diff options
66 files changed, 1876 insertions, 540 deletions
diff --git a/configure.ac b/configure.ac index 4d1c96606f..36061c0ba9 100644 --- a/configure.ac +++ b/configure.ac @@ -1220,9 +1220,9 @@ AM_CONDITIONAL(ENABLE_NETWORKD, [test "x$have_networkd" = "xyes"]) # ------------------------------------------------------------------------------ have_efi=no -AC_ARG_ENABLE(efi, AS_HELP_STRING([--disable-efi], [disable EFI support])) +AC_ARG_ENABLE(efi, AS_HELP_STRING([--disable-efi], [disable systemd-boot and bootctl (EFI support)])) if test "x$enable_efi" != "xno"; then - AC_DEFINE(ENABLE_EFI, 1, [Define if EFI support is to be enabled]) + AC_DEFINE(ENABLE_EFI, 1, [Define if systemd-boot and bootctl are to be enabled]) have_efi=yes fi AM_CONDITIONAL(ENABLE_EFI, [test "x$have_efi" = "xyes"]) diff --git a/hwdb/60-evdev.hwdb b/hwdb/60-evdev.hwdb index 4d14a6a2f4..e738ff7be5 100644 --- a/hwdb/60-evdev.hwdb +++ b/hwdb/60-evdev.hwdb @@ -134,6 +134,13 @@ evdev:name:AlpsPS/2 ALPS DualPoint TouchPad:dmi:bvn*:bvr*:bd*:svnDellInc.:pnLati EVDEV_ABS_35=76:1815:22 EVDEV_ABS_36=131:1330:30 +# Dell Precision 5510 +evdev:name:SynPS/2 Synaptics TouchPad:dmi:bvn*:bvr*:bd*:svnDellInc.:pnPrecision5510* + EVDEV_ABS_00=::42 + EVDEV_ABS_01=::43 + EVDEV_ABS_35=::42 + EVDEV_ABS_36=::43 + # Dell Precision M4700 evdev:name:AlpsPS/2 ALPS DualPoint TouchPad:dmi:*svnDellInc.:pnPrecisionM4700* EVDEV_ABS_00=0:1960:24 diff --git a/hwdb/60-keyboard.hwdb b/hwdb/60-keyboard.hwdb index 1176881650..25caa60626 100644 --- a/hwdb/60-keyboard.hwdb +++ b/hwdb/60-keyboard.hwdb @@ -758,22 +758,22 @@ evdev:input:b0003v046DpC52D* # Internet Navigator evdev:input:b0003v046DpC309* - KEYBOARD_KEY_90001=chat # Messenger/SMS - KEYBOARD_KEY_90002=camera # webcam - KEYBOARD_KEY_90003=prog1 # iTouch - KEYBOARD_KEY_90004=shop # Shopping - KEYBOARD_KEY_c0201=new # New (F1) - KEYBOARD_KEY_c0289=reply # Reply mail (F2) - KEYBOARD_KEY_c028b=forwardmail # Forward mail (F3) - KEYBOARD_KEY_c028c=send # Send (F4) - KEYBOARD_KEY_c021a=undo # Undo (F5). - KEYBOARD_KEY_c0279=redo # Redo (F6). - KEYBOARD_KEY_c0208=print # Print (F7) - KEYBOARD_KEY_c0207=save # Save (F8) - KEYBOARD_KEY_c0194=file # My Computer (F9) - KEYBOARD_KEY_c01a7=documents # My Documents (F10) - KEYBOARD_KEY_c01b6=images # My Pictures (F11) ?? - KEYBOARD_KEY_c01b7=sound # My Music (F12) ?? + KEYBOARD_KEY_90001=chat # Messenger/SMS + KEYBOARD_KEY_90002=camera # webcam + KEYBOARD_KEY_90003=prog1 # iTouch + KEYBOARD_KEY_90004=shop # Shopping + KEYBOARD_KEY_c0201=new # New (F1) + KEYBOARD_KEY_c0289=reply # Reply mail (F2) + KEYBOARD_KEY_c028b=forwardmail # Forward mail (F3) + KEYBOARD_KEY_c028c=send # Send (F4) + KEYBOARD_KEY_c021a=undo # Undo (F5) + KEYBOARD_KEY_c0279=redo # Redo (F6) + KEYBOARD_KEY_c0208=print # Print (F7) + KEYBOARD_KEY_c0207=save # Save (F8) + KEYBOARD_KEY_c0194=file # My Computer (F9) + KEYBOARD_KEY_c01a7=documents # My Documents (F10) + KEYBOARD_KEY_c01b6=images # My Pictures (F11) ?? + KEYBOARD_KEY_c01b7=sound # My Music (F12) ?? ########################################################### @@ -803,7 +803,7 @@ evdev:atkbd:dmi:bvn*:bvr*:svnMedion*:pnAkoya*:pvr* KEYBOARD_KEY_b0=!volumeup KEYBOARD_KEY_19=!p KEYBOARD_KEY_df=sleep - + # FID2060 evdev:atkbd:dmi:bvn*:bvr*:bd*:svnMEDION*:pn*FID2060*:pvr* KEYBOARD_KEY_6b=channeldown # Thottle Down diff --git a/hwdb/70-mouse.hwdb b/hwdb/70-mouse.hwdb index d8215a6179..56e36af0e5 100644 --- a/hwdb/70-mouse.hwdb +++ b/hwdb/70-mouse.hwdb @@ -47,6 +47,7 @@ # ID_INPUT_TRACKBALL # MOUSE_DPI # MOUSE_WHEEL_CLICK_ANGLE +# MOUSE_WHEEL_CLICK_ANGLE_HORIZONTAL # ######################################### # ID_INPUT_TRACKBALL # @@ -104,6 +105,14 @@ # # Most mice have a 15 degree click stop (24 clicks per full rotation). # +######################################### +# MOUSE_WHEEL_CLICK_ANGLE_HORIZONTAL # +######################################### +# +# Identical to MOUSE_WHEEL_CLICK_ANGLE but for the horizontal scroll wheel. +# This property may only be specified if the angle for the horizontal +# scroll wheel differs from the vertical wheel. If so, *both* click angles +# must be specified. # # Sort by brand, type (usb, bluetooth), DPI, frequency. @@ -344,9 +353,14 @@ mouse:usb:v046dpc52b:name:Logitech Unifying Device. Wireless PID:402d: # Logitech Performance MX mouse:usb:v046dp101a:name:Logitech Performance MX: + MOUSE_DPI=1000@166 + # Logitech MX Master +# Horiz wheel has 14 stops, angle is rounded up mouse:usb:v046dp4041:name:Logitech MX Master: MOUSE_DPI=1000@166 + MOUSE_WHEEL_CLICK_ANGLE=15 + MOUSE_WHEEL_CLICK_ANGLE_HORIZONTAL=26 # Logitech MK260 Wireless Combo Receiver aka M-R0011 mouse:usb:v046dpc52e:name:Logitech USB Receiver: diff --git a/hwdb/parse_hwdb.py b/hwdb/parse_hwdb.py index 99d034b4e0..f55562250d 100755 --- a/hwdb/parse_hwdb.py +++ b/hwdb/parse_hwdb.py @@ -82,6 +82,7 @@ def property_grammar(): setting = Optional('*')('DEFAULT') + INTEGER('DPI') + Suppress('@') + INTEGER('HZ') props = (('MOUSE_DPI', Group(OneOrMore(setting('SETTINGS*')))), ('MOUSE_WHEEL_CLICK_ANGLE', INTEGER), + ('MOUSE_WHEEL_CLICK_ANGLE_HORIZONTAL', INTEGER), ('ID_INPUT_TRACKBALL', Literal('1')), ('POINTINGSTICK_SENSITIVITY', INTEGER), ('POINTINGSTICK_CONST_ACCEL', REAL), diff --git a/man/hostnamectl.xml b/man/hostnamectl.xml index 60004e9d04..9e1b593e6d 100644 --- a/man/hostnamectl.xml +++ b/man/hostnamectl.xml @@ -71,10 +71,9 @@ set, and is valid (something other than localhost), then the transient hostname is not used.</para> - <para>Note that the pretty hostname has little restrictions on the - characters used, while the static and transient hostnames are - limited to the usually accepted characters of Internet domain - names.</para> + <para>Note that the pretty hostname has little restrictions on the characters and length used, while the static and + transient hostnames are limited to the usually accepted characters of Internet domain names, and 64 characters at + maximum (the latter being a Linux limitation).</para> <para>The static hostname is stored in <filename>/etc/hostname</filename>, see @@ -107,15 +106,11 @@ <term><option>--transient</option></term> <term><option>--pretty</option></term> - <listitem><para>If <command>status</command> is used (or no - explicit command is given) and one of those fields is given, - <command>hostnamectl</command> will print out just this - selected hostname.</para> + <listitem><para>If <command>status</command> is invoked (or no explicit command is given) and one of these + switches is specified, <command>hostnamectl</command> will print out just this selected hostname.</para> - <para>If used with <command>set-hostname</command>, only the - selected hostname(s) will be updated. When more than one of - those options is used, all the specified hostnames will be - updated. </para></listitem> + <para>If used with <command>set-hostname</command>, only the selected hostname(s) will be updated. When more + than one of these switches are specified, all the specified hostnames will be updated. </para></listitem> </varlistentry> <xi:include href="user-system-options.xml" xpointer="host" /> @@ -139,22 +134,14 @@ <varlistentry> <term><command>set-hostname <replaceable>NAME</replaceable></command></term> - <listitem><para>Set the system hostname to - <replaceable>NAME</replaceable>. By default, this will alter - the pretty, the static, and the transient hostname alike; - however, if one or more of <option>--static</option>, - <option>--transient</option>, <option>--pretty</option> are - used, only the selected hostnames are changed. If the pretty - hostname is being set, and static or transient are being set - as well, the specified hostname will be simplified in regards - to the character set used before the latter are updated. This - is done by replacing spaces with <literal>-</literal> and - removing special characters. This ensures that the pretty and - the static hostname are always closely related while still - following the validity rules of the specific name. This - simplification of the hostname string is not done if only the - transient and/or static host names are set, and the pretty - host name is left untouched.</para> + <listitem><para>Set the system hostname to <replaceable>NAME</replaceable>. By default, this will alter the + pretty, the static, and the transient hostname alike; however, if one or more of <option>--static</option>, + <option>--transient</option>, <option>--pretty</option> are used, only the selected hostnames are changed. If + the pretty hostname is being set, and static or transient are being set as well, the specified hostname will be + simplified in regards to the character set used before the latter are updated. This is done by removing special + characters and spaces. This ensures that the pretty and the static hostname are always closely related while + still following the validity rules of the specific name. This simplification of the hostname string is not done + if only the transient and/or static host names are set, and the pretty host name is left untouched.</para> <para>Pass the empty string <literal></literal> as the hostname to reset the selected hostnames to their default diff --git a/man/journalctl.xml b/man/journalctl.xml index c448a29a51..63b4a267b8 100644 --- a/man/journalctl.xml +++ b/man/journalctl.xml @@ -659,10 +659,12 @@ <term><option>--root=<replaceable>ROOT</replaceable></option></term> <listitem><para>Takes a directory path as an argument. If - specified, journalctl will operate on catalog file hierarchy + specified, journalctl will operate on journal directories and catalog file hierarchy underneath the specified directory instead of the root directory (e.g. <option>--update-catalog</option> will create - <filename><replaceable>ROOT</replaceable>/var/lib/systemd/catalog/database</filename>). + <filename><replaceable>ROOT</replaceable>/var/lib/systemd/catalog/database</filename>, + and journal files under <filename><replaceable>ROOT</replaceable>/run/journal</filename> + or <filename><replaceable>ROOT</replaceable>/var/log/journal</filename> will be displayed). </para></listitem> </varlistentry> diff --git a/man/logind.conf.xml b/man/logind.conf.xml index adba5a4131..9b0e181849 100644 --- a/man/logind.conf.xml +++ b/man/logind.conf.xml @@ -211,7 +211,7 @@ <term><varname>HandleLidSwitch=</varname></term> <term><varname>HandleLidSwitchDocked=</varname></term> - <listitem><para>Controls whether logind shall handle the + <listitem><para>Controls how logind shall handle the system power and sleep keys and the lid switch to trigger actions such as system power-off or suspend. Can be one of <literal>ignore</literal>, @@ -240,7 +240,16 @@ docking station, or if more than one display is connected, the action specified by <varname>HandleLidSwitchDocked=</varname> occurs; otherwise the <varname>HandleLidSwitch=</varname> - action occurs.</para></listitem> + action occurs.</para> + + <para>A different application may disable logind's handling of system power and + sleep keys and the lid switch by taking a low-level inhibitor lock + ("handle-power-key", "handle-suspend-key", "handle-hibernate-key", + "handle-lid-switch"). This is most commonly used by graphical desktop environments + to take over suspend and hibernation handling, and to use their own configuration + mechanisms. If a low-level inhibitor lock is taken, logind will not take any + action when that key or switch is triggered and the <varname>Handle*=</varname> + settings are irrelevant.</para></listitem> </varlistentry> <varlistentry> @@ -249,21 +258,22 @@ <term><varname>HibernateKeyIgnoreInhibited=</varname></term> <term><varname>LidSwitchIgnoreInhibited=</varname></term> - <listitem><para>Controls whether actions triggered by the - power and sleep keys and the lid switch are subject to - inhibitor locks. These settings take boolean arguments. If - <literal>no</literal>, the inhibitor locks taken by - applications in order to block the requested operation are - respected. If <literal>yes</literal>, the requested operation - is executed in any case. + <listitem><para>Controls whether actions that <command>systemd-logind</command> + takes when the power and sleep keys and the lid switch are triggered are subject + to high-level inhibitor locks ("shutdown", "sleep", "idle"). Low level inhibitor + locks ("handle-*-key"), are always honoured, irrespective of this setting.</para> + + <para>These settings take boolean arguments. If <literal>no</literal>, the + inhibitor locks taken by applications are respected. If <literal>yes</literal>, + "shutdown", "sleep", and "idle" inhibitor locks are ignored. <varname>PowerKeyIgnoreInhibited=</varname>, - <varname>SuspendKeyIgnoreInhibited=</varname> and - <varname>HibernateKeyIgnoreInhibited=</varname> default to - <literal>no</literal>. - <varname>LidSwitchIgnoreInhibited=</varname> defaults to - <literal>yes</literal>. This means that the lid switch does - not respect suspend blockers by default, but the power and - sleep keys do. </para></listitem> + <varname>SuspendKeyIgnoreInhibited=</varname>, and + <varname>HibernateKeyIgnoreInhibited=</varname> default to <literal>no</literal>. + <varname>LidSwitchIgnoreInhibited=</varname> defaults to <literal>yes</literal>. + This means that when <command>systemd-logind</command> is handling events by + itself (no low level inhibitor locks are taken by another application), the lid + switch does not respect suspend blockers by default, but the power and sleep keys + do.</para></listitem> </varlistentry> <varlistentry> @@ -318,8 +328,9 @@ <listitem><para>Sets the maximum number of OS tasks each user may run concurrently. This controls the <varname>TasksMax=</varname> setting of the per-user slice unit, see <citerefentry><refentrytitle>systemd.resource-control</refentrytitle><manvolnum>5</manvolnum></citerefentry> - for details. Defaults to 33%, which equals 10813 with the kernel's defaults on the host, but might be smaller - in OS containers.</para></listitem> + for details. If assigned the special value <literal>infinity</literal>, no tasks limit is applied. + Defaults to 33%, which equals 10813 with the kernel's defaults on the host, but might be smaller in + OS containers.</para></listitem> </varlistentry> <varlistentry> diff --git a/man/sd_journal_open.xml b/man/sd_journal_open.xml index 153af2387f..74e67023b5 100644 --- a/man/sd_journal_open.xml +++ b/man/sd_journal_open.xml @@ -129,10 +129,13 @@ <para><function>sd_journal_open_directory()</function> is similar to <function>sd_journal_open()</function> but takes an absolute directory path as argument. All journal files in this directory will be opened and interleaved - automatically. This call also takes a flags argument. The only flags parameter accepted by this call is - <constant>SD_JOURNAL_OS_ROOT</constant>. If specified, the journal files are searched below the usual - <filename>/var/log/journal</filename> and <filename>/run/log/journal</filename> relative to the specified path, - instead of directly beneath it.</para> + automatically. This call also takes a flags argument. The flags parameters accepted by this call are + <constant>SD_JOURNAL_OS_ROOT</constant>, <constant>SD_JOURNAL_SYSTEM</constant>, and + <constant>SD_JOURNAL_CURRENT_USER</constant>. If <constant>SD_JOURNAL_OS_ROOT</constant> is specified, journal + files are searched for below the usual <filename>/var/log/journal</filename> and + <filename>/run/log/journal</filename> relative to the specified path, instead of directly beneath it. + The other two flags limit which files are opened, the same as for <function>sd_journal_open()</function>. + </para> <para><function>sd_journal_open_directory_fd()</function> is similar to <function>sd_journal_open_directory()</function>, but takes a file descriptor referencing a directory in the file diff --git a/man/systemctl.xml b/man/systemctl.xml index 78607c9ba3..fde4f4f3bb 100644 --- a/man/systemctl.xml +++ b/man/systemctl.xml @@ -1681,20 +1681,15 @@ kobject-uevent 1 systemd-udevd-kernel.socket systemd-udevd.service <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> + <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 which is loaded from the actual host + volume. 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, equal to the empty string or identical to the path to the systemd binary, 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 phase.</para> </listitem> </varlistentry> diff --git a/man/systemd-inhibit.xml b/man/systemd-inhibit.xml index 9d85908f97..ce169960d8 100644 --- a/man/systemd-inhibit.xml +++ b/man/systemd-inhibit.xml @@ -61,7 +61,7 @@ <title>Description</title> <para><command>systemd-inhibit</command> may be used to execute a - program with a shutdown, sleep or idle inhibitor lock taken. The + program with a shutdown, sleep, or idle inhibitor lock taken. The lock will be acquired before the specified command line is executed and released afterwards.</para> diff --git a/man/systemd-logind.service.xml b/man/systemd-logind.service.xml index 5733e42cd1..f0bdb1c756 100644 --- a/man/systemd-logind.service.xml +++ b/man/systemd-logind.service.xml @@ -84,7 +84,7 @@ management</para></listitem> </itemizedlist> - <para>User sessions are registered in logind via the + <para>User sessions are registered with logind via the <citerefentry><refentrytitle>pam_systemd</refentrytitle><manvolnum>8</manvolnum></citerefentry> PAM module.</para> diff --git a/man/systemd.exec.xml b/man/systemd.exec.xml index bf82326096..bcedebd5bb 100644 --- a/man/systemd.exec.xml +++ b/man/systemd.exec.xml @@ -160,14 +160,14 @@ use. However, UID/GIDs are recycled after a unit is terminated. Care should be taken that any processes running as part of a unit for which dynamic users/groups are enabled do not leave files or directories owned by these users/groups around, as a different unit might get the same UID/GID assigned later on, and thus gain access to - these files or directories. If <varname>DynamicUser=</varname> is enabled, <varname>PrivateTmp=</varname> is - implied. This ensures that the lifetime of temporary files created by the executed processes is bound to the - runtime of the service, and hence the lifetime of the dynamic user/group. Since <filename>/tmp</filename> and - <filename>/var/tmp</filename> are usually the only world-writable directories on a system this ensures that a - unit making use of dynamic user/group allocation cannot leave files around after unit termination. Use - <varname>RuntimeDirectory=</varname> (see below) in order to assign a writable runtime directory to a service, - owned by the dynamic user/group and removed automatically when the unit is terminated. Defaults to - off.</para></listitem> + these files or directories. If <varname>DynamicUser=</varname> is enabled, <varname>RemoveIPC=</varname> and + <varname>PrivateTmp=</varname> are implied. This ensures that the lifetime of IPC objects and temporary files + created by the executed processes is bound to the runtime of the service, and hence the lifetime of the dynamic + user/group. Since <filename>/tmp</filename> and <filename>/var/tmp</filename> are usually the only + world-writable directories on a system this ensures that a unit making use of dynamic user/group allocation + cannot leave files around after unit termination. Use <varname>RuntimeDirectory=</varname> (see below) in order + to assign a writable runtime directory to a service, owned by the dynamic user/group and removed automatically + when the unit is terminated. Defaults to off.</para></listitem> </varlistentry> <varlistentry> @@ -186,6 +186,18 @@ </varlistentry> <varlistentry> + <term><varname>RemoveIPC=</varname></term> + + <listitem><para>Takes a boolean parameter. If set, all System V and POSIX IPC objects owned by the user and + group the processes of this unit are run as are removed when the unit is stopped. This setting only has an + effect if at least one of <varname>User=</varname>, <varname>Group=</varname> and + <varname>DynamicUser=</varname> are used. It has no effect on IPC objects owned by the root user. Specifically, + this removes System V semaphores, as well as System V and POSIX shared memory segments and message queues. If + multiple units use the same user or group the IPC objects are removed when the last of these units is + stopped. This setting is implied if <varname>DynamicUser=</varname> is set.</para></listitem> + </varlistentry> + + <varlistentry> <term><varname>Nice=</varname></term> <listitem><para>Sets the default nice level (scheduling @@ -920,27 +932,19 @@ <varlistentry> <term><varname>PrivateTmp=</varname></term> - <listitem><para>Takes a boolean argument. If true, sets up a - new file system namespace for the executed processes and - mounts private <filename>/tmp</filename> and - <filename>/var/tmp</filename> directories inside it that is - not shared by processes outside of the namespace. This is - useful to secure access to temporary files of the process, but - makes sharing between processes via <filename>/tmp</filename> - or <filename>/var/tmp</filename> impossible. If this is - enabled, all temporary files created by a service in these - directories will be removed after the service is stopped. - Defaults to false. It is possible to run two or more units - within the same private <filename>/tmp</filename> and - <filename>/var/tmp</filename> namespace by using the + <listitem><para>Takes a boolean argument. If true, sets up a new file system namespace for the executed + processes and mounts private <filename>/tmp</filename> and <filename>/var/tmp</filename> directories inside it + that is not shared by processes outside of the namespace. This is useful to secure access to temporary files of + the process, but makes sharing between processes via <filename>/tmp</filename> or <filename>/var/tmp</filename> + impossible. If this is enabled, all temporary files created by a service in these directories will be removed + after the service is stopped. Defaults to false. It is possible to run two or more units within the same + private <filename>/tmp</filename> and <filename>/var/tmp</filename> namespace by using the <varname>JoinsNamespaceOf=</varname> directive, see - <citerefentry><refentrytitle>systemd.unit</refentrytitle><manvolnum>5</manvolnum></citerefentry> - for details. Note that using this setting will disconnect - propagation of mounts from the service to the host - (propagation in the opposite direction continues to work). - This means that this setting may not be used for services - which shall be able to install mount points in the main mount - namespace.</para></listitem> + <citerefentry><refentrytitle>systemd.unit</refentrytitle><manvolnum>5</manvolnum></citerefentry> for + details. Note that using this setting will disconnect propagation of mounts from the service to the host + (propagation in the opposite direction continues to work). This means that this setting may not be used for + services which shall be able to install mount points in the main mount namespace. This setting is implied if + <varname>DynamicUser=</varname> is set.</para></listitem> </varlistentry> <varlistentry> diff --git a/man/systemd.resource-control.xml b/man/systemd.resource-control.xml index 0e98ca78b8..84dbfa2ff3 100644 --- a/man/systemd.resource-control.xml +++ b/man/systemd.resource-control.xml @@ -106,13 +106,21 @@ <para> <variablelist> + <varlistentry> - <term><option>IO</option></term> + <term><option>CPU</option></term> <listitem> - <para><varname>IO</varname> prefixed settings are superset of and replace <varname>BlockIO</varname> - prefixed ones. On unified hierarchy, IO resource control also applies to buffered writes.</para> + <para>Due to the lack of consensus in the kernel community, the CPU controller support on the unified + cgroup hierarchy requires out-of-tree kernel patches. See <ulink + url="https://git.kernel.org/cgit/linux/kernel/git/tj/cgroup.git/tree/Documentation/cgroup-v2-cpu.txt?h=cgroup-v2-cpu">cgroup-v2-cpu.txt</ulink>.</para> + + <para><varname>CPUWeight=</varname> and <varname>StartupCPUWeight=</varname> replace + <varname>CPUShares=</varname> and <varname>StartupCPUShares=</varname>, respectively.</para> + + <para>The <literal>cpuacct</literal> controller does not exist separately on the unified hierarchy.</para> </listitem> </varlistentry> + <varlistentry> <term><option>Memory</option></term> <listitem> @@ -120,6 +128,15 @@ and <varname>MemoryHigh=</varname> are effective only on unified hierarchy.</para> </listitem> </varlistentry> + + <varlistentry> + <term><option>IO</option></term> + <listitem> + <para><varname>IO</varname> prefixed settings are superset of and replace <varname>BlockIO</varname> + prefixed ones. On unified hierarchy, IO resource control also applies to buffered writes.</para> + </listitem> + </varlistentry> + </variablelist> </para> @@ -160,30 +177,49 @@ </varlistentry> <varlistentry> + <term><varname>CPUWeight=<replaceable>weight</replaceable></varname></term> + <term><varname>StartupCPUWeight=<replaceable>weight</replaceable></varname></term> + + <listitem> + <para>Assign the specified CPU time weight to the processes executed, if the unified control group hierarchy + is used on the system. These options take an integer value and control the <literal>cpu.weight</literal> + control group attribute. The allowed range is 1 to 10000. Defaults to 100. For details about this control + group attribute, see <ulink + url="https://www.kernel.org/doc/Documentation/cgroup-v2.txt">cgroup-v2.txt</ulink> and <ulink + url="https://www.kernel.org/doc/Documentation/scheduler/sched-design-CFS.txt">sched-design-CFS.txt</ulink>. + The available CPU time is split up among all units within one slice relative to their CPU time weight.</para> + + <para>While <varname>StartupCPUWeight=</varname> only applies to the startup phase of the system, + <varname>CPUWeight=</varname> applies to normal runtime of the system, and if the former is not set also to + the startup phase. Using <varname>StartupCPUWeight=</varname> allows prioritizing specific services at + boot-up differently than during normal runtime.</para> + + <para>Implies <literal>CPUAccounting=true</literal>.</para> + + <para>These settings are supported only if the unified control group hierarchy is used.</para> + </listitem> + </varlistentry> + + <varlistentry> <term><varname>CPUShares=<replaceable>weight</replaceable></varname></term> <term><varname>StartupCPUShares=<replaceable>weight</replaceable></varname></term> <listitem> - <para>Assign the specified CPU time share weight to the - processes executed. These options take an integer value and - control the <literal>cpu.shares</literal> control group - attribute. The allowed range is 2 to 262144. Defaults to - 1024. For details about this control group attribute, see - <ulink + <para>Assign the specified CPU time share weight to the processes executed. These options take an integer + value and control the <literal>cpu.shares</literal> control group attribute. The allowed range is 2 to + 262144. Defaults to 1024. For details about this control group attribute, see <ulink url="https://www.kernel.org/doc/Documentation/scheduler/sched-design-CFS.txt">sched-design-CFS.txt</ulink>. - The available CPU time is split up among all units within - one slice relative to their CPU time share weight.</para> + The available CPU time is split up among all units within one slice relative to their CPU time share + weight.</para> - <para>While <varname>StartupCPUShares=</varname> only - applies to the startup phase of the system, - <varname>CPUShares=</varname> applies to normal runtime of - the system, and if the former is not set also to the startup - phase. Using <varname>StartupCPUShares=</varname> allows - prioritizing specific services at boot-up differently than - during normal runtime.</para> + <para>While <varname>StartupCPUShares=</varname> only applies to the startup phase of the system, + <varname>CPUShares=</varname> applies to normal runtime of the system, and if the former is not set also to + the startup phase. Using <varname>StartupCPUShares=</varname> allows prioritizing specific services at + boot-up differently than during normal runtime.</para> - <para>These options imply - <literal>CPUAccounting=true</literal>.</para> + <para>Implies <literal>CPUAccounting=true</literal>.</para> + + <para>These settings are supported only if the legacy control group hierarchy is used.</para> </listitem> </varlistentry> @@ -191,22 +227,20 @@ <term><varname>CPUQuota=</varname></term> <listitem> - <para>Assign the specified CPU time quota to the processes - executed. Takes a percentage value, suffixed with "%". The - percentage specifies how much CPU time the unit shall get at - maximum, relative to the total CPU time available on one - CPU. Use values > 100% for allotting CPU time on more than - one CPU. This controls the - <literal>cpu.cfs_quota_us</literal> control group - attribute. For details about this control group attribute, - see <ulink + <para>Assign the specified CPU time quota to the processes executed. Takes a percentage value, suffixed with + "%". The percentage specifies how much CPU time the unit shall get at maximum, relative to the total CPU time + available on one CPU. Use values > 100% for allotting CPU time on more than one CPU. This controls the + <literal>cpu.max</literal> attribute on the unified control group hierarchy and + <literal>cpu.cfs_quota_us</literal> on legacy. For details about these control group attributes, see <ulink + url="https://www.kernel.org/doc/Documentation/cgroup-v2.txt">cgroup-v2.txt</ulink> and <ulink url="https://www.kernel.org/doc/Documentation/scheduler/sched-design-CFS.txt">sched-design-CFS.txt</ulink>.</para> - <para>Example: <varname>CPUQuota=20%</varname> ensures that - the executed processes will never get more than 20% CPU time - on one CPU.</para> + <para>Example: <varname>CPUQuota=20%</varname> ensures that the executed processes will never get more than + 20% CPU time on one CPU.</para> <para>Implies <literal>CPUAccounting=true</literal>.</para> + + <para>This setting is supported on both unified and legacy control group hierarchies.</para> </listitem> </varlistentry> diff --git a/rules/60-persistent-storage.rules b/rules/60-persistent-storage.rules index d7bbbf9866..c13d05cdb1 100644 --- a/rules/60-persistent-storage.rules +++ b/rules/60-persistent-storage.rules @@ -19,6 +19,12 @@ ENV{DEVTYPE}=="partition", IMPORT{parent}="ID_*" KERNEL=="nvme*[0-9]n*[0-9]", ATTR{wwid}=="?*", SYMLINK+="disk/by-id/nvme-$attr{wwid}" KERNEL=="nvme*[0-9]n*[0-9]p*[0-9]", ENV{DEVTYPE}=="partition", ATTRS{wwid}=="?*", SYMLINK+="disk/by-id/nvme-$attr{wwid}-part%n" +KERNEL=="nvme*[0-9]n*[0-9]", ENV{DEVTYPE}=="disk", ATTRS{serial}=="?*", ENV{ID_SERIAL_SHORT}="$attr{serial}" +KERNEL=="nvme*[0-9]n*[0-9]", ENV{DEVTYPE}=="disk", ATTRS{model}=="?*", ENV{ID_SERIAL_SHORT}=="?*", ENV{ID_SERIAL}="$attr{model}_$env{ID_SERIAL_SHORT}", SYMLINK+="disk/by-id/nvme-$env{ID_SERIAL}" + +KERNEL=="nvme*[0-9]n*[0-9]p*[0-9]", ENV{DEVTYPE}=="partition", ATTRS{serial}=="?*", ENV{ID_SERIAL_SHORT}="$attr{serial}" +KERNEL=="nvme*[0-9]n*[0-9]p*[0-9]", ENV{DEVTYPE}=="partition", ATTRS{model}=="?*", ENV{ID_SERIAL_SHORT}=="?*", ENV{ID_SERIAL}="$attr{model}_$env{ID_SERIAL_SHORT}", SYMLINK+="disk/by-id/nvme-$env{ID_SERIAL}-part%n" + # virtio-blk KERNEL=="vd*[!0-9]", ATTRS{serial}=="?*", ENV{ID_SERIAL}="$attr{serial}", SYMLINK+="disk/by-id/virtio-$env{ID_SERIAL}" KERNEL=="vd*[0-9]", ATTRS{serial}=="?*", ENV{ID_SERIAL}="$attr{serial}", SYMLINK+="disk/by-id/virtio-$env{ID_SERIAL}-part%n" diff --git a/shell-completion/zsh/_journalctl b/shell-completion/zsh/_journalctl index 2271f7fa9c..ef67fcf2a0 100644 --- a/shell-completion/zsh/_journalctl +++ b/shell-completion/zsh/_journalctl @@ -33,7 +33,7 @@ _journal_none() { _journal_fields() { local -a _fields cmd - cmd=("journalctl" "-F ${@[-1]}" "2>/dev/null" ) + cmd=("journalctl $_sys_service_mgr" "-F ${@[-1]}" "2>/dev/null" ) _fields=$(_call_program fields $cmd[@]) _fields=${_fields//'\'/'\\'} _fields=${_fields//':'/'\:'} @@ -51,8 +51,31 @@ _journal_boots() { "bootid:boot ids:compadd -a _bootid" } -local -a _modes; _modes=("--user" "--system") -local _sys_service_mgr=${${words:*_modes}[(R)(${(j.|.)_modes})]:---system} +# Build arguments for "journalctl" to be used in completion. +# Use both --user and --system modes, they are not exclusive. +local -a _modes; _modes=(--user --system) +local -a _modes_with_arg; _modes_with_arg=(--directory -D --file -M --machine --root) +typeset -a _sys_service_mgr +local w k v i=0 n=$#words +while (( i++ < n )); do + w=$words[$i] + if (( $_modes[(I)$w] )); then + _sys_service_mgr+=($w) + else + # Handle options with arguments. "--key=value" and "--key value". + k=${w%%=*} + if (( ${_modes_with_arg[(I)$k]} )); then + v=${w#*=} + if [[ "$k" != "$w" ]]; then + # "--key=value" style. + _sys_service_mgr+=($w) + else + # "--key value" style. + _sys_service_mgr+=($w ${words[((++i))]}) + fi + fi + fi +done _arguments -s \ {-h,--help}'[Show this help]' \ '--version[Show package version]' \ @@ -82,10 +105,10 @@ _arguments -s \ {-F,--field=}'[List all values a certain field takes]:Fields:_list_fields' \ '--system[Show system and kernel messages]' \ '--user[Show messages from user services]' \ - {-M+,--machine=}'[Operate on local container]:machines:_sd_machines' \ - {-D+,--directory=}'[Show journal files from directory]:directories:_directories' \ - '--file=[Operate on specified journal files]:file:_files' \ - '--root=[Operate on catalog hierarchy under specified directory]:directories:_directories' \ + '(--directory -D -M --machine --root --file)'{-M+,--machine=}'[Operate on local container]:machines:_sd_machines' \ + '(--directory -D -M --machine --root --file)'{-D+,--directory=}'[Show journal files from directory]:directories:_directories' \ + '(--directory -D -M --machine --root --file)''--root=[Operate on catalog hierarchy under specified directory]:directories:_directories' \ + '(--directory -D -M --machine --root)--file=[Operate on specified journal files]:file:_files' \ '--new-id128[Generate a new 128 Bit ID]' \ '--header[Show journal header information]' \ '--disk-usage[Show total disk usage]' \ diff --git a/shell-completion/zsh/_systemctl.in b/shell-completion/zsh/_systemctl.in index 69f643303d..b525286932 100644 --- a/shell-completion/zsh/_systemctl.in +++ b/shell-completion/zsh/_systemctl.in @@ -351,8 +351,10 @@ _job_modes() { _values -s , "${_modes[@]}" } +# Build arguments for "systemctl" to be used in completion. local -a _modes; _modes=("--user" "--system") -local _sys_service_mgr=${${words:*_modes}[(R)(${(j.|.)_modes})]:---system} +# Use the last mode (they are exclusive and the last one is used). +local _sys_service_mgr=${${words:*_modes}[(R)(${(j.|.)_modes})]} _arguments -s \ {-h,--help}'[Show help]' \ '--version[Show package version]' \ diff --git a/src/basic/cgroup-util.c b/src/basic/cgroup-util.c index 302b958d0d..25ef8a5c76 100644 --- a/src/basic/cgroup-util.c +++ b/src/basic/cgroup-util.c @@ -1905,6 +1905,49 @@ int cg_get_attribute(const char *controller, const char *path, const char *attri return read_one_line_file(p, ret); } +int cg_get_keyed_attribute(const char *controller, const char *path, const char *attribute, const char **keys, char **values) { + _cleanup_free_ char *filename = NULL, *content = NULL; + char *line, *p; + int i, r; + + for (i = 0; keys[i]; i++) + values[i] = NULL; + + r = cg_get_path(controller, path, attribute, &filename); + if (r < 0) + return r; + + r = read_full_file(filename, &content, NULL); + if (r < 0) + return r; + + p = content; + while ((line = strsep(&p, "\n"))) { + char *key; + + key = strsep(&line, " "); + + for (i = 0; keys[i]; i++) { + if (streq(key, keys[i])) { + values[i] = strdup(line); + break; + } + } + } + + for (i = 0; keys[i]; i++) { + if (!values[i]) { + for (i = 0; keys[i]; i++) { + free(values[i]); + values[i] = NULL; + } + return -ENOENT; + } + } + + return 0; +} + int cg_create_everywhere(CGroupMask supported, CGroupMask mask, const char *path) { CGroupController c; int r, unified; @@ -2101,10 +2144,10 @@ int cg_mask_supported(CGroupMask *ret) { mask |= CGROUP_CONTROLLER_TO_MASK(v); } - /* Currently, we only support the memory, io and pids + /* Currently, we support the cpu, memory, io and pids * controller in the unified hierarchy, mask * everything else off. */ - mask &= CGROUP_MASK_MEMORY | CGROUP_MASK_IO | CGROUP_MASK_PIDS; + mask &= CGROUP_MASK_CPU | CGROUP_MASK_MEMORY | CGROUP_MASK_IO | CGROUP_MASK_PIDS; } else { CGroupController c; diff --git a/src/basic/cgroup-util.h b/src/basic/cgroup-util.h index ec5c715987..f1617a16be 100644 --- a/src/basic/cgroup-util.h +++ b/src/basic/cgroup-util.h @@ -112,6 +112,10 @@ static inline bool CGROUP_BLKIO_WEIGHT_IS_OK(uint64_t x) { (x >= CGROUP_BLKIO_WEIGHT_MIN && x <= CGROUP_BLKIO_WEIGHT_MAX); } +/* Default resource limits */ +#define DEFAULT_TASKS_MAX_PERCENTAGE 15U /* 15% of PIDs, 4915 on default settings */ +#define DEFAULT_USER_TASKS_MAX_PERCENTAGE 33U /* 33% of PIDs, 10813 on default settings */ + /* * General rules: * @@ -169,6 +173,7 @@ int cg_create_and_attach(const char *controller, const char *path, pid_t pid); int cg_set_attribute(const char *controller, const char *path, const char *attribute, const char *value); int cg_get_attribute(const char *controller, const char *path, const char *attribute, char **ret); +int cg_get_keyed_attribute(const char *controller, const char *path, const char *attribute, const char **keys, char **values); int cg_set_group_access(const char *controller, const char *path, mode_t mode, uid_t uid, gid_t gid); int cg_set_task_access(const char *controller, const char *path, mode_t mode, uid_t uid, gid_t gid); diff --git a/src/basic/hostname-util.c b/src/basic/hostname-util.c index 13c3bb6446..e44a357287 100644 --- a/src/basic/hostname-util.c +++ b/src/basic/hostname-util.c @@ -163,7 +163,6 @@ char* hostname_cleanup(char *s) { *(d++) = *p; dot = false; } - } if (dot && d > s) diff --git a/src/basic/user-util.h b/src/basic/user-util.h index 36f71fb004..f569363811 100644 --- a/src/basic/user-util.h +++ b/src/basic/user-util.h @@ -20,6 +20,7 @@ ***/ #include <stdbool.h> +#include <stdint.h> #include <sys/types.h> #include <unistd.h> @@ -57,8 +58,19 @@ int take_etc_passwd_lock(const char *root); #define UID_INVALID ((uid_t) -1) #define GID_INVALID ((gid_t) -1) -/* The following macros add 1 when converting things, since UID 0 is a - * valid UID, while the pointer NULL is special */ +/* Let's pick a UIDs within the 16bit range, so that we are compatible with containers using 16bit + * user namespacing. At least on Fedora normal users are allocated until UID 60000, hence do not + * allocate from below this. Also stay away from the upper end of the range as that is often used + * for overflow/nobody users. */ +#define DYNAMIC_UID_MIN ((uid_t) UINT32_C(0x0000EF00)) +#define DYNAMIC_UID_MAX ((uid_t) UINT32_C(0x0000FFEF)) + +static inline bool uid_is_dynamic(uid_t uid) { + return DYNAMIC_UID_MIN <= uid && uid <= DYNAMIC_UID_MAX; +} + +/* The following macros add 1 when converting things, since UID 0 is a valid UID, while the pointer + * NULL is special */ #define PTR_TO_UID(p) ((uid_t) (((uintptr_t) (p))-1)) #define UID_TO_PTR(u) ((void*) (((uintptr_t) (u))+1)) diff --git a/src/cgtop/cgtop.c b/src/cgtop/cgtop.c index c67b328b38..6045ae0437 100644 --- a/src/cgtop/cgtop.c +++ b/src/cgtop/cgtop.c @@ -208,24 +208,47 @@ static int process( if (g->n_tasks > 0) g->n_tasks_valid = true; - } else if (streq(controller, "cpuacct") && cg_unified() <= 0) { + } else if (streq(controller, "cpu") || streq(controller, "cpuacct")) { _cleanup_free_ char *p = NULL, *v = NULL; uint64_t new_usage; nsec_t timestamp; - r = cg_get_path(controller, path, "cpuacct.usage", &p); - if (r < 0) - return r; + if (cg_unified() > 0) { + const char *keys[] = { "usage_usec", NULL }; + _cleanup_free_ char *val = NULL; - r = read_one_line_file(p, &v); - if (r == -ENOENT) - return 0; - if (r < 0) - return r; + if (!streq(controller, "cpu")) + return 0; - r = safe_atou64(v, &new_usage); - if (r < 0) - return r; + r = cg_get_keyed_attribute("cpu", path, "cpu.stat", keys, &val); + if (r == -ENOENT) + return 0; + if (r < 0) + return r; + + r = safe_atou64(val, &new_usage); + if (r < 0) + return r; + + new_usage *= NSEC_PER_USEC; + } else { + if (!streq(controller, "cpuacct")) + return 0; + + r = cg_get_path(controller, path, "cpuacct.usage", &p); + if (r < 0) + return r; + + r = read_one_line_file(p, &v); + if (r == -ENOENT) + return 0; + if (r < 0) + return r; + + r = safe_atou64(v, &new_usage); + if (r < 0) + return r; + } timestamp = now_nsec(CLOCK_MONOTONIC); @@ -449,6 +472,9 @@ static int refresh(const char *root, Hashmap *a, Hashmap *b, unsigned iteration) r = refresh_one(SYSTEMD_CGROUP_CONTROLLER, root, a, b, iteration, 0, NULL); if (r < 0) return r; + r = refresh_one("cpu", root, a, b, iteration, 0, NULL); + if (r < 0) + return r; r = refresh_one("cpuacct", root, a, b, iteration, 0, NULL); if (r < 0) return r; diff --git a/src/core/cgroup.c b/src/core/cgroup.c index c19e43f571..ca3c3366f3 100644 --- a/src/core/cgroup.c +++ b/src/core/cgroup.c @@ -57,9 +57,12 @@ void cgroup_context_init(CGroupContext *c) { /* Initialize everything to the kernel defaults, assuming the * structure is preinitialized to 0 */ + c->cpu_weight = CGROUP_WEIGHT_INVALID; + c->startup_cpu_weight = CGROUP_WEIGHT_INVALID; + c->cpu_quota_per_sec_usec = USEC_INFINITY; + c->cpu_shares = CGROUP_CPU_SHARES_INVALID; c->startup_cpu_shares = CGROUP_CPU_SHARES_INVALID; - c->cpu_quota_per_sec_usec = USEC_INFINITY; c->memory_high = CGROUP_LIMIT_MAX; c->memory_max = CGROUP_LIMIT_MAX; @@ -158,6 +161,8 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { "%sBlockIOAccounting=%s\n" "%sMemoryAccounting=%s\n" "%sTasksAccounting=%s\n" + "%sCPUWeight=%" PRIu64 "\n" + "%sStartupCPUWeight=%" PRIu64 "\n" "%sCPUShares=%" PRIu64 "\n" "%sStartupCPUShares=%" PRIu64 "\n" "%sCPUQuotaPerSecSec=%s\n" @@ -177,6 +182,8 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { prefix, yes_no(c->blockio_accounting), prefix, yes_no(c->memory_accounting), prefix, yes_no(c->tasks_accounting), + prefix, c->cpu_weight, + prefix, c->startup_cpu_weight, prefix, c->cpu_shares, prefix, c->startup_cpu_shares, prefix, format_timespan(u, sizeof(u), c->cpu_quota_per_sec_usec, 1), @@ -382,6 +389,95 @@ fail: return -errno; } +static bool cgroup_context_has_cpu_weight(CGroupContext *c) { + return c->cpu_weight != CGROUP_WEIGHT_INVALID || + c->startup_cpu_weight != CGROUP_WEIGHT_INVALID; +} + +static bool cgroup_context_has_cpu_shares(CGroupContext *c) { + return c->cpu_shares != CGROUP_CPU_SHARES_INVALID || + c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID; +} + +static uint64_t cgroup_context_cpu_weight(CGroupContext *c, ManagerState state) { + if (IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && + c->startup_cpu_weight != CGROUP_WEIGHT_INVALID) + return c->startup_cpu_weight; + else if (c->cpu_weight != CGROUP_WEIGHT_INVALID) + return c->cpu_weight; + else + return CGROUP_WEIGHT_DEFAULT; +} + +static uint64_t cgroup_context_cpu_shares(CGroupContext *c, ManagerState state) { + if (IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && + c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID) + return c->startup_cpu_shares; + else if (c->cpu_shares != CGROUP_CPU_SHARES_INVALID) + return c->cpu_shares; + else + return CGROUP_CPU_SHARES_DEFAULT; +} + +static void cgroup_apply_unified_cpu_config(Unit *u, uint64_t weight, uint64_t quota) { + char buf[MAX(DECIMAL_STR_MAX(uint64_t) + 1, (DECIMAL_STR_MAX(usec_t) + 1) * 2)]; + int r; + + xsprintf(buf, "%" PRIu64 "\n", weight); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.weight", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.weight: %m"); + + if (quota != USEC_INFINITY) + xsprintf(buf, USEC_FMT " " USEC_FMT "\n", + quota * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC, CGROUP_CPU_QUOTA_PERIOD_USEC); + else + xsprintf(buf, "max " USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC); + + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.max", buf); + + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.max: %m"); +} + +static void cgroup_apply_legacy_cpu_config(Unit *u, uint64_t shares, uint64_t quota) { + char buf[MAX(DECIMAL_STR_MAX(uint64_t), DECIMAL_STR_MAX(usec_t)) + 1]; + int r; + + xsprintf(buf, "%" PRIu64 "\n", shares); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.shares", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.shares: %m"); + + xsprintf(buf, USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.cfs_period_us", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.cfs_period_us: %m"); + + if (quota != USEC_INFINITY) { + xsprintf(buf, USEC_FMT "\n", quota * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.cfs_quota_us", buf); + } else + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.cfs_quota_us", "-1"); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.cfs_quota_us: %m"); +} + +static uint64_t cgroup_cpu_shares_to_weight(uint64_t shares) { + return CLAMP(shares * CGROUP_WEIGHT_DEFAULT / CGROUP_CPU_SHARES_DEFAULT, + CGROUP_WEIGHT_MIN, CGROUP_WEIGHT_MAX); +} + +static uint64_t cgroup_cpu_weight_to_shares(uint64_t weight) { + return CLAMP(weight * CGROUP_CPU_SHARES_DEFAULT / CGROUP_WEIGHT_DEFAULT, + CGROUP_CPU_SHARES_MIN, CGROUP_CPU_SHARES_MAX); +} + static bool cgroup_context_has_io_config(CGroupContext *c) { return c->io_accounting || c->io_weight != CGROUP_WEIGHT_INVALID || @@ -566,30 +662,42 @@ static void cgroup_context_apply(Unit *u, CGroupMask mask, ManagerState state) { * and missing cgroups, i.e. EROFS and ENOENT. */ if ((mask & CGROUP_MASK_CPU) && !is_root) { - char buf[MAX(DECIMAL_STR_MAX(uint64_t), DECIMAL_STR_MAX(usec_t)) + 1]; + bool has_weight = cgroup_context_has_cpu_weight(c); + bool has_shares = cgroup_context_has_cpu_shares(c); - sprintf(buf, "%" PRIu64 "\n", - IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID ? c->startup_cpu_shares : - c->cpu_shares != CGROUP_CPU_SHARES_INVALID ? c->cpu_shares : CGROUP_CPU_SHARES_DEFAULT); - r = cg_set_attribute("cpu", path, "cpu.shares", buf); - if (r < 0) - log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set cpu.shares: %m"); + if (cg_unified() > 0) { + uint64_t weight; - sprintf(buf, USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC); - r = cg_set_attribute("cpu", path, "cpu.cfs_period_us", buf); - if (r < 0) - log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set cpu.cfs_period_us: %m"); + if (has_weight) + weight = cgroup_context_cpu_weight(c, state); + else if (has_shares) { + uint64_t shares = cgroup_context_cpu_shares(c, state); - if (c->cpu_quota_per_sec_usec != USEC_INFINITY) { - sprintf(buf, USEC_FMT "\n", c->cpu_quota_per_sec_usec * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC); - r = cg_set_attribute("cpu", path, "cpu.cfs_quota_us", buf); - } else - r = cg_set_attribute("cpu", path, "cpu.cfs_quota_us", "-1"); - if (r < 0) - log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set cpu.cfs_quota_us: %m"); + weight = cgroup_cpu_shares_to_weight(shares); + + log_cgroup_compat(u, "Applying [Startup]CpuShares %" PRIu64 " as [Startup]CpuWeight %" PRIu64 " on %s", + shares, weight, path); + } else + weight = CGROUP_WEIGHT_DEFAULT; + + cgroup_apply_unified_cpu_config(u, weight, c->cpu_quota_per_sec_usec); + } else { + uint64_t shares; + + if (has_shares) + shares = cgroup_context_cpu_shares(c, state); + else if (has_weight) { + uint64_t weight = cgroup_context_cpu_weight(c, state); + + shares = cgroup_cpu_weight_to_shares(weight); + + log_cgroup_compat(u, "Applying [Startup]CpuWeight %" PRIu64 " as [Startup]CpuShares %" PRIu64 " on %s", + weight, shares, path); + } else + shares = CGROUP_CPU_SHARES_DEFAULT; + + cgroup_apply_legacy_cpu_config(u, shares, c->cpu_quota_per_sec_usec); + } } if (mask & CGROUP_MASK_IO) { @@ -844,7 +952,7 @@ static void cgroup_context_apply(Unit *u, CGroupMask mask, ManagerState state) { if ((mask & CGROUP_MASK_PIDS) && !is_root) { - if (c->tasks_max != (uint64_t) -1) { + if (c->tasks_max != CGROUP_LIMIT_MAX) { char buf[DECIMAL_STR_MAX(uint64_t) + 2]; sprintf(buf, "%" PRIu64 "\n", c->tasks_max); @@ -864,8 +972,8 @@ CGroupMask cgroup_context_get_mask(CGroupContext *c) { /* Figure out which controllers we need */ if (c->cpu_accounting || - c->cpu_shares != CGROUP_CPU_SHARES_INVALID || - c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID || + cgroup_context_has_cpu_weight(c) || + cgroup_context_has_cpu_shares(c) || c->cpu_quota_per_sec_usec != USEC_INFINITY) mask |= CGROUP_MASK_CPUACCT | CGROUP_MASK_CPU; @@ -1890,18 +1998,37 @@ static int unit_get_cpu_usage_raw(Unit *u, nsec_t *ret) { if (!u->cgroup_path) return -ENODATA; - if ((u->cgroup_realized_mask & CGROUP_MASK_CPUACCT) == 0) - return -ENODATA; + if (cg_unified() > 0) { + const char *keys[] = { "usage_usec", NULL }; + _cleanup_free_ char *val = NULL; + uint64_t us; - r = cg_get_attribute("cpuacct", u->cgroup_path, "cpuacct.usage", &v); - if (r == -ENOENT) - return -ENODATA; - if (r < 0) - return r; + if ((u->cgroup_realized_mask & CGROUP_MASK_CPU) == 0) + return -ENODATA; - r = safe_atou64(v, &ns); - if (r < 0) - return r; + r = cg_get_keyed_attribute("cpu", u->cgroup_path, "cpu.stat", keys, &val); + if (r < 0) + return r; + + r = safe_atou64(val, &us); + if (r < 0) + return r; + + ns = us * NSEC_PER_USEC; + } else { + if ((u->cgroup_realized_mask & CGROUP_MASK_CPUACCT) == 0) + return -ENODATA; + + r = cg_get_attribute("cpuacct", u->cgroup_path, "cpuacct.usage", &v); + if (r == -ENOENT) + return -ENODATA; + if (r < 0) + return r; + + r = safe_atou64(v, &ns); + if (r < 0) + return r; + } *ret = ns; return 0; @@ -1915,8 +2042,8 @@ int unit_get_cpu_usage(Unit *u, nsec_t *ret) { if (r < 0) return r; - if (ns > u->cpuacct_usage_base) - ns -= u->cpuacct_usage_base; + if (ns > u->cpu_usage_base) + ns -= u->cpu_usage_base; else ns = 0; @@ -1932,11 +2059,11 @@ int unit_reset_cpu_usage(Unit *u) { r = unit_get_cpu_usage_raw(u, &ns); if (r < 0) { - u->cpuacct_usage_base = 0; + u->cpu_usage_base = 0; return r; } - u->cpuacct_usage_base = ns; + u->cpu_usage_base = ns; return 0; } diff --git a/src/core/cgroup.h b/src/core/cgroup.h index a57403e79f..2fe9cc4039 100644 --- a/src/core/cgroup.h +++ b/src/core/cgroup.h @@ -89,6 +89,10 @@ struct CGroupContext { bool tasks_accounting; /* For unified hierarchy */ + uint64_t cpu_weight; + uint64_t startup_cpu_weight; + usec_t cpu_quota_per_sec_usec; + uint64_t io_weight; uint64_t startup_io_weight; LIST_HEAD(CGroupIODeviceWeight, io_device_weights); @@ -101,7 +105,6 @@ struct CGroupContext { /* For legacy hierarchies */ uint64_t cpu_shares; uint64_t startup_cpu_shares; - usec_t cpu_quota_per_sec_usec; uint64_t blockio_weight; uint64_t startup_blockio_weight; diff --git a/src/core/dbus-cgroup.c b/src/core/dbus-cgroup.c index 85b0c86a2f..2ca80d2996 100644 --- a/src/core/dbus-cgroup.c +++ b/src/core/dbus-cgroup.c @@ -210,6 +210,8 @@ const sd_bus_vtable bus_cgroup_vtable[] = { SD_BUS_VTABLE_START(0), SD_BUS_PROPERTY("Delegate", "b", bus_property_get_bool, offsetof(CGroupContext, delegate), 0), SD_BUS_PROPERTY("CPUAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, cpu_accounting), 0), + SD_BUS_PROPERTY("CPUWeight", "t", NULL, offsetof(CGroupContext, cpu_weight), 0), + SD_BUS_PROPERTY("StartupCPUWeight", "t", NULL, offsetof(CGroupContext, startup_cpu_weight), 0), SD_BUS_PROPERTY("CPUShares", "t", NULL, offsetof(CGroupContext, cpu_shares), 0), SD_BUS_PROPERTY("StartupCPUShares", "t", NULL, offsetof(CGroupContext, startup_cpu_shares), 0), SD_BUS_PROPERTY("CPUQuotaPerSecUSec", "t", bus_property_get_usec, offsetof(CGroupContext, cpu_quota_per_sec_usec), 0), @@ -303,6 +305,50 @@ int bus_cgroup_set_property( return 1; + } else if (streq(name, "CPUWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "CPUWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->cpu_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (weight == CGROUP_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "CPUWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "CPUWeight=%" PRIu64, weight); + } + + return 1; + + } else if (streq(name, "StartupCPUWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupCPUWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->startup_cpu_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (weight == CGROUP_CPU_SHARES_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupCPUWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupCPUWeight=%" PRIu64, weight); + } + + return 1; + } else if (streq(name, "CPUShares")) { uint64_t shares; diff --git a/src/core/dbus-execute.c b/src/core/dbus-execute.c index e35d3ccd2e..7e33a2d201 100644 --- a/src/core/dbus-execute.c +++ b/src/core/dbus-execute.c @@ -695,6 +695,7 @@ const sd_bus_vtable bus_exec_vtable[] = { SD_BUS_PROPERTY("User", "s", NULL, offsetof(ExecContext, user), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("Group", "s", NULL, offsetof(ExecContext, group), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("DynamicUser", "b", bus_property_get_bool, offsetof(ExecContext, dynamic_user), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RemoveIPC", "b", bus_property_get_bool, offsetof(ExecContext, remove_ipc), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("SupplementaryGroups", "as", NULL, offsetof(ExecContext, supplementary_groups), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("PAMName", "s", NULL, offsetof(ExecContext, pam_name), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("ReadWriteDirectories", "as", NULL, offsetof(ExecContext, read_write_paths), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), @@ -1071,7 +1072,7 @@ int bus_exec_context_set_transient_property( "IgnoreSIGPIPE", "TTYVHangup", "TTYReset", "PrivateTmp", "PrivateDevices", "PrivateNetwork", "PrivateUsers", "NoNewPrivileges", "SyslogLevelPrefix", "MemoryDenyWriteExecute", - "RestrictRealtime", "DynamicUser")) { + "RestrictRealtime", "DynamicUser", "RemoveIPC")) { int b; r = sd_bus_message_read(message, "b", &b); @@ -1103,6 +1104,8 @@ int bus_exec_context_set_transient_property( c->restrict_realtime = b; else if (streq(name, "DynamicUser")) c->dynamic_user = b; + else if (streq(name, "RemoveIPC")) + c->remove_ipc = b; unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, yes_no(b)); } diff --git a/src/core/dbus-mount.c b/src/core/dbus-mount.c index b4bbee0648..3c6bda4073 100644 --- a/src/core/dbus-mount.c +++ b/src/core/dbus-mount.c @@ -117,6 +117,8 @@ const sd_bus_vtable bus_mount_vtable[] = { SD_BUS_PROPERTY("DirectoryMode", "u", bus_property_get_mode, offsetof(Mount, directory_mode), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("SloppyOptions", "b", bus_property_get_bool, offsetof(Mount, sloppy_options), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Mount, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), BUS_EXEC_COMMAND_VTABLE("ExecMount", offsetof(Mount, exec_command[MOUNT_EXEC_MOUNT]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_VTABLE("ExecUnmount", offsetof(Mount, exec_command[MOUNT_EXEC_UNMOUNT]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_VTABLE("ExecRemount", offsetof(Mount, exec_command[MOUNT_EXEC_REMOUNT]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), diff --git a/src/core/dbus-service.c b/src/core/dbus-service.c index fab3677a01..3c55e0f7fe 100644 --- a/src/core/dbus-service.c +++ b/src/core/dbus-service.c @@ -50,11 +50,6 @@ const sd_bus_vtable bus_service_vtable[] = { SD_BUS_PROPERTY("RuntimeMaxUSec", "t", bus_property_get_usec, offsetof(Service, runtime_max_usec), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("WatchdogUSec", "t", bus_property_get_usec, offsetof(Service, watchdog_usec), SD_BUS_VTABLE_PROPERTY_CONST), BUS_PROPERTY_DUAL_TIMESTAMP("WatchdogTimestamp", offsetof(Service, watchdog_timestamp), 0), - /* The following four are obsolete, and thus marked hidden here. They moved into the Unit interface */ - SD_BUS_PROPERTY("StartLimitInterval", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), - SD_BUS_PROPERTY("StartLimitBurst", "u", bus_property_get_unsigned, offsetof(Unit, start_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), - SD_BUS_PROPERTY("StartLimitAction", "s", property_get_failure_action, offsetof(Unit, start_limit_action), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), - SD_BUS_PROPERTY("RebootArgument", "s", NULL, offsetof(Unit, reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), SD_BUS_PROPERTY("FailureAction", "s", property_get_failure_action, offsetof(Service, failure_action), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("PermissionsStartOnly", "b", bus_property_get_bool, offsetof(Service, permissions_start_only), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("RootDirectoryStartOnly", "b", bus_property_get_bool, offsetof(Service, root_directory_start_only), SD_BUS_VTABLE_PROPERTY_CONST), @@ -70,6 +65,9 @@ const sd_bus_vtable bus_service_vtable[] = { SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Service, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), SD_BUS_PROPERTY("USBFunctionDescriptors", "s", NULL, offsetof(Service, usb_function_descriptors), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), SD_BUS_PROPERTY("USBFunctionStrings", "s", NULL, offsetof(Service, usb_function_strings), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_EXEC_STATUS_VTABLE("ExecMain", offsetof(Service, main_exec_status), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPre", offsetof(Service, exec_command[SERVICE_EXEC_START_PRE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStart", offsetof(Service, exec_command[SERVICE_EXEC_START]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), @@ -77,6 +75,12 @@ const sd_bus_vtable bus_service_vtable[] = { BUS_EXEC_COMMAND_LIST_VTABLE("ExecReload", offsetof(Service, exec_command[SERVICE_EXEC_RELOAD]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStop", offsetof(Service, exec_command[SERVICE_EXEC_STOP]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStopPost", offsetof(Service, exec_command[SERVICE_EXEC_STOP_POST]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + + /* The following four are obsolete, and thus marked hidden here. They moved into the Unit interface */ + SD_BUS_PROPERTY("StartLimitInterval", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("StartLimitBurst", "u", bus_property_get_unsigned, offsetof(Unit, start_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("StartLimitAction", "s", property_get_failure_action, offsetof(Unit, start_limit_action), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RebootArgument", "s", NULL, offsetof(Unit, reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), SD_BUS_VTABLE_END }; diff --git a/src/core/dbus-socket.c b/src/core/dbus-socket.c index 9a071a1355..21adb64e15 100644 --- a/src/core/dbus-socket.c +++ b/src/core/dbus-socket.c @@ -152,6 +152,8 @@ const sd_bus_vtable bus_socket_vtable[] = { SD_BUS_PROPERTY("SocketProtocol", "i", bus_property_get_int, offsetof(Socket, socket_protocol), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("TriggerLimitIntervalUSec", "t", bus_property_get_usec, offsetof(Socket, trigger_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("TriggerLimitBurst", "u", bus_property_get_unsigned, offsetof(Socket, trigger_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPre", offsetof(Socket, exec_command[SOCKET_EXEC_START_PRE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPost", offsetof(Socket, exec_command[SOCKET_EXEC_START_POST]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_LIST_VTABLE("ExecStopPre", offsetof(Socket, exec_command[SOCKET_EXEC_STOP_PRE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), diff --git a/src/core/dbus-swap.c b/src/core/dbus-swap.c index 292f8738c6..85a2c26b98 100644 --- a/src/core/dbus-swap.c +++ b/src/core/dbus-swap.c @@ -84,6 +84,8 @@ const sd_bus_vtable bus_swap_vtable[] = { SD_BUS_PROPERTY("TimeoutUSec", "t", bus_property_get_usec, offsetof(Swap, timeout_usec), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("ControlPID", "u", bus_property_get_pid, offsetof(Swap, control_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Swap, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), BUS_EXEC_COMMAND_VTABLE("ExecActivate", offsetof(Swap, exec_command[SWAP_EXEC_ACTIVATE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), BUS_EXEC_COMMAND_VTABLE("ExecDeactivate", offsetof(Swap, exec_command[SWAP_EXEC_DEACTIVATE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), SD_BUS_VTABLE_END diff --git a/src/core/dbus-unit.c b/src/core/dbus-unit.c index b55d2cf735..89e56a2e51 100644 --- a/src/core/dbus-unit.c +++ b/src/core/dbus-unit.c @@ -660,10 +660,6 @@ const sd_bus_vtable bus_unit_vtable[] = { SD_BUS_PROPERTY("PropagatesReloadTo", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_PROPAGATES_RELOAD_TO]), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("ReloadPropagatedFrom", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_RELOAD_PROPAGATED_FROM]), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("JoinsNamespaceOf", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_JOINS_NAMESPACE_OF]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("RequiresOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), - SD_BUS_PROPERTY("RequisiteOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), - SD_BUS_PROPERTY("RequiredByOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), - SD_BUS_PROPERTY("RequisiteOfOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), SD_BUS_PROPERTY("RequiresMountsFor", "as", NULL, offsetof(Unit, requires_mounts_for), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("Documentation", "as", NULL, offsetof(Unit, documentation), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("Description", "s", property_get_description, 0, SD_BUS_VTABLE_PROPERTY_CONST), @@ -705,7 +701,6 @@ const sd_bus_vtable bus_unit_vtable[] = { SD_BUS_PROPERTY("LoadError", "(ss)", property_get_load_error, 0, SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("Transient", "b", bus_property_get_bool, offsetof(Unit, transient), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("StartLimitIntervalSec", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("StartLimitInterval", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), /* obsolete alias name */ SD_BUS_PROPERTY("StartLimitBurst", "u", bus_property_get_unsigned, offsetof(Unit, start_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("StartLimitAction", "s", property_get_failure_action, offsetof(Unit, start_limit_action), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("RebootArgument", "s", NULL, offsetof(Unit, reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST), @@ -721,6 +716,12 @@ const sd_bus_vtable bus_unit_vtable[] = { SD_BUS_METHOD("ResetFailed", NULL, NULL, bus_unit_method_reset_failed, SD_BUS_VTABLE_UNPRIVILEGED), SD_BUS_METHOD("SetProperties", "ba(sv)", NULL, bus_unit_method_set_properties, SD_BUS_VTABLE_UNPRIVILEGED), + /* Obsolete properties or obsolete alias names */ + SD_BUS_PROPERTY("RequiresOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RequisiteOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RequiredByOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RequisiteOfOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("StartLimitInterval", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), SD_BUS_VTABLE_END }; diff --git a/src/core/dynamic-user.c b/src/core/dynamic-user.c index 8035bee231..310aaa94e1 100644 --- a/src/core/dynamic-user.c +++ b/src/core/dynamic-user.c @@ -31,14 +31,8 @@ #include "user-util.h" #include "fileio.h" -/* Let's pick a UIDs within the 16bit range, so that we are compatible with containers using 16bit user namespacing. At - * least on Fedora normal users are allocated until UID 60000, hence do not allocate from below this. Also stay away - * from the upper end of the range as that is often used for overflow/nobody users. */ -#define UID_PICK_MIN ((uid_t) UINT32_C(0x0000EF00)) -#define UID_PICK_MAX ((uid_t) UINT32_C(0x0000FFEF)) - /* Takes a value generated randomly or by hashing and turns it into a UID in the right range */ -#define UID_CLAMP_INTO_RANGE(rnd) (((uid_t) (rnd) % (UID_PICK_MAX - UID_PICK_MIN + 1)) + UID_PICK_MIN) +#define UID_CLAMP_INTO_RANGE(rnd) (((uid_t) (rnd) % (DYNAMIC_UID_MAX - DYNAMIC_UID_MIN + 1)) + DYNAMIC_UID_MIN) static DynamicUser* dynamic_user_free(DynamicUser *d) { if (!d) @@ -150,6 +144,45 @@ int dynamic_user_acquire(Manager *m, const char *name, DynamicUser** ret) { return 1; } +static int make_uid_symlinks(uid_t uid, const char *name, bool b) { + + char path1[strlen("/run/systemd/dynamic-uid/direct:") + DECIMAL_STR_MAX(uid_t) + 1]; + const char *path2; + int r = 0, k; + + /* Add direct additional symlinks for direct lookups of dynamic UIDs and their names by userspace code. The + * only reason we have this is because dbus-daemon cannot use D-Bus for resolving users and groups (since it + * would be its own client then). We hence keep these world-readable symlinks in place, so that the + * unprivileged dbus user can read the mappings when it needs them via these symlinks instead of having to go + * via the bus. Ideally, we'd use the lock files we keep for this anyway, but we can't since we use BSD locks + * on them and as those may be taken by any user with read access we can't make them world-readable. */ + + xsprintf(path1, "/run/systemd/dynamic-uid/direct:" UID_FMT, uid); + if (unlink(path1) < 0 && errno != ENOENT) + r = -errno; + + if (b && symlink(name, path1) < 0) { + k = log_warning_errno(errno, "Failed to symlink \"%s\": %m", path1); + if (r == 0) + r = k; + } + + path2 = strjoina("/run/systemd/dynamic-uid/direct:", name); + if (unlink(path2) < 0 && errno != ENOENT) { + k = -errno; + if (r == 0) + r = k; + } + + if (b && symlink(path1 + strlen("/run/systemd/dynamic-uid/direct:"), path2) < 0) { + k = log_warning_errno(errno, "Failed to symlink \"%s\": %m", path2); + if (r == 0) + r = k; + } + + return r; +} + static int pick_uid(const char *name, uid_t *ret_uid) { static const uint8_t hash_key[] = { @@ -175,7 +208,7 @@ static int pick_uid(const char *name, uid_t *ret_uid) { if (--n_tries <= 0) /* Give up retrying eventually */ return -EBUSY; - if (candidate < UID_PICK_MIN || candidate > UID_PICK_MAX) + if (!uid_is_dynamic(candidate)) goto next; xsprintf(lock_path, "/run/systemd/dynamic-uid/" UID_FMT, candidate); @@ -223,6 +256,7 @@ static int pick_uid(const char *name, uid_t *ret_uid) { } (void) ftruncate(lock_fd, l); + (void) make_uid_symlinks(candidate, name, true); /* also add direct lookup symlinks */ *ret_uid = candidate; r = lock_fd; @@ -324,14 +358,16 @@ static int dynamic_user_push(DynamicUser *d, uid_t uid, int lock_fd) { return 0; } -static void unlink_uid_lock(int lock_fd, uid_t uid) { +static void unlink_uid_lock(int lock_fd, uid_t uid, const char *name) { char lock_path[strlen("/run/systemd/dynamic-uid/") + DECIMAL_STR_MAX(uid_t) + 1]; if (lock_fd < 0) return; xsprintf(lock_path, "/run/systemd/dynamic-uid/" UID_FMT, uid); - (void) unlink_noerrno(lock_path); + (void) unlink(lock_path); + + (void) make_uid_symlinks(uid, name, false); /* remove direct lookup symlinks */ } int dynamic_user_realize(DynamicUser *d, uid_t *ret) { @@ -399,7 +435,7 @@ int dynamic_user_realize(DynamicUser *d, uid_t *ret) { /* So, we found a working UID/lock combination. Let's see if we actually still need it. */ if (lockf(d->storage_socket[0], F_LOCK, 0) < 0) { - unlink_uid_lock(uid_lock_fd, uid); + unlink_uid_lock(uid_lock_fd, uid, d->name); return -errno; } @@ -407,7 +443,7 @@ int dynamic_user_realize(DynamicUser *d, uid_t *ret) { if (r < 0) { if (r != -EAGAIN) { /* OK, something bad happened, let's get rid of the bits we acquired. */ - unlink_uid_lock(uid_lock_fd, uid); + unlink_uid_lock(uid_lock_fd, uid, d->name); goto finish; } @@ -416,7 +452,7 @@ int dynamic_user_realize(DynamicUser *d, uid_t *ret) { /* Hmm, so as it appears there's now something stored in the storage socket. Throw away what we * acquired, and use what's stored now. */ - unlink_uid_lock(uid_lock_fd, uid); + unlink_uid_lock(uid_lock_fd, uid, d->name); safe_close(uid_lock_fd); uid = new_uid; @@ -513,7 +549,7 @@ static int dynamic_user_close(DynamicUser *d) { goto finish; /* This dynamic user was realized and dynamically allocated. In this case, let's remove the lock file. */ - unlink_uid_lock(lock_fd, uid); + unlink_uid_lock(lock_fd, uid, d->name); r = 1; finish: @@ -634,11 +670,8 @@ int dynamic_user_lookup_uid(Manager *m, uid_t uid, char **ret) { assert(m); assert(ret); - /* A friendly way to translate a dynamic user's UID into a his name. */ - - if (uid < UID_PICK_MIN) - return -ESRCH; - if (uid > UID_PICK_MAX) + /* A friendly way to translate a dynamic user's UID into a name. */ + if (!uid_is_dynamic(uid)) return -ESRCH; xsprintf(lock_path, "/run/systemd/dynamic-uid/" UID_FMT, uid); diff --git a/src/core/execute.c b/src/core/execute.c index 6019df7ea6..0af8eb5a02 100644 --- a/src/core/execute.c +++ b/src/core/execute.c @@ -91,6 +91,7 @@ #include "selinux-util.h" #include "signal-util.h" #include "smack-util.h" +#include "special.h" #include "string-table.h" #include "string-util.h" #include "strv.h" @@ -1384,6 +1385,7 @@ static void do_idle_pipe_dance(int idle_pipe[4]) { } static int build_environment( + Unit *u, const ExecContext *c, const ExecParameters *p, unsigned n_fds, @@ -1401,7 +1403,7 @@ static int build_environment( assert(c); assert(ret); - our_env = new0(char*, 12); + our_env = new0(char*, 13); if (!our_env) return -ENOMEM; @@ -1436,6 +1438,16 @@ static int build_environment( our_env[n_env++] = x; } + /* If this is D-Bus, tell the nss-systemd module, since it relies on being able to use D-Bus look up dynamic + * users via PID 1, possibly dead-locking the dbus daemon. This way it will not use D-Bus to resolve names, but + * check the database directly. */ + if (unit_has_name(u, SPECIAL_DBUS_SERVICE)) { + x = strdup("SYSTEMD_NSS_BYPASS_BUS=1"); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + if (home) { x = strappend("HOME=", home); if (!x) @@ -1723,11 +1735,12 @@ static int close_remaining_fds( const ExecParameters *params, ExecRuntime *runtime, DynamicCreds *dcreds, + int user_lookup_fd, int socket_fd, int *fds, unsigned n_fds) { unsigned n_dont_close = 0; - int dont_close[n_fds + 11]; + int dont_close[n_fds + 12]; assert(params); @@ -1755,9 +1768,40 @@ static int close_remaining_fds( append_socket_pair(dont_close, &n_dont_close, dcreds->group->storage_socket); } + if (user_lookup_fd >= 0) + dont_close[n_dont_close++] = user_lookup_fd; + return close_all_fds(dont_close, n_dont_close); } +static int send_user_lookup( + Unit *unit, + int user_lookup_fd, + uid_t uid, + gid_t gid) { + + assert(unit); + + /* Send the resolved UID/GID to PID 1 after we learnt it. We send a single datagram, containing the UID/GID + * data as well as the unit name. Note that we suppress sending this if no user/group to resolve was + * specified. */ + + if (user_lookup_fd < 0) + return 0; + + if (!uid_is_valid(uid) && !gid_is_valid(gid)) + return 0; + + if (writev(user_lookup_fd, + (struct iovec[]) { + { .iov_base = &uid, .iov_len = sizeof(uid) }, + { .iov_base = &gid, .iov_len = sizeof(gid) }, + { .iov_base = unit->id, .iov_len = strlen(unit->id) }}, 3) < 0) + return -errno; + + return 0; +} + static int exec_child( Unit *unit, ExecCommand *command, @@ -1769,6 +1813,7 @@ static int exec_child( int socket_fd, int *fds, unsigned n_fds, char **files_env, + int user_lookup_fd, int *exit_status) { _cleanup_strv_free_ char **our_env = NULL, **pass_env = NULL, **accum_env = NULL, **final_argv = NULL; @@ -1815,7 +1860,7 @@ static int exec_child( log_forget_fds(); - r = close_remaining_fds(params, runtime, dcreds, socket_fd, fds, n_fds); + r = close_remaining_fds(params, runtime, dcreds, user_lookup_fd, socket_fd, fds, n_fds); if (r < 0) { *exit_status = EXIT_FDS; return r; @@ -1862,7 +1907,7 @@ static int exec_child( return r; } - if (uid == UID_INVALID || gid == GID_INVALID) { + if (!uid_is_valid(uid) || !gid_is_valid(gid)) { *exit_status = EXIT_USER; return -ESRCH; } @@ -1902,6 +1947,14 @@ static int exec_child( } } + r = send_user_lookup(unit, user_lookup_fd, uid, gid); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + + user_lookup_fd = safe_close(user_lookup_fd); + /* If a socket is connected to STDIN/STDOUT/STDERR, we * must sure to drop O_NONBLOCK */ if (socket_fd >= 0) @@ -2059,6 +2112,7 @@ static int exec_child( } r = build_environment( + unit, context, params, n_fds, @@ -2501,6 +2555,7 @@ int exec_spawn(Unit *unit, socket_fd, fds, n_fds, files_env, + unit->manager->user_lookup_fds[1], &exit_status); if (r < 0) { log_open(); diff --git a/src/core/execute.h b/src/core/execute.h index 106154f81a..6082c42aba 100644 --- a/src/core/execute.h +++ b/src/core/execute.h @@ -178,6 +178,7 @@ struct ExecContext { bool no_new_privileges; bool dynamic_user; + bool remove_ipc; /* This is not exposed to the user but available * internally. We need it to make sure that whenever we spawn diff --git a/src/core/load-fragment-gperf.gperf.m4 b/src/core/load-fragment-gperf.gperf.m4 index 251155b428..05fe0df7e3 100644 --- a/src/core/load-fragment-gperf.gperf.m4 +++ b/src/core/load-fragment-gperf.gperf.m4 @@ -122,6 +122,8 @@ $1.KillSignal, config_parse_signal, 0, m4_define(`CGROUP_CONTEXT_CONFIG_ITEMS', `$1.Slice, config_parse_unit_slice, 0, 0 $1.CPUAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.cpu_accounting) +$1.CPUWeight, config_parse_cpu_weight, 0, offsetof($1, cgroup_context.cpu_weight) +$1.StartupCPUWeight, config_parse_cpu_weight, 0, offsetof($1, cgroup_context.startup_cpu_weight) $1.CPUShares, config_parse_cpu_shares, 0, offsetof($1, cgroup_context.cpu_shares) $1.StartupCPUShares, config_parse_cpu_shares, 0, offsetof($1, cgroup_context.startup_cpu_shares) $1.CPUQuota, config_parse_cpu_quota, 0, offsetof($1, cgroup_context) diff --git a/src/core/load-fragment.c b/src/core/load-fragment.c index 420f368689..d5185cf6a0 100644 --- a/src/core/load-fragment.c +++ b/src/core/load-fragment.c @@ -2851,6 +2851,34 @@ int config_parse_unit_slice( DEFINE_CONFIG_PARSE_ENUM(config_parse_device_policy, cgroup_device_policy, CGroupDevicePolicy, "Failed to parse device policy"); +int config_parse_cpu_weight( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *weight = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + r = cg_weight_parse(rvalue, weight); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "CPU weight '%s' invalid. Ignoring.", rvalue); + return 0; + } + + return 0; +} + int config_parse_cpu_shares( const char *unit, const char *filename, @@ -2971,30 +2999,36 @@ int config_parse_tasks_max( void *data, void *userdata) { - uint64_t *tasks_max = data, u; + uint64_t *tasks_max = data, v; + Unit *u = userdata; int r; - if (isempty(rvalue) || streq(rvalue, "infinity")) { - *tasks_max = (uint64_t) -1; + if (isempty(rvalue)) { + *tasks_max = u->manager->default_tasks_max; + return 0; + } + + if (streq(rvalue, "infinity")) { + *tasks_max = CGROUP_LIMIT_MAX; return 0; } r = parse_percent(rvalue); if (r < 0) { - r = safe_atou64(rvalue, &u); + r = safe_atou64(rvalue, &v); if (r < 0) { log_syntax(unit, LOG_ERR, filename, line, r, "Maximum tasks value '%s' invalid. Ignoring.", rvalue); return 0; } } else - u = system_tasks_max_scale(r, 100U); + v = system_tasks_max_scale(r, 100U); - if (u <= 0 || u >= UINT64_MAX) { + if (v <= 0 || v >= UINT64_MAX) { log_syntax(unit, LOG_ERR, filename, line, 0, "Maximum tasks value '%s' out of range. Ignoring.", rvalue); return 0; } - *tasks_max = u; + *tasks_max = v; return 0; } @@ -4191,6 +4225,7 @@ void unit_dump_config_items(FILE *f) { { config_parse_address_families, "FAMILIES" }, #endif { config_parse_cpu_shares, "SHARES" }, + { config_parse_cpu_weight, "WEIGHT" }, { config_parse_memory_limit, "LIMIT" }, { config_parse_device_allow, "DEVICE" }, { config_parse_device_policy, "POLICY" }, diff --git a/src/core/load-fragment.h b/src/core/load-fragment.h index 213bce55a7..8b688740cf 100644 --- a/src/core/load-fragment.h +++ b/src/core/load-fragment.h @@ -81,6 +81,7 @@ int config_parse_syscall_errno(const char *unit, const char *filename, unsigned int config_parse_environ(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_pass_environ(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_unit_slice(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_cpu_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_cpu_shares(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_memory_limit(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); int config_parse_tasks_max(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); diff --git a/src/core/main.c b/src/core/main.c index 02324d325e..c7c6df3447 100644 --- a/src/core/main.c +++ b/src/core/main.c @@ -1559,7 +1559,7 @@ int main(int argc, char *argv[]) { (void) reset_all_signal_handlers(); (void) ignore_signals(SIGNALS_IGNORE, -1); - arg_default_tasks_max = system_tasks_max_scale(15U, 100U); /* 15% the system PIDs equals 4915 by default. */ + arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U); if (parse_config_file() < 0) { error_message = "Failed to parse config file"; diff --git a/src/core/manager.c b/src/core/manager.c index c20e185d78..6f2477eef4 100644 --- a/src/core/manager.c +++ b/src/core/manager.c @@ -45,6 +45,7 @@ #include "bus-error.h" #include "bus-kernel.h" #include "bus-util.h" +#include "clean-ipc.h" #include "dbus-job.h" #include "dbus-manager.h" #include "dbus-unit.h" @@ -81,6 +82,7 @@ #include "transaction.h" #include "umask-util.h" #include "unit-name.h" +#include "user-util.h" #include "util.h" #include "virt.h" #include "watchdog.h" @@ -98,6 +100,7 @@ static int manager_dispatch_cgroups_agent_fd(sd_event_source *source, int fd, ui static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); static int manager_dispatch_time_change_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); static int manager_dispatch_idle_pipe_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_user_lookup_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); static int manager_dispatch_jobs_in_progress(sd_event_source *source, usec_t usec, void *userdata); static int manager_dispatch_run_queue(sd_event_source *source, void *userdata); static int manager_run_generators(Manager *m); @@ -590,6 +593,8 @@ int manager_new(UnitFileScope scope, bool test_run, Manager **_m) { m->dev_autofs_fd = m->private_listen_fd = m->kdbus_fd = m->cgroup_inotify_fd = m->ask_password_inotify_fd = -1; + m->user_lookup_fds[0] = m->user_lookup_fds[1] = -1; + m->current_job_id = 1; /* start as id #1, so that we can leave #0 around as "null-like" value */ m->have_ask_password = -EINVAL; /* we don't know */ @@ -812,6 +817,59 @@ static int manager_setup_cgroups_agent(Manager *m) { return 0; } +static int manager_setup_user_lookup_fd(Manager *m) { + int r; + + assert(m); + + /* Set up the socket pair used for passing UID/GID resolution results from forked off processes to PID + * 1. Background: we can't do name lookups (NSS) from PID 1, since it might involve IPC and thus activation, + * and we might hence deadlock on ourselves. Hence we do all user/group lookups asynchronously from the forked + * off processes right before executing the binaries to start. In order to be able to clean up any IPC objects + * created by a unit (see RemoveIPC=) we need to know in PID 1 the used UID/GID of the executed processes, + * hence we establish this communication channel so that forked off processes can pass their UID/GID + * information back to PID 1. The forked off processes send their resolved UID/GID to PID 1 in a simple + * datagram, along with their unit name, so that we can share one communication socket pair among all units for + * this purpose. + * + * You might wonder why we need a communication channel for this that is independent of the usual notification + * socket scheme (i.e. $NOTIFY_SOCKET). The primary difference is about trust: data sent via the $NOTIFY_SOCKET + * channel is only accepted if it originates from the right unit and if reception was enabled for it. The user + * lookup socket OTOH is only accessible by PID 1 and its children until they exec(), and always available. + * + * Note that this function is called under two circumstances: when we first initialize (in which case we + * allocate both the socket pair and the event source to listen on it), and when we deserialize after a reload + * (in which case the socket pair already exists but we still need to allocate the event source for it). */ + + if (m->user_lookup_fds[0] < 0) { + + /* Free all secondary fields */ + safe_close_pair(m->user_lookup_fds); + m->user_lookup_event_source = sd_event_source_unref(m->user_lookup_event_source); + + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, m->user_lookup_fds) < 0) + return log_error_errno(errno, "Failed to allocate user lookup socket: %m"); + + (void) fd_inc_rcvbuf(m->user_lookup_fds[0], NOTIFY_RCVBUF_SIZE); + } + + if (!m->user_lookup_event_source) { + r = sd_event_add_io(m->event, &m->user_lookup_event_source, m->user_lookup_fds[0], EPOLLIN, manager_dispatch_user_lookup_fd, m); + if (r < 0) + return log_error_errno(errno, "Failed to allocate user lookup event source: %m"); + + /* Process even earlier than the notify event source, so that we always know first about valid UID/GID + * resolutions */ + r = sd_event_source_set_priority(m->user_lookup_event_source, SD_EVENT_PRIORITY_NORMAL-8); + if (r < 0) + return log_error_errno(errno, "Failed to set priority ot user lookup event source: %m"); + + (void) sd_event_source_set_description(m->user_lookup_event_source, "user-lookup"); + } + + return 0; +} + static int manager_connect_bus(Manager *m, bool reexecuting) { bool try_bus_connect; @@ -853,8 +911,7 @@ enum { _GC_OFFSET_MAX }; -static void unit_gc_mark_good(Unit *u, unsigned gc_marker) -{ +static void unit_gc_mark_good(Unit *u, unsigned gc_marker) { Iterator i; Unit *other; @@ -1021,12 +1078,14 @@ Manager* manager_free(Manager *m) { sd_event_source_unref(m->time_change_event_source); sd_event_source_unref(m->jobs_in_progress_event_source); sd_event_source_unref(m->run_queue_event_source); + sd_event_source_unref(m->user_lookup_event_source); safe_close(m->signal_fd); safe_close(m->notify_fd); safe_close(m->cgroups_agent_fd); safe_close(m->time_change_fd); safe_close(m->kdbus_fd); + safe_close_pair(m->user_lookup_fds); manager_close_ask_password(m); @@ -1052,6 +1111,9 @@ Manager* manager_free(Manager *m) { assert(hashmap_isempty(m->units_requiring_mounts_for)); hashmap_free(m->units_requiring_mounts_for); + hashmap_free(m->uid_refs); + hashmap_free(m->gid_refs); + free(m); return NULL; } @@ -1221,6 +1283,10 @@ int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { if (q < 0 && r == 0) r = q; + q = manager_setup_user_lookup_fd(m); + if (q < 0 && r == 0) + r = q; + /* We might have deserialized the kdbus control fd, but if we * didn't, then let's create the bus now. */ manager_connect_bus(m, !!serialization); @@ -1232,6 +1298,10 @@ int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { /* Release any dynamic users no longer referenced */ dynamic_user_vacuum(m, true); + /* Release any references to UIDs/GIDs no longer referenced, and destroy any IPC owned by them */ + manager_vacuum_uid_refs(m); + manager_vacuum_gid_refs(m); + if (serialization) { assert(m->n_reloading > 0); m->n_reloading--; @@ -2396,6 +2466,20 @@ int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { fprintf(f, "cgroups-agent-fd=%i\n", copy); } + if (m->user_lookup_fds[0] >= 0) { + int copy0, copy1; + + copy0 = fdset_put_dup(fds, m->user_lookup_fds[0]); + if (copy0 < 0) + return copy0; + + copy1 = fdset_put_dup(fds, m->user_lookup_fds[1]); + if (copy1 < 0) + return copy1; + + fprintf(f, "user-lookup=%i %i\n", copy0, copy1); + } + if (m->kdbus_fd >= 0) { int copy; @@ -2412,6 +2496,9 @@ int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { if (r < 0) return r; + manager_serialize_uid_refs(m, f); + manager_serialize_gid_refs(m, f); + fputc('\n', f); HASHMAP_FOREACH_KEY(u, t, m->units, i) { @@ -2578,6 +2665,18 @@ int manager_deserialize(Manager *m, FILE *f, FDSet *fds) { m->cgroups_agent_fd = fdset_remove(fds, fd); } + } else if (startswith(l, "user-lookup=")) { + int fd0, fd1; + + if (sscanf(l + 12, "%i %i", &fd0, &fd1) != 2 || fd0 < 0 || fd1 < 0 || fd0 == fd1 || !fdset_contains(fds, fd0) || !fdset_contains(fds, fd1)) + log_debug("Failed to parse user lookup fd: %s", l + 12); + else { + m->user_lookup_event_source = sd_event_source_unref(m->user_lookup_event_source); + safe_close_pair(m->user_lookup_fds); + m->user_lookup_fds[0] = fdset_remove(fds, fd0); + m->user_lookup_fds[1] = fdset_remove(fds, fd1); + } + } else if (startswith(l, "kdbus-fd=")) { int fd; @@ -2590,6 +2689,10 @@ int manager_deserialize(Manager *m, FILE *f, FDSet *fds) { } else if (startswith(l, "dynamic-user=")) dynamic_user_deserialize_one(m, l + 13, fds); + else if (startswith(l, "destroy-ipc-uid=")) + manager_deserialize_uid_refs_one(m, l + 16); + else if (startswith(l, "destroy-ipc-gid=")) + manager_deserialize_gid_refs_one(m, l + 16); else { int k; @@ -2672,6 +2775,8 @@ int manager_reload(Manager *m) { lookup_paths_flush_generator(&m->lookup_paths); lookup_paths_free(&m->lookup_paths); dynamic_user_vacuum(m, false); + m->uid_refs = hashmap_free(m->uid_refs); + m->gid_refs = hashmap_free(m->gid_refs); q = lookup_paths_init(&m->lookup_paths, m->unit_file_scope, 0, NULL); if (q < 0 && r >= 0) @@ -2705,12 +2810,20 @@ int manager_reload(Manager *m) { if (q < 0 && r >= 0) r = q; + q = manager_setup_user_lookup_fd(m); + if (q < 0 && r >= 0) + r = q; + /* Third, fire things up! */ manager_coldplug(m); /* Release any dynamic users no longer referenced */ dynamic_user_vacuum(m, true); + /* Release any references to UIDs/GIDs no longer referenced, and destroy any IPC owned by them */ + manager_vacuum_uid_refs(m); + manager_vacuum_gid_refs(m); + /* Sync current state of bus names with our set of listening units */ if (m->api_bus) manager_sync_bus_names(m, m->api_bus); @@ -3144,6 +3257,300 @@ ManagerState manager_state(Manager *m) { return MANAGER_RUNNING; } +#define DESTROY_IPC_FLAG (UINT32_C(1) << 31) + +static void manager_unref_uid_internal( + Manager *m, + Hashmap **uid_refs, + uid_t uid, + bool destroy_now, + int (*_clean_ipc)(uid_t uid)) { + + uint32_t c, n; + + assert(m); + assert(uid_refs); + assert(uid_is_valid(uid)); + assert(_clean_ipc); + + /* A generic implementation, covering both manager_unref_uid() and manager_unref_gid(), under the assumption + * that uid_t and gid_t are actually defined the same way, with the same validity rules. + * + * We store a hashmap where the UID/GID is they key and the value is a 32bit reference counter, whose highest + * bit is used as flag for marking UIDs/GIDs whose IPC objects to remove when the last reference to the UID/GID + * is dropped. The flag is set to on, once at least one reference from a unit where RemoveIPC= is set is added + * on a UID/GID. It is reset when the UID's/GID's reference counter drops to 0 again. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (uid == 0) /* We don't keep track of root, and will never destroy it */ + return; + + c = PTR_TO_UINT32(hashmap_get(*uid_refs, UID_TO_PTR(uid))); + + n = c & ~DESTROY_IPC_FLAG; + assert(n > 0); + n--; + + if (destroy_now && n == 0) { + hashmap_remove(*uid_refs, UID_TO_PTR(uid)); + + if (c & DESTROY_IPC_FLAG) { + log_debug("%s " UID_FMT " is no longer referenced, cleaning up its IPC.", + _clean_ipc == clean_ipc_by_uid ? "UID" : "GID", + uid); + (void) _clean_ipc(uid); + } + } else { + c = n | (c & DESTROY_IPC_FLAG); + assert_se(hashmap_update(*uid_refs, UID_TO_PTR(uid), UINT32_TO_PTR(c)) >= 0); + } +} + +void manager_unref_uid(Manager *m, uid_t uid, bool destroy_now) { + manager_unref_uid_internal(m, &m->uid_refs, uid, destroy_now, clean_ipc_by_uid); +} + +void manager_unref_gid(Manager *m, gid_t gid, bool destroy_now) { + manager_unref_uid_internal(m, &m->gid_refs, (uid_t) gid, destroy_now, clean_ipc_by_gid); +} + +static int manager_ref_uid_internal( + Manager *m, + Hashmap **uid_refs, + uid_t uid, + bool clean_ipc) { + + uint32_t c, n; + int r; + + assert(m); + assert(uid_refs); + assert(uid_is_valid(uid)); + + /* A generic implementation, covering both manager_ref_uid() and manager_ref_gid(), under the assumption + * that uid_t and gid_t are actually defined the same way, with the same validity rules. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (uid == 0) /* We don't keep track of root, and will never destroy it */ + return 0; + + r = hashmap_ensure_allocated(uid_refs, &trivial_hash_ops); + if (r < 0) + return r; + + c = PTR_TO_UINT32(hashmap_get(*uid_refs, UID_TO_PTR(uid))); + + n = c & ~DESTROY_IPC_FLAG; + n++; + + if (n & DESTROY_IPC_FLAG) /* check for overflow */ + return -EOVERFLOW; + + c = n | (c & DESTROY_IPC_FLAG) | (clean_ipc ? DESTROY_IPC_FLAG : 0); + + return hashmap_replace(*uid_refs, UID_TO_PTR(uid), UINT32_TO_PTR(c)); +} + +int manager_ref_uid(Manager *m, uid_t uid, bool clean_ipc) { + return manager_ref_uid_internal(m, &m->uid_refs, uid, clean_ipc); +} + +int manager_ref_gid(Manager *m, gid_t gid, bool clean_ipc) { + return manager_ref_uid_internal(m, &m->gid_refs, (uid_t) gid, clean_ipc); +} + +static void manager_vacuum_uid_refs_internal( + Manager *m, + Hashmap **uid_refs, + int (*_clean_ipc)(uid_t uid)) { + + Iterator i; + void *p, *k; + + assert(m); + assert(uid_refs); + assert(_clean_ipc); + + HASHMAP_FOREACH_KEY(p, k, *uid_refs, i) { + uint32_t c, n; + uid_t uid; + + uid = PTR_TO_UID(k); + c = PTR_TO_UINT32(p); + + n = c & ~DESTROY_IPC_FLAG; + if (n > 0) + continue; + + if (c & DESTROY_IPC_FLAG) { + log_debug("Found unreferenced %s " UID_FMT " after reload/reexec. Cleaning up.", + _clean_ipc == clean_ipc_by_uid ? "UID" : "GID", + uid); + (void) _clean_ipc(uid); + } + + assert_se(hashmap_remove(*uid_refs, k) == p); + } +} + +void manager_vacuum_uid_refs(Manager *m) { + manager_vacuum_uid_refs_internal(m, &m->uid_refs, clean_ipc_by_uid); +} + +void manager_vacuum_gid_refs(Manager *m) { + manager_vacuum_uid_refs_internal(m, &m->gid_refs, clean_ipc_by_gid); +} + +static void manager_serialize_uid_refs_internal( + Manager *m, + FILE *f, + Hashmap **uid_refs, + const char *field_name) { + + Iterator i; + void *p, *k; + + assert(m); + assert(f); + assert(uid_refs); + assert(field_name); + + /* Serialize the UID reference table. Or actually, just the IPC destruction flag of it, as the actual counter + * of it is better rebuild after a reload/reexec. */ + + HASHMAP_FOREACH_KEY(p, k, *uid_refs, i) { + uint32_t c; + uid_t uid; + + uid = PTR_TO_UID(k); + c = PTR_TO_UINT32(p); + + if (!(c & DESTROY_IPC_FLAG)) + continue; + + fprintf(f, "%s=" UID_FMT "\n", field_name, uid); + } +} + +void manager_serialize_uid_refs(Manager *m, FILE *f) { + manager_serialize_uid_refs_internal(m, f, &m->uid_refs, "destroy-ipc-uid"); +} + +void manager_serialize_gid_refs(Manager *m, FILE *f) { + manager_serialize_uid_refs_internal(m, f, &m->gid_refs, "destroy-ipc-gid"); +} + +static void manager_deserialize_uid_refs_one_internal( + Manager *m, + Hashmap** uid_refs, + const char *value) { + + uid_t uid; + uint32_t c; + int r; + + assert(m); + assert(uid_refs); + assert(value); + + r = parse_uid(value, &uid); + if (r < 0 || uid == 0) { + log_debug("Unable to parse UID reference serialization"); + return; + } + + r = hashmap_ensure_allocated(uid_refs, &trivial_hash_ops); + if (r < 0) { + log_oom(); + return; + } + + c = PTR_TO_UINT32(hashmap_get(*uid_refs, UID_TO_PTR(uid))); + if (c & DESTROY_IPC_FLAG) + return; + + c |= DESTROY_IPC_FLAG; + + r = hashmap_replace(*uid_refs, UID_TO_PTR(uid), UINT32_TO_PTR(c)); + if (r < 0) { + log_debug("Failed to add UID reference entry"); + return; + } +} + +void manager_deserialize_uid_refs_one(Manager *m, const char *value) { + manager_deserialize_uid_refs_one_internal(m, &m->uid_refs, value); +} + +void manager_deserialize_gid_refs_one(Manager *m, const char *value) { + manager_deserialize_uid_refs_one_internal(m, &m->gid_refs, value); +} + +int manager_dispatch_user_lookup_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + struct buffer { + uid_t uid; + gid_t gid; + char unit_name[UNIT_NAME_MAX+1]; + } _packed_ buffer; + + Manager *m = userdata; + ssize_t l; + size_t n; + Unit *u; + + assert_se(source); + assert_se(m); + + /* Invoked whenever a child process succeeded resolving its user/group to use and sent us the resulting UID/GID + * in a datagram. We parse the datagram here and pass it off to the unit, so that it can add a reference to the + * UID/GID so that it can destroy the UID/GID's IPC objects when the reference counter drops to 0. */ + + l = recv(fd, &buffer, sizeof(buffer), MSG_DONTWAIT); + if (l < 0) { + if (errno == EINTR || errno == EAGAIN) + return 0; + + return log_error_errno(errno, "Failed to read from user lookup fd: %m"); + } + + if ((size_t) l <= offsetof(struct buffer, unit_name)) { + log_warning("Received too short user lookup message, ignoring."); + return 0; + } + + if ((size_t) l > offsetof(struct buffer, unit_name) + UNIT_NAME_MAX) { + log_warning("Received too long user lookup message, ignoring."); + return 0; + } + + if (!uid_is_valid(buffer.uid) && !gid_is_valid(buffer.gid)) { + log_warning("Got user lookup message with invalid UID/GID pair, ignoring."); + return 0; + } + + n = (size_t) l - offsetof(struct buffer, unit_name); + if (memchr(buffer.unit_name, 0, n)) { + log_warning("Received lookup message with embedded NUL character, ignoring."); + return 0; + } + + buffer.unit_name[n] = 0; + u = manager_get_unit(m, buffer.unit_name); + if (!u) { + log_debug("Got user lookup message but unit doesn't exist, ignoring."); + return 0; + } + + log_unit_debug(u, "User lookup succeeded: uid=" UID_FMT " gid=" GID_FMT, buffer.uid, buffer.gid); + + unit_notify_user_lookup(u, buffer.uid, buffer.gid); + return 0; +} + static const char *const manager_state_table[_MANAGER_STATE_MAX] = { [MANAGER_INITIALIZING] = "initializing", [MANAGER_STARTING] = "starting", diff --git a/src/core/manager.h b/src/core/manager.h index c681d5dc46..b9f2e4b5a1 100644 --- a/src/core/manager.h +++ b/src/core/manager.h @@ -143,6 +143,9 @@ struct Manager { sd_event_source *jobs_in_progress_event_source; + int user_lookup_fds[2]; + sd_event_source *user_lookup_event_source; + UnitFileScope unit_file_scope; LookupPaths lookup_paths; Set *unit_path_cache; @@ -234,7 +237,6 @@ struct Manager { bool dispatching_dbus_queue:1; bool taint_usr:1; - bool test_run:1; /* If non-zero, exit with the following value when the systemd @@ -301,6 +303,10 @@ struct Manager { /* Dynamic users/groups, indexed by their name */ Hashmap *dynamic_users; + /* Keep track of all UIDs and GIDs any of our services currently use. This is useful for the RemoveIPC= logic. */ + Hashmap *uid_refs; + Hashmap *gid_refs; + /* When the user hits C-A-D more than 7 times per 2s, reboot immediately... */ RateLimit ctrl_alt_del_ratelimit; @@ -378,5 +384,20 @@ ManagerState manager_state(Manager *m); int manager_update_failed_units(Manager *m, Unit *u, bool failed); +void manager_unref_uid(Manager *m, uid_t uid, bool destroy_now); +int manager_ref_uid(Manager *m, uid_t uid, bool clean_ipc); + +void manager_unref_gid(Manager *m, gid_t gid, bool destroy_now); +int manager_ref_gid(Manager *m, gid_t gid, bool destroy_now); + +void manager_vacuum_uid_refs(Manager *m); +void manager_vacuum_gid_refs(Manager *m); + +void manager_serialize_uid_refs(Manager *m, FILE *f); +void manager_deserialize_uid_refs_one(Manager *m, const char *value); + +void manager_serialize_gid_refs(Manager *m, FILE *f); +void manager_deserialize_gid_refs_one(Manager *m, const char *value); + const char *manager_state_to_string(ManagerState m) _const_; ManagerState manager_state_from_string(const char *s) _pure_; diff --git a/src/core/mount.c b/src/core/mount.c index f3ccf6d48a..f2ac8d171f 100644 --- a/src/core/mount.c +++ b/src/core/mount.c @@ -769,6 +769,8 @@ static void mount_enter_dead(Mount *m, MountResult f) { exec_context_destroy_runtime_directory(&m->exec_context, manager_get_runtime_prefix(UNIT(m)->manager)); + unit_unref_uid_gid(UNIT(m), true); + dynamic_creds_destroy(&m->dynamic_creds); } diff --git a/src/core/service.c b/src/core/service.c index 4a37702f52..1951ba9222 100644 --- a/src/core/service.c +++ b/src/core/service.c @@ -1471,6 +1471,9 @@ static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) /* Also, remove the runtime directory */ exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + /* Get rid of the IPC bits of the user */ + unit_unref_uid_gid(UNIT(s), true); + /* Release the user, and destroy it if we are the only remaining owner */ dynamic_creds_destroy(&s->dynamic_creds); diff --git a/src/core/socket.c b/src/core/socket.c index 50872e8366..70d55dd9ed 100644 --- a/src/core/socket.c +++ b/src/core/socket.c @@ -1905,6 +1905,8 @@ static void socket_enter_dead(Socket *s, SocketResult f) { exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + unit_unref_uid_gid(UNIT(s), true); + dynamic_creds_destroy(&s->dynamic_creds); } diff --git a/src/core/swap.c b/src/core/swap.c index 2c802da3b5..fb222b6858 100644 --- a/src/core/swap.c +++ b/src/core/swap.c @@ -683,6 +683,8 @@ static void swap_enter_dead(Swap *s, SwapResult f) { exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + unit_unref_uid_gid(UNIT(s), true); + dynamic_creds_destroy(&s->dynamic_creds); } diff --git a/src/core/unit.c b/src/core/unit.c index ff7c562fba..4b8d81c3f1 100644 --- a/src/core/unit.c +++ b/src/core/unit.c @@ -100,7 +100,8 @@ Unit *unit_new(Manager *m, size_t size) { u->on_failure_job_mode = JOB_REPLACE; u->cgroup_inotify_wd = -1; u->job_timeout = USEC_INFINITY; - u->sigchldgen = 0; + u->ref_uid = UID_INVALID; + u->ref_gid = GID_INVALID; RATELIMIT_INIT(u->start_limit, m->default_start_limit_interval, m->default_start_limit_burst); RATELIMIT_INIT(u->auto_stop_ratelimit, 10 * USEC_PER_SEC, 16); @@ -550,6 +551,8 @@ void unit_free(Unit *u) { unit_release_cgroup(u); + unit_unref_uid_gid(u, false); + (void) manager_update_failed_units(u->manager, u, false); set_remove(u->manager->startup_units, u); @@ -2608,12 +2611,17 @@ int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { unit_serialize_item(u, f, "assert-result", yes_no(u->assert_result)); unit_serialize_item(u, f, "transient", yes_no(u->transient)); - unit_serialize_item_format(u, f, "cpuacct-usage-base", "%" PRIu64, u->cpuacct_usage_base); + unit_serialize_item_format(u, f, "cpu-usage-base", "%" PRIu64, u->cpu_usage_base); if (u->cgroup_path) unit_serialize_item(u, f, "cgroup", u->cgroup_path); unit_serialize_item(u, f, "cgroup-realized", yes_no(u->cgroup_realized)); + if (uid_is_valid(u->ref_uid)) + unit_serialize_item_format(u, f, "ref-uid", UID_FMT, u->ref_uid); + if (gid_is_valid(u->ref_gid)) + unit_serialize_item_format(u, f, "ref-gid", GID_FMT, u->ref_gid); + if (serialize_jobs) { if (u->job) { fprintf(f, "job\n"); @@ -2824,9 +2832,9 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { continue; - } else if (streq(l, "cpuacct-usage-base")) { + } else if (streq(l, "cpu-usage-base") || streq(l, "cpuacct-usage-base")) { - r = safe_atou64(v, &u->cpuacct_usage_base); + r = safe_atou64(v, &u->cpu_usage_base); if (r < 0) log_unit_debug(u, "Failed to parse CPU usage %s, ignoring.", v); @@ -2851,6 +2859,28 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { u->cgroup_realized = b; continue; + + } else if (streq(l, "ref-uid")) { + uid_t uid; + + r = parse_uid(v, &uid); + if (r < 0) + log_unit_debug(u, "Failed to parse referenced UID %s, ignoring.", v); + else + unit_ref_uid_gid(u, uid, GID_INVALID); + + continue; + + } else if (streq(l, "ref-gid")) { + gid_t gid; + + r = parse_gid(v, &gid); + if (r < 0) + log_unit_debug(u, "Failed to parse referenced GID %s, ignoring.", v); + else + unit_ref_uid_gid(u, UID_INVALID, gid); + + continue; } if (unit_can_serialize(u)) { @@ -3310,6 +3340,7 @@ int unit_patch_contexts(Unit *u) { } ec->private_tmp = true; + ec->remove_ipc = true; } } @@ -3932,3 +3963,144 @@ pid_t unit_main_pid(Unit *u) { return 0; } + +static void unit_unref_uid_internal( + Unit *u, + uid_t *ref_uid, + bool destroy_now, + void (*_manager_unref_uid)(Manager *m, uid_t uid, bool destroy_now)) { + + assert(u); + assert(ref_uid); + assert(_manager_unref_uid); + + /* Generic implementation of both unit_unref_uid() and unit_unref_gid(), under the assumption that uid_t and + * gid_t are actually the same time, with the same validity rules. + * + * Drops a reference to UID/GID from a unit. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (!uid_is_valid(*ref_uid)) + return; + + _manager_unref_uid(u->manager, *ref_uid, destroy_now); + *ref_uid = UID_INVALID; +} + +void unit_unref_uid(Unit *u, bool destroy_now) { + unit_unref_uid_internal(u, &u->ref_uid, destroy_now, manager_unref_uid); +} + +void unit_unref_gid(Unit *u, bool destroy_now) { + unit_unref_uid_internal(u, (uid_t*) &u->ref_gid, destroy_now, manager_unref_gid); +} + +static int unit_ref_uid_internal( + Unit *u, + uid_t *ref_uid, + uid_t uid, + bool clean_ipc, + int (*_manager_ref_uid)(Manager *m, uid_t uid, bool clean_ipc)) { + + int r; + + assert(u); + assert(ref_uid); + assert(uid_is_valid(uid)); + assert(_manager_ref_uid); + + /* Generic implementation of both unit_ref_uid() and unit_ref_guid(), under the assumption that uid_t and gid_t + * are actually the same type, and have the same validity rules. + * + * Adds a reference on a specific UID/GID to this unit. Each unit referencing the same UID/GID maintains a + * reference so that we can destroy the UID/GID's IPC resources as soon as this is requested and the counter + * drops to zero. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (*ref_uid == uid) + return 0; + + if (uid_is_valid(*ref_uid)) /* Already set? */ + return -EBUSY; + + r = _manager_ref_uid(u->manager, uid, clean_ipc); + if (r < 0) + return r; + + *ref_uid = uid; + return 1; +} + +int unit_ref_uid(Unit *u, uid_t uid, bool clean_ipc) { + return unit_ref_uid_internal(u, &u->ref_uid, uid, clean_ipc, manager_ref_uid); +} + +int unit_ref_gid(Unit *u, gid_t gid, bool clean_ipc) { + return unit_ref_uid_internal(u, (uid_t*) &u->ref_gid, (uid_t) gid, clean_ipc, manager_ref_gid); +} + +static int unit_ref_uid_gid_internal(Unit *u, uid_t uid, gid_t gid, bool clean_ipc) { + int r = 0, q = 0; + + assert(u); + + /* Reference both a UID and a GID in one go. Either references both, or neither. */ + + if (uid_is_valid(uid)) { + r = unit_ref_uid(u, uid, clean_ipc); + if (r < 0) + return r; + } + + if (gid_is_valid(gid)) { + q = unit_ref_gid(u, gid, clean_ipc); + if (q < 0) { + if (r > 0) + unit_unref_uid(u, false); + + return q; + } + } + + return r > 0 || q > 0; +} + +int unit_ref_uid_gid(Unit *u, uid_t uid, gid_t gid) { + ExecContext *c; + int r; + + assert(u); + + c = unit_get_exec_context(u); + + r = unit_ref_uid_gid_internal(u, uid, gid, c ? c->remove_ipc : false); + if (r < 0) + return log_unit_warning_errno(u, r, "Couldn't add UID/GID reference to unit, proceeding without: %m"); + + return r; +} + +void unit_unref_uid_gid(Unit *u, bool destroy_now) { + assert(u); + + unit_unref_uid(u, destroy_now); + unit_unref_gid(u, destroy_now); +} + +void unit_notify_user_lookup(Unit *u, uid_t uid, gid_t gid) { + int r; + + assert(u); + + /* This is invoked whenever one of the forked off processes let's us know the UID/GID its user name/group names + * resolved to. We keep track of which UID/GID is currently assigned in order to be able to destroy its IPC + * objects when no service references the UID/GID anymore. */ + + r = unit_ref_uid_gid(u, uid, gid); + if (r > 0) + bus_unit_send_change_signal(u); +} diff --git a/src/core/unit.h b/src/core/unit.h index 47eb8d50a6..53875653d7 100644 --- a/src/core/unit.h +++ b/src/core/unit.h @@ -180,12 +180,16 @@ struct Unit { /* Make sure we never enter endless loops with the check unneeded logic, or the BindsTo= logic */ RateLimit auto_stop_ratelimit; + /* Reference to a specific UID/GID */ + uid_t ref_uid; + gid_t ref_gid; + /* Cached unit file state and preset */ UnitFileState unit_file_state; int unit_file_preset; - /* Where the cpuacct.usage cgroup counter was at the time the unit was started */ - nsec_t cpuacct_usage_base; + /* Where the cpu.stat or cpuacct.usage was at the time the unit was started */ + nsec_t cpu_usage_base; /* Counterparts in the cgroup filesystem */ char *cgroup_path; @@ -195,8 +199,6 @@ struct Unit { CGroupMask cgroup_members_mask; int cgroup_inotify_wd; - uint32_t cgroup_netclass_id; - /* How to start OnFailure units */ JobMode on_failure_job_mode; @@ -373,8 +375,7 @@ struct UnitVTable { /* Called whenever a process of this unit sends us a message */ void (*notify_message)(Unit *u, pid_t pid, char **tags, FDSet *fds); - /* Called whenever a name this Unit registered for comes or - * goes away. */ + /* Called whenever a name this Unit registered for comes or goes away. */ void (*bus_name_owner_change)(Unit *u, const char *name, const char *old_owner, const char *new_owner); /* Called for each property that is being set */ @@ -623,6 +624,17 @@ int unit_fail_if_symlink(Unit *u, const char* where); int unit_start_limit_test(Unit *u); +void unit_unref_uid(Unit *u, bool destroy_now); +int unit_ref_uid(Unit *u, uid_t uid, bool clean_ipc); + +void unit_unref_gid(Unit *u, bool destroy_now); +int unit_ref_gid(Unit *u, gid_t gid, bool clean_ipc); + +int unit_ref_uid_gid(Unit *u, uid_t uid, gid_t gid); +void unit_unref_uid_gid(Unit *u, bool destroy_now); + +void unit_notify_user_lookup(Unit *u, uid_t uid, gid_t gid); + /* Macros which append UNIT= or USER_UNIT= to the message */ #define log_unit_full(unit, level, error, ...) \ diff --git a/src/hostname/hostnamectl.c b/src/hostname/hostnamectl.c index c16a324232..4795324667 100644 --- a/src/hostname/hostnamectl.c +++ b/src/hostname/hostnamectl.c @@ -251,7 +251,7 @@ static int set_simple_string(sd_bus *bus, const char *method, const char *value) static int set_hostname(sd_bus *bus, char **args, unsigned n) { _cleanup_free_ char *h = NULL; - char *hostname = args[1]; + const char *hostname = args[1]; int r; assert(args); @@ -263,27 +263,29 @@ static int set_hostname(sd_bus *bus, char **args, unsigned n) { if (arg_pretty) { const char *p; - /* If the passed hostname is already valid, then - * assume the user doesn't know anything about pretty - * hostnames, so let's unset the pretty hostname, and - * just set the passed hostname as static/dynamic + /* If the passed hostname is already valid, then assume the user doesn't know anything about pretty + * hostnames, so let's unset the pretty hostname, and just set the passed hostname as static/dynamic * hostname. */ - - if (arg_static && hostname_is_valid(hostname, true)) { - p = ""; - /* maybe get rid of trailing dot */ - hostname = hostname_cleanup(hostname); - } else { - p = h = strdup(hostname); - if (!p) - return log_oom(); - - hostname_cleanup(hostname); - } + if (arg_static && hostname_is_valid(hostname, true)) + p = ""; /* No pretty hostname (as it is redundant), just a static one */ + else + p = hostname; /* Use the passed name as pretty hostname */ r = set_simple_string(bus, "SetPrettyHostname", p); if (r < 0) return r; + + /* Now that we set the pretty hostname, let's clean up the parameter and use that as static + * hostname. If the hostname was already valid as static hostname, this will only chop off the trailing + * dot if there is one. If it was not valid, then it will be made fully valid by truncating, dropping + * multiple dots, and and dropping weird chars. Note that we clean the name up only if we also are + * supposed to set the pretty name. If the pretty name is not being set we assume the user knows what + * he does and pass the name as-is. */ + h = strdup(hostname); + if (!h) + return log_oom(); + + hostname = hostname_cleanup(h); /* Use the cleaned up name as static hostname */ } if (arg_static) { diff --git a/src/journal/journalctl.c b/src/journal/journalctl.c index 6f841efb69..381e219390 100644 --- a/src/journal/journalctl.c +++ b/src/journal/journalctl.c @@ -310,7 +310,7 @@ static void help(void) { " -m --merge Show entries from all available journals\n" " -D --directory=PATH Show journal files from directory\n" " --file=PATH Show journal file\n" - " --root=ROOT Operate on catalog files below a root directory\n" + " --root=ROOT Operate on files below a root directory\n" #ifdef HAVE_GCRYPT " --interval=TIME Time interval for changing the FSS sealing key\n" " --verify-key=KEY Specify FSS verification key\n" @@ -848,8 +848,8 @@ static int parse_argv(int argc, char *argv[]) { if (arg_follow && !arg_no_tail && !arg_since && arg_lines == ARG_LINES_DEFAULT) arg_lines = 10; - if (!!arg_directory + !!arg_file + !!arg_machine > 1) { - log_error("Please specify either -D/--directory= or --file= or -M/--machine=, not more than one."); + if (!!arg_directory + !!arg_file + !!arg_machine + !!arg_root > 1) { + log_error("Please specify at most one of -D/--directory=, --file=, -M/--machine=, --root."); return -EINVAL; } @@ -1267,7 +1267,7 @@ static int add_boot(sd_journal *j) { * We can do this only when we logs are coming from the current machine, * so take the slow path if log location is specified. */ if (arg_boot_offset == 0 && sd_id128_is_null(arg_boot_id) && - !arg_directory && !arg_file) + !arg_directory && !arg_file && !arg_root) return add_match_this_boot(j, arg_machine); @@ -2161,6 +2161,8 @@ int main(int argc, char *argv[]) { if (arg_directory) r = sd_journal_open_directory(&j, arg_directory, arg_journal_type); + else if (arg_root) + r = sd_journal_open_directory(&j, arg_root, arg_journal_type | SD_JOURNAL_OS_ROOT); else if (arg_file_stdin) { int ifd = STDIN_FILENO; r = sd_journal_open_files_fd(&j, &ifd, 1, 0); diff --git a/src/journal/journald-server.c b/src/journal/journald-server.c index 587c343b31..2a043a95b1 100644 --- a/src/journal/journald-server.c +++ b/src/journal/journald-server.c @@ -267,6 +267,82 @@ static int open_journal( return r; } +static int system_journal_open(Server *s, bool flush_requested) { + const char *fn; + int r = 0; + + if (!s->system_journal && + (s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) && + (flush_requested + || access("/run/systemd/journal/flushed", F_OK) >= 0)) { + + /* If in auto mode: first try to create the machine + * path, but not the prefix. + * + * If in persistent mode: create /var/log/journal and + * the machine path */ + + if (s->storage == STORAGE_PERSISTENT) + (void) mkdir_p("/var/log/journal/", 0755); + + fn = strjoina("/var/log/journal/", SERVER_MACHINE_ID(s)); + (void) mkdir(fn, 0755); + + fn = strjoina(fn, "/system.journal"); + r = open_journal(s, true, fn, O_RDWR|O_CREAT, s->seal, &s->system_metrics, &s->system_journal); + if (r >= 0) { + server_add_acls(s->system_journal, 0); + (void) determine_space_for(s, &s->system_metrics, "/var/log/journal/", "System journal", true, true, NULL, NULL); + } else if (r < 0) { + if (r != -ENOENT && r != -EROFS) + log_warning_errno(r, "Failed to open system journal: %m"); + + r = 0; + } + } + + if (!s->runtime_journal && + (s->storage != STORAGE_NONE)) { + + fn = strjoina("/run/log/journal/", SERVER_MACHINE_ID(s), "/system.journal"); + + if (s->system_journal) { + + /* Try to open the runtime journal, but only + * if it already exists, so that we can flush + * it into the system journal */ + + r = open_journal(s, false, fn, O_RDWR, false, &s->runtime_metrics, &s->runtime_journal); + if (r < 0) { + if (r != -ENOENT) + log_warning_errno(r, "Failed to open runtime journal: %m"); + + r = 0; + } + + } else { + + /* OK, we really need the runtime journal, so create + * it if necessary. */ + + (void) mkdir("/run/log", 0755); + (void) mkdir("/run/log/journal", 0755); + (void) mkdir_parents(fn, 0750); + + r = open_journal(s, true, fn, O_RDWR|O_CREAT, false, &s->runtime_metrics, &s->runtime_journal); + if (r < 0) + return log_error_errno(r, "Failed to open runtime journal: %m"); + } + + if (s->runtime_journal) { + server_add_acls(s->runtime_journal, 0); + (void) determine_space_for(s, &s->runtime_metrics, "/run/log/journal/", "Runtime journal", true, true, NULL, NULL); + } + } + + return r; +} + static JournalFile* find_journal(Server *s, uid_t uid) { _cleanup_free_ char *p = NULL; int r; @@ -275,6 +351,17 @@ static JournalFile* find_journal(Server *s, uid_t uid) { assert(s); + /* A rotate that fails to create the new journal (ENOSPC) leaves the + * rotated journal as NULL. Unless we revisit opening, even after + * space is made available we'll continue to return NULL indefinitely. + * + * system_journal_open() is a noop if the journals are already open, so + * we can just call it here to recover from failed rotates (or anything + * else that's left the journals as NULL). + * + * Fixes https://github.com/systemd/systemd/issues/3968 */ + (void) system_journal_open(s, false); + /* We split up user logs only on /var, not on /run. If the * runtime file is open, we write to it exclusively, in order * to guarantee proper order as soon as we flush /run to @@ -283,7 +370,7 @@ static JournalFile* find_journal(Server *s, uid_t uid) { if (s->runtime_journal) return s->runtime_journal; - if (uid <= SYSTEM_UID_MAX) + if (uid <= SYSTEM_UID_MAX || uid_is_dynamic(uid)) return s->system_journal; r = sd_id128_get_machine(&machine); @@ -979,83 +1066,6 @@ finish: dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id, priority, object_pid); } - -static int system_journal_open(Server *s, bool flush_requested) { - const char *fn; - int r = 0; - - if (!s->system_journal && - (s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) && - (flush_requested - || access("/run/systemd/journal/flushed", F_OK) >= 0)) { - - /* If in auto mode: first try to create the machine - * path, but not the prefix. - * - * If in persistent mode: create /var/log/journal and - * the machine path */ - - if (s->storage == STORAGE_PERSISTENT) - (void) mkdir_p("/var/log/journal/", 0755); - - fn = strjoina("/var/log/journal/", SERVER_MACHINE_ID(s)); - (void) mkdir(fn, 0755); - - fn = strjoina(fn, "/system.journal"); - r = open_journal(s, true, fn, O_RDWR|O_CREAT, s->seal, &s->system_metrics, &s->system_journal); - if (r >= 0) { - server_add_acls(s->system_journal, 0); - (void) determine_space_for(s, &s->system_metrics, "/var/log/journal/", "System journal", true, true, NULL, NULL); - } else if (r < 0) { - if (r != -ENOENT && r != -EROFS) - log_warning_errno(r, "Failed to open system journal: %m"); - - r = 0; - } - } - - if (!s->runtime_journal && - (s->storage != STORAGE_NONE)) { - - fn = strjoina("/run/log/journal/", SERVER_MACHINE_ID(s), "/system.journal"); - - if (s->system_journal) { - - /* Try to open the runtime journal, but only - * if it already exists, so that we can flush - * it into the system journal */ - - r = open_journal(s, false, fn, O_RDWR, false, &s->runtime_metrics, &s->runtime_journal); - if (r < 0) { - if (r != -ENOENT) - log_warning_errno(r, "Failed to open runtime journal: %m"); - - r = 0; - } - - } else { - - /* OK, we really need the runtime journal, so create - * it if necessary. */ - - (void) mkdir("/run/log", 0755); - (void) mkdir("/run/log/journal", 0755); - (void) mkdir_parents(fn, 0750); - - r = open_journal(s, true, fn, O_RDWR|O_CREAT, false, &s->runtime_metrics, &s->runtime_journal); - if (r < 0) - return log_error_errno(r, "Failed to open runtime journal: %m"); - } - - if (s->runtime_journal) { - server_add_acls(s->runtime_journal, 0); - (void) determine_space_for(s, &s->runtime_metrics, "/run/log/journal/", "Runtime journal", true, true, NULL, NULL); - } - } - - return r; -} - int server_flush_to_var(Server *s) { sd_id128_t machine; sd_journal *j = NULL; diff --git a/src/journal/sd-journal.c b/src/journal/sd-journal.c index 2a3824d0e8..98c8a47afe 100644 --- a/src/journal/sd-journal.c +++ b/src/journal/sd-journal.c @@ -1719,9 +1719,16 @@ static sd_journal *journal_new(int flags, const char *path) { j->data_threshold = DEFAULT_DATA_THRESHOLD; if (path) { - j->path = strdup(path); - if (!j->path) + char *t; + + t = strdup(path); + if (!t) goto fail; + + if (flags & SD_JOURNAL_OS_ROOT) + j->prefix = t; + else + j->path = t; } j->files = ordered_hashmap_new(&string_hash_ops); @@ -1737,12 +1744,17 @@ fail: return NULL; } +#define OPEN_ALLOWED_FLAGS \ + (SD_JOURNAL_LOCAL_ONLY | \ + SD_JOURNAL_RUNTIME_ONLY | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER) + _public_ int sd_journal_open(sd_journal **ret, int flags) { sd_journal *j; int r; assert_return(ret, -EINVAL); - assert_return((flags & ~(SD_JOURNAL_LOCAL_ONLY|SD_JOURNAL_RUNTIME_ONLY|SD_JOURNAL_SYSTEM|SD_JOURNAL_CURRENT_USER)) == 0, -EINVAL); + assert_return((flags & ~OPEN_ALLOWED_FLAGS) == 0, -EINVAL); j = journal_new(flags, NULL); if (!j) @@ -1761,6 +1773,9 @@ fail: return r; } +#define OPEN_CONTAINER_ALLOWED_FLAGS \ + (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM) + _public_ int sd_journal_open_container(sd_journal **ret, const char *machine, int flags) { _cleanup_free_ char *root = NULL, *class = NULL; sd_journal *j; @@ -1772,7 +1787,7 @@ _public_ int sd_journal_open_container(sd_journal **ret, const char *machine, in assert_return(machine, -EINVAL); assert_return(ret, -EINVAL); - assert_return((flags & ~(SD_JOURNAL_LOCAL_ONLY|SD_JOURNAL_SYSTEM)) == 0, -EINVAL); + assert_return((flags & ~OPEN_CONTAINER_ALLOWED_FLAGS) == 0, -EINVAL); assert_return(machine_name_is_valid(machine), -EINVAL); p = strjoina("/run/systemd/machines/", machine); @@ -1787,13 +1802,10 @@ _public_ int sd_journal_open_container(sd_journal **ret, const char *machine, in if (!streq_ptr(class, "container")) return -EIO; - j = journal_new(flags, NULL); + j = journal_new(flags, root); if (!j) return -ENOMEM; - j->prefix = root; - root = NULL; - r = add_search_paths(j); if (r < 0) goto fail; @@ -1806,13 +1818,17 @@ fail: return r; } +#define OPEN_DIRECTORY_ALLOWED_FLAGS \ + (SD_JOURNAL_OS_ROOT | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER ) + _public_ int sd_journal_open_directory(sd_journal **ret, const char *path, int flags) { sd_journal *j; int r; assert_return(ret, -EINVAL); assert_return(path, -EINVAL); - assert_return((flags & ~SD_JOURNAL_OS_ROOT) == 0, -EINVAL); + assert_return((flags & ~OPEN_DIRECTORY_ALLOWED_FLAGS) == 0, -EINVAL); j = journal_new(flags, path); if (!j) @@ -1861,6 +1877,10 @@ fail: return r; } +#define OPEN_DIRECTORY_FD_ALLOWED_FLAGS \ + (SD_JOURNAL_OS_ROOT | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER ) + _public_ int sd_journal_open_directory_fd(sd_journal **ret, int fd, int flags) { sd_journal *j; struct stat st; @@ -1868,7 +1888,7 @@ _public_ int sd_journal_open_directory_fd(sd_journal **ret, int fd, int flags) { assert_return(ret, -EINVAL); assert_return(fd >= 0, -EBADF); - assert_return((flags & ~SD_JOURNAL_OS_ROOT) == 0, -EINVAL); + assert_return((flags & ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS) == 0, -EINVAL); if (fstat(fd, &st) < 0) return -errno; diff --git a/src/libudev/libudev-device.c b/src/libudev/libudev-device.c index 995bf56586..c5f36725dc 100644 --- a/src/libudev/libudev-device.c +++ b/src/libudev/libudev-device.c @@ -495,7 +495,7 @@ _public_ struct udev_device *udev_device_get_parent_with_subsystem_devtype(struc return NULL; } - /* then walk the chain of udev_device parents until the correspanding + /* then walk the chain of udev_device parents until the corresponding one is found */ while ((udev_device = udev_device_get_parent(udev_device))) { if (udev_device->device == parent) diff --git a/src/login/logind-action.c b/src/login/logind-action.c index 8ef48dbaa1..a950409254 100644 --- a/src/login/logind-action.c +++ b/src/login/logind-action.c @@ -85,7 +85,7 @@ int manager_handle_action( } /* If the key handling is inhibited, don't do anything */ - if (!ignore_inhibited && inhibit_key > 0) { + if (inhibit_key > 0) { if (manager_is_inhibited(m, inhibit_key, INHIBIT_BLOCK, NULL, true, false, 0, NULL)) { log_debug("Refusing operation, %s is inhibited.", inhibit_what_to_string(inhibit_key)); return 0; diff --git a/src/login/logind-user.c b/src/login/logind-user.c index 63363035e7..e0e73b034d 100644 --- a/src/login/logind-user.c +++ b/src/login/logind-user.c @@ -26,6 +26,7 @@ #include "bus-common-errors.h" #include "bus-error.h" #include "bus-util.h" +#include "cgroup-util.h" #include "clean-ipc.h" #include "conf-parser.h" #include "escape.h" @@ -612,9 +613,14 @@ int user_finalize(User *u) { if (k < 0) r = k; - /* Clean SysV + POSIX IPC objects */ - if (u->manager->remove_ipc) { - k = clean_ipc(u->uid); + /* Clean SysV + POSIX IPC objects, but only if this is not a system user. Background: in many setups cronjobs + * are run in full PAM and thus logind sessions, even if the code run doesn't belong to actual users but to + * system components. Since enable RemoveIPC= globally for all users, we need to be a bit careful with such + * cases, as we shouldn't accidentally remove a system service's IPC objects while it is running, just because + * a cronjob running as the same user just finished. Hence: exclude system users generally from IPC clean-up, + * and do it only for normal users. */ + if (u->manager->remove_ipc && u->uid > SYSTEM_UID_MAX) { + k = clean_ipc_by_uid(u->uid); if (k < 0) r = k; } @@ -891,7 +897,17 @@ int config_parse_user_tasks_max( assert(rvalue); assert(data); - /* First, try to parse as percentage */ + if (isempty(rvalue)) { + *m = system_tasks_max_scale(DEFAULT_USER_TASKS_MAX_PERCENTAGE, 100U); + return 0; + } + + if (streq(rvalue, "infinity")) { + *m = CGROUP_LIMIT_MAX; + return 0; + } + + /* Try to parse as percentage */ r = parse_percent(rvalue); if (r >= 0) k = system_tasks_max_scale(r, 100U); diff --git a/src/login/logind.c b/src/login/logind.c index 5ce36d28c7..bbbf4aef57 100644 --- a/src/login/logind.c +++ b/src/login/logind.c @@ -38,6 +38,7 @@ #include "signal-util.h" #include "strv.h" #include "udev-util.h" +#include "cgroup-util.h" static void manager_free(Manager *m); @@ -62,7 +63,7 @@ static void manager_reset_config(Manager *m) { m->idle_action = HANDLE_IGNORE; m->runtime_dir_size = physical_memory_scale(10U, 100U); /* 10% */ - m->user_tasks_max = system_tasks_max_scale(33U, 100U); /* 33% */ + m->user_tasks_max = system_tasks_max_scale(DEFAULT_USER_TASKS_MAX_PERCENTAGE, 100U); /* 33% */ m->sessions_max = 8192; m->inhibitors_max = 8192; diff --git a/src/network/networkd-route.c b/src/network/networkd-route.c index cedaf47cf8..b197a5da6b 100644 --- a/src/network/networkd-route.c +++ b/src/network/networkd-route.c @@ -41,7 +41,7 @@ int route_new(Route **ret) { route->family = AF_UNSPEC; route->scope = RT_SCOPE_UNIVERSE; route->protocol = RTPROT_UNSPEC; - route->table = RT_TABLE_DEFAULT; + route->table = RT_TABLE_MAIN; route->lifetime = USEC_INFINITY; *ret = route; @@ -322,7 +322,8 @@ int route_add( } else return r; - *ret = route; + if (ret) + *ret = route; return 0; } @@ -440,20 +441,14 @@ static int route_expire_callback(sd_netlink *rtnl, sd_netlink_message *m, void * assert(m); assert(link); assert(link->ifname); - assert(link->link_messages > 0); if (IN_SET(link->state, LINK_STATE_FAILED, LINK_STATE_LINGER)) return 1; - link->link_messages--; - r = sd_netlink_message_get_errno(m); if (r < 0 && r != -EEXIST) log_link_warning_errno(link, r, "could not remove route: %m"); - if (link->link_messages == 0) - log_link_debug(link, "route removed"); - return 1; } @@ -466,11 +461,8 @@ int route_expire_handler(sd_event_source *s, uint64_t usec, void *userdata) { r = route_remove(route, route->link, route_expire_callback); if (r < 0) log_warning_errno(r, "Could not remove route: %m"); - else { - /* route may not be exist in kernel. If we fail still remove it */ - route->link->link_messages++; + else route_free(route); - } return 1; } @@ -557,14 +549,12 @@ int route_configure( if (r < 0) return log_error_errno(r, "Could not set flags: %m"); - if (route->table != RT_TABLE_DEFAULT) { - + if (route->table != RT_TABLE_MAIN) { if (route->table < 256) { r = sd_rtnl_message_route_set_table(req, route->table); if (r < 0) return log_error_errno(r, "Could not set route table: %m"); } else { - r = sd_rtnl_message_route_set_table(req, RT_TABLE_UNSPEC); if (r < 0) return log_error_errno(r, "Could not set route table: %m"); diff --git a/src/nss-mymachines/nss-mymachines.c b/src/nss-mymachines/nss-mymachines.c index 8d57b26cbc..895f61c462 100644 --- a/src/nss-mymachines/nss-mymachines.c +++ b/src/nss-mymachines/nss-mymachines.c @@ -25,6 +25,7 @@ #include "alloc-util.h" #include "bus-common-errors.h" +#include "env-util.h" #include "hostname-util.h" #include "in-addr-util.h" #include "macro.h" @@ -434,6 +435,12 @@ enum nss_status _nss_mymachines_getpwnam_r( if (!machine_name_is_valid(machine)) goto not_found; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) + /* Make sure we can't deadlock if we are invoked by dbus-daemon. This way, it won't be able to resolve + * these UIDs, but that should be unproblematic as containers should never be able to connect to a bus + * running on the host. */ + goto not_found; + r = sd_bus_open_system(&bus); if (r < 0) goto fail; @@ -514,6 +521,9 @@ enum nss_status _nss_mymachines_getpwuid_r( if (uid < HOST_UID_LIMIT) goto not_found; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) + goto not_found; + r = sd_bus_open_system(&bus); if (r < 0) goto fail; @@ -605,6 +615,9 @@ enum nss_status _nss_mymachines_getgrnam_r( if (!machine_name_is_valid(machine)) goto not_found; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) + goto not_found; + r = sd_bus_open_system(&bus); if (r < 0) goto fail; @@ -682,6 +695,9 @@ enum nss_status _nss_mymachines_getgrgid_r( if (gid < HOST_GID_LIMIT) goto not_found; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) + goto not_found; + r = sd_bus_open_system(&bus); if (r < 0) goto fail; diff --git a/src/nss-systemd/nss-systemd.c b/src/nss-systemd/nss-systemd.c index 7078c0c50c..17d04e958d 100644 --- a/src/nss-systemd/nss-systemd.c +++ b/src/nss-systemd/nss-systemd.c @@ -21,11 +21,14 @@ #include "sd-bus.h" +#include "alloc-util.h" #include "bus-common-errors.h" #include "env-util.h" +#include "fs-util.h" #include "macro.h" #include "nss-util.h" #include "signal-util.h" +#include "stdio-util.h" #include "string-util.h" #include "user-util.h" #include "util.h" @@ -75,15 +78,50 @@ static const struct group nobody_group = { NSS_GETPW_PROTOTYPES(systemd); NSS_GETGR_PROTOTYPES(systemd); +static int direct_lookup_name(const char *name, uid_t *ret) { + _cleanup_free_ char *s = NULL; + const char *path; + int r; + + assert(name); + + /* Normally, we go via the bus to resolve names. That has the benefit that it is available from any mount + * namespace and subject to proper authentication. However, there's one problem: if our module is called from + * dbus-daemon itself we really can't use D-Bus to communicate. In this case, resort to a client-side hack, + * and look for the dynamic names directly. This is pretty ugly, but breaks the cyclic dependency. */ + + path = strjoina("/run/systemd/dynamic-uid/direct:", name); + r = readlink_malloc(path, &s); + if (r < 0) + return r; + + return parse_uid(s, ret); +} + +static int direct_lookup_uid(uid_t uid, char **ret) { + char path[strlen("/run/systemd/dynamic-uid/direct:") + DECIMAL_STR_MAX(uid_t) + 1], *s; + int r; + + xsprintf(path, "/run/systemd/dynamic-uid/direct:" UID_FMT, uid); + + r = readlink_malloc(path, &s); + if (r < 0) + return r; + if (!valid_user_group_name(s)) { /* extra safety check */ + free(s); + return -EINVAL; + } + + *ret = s; + return 0; +} + enum nss_status _nss_systemd_getpwnam_r( const char *name, struct passwd *pwd, char *buffer, size_t buflen, int *errnop) { - _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; - _cleanup_(sd_bus_message_unrefp) sd_bus_message* reply = NULL; - _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL; uint32_t translated; size_t l; int r; @@ -114,30 +152,45 @@ enum nss_status _nss_systemd_getpwnam_r( if (getenv_bool("SYSTEMD_NSS_DYNAMIC_BYPASS") > 0) goto not_found; - r = sd_bus_open_system(&bus); - if (r < 0) - goto fail; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) { - r = sd_bus_call_method(bus, - "org.freedesktop.systemd1", - "/org/freedesktop/systemd1", - "org.freedesktop.systemd1.Manager", - "LookupDynamicUserByName", - &error, - &reply, - "s", - name); - if (r < 0) { - if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + /* Access the dynamic UID allocation directly if we are called from dbus-daemon, see above. */ + r = direct_lookup_name(name, (uid_t*) &translated); + if (r == -ENOENT) goto not_found; - - goto fail; + if (r < 0) + goto fail; + + } else { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message* reply = NULL; + _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL; + + r = sd_bus_open_system(&bus); + if (r < 0) + goto fail; + + r = sd_bus_call_method(bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "LookupDynamicUserByName", + &error, + &reply, + "s", + name); + if (r < 0) { + if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + goto not_found; + + goto fail; + } + + r = sd_bus_message_read(reply, "u", &translated); + if (r < 0) + goto fail; } - r = sd_bus_message_read(reply, "u", &translated); - if (r < 0) - goto fail; - l = strlen(name); if (buflen < l+1) { *errnop = ENOMEM; @@ -175,6 +228,7 @@ enum nss_status _nss_systemd_getpwuid_r( _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; _cleanup_(sd_bus_message_unrefp) sd_bus_message* reply = NULL; _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL; + _cleanup_free_ char *direct = NULL; const char *translated; size_t l; int r; @@ -204,30 +258,42 @@ enum nss_status _nss_systemd_getpwuid_r( if (getenv_bool("SYSTEMD_NSS_DYNAMIC_BYPASS") > 0) goto not_found; - r = sd_bus_open_system(&bus); - if (r < 0) - goto fail; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) { - r = sd_bus_call_method(bus, - "org.freedesktop.systemd1", - "/org/freedesktop/systemd1", - "org.freedesktop.systemd1.Manager", - "LookupDynamicUserByUID", - &error, - &reply, - "u", - (uint32_t) uid); - if (r < 0) { - if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + r = direct_lookup_uid(uid, &direct); + if (r == -ENOENT) goto not_found; - - goto fail; + if (r < 0) + goto fail; + + translated = direct; + + } else { + r = sd_bus_open_system(&bus); + if (r < 0) + goto fail; + + r = sd_bus_call_method(bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "LookupDynamicUserByUID", + &error, + &reply, + "u", + (uint32_t) uid); + if (r < 0) { + if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + goto not_found; + + goto fail; + } + + r = sd_bus_message_read(reply, "s", &translated); + if (r < 0) + goto fail; } - r = sd_bus_message_read(reply, "s", &translated); - if (r < 0) - goto fail; - l = strlen(translated) + 1; if (buflen < l) { *errnop = ENOMEM; @@ -262,9 +328,6 @@ enum nss_status _nss_systemd_getgrnam_r( char *buffer, size_t buflen, int *errnop) { - _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; - _cleanup_(sd_bus_message_unrefp) sd_bus_message* reply = NULL; - _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL; uint32_t translated; size_t l; int r; @@ -294,30 +357,45 @@ enum nss_status _nss_systemd_getgrnam_r( if (getenv_bool("SYSTEMD_NSS_DYNAMIC_BYPASS") > 0) goto not_found; - r = sd_bus_open_system(&bus); - if (r < 0) - goto fail; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) { - r = sd_bus_call_method(bus, - "org.freedesktop.systemd1", - "/org/freedesktop/systemd1", - "org.freedesktop.systemd1.Manager", - "LookupDynamicUserByName", - &error, - &reply, - "s", - name); - if (r < 0) { - if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + /* Access the dynamic GID allocation directly if we are called from dbus-daemon, see above. */ + r = direct_lookup_name(name, (uid_t*) &translated); + if (r == -ENOENT) goto not_found; - - goto fail; + if (r < 0) + goto fail; + } else { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message* reply = NULL; + _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL; + + r = sd_bus_open_system(&bus); + if (r < 0) + goto fail; + + r = sd_bus_call_method(bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "LookupDynamicUserByName", + &error, + &reply, + "s", + name); + if (r < 0) { + if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + goto not_found; + + goto fail; + } + + r = sd_bus_message_read(reply, "u", &translated); + if (r < 0) + goto fail; } - r = sd_bus_message_read(reply, "u", &translated); - if (r < 0) - goto fail; - l = sizeof(char*) + strlen(name) + 1; if (buflen < l) { *errnop = ENOMEM; @@ -353,6 +431,7 @@ enum nss_status _nss_systemd_getgrgid_r( _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; _cleanup_(sd_bus_message_unrefp) sd_bus_message* reply = NULL; _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL; + _cleanup_free_ char *direct = NULL; const char *translated; size_t l; int r; @@ -382,30 +461,41 @@ enum nss_status _nss_systemd_getgrgid_r( if (getenv_bool("SYSTEMD_NSS_DYNAMIC_BYPASS") > 0) goto not_found; - r = sd_bus_open_system(&bus); - if (r < 0) - goto fail; + if (getenv_bool("SYSTEMD_NSS_BYPASS_BUS") > 0) { - r = sd_bus_call_method(bus, - "org.freedesktop.systemd1", - "/org/freedesktop/systemd1", - "org.freedesktop.systemd1.Manager", - "LookupDynamicUserByUID", - &error, - &reply, - "u", - (uint32_t) gid); - if (r < 0) { - if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + r = direct_lookup_uid(gid, &direct); + if (r == -ENOENT) goto not_found; - - goto fail; + if (r < 0) + goto fail; + + translated = direct; + } else { + r = sd_bus_open_system(&bus); + if (r < 0) + goto fail; + + r = sd_bus_call_method(bus, + "org.freedesktop.systemd1", + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "LookupDynamicUserByUID", + &error, + &reply, + "u", + (uint32_t) gid); + if (r < 0) { + if (sd_bus_error_has_name(&error, BUS_ERROR_NO_SUCH_DYNAMIC_USER)) + goto not_found; + + goto fail; + } + + r = sd_bus_message_read(reply, "s", &translated); + if (r < 0) + goto fail; } - r = sd_bus_message_read(reply, "s", &translated); - if (r < 0) - goto fail; - l = sizeof(char*) + strlen(translated) + 1; if (buflen < l) { *errnop = ENOMEM; diff --git a/src/shared/bus-unit-util.c b/src/shared/bus-unit-util.c index 7774d607c7..ab30afb527 100644 --- a/src/shared/bus-unit-util.c +++ b/src/shared/bus-unit-util.c @@ -204,7 +204,7 @@ int bus_append_unit_property_assignment(sd_bus_message *m, const char *assignmen "IgnoreSIGPIPE", "TTYVHangup", "TTYReset", "RemainAfterExit", "PrivateTmp", "PrivateDevices", "PrivateNetwork", "PrivateUsers", "NoNewPrivileges", "SyslogLevelPrefix", "Delegate", "RemainAfterElapse", "MemoryDenyWriteExecute", - "RestrictRealtime", "DynamicUser")) { + "RestrictRealtime", "DynamicUser", "RemoveIPC")) { r = parse_boolean(eq); if (r < 0) @@ -212,6 +212,17 @@ int bus_append_unit_property_assignment(sd_bus_message *m, const char *assignmen r = sd_bus_message_append(m, "v", "b", r); + } else if (STR_IN_SET(field, "CPUWeight", "StartupCPUWeight")) { + uint64_t u; + + r = cg_weight_parse(eq, &u); + if (r < 0) { + log_error("Failed to parse %s value %s.", field, eq); + return -EINVAL; + } + + r = sd_bus_message_append(m, "v", "t", u); + } else if (STR_IN_SET(field, "CPUShares", "StartupCPUShares")) { uint64_t u; diff --git a/src/shared/clean-ipc.c b/src/shared/clean-ipc.c index a3ac7aeb82..d5db604f03 100644 --- a/src/shared/clean-ipc.c +++ b/src/shared/clean-ipc.c @@ -41,8 +41,20 @@ #include "macro.h" #include "string-util.h" #include "strv.h" +#include "user-util.h" -static int clean_sysvipc_shm(uid_t delete_uid) { +static bool match_uid_gid(uid_t subject_uid, gid_t subject_gid, uid_t delete_uid, gid_t delete_gid) { + + if (uid_is_valid(delete_uid) && subject_uid == delete_uid) + return true; + + if (gid_is_valid(delete_gid) && subject_gid == delete_gid) + return true; + + return false; +} + +static int clean_sysvipc_shm(uid_t delete_uid, gid_t delete_gid) { _cleanup_fclose_ FILE *f = NULL; char line[LINE_MAX]; bool first = true; @@ -77,7 +89,7 @@ static int clean_sysvipc_shm(uid_t delete_uid) { if (n_attached > 0) continue; - if (uid != delete_uid) + if (!match_uid_gid(uid, gid, delete_uid, delete_gid)) continue; if (shmctl(shmid, IPC_RMID, NULL) < 0) { @@ -89,7 +101,8 @@ static int clean_sysvipc_shm(uid_t delete_uid) { ret = log_warning_errno(errno, "Failed to remove SysV shared memory segment %i: %m", shmid); - } + } else + log_debug("Removed SysV shared memory segment %i.", shmid); } return ret; @@ -98,7 +111,7 @@ fail: return log_warning_errno(errno, "Failed to read /proc/sysvipc/shm: %m"); } -static int clean_sysvipc_sem(uid_t delete_uid) { +static int clean_sysvipc_sem(uid_t delete_uid, gid_t delete_gid) { _cleanup_fclose_ FILE *f = NULL; char line[LINE_MAX]; bool first = true; @@ -128,7 +141,7 @@ static int clean_sysvipc_sem(uid_t delete_uid) { &semid, &uid, &gid, &cuid, &cgid) != 5) continue; - if (uid != delete_uid) + if (!match_uid_gid(uid, gid, delete_uid, delete_gid)) continue; if (semctl(semid, 0, IPC_RMID) < 0) { @@ -140,7 +153,8 @@ static int clean_sysvipc_sem(uid_t delete_uid) { ret = log_warning_errno(errno, "Failed to remove SysV semaphores object %i: %m", semid); - } + } else + log_debug("Removed SysV semaphore %i.", semid); } return ret; @@ -149,7 +163,7 @@ fail: return log_warning_errno(errno, "Failed to read /proc/sysvipc/sem: %m"); } -static int clean_sysvipc_msg(uid_t delete_uid) { +static int clean_sysvipc_msg(uid_t delete_uid, gid_t delete_gid) { _cleanup_fclose_ FILE *f = NULL; char line[LINE_MAX]; bool first = true; @@ -180,7 +194,7 @@ static int clean_sysvipc_msg(uid_t delete_uid) { &msgid, &cpid, &lpid, &uid, &gid, &cuid, &cgid) != 7) continue; - if (uid != delete_uid) + if (!match_uid_gid(uid, gid, delete_uid, delete_gid)) continue; if (msgctl(msgid, IPC_RMID, NULL) < 0) { @@ -192,7 +206,8 @@ static int clean_sysvipc_msg(uid_t delete_uid) { ret = log_warning_errno(errno, "Failed to remove SysV message queue %i: %m", msgid); - } + } else + log_debug("Removed SysV message queue %i.", msgid); } return ret; @@ -201,13 +216,13 @@ fail: return log_warning_errno(errno, "Failed to read /proc/sysvipc/msg: %m"); } -static int clean_posix_shm_internal(DIR *dir, uid_t uid) { +static int clean_posix_shm_internal(DIR *dir, uid_t uid, gid_t gid) { struct dirent *de; int ret = 0, r; assert(dir); - FOREACH_DIRENT(de, dir, goto fail) { + FOREACH_DIRENT_ALL(de, dir, goto fail) { struct stat st; if (STR_IN_SET(de->d_name, "..", ".")) @@ -217,12 +232,11 @@ static int clean_posix_shm_internal(DIR *dir, uid_t uid) { if (errno == ENOENT) continue; - log_warning_errno(errno, "Failed to stat() POSIX shared memory segment %s: %m", de->d_name); - ret = -errno; + ret = log_warning_errno(errno, "Failed to stat() POSIX shared memory segment %s: %m", de->d_name); continue; } - if (st.st_uid != uid) + if (!match_uid_gid(st.st_uid, st.st_gid, uid, gid)) continue; if (S_ISDIR(st.st_mode)) { @@ -230,12 +244,10 @@ static int clean_posix_shm_internal(DIR *dir, uid_t uid) { kid = xopendirat(dirfd(dir), de->d_name, O_NOFOLLOW|O_NOATIME); if (!kid) { - if (errno != ENOENT) { - log_warning_errno(errno, "Failed to enter shared memory directory %s: %m", de->d_name); - ret = -errno; - } + if (errno != ENOENT) + ret = log_warning_errno(errno, "Failed to enter shared memory directory %s: %m", de->d_name); } else { - r = clean_posix_shm_internal(kid, uid); + r = clean_posix_shm_internal(kid, uid, gid); if (r < 0) ret = r; } @@ -245,9 +257,9 @@ static int clean_posix_shm_internal(DIR *dir, uid_t uid) { if (errno == ENOENT) continue; - log_warning_errno(errno, "Failed to remove POSIX shared memory directory %s: %m", de->d_name); - ret = -errno; - } + ret = log_warning_errno(errno, "Failed to remove POSIX shared memory directory %s: %m", de->d_name); + } else + log_debug("Removed POSIX shared memory directory %s", de->d_name); } else { if (unlinkat(dirfd(dir), de->d_name, 0) < 0) { @@ -255,20 +267,19 @@ static int clean_posix_shm_internal(DIR *dir, uid_t uid) { if (errno == ENOENT) continue; - log_warning_errno(errno, "Failed to remove POSIX shared memory segment %s: %m", de->d_name); - ret = -errno; - } + ret = log_warning_errno(errno, "Failed to remove POSIX shared memory segment %s: %m", de->d_name); + } else + log_debug("Removed POSIX shared memory segment %s", de->d_name); } } return ret; fail: - log_warning_errno(errno, "Failed to read /dev/shm: %m"); - return -errno; + return log_warning_errno(errno, "Failed to read /dev/shm: %m"); } -static int clean_posix_shm(uid_t uid) { +static int clean_posix_shm(uid_t uid, gid_t gid) { _cleanup_closedir_ DIR *dir = NULL; dir = opendir("/dev/shm"); @@ -279,10 +290,10 @@ static int clean_posix_shm(uid_t uid) { return log_warning_errno(errno, "Failed to open /dev/shm: %m"); } - return clean_posix_shm_internal(dir, uid); + return clean_posix_shm_internal(dir, uid, gid); } -static int clean_posix_mq(uid_t uid) { +static int clean_posix_mq(uid_t uid, gid_t gid) { _cleanup_closedir_ DIR *dir = NULL; struct dirent *de; int ret = 0; @@ -295,7 +306,7 @@ static int clean_posix_mq(uid_t uid) { return log_warning_errno(errno, "Failed to open /dev/mqueue: %m"); } - FOREACH_DIRENT(de, dir, goto fail) { + FOREACH_DIRENT_ALL(de, dir, goto fail) { struct stat st; char fn[1+strlen(de->d_name)+1]; @@ -312,7 +323,7 @@ static int clean_posix_mq(uid_t uid) { continue; } - if (st.st_uid != uid) + if (!match_uid_gid(st.st_uid, st.st_gid, uid, gid)) continue; fn[0] = '/'; @@ -325,7 +336,8 @@ static int clean_posix_mq(uid_t uid) { ret = log_warning_errno(errno, "Failed to unlink POSIX message queue %s: %m", fn); - } + } else + log_debug("Removed POSIX message queue %s", fn); } return ret; @@ -334,32 +346,44 @@ fail: return log_warning_errno(errno, "Failed to read /dev/mqueue: %m"); } -int clean_ipc(uid_t uid) { +int clean_ipc(uid_t uid, gid_t gid) { int ret = 0, r; - /* Refuse to clean IPC of the root and system users */ - if (uid <= SYSTEM_UID_MAX) + /* Anything to do? */ + if (!uid_is_valid(uid) && !gid_is_valid(gid)) + return 0; + + /* Refuse to clean IPC of the root user */ + if (uid == 0 && gid == 0) return 0; - r = clean_sysvipc_shm(uid); + r = clean_sysvipc_shm(uid, gid); if (r < 0) ret = r; - r = clean_sysvipc_sem(uid); + r = clean_sysvipc_sem(uid, gid); if (r < 0) ret = r; - r = clean_sysvipc_msg(uid); + r = clean_sysvipc_msg(uid, gid); if (r < 0) ret = r; - r = clean_posix_shm(uid); + r = clean_posix_shm(uid, gid); if (r < 0) ret = r; - r = clean_posix_mq(uid); + r = clean_posix_mq(uid, gid); if (r < 0) ret = r; return ret; } + +int clean_ipc_by_uid(uid_t uid) { + return clean_ipc(uid, GID_INVALID); +} + +int clean_ipc_by_gid(gid_t gid) { + return clean_ipc(UID_INVALID, gid); +} diff --git a/src/shared/clean-ipc.h b/src/shared/clean-ipc.h index 44a83afcf7..6ca57f44fd 100644 --- a/src/shared/clean-ipc.h +++ b/src/shared/clean-ipc.h @@ -21,4 +21,6 @@ #include <sys/types.h> -int clean_ipc(uid_t uid); +int clean_ipc(uid_t uid, gid_t gid); +int clean_ipc_by_uid(uid_t uid); +int clean_ipc_by_gid(gid_t gid); diff --git a/src/sysv-generator/sysv-generator.c b/src/sysv-generator/sysv-generator.c index 3ed8f23ff9..39821687b9 100644 --- a/src/sysv-generator/sysv-generator.c +++ b/src/sysv-generator/sysv-generator.c @@ -247,7 +247,7 @@ static char *sysv_translate_name(const char *name) { return res; } -static int sysv_translate_facility(const char *name, const char *filename, char **ret) { +static int sysv_translate_facility(SysvStub *s, unsigned line, const char *name, char **ret) { /* We silently ignore the $ prefix here. According to the LSB * spec it simply indicates whether something is a @@ -266,15 +266,18 @@ static int sysv_translate_facility(const char *name, const char *filename, char "time", SPECIAL_TIME_SYNC_TARGET, }; + const char *filename; char *filename_no_sh, *e, *m; const char *n; unsigned i; int r; assert(name); - assert(filename); + assert(s); assert(ret); + filename = basename(s->path); + n = *name == '$' ? name + 1 : name; for (i = 0; i < ELEMENTSOF(table); i += 2) { @@ -299,7 +302,7 @@ static int sysv_translate_facility(const char *name, const char *filename, char if (*name == '$') { r = unit_name_build(n, NULL, ".target", ret); if (r < 0) - return log_error_errno(r, "Failed to build name: %m"); + return log_error_errno(r, "[%s:%u] Could not build name for facility %s: %m", s->path, line, name); return r; } @@ -337,11 +340,11 @@ static int handle_provides(SysvStub *s, unsigned line, const char *full_text, co r = extract_first_word(&text, &word, NULL, EXTRACT_QUOTES|EXTRACT_RELAX); if (r < 0) - return log_error_errno(r, "Failed to parse word from provides string: %m"); + return log_error_errno(r, "[%s:%u] Failed to parse word from provides string: %m", s->path, line); if (r == 0) break; - r = sysv_translate_facility(word, basename(s->path), &m); + r = sysv_translate_facility(s, line, word, &m); if (r <= 0) /* continue on error */ continue; @@ -403,11 +406,11 @@ static int handle_dependencies(SysvStub *s, unsigned line, const char *full_text r = extract_first_word(&text, &word, NULL, EXTRACT_QUOTES|EXTRACT_RELAX); if (r < 0) - return log_error_errno(r, "Failed to parse word from provides string: %m"); + return log_error_errno(r, "[%s:%u] Failed to parse word from provides string: %m", s->path, line); if (r == 0) break; - r = sysv_translate_facility(word, basename(s->path), &m); + r = sysv_translate_facility(s, line, word, &m); if (r <= 0) /* continue on error */ continue; diff --git a/src/test/test-condition.c b/src/test/test-condition.c index 66003aa6bd..6f7d71ef9a 100644 --- a/src/test/test-condition.c +++ b/src/test/test-condition.c @@ -37,66 +37,82 @@ static void test_condition_test_path(void) { Condition *condition; condition = condition_new(CONDITION_PATH_EXISTS, "/bin/sh", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_EXISTS, "/bin/s?", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_EXISTS_GLOB, "/bin/s?", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_EXISTS_GLOB, "/bin/s?", false, true); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_EXISTS, "/thiscertainlywontexist", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_EXISTS, "/thiscertainlywontexist", false, true); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_IS_DIRECTORY, "/bin", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_DIRECTORY_NOT_EMPTY, "/bin", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_FILE_NOT_EMPTY, "/bin/sh", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_FILE_IS_EXECUTABLE, "/bin/sh", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_FILE_IS_EXECUTABLE, "/etc/passwd", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_IS_MOUNT_POINT, "/proc", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_IS_MOUNT_POINT, "/", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_IS_MOUNT_POINT, "/bin", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_IS_READ_WRITE, "/tmp", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_PATH_IS_SYMBOLIC_LINK, "/dev/stdout", false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); } @@ -105,38 +121,44 @@ static void test_condition_test_ac_power(void) { Condition *condition; condition = condition_new(CONDITION_AC_POWER, "true", false, false); + assert_se(condition); assert_se(condition_test(condition) == on_ac_power()); condition_free(condition); condition = condition_new(CONDITION_AC_POWER, "false", false, false); + assert_se(condition); assert_se(condition_test(condition) != on_ac_power()); condition_free(condition); condition = condition_new(CONDITION_AC_POWER, "false", false, true); + assert_se(condition); assert_se(condition_test(condition) == on_ac_power()); condition_free(condition); } static void test_condition_test_host(void) { + _cleanup_free_ char *hostname = NULL; + char sid[SD_ID128_STRING_MAX]; Condition *condition; sd_id128_t id; int r; - char sid[SD_ID128_STRING_MAX]; - _cleanup_free_ char *hostname = NULL; r = sd_id128_get_machine(&id); assert_se(r >= 0); assert_se(sd_id128_to_string(id, sid)); condition = condition_new(CONDITION_HOST, sid, false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_HOST, "garbage value jjjjjjjjjjjjjj", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_HOST, sid, false, true); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); @@ -148,6 +170,7 @@ static void test_condition_test_host(void) { log_notice("hostname is an id128, skipping test"); else { condition = condition_new(CONDITION_HOST, hostname, false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); } @@ -165,14 +188,17 @@ static void test_condition_test_architecture(void) { assert_se(sa); condition = condition_new(CONDITION_ARCHITECTURE, sa, false, false); + assert_se(condition); assert_se(condition_test(condition) > 0); condition_free(condition); condition = condition_new(CONDITION_ARCHITECTURE, "garbage value", false, false); + assert_se(condition); assert_se(condition_test(condition) == 0); condition_free(condition); condition = condition_new(CONDITION_ARCHITECTURE, sa, false, true); + assert_se(condition); assert_se(condition_test(condition) == 0); condition_free(condition); } @@ -181,10 +207,12 @@ static void test_condition_test_kernel_command_line(void) { Condition *condition; condition = condition_new(CONDITION_KERNEL_COMMAND_LINE, "thisreallyshouldntbeonthekernelcommandline", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_KERNEL_COMMAND_LINE, "andthis=neither", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); } @@ -193,10 +221,12 @@ static void test_condition_test_null(void) { Condition *condition; condition = condition_new(CONDITION_NULL, NULL, false, false); + assert_se(condition); assert_se(condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_NULL, NULL, false, true); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); } @@ -205,31 +235,36 @@ static void test_condition_test_security(void) { Condition *condition; condition = condition_new(CONDITION_SECURITY, "garbage oifdsjfoidsjoj", false, false); + assert_se(condition); assert_se(!condition_test(condition)); condition_free(condition); condition = condition_new(CONDITION_SECURITY, "selinux", false, true); + assert_se(condition); assert_se(condition_test(condition) != mac_selinux_have()); condition_free(condition); condition = condition_new(CONDITION_SECURITY, "ima", false, false); + assert_se(condition); assert_se(condition_test(condition) == use_ima()); condition_free(condition); condition = condition_new(CONDITION_SECURITY, "apparmor", false, false); + assert_se(condition); assert_se(condition_test(condition) == mac_apparmor_use()); condition_free(condition); condition = condition_new(CONDITION_SECURITY, "smack", false, false); + assert_se(condition); assert_se(condition_test(condition) == mac_smack_use()); condition_free(condition); condition = condition_new(CONDITION_SECURITY, "audit", false, false); + assert_se(condition); assert_se(condition_test(condition) == use_audit()); condition_free(condition); } - int main(int argc, char *argv[]) { log_parse_environment(); log_open(); diff --git a/src/test/test-hostname-util.c b/src/test/test-hostname-util.c index 17fde9f27e..1c3d13ed1d 100644 --- a/src/test/test-hostname-util.c +++ b/src/test/test-hostname-util.c @@ -42,6 +42,7 @@ static void test_hostname_is_valid(void) { assert_se(!hostname_is_valid("foo..bar", false)); assert_se(!hostname_is_valid("foo.bar..", false)); assert_se(!hostname_is_valid("xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx", false)); + assert_se(!hostname_is_valid("au-xph5-rvgrdsb5hcxc-47et3a5vvkrc-server-wyoz4elpdpe3.openstack.local", false)); assert_se(hostname_is_valid("foobar", true)); assert_se(hostname_is_valid("foobar.com", true)); diff --git a/src/test/test-ipcrm.c b/src/test/test-ipcrm.c index c5bcaf47bb..551eba7215 100644 --- a/src/test/test-ipcrm.c +++ b/src/test/test-ipcrm.c @@ -32,5 +32,5 @@ int main(int argc, char *argv[]) { return EXIT_FAILURE; } - return clean_ipc(uid) < 0 ? EXIT_FAILURE : EXIT_SUCCESS; + return clean_ipc_by_uid(uid) < 0 ? EXIT_FAILURE : EXIT_SUCCESS; } diff --git a/test/udev-test.pl b/test/udev-test.pl index 129af854f1..9723386b23 100755 --- a/test/udev-test.pl +++ b/test/udev-test.pl @@ -37,7 +37,7 @@ my $EXIT_TEST_SKIP = 77; my $rules_10k_tags = ""; for (my $i = 1; $i <= 10000; ++$i) { - $rules_10k_tags .= 'KERNEL=="sda", TAG+="test' . $i . "\"\n"; + $rules_10k_tags .= 'KERNEL=="sda", TAG+="test' . $i . "\"\n"; } my @tests = ( @@ -1596,6 +1596,6 @@ system("umount", "$udev_tmpfs"); rmdir($udev_tmpfs); if ($error > 0) { - exit(1); + exit(1); } exit(0); diff --git a/units/systemd-random-seed.service.in b/units/systemd-random-seed.service.in index 115233268d..b244a8ce43 100644 --- a/units/systemd-random-seed.service.in +++ b/units/systemd-random-seed.service.in @@ -13,6 +13,7 @@ RequiresMountsFor=@RANDOM_SEED@ Conflicts=shutdown.target After=systemd-remount-fs.service Before=sysinit.target shutdown.target +ConditionVirtualization=!container [Service] Type=oneshot |