rescue mode\n" + " q, Q Reload init daemon configuration\n" + " u, U Reexecute init daemon\n", + program_invocation_short_name); +} + +static void runlevel_help(void) { + printf("%s [OPTIONS...]\n\n" + "Prints the previous and current runlevel of the init system.\n\n" + " --help Show this help\n", + program_invocation_short_name); +} + +static void help_types(void) { + int i; + + if (!arg_no_legend) + puts("Available unit types:"); + for (i = 0; i < _UNIT_TYPE_MAX; i++) + puts(unit_type_to_string(i)); +} + +static void help_states(void) { + int i; + + if (!arg_no_legend) + puts("Available unit load states:"); + for (i = 0; i < _UNIT_LOAD_STATE_MAX; i++) + puts(unit_load_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable unit active states:"); + for (i = 0; i < _UNIT_ACTIVE_STATE_MAX; i++) + puts(unit_active_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable automount unit substates:"); + for (i = 0; i < _AUTOMOUNT_STATE_MAX; i++) + puts(automount_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable busname unit substates:"); + for (i = 0; i < _BUSNAME_STATE_MAX; i++) + puts(busname_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable device unit substates:"); + for (i = 0; i < _DEVICE_STATE_MAX; i++) + puts(device_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable mount unit substates:"); + for (i = 0; i < _MOUNT_STATE_MAX; i++) + puts(mount_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable path unit substates:"); + for (i = 0; i < _PATH_STATE_MAX; i++) + puts(path_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable scope unit substates:"); + for (i = 0; i < _SCOPE_STATE_MAX; i++) + puts(scope_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable service unit substates:"); + for (i = 0; i < _SERVICE_STATE_MAX; i++) + puts(service_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable slice unit substates:"); + for (i = 0; i < _SLICE_STATE_MAX; i++) + puts(slice_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable socket unit substates:"); + for (i = 0; i < _SOCKET_STATE_MAX; i++) + puts(socket_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable swap unit substates:"); + for (i = 0; i < _SWAP_STATE_MAX; i++) + puts(swap_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable target unit substates:"); + for (i = 0; i < _TARGET_STATE_MAX; i++) + puts(target_state_to_string(i)); + + if (!arg_no_legend) + puts("\nAvailable timer unit substates:"); + for (i = 0; i < _TIMER_STATE_MAX; i++) + puts(timer_state_to_string(i)); +} + +static int systemctl_parse_argv(int argc, char *argv[]) { + + enum { + ARG_FAIL = 0x100, + ARG_REVERSE, + ARG_AFTER, + ARG_BEFORE, + ARG_SHOW_TYPES, + ARG_IRREVERSIBLE, + ARG_IGNORE_DEPENDENCIES, + ARG_VALUE, + ARG_VERSION, + ARG_USER, + ARG_SYSTEM, + ARG_GLOBAL, + ARG_NO_BLOCK, + ARG_NO_LEGEND, + ARG_NO_PAGER, + ARG_NO_WALL, + ARG_ROOT, + ARG_NO_RELOAD, + ARG_KILL_WHO, + ARG_NO_ASK_PASSWORD, + ARG_FAILED, + ARG_RUNTIME, + ARG_FORCE, + ARG_PLAIN, + ARG_STATE, + ARG_JOB_MODE, + ARG_PRESET_MODE, + ARG_FIRMWARE_SETUP, + ARG_NOW, + ARG_MESSAGE, + }; + + static const struct option options[] = { + { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, + { "type", required_argument, NULL, 't' }, + { "property", required_argument, NULL, 'p' }, + { "all", no_argument, NULL, 'a' }, + { "reverse", no_argument, NULL, ARG_REVERSE }, + { "after", no_argument, NULL, ARG_AFTER }, + { "before", no_argument, NULL, ARG_BEFORE }, + { "show-types", no_argument, NULL, ARG_SHOW_TYPES }, + { "failed", no_argument, NULL, ARG_FAILED }, /* compatibility only */ + { "full", no_argument, NULL, 'l' }, + { "job-mode", required_argument, NULL, ARG_JOB_MODE }, + { "fail", no_argument, NULL, ARG_FAIL }, /* compatibility only */ + { "irreversible", no_argument, NULL, ARG_IRREVERSIBLE }, /* compatibility only */ + { "ignore-dependencies", no_argument, NULL, ARG_IGNORE_DEPENDENCIES }, /* compatibility only */ + { "ignore-inhibitors", no_argument, NULL, 'i' }, + { "value", no_argument, NULL, ARG_VALUE }, + { "user", no_argument, NULL, ARG_USER }, + { "system", no_argument, NULL, ARG_SYSTEM }, + { "global", no_argument, NULL, ARG_GLOBAL }, + { "no-block", no_argument, NULL, ARG_NO_BLOCK }, + { "no-legend", no_argument, NULL, ARG_NO_LEGEND }, + { "no-pager", no_argument, NULL, ARG_NO_PAGER }, + { "no-wall", no_argument, NULL, ARG_NO_WALL }, + { "quiet", no_argument, NULL, 'q' }, + { "root", required_argument, NULL, ARG_ROOT }, + { "force", no_argument, NULL, ARG_FORCE }, + { "no-reload", no_argument, NULL, ARG_NO_RELOAD }, + { "kill-who", required_argument, NULL, ARG_KILL_WHO }, + { "signal", required_argument, NULL, 's' }, + { "no-ask-password", no_argument, NULL, ARG_NO_ASK_PASSWORD }, + { "host", required_argument, NULL, 'H' }, + { "machine", required_argument, NULL, 'M' }, + { "runtime", no_argument, NULL, ARG_RUNTIME }, + { "lines", required_argument, NULL, 'n' }, + { "output", required_argument, NULL, 'o' }, + { "plain", no_argument, NULL, ARG_PLAIN }, + { "state", required_argument, NULL, ARG_STATE }, + { "recursive", no_argument, NULL, 'r' }, + { "preset-mode", required_argument, NULL, ARG_PRESET_MODE }, + { "firmware-setup", no_argument, NULL, ARG_FIRMWARE_SETUP }, + { "now", no_argument, NULL, ARG_NOW }, + { "message", required_argument, NULL, ARG_MESSAGE }, + {} + }; + + const char *p; + int c, r; + + assert(argc >= 0); + assert(argv); + + /* we default to allowing interactive authorization only in systemctl (not in the legacy commands) */ + arg_ask_password = true; + + while ((c = getopt_long(argc, argv, "ht:p:alqfs:H:M:n:o:ir", options, NULL)) >= 0) + + switch (c) { + + case 'h': + systemctl_help(); + return 0; + + case ARG_VERSION: + return version(); + + case 't': { + if (isempty(optarg)) { + log_error("--type requires arguments."); + return -EINVAL; + } + + p = optarg; + for (;;) { + _cleanup_free_ char *type = NULL; + + r = extract_first_word(&p, &type, ",", 0); + if (r < 0) + return log_error_errno(r, "Failed to parse type: %s", optarg); + + if (r == 0) + break; + + if (streq(type, "help")) { + help_types(); + return 0; + } + + if (unit_type_from_string(type) >= 0) { + if (strv_push(&arg_types, type) < 0) + return log_oom(); + type = NULL; + continue; + } + + /* It's much nicer to use --state= for + * load states, but let's support this + * in --types= too for compatibility + * with old versions */ + if (unit_load_state_from_string(type) >= 0) { + if (strv_push(&arg_states, type) < 0) + return log_oom(); + type = NULL; + continue; + } + + log_error("Unknown unit type or load state '%s'.", type); + log_info("Use -t help to see a list of allowed values."); + return -EINVAL; + } + + break; + } + + case 'p': { + /* Make sure that if the empty property list + was specified, we won't show any properties. */ + if (isempty(optarg) && !arg_properties) { + arg_properties = new0(char*, 1); + if (!arg_properties) + return log_oom(); + } else { + p = optarg; + for (;;) { + _cleanup_free_ char *prop = NULL; + + r = extract_first_word(&p, &prop, ",", 0); + if (r < 0) + return log_error_errno(r, "Failed to parse property: %s", optarg); + + if (r == 0) + break; + + if (strv_push(&arg_properties, prop) < 0) + return log_oom(); + + prop = NULL; + } + } + + /* If the user asked for a particular + * property, show it to him, even if it is + * empty. */ + arg_all = true; + + break; + } + + case 'a': + arg_all = true; + break; + + case ARG_REVERSE: + arg_dependency = DEPENDENCY_REVERSE; + break; + + case ARG_AFTER: + arg_dependency = DEPENDENCY_AFTER; + break; + + case ARG_BEFORE: + arg_dependency = DEPENDENCY_BEFORE; + break; + + case ARG_SHOW_TYPES: + arg_show_types = true; + break; + + case ARG_VALUE: + arg_value = true; + break; + + case ARG_JOB_MODE: + arg_job_mode = optarg; + break; + + case ARG_FAIL: + arg_job_mode = "fail"; + break; + + case ARG_IRREVERSIBLE: + arg_job_mode = "replace-irreversibly"; + break; + + case ARG_IGNORE_DEPENDENCIES: + arg_job_mode = "ignore-dependencies"; + break; + + case ARG_USER: + arg_scope = UNIT_FILE_USER; + break; + + case ARG_SYSTEM: + arg_scope = UNIT_FILE_SYSTEM; + break; + + case ARG_GLOBAL: + arg_scope = UNIT_FILE_GLOBAL; + break; + + case ARG_NO_BLOCK: + arg_no_block = true; + break; + + case ARG_NO_LEGEND: + arg_no_legend = true; + break; + + case ARG_NO_PAGER: + arg_no_pager = true; + break; + + case ARG_NO_WALL: + arg_no_wall = true; + break; + + case ARG_ROOT: + r = parse_path_argument_and_warn(optarg, false, &arg_root); + if (r < 0) + return r; + break; + + case 'l': + arg_full = true; + break; + + case ARG_FAILED: + if (strv_extend(&arg_states, "failed") < 0) + return log_oom(); + + break; + + case 'q': + arg_quiet = true; + break; + + case ARG_FORCE: + arg_force++; + break; + + case 'f': + arg_force++; + break; + + case ARG_NO_RELOAD: + arg_no_reload = true; + break; + + case ARG_KILL_WHO: + arg_kill_who = optarg; + break; + + case 's': + arg_signal = signal_from_string_try_harder(optarg); + if (arg_signal < 0) { + log_error("Failed to parse signal string %s.", optarg); + return -EINVAL; + } + break; + + case ARG_NO_ASK_PASSWORD: + arg_ask_password = false; + break; + + case 'H': + arg_transport = BUS_TRANSPORT_REMOTE; + arg_host = optarg; + break; + + case 'M': + arg_transport = BUS_TRANSPORT_MACHINE; + arg_host = optarg; + break; + + case ARG_RUNTIME: + arg_runtime = true; + break; + + case 'n': + if (safe_atou(optarg, &arg_lines) < 0) { + log_error("Failed to parse lines '%s'", optarg); + return -EINVAL; + } + break; + + case 'o': + arg_output = output_mode_from_string(optarg); + if (arg_output < 0) { + log_error("Unknown output '%s'.", optarg); + return -EINVAL; + } + break; + + case 'i': + arg_ignore_inhibitors = true; + break; + + case ARG_PLAIN: + arg_plain = true; + break; + + case ARG_FIRMWARE_SETUP: + arg_firmware_setup = true; + break; + + case ARG_STATE: { + if (isempty(optarg)) { + log_error("--signal requires arguments."); + return -EINVAL; + } + + p = optarg; + for (;;) { + _cleanup_free_ char *s = NULL; + + r = extract_first_word(&p, &s, ",", 0); + if (r < 0) + return log_error_errno(r, "Failed to parse signal: %s", optarg); + + if (r == 0) + break; + + if (streq(s, "help")) { + help_states(); + return 0; + } + + if (strv_push(&arg_states, s) < 0) + return log_oom(); + + s = NULL; + } + break; + } + + case 'r': + if (geteuid() != 0) { + log_error("--recursive requires root privileges."); + return -EPERM; + } + + arg_recursive = true; + break; + + case ARG_PRESET_MODE: + + arg_preset_mode = unit_file_preset_mode_from_string(optarg); + if (arg_preset_mode < 0) { + log_error("Failed to parse preset mode: %s.", optarg); + return -EINVAL; + } + + break; + + case ARG_NOW: + arg_now = true; + break; + + case ARG_MESSAGE: + if (strv_extend(&arg_wall, optarg) < 0) + return log_oom(); + break; + + case '?': + return -EINVAL; + + default: + assert_not_reached("Unhandled option"); + } + + if (arg_transport != BUS_TRANSPORT_LOCAL && arg_scope != UNIT_FILE_SYSTEM) { + log_error("Cannot access user instance remotely."); + return -EINVAL; + } + + return 1; +} + +static int halt_parse_argv(int argc, char *argv[]) { + + enum { + ARG_HELP = 0x100, + ARG_HALT, + ARG_REBOOT, + ARG_NO_WALL + }; + + static const struct option options[] = { + { "help", no_argument, NULL, ARG_HELP }, + { "halt", no_argument, NULL, ARG_HALT }, + { "poweroff", no_argument, NULL, 'p' }, + { "reboot", no_argument, NULL, ARG_REBOOT }, + { "force", no_argument, NULL, 'f' }, + { "wtmp-only", no_argument, NULL, 'w' }, + { "no-wtmp", no_argument, NULL, 'd' }, + { "no-sync", no_argument, NULL, 'n' }, + { "no-wall", no_argument, NULL, ARG_NO_WALL }, + {} + }; + + int c, r, runlevel; + + assert(argc >= 0); + assert(argv); + + if (utmp_get_runlevel(&runlevel, NULL) >= 0) + if (runlevel == '0' || runlevel == '6') + arg_force = 2; + + while ((c = getopt_long(argc, argv, "pfwdnih", options, NULL)) >= 0) + switch (c) { + + case ARG_HELP: + halt_help(); + return 0; + + case ARG_HALT: + arg_action = ACTION_HALT; + break; + + case 'p': + if (arg_action != ACTION_REBOOT) + arg_action = ACTION_POWEROFF; + break; + + case ARG_REBOOT: + arg_action = ACTION_REBOOT; + break; + + case 'f': + arg_force = 2; + break; + + case 'w': + arg_dry = true; + break; + + case 'd': + arg_no_wtmp = true; + break; + + case 'n': + arg_no_sync = true; + break; + + case ARG_NO_WALL: + arg_no_wall = true; + break; + + case 'i': + case 'h': + /* Compatibility nops */ + break; + + case '?': + return -EINVAL; + + default: + assert_not_reached("Unhandled option"); + } + + if (arg_action == ACTION_REBOOT && (argc == optind || argc == optind + 1)) { + r = update_reboot_parameter_and_warn(argc == optind + 1 ? argv[optind] : NULL); + if (r < 0) + return r; + } else if (optind < argc) { + log_error("Too many arguments."); + return -EINVAL; + } + + return 1; +} + +static int parse_shutdown_time_spec(const char *t, usec_t *_u) { + assert(t); + assert(_u); + + if (streq(t, "now")) + *_u = 0; + else if (!strchr(t, ':')) { + uint64_t u; + + if (safe_atou64(t, &u) < 0) + return -EINVAL; + + *_u = now(CLOCK_REALTIME) + USEC_PER_MINUTE * u; + } else { + char *e = NULL; + long hour, minute; + struct tm tm = {}; + time_t s; + usec_t n; + + errno = 0; + hour = strtol(t, &e, 10); + if (errno > 0 || *e != ':' || hour < 0 || hour > 23) + return -EINVAL; + + minute = strtol(e+1, &e, 10); + if (errno > 0 || *e != 0 || minute < 0 || minute > 59) + return -EINVAL; + + n = now(CLOCK_REALTIME); + s = (time_t) (n / USEC_PER_SEC); + + assert_se(localtime_r(&s, &tm)); + + tm.tm_hour = (int) hour; + tm.tm_min = (int) minute; + tm.tm_sec = 0; + + assert_se(s = mktime(&tm)); + + *_u = (usec_t) s * USEC_PER_SEC; + + while (*_u <= n) + *_u += USEC_PER_DAY; + } + + return 0; +} + +static int shutdown_parse_argv(int argc, char *argv[]) { + + enum { + ARG_HELP = 0x100, + ARG_NO_WALL + }; + + static const struct option options[] = { + { "help", no_argument, NULL, ARG_HELP }, + { "halt", no_argument, NULL, 'H' }, + { "poweroff", no_argument, NULL, 'P' }, + { "reboot", no_argument, NULL, 'r' }, + { "kexec", no_argument, NULL, 'K' }, /* not documented extension */ + { "no-wall", no_argument, NULL, ARG_NO_WALL }, + {} + }; + + char **wall = NULL; + int c, r; + + assert(argc >= 0); + assert(argv); + + while ((c = getopt_long(argc, argv, "HPrhkKtafFc", options, NULL)) >= 0) + switch (c) { + + case ARG_HELP: + shutdown_help(); + return 0; + + case 'H': + arg_action = ACTION_HALT; + break; + + case 'P': + arg_action = ACTION_POWEROFF; + break; + + case 'r': + if (kexec_loaded()) + arg_action = ACTION_KEXEC; + else + arg_action = ACTION_REBOOT; + break; + + case 'K': + arg_action = ACTION_KEXEC; + break; + + case 'h': + if (arg_action != ACTION_HALT) + arg_action = ACTION_POWEROFF; + break; + + case 'k': + arg_dry = true; + break; + + case ARG_NO_WALL: + arg_no_wall = true; + break; + + case 't': + case 'a': + case 'f': + case 'F': + /* Compatibility nops */ + break; + + case 'c': + arg_action = ACTION_CANCEL_SHUTDOWN; + break; + + case '?': + return -EINVAL; + + default: + assert_not_reached("Unhandled option"); + } + + if (argc > optind && arg_action != ACTION_CANCEL_SHUTDOWN) { + r = parse_shutdown_time_spec(argv[optind], &arg_when); + if (r < 0) { + log_error("Failed to parse time specification: %s", argv[optind]); + return r; + } + } else + arg_when = now(CLOCK_REALTIME) + USEC_PER_MINUTE; + + if (argc > optind && arg_action == ACTION_CANCEL_SHUTDOWN) + /* No time argument for shutdown cancel */ + wall = argv + optind; + else if (argc > optind + 1) + /* We skip the time argument */ + wall = argv + optind + 1; + + if (wall) { + arg_wall = strv_copy(wall); + if (!arg_wall) + return log_oom(); + } + + optind = argc; + + return 1; +} + +static int telinit_parse_argv(int argc, char *argv[]) { + + enum { + ARG_HELP = 0x100, + ARG_NO_WALL + }; + + static const struct option options[] = { + { "help", no_argument, NULL, ARG_HELP }, + { "no-wall", no_argument, NULL, ARG_NO_WALL }, + {} + }; + + static const struct { + char from; + enum action to; + } table[] = { + { '0', ACTION_POWEROFF }, + { '6', ACTION_REBOOT }, + { '1', ACTION_RESCUE }, + { '2', ACTION_RUNLEVEL2 }, + { '3', ACTION_RUNLEVEL3 }, + { '4', ACTION_RUNLEVEL4 }, + { '5', ACTION_RUNLEVEL5 }, + { 's', ACTION_RESCUE }, + { 'S', ACTION_RESCUE }, + { 'q', ACTION_RELOAD }, + { 'Q', ACTION_RELOAD }, + { 'u', ACTION_REEXEC }, + { 'U', ACTION_REEXEC } + }; + + unsigned i; + int c; + + assert(argc >= 0); + assert(argv); + + while ((c = getopt_long(argc, argv, "", options, NULL)) >= 0) + switch (c) { + + case ARG_HELP: + telinit_help(); + return 0; + + case ARG_NO_WALL: + arg_no_wall = true; + break; + + case '?': + return -EINVAL; + + default: + assert_not_reached("Unhandled option"); + } + + if (optind >= argc) { + log_error("%s: required argument missing.", program_invocation_short_name); + return -EINVAL; + } + + if (optind + 1 < argc) { + log_error("Too many arguments."); + return -EINVAL; + } + + if (strlen(argv[optind]) != 1) { + log_error("Expected single character argument."); + return -EINVAL; + } + + for (i = 0; i < ELEMENTSOF(table); i++) + if (table[i].from == argv[optind][0]) + break; + + if (i >= ELEMENTSOF(table)) { + log_error("Unknown command '%s'.", argv[optind]); + return -EINVAL; + } + + arg_action = table[i].to; + + optind++; + + return 1; +} + +static int runlevel_parse_argv(int argc, char *argv[]) { + + enum { + ARG_HELP = 0x100, + }; + + static const struct option options[] = { + { "help", no_argument, NULL, ARG_HELP }, + {} + }; + + int c; + + assert(argc >= 0); + assert(argv); + + while ((c = getopt_long(argc, argv, "", options, NULL)) >= 0) + switch (c) { + + case ARG_HELP: + runlevel_help(); + return 0; + + case '?': + return -EINVAL; + + default: + assert_not_reached("Unhandled option"); + } + + if (optind < argc) { + log_error("Too many arguments."); + return -EINVAL; + } + + return 1; +} + +static int parse_argv(int argc, char *argv[]) { + assert(argc >= 0); + assert(argv); + + if (program_invocation_short_name) { + + if (strstr(program_invocation_short_name, "halt")) { + arg_action = ACTION_HALT; + return halt_parse_argv(argc, argv); + } else if (strstr(program_invocation_short_name, "poweroff")) { + arg_action = ACTION_POWEROFF; + return halt_parse_argv(argc, argv); + } else if (strstr(program_invocation_short_name, "reboot")) { + if (kexec_loaded()) + arg_action = ACTION_KEXEC; + else + arg_action = ACTION_REBOOT; + return halt_parse_argv(argc, argv); + } else if (strstr(program_invocation_short_name, "shutdown")) { + arg_action = ACTION_POWEROFF; + return shutdown_parse_argv(argc, argv); + } else if (strstr(program_invocation_short_name, "init")) { + + if (sd_booted() > 0) { + arg_action = _ACTION_INVALID; + return telinit_parse_argv(argc, argv); + } else { + /* Hmm, so some other init system is + * running, we need to forward this + * request to it. For now we simply + * guess that it is Upstart. */ + + execv(TELINIT, argv); + + log_error("Couldn't find an alternative telinit implementation to spawn."); + return -EIO; + } + + } else if (strstr(program_invocation_short_name, "runlevel")) { + arg_action = ACTION_RUNLEVEL; + return runlevel_parse_argv(argc, argv); + } + } + + arg_action = ACTION_SYSTEMCTL; + return systemctl_parse_argv(argc, argv); +} + +#ifdef HAVE_SYSV_COMPAT +_pure_ static int action_to_runlevel(void) { + + static const char table[_ACTION_MAX] = { + [ACTION_HALT] = '0', + [ACTION_POWEROFF] = '0', + [ACTION_REBOOT] = '6', + [ACTION_RUNLEVEL2] = '2', + [ACTION_RUNLEVEL3] = '3', + [ACTION_RUNLEVEL4] = '4', + [ACTION_RUNLEVEL5] = '5', + [ACTION_RESCUE] = '1' + }; + + assert(arg_action < _ACTION_MAX); + + return table[arg_action]; +} +#endif + +static int talk_initctl(void) { +#ifdef HAVE_SYSV_COMPAT + struct init_request request = { + .magic = INIT_MAGIC, + .sleeptime = 0, + .cmd = INIT_CMD_RUNLVL + }; + + _cleanup_close_ int fd = -1; + char rl; + int r; + + rl = action_to_runlevel(); + if (!rl) + return 0; + + request.runlevel = rl; + + fd = open(INIT_FIFO, O_WRONLY|O_NDELAY|O_CLOEXEC|O_NOCTTY); + if (fd < 0) { + if (errno == ENOENT) + return 0; + + return log_error_errno(errno, "Failed to open "INIT_FIFO": %m"); + } + + r = loop_write(fd, &request, sizeof(request), false); + if (r < 0) + return log_error_errno(r, "Failed to write to "INIT_FIFO": %m"); + + return 1; +#else + return 0; +#endif +} + +static int systemctl_main(int argc, char *argv[]) { + + static const Verb verbs[] = { + { "list-units", VERB_ANY, VERB_ANY, VERB_DEFAULT|VERB_NOCHROOT, list_units }, + { "list-unit-files", VERB_ANY, VERB_ANY, 0, list_unit_files }, + { "list-sockets", VERB_ANY, VERB_ANY, VERB_NOCHROOT, list_sockets }, + { "list-timers", VERB_ANY, VERB_ANY, VERB_NOCHROOT, list_timers }, + { "list-jobs", VERB_ANY, VERB_ANY, VERB_NOCHROOT, list_jobs }, + { "list-machines", VERB_ANY, VERB_ANY, VERB_NOCHROOT, list_machines }, + { "clear-jobs", VERB_ANY, 1, VERB_NOCHROOT, daemon_reload }, + { "cancel", VERB_ANY, VERB_ANY, VERB_NOCHROOT, cancel_job }, + { "start", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "stop", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "condstop", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, /* For compatibility with ALTLinux */ + { "reload", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "restart", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "try-restart", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "reload-or-restart", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "reload-or-try-restart", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, /* For compatbility with old systemctl <= 228 */ + { "try-reload-or-restart", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, + { "force-reload", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, /* For compatibility with SysV */ + { "condreload", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, /* For compatibility with ALTLinux */ + { "condrestart", 2, VERB_ANY, VERB_NOCHROOT, start_unit }, /* For compatibility with RH */ + { "isolate", 2, 2, VERB_NOCHROOT, start_unit }, + { "kill", 2, VERB_ANY, VERB_NOCHROOT, kill_unit }, + { "is-active", 2, VERB_ANY, VERB_NOCHROOT, check_unit_active }, + { "check", 2, VERB_ANY, VERB_NOCHROOT, check_unit_active }, + { "is-failed", 2, VERB_ANY, VERB_NOCHROOT, check_unit_failed }, + { "show", VERB_ANY, VERB_ANY, VERB_NOCHROOT, show }, + { "cat", 2, VERB_ANY, VERB_NOCHROOT, cat }, + { "status", VERB_ANY, VERB_ANY, VERB_NOCHROOT, show }, + { "help", VERB_ANY, VERB_ANY, VERB_NOCHROOT, show }, + { "daemon-reload", VERB_ANY, 1, VERB_NOCHROOT, daemon_reload }, + { "daemon-reexec", VERB_ANY, 1, VERB_NOCHROOT, daemon_reload }, + { "show-environment", VERB_ANY, 1, VERB_NOCHROOT, show_environment }, + { "set-environment", 2, VERB_ANY, VERB_NOCHROOT, set_environment }, + { "unset-environment", 2, VERB_ANY, VERB_NOCHROOT, set_environment }, + { "import-environment", VERB_ANY, VERB_ANY, VERB_NOCHROOT, import_environment}, + { "halt", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "poweroff", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "reboot", VERB_ANY, 2, VERB_NOCHROOT, start_special }, + { "kexec", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "suspend", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "hibernate", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "hybrid-sleep", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "default", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "rescue", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "emergency", VERB_ANY, 1, VERB_NOCHROOT, start_special }, + { "exit", VERB_ANY, 2, VERB_NOCHROOT, start_special }, + { "reset-failed", VERB_ANY, VERB_ANY, VERB_NOCHROOT, reset_failed }, + { "enable", 2, VERB_ANY, 0, enable_unit }, + { "disable", 2, VERB_ANY, 0, enable_unit }, + { "is-enabled", 2, VERB_ANY, 0, unit_is_enabled }, + { "reenable", 2, VERB_ANY, 0, enable_unit }, + { "preset", 2, VERB_ANY, 0, enable_unit }, + { "preset-all", VERB_ANY, 1, 0, preset_all }, + { "mask", 2, VERB_ANY, 0, enable_unit }, + { "unmask", 2, VERB_ANY, 0, enable_unit }, + { "link", 2, VERB_ANY, 0, enable_unit }, + { "revert", 2, VERB_ANY, 0, enable_unit }, + { "switch-root", 2, VERB_ANY, VERB_NOCHROOT, switch_root }, + { "list-dependencies", VERB_ANY, 2, VERB_NOCHROOT, list_dependencies }, + { "set-default", 2, 2, 0, set_default }, + { "get-default", VERB_ANY, 1, 0, get_default, }, + { "set-property", 3, VERB_ANY, VERB_NOCHROOT, set_property }, + { "is-system-running", VERB_ANY, 1, 0, is_system_running }, + { "add-wants", 3, VERB_ANY, 0, add_dependency }, + { "add-requires", 3, VERB_ANY, 0, add_dependency }, + { "edit", 2, VERB_ANY, VERB_NOCHROOT, edit }, + {} + }; + + return dispatch_verb(argc, argv, verbs, NULL); +} + +static int reload_with_fallback(void) { + + /* First, try systemd via D-Bus. */ + if (daemon_reload(0, NULL, NULL) >= 0) + return 0; + + /* Nothing else worked, so let's try signals */ + assert(arg_action == ACTION_RELOAD || arg_action == ACTION_REEXEC); + + if (kill(1, arg_action == ACTION_RELOAD ? SIGHUP : SIGTERM) < 0) + return log_error_errno(errno, "kill() failed: %m"); + + return 0; +} + +static int start_with_fallback(void) { + + /* First, try systemd via D-Bus. */ + if (start_unit(0, NULL, NULL) >= 0) + return 0; + + /* Nothing else worked, so let's try + * /dev/initctl */ + if (talk_initctl() > 0) + return 0; + + log_error("Failed to talk to init daemon."); + return -EIO; +} + +static int halt_now(enum action a) { + int r; + + /* The kernel will automaticall flush ATA disks and suchlike + * on reboot(), but the file systems need to be synce'd + * explicitly in advance. */ + if (!arg_no_sync) + (void) sync(); + + /* Make sure C-A-D is handled by the kernel from this point + * on... */ + (void) reboot(RB_ENABLE_CAD); + + switch (a) { + + case ACTION_HALT: + log_info("Halting."); + (void) reboot(RB_HALT_SYSTEM); + return -errno; + + case ACTION_POWEROFF: + log_info("Powering off."); + (void) reboot(RB_POWER_OFF); + return -errno; + + case ACTION_KEXEC: + case ACTION_REBOOT: { + _cleanup_free_ char *param = NULL; + + r = read_one_line_file("/run/systemd/reboot-param", ¶m); + if (r < 0) + log_warning_errno(r, "Failed to read reboot parameter file: %m"); + + if (!isempty(param)) { + log_info("Rebooting with argument '%s'.", param); + (void) syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, LINUX_REBOOT_CMD_RESTART2, param); + log_warning_errno(errno, "Failed to reboot with parameter, retrying without: %m"); + } + + log_info("Rebooting."); + (void) reboot(RB_AUTOBOOT); + return -errno; + } + + default: + assert_not_reached("Unknown action."); + } +} + +static int logind_schedule_shutdown(void) { + +#ifdef HAVE_LOGIND + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + char date[FORMAT_TIMESTAMP_MAX]; + const char *action; + sd_bus *bus; + int r; + + (void) logind_set_wall_message(); + + r = acquire_bus(BUS_FULL, &bus); + if (r < 0) + return r; + + switch (arg_action) { + case ACTION_HALT: + action = "halt"; + break; + case ACTION_POWEROFF: + action = "poweroff"; + break; + case ACTION_KEXEC: + action = "kexec"; + break; + case ACTION_EXIT: + action = "exit"; + break; + case ACTION_REBOOT: + default: + action = "reboot"; + break; + } + + if (arg_dry) + action = strjoina("dry-", action); + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "ScheduleShutdown", + &error, + NULL, + "st", + action, + arg_when); + if (r < 0) + return log_warning_errno(r, "Failed to call ScheduleShutdown in logind, proceeding with immediate shutdown: %s", bus_error_message(&error, r)); + + log_info("Shutdown scheduled for %s, use 'shutdown -c' to cancel.", format_timestamp(date, sizeof(date), arg_when)); + return 0; +#else + log_error("Cannot schedule shutdown without logind support, proceeding with immediate shutdown."); + return -ENOSYS; +#endif +} + +static int halt_main(void) { + int r; + + r = logind_check_inhibitors(arg_action); + if (r < 0) + return r; + + if (arg_when > 0) + return logind_schedule_shutdown(); + + if (geteuid() != 0) { + if (arg_dry || arg_force > 0) { + log_error("Must be root."); + return -EPERM; + } + + /* Try logind if we are a normal user and no special + * mode applies. Maybe PolicyKit allows us to shutdown + * the machine. */ + if (IN_SET(arg_action, ACTION_POWEROFF, ACTION_REBOOT)) { + r = logind_reboot(arg_action); + if (r >= 0) + return r; + if (IN_SET(r, -EOPNOTSUPP, -EINPROGRESS)) + /* requested operation is not + * supported on the local system or + * already in progress */ + return r; + /* on all other errors, try low-level operation */ + } + } + + if (!arg_dry && !arg_force) + return start_with_fallback(); + + assert(geteuid() == 0); + + if (!arg_no_wtmp) { + if (sd_booted() > 0) + log_debug("Not writing utmp record, assuming that systemd-update-utmp is used."); + else { + r = utmp_put_shutdown(); + if (r < 0) + log_warning_errno(r, "Failed to write utmp record: %m"); + } + } + + if (arg_dry) + return 0; + + r = halt_now(arg_action); + return log_error_errno(r, "Failed to reboot: %m"); +} + +static int runlevel_main(void) { + int r, runlevel, previous; + + r = utmp_get_runlevel(&runlevel, &previous); + if (r < 0) { + puts("unknown"); + return r; + } + + printf("%c %c\n", + previous <= 0 ? 'N' : previous, + runlevel <= 0 ? 'N' : runlevel); + + return 0; +} + +static int logind_cancel_shutdown(void) { +#ifdef HAVE_LOGIND + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + sd_bus *bus; + int r; + + r = acquire_bus(BUS_FULL, &bus); + if (r < 0) + return r; + + (void) logind_set_wall_message(); + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "CancelScheduledShutdown", + &error, + NULL, NULL); + if (r < 0) + return log_warning_errno(r, "Failed to talk to logind, shutdown hasn't been cancelled: %s", bus_error_message(&error, r)); + + return 0; +#else + log_error("Not compiled with logind support, cannot cancel scheduled shutdowns."); + return -ENOSYS; +#endif +} + +int main(int argc, char*argv[]) { + int r; + + setlocale(LC_ALL, ""); + log_parse_environment(); + log_open(); + + /* Explicitly not on_tty() to avoid setting cached value. + * This becomes relevant for piping output which might be + * ellipsized. */ + original_stdout_is_tty = isatty(STDOUT_FILENO); + + r = parse_argv(argc, argv); + if (r <= 0) + goto finish; + + if (arg_action != ACTION_SYSTEMCTL && running_in_chroot() > 0) { + log_info("Running in chroot, ignoring request."); + r = 0; + goto finish; + } + + /* systemctl_main() will print an error message for the bus + * connection, but only if it needs to */ + + switch (arg_action) { + + case ACTION_SYSTEMCTL: + r = systemctl_main(argc, argv); + break; + + case ACTION_HALT: + case ACTION_POWEROFF: + case ACTION_REBOOT: + case ACTION_KEXEC: + r = halt_main(); + break; + + case ACTION_RUNLEVEL2: + case ACTION_RUNLEVEL3: + case ACTION_RUNLEVEL4: + case ACTION_RUNLEVEL5: + case ACTION_RESCUE: + case ACTION_EMERGENCY: + case ACTION_DEFAULT: + r = start_with_fallback(); + break; + + case ACTION_RELOAD: + case ACTION_REEXEC: + r = reload_with_fallback(); + break; + + case ACTION_CANCEL_SHUTDOWN: + r = logind_cancel_shutdown(); + break; + + case ACTION_RUNLEVEL: + r = runlevel_main(); + break; + + case _ACTION_INVALID: + default: + assert_not_reached("Unknown action"); + } + +finish: + pager_close(); + ask_password_agent_close(); + polkit_agent_close(); + + strv_free(arg_types); + strv_free(arg_states); + strv_free(arg_properties); + + strv_free(arg_wall); + free(arg_root); + + release_busses(); + + /* Note that we return r here, not EXIT_SUCCESS, so that we can implement the LSB-like return codes */ + + return r < 0 ? EXIT_FAILURE : r; +} diff --git a/src/grp-system/systemctl/systemctl.completion.bash.in b/src/grp-system/systemctl/systemctl.completion.bash.in new file mode 100644 index 0000000000..6f2b3f122c --- /dev/null +++ b/src/grp-system/systemctl/systemctl.completion.bash.in @@ -0,0 +1,284 @@ +# systemctl(1) completion -*- shell-script -*- +# +# This file is part of systemd. +# +# Copyright 2010 Ran Benita +# +# systemd is free software; you can redistribute it and/or modify it +# 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 +# (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 +# General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with systemd; If not, see <http://www.gnu.org/licenses/>. + +__systemctl() { + local mode=$1; shift 1 + systemctl $mode --full --no-legend "$@" +} + +__systemd_properties() { + local mode=$1 + { __systemctl $mode show --all; + @rootlibexecdir@/systemd --dump-configuration-items; } | + while IFS='=' read -r key value; do + [[ $value ]] && echo "$key" + done +} + +__contains_word () { + local w word=$1; shift + for w in "$@"; do + [[ $w = "$word" ]] && return + done +} + +__filter_units_by_property () { + local mode=$1 property=$2 value=$3 ; shift 3 + local units=("$@") + local props + IFS=$'\n' read -rd '' -a props < \ + <(__systemctl $mode show --property "$property" -- "${units[@]}") + for ((i=0; $i < ${#units[*]}; i++)); do + if [[ "${props[i]}" = "$property=$value" ]]; then + echo " ${units[i]}" + fi + done +} + +__get_all_units () { { __systemctl $1 list-unit-files; __systemctl $1 list-units --all; } \ + | { while read -r a b; do [[ $a =~ @\. ]] || echo " $a"; done; }; } +__get_template_names () { __systemctl $1 list-unit-files \ + | { while read -r a b; do [[ $a =~ @\. ]] && echo " ${a%%@.*}@"; done; }; } + +__get_active_units () { __systemctl $1 list-units \ + | { while read -r a b; do echo " $a"; done; }; } +__get_startable_units () { + # find startable inactive units + __filter_units_by_property $mode ActiveState inactive $( + __filter_units_by_property $mode CanStart yes $( + __systemctl $mode list-unit-files --state enabled,disabled,static | \ + { while read -r a b; do [[ $a =~ @\. ]] || echo " $a"; done; } + __systemctl $mode list-units --state inactive,failed | \ + { while read -r a b; do echo " $a"; done; } )) +} +__get_restartable_units () { + # filter out masked and not-found + __filter_units_by_property $mode CanStart yes $( + __systemctl $mode list-unit-files --state enabled,disabled,static | \ + { while read -r a b; do [[ $a =~ @\. ]] || echo " $a"; done; } + __systemctl $mode list-units | \ + { while read -r a b; do echo " $a"; done; } ) +} +__get_failed_units () { __systemctl $1 list-units \ + | { while read -r a b c d; do [[ $c == "failed" ]] && echo " $a"; done; }; } +__get_enabled_units () { __systemctl $1 list-unit-files \ + | { while read -r a b c ; do [[ $b == "enabled" ]] && echo " $a"; done; }; } +__get_disabled_units () { __systemctl $1 list-unit-files \ + | { while read -r a b c ; do [[ $b == "disabled" ]] && echo " $a"; done; }; } +__get_masked_units () { __systemctl $1 list-unit-files \ + | { while read -r a b c ; do [[ $b == "masked" ]] && echo " $a"; done; }; } +__get_all_unit_files () { { __systemctl $1 list-unit-files; } | { while read -r a b; do echo " $a"; done; }; } + +__get_machines() { + local a b + { machinectl list-images --no-legend --no-pager; machinectl list --no-legend --no-pager; } | \ + { while read a b; do echo " $a"; done; } +} + +_systemctl () { + local cur=${COMP_WORDS[COMP_CWORD]} prev=${COMP_WORDS[COMP_CWORD-1]} + local i verb comps mode + + local -A OPTS=( + [STANDALONE]='--all -a --reverse --after --before --defaults --force -f --full -l --global + --help -h --no-ask-password --no-block --no-legend --no-pager --no-reload --no-wall + --quiet -q --privileged -P --system --user --version --runtime --recursive -r --firmware-setup + --show-types -i --ignore-inhibitors --plain' + [ARG]='--host -H --kill-who --property -p --signal -s --type -t --state --job-mode --root + --preset-mode -n --lines -o --output -M --machine' + ) + + if __contains_word "--user" ${COMP_WORDS[*]}; then + mode=--user + elif __contains_word "--global" ${COMP_WORDS[*]}; then + mode=--user + else + mode=--system + fi + + if __contains_word "$prev" ${OPTS[ARG]}; then + case $prev in + --signal|-s) + _signals + return + ;; + --type|-t) + comps=$(__systemctl $mode -t help) + ;; + --state) + comps=$(__systemctl $mode --state=help) + ;; + --job-mode) + comps='fail replace replace-irreversibly isolate + ignore-dependencies ignore-requirements flush' + ;; + --kill-who) + comps='all control main' + ;; + --root) + comps=$(compgen -A directory -- "$cur" ) + compopt -o filenames + ;; + --host|-H) + comps=$(compgen -A hostname) + ;; + --property|-p) + comps=$(__systemd_properties $mode) + ;; + --preset-mode) + comps='full enable-only disable-only' + ;; + --output|-o) + comps='short short-iso short-precise short-monotonic verbose export json + json-pretty json-sse cat' + ;; + --machine|-M) + comps=$( __get_machines ) + ;; + esac + COMPREPLY=( $(compgen -W '$comps' -- "$cur") ) + return 0 + fi + + if [[ "$cur" = -* ]]; then + COMPREPLY=( $(compgen -W '${OPTS[*]}' -- "$cur") ) + return 0 + fi + + local -A VERBS=( + [ALL_UNITS]='is-active is-failed is-enabled status show cat mask preset help list-dependencies edit set-property' + [ENABLED_UNITS]='disable' + [DISABLED_UNITS]='enable' + [REENABLABLE_UNITS]='reenable' + [FAILED_UNITS]='reset-failed' + [STARTABLE_UNITS]='start' + [STOPPABLE_UNITS]='stop condstop kill try-restart condrestart' + [ISOLATABLE_UNITS]='isolate' + [RELOADABLE_UNITS]='reload condreload try-reload-or-restart force-reload' + [RESTARTABLE_UNITS]='restart reload-or-restart' + [TARGET_AND_UNITS]='add-wants add-requires' + [MASKED_UNITS]='unmask' + [JOBS]='cancel' + [ENVS]='set-environment unset-environment' + [STANDALONE]='daemon-reexec daemon-reload default + emergency exit halt hibernate hybrid-sleep kexec list-jobs + list-sockets list-timers list-units list-unit-files poweroff + reboot rescue show-environment suspend get-default + is-system-running' + [FILE]='link switch-root' + [TARGETS]='set-default' + ) + + for ((i=0; i < COMP_CWORD; i++)); do + if __contains_word "${COMP_WORDS[i]}" ${VERBS[*]} && + ! __contains_word "${COMP_WORDS[i-1]}" ${OPTS[ARG]}; then + verb=${COMP_WORDS[i]} + break + fi + done + + if [[ -z $verb ]]; then + comps="${VERBS[*]}" + + elif __contains_word "$verb" ${VERBS[ALL_UNITS]}; then + comps=$( __get_all_units $mode ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[ENABLED_UNITS]}; then + comps=$( __get_enabled_units $mode ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[DISABLED_UNITS]}; then + comps=$( __get_disabled_units $mode; + __get_template_names $mode) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[REENABLABLE_UNITS]}; then + comps=$( __get_disabled_units $mode; + __get_enabled_units $mode; + __get_template_names $mode) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[STARTABLE_UNITS]}; then + comps=$( __get_startable_units $mode; + __get_template_names $mode) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[RESTARTABLE_UNITS]}; then + comps=$( __get_restartable_units $mode; + __get_template_names $mode) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[STOPPABLE_UNITS]}; then + comps=$( __filter_units_by_property $mode CanStop yes \ + $( __get_active_units $mode ) ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[RELOADABLE_UNITS]}; then + comps=$( __filter_units_by_property $mode CanReload yes \ + $( __get_active_units $mode ) ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[ISOLATABLE_UNITS]}; then + comps=$( __filter_units_by_property $mode AllowIsolate yes \ + $( __get_all_units $mode ) ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[FAILED_UNITS]}; then + comps=$( __get_failed_units $mode ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[MASKED_UNITS]}; then + comps=$( __get_masked_units $mode ) + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[TARGET_AND_UNITS]}; then + if __contains_word "$prev" ${VERBS[TARGET_AND_UNITS]} \ + || __contains_word "$prev" ${OPTS[STANDALONE]}; then + comps=$( __systemctl $mode list-unit-files --type target --all \ + | { while read -r a b; do echo " $a"; done; } ) + else + comps=$( __get_all_unit_files $mode ) + fi + compopt -o filenames + + elif __contains_word "$verb" ${VERBS[STANDALONE]}; then + comps='' + + elif __contains_word "$verb" ${VERBS[JOBS]}; then + comps=$( __systemctl $mode list-jobs | { while read -r a b; do echo " $a"; done; } ) + + elif __contains_word "$verb" ${VERBS[ENVS]}; then + comps=$( __systemctl $mode show-environment \ + | while read -r line; do echo " ${line%%=*}=";done ) + compopt -o nospace + + elif __contains_word "$verb" ${VERBS[FILE]}; then + comps=$( compgen -A file -- "$cur" ) + compopt -o filenames + elif __contains_word "$verb" ${VERBS[TARGETS]}; then + comps=$( __systemctl $mode list-unit-files --type target --full --all \ + | { while read -r a b; do echo " $a"; done; } ) + fi + + COMPREPLY=( $(compgen -o filenames -W '$comps' -- "$cur") ) + return 0 +} + +complete -F _systemctl systemctl diff --git a/src/grp-system/systemctl/systemctl.completion.zsh.in b/src/grp-system/systemctl/systemctl.completion.zsh.in new file mode 100644 index 0000000000..44c31b7833 --- /dev/null +++ b/src/grp-system/systemctl/systemctl.completion.zsh.in @@ -0,0 +1,391 @@ +#compdef systemctl + +(( $+functions[_systemctl_command] )) || _systemctl_command() +{ + local -a _systemctl_cmds + _systemctl_cmds=( + "list-sockets:List sockets" + "list-timers:List timers" + "list-units:List units" + "start:Start (activate) one or more units" + "stop:Stop (deactivate) one or more units" + "reload:Reload one or more units" + "restart:Start or restart one or more units" + "condrestart:Restart one or more units if active" + "try-restart:Restart one or more units if active" + "reload-or-restart:Reload one or more units if possible, otherwise start or restart" + "force-reload:Reload one or more units if possible, otherwise restart if active" + "hibernate:Hibernate the system" + "hybrid-sleep:Hibernate and suspend the system" + "try-reload-or-restart:Reload one or more units if possible, otherwise restart if active" + "isolate:Start one unit and stop all others" + "kill:Send signal to processes of a unit" + "is-active:Check whether units are active" + "is-failed:Check whether units are failed" + "status:Show runtime status of one or more units" + "show:Show properties of one or more units/jobs or the manager" + "cat:Show the source unit files and drop-ins" + "reset-failed:Reset failed state for all, one, or more units" + "list-unit-files:List installed unit files" + "enable:Enable one or more unit files" + "disable:Disable one or more unit files" + "reenable:Reenable one or more unit files" + "preset:Enable/disable one or more unit files based on preset configuration" + "set-default:Set the default target" + "get-default:Query the default target" + "edit:Edit one or more unit files" + "is-system-running:Query overall status of the system" + "help:Show documentation for specified units" + "list-dependencies:Show unit dependency tree" + "mask:Mask one or more units" + "unmask:Unmask one or more units" + "link:Link one or more units files into the search path" + "is-enabled:Check whether unit files are enabled" + "list-jobs:List jobs" + "cancel:Cancel all, one, or more jobs" + "show-environment:Dump environment" + "set-environment:Set one or more environment variables" + "unset-environment:Unset one or more environment variables" + "daemon-reload:Reload systemd manager configuration" + "daemon-reexec:Reexecute systemd manager" + "default:Enter system default mode" + "rescue:Enter system rescue mode" + "emergency:Enter system emergency mode" + "halt:Shut down and halt the system" + "suspend:Suspend the system" + "poweroff:Shut down and power-off the system" + "reboot:Shut down and reboot the system" + "kexec:Shut down and reboot the system with kexec" + "exit:Ask for user instance termination" + "switch-root:Change root directory" + ) + + if (( CURRENT == 1 )); then + _describe -t commands 'systemctl command' _systemctl_cmds || compadd "$@" + else + local curcontext="$curcontext" expl + + cmd="${${_systemctl_cmds[(r)$words[1]:*]%%:*}}" + # Deal with any aliases + case $cmd in + condrestart) cmd="try-restart";; + force-reload) cmd="try-reload-or-restart";; + esac + + if (( $#cmd )); then + curcontext="${curcontext%:*:*}:systemctl-${cmd}:" + + local update_policy + zstyle -s ":completion:${curcontext}:" cache-policy update_policy + if [[ -z "$update_policy" ]]; then + zstyle ":completion:${curcontext}:" cache-policy _systemctl_caching_policy + fi + + _call_function ret _systemctl_$cmd || _message 'no more arguments' + else + _message "unknown systemctl command: $words[1]" + fi + return ret + fi +} + +__systemctl() +{ + systemctl $_sys_service_mgr --full --no-legend --no-pager "$@" +} + + +# Fills the unit list +_systemctl_all_units() +{ + if ( [[ ${+_sys_all_units} -eq 0 ]] || _cache_invalid SYS_ALL_UNITS ) && + ! _retrieve_cache SYS_ALL_UNITS; + then + _sys_all_units=( ${${(f)"$(__systemctl list-units --all)"}%% *} ) + _store_cache SYS_ALL_UNITS _sys_all_units + fi +} + +# Fills the unit list including all file units +_systemctl_really_all_units() +{ + local -a all_unit_files; + local -a really_all_units; + if ( [[ ${+_sys_really_all_units} -eq 0 ]] || _cache_invalid SYS_REALLY_ALL_UNITS ) && + ! _retrieve_cache SYS_REALLY_ALL_UNITS; + then + all_unit_files=( ${${(f)"$(__systemctl list-unit-files)"}%% *} ) + _systemctl_all_units + really_all_units=($_sys_all_units $all_unit_files) + _sys_really_all_units=(${(u)really_all_units}) + _store_cache SYS_REALLY_ALL_UNITS _sys_really_all_units + fi +} + +_filter_units_by_property() { + local property=$1 value=$2 ; shift ; shift + local -a units ; units=($*) + local props + for props in ${(ps:\n\n:)"$(_call_program units "$service show --no-pager --property="Id,$property" -- ${units} 2>/dev/null")"}; do + props=(${(f)props}) + if [[ "${props[2]}" = "$property=$value" ]]; then + echo -E - " ${props[1]#Id=}" + fi + done +} + +_systemctl_get_template_names() { echo -E - ${^${(M)${(f)"$(__systemctl list-unit-files)"}##*@.[^[:space:]]##}%%@.*}\@ } + + +_systemctl_active_units() {_sys_active_units=( ${${(f)"$(__systemctl list-units)"}%% *} )} + +_systemctl_startable_units(){ + _sys_startable_units=( $( _filter_units_by_property ActiveState inactive $( + _filter_units_by_property CanStart yes $( + __systemctl $mode list-unit-files --state enabled,disabled,static | \ + { while read -r a b; do [[ $a =~ @\. ]] || echo -E - " $a"; done; } + __systemctl $mode list-units --state inactive,failed | \ + { while read -r a b; do echo -E - " $a"; done; } )) ) ) +} + +_systemctl_restartable_units(){ + _sys_restartable_units=( $(_filter_units_by_property CanStart yes $( + __systemctl $mode list-unit-files --state enabled,disabled,static | \ + { while read -r a b; do [[ $a =~ @\. ]] || echo -E - " $a"; done; } + __systemctl $mode list-units | \ + { while read -r a b; do echo -E - " $a"; done; } )) ) +} + +_systemctl_failed_units() {_sys_failed_units=( ${${(f)"$(__systemctl list-units --state=failed)"}%% *} ) } +_systemctl_unit_state() { typeset -gA _sys_unit_state; _sys_unit_state=( $(__systemctl list-unit-files) ) } + +local fun +# Completion functions for ALL_UNITS +for fun in is-active is-failed is-enabled status show cat mask preset help list-dependencies edit ; do + (( $+functions[_systemctl_$fun] )) || _systemctl_$fun() + { + _systemctl_really_all_units + _wanted systemd-units expl unit \ + compadd "$@" -a - _sys_really_all_units + } +done + +# Completion functions for ENABLED_UNITS +(( $+functions[_systemctl_disable] )) || _systemctl_disable() +{ + local _sys_unit_state; _systemctl_unit_state + _wanted systemd-units expl 'enabled unit' \ + compadd "$@" - ${(k)_sys_unit_state[(R)enabled]} +} + +(( $+functions[_systemctl_reenable] )) || _systemctl_reenable() +{ + local _sys_unit_state; _systemctl_unit_state + _wanted systemd-units expl 'enabled/disabled unit' \ + compadd "$@" - ${(k)_sys_unit_state[(R)(enabled|disabled)]} $(_systemctl_get_template_names) +} + +# Completion functions for DISABLED_UNITS +(( $+functions[_systemctl_enable] )) || _systemctl_enable() +{ + local _sys_unit_state; _systemctl_unit_state + _wanted systemd-units expl 'disabled unit' \ + compadd "$@" - ${(k)_sys_unit_state[(R)disabled]} $(_systemctl_get_template_names) +} + +# Completion functions for FAILED_UNITS +(( $+functions[_systemctl_reset-failed] )) || _systemctl_reset-failed() +{ + local _sys_failed_units; _systemctl_failed_units + _wanted systemd-units expl 'failed unit' \ + compadd "$@" -a - _sys_failed_units || _message "no failed unit found" +} + +# Completion functions for STARTABLE_UNITS +(( $+functions[_systemctl_start] )) || _systemctl_start() +{ + local _sys_startable_units; _systemctl_startable_units + _wanted systemd-units expl 'startable unit' \ + compadd "$@" - ${_sys_startable_units[*]} $(_systemctl_get_template_names) +} + +# Completion functions for STOPPABLE_UNITS +for fun in stop kill try-restart condrestart ; do + (( $+functions[_systemctl_$fun] )) || _systemctl_$fun() + { + local _sys_active_units; _systemctl_active_units + _wanted systemd-units expl 'stoppable unit' \ + compadd "$@" - $( _filter_units_by_property CanStop yes \ + ${_sys_active_units[*]} ) + } +done + +# Completion functions for ISOLATABLE_UNITS +(( $+functions[_systemctl_isolate] )) || _systemctl_isolate() +{ + _systemctl_all_units + _wanted systemd-units expl 'isolatable unit' \ + compadd "$@" - $( _filter_units_by_property AllowIsolate yes \ + ${_sys_all_units[*]} ) +} + +# Completion functions for RELOADABLE_UNITS +for fun in reload try-reload-or-restart force-reload ; do + (( $+functions[_systemctl_$fun] )) || _systemctl_$fun() + { + local _sys_active_units; _systemctl_active_units + _wanted systemd-units expl 'reloadable unit' \ + compadd "$@" - $( _filter_units_by_property CanReload yes \ + ${_sys_active_units[*]} ) + } +done + +# Completion functions for RESTARTABLE_UNITS +for fun in restart reload-or-restart ; do + (( $+functions[_systemctl_$fun] )) || _systemctl_$fun() + { + local _sys_restartable_units; _systemctl_restartable_units + _wanted systemd-units expl 'restartable unit' \ + compadd "$@" - ${_sys_restartable_units[*]} $(_systemctl_get_template_names) + } +done + +# Completion functions for MASKED_UNITS +(( $+functions[_systemctl_unmask] )) || _systemctl_unmask() +{ + local _sys_unit_state; _systemctl_unit_state + _wanted systemd-units expl 'masked unit' \ + compadd "$@" - ${(k)_sys_unit_state[(R)masked]} || _message "no masked units found" +} + +# Completion functions for JOBS +(( $+functions[_systemctl_cancel] )) || _systemctl_cancel() +{ + _wanted systemd-jobs expl job \ + compadd "$@" - ${${(f)"$(__systemctl list-jobs)"}%% *} || + _message "no jobs found" +} + +# Completion functions for TARGETS +(( $+functions[_systemctl_set-default] )) || _systemctl_set-default() +{ + _wanted systemd-targets expl target \ + compadd "$@" - ${${(f)"$(__systemctl list-unit-files --type target --all)"}%% *} || + _message "no targets found" +} + +# Completion functions for ENVS +for fun in set-environment unset-environment ; do + (( $+functions[_systemctl_$fun] )) || _systemctl_$fun() + { + local fun=$0 ; fun=${fun##_systemctl_} + local suf + if [[ "${fun}" = "set-environment" ]]; then + suf='-S=' + fi + _wanted systemd-environment expl 'environment variable' \ + compadd "$@" ${suf} - ${${(f)"$(systemctl show-environment)"}%%=*} + } +done + +(( $+functions[_systemctl_link] )) || _systemctl_link() { + _sd_unit_files +} + +(( $+functions[_systemctl_switch-root] )) || _systemctl_switch-root() { + _files +} + +# no systemctl completion for: +# [STANDALONE]='daemon-reexec daemon-reload default +# emergency exit halt kexec list-jobs list-units +# list-unit-files poweroff reboot rescue show-environment' + +_systemctl_caching_policy() +{ + local _sysunits + local -a oldcache + + # rebuild if cache is more than a day old + oldcache=( "$1"(mh+1) ) + (( $#oldcache )) && return 0 + + _sysunits=(${${(f)"$(__systemctl --all)"}%% *}) + + if (( $#_sysunits )); then + for unit in $_sysunits; do + [[ "$unit" -nt "$1" ]] && return 0 + done + fi + + return 1 +} + +_unit_states() { + local -a _states + _states=("${(fo)$(__systemctl --state=help)}") + _values -s , "${_states[@]}" +} + +_unit_types() { + local -a _types + _types=("${(fo)$(__systemctl -t help)}") + _values -s , "${_types[@]}" +} + +_unit_properties() { + if ( [[ ${+_sys_all_properties} -eq 0 ]] || _cache_invalid SYS_ALL_PROPERTIES ) && + ! _retrieve_cache SYS_ALL_PROPERTIES; + then + _sys_all_properties=( ${${(M)${(f)"$(__systemctl show --all; + @rootlibexecdir@/systemd --dump-configuration-items)"}##[[:alnum:]]##=*}%%=*} + ) + _store_cache SYS_ALL_PROPRTIES _sys_all_properties + fi + _values -s , "${_sys_all_properties[@]}" +} + +_job_modes() { + local -a _modes + _modes=(fail replace replace-irreversibly isolate ignore-dependencies ignore-requirements flush) + _values -s , "${_modes[@]}" +} + +local -a _modes; _modes=("--user" "--system") +local _sys_service_mgr=${${words:*_modes}[(R)(${(j.|.)_modes})]:---system} +_arguments -s \ + {-h,--help}'[Show help]' \ + '--version[Show package version]' \ + {-t+,--type=}'[List only units of a particular type]:unit type:_unit_types' \ + '--state=[Display units in the specified state]:unit state:_unit_states' \ + '--job-mode=[Specify how to deal with other jobs]:mode:_job_modes' \ + {-p+,--property=}'[Show only properties by specific name]:unit property:_unit_properties' \ + {-a,--all}'[Show all units/properties, including dead/empty ones]' \ + '--reverse[Show reverse dependencies]' \ + '--after[Show units ordered after]' \ + '--before[Show units ordered before]' \ + {-l,--full}"[Don't ellipsize unit names on output]" \ + '--show-types[When showing sockets, show socket type]' \ + {-i,--ignore-inhibitors}'[When executing a job, ignore jobs dependencies]' \ + {-q,--quiet}'[Suppress output]' \ + '--no-block[Do not wait until operation finished]' \ + '--no-legend[Do not print a legend, i.e. the column headers and the footer with hints]' \ + '--no-pager[Do not pipe output into a pager]' \ + '--system[Connect to system manager]' \ + '--user[Connect to user service manager]' \ + "--no-wall[Don't send wall message before halt/power-off/reboot]" \ + '--global[Enable/disable unit files globally]' \ + "--no-reload[When enabling/disabling unit files, don't reload daemon configuration]" \ + '--no-ask-password[Do not ask for system passwords]' \ + '--kill-who=[Who to send signal to]:killwho:(main control all)' \ + {-s+,--signal=}'[Which signal to send]:signal:_signals' \ + {-f,--force}'[When enabling unit files, override existing symlinks. When shutting down, execute action immediately]' \ + '--root=[Enable unit files in the specified root directory]:directory:_directories' \ + '--runtime[Enable unit files only temporarily until next reboot]' \ + {-H+,--host=}'[Operate on remote host]:userathost:_sd_hosts_or_user_at_host' \ + {-P,--privileged}'[Acquire privileges before execution]' \ + {-n+,--lines=}'[Journal entries to show]:number of entries' \ + {-o+,--output=}'[Change journal output mode]:modes:_sd_outputmodes' \ + '--firmware-setup[Tell the firmware to show the setup menu on next boot]' \ + '--plain[When used with list-dependencies, print output as a list]' \ + '*::systemctl command:_systemctl_command' diff --git a/src/grp-system/systemctl/systemctl.xml b/src/grp-system/systemctl/systemctl.xml new file mode 100644 index 0000000000..991e9bafaf --- /dev/null +++ b/src/grp-system/systemctl/systemctl.xml @@ -0,0 +1,1838 @@ +<?xml version='1.0'?> <!--*- Mode: nxml; nxml-child-indent: 2; indent-tabs-mode: nil -*--> +<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" +"http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd" [ +<!ENTITY % entities SYSTEM "custom-entities.ent" > +%entities; +]> + +<!-- + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + 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 + (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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +--> + +<refentry id="systemctl" + xmlns:xi="http://www.w3.org/2001/XInclude"> + + <refentryinfo> + <title>systemctl</title> + <productname>systemd</productname> + + <authorgroup> + <author> + <contrib>Developer</contrib> + <firstname>Lennart</firstname> + <surname>Poettering</surname> + <email>lennart@poettering.net</email> + </author> + </authorgroup> + </refentryinfo> + + <refmeta> + <refentrytitle>systemctl</refentrytitle> + <manvolnum>1</manvolnum> + </refmeta> + + <refnamediv> + <refname>systemctl</refname> + <refpurpose>Control the systemd system and service manager</refpurpose> + </refnamediv> + + <refsynopsisdiv> + <cmdsynopsis> + <command>systemctl</command> + <arg choice="opt" rep="repeat">OPTIONS</arg> + <arg choice="plain">COMMAND</arg> + <arg choice="opt" rep="repeat">NAME</arg> + </cmdsynopsis> + </refsynopsisdiv> + + <refsect1> + <title>Description</title> + + <para><command>systemctl</command> may be used to introspect and + control the state of the <literal>systemd</literal> system and + service manager. Please refer to + <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry> + for an introduction into the basic concepts and functionality this + tool manages.</para> + </refsect1> + + <refsect1> + <title>Options</title> + + <para>The following options are understood:</para> + + <variablelist> + <varlistentry> + <term><option>-t</option></term> + <term><option>--type=</option></term> + + <listitem> + <para>The argument should be a comma-separated list of unit + types such as <option>service</option> and + <option>socket</option>. + </para> + + <para>If one of the arguments is a unit type, when listing + units, limit display to certain unit types. Otherwise, units + of all types will be shown.</para> + + <para>As a special case, if one of the arguments is + <option>help</option>, a list of allowed values will be + printed and the program will exit.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--state=</option></term> + + <listitem> + <para>The argument should be a comma-separated list of unit + LOAD, SUB, or ACTIVE states. When listing units, show only + those in the specified states. Use <option>--state=failed</option> + to show only failed units.</para> + + <para>As a special case, if one of the arguments is + <option>help</option>, a list of allowed values will be + printed and the program will exit.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-p</option></term> + <term><option>--property=</option></term> + + <listitem> + <para>When showing unit/job/manager properties with the + <command>show</command> command, limit display to properties + specified in the argument. The argument should be a + comma-separated list of property names, such as + <literal>MainPID</literal>. Unless specified, all known + properties are shown. If specified more than once, all + properties with the specified names are shown. Shell + completion is implemented for property names.</para> + + <para>For the manager itself, + <command>systemctl show</command> will show all available + properties. Those properties are documented in + <citerefentry><refentrytitle>systemd-system.conf</refentrytitle><manvolnum>5</manvolnum></citerefentry>. + </para> + + <para>Properties for units vary by unit type, so showing any + unit (even a non-existent one) is a way to list properties + pertaining to this type. Similarly, showing any job will list + properties pertaining to all jobs. Properties for units are + documented in + <citerefentry><refentrytitle>systemd.unit</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + and the pages for individual unit types + <citerefentry><refentrytitle>systemd.service</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.socket</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + etc.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-a</option></term> + <term><option>--all</option></term> + + <listitem> + <para>When listing units, show all loaded units, regardless + of their state, including inactive units. When showing + unit/job/manager properties, show all properties regardless + whether they are set or not.</para> + <para>To list all units installed on the system, use the + <command>list-unit-files</command> command instead.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-r</option></term> + <term><option>--recursive</option></term> + + <listitem> + <para>When listing units, also show units of local + containers. Units of local containers will be prefixed with + the container name, separated by a single colon character + (<literal>:</literal>).</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--reverse</option></term> + + <listitem> + <para>Show reverse dependencies between units with + <command>list-dependencies</command>, i.e. follow + dependencies of type <varname>WantedBy=</varname>, + <varname>RequiredBy=</varname>, + <varname>PartOf=</varname>, <varname>BoundBy=</varname>, + instead of <varname>Wants=</varname> and similar. + </para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--after</option></term> + + <listitem> + <para>With <command>list-dependencies</command>, show the + units that are ordered before the specified unit. In other + words, recursively list units following the + <varname>After=</varname> dependency.</para> + + <para>Note that any <varname>After=</varname> dependency is + automatically mirrored to create a + <varname>Before=</varname> dependency. Temporal dependencies + may be specified explicitly, but are also created implicitly + for units which are <varname>WantedBy=</varname> targets + (see + <citerefentry><refentrytitle>systemd.target</refentrytitle><manvolnum>5</manvolnum></citerefentry>), + and as a result of other directives (for example + <varname>RequiresMountsFor=</varname>). Both explicitly + and implicitly introduced dependencies are shown with + <command>list-dependencies</command>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--before</option></term> + + <listitem> + <para>With <command>list-dependencies</command>, show the + units that are ordered after the specified unit. In other + words, recursively list units following the + <varname>Before=</varname> dependency.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-l</option></term> + <term><option>--full</option></term> + + <listitem> + <para>Do not ellipsize unit names, process tree entries, + journal output, or truncate unit descriptions in the output + of <command>status</command>, <command>list-units</command>, + <command>list-jobs</command>, and + <command>list-timers</command>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--value</option></term> + + <listitem> + <para>When printing properties with <command>show</command>, + only print the value, and skip the property name and + <literal>=</literal>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--show-types</option></term> + + <listitem> + <para>When showing sockets, show the type of the socket.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--job-mode=</option></term> + + <listitem> + <para>When queuing a new job, this option controls how to deal with + already queued jobs. It takes one of <literal>fail</literal>, + <literal>replace</literal>, + <literal>replace-irreversibly</literal>, + <literal>isolate</literal>, + <literal>ignore-dependencies</literal>, + <literal>ignore-requirements</literal> or + <literal>flush</literal>. Defaults to + <literal>replace</literal>, except when the + <command>isolate</command> command is used which implies the + <literal>isolate</literal> job mode.</para> + + <para>If <literal>fail</literal> is specified and a requested + operation conflicts with a pending job (more specifically: + causes an already pending start job to be reversed into a stop + job or vice versa), cause the operation to fail.</para> + + <para>If <literal>replace</literal> (the default) is + specified, any conflicting pending job will be replaced, as + necessary.</para> + + <para>If <literal>replace-irreversibly</literal> is specified, + operate like <literal>replace</literal>, but also mark the new + jobs as irreversible. This prevents future conflicting + transactions from replacing these jobs (or even being enqueued + while the irreversible jobs are still pending). Irreversible + jobs can still be cancelled using the <command>cancel</command> + command.</para> + + <para><literal>isolate</literal> is only valid for start + operations and causes all other units to be stopped when the + specified unit is started. This mode is always used when the + <command>isolate</command> command is used.</para> + + <para><literal>flush</literal> will cause all queued jobs to + be canceled when the new job is enqueued.</para> + + <para>If <literal>ignore-dependencies</literal> is specified, + then all unit dependencies are ignored for this new job and + the operation is executed immediately. If passed, no required + units of the unit passed will be pulled in, and no ordering + dependencies will be honored. This is mostly a debugging and + rescue tool for the administrator and should not be used by + applications.</para> + + <para><literal>ignore-requirements</literal> is similar to + <literal>ignore-dependencies</literal>, but only causes the + requirement dependencies to be ignored, the ordering + dependencies will still be honoured.</para> + </listitem> + + </varlistentry> + + <varlistentry> + <term><option>--fail</option></term> + + <listitem> + <para>Shorthand for <option>--job-mode=</option>fail.</para> + <para>When used with the <command>kill</command> command, + if no units were killed, the operation results in an error. + </para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-i</option></term> + <term><option>--ignore-inhibitors</option></term> + + <listitem> + <para>When system shutdown or a sleep state is requested, + ignore inhibitor locks. Applications can establish inhibitor + locks to avoid that certain important operations (such as CD + burning or suchlike) are interrupted by system shutdown or a + sleep state. Any user may take these locks and privileged + users may override these locks. If any locks are taken, + shutdown and sleep state requests will normally fail + (regardless of whether privileged or not) and a list of active locks + is printed. However, if <option>--ignore-inhibitors</option> + is specified, the locks are ignored and not printed, and the + operation attempted anyway, possibly requiring additional + privileges.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-q</option></term> + <term><option>--quiet</option></term> + + <listitem> + <para>Suppress printing of the results of various commands + and also the hints about truncated log lines. This does not + suppress output of commands for which the printed output is + the only result (like <command>show</command>). Errors are + always printed.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--no-block</option></term> + + <listitem> + <para>Do not synchronously wait for the requested operation + to finish. If this is not specified, the job will be + verified, enqueued and <command>systemctl</command> will + wait until the unit's start-up is completed. By passing this + argument, it is only verified and enqueued.</para> + </listitem> + </varlistentry> + + <xi:include href="user-system-options.xml" xpointer="user" /> + <xi:include href="user-system-options.xml" xpointer="system" /> + + <!-- we do not document -failed here, as it has been made + redundant by -state=failed, which it predates. To keep + things simple, we only document the new switch, while + keeping the old one around for compatibility only. --> + + <varlistentry> + <term><option>--no-wall</option></term> + + <listitem> + <para>Do not send wall message before halt, power-off, + reboot.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--global</option></term> + + <listitem> + <para>When used with <command>enable</command> and + <command>disable</command>, operate on the global user + configuration directory, thus enabling or disabling a unit + file globally for all future logins of all users.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--no-reload</option></term> + + <listitem> + <para>When used with <command>enable</command> and + <command>disable</command>, do not implicitly reload daemon + configuration after executing the changes.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--no-ask-password</option></term> + + <listitem> + <para>When used with <command>start</command> and related + commands, disables asking for passwords. Background services + may require input of a password or passphrase string, for + example to unlock system hard disks or cryptographic + certificates. Unless this option is specified and the + command is invoked from a terminal, + <command>systemctl</command> will query the user on the + terminal for the necessary secrets. Use this option to + switch this behavior off. In this case, the password must be + supplied by some other means (for example graphical password + agents) or the service might fail. This also disables + querying the user for authentication for privileged + operations.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--kill-who=</option></term> + + <listitem> + <para>When used with <command>kill</command>, choose which + processes to send a signal to. Must be one of + <option>main</option>, <option>control</option> or + <option>all</option> to select whether to kill only the main + process, the control process or all processes of the + unit. The main process of the unit is the one that defines + the life-time of it. A control process of a unit is one that + is invoked by the manager to induce state changes of it. For + example, all processes started due to the + <varname>ExecStartPre=</varname>, + <varname>ExecStop=</varname> or + <varname>ExecReload=</varname> settings of service units are + control processes. Note that there is only one control + process per unit at a time, as only one state change is + executed at a time. For services of type + <varname>Type=forking</varname>, the initial process started + by the manager for <varname>ExecStart=</varname> is a + control process, while the process ultimately forked off by + that one is then considered the main process of the unit (if + it can be determined). This is different for service units + of other types, where the process forked off by the manager + for <varname>ExecStart=</varname> is always the main process + itself. A service unit consists of zero or one main process, + zero or one control process plus any number of additional + processes. Not all unit types manage processes of these + types however. For example, for mount units, control processes + are defined (which are the invocations of + <filename>&MOUNT_PATH;</filename> and + <filename>&UMOUNT_PATH;</filename>), but no main process + is defined. If omitted, defaults to + <option>all</option>.</para> + </listitem> + + </varlistentry> + + <varlistentry> + <term><option>-s</option></term> + <term><option>--signal=</option></term> + + <listitem> + <para>When used with <command>kill</command>, choose which + signal to send to selected processes. Must be one of the + well-known signal specifiers such as <constant>SIGTERM</constant>, <constant>SIGINT</constant> or + <constant>SIGSTOP</constant>. If omitted, defaults to + <option>SIGTERM</option>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-f</option></term> + <term><option>--force</option></term> + + <listitem> + <para>When used with <command>enable</command>, overwrite + any existing conflicting symlinks.</para> + + <para>When used with <command>halt</command>, + <command>poweroff</command>, <command>reboot</command> or + <command>kexec</command>, execute the selected operation + without shutting down all units. However, all processes will + be killed forcibly and all file systems are unmounted or + remounted read-only. This is hence a drastic but relatively + safe option to request an immediate reboot. If + <option>--force</option> is specified twice for these + operations, they will be executed immediately without + terminating any processes or unmounting any file + systems. Warning: specifying <option>--force</option> twice + with any of these operations might result in data + loss.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--message=</option></term> + + <listitem> + <para>When used with <command>halt</command>, + <command>poweroff</command>, <command>reboot</command> or + <command>kexec</command>, set a short message explaining the reason + for the operation. The message will be logged together with the + default shutdown message.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--now</option></term> + + <listitem> + <para>When used with <command>enable</command>, the units + will also be started. When used with <command>disable</command> or + <command>mask</command>, the units will also be stopped. The start + or stop operation is only carried out when the respective enable or + disable operation has been successful.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--root=</option></term> + + <listitem> + <para>When used with + <command>enable</command>/<command>disable</command>/<command>is-enabled</command> + (and related commands), use an alternate root path when + looking for unit files.</para> + </listitem> + + </varlistentry> + + <varlistentry> + <term><option>--runtime</option></term> + + <listitem> + <para>When used with <command>enable</command>, + <command>disable</command>, <command>edit</command>, + (and related commands), make changes only temporarily, so + that they are lost on the next reboot. This will have the + effect that changes are not made in subdirectories of + <filename>/etc</filename> but in <filename>/run</filename>, + with identical immediate effects, however, since the latter + is lost on reboot, the changes are lost too.</para> + + <para>Similarly, when used with + <command>set-property</command>, make changes only + temporarily, so that they are lost on the next + reboot.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--preset-mode=</option></term> + + <listitem> + <para>Takes one of <literal>full</literal> (the default), + <literal>enable-only</literal>, + <literal>disable-only</literal>. When used with the + <command>preset</command> or <command>preset-all</command> + commands, controls whether units shall be disabled and + enabled according to the preset rules, or only enabled, or + only disabled.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-n</option></term> + <term><option>--lines=</option></term> + + <listitem> + <para>When used with <command>status</command>, controls the + number of journal lines to show, counting from the most + recent ones. Takes a positive integer argument. Defaults to + 10.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>-o</option></term> + <term><option>--output=</option></term> + + <listitem> + <para>When used with <command>status</command>, controls the + formatting of the journal entries that are shown. For the + available choices, see + <citerefentry><refentrytitle>journalctl</refentrytitle><manvolnum>1</manvolnum></citerefentry>. + Defaults to <literal>short</literal>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--firmware-setup</option></term> + + <listitem> + <para>When used with the <command>reboot</command> command, + indicate to the system's firmware to boot into setup + mode. Note that this is currently only supported on some EFI + systems and only if the system was booted in EFI + mode.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><option>--plain</option></term> + + <listitem> + <para>When used with <command>list-dependencies</command>, + <command>list-units</command> or <command>list-machines</command>, the + the output is printed as a list instead of a tree, and the bullet + circles are omitted.</para> + </listitem> + </varlistentry> + + <xi:include href="user-system-options.xml" xpointer="host" /> + <xi:include href="user-system-options.xml" xpointer="machine" /> + + <xi:include href="standard-options.xml" xpointer="no-pager" /> + <xi:include href="standard-options.xml" xpointer="no-legend" /> + <xi:include href="standard-options.xml" xpointer="help" /> + <xi:include href="standard-options.xml" xpointer="version" /> + </variablelist> + </refsect1> + + <refsect1> + <title>Commands</title> + + <para>The following commands are understood:</para> + + <refsect2> + <title>Unit Commands</title> + + <variablelist> + <varlistentry> + <term><command>list-units <optional><replaceable>PATTERN</replaceable>...</optional></command></term> + + <listitem> + <para>List known units (subject to limitations specified + with <option>-t</option>). If one or more + <replaceable>PATTERN</replaceable>s are specified, only + units matching one of them are shown.</para> + + <para>This is the default command.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>list-sockets <optional><replaceable>PATTERN</replaceable>...</optional></command></term> + + <listitem> + <para>List socket units ordered by listening address. + If one or more <replaceable>PATTERN</replaceable>s are + specified, only socket units matching one of them are + shown. Produces output similar to + <programlisting> +LISTEN UNIT ACTIVATES +/dev/initctl systemd-initctl.socket systemd-initctl.service +... +[::]:22 sshd.socket sshd.service +kobject-uevent 1 systemd-udevd-kernel.socket systemd-udevd.service + +5 sockets listed.</programlisting> + Note: because the addresses might contains spaces, this output + is not suitable for programmatic consumption. + </para> + + <para>See also the options <option>--show-types</option>, + <option>--all</option>, and <option>--state=</option>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>list-timers <optional><replaceable>PATTERN</replaceable>...</optional></command></term> + + <listitem> + <para>List timer units ordered by the time they elapse + next. If one or more <replaceable>PATTERN</replaceable>s + are specified, only units matching one of them are shown. + </para> + + <para>See also the options <option>--all</option> and + <option>--state=</option>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>start <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Start (activate) one or more units specified on the + command line.</para> + + <para>Note that glob patterns operate on the set of primary names of currently loaded units. Units which + are not active and are not in a failed state usually are not loaded, and will not be matched by any + pattern. In addition, in case of instantiated units, systemd is often unaware of the instance name until + the instance has been started. Therefore, using glob patterns with <command>start</command> has limited + usefulness. Also, secondary alias names of units are not considered.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>stop <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Stop (deactivate) one or more units specified on the + command line.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>reload <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Asks all units listed on the command line to reload + their configuration. Note that this will reload the + service-specific configuration, not the unit configuration + file of systemd. If you want systemd to reload the + configuration file of a unit, use the + <command>daemon-reload</command> command. In other words: + for the example case of Apache, this will reload Apache's + <filename>httpd.conf</filename> in the web server, not the + <filename>apache.service</filename> systemd unit + file.</para> + + <para>This command should not be confused with the + <command>daemon-reload</command> command.</para> + </listitem> + + </varlistentry> + <varlistentry> + <term><command>restart <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Restart one or more units specified on the command + line. If the units are not running yet, they will be + started.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>try-restart <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Restart one or more units specified on the command + line if the units are running. This does nothing if units are not + running.</para> + <!-- Note that we don't document condrestart here, as that is just compatibility support, and we generally + don't document that. --> + </listitem> + </varlistentry> + <varlistentry> + <term><command>reload-or-restart <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Reload one or more units if they support it. If not, + restart them instead. If the units are not running yet, they + will be started.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>try-reload-or-restart <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Reload one or more units if they support it. If not, + restart them instead. This does nothing if the units are not + running.</para> + <!-- Note that we don't document force-reload here, as that is just compatibility support, and we generally + don't document that. --> + </listitem> + </varlistentry> + <varlistentry> + <term><command>isolate <replaceable>NAME</replaceable></command></term> + + <listitem> + <para>Start the unit specified on the command line and its + dependencies and stop all others. If a unit name with no + extension is given, an extension of + <literal>.target</literal> will be assumed.</para> + + <para>This is similar to changing the runlevel in a + traditional init system. The <command>isolate</command> + command will immediately stop processes that are not enabled + in the new unit, possibly including the graphical + environment or terminal you are currently using.</para> + + <para>Note that this is allowed only on units where + <option>AllowIsolate=</option> is enabled. See + <citerefentry><refentrytitle>systemd.unit</refentrytitle><manvolnum>5</manvolnum></citerefentry> + for details.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>kill <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Send a signal to one or more processes of the + unit. Use <option>--kill-who=</option> to select which + process to kill. Use <option>--signal=</option> to select + the signal to send.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>is-active <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Check whether any of the specified units are active + (i.e. running). Returns an exit code + <constant>0</constant> if at least one is active, or + non-zero otherwise. Unless <option>--quiet</option> is + specified, this will also print the current unit state to + standard output.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>is-failed <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Check whether any of the specified units are in a + "failed" state. Returns an exit code + <constant>0</constant> if at least one has failed, + non-zero otherwise. Unless <option>--quiet</option> is + specified, this will also print the current unit state to + standard output.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>status</command> <optional><replaceable>PATTERN</replaceable>...|<replaceable>PID</replaceable>...]</optional></term> + + <listitem> + <para>Show terse runtime status information about one or + more units, followed by most recent log data from the + journal. If no units are specified, show system status. If + combined with <option>--all</option>, also show the status of + all units (subject to limitations specified with + <option>-t</option>). If a PID is passed, show information + about the unit the process belongs to.</para> + + <para>This function is intended to generate human-readable + output. If you are looking for computer-parsable output, + use <command>show</command> instead. By default, this + function only shows 10 lines of output and ellipsizes + lines to fit in the terminal window. This can be changed + with <option>--lines</option> and <option>--full</option>, + see above. In addition, <command>journalctl + --unit=<replaceable>NAME</replaceable></command> or + <command>journalctl + --user-unit=<replaceable>NAME</replaceable></command> use + a similar filter for messages and might be more + convenient. + </para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>show</command> <optional><replaceable>PATTERN</replaceable>...|<replaceable>JOB</replaceable>...</optional></term> + + <listitem> + <para>Show properties of one or more units, jobs, or the + manager itself. If no argument is specified, properties of + the manager will be shown. If a unit name is specified, + properties of the unit is shown, and if a job ID is + specified, properties of the job is shown. By default, empty + properties are suppressed. Use <option>--all</option> to + show those too. To select specific properties to show, use + <option>--property=</option>. This command is intended to be + used whenever computer-parsable output is required. Use + <command>status</command> if you are looking for formatted + human-readable output.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>cat <replaceable>PATTERN</replaceable>...</command></term> + + <listitem> + <para>Show backing files of one or more units. Prints the + "fragment" and "drop-ins" (source files) of units. Each + file is preceded by a comment which includes the file + name.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>set-property <replaceable>NAME</replaceable> <replaceable>ASSIGNMENT</replaceable>...</command></term> + + <listitem> + <para>Set the specified unit properties at runtime where + this is supported. This allows changing configuration + parameter properties such as resource control settings at + runtime. Not all properties may be changed at runtime, but + many resource control settings (primarily those in + <citerefentry><refentrytitle>systemd.resource-control</refentrytitle><manvolnum>5</manvolnum></citerefentry>) + may. The changes are applied instantly, and stored on disk + for future boots, unless <option>--runtime</option> is + passed, in which case the settings only apply until the + next reboot. The syntax of the property assignment follows + closely the syntax of assignments in unit files.</para> + + <para>Example: <command>systemctl set-property foobar.service CPUShares=777</command></para> + + <para>If the specified unit appears to be inactive, the + changes will be only stored on disk as described + previously hence they will be effective when the unit will + be started.</para> + + <para>Note that this command allows changing multiple + properties at the same time, which is preferable over + setting them individually. Like unit file configuration + settings, assigning the empty list to list parameters will + reset the list.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>help <replaceable>PATTERN</replaceable>...|<replaceable>PID</replaceable>...</command></term> + + <listitem> + <para>Show manual pages for one or more units, if + available. If a PID is given, the manual pages for the unit + the process belongs to are shown.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>reset-failed [<replaceable>PATTERN</replaceable>...]</command></term> + + <listitem> + <para>Reset the <literal>failed</literal> state of the + specified units, or if no unit name is passed, reset the state of all + units. When a unit fails in some way (i.e. process exiting + with non-zero error code, terminating abnormally or timing + out), it will automatically enter the + <literal>failed</literal> state and its exit code and status + is recorded for introspection by the administrator until the + service is restarted or reset with this command.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term> + <command>list-dependencies</command> + <optional><replaceable>NAME</replaceable></optional> + </term> + + <listitem> + <para>Shows units required and wanted by the specified + unit. This recursively lists units following the + <varname>Requires=</varname>, + <varname>Requisite=</varname>, + <varname>ConsistsOf=</varname>, + <varname>Wants=</varname>, <varname>BindsTo=</varname> + dependencies. If no unit is specified, + <filename>default.target</filename> is implied.</para> + + <para>By default, only target units are recursively + expanded. When <option>--all</option> is passed, all other + units are recursively expanded as well.</para> + + <para>Options <option>--reverse</option>, + <option>--after</option>, <option>--before</option> + may be used to change what types of dependencies + are shown.</para> + </listitem> + </varlistentry> + </variablelist> + </refsect2> + + <refsect2> + <title>Unit File Commands</title> + + <variablelist> + <varlistentry> + <term><command>list-unit-files <optional><replaceable>PATTERN...</replaceable></optional></command></term> + + <listitem> + <para>List installed unit files and their enablement state + (as reported by <command>is-enabled</command>). If one or + more <replaceable>PATTERN</replaceable>s are specified, + only units whose filename (just the last component of the + path) matches one of them are shown.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>enable <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Enable one or more unit files or unit file instances, + as specified on the command line. This will create a number + of symlinks as encoded in the <literal>[Install]</literal> + sections of the unit files. After the symlinks have been + created, the systemd configuration is reloaded (in a way that + is equivalent to <command>daemon-reload</command>) to ensure + the changes are taken into account immediately. Note that + this does <emphasis>not</emphasis> have the effect of also + starting any of the units being enabled. If this + is desired, either <option>--now</option> should be used + together with this command, or an additional <command>start</command> + command must be invoked for the unit. Also note that, in case of + instance enablement, symlinks named the same as instances + are created in the install location, however they all point to the + same template unit file.</para> + + <para>This command will print the actions executed. This + output may be suppressed by passing <option>--quiet</option>. + </para> + + <para>Note that this operation creates only the suggested + symlinks for the units. While this command is the + recommended way to manipulate the unit configuration + directory, the administrator is free to make additional + changes manually by placing or removing symlinks in the + directory. This is particularly useful to create + configurations that deviate from the suggested default + installation. In this case, the administrator must make sure + to invoke <command>daemon-reload</command> manually as + necessary to ensure the changes are taken into account. + </para> + + <para>Enabling units should not be confused with starting + (activating) units, as done by the <command>start</command> + command. Enabling and starting units is orthogonal: units + may be enabled without being started and started without + being enabled. Enabling simply hooks the unit into various + suggested places (for example, so that the unit is + automatically started on boot or when a particular kind of + hardware is plugged in). Starting actually spawns the daemon + process (in case of service units), or binds the socket (in + case of socket units), and so on.</para> + + <para>Depending on whether <option>--system</option>, + <option>--user</option>, <option>--runtime</option>, + or <option>--global</option> is specified, this enables the unit + for the system, for the calling user only, for only this boot of + the system, or for all future logins of all users, or only this + boot. Note that in the last case, no systemd daemon + configuration is reloaded.</para> + + <para>Using <command>enable</command> on masked units + results in an error.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>disable <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Disables one or more units. This removes all symlinks + to the specified unit files from the unit configuration + directory, and hence undoes the changes made by + <command>enable</command>. Note however that this removes + all symlinks to the unit files (i.e. including manual + additions), not just those actually created by + <command>enable</command>. This call implicitly reloads the + systemd daemon configuration after completing the disabling + of the units. Note that this command does not implicitly + stop the units that are being disabled. If this is desired, either + <option>--now</option> should be used together with this command, or + an additional <command>stop</command> command should be executed + afterwards.</para> + + <para>This command will print the actions executed. This + output may be suppressed by passing <option>--quiet</option>. + </para> + + <para>This command honors <option>--system</option>, + <option>--user</option>, <option>--runtime</option> and + <option>--global</option> in a similar way as + <command>enable</command>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>reenable <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Reenable one or more unit files, as specified on the + command line. This is a combination of + <command>disable</command> and <command>enable</command> and + is useful to reset the symlinks a unit is enabled with to + the defaults configured in the <literal>[Install]</literal> + section of the unit file.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>preset <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Reset the enable/disable status one or more unit files, as specified on + the command line, to the defaults configured in the preset policy files. This + has the same effect as <command>disable</command> or + <command>enable</command>, depending how the unit is listed in the preset + files.</para> + + <para>Use <option>--preset-mode=</option> to control whether units shall be + enabled and disabled, or only enabled, or only disabled.</para> + + <para>If the unit carries no install information, it will be silently ignored + by this command.</para> + + <para>For more information on the preset policy format, see + <citerefentry><refentrytitle>systemd.preset</refentrytitle><manvolnum>5</manvolnum></citerefentry>. + For more information on the concept of presets, please consult the + <ulink url="http://freedesktop.org/wiki/Software/systemd/Preset">Preset</ulink> + document.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>preset-all</command></term> + + <listitem> + <para>Resets all installed unit files to the defaults + configured in the preset policy file (see above).</para> + + <para>Use <option>--preset-mode=</option> to control + whether units shall be enabled and disabled, or only + enabled, or only disabled.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>is-enabled <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Checks whether any of the specified unit files are + enabled (as with <command>enable</command>). Returns an + exit code of 0 if at least one is enabled, non-zero + otherwise. Prints the current enable status (see table). + To suppress this output, use <option>--quiet</option>. + </para> + + <table> + <title> + <command>is-enabled</command> output + </title> + + <tgroup cols='3'> + <thead> + <row> + <entry>Name</entry> + <entry>Description</entry> + <entry>Exit Code</entry> + </row> + </thead> + <tbody> + <row> + <entry><literal>enabled</literal></entry> + <entry morerows='1'>Enabled via <filename>.wants/</filename>, <filename>.requires/</filename> or alias symlinks (permanently in <filename>/etc/systemd/system/</filename>, or transiently in <filename>/run/systemd/system/</filename>).</entry> + <entry morerows='1'>0</entry> + </row> + <row> + <entry><literal>enabled-runtime</literal></entry> + </row> + <row> + <entry><literal>linked</literal></entry> + <entry morerows='1'>Made available through one or more symlinks to the unit file (permanently in <filename>/etc/systemd/system/</filename> or transiently in <filename>/run/systemd/system/</filename>), even though the unit file might reside outside of the unit file search path.</entry> + <entry morerows='1'>> 0</entry> + </row> + <row> + <entry><literal>linked-runtime</literal></entry> + </row> + <row> + <entry><literal>masked</literal></entry> + <entry morerows='1'>Completely disabled, so that any start operation on it fails (permanently in <filename>/etc/systemd/system/</filename> or transiently in <filename>/run/systemd/systemd/</filename>).</entry> + <entry morerows='1'>> 0</entry> + </row> + <row> + <entry><literal>masked-runtime</literal></entry> + </row> + <row> + <entry><literal>static</literal></entry> + <entry>The unit file is not enabled, and has no provisions for enabling in the <literal>[Install]</literal> unit file section.</entry> + <entry>0</entry> + </row> + <row> + <entry><literal>indirect</literal></entry> + <entry>The unit file itself is not enabled, but it has a non-empty <varname>Also=</varname> setting in the <literal>[Install]</literal> unit file section, listing other unit files that might be enabled.</entry> + <entry>0</entry> + </row> + <row> + <entry><literal>disabled</literal></entry> + <entry>The unit file is not enabled, but contains an <literal>[Install]</literal> section with installation instructions.</entry> + <entry>> 0</entry> + </row> + <row> + <entry><literal>generated</literal></entry> + <entry>The unit file was generated dynamically via a generator tool. See <citerefentry><refentrytitle>systemd.generator</refentrytitle><manvolnum>7</manvolnum></citerefentry>. Generated unit files may not be enabled, they are enabled implicitly by their generator.</entry> + <entry>0</entry> + </row> + <row> + <entry><literal>transient</literal></entry> + <entry>The unit file has been created dynamically with the runtime API. Transient units may not be enabled.</entry> + <entry>0</entry> + </row> + <row> + <entry><literal>bad</literal></entry> + <entry>The unit file is invalid or another error occurred. Note that <command>is-enabled</command> will not actually return this state, but print an error message instead. However the unit file listing printed by <command>list-unit-files</command> might show it.</entry> + <entry>> 0</entry> + </row> + </tbody> + </tgroup> + </table> + + </listitem> + </varlistentry> + + <varlistentry> + <term><command>mask <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Mask one or more unit files, as specified on the + command line. This will link these units to + <filename>/dev/null</filename>, making it impossible to + start them. This is a stronger version of + <command>disable</command>, since it prohibits all kinds of + activation of the unit, including enablement and manual + activation. Use this option with care. This honors the + <option>--runtime</option> option to only mask temporarily + until the next reboot of the system. The <option>--now</option> + option can be used to ensure that the units are also stopped.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>unmask <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Unmask one or more unit files, as specified on the + command line. This will undo the effect of + <command>mask</command>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>link <replaceable>FILENAME</replaceable>...</command></term> + + <listitem> + <para>Link a unit file that is not in the unit file search + paths into the unit file search path. This requires an + absolute path to a unit file. The effect of this can be + undone with <command>disable</command>. The effect of this + command is that a unit file is available for + <command>start</command> and other commands although it + is not installed directly in the unit search path.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>revert <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Revert one or more unit files to their vendor versions. This command removes drop-in configuration + files that modify the specified units, as well as any user-configured unit file that overrides a matching + vendor supplied unit file. Specifically, for a unit <literal>foo.service</literal> the matching directories + <literal>foo.service.d/</literal> with all their contained files are removed, both below the persistent and + runtime configuration directories (i.e. below <filename>/etc/systemd/system</filename> and + <filename>/run/systemd/system</filename>); if the unit file has a vendor-supplied version (i.e. a unit file + located below <filename>/usr</filename>) any matching peristent or runtime unit file that overrides it is + removed, too. Note that if a unit file has no vendor-supplied version (i.e. is only defined below + <filename>/etc/systemd/system</filename> or <filename>/run/systemd/system</filename>, but not in a unit + file stored below <filename>/usr</filename>), then it is not removed. Also, if a unit is masked, it is + unmasked.</para> + + <para>Effectively, this command may be used to undo all changes made with <command>systemctl + edit</command>, <command>systemctl set-property</command> and <command>systemctl mask</command> and puts + the original unit file with its settings back in effect.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>add-wants <replaceable>TARGET</replaceable> + <replaceable>NAME</replaceable>...</command></term> + <term><command>add-requires <replaceable>TARGET</replaceable> + <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Adds <literal>Wants=</literal> or <literal>Requires=</literal> + dependencies, respectively, to the specified + <replaceable>TARGET</replaceable> for one or more units. </para> + + <para>This command honors <option>--system</option>, + <option>--user</option>, <option>--runtime</option> and + <option>--global</option> in a way similar to + <command>enable</command>.</para> + + </listitem> + </varlistentry> + + <varlistentry> + <term><command>edit <replaceable>NAME</replaceable>...</command></term> + + <listitem> + <para>Edit a drop-in snippet or a whole replacement file if + <option>--full</option> is specified, to extend or override the + specified unit.</para> + + <para>Depending on whether <option>--system</option> (the default), + <option>--user</option>, or <option>--global</option> is specified, + this command creates a drop-in file for each unit either for the system, + for the calling user, or for all futures logins of all users. Then, + the editor (see the "Environment" section below) is invoked on + temporary files which will be written to the real location if the + editor exits successfully.</para> + + <para>If <option>--full</option> is specified, this will copy the + original units instead of creating drop-in files.</para> + + <para>If <option>--runtime</option> is specified, the changes will + be made temporarily in <filename>/run</filename> and they will be + lost on the next reboot.</para> + + <para>If the temporary file is empty upon exit, the modification of + the related unit is canceled.</para> + + <para>After the units have been edited, systemd configuration is + reloaded (in a way that is equivalent to <command>daemon-reload</command>). + </para> + + <para>Note that this command cannot be used to remotely edit units + and that you cannot temporarily edit units which are in + <filename>/etc</filename>, since they take precedence over + <filename>/run</filename>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>get-default</command></term> + + <listitem> + <para>Return the default target to boot into. This returns + the target unit name <filename>default.target</filename> + is aliased (symlinked) to.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>set-default <replaceable>NAME</replaceable></command></term> + + <listitem> + <para>Set the default target to boot into. This sets + (symlinks) the <filename>default.target</filename> alias + to the given target unit.</para> + </listitem> + </varlistentry> + + </variablelist> + </refsect2> + + <refsect2> + <title>Machine Commands</title> + + <variablelist> + <varlistentry> + <term><command>list-machines <optional><replaceable>PATTERN</replaceable>...</optional></command></term> + + <listitem> + <para>List the host and all running local containers with + their state. If one or more + <replaceable>PATTERN</replaceable>s are specified, only + containers matching one of them are shown. + </para> + </listitem> + </varlistentry> + </variablelist> + </refsect2> + + <refsect2> + <title>Job Commands</title> + + <variablelist> + <varlistentry> + <term><command>list-jobs <optional><replaceable>PATTERN...</replaceable></optional></command></term> + + <listitem> + <para>List jobs that are in progress. If one or more + <replaceable>PATTERN</replaceable>s are specified, only + jobs for units matching one of them are shown.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>cancel <replaceable>JOB</replaceable>...</command></term> + + <listitem> + <para>Cancel one or more jobs specified on the command line + by their numeric job IDs. If no job ID is specified, cancel + all pending jobs.</para> + </listitem> + </varlistentry> + </variablelist> + </refsect2> + + <refsect2> + <title>Environment Commands</title> + + <variablelist> + <varlistentry> + <term><command>show-environment</command></term> + + <listitem> + <para>Dump the systemd manager environment block. The + environment block will be dumped in straight-forward form + suitable for sourcing into a shell script. This environment + block will be passed to all processes the manager + spawns.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>set-environment <replaceable>VARIABLE=VALUE</replaceable>...</command></term> + + <listitem> + <para>Set one or more systemd manager environment variables, + as specified on the command line.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>unset-environment <replaceable>VARIABLE</replaceable>...</command></term> + + <listitem> + <para>Unset one or more systemd manager environment + variables. If only a variable name is specified, it will be + removed regardless of its value. If a variable and a value + are specified, the variable is only removed if it has the + specified value.</para> + </listitem> + </varlistentry> + <varlistentry> + <term> + <command>import-environment</command> + <optional><replaceable>VARIABLE...</replaceable></optional> + </term> + + <listitem> + <para>Import all, one or more environment variables set on + the client into the systemd manager environment block. If + no arguments are passed, the entire environment block is + imported. Otherwise, a list of one or more environment + variable names should be passed, whose client-side values + are then imported into the manager's environment + block.</para> + </listitem> + </varlistentry> + </variablelist> + </refsect2> + + <refsect2> + <title>Manager Lifecycle Commands</title> + + <variablelist> + <varlistentry> + <term><command>daemon-reload</command></term> + + <listitem> + <para>Reload the systemd manager configuration. This will + rerun all generators (see + <citerefentry><refentrytitle>systemd.generator</refentrytitle><manvolnum>7</manvolnum></citerefentry>), + reload all unit files, and recreate the entire dependency + tree. While the daemon is being reloaded, all sockets + systemd listens on behalf of user configuration will stay + accessible.</para> + + <para>This command should not be confused with the + <command>reload</command> command.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>daemon-reexec</command></term> + + <listitem> + <para>Reexecute the systemd manager. This will serialize the + manager state, reexecute the process and deserialize the + state again. This command is of little use except for + debugging and package upgrades. Sometimes, it might be + helpful as a heavy-weight <command>daemon-reload</command>. + While the daemon is being reexecuted, all sockets systemd listening + on behalf of user configuration will stay accessible. + </para> + </listitem> + </varlistentry> + </variablelist> + </refsect2> + + <refsect2> + <title>System Commands</title> + + <variablelist> + <varlistentry> + <term><command>is-system-running</command></term> + + <listitem> + <para>Checks whether the system is operational. This + returns success (exit code 0) when the system is fully up + and running, specifically not in startup, shutdown or + maintenance mode, and with no failed services. Failure is + returned otherwise (exit code non-zero). In addition, the + current state is printed in a short string to standard + output, see the table below. Use <option>--quiet</option> to + suppress this output.</para> + + <table> + <title><command>is-system-running</command> output</title> + <tgroup cols='3'> + <colspec colname='name'/> + <colspec colname='description'/> + <colspec colname='exit-code'/> + <thead> + <row> + <entry>Name</entry> + <entry>Description</entry> + <entry>Exit Code</entry> + </row> + </thead> + <tbody> + <row> + <entry><varname>initializing</varname></entry> + <entry><para>Early bootup, before + <filename>basic.target</filename> is reached + or the <varname>maintenance</varname> state entered. + </para></entry> + <entry>> 0</entry> + </row> + <row> + <entry><varname>starting</varname></entry> + <entry><para>Late bootup, before the job queue + becomes idle for the first time, or one of the + rescue targets are reached.</para></entry> + <entry>> 0</entry> + </row> + <row> + <entry><varname>running</varname></entry> + <entry><para>The system is fully + operational.</para></entry> + <entry>0</entry> + </row> + <row> + <entry><varname>degraded</varname></entry> + <entry><para>The system is operational but one or more + units failed.</para></entry> + <entry>> 0</entry> + </row> + <row> + <entry><varname>maintenance</varname></entry> + <entry><para>The rescue or emergency target is + active.</para></entry> + <entry>> 0</entry> + </row> + <row> + <entry><varname>stopping</varname></entry> + <entry><para>The manager is shutting + down.</para></entry> + <entry>> 0</entry> + </row> + <row> + <entry><varname>offline</varname></entry> + <entry><para>The manager is not + running. Specifically, this is the operational + state if an incompatible program is running as + system manager (PID 1).</para></entry> + <entry>> 0</entry> + </row> + <row> + <entry><varname>unknown</varname></entry> + <entry><para>The operational state could not be + determined, due to lack of resources or another + error cause.</para></entry> + <entry>> 0</entry> + </row> + </tbody> + </tgroup> + </table> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>default</command></term> + + <listitem> + <para>Enter default mode. This is mostly equivalent to + <command>isolate default.target</command>.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>rescue</command></term> + + <listitem> + <para>Enter rescue mode. This is mostly equivalent to + <command>isolate rescue.target</command>, but also prints a + wall message to all users.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>emergency</command></term> + + <listitem> + <para>Enter emergency mode. This is mostly equivalent to + <command>isolate emergency.target</command>, but also prints + a wall message to all users.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>halt</command></term> + + <listitem> + <para>Shut down and halt the system. This is mostly equivalent to + <command>start halt.target --job-mode=replace-irreversibly</command>, but also + prints a wall message to all users. If combined with + <option>--force</option>, shutdown of all running services is + skipped, however all processes are killed and all file + systems are unmounted or mounted read-only, immediately + followed by the system halt. If <option>--force</option> is + specified twice, the operation is immediately executed + without terminating any processes or unmounting any file + systems. This may result in data loss.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>poweroff</command></term> + + <listitem> + <para>Shut down and power-off the system. This is mostly + equivalent to <command>start poweroff.target --job-mode=replace-irreversibly</command>, + but also prints a wall message to all users. If combined with + <option>--force</option>, shutdown of all running services is + skipped, however all processes are killed and all file + systems are unmounted or mounted read-only, immediately + followed by the powering off. If <option>--force</option> is + specified twice, the operation is immediately executed + without terminating any processes or unmounting any file + systems. This may result in data loss.</para> + </listitem> + </varlistentry> + <varlistentry> + <term><command>reboot <optional><replaceable>arg</replaceable></optional></command></term> + + <listitem> + <para>Shut down and reboot the system. This is mostly + equivalent to <command>start reboot.target --job-mode=replace-irreversibly</command>, + but also prints a wall message to all users. If combined with + <option>--force</option>, shutdown of all running services is + skipped, however all processes are killed and all file + systems are unmounted or mounted read-only, immediately + followed by the reboot. If <option>--force</option> is + specified twice, the operation is immediately executed + without terminating any processes or unmounting any file + systems. This may result in data loss.</para> + + <para>If the optional argument + <replaceable>arg</replaceable> is given, it will be passed + as the optional argument to the + <citerefentry><refentrytitle>reboot</refentrytitle><manvolnum>2</manvolnum></citerefentry> + system call. The value is architecture and firmware + specific. As an example, <literal>recovery</literal> might + be used to trigger system recovery, and + <literal>fota</literal> might be used to trigger a + <quote>firmware over the air</quote> update.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>kexec</command></term> + + <listitem> + <para>Shut down and reboot the system via kexec. This is + mostly equivalent to <command>start kexec.target --job-mode=replace-irreversibly</command>, + but also prints a wall message to all users. If combined + with <option>--force</option>, shutdown of all running + services is skipped, however all processes are killed and + all file systems are unmounted or mounted read-only, + immediately followed by the reboot.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>exit <optional><replaceable>EXIT_CODE</replaceable></optional></command></term> + + <listitem> + <para>Ask the systemd manager to quit. This is only + supported for user service managers (i.e. in conjunction + with the <option>--user</option> option) or in containers + and is equivalent to <command>poweroff</command> otherwise.</para> + + <para>The systemd manager can exit with a non-zero exit + code if the optional argument + <replaceable>EXIT_CODE</replaceable> is given.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>switch-root <replaceable>ROOT</replaceable> <optional><replaceable>INIT</replaceable></optional></command></term> + + <listitem> + <para>Switches to a different root directory and executes a + new system manager process below it. This is intended for + usage in initial RAM disks ("initrd"), and will transition + from the initrd's system manager process (a.k.a. "init" + process) to the main system manager process. This call takes two + arguments: the directory that is to become the new root directory, and + the path to the new system manager binary below it to + execute as PID 1. If the latter is omitted or the empty + string, a systemd binary will automatically be searched for + and used as init. If the system manager path is omitted or + equal to the empty string, the state of the initrd's system + manager process is passed to the main system manager, which + allows later introspection of the state of the services + involved in the initrd boot.</para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>suspend</command></term> + + <listitem> + <para>Suspend the system. This will trigger activation of + the special <filename>suspend.target</filename> target. + </para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>hibernate</command></term> + + <listitem> + <para>Hibernate the system. This will trigger activation of + the special <filename>hibernate.target</filename> target. + </para> + </listitem> + </varlistentry> + + <varlistentry> + <term><command>hybrid-sleep</command></term> + + <listitem> + <para>Hibernate and suspend the system. This will trigger + activation of the special + <filename>hybrid-sleep.target</filename> target.</para> + </listitem> + </varlistentry> + </variablelist> + </refsect2> + + <refsect2> + <title>Parameter Syntax</title> + + <para>Unit commands listed above take either a single unit name (designated as <replaceable>NAME</replaceable>), + or multiple unit specifications (designated as <replaceable>PATTERN</replaceable>...). In the first case, the + unit name with or without a suffix must be given. If the suffix is not specified (unit name is "abbreviated"), + systemctl will append a suitable suffix, <literal>.service</literal> by default, and a type-specific suffix in + case of commands which operate only on specific unit types. For example, + <programlisting># systemctl start sshd</programlisting> and + <programlisting># systemctl start sshd.service</programlisting> + are equivalent, as are + <programlisting># systemctl isolate default</programlisting> + and + <programlisting># systemctl isolate default.target</programlisting> + Note that (absolute) paths to device nodes are automatically converted to device unit names, and other (absolute) + paths to mount unit names. + <programlisting># systemctl status /dev/sda +# systemctl status /home</programlisting> + are equivalent to: + <programlisting># systemctl status dev-sda.device +# systemctl status home.mount</programlisting> + In the second case, shell-style globs will be matched against the primary names of all currently loaded units; + literal unit names, with or without a suffix, will be treated as in the first case. This means that literal unit + names always refer to exactly one unit, but globs may match zero units and this is not considered an + error.</para> + + <para>Glob patterns use + <citerefentry project='man-pages'><refentrytitle>fnmatch</refentrytitle><manvolnum>3</manvolnum></citerefentry>, + so normal shell-style globbing rules are used, and + <literal>*</literal>, <literal>?</literal>, + <literal>[]</literal> may be used. See + <citerefentry project='man-pages'><refentrytitle>glob</refentrytitle><manvolnum>7</manvolnum></citerefentry> + for more details. The patterns are matched against the primary names of + currently loaded units, and patterns which do not match anything + are silently skipped. For example: + <programlisting># systemctl stop sshd@*.service</programlisting> + will stop all <filename>sshd@.service</filename> instances. Note that alias names of units, and units that aren't + loaded are not considered for glob expansion. + </para> + + <para>For unit file commands, the specified <replaceable>NAME</replaceable> should be the name of the unit file + (possibly abbreviated, see above), or the absolute path to the unit file: + <programlisting># systemctl enable foo.service</programlisting> + or + <programlisting># systemctl link /path/to/foo.service</programlisting> + </para> + </refsect2> + + </refsect1> + + <refsect1> + <title>Exit status</title> + + <para>On success, 0 is returned, a non-zero failure + code otherwise.</para> + </refsect1> + + <refsect1> + <title>Environment</title> + + <variablelist class='environment-variables'> + <varlistentry> + <term><varname>$SYSTEMD_EDITOR</varname></term> + + <listitem><para>Editor to use when editing units; overrides + <varname>$EDITOR</varname> and <varname>$VISUAL</varname>. If neither + <varname>$SYSTEMD_EDITOR</varname> nor <varname>$EDITOR</varname> nor + <varname>$VISUAL</varname> are present or if it is set to an empty + string or if their execution failed, systemctl will try to execute well + known editors in this order: + <citerefentry project='die-net'><refentrytitle>editor</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry project='die-net'><refentrytitle>nano</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry project='die-net'><refentrytitle>vim</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry project='die-net'><refentrytitle>vi</refentrytitle><manvolnum>1</manvolnum></citerefentry>. + </para></listitem> + </varlistentry> + </variablelist> + <xi:include href="less-variables.xml" xpointer="pager"/> + <xi:include href="less-variables.xml" xpointer="less"/> + </refsect1> + + <refsect1> + <title>See Also</title> + <para> + <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>journalctl</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>loginctl</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>machinectl</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.unit</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.resource-control</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.special</refentrytitle><manvolnum>7</manvolnum></citerefentry>, + <citerefentry project='man-pages'><refentrytitle>wall</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.preset</refentrytitle><manvolnum>5</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemd.generator</refentrytitle><manvolnum>7</manvolnum></citerefentry>, + <citerefentry project='man-pages'><refentrytitle>glob</refentrytitle><manvolnum>7</manvolnum></citerefentry> + </para> + </refsect1> + +</refentry> diff --git a/src/grp-system/systemctl/systemd-sysv-install.SKELETON b/src/grp-system/systemctl/systemd-sysv-install.SKELETON new file mode 100755 index 0000000000..a53a3e6221 --- /dev/null +++ b/src/grp-system/systemctl/systemd-sysv-install.SKELETON @@ -0,0 +1,47 @@ +#!/bin/sh +# This script is called by "systemctl enable/disable" when the given unit is a +# SysV init.d script. It needs to call the distribution's mechanism for +# enabling/disabling those, such as chkconfig, update-rc.d, or similar. This +# can optionally take a --root argument for enabling a SysV init script +# in a chroot or similar. +set -e + +usage() { + echo "Usage: $0 [--root=path] enable|disable|is-enabled <sysv script name>" >&2 + exit 1 +} + +# parse options +eval set -- "$(getopt -o r: --long root: -- "$@")" +while true; do + case "$1" in + -r|--root) + ROOT="$2" + shift 2 ;; + --) shift ; break ;; + *) usage ;; + esac +done + +NAME="$2" +[ -n "$NAME" ] || usage + +case "$1" in + enable) + # call the command to enable SysV init script $NAME here + # (consider optional $ROOT) + echo "IMPLEMENT ME: enabling SysV init.d script $NAME" + ;; + disable) + # call the command to disable SysV init script $NAME here + # (consider optional $ROOT) + echo "IMPLEMENT ME: disabling SysV init.d script $NAME" + ;; + is-enabled) + # exit with 0 if $NAME is enabled, non-zero if it is disabled + # (consider optional $ROOT) + echo "IMPLEMENT ME: checking SysV init.d script $NAME" + ;; + *) + usage ;; +esac diff --git a/src/grp-system/systemctl/telinit.xml b/src/grp-system/systemctl/telinit.xml new file mode 100644 index 0000000000..02d31fbd46 --- /dev/null +++ b/src/grp-system/systemctl/telinit.xml @@ -0,0 +1,179 @@ +<?xml version='1.0'?> <!--*-nxml-*--> +<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN" + "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd"> + +<!-- + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + 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 + (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 + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +--> + +<refentry id="telinit" + xmlns:xi="http://www.w3.org/2001/XInclude"> + + <refentryinfo> + <title>telinit</title> + <productname>systemd</productname> + + <authorgroup> + <author> + <contrib>Developer</contrib> + <firstname>Lennart</firstname> + <surname>Poettering</surname> + <email>lennart@poettering.net</email> + </author> + </authorgroup> + </refentryinfo> + + <refmeta> + <refentrytitle>telinit</refentrytitle> + <manvolnum>8</manvolnum> + </refmeta> + + <refnamediv> + <refname>telinit</refname> + <refpurpose>Change SysV runlevel</refpurpose> + </refnamediv> + + <refsynopsisdiv> + <cmdsynopsis> + <command>telinit <arg choice="opt" rep="repeat">OPTIONS</arg> <arg choice="req">COMMAND</arg></command> + </cmdsynopsis> + </refsynopsisdiv> + + <refsect1> + <title>Description</title> + + <para><command>telinit</command> may be used to change the SysV + system runlevel. Since the concept of SysV runlevels is obsolete + the runlevel requests will be transparently translated into + systemd unit activation requests.</para> + + </refsect1> + + <refsect1> + <title>Options</title> + + <para>The following options are understood:</para> + + <variablelist> + <varlistentry> + <term><option>--help</option></term> + + <xi:include href="standard-options.xml" xpointer="help-text" /> + </varlistentry> + + <varlistentry> + <term><option>--no-wall</option></term> + + <listitem><para>Do not send wall message before + reboot/halt/power-off.</para></listitem> + </varlistentry> + </variablelist> + + <para>The following commands are understood:</para> + + <variablelist> + <varlistentry> + <term><command>0</command></term> + + <listitem><para>Power-off the machine. This is translated into + an activation request for <filename>poweroff.target</filename> + and is equivalent to <command>systemctl + poweroff</command>.</para></listitem> + </varlistentry> + + <varlistentry> + <term><command>6</command></term> + + <listitem><para>Reboot the machine. This is translated into an + activation request for <filename>reboot.target</filename> and + is equivalent to <command>systemctl + reboot</command>.</para></listitem> + </varlistentry> + + <varlistentry> + <term><command>2</command></term> + <term><command>3</command></term> + <term><command>4</command></term> + <term><command>5</command></term> + + <listitem><para>Change the SysV runlevel. This is translated + into an activation request for + <filename>runlevel2.target</filename>, + <filename>runlevel3.target</filename>, ... and is equivalent + to <command>systemctl isolate runlevel2.target</command>, + <command>systemctl isolate runlevel3.target</command>, + ...</para></listitem> + </varlistentry> + + <varlistentry> + <term><command>1</command></term> + <term><command>s</command></term> + <term><command>S</command></term> + + <listitem><para>Change into system rescue mode. This is + translated into an activation request for + <filename>rescue.target</filename> and is equivalent to + <command>systemctl rescue</command>.</para></listitem> + </varlistentry> + + <varlistentry> + <term><command>q</command></term> + <term><command>Q</command></term> + + <listitem><para>Reload daemon configuration. This is + equivalent to <command>systemctl + daemon-reload</command>.</para></listitem> + </varlistentry> + + <varlistentry> + <term><command>u</command></term> + <term><command>U</command></term> + + <listitem><para>Serialize state, reexecute daemon and + deserialize state again. This is equivalent to + <command>systemctl daemon-reexec</command>.</para></listitem> + </varlistentry> + + </variablelist> + </refsect1> + + <refsect1> + <title>Exit status</title> + + <para>On success, 0 is returned, a non-zero failure + code otherwise.</para> + </refsect1> + + <refsect1> + <title>Notes</title> + + <para>This is a legacy command available for compatibility only. + It should not be used anymore, as the concept of runlevels is + obsolete.</para> + </refsect1> + + <refsect1> + <title>See Also</title> + <para> + <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry><refentrytitle>systemctl</refentrytitle><manvolnum>1</manvolnum></citerefentry>, + <citerefentry project='man-pages'><refentrytitle>wall</refentrytitle><manvolnum>1</manvolnum></citerefentry> + </para> + </refsect1> + +</refentry> |