1 systemd System and Service Manager
5 * A new tool systemd-sysusers has been added. This tool
6 creates system users and groups in /etc/passwd and
7 /etc/group, based on static declarative system user/group
8 definitions in /usr/lib/sysusers.d/. This is useful to
9 enable factory resets and volatile systems that boot up with
10 an empty /etc directory, and thus need system users and
11 groups created during early boot. systemd now also ships
12 with two default sysusers.d/ files for the most basic
13 users and groups systemd and the core operating system
16 * A new tmpfiles snippet has been added that rebuilds the
17 essential files in /etc on boot, should they be missing.
19 * A directive for ensuring automatic clean-up of
20 /var/cache/man/ has been removed from the default
21 configuration. This line should now be shipped by the man
22 implementation. The necessary change has been made to the
23 man-db implementation. Note that you need to update your man
24 implementation to one that ships this line, otherwise no
25 automatic clean-up of /var/cache/man will take place.
27 * A new condition ConditionNeedsUpdate= has been added that
28 may conditionalize services to only run when /etc or /var
29 are "older" than the vendor operating system resources in
30 /usr. This is useful for reconstructing or updating /etc
31 after an offline update of /usr or a factory reset, on the
32 next reboot. Services that want to run once after such an
33 update or reset should use this condition and order
34 themselves before the new systemd-update-done.service, which
35 will mark the two directories as fully updated. A number of
36 service files have been added making use of this, to rebuild
37 the udev hardware database, the journald message catalog and
38 dynamic loader cache (ldconfig). The systemd-sysusers tool
39 described above also makes use of this now. With this in
40 place it is now possible to start up a minimal operating
41 system with /etc empty cleanly. Fore more information on the
42 concepts involved see this recent blog story:
44 http://0pointer.de/blog/projects/stateless.html
46 * A new system group "input" has been introduced, and all
47 input device nodes get this group assigned. This is useful
48 for system-level software to get access to input devices. It
49 complements what is already done for "audio" and "video".
51 * systemd-networkd learnt minimal DHCPv4 server support in
52 addition to the existing DHCPv4 client support. It also
53 learnt DHCPv6 client and IPv6 Router Solicitation client
54 support. The DHCPv4 client gained support for static routes
55 passed in from the server. Note that the [DHCPv4] section
56 known in older systemd-networkd versions has been renamed to
57 [DHCP] and is now also used by the DHCPv6 client. Existing
58 .network files using settings of this section need to be
61 * networkd gained support for vxlan virtual networks.
63 * networkd gained support for automatic allocation of address
64 ranges for interfaces from a system-wide pool of
65 addresses. This is useful for dynamically managing a large
66 number of interfaces with a single network configuration
67 file. In particular this is useful to easily assign
68 appropriate IP addresses to the veth links of a large number
71 * RPM macros for processing sysusers, sysctl and binfmt
72 drop-in snippets at package installation time have been
75 * The /etc/os-release file should now be placed in
76 /usr/lib/os-release. The old location is automatically
77 created as symlink. /usr/lib is the more appropriate
78 location of this file, since it shall actually describe the
79 vendor operating system shipped in /usr, and not the
80 configuration stored in /etc.
82 * .mount units gained a new boolean SloppyOptions= setting
83 that maps to mount(8)'s -s option which enables permissive
84 parsing of unknown mount options.
86 * tmpfiles learnt a new "L+" directive which creates a symlink
87 but (unlike "L") deletes a pre-existing file first, should
88 it already exist and not already be the correct
89 symlink. Similar, "b+", "c+" and "p+" directives have been
90 added as well, which create block and character devices, as
91 well as fifos in the filesystem, possibly removing any
92 pre-existing files of different types.
94 * For tmpfiles' "L", "L+", "C" and "C+" directives the final
95 'argument' field (which so far specified the source to
96 symlink/copy the files from) is now optional. If ommited the
97 same file os copied from /usr/share/factory/ suffixed by the
98 full destination path. This is useful for populating /etc
99 with essential files, by copying them from vendor defaults
100 shipped in /usr/share/factory/etc.
102 * A new command "systemctl preset-all" has been added that
103 applies the service preset settings to all installed unit
104 files. A new switch --preset-mode= has been added that
105 controls whether only enable or only disable operations
108 * A new command "systemctl is-system-running" has been added
109 that allows checking the overall state of the system, for
110 example whether it is fully up an running.
112 * When the system boots up with an empty /etc, the equivalent
113 to "systemctl preset-all" is executed during early boot, to
114 make sure all default services are enabled after a factory
117 * systemd now contains a minimal preset file that enables the
118 most basic services systemd ships by default.
120 * Unit files' [Install] section gained a new DefaultInstance=
121 field for defining the default instance to create if a
122 template unit is enabled with no instance specified.
124 * A new passive target cryptsetup-pre.target has been added
125 that may be used by services that need to make they run and
126 finish before the first LUKS cryptographic device is set up.
128 * The /dev/loop-control and /dev/btrfs-control device nodes
129 are now owned by the "disk" group by default, opening up
130 access to this group.
132 * systemd-coredump will now automatically generate a
133 stack trace of all core dumps taking place on the system,
134 based on elfutils' libdw library. This stack trace is logged
137 * systemd-coredump may now optionally store coredumps directly
138 on disk (in /var/lib/systemd/coredump, possibly compressed),
139 instead of storing them unconditionally in the journal. This
140 mode is the new default. A new configuration file
141 /etc/systemd/coredump.conf has been added to configure this
142 and other parameters of systemd-coredump.
144 * coredumpctl gained a new "info" verb to show details about a
145 specific coredump. A new switch "-1" has also been added
146 that makes sure to only show information about the most
147 recent entry instead of all entries. Also, as the tool is
148 generally useful now the "systemd-" prefix of the binary
149 name has been removed. Distributions that want to maintain
150 compatibility with the old name should add a symlink from
151 the old name to the new name.
153 * journald's SplitMode= now defaults to "uid". This makes sure
154 that unpriviliged users can access their own coredumps with
155 coredumpctl without restrictions.
157 * New kernel command line options "systemd.wants=" (for
158 pulling an additional unit during boot), "systemd.mask="
159 (for masking a specific unit for the boot), and
160 "systemd.debug-shell" (for enabling the debug shell on tty9)
161 have been added. This is implemented in the new generator
162 "systemd-debug-generator".
164 * systemd-nspawn will now by default filter a couple of
165 syscalls for containers, among them those required for
166 kernel module loading, direct x86 IO port access, swap
167 management, and kexec. Most importantly though
168 open_by_handle_at() is now prohibited for containers,
169 closing a hole similar to a recently discussed vulnerability
170 in docker regarding access to files on file hierarchies the
171 container should normally not have access to. Note that for
172 nspawn we generally make no security claims anyway (and
173 this is explicitly documented in the man page), so this is
174 just a fix for one of the most obvious problems.
176 * A new man page file-hierarchy(7) has been added that
177 contains a minimized, modernized version of the file system
178 layout systemd expects, similar in style to the FHS
179 specification or hier(5).
181 * Automatic time-based clean-up of $XDG_RUNTIME_DIR is no
182 longer done. Since the directory now has a per-user size
183 limit, and is cleaned on logout this appears unnecessary,
184 in particular since this now brings the lifecycle of this
185 directory closer in line with how IPC objects are handled.
187 * systemd.pc now exports a number of additional directories,
188 including $libdir (which is useful to identify the library
189 path for the primary architecture of the system), and a
190 couple of drop-in directories.
192 * udev's predictable network interface names now use the dev_port
193 sysfs attribute, introduced in linux 3.15 instead of dev_id to
194 distinguish between ports of the same PCI function. dev_id should
195 only be used for ports using the same HW address, hence the need
200 * As an experimental feature, udev now tries to lock the
201 disk device node (flock(LOCK_SH|LOCK_NB)) while it
202 executes events for the disk or any of its partitions.
203 Applications like partitioning programs can lock the
204 disk device node (flock(LOCK_EX)) and claim temporary
205 device ownership that way; udev will entirely skip all event
206 handling for this disk and its partitions. If the disk
207 was opened for writing, the close will trigger a partition
208 table rescan in udev's "watch" facility, and if needed
209 synthesize "change" events for the disk and all its partitions.
210 This is now unconditionally enabled, and if it turns out to
211 cause major problems, we might turn it on only for specific
212 devices, or might need to disable it entirely. Device Mapper
213 devices are excluded from this logic.
215 * We temporarily dropped the "-l" switch for fsck invocations,
216 since they collide with the flock() logic above. util-linux
217 upstream has been changed already to avoid this conflict,
218 and we will readd "-l" as soon as util-linux with this
219 change has been released.
221 * The dependency on libattr has been removed. Since a long
222 time, the extended attribute calls have moved to glibc, and
223 libattr is thus unnecessary.
225 * Virtualization detection works without priviliges now. This
226 means the systemd-detect-virt binary no longer requires
227 CAP_SYS_PTRACE file capabilities, and our daemons can run
228 with fewer privileges.
230 * systemd-networkd now runs under its own "systemd-network"
231 user. It retains the CAP_NET_ADMIN, CAP_NET_BIND_SERVICE,
232 CAP_NET_BROADCAST, CAP_NET_RAW capabilities though, but
233 loses the ability to write to files owned by root this way.
235 * Similar, systemd-resolved now runs under its own
236 "systemd-resolve" user with no capabilities remaining.
238 * Similar, systemd-bus-proxyd now runs under its own
239 "systemd-bus-proxy" user with only CAP_IPC_OWNER remaining.
241 * systemd-networkd gained support for setting up "veth"
242 virtual ethernet devices for container connectivity, as well
243 as GRE and VTI tunnels.
245 * systemd-networkd will no longer automatically attempt to
246 manually load kernel modules necessary for certain tunnel
247 transports. Instead, it is assumed the kernel loads them
248 automatically when required. This only works correctly on
249 very new kernels. On older kernels, please consider adding
250 the kernel modules to /etc/modules-load.d/ as a work-around.
252 * The resolv.conf file systemd-resolved generates has been
253 moved to /run/systemd/resolve/. If you have a symlink from
254 /etc/resolv.conf, it might be necessary to correct it.
256 * Two new service settings, ProtectedHome= and ProtectedSystem=,
257 have been added. When enabled, they will make the user data
258 (such as /home) inaccessible or read-only and the system
259 (such as /usr) read-only, for specific services. This allows
260 very light-weight per-service sandboxing to avoid
261 modifications of user data or system files from
262 services. These two new switches have been enabled for all
263 of systemd's long-running services, where appropriate.
265 * Socket units gained new SocketUser= and SocketGroup=
266 settings to set the owner user and group of AF_UNIX sockets
267 and FIFOs in the file system.
269 * Socket units gained a new RemoveOnStop= setting. If enabled,
270 all FIFOS and sockets in the file system will be removed
271 when the specific socket unit is stopped.
273 * Socket units gained a new Symlinks= setting. It takes a list
274 of symlinks to create to file system sockets or FIFOs
275 created by the specific Unix sockets. This is useful to
276 manage symlinks to socket nodes with the same life-cycle as
279 * The /dev/log socket and /dev/initctl FIFO have been moved to
280 /run, and have been replaced by symlinks. This allows
281 connecting to these facilities even if PrivateDevices=yes is
282 used for a service (which makes /dev/log itself unavailable,
283 but /run is left). This also has the benefit of ensuring
284 that /dev only contains device nodes, directories and
285 symlinks, and nothing else.
287 * sd-daemon gained two new calls sd_pid_notify() and
288 sd_pid_notifyf(). They are similar to sd_notify() and
289 sd_notifyf(), but allow overriding of the source PID of
290 notification messages if permissions permit this. This is
291 useful to send notify messages on behalf of a different
292 process (for example, the parent process). The
293 systemd-notify tool has been updated to make use of this
294 when sending messages (so that notification messages now
295 originate from the shell script invoking systemd-notify and
296 not the systemd-notify process itself. This should minimize
297 a race where systemd fails to associate notification
298 messages to services when the originating process already
301 * A new "on-abnormal" setting for Restart= has been added. If
302 set, it will result in automatic restarts on all "abnormal"
303 reasons for a process to exit, which includes unclean
304 signals, core dumps, timeouts and watchdog timeouts, but
305 does not include clean and unclean exit codes or clean
306 signals. Restart=on-abnormal is an alternative for
307 Restart=on-failure for services that shall be able to
308 terminate and avoid restarts on certain errors, by
309 indicating so with an unclean exit code. Restart=on-failure
310 or Restart=on-abnormal is now the recommended setting for
311 all long-running services.
313 * If the InaccessibleDirectories= service setting points to a
314 mount point (or if there are any submounts contained within
315 it), it is now attempted to completely unmount it, to make
316 the file systems truly unavailable for the respective
319 * The ReadOnlyDirectories= service setting and
320 systemd-nspawn's --read-only parameter are now recursively
321 applied to all submounts, too.
323 * Mount units may now be created transiently via the bus APIs.
325 * The support for SysV and LSB init scripts has been removed
326 from the systemd daemon itself. Instead, it is now
327 implemented as a generator that creates native systemd units
328 from these scripts when needed. This enables us to remove a
329 substantial amount of legacy code from PID 1, following the
330 fact that many distributions only ship a very small number
331 of LSB/SysV init scripts nowadays.
333 * Priviliged Xen (dom0) domains are not considered
334 virtualization anymore by the virtualization detection
335 logic. After all, they generally have unrestricted access to
336 the hardware and usually are used to manage the unprivileged
339 * systemd-tmpfiles gained a new "C" line type, for copying
340 files or entire directories.
342 * systemd-tmpfiles "m" lines are now fully equivalent to "z"
343 lines. So far, they have been non-globbing versions of the
344 latter, and have thus been redundant. In future, it is
345 recommended to only use "z". "m" has hence been removed
346 from the documentation, even though it stays supported.
348 * A tmpfiles snippet to recreate the most basic structure in
349 /var has been added. This is enough to create the /var/run →
350 /run symlink and create a couple of structural
351 directories. This allows systems to boot up with an empty or
352 volatile /var. Of course, while with this change, the core OS
353 now is capable with dealing with a volatile /var, not all
354 user services are ready for it. However, we hope that sooner
355 or later, many service daemons will be changed upstream so
356 that they are able to automatically create their necessary
357 directories in /var at boot, should they be missing. This is
358 the first step to allow state-less systems that only require
359 the vendor image for /usr to boot.
361 * systemd-nspawn has gained a new --tmpfs= switch to mount an
362 empty tmpfs instance to a specific directory. This is
363 particularly useful for making use of the automatic
364 reconstruction of /var (see above), by passing --tmpfs=/var.
366 * Access modes specified in tmpfiles snippets may now be
367 prefixed with "~", which indicates that they shall be masked
368 by whether the existing file or directly is currently
369 writable, readable or executable at all. Also, if specified,
370 the sgid/suid/sticky bits will be masked for all
373 * A new passive target unit "network-pre.target" has been
374 added which is useful for services that shall run before any
375 network is configured, for example firewall scripts.
377 * The "floppy" group that previously owned the /dev/fd*
378 devices is no longer used. The "disk" group is now used
379 instead. Distributions should probably deprecate usage of
382 Contributions from: Camilo Aguilar, Christian Hesse, Colin Ian
383 King, Cristian Rodríguez, Daniel Buch, Dave Reisner, David
384 Strauss, Denis Tikhomirov, John, Jonathan Liu, Kay Sievers,
385 Lennart Poettering, Mantas Mikulėnas, Mark Eichin, Ronny
386 Chevalier, Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel
387 Andersen, Tom Gundersen, Umut Tezduyar Lindskog, Zbigniew
390 -- Berlin, 2014-06-11
394 * A new "systemd-timesyncd" daemon has been added for
395 synchronizing the system clock across the network. It
396 implements an SNTP client. In contrast to NTP
397 implementations such as chrony or the NTP reference server,
398 this only implements a client side, and does not bother with
399 the full NTP complexity, focusing only on querying time from
400 one remote server and synchronizing the local clock to
401 it. Unless you intend to serve NTP to networked clients or
402 want to connect to local hardware clocks, this simple NTP
403 client should be more than appropriate for most
404 installations. The daemon runs with minimal privileges, and
405 has been hooked up with networkd to only operate when
406 network connectivity is available. The daemon saves the
407 current clock to disk every time a new NTP sync has been
408 acquired, and uses this to possibly correct the system clock
409 early at bootup, in order to accommodate for systems that
410 lack an RTC such as the Raspberry Pi and embedded devices,
411 and to make sure that time monotonically progresses on these
412 systems, even if it is not always correct. To make use of
413 this daemon, a new system user and group "systemd-timesync"
414 needs to be created on installation of systemd.
416 * The queue "seqnum" interface of libudev has been disabled, as
417 it was generally incompatible with device namespacing as
418 sequence numbers of devices go "missing" if the devices are
419 part of a different namespace.
421 * "systemctl list-timers" and "systemctl list-sockets" gained
422 a --recursive switch for showing units of these types also
423 for all local containers, similar in style to the already
424 supported --recursive switch for "systemctl list-units".
426 * A new RebootArgument= setting has been added for service
427 units, which may be used to specify a kernel reboot argument
428 to use when triggering reboots with StartLimitAction=.
430 * A new FailureAction= setting has been added for service
431 units which may be used to specify an operation to trigger
432 when a service fails. This works similarly to
433 StartLimitAction=, but unlike it, controls what is done
434 immediately rather than only after several attempts to
435 restart the service in question.
437 * hostnamed got updated to also expose the kernel name,
438 release, and version on the bus. This is useful for
439 executing commands like hostnamectl with the -H switch.
440 systemd-analyze makes use of this to properly display
441 details when running non-locally.
443 * The bootchart tool can now show cgroup information in the
446 * The CFS CPU quota cgroup attribute is now exposed for
447 services. The new CPUQuota= switch has been added for this
448 which takes a percentage value. Setting this will have the
449 result that a service may never get more CPU time than the
450 specified percentage, even if the machine is otherwise idle.
452 * systemd-networkd learned IPIP and SIT tunnel support.
454 * LSB init scripts exposing a dependency on $network will now
455 get a dependency on network-online.target rather than simply
456 network.target. This should bring LSB handling closer to
457 what it was on SysV systems.
459 * A new fsck.repair= kernel option has been added to control
460 how fsck shall deal with unclean file systems at boot.
462 * The (.ini) configuration file parser will now silently
463 ignore sections whose name begins with "X-". This may be
464 used to maintain application-specific extension sections in unit
467 * machined gained a new API to query the IP addresses of
468 registered containers. "machinectl status" has been updated
469 to show these addresses in its output.
471 * A new call sd_uid_get_display() has been added to the
472 sd-login APIs for querying the "primary" session of a
473 user. The "primary" session of the user is elected from the
474 user's sessions and generally a graphical session is
475 preferred over a text one.
477 * A minimal systemd-resolved daemon has been added. It
478 currently simply acts as a companion to systemd-networkd and
479 manages resolv.conf based on per-interface DNS
480 configuration, possibly supplied via DHCP. In the long run
481 we hope to extend this into a local DNSSEC enabled DNS and
484 * The systemd-networkd-wait-online tool is now enabled by
485 default. It will delay network-online.target until a network
486 connection has been configured. The tool primarily integrates
487 with networkd, but will also make a best effort to make sense
488 of network configuration performed in some other way.
490 * Two new service options StartupCPUShares= and
491 StartupBlockIOWeight= have been added that work similarly to
492 CPUShares= and BlockIOWeight= however only apply during
493 system startup. This is useful to prioritize certain services
494 differently during bootup than during normal runtime.
496 * hostnamed has been changed to prefer the statically
497 configured hostname in /etc/hostname (unless set to
498 'localhost' or empty) over any dynamic one supplied by
499 dhcp. With this change, the rules for picking the hostname
500 match more closely the rules of other configuration settings
501 where the local administrator's configuration in /etc always
502 overrides any other settings.
504 Contributions fron: Ali H. Caliskan, Alison Chaiken, Bas van
505 den Berg, Brandon Philips, Cristian Rodríguez, Daniel Buch,
506 Dan Kilman, Dave Reisner, David Härdeman, David Herrmann,
507 David Strauss, Dimitris Spingos, Djalal Harouni, Eelco
508 Dolstra, Evan Nemerson, Florian Albrechtskirchinger, Greg
509 Kroah-Hartman, Harald Hoyer, Holger Hans Peter Freyther, Jan
510 Engelhardt, Jani Nikula, Jason St. John, Jeffrey Clark,
511 Jonathan Boulle, Kay Sievers, Lennart Poettering, Lukas
512 Nykryn, Lukasz Skalski, Łukasz Stelmach, Mantas Mikulėnas,
513 Marcel Holtmann, Martin Pitt, Matthew Monaco, Michael
514 Marineau, Michael Olbrich, Michal Sekletar, Mike Gilbert, Nis
515 Martensen, Patrik Flykt, Philip Lorenz, poma, Ray Strode,
516 Reyad Attiyat, Robert Milasan, Scott Thrasher, Stef Walter,
517 Steven Siloti, Susant Sahani, Tanu Kaskinen, Thomas Bächler,
518 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar
519 Lindskog, WaLyong Cho, Will Woods, Zbigniew
522 -- Beijing, 2014-05-28
526 * When restoring the screen brightness at boot, stay away from
527 the darkest setting or from the lowest 5% of the available
528 range, depending on which is the larger value of both. This
529 should effectively protect the user from rebooting into a
530 black screen, should the brightness have been set to minimum
533 * sd-login gained a new sd_machine_get_class() call to
534 determine the class ("vm" or "container") of a machine
535 registered with machined.
537 * sd-login gained new calls
538 sd_peer_get_{session,owner_uid,unit,user_unit,slice,machine_name}(),
539 to query the identity of the peer of a local AF_UNIX
540 connection. They operate similarly to their sd_pid_get_xyz()
543 * PID 1 will now maintain a system-wide system state engine
544 with the states "starting", "running", "degraded",
545 "maintenance", "stopping". These states are bound to system
546 startup, normal runtime, runtime with at least one failed
547 service, rescue/emergency mode and system shutdown. This
548 state is shown in the "systemctl status" output when no unit
549 name is passed. It is useful to determine system state, in
550 particularly when doing so for many systems or containers at
553 * A new command "list-machines" has been added to "systemctl"
554 that lists all local OS containers and shows their system
555 state (see above), if systemd runs inside of them.
557 * systemctl gained a new "-r" switch to recursively enumerate
558 units on all local containers, when used with the
559 "list-unit" command (which is the default one that is
560 executed when no parameters are specified).
562 * The GPT automatic partition discovery logic will now honour
563 two GPT partition flags: one may be set on a partition to
564 cause it to be mounted read-only, and the other may be set
565 on a partition to ignore it during automatic discovery.
567 * Two new GPT type UUIDs have been added for automatic root
568 partition discovery, for 32-bit and 64-bit ARM. This is not
569 particularly useful for discovering the root directory on
570 these architectures during bare-metal boots (since UEFI is
571 not common there), but still very useful to allow booting of
572 ARM disk images in nspawn with the -i option.
574 * MAC addresses of interfaces created with nspawn's
575 --network-interface= switch will now be generated from the
576 machine name, and thus be stable between multiple invocations
579 * logind will now automatically remove all IPC objects owned
580 by a user if she or he fully logs out. This makes sure that
581 users who are logged out cannot continue to consume IPC
582 resources. This covers SysV memory, semaphores and message
583 queues as well as POSIX shared memory and message
584 queues. Traditionally, SysV and POSIX IPC had no life-cycle
585 limits. With this functionality, that is corrected. This may
586 be turned off by using the RemoveIPC= switch of logind.conf.
588 * The systemd-machine-id-setup and tmpfiles tools gained a
589 --root= switch to operate on a specific root directory,
592 * journald can now forward logged messages to the TTYs of all
593 logged in users ("wall"). This is the default for all
594 emergency messages now.
596 * A new tool systemd-journal-remote has been added to stream
597 journal log messages across the network.
599 * /sys/fs/cgroup/ is now mounted read-only after all cgroup
600 controller trees are mounted into it. Note that the
601 directories mounted beneath it are not read-only. This is a
602 security measure and is particularly useful because glibc
603 actually includes a search logic to pick any tmpfs it can
604 find to implement shm_open() if /dev/shm is not available
605 (which it might very well be in namespaced setups).
607 * machinectl gained a new "poweroff" command to cleanly power
608 down a local OS container.
610 * The PrivateDevices= unit file setting will now also drop the
611 CAP_MKNOD capability from the capability bound set, and
612 imply DevicePolicy=closed.
614 * PrivateDevices=, PrivateNetwork= and PrivateTmp= is now used
615 comprehensively on all long-running systemd services where
618 * systemd-udevd will now run in a disassociated mount
619 namespace. To mount directories from udev rules, make sure to
620 pull in mount units via SYSTEMD_WANTS properties.
622 * The kdbus support gained support for uploading policy into
623 the kernel. sd-bus gained support for creating "monitoring"
624 connections that can eavesdrop into all bus communication
625 for debugging purposes.
627 * Timestamps may now be specified in seconds since the UNIX
628 epoch Jan 1st, 1970 by specifying "@" followed by the value
631 * Native tcpwrap support in systemd has been removed. tcpwrap
632 is old code, not really maintained anymore and has serious
633 shortcomings, and better options such as firewalls
634 exist. For setups that require tcpwrap usage, please
635 consider invoking your socket-activated service via tcpd,
636 like on traditional inetd.
638 * A new system.conf configuration option
639 DefaultTimerAccuracySec= has been added that controls the
640 default AccuracySec= setting of .timer units.
642 * Timer units gained a new WakeSystem= switch. If enabled,
643 timers configured this way will cause the system to resume
644 from system suspend (if the system supports that, which most
647 * Timer units gained a new Persistent= switch. If enabled,
648 timers configured this way will save to disk when they have
649 been last triggered. This information is then used on next
650 reboot to possible execute overdue timer events, that
651 could not take place because the system was powered off.
652 This enables simple anacron-like behaviour for timer units.
654 * systemctl's "list-timers" will now also list the time a
655 timer unit was last triggered in addition to the next time
656 it will be triggered.
658 * systemd-networkd will now assign predictable IPv4LL
659 addresses to its local interfaces.
661 Contributions from: Brandon Philips, Daniel Buch, Daniel Mack,
662 Dave Reisner, David Herrmann, Gerd Hoffmann, Greg
663 Kroah-Hartman, Hendrik Brueckner, Jason St. John, Josh
664 Triplett, Kay Sievers, Lennart Poettering, Marc-Antoine
665 Perennou, Michael Marineau, Michael Olbrich, Miklos Vajna,
666 Patrik Flykt, poma, Sebastian Thorarensen, Thomas Bächler,
667 Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom Gundersen,
668 Umut Tezduyar Lindskog, Wieland Hoffmann, Zbigniew
671 -- Berlin, 2014-03-25
675 * A new unit file setting RestrictAddressFamilies= has been
676 added to restrict which socket address families unit
677 processes gain access to. This takes address family names
678 like "AF_INET" or "AF_UNIX", and is useful to minimize the
679 attack surface of services via exotic protocol stacks. This
680 is built on seccomp system call filters.
682 * Two new unit file settings RuntimeDirectory= and
683 RuntimeDirectoryMode= have been added that may be used to
684 manage a per-daemon runtime directories below /run. This is
685 an alternative for setting up directory permissions with
686 tmpfiles snippets, and has the advantage that the runtime
687 directory's lifetime is bound to the daemon runtime and that
688 the daemon starts up with an empty directory each time. This
689 is particularly useful when writing services that drop
690 privileges using the User= or Group= setting.
692 * The DeviceAllow= unit setting now supports globbing for
693 matching against device group names.
695 * The systemd configuration file system.conf gained new
696 settings DefaultCPUAccounting=, DefaultBlockIOAccounting=,
697 DefaultMemoryAccounting= to globally turn on/off accounting
698 for specific resources (cgroups) for all units. These
699 settings may still be overridden individually in each unit
702 * systemd-gpt-auto-generator is now able to discover /srv and
703 root partitions in addition to /home and swap partitions. It
704 also supports LUKS-encrypted partitions now. With this in
705 place, automatic discovery of partitions to mount following
706 the Discoverable Partitions Specification
707 (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec)
708 is now a lot more complete. This allows booting without
709 /etc/fstab and without root= on the kernel command line on
710 systems prepared appropriately.
712 * systemd-nspawn gained a new --image= switch which allows
713 booting up disk images and Linux installations on any block
714 device that follow the Discoverable Partitions Specification
715 (see above). This means that installations made with
716 appropriately updated installers may now be started and
717 deployed using container managers, completely
718 unmodified. (We hope that libvirt-lxc will add support for
719 this feature soon, too.)
721 * systemd-nspawn gained a new --network-macvlan= setting to
722 set up a private macvlan interface for the
723 container. Similarly, systemd-networkd gained a new
724 Kind=macvlan setting in .netdev files.
726 * systemd-networkd now supports configuring local addresses
729 * A new tool systemd-network-wait-online has been added to
730 synchronously wait for network connectivity using
733 * The sd-bus.h bus API gained a new sd_bus_track object for
734 tracking the life-cycle of bus peers. Note that sd-bus.h is
735 still not a public API though (unless you specify
736 --enable-kdbus on the configure command line, which however
737 voids your warranty and you get no API stability guarantee).
739 * The $XDG_RUNTIME_DIR runtime directories for each user are
740 now individual tmpfs instances, which has the benefit of
741 introducing separate pools for each user, with individual
742 size limits, and thus making sure that unprivileged clients
743 can no longer negatively impact the system or other users by
744 filling up their $XDG_RUNTIME_DIR. A new logind.conf setting
745 RuntimeDirectorySize= has been introduced that allows
746 controlling the default size limit for all users. It
747 defaults to 10% of the available physical memory. This is no
748 replacement for quotas on tmpfs though (which the kernel
749 still does not support), as /dev/shm and /tmp are still
750 shared resources used by both the system and unprivileged
753 * logind will now automatically turn off automatic suspending
754 on laptop lid close when more than one display is
755 connected. This was previously expected to be implemented
756 individually in desktop environments (such as GNOME),
757 however has been added to logind now, in order to fix a
758 boot-time race where a desktop environment might not have
759 been started yet and thus not been able to take an inhibitor
760 lock at the time where logind already suspends the system
763 * logind will now wait at least 30s after each system
764 suspend/resume cycle, and 3min after system boot before
765 suspending the system due to a closed laptop lid. This
766 should give USB docking stations and similar enough time to
767 be probed and configured after system resume and boot in
768 order to then act as suspend blocker.
770 * systemd-run gained a new --property= setting which allows
771 initialization of resource control properties (and others)
772 for the created scope or service unit. Example: "systemd-run
773 --property=BlockIOWeight=10 updatedb" may be used to run
774 updatedb at a low block IO scheduling weight.
776 * systemd-run's --uid=, --gid=, --setenv=, --setenv= switches
777 now also work in --scope mode.
779 * When systemd is compiled with kdbus support, basic support
780 for enforced policies is now in place. (Note that enabling
781 kdbus still voids your warranty and no API compatibility
784 Contributions from: Andrey Borzenkov, Ansgar Burchardt, Armin
785 K., Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
786 Harald Hoyer, Henrik Grindal Bakken, Jasper St. Pierre, Kay
787 Sievers, Kieran Clancy, Lennart Poettering, Lukas Nykryn,
788 Mantas Mikulėnas, Marcel Holtmann, Mark Oteiza, Martin Pitt,
789 Mike Gilbert, Peter Rajnoha, poma, Samuli Suominen, Stef
790 Walter, Susant Sahani, Tero Roponen, Thomas Andersen, Thomas
791 Bächler, Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom
792 Gundersen, Umut Tezduyar Lindskog, Uoti Urpala, Zachary Cook,
793 Zbigniew Jędrzejewski-Szmek
795 -- Berlin, 2014-03-12
799 * systemd will now relabel /dev after loading the SMACK policy
800 according to SMACK rules.
802 * A new unit file option AppArmorProfile= has been added to
803 set the AppArmor profile for the processes of a unit.
805 * A new condition check ConditionArchitecture= has been added
806 to conditionalize units based on the system architecture, as
807 reported by uname()'s "machine" field.
809 * systemd-networkd now supports matching on the system
810 virtualization, architecture, kernel command line, host name
813 * logind is now a lot more aggressive when suspending the
814 machine due to a closed laptop lid. Instead of acting only
815 on the lid close action, it will continuously watch the lid
816 status and act on it. This is useful for laptops where the
817 power button is on the outside of the chassis so that it can
818 be reached without opening the lid (such as the Lenovo
819 Yoga). On those machines, logind will now immediately
820 re-suspend the machine if the power button has been
821 accidentally pressed while the laptop was suspended and in a
824 * logind will now watch SW_DOCK switches and inhibit reaction
825 to the lid switch if it is pressed. This means that logind
826 will not suspend the machine anymore if the lid is closed
827 and the system is docked, if the laptop supports SW_DOCK
828 notifications via the input layer. Note that ACPI docking
829 stations do not generate this currently. Also note that this
830 logic is usually not fully sufficient and Desktop
831 Environments should take a lid switch inhibitor lock when an
832 external display is connected, as systemd will not watch
835 * nspawn will now make use of the devices cgroup controller by
836 default, and only permit creation of and access to the usual
837 API device nodes like /dev/null or /dev/random, as well as
838 access to (but not creation of) the pty devices.
840 * We will now ship a default .network file for
841 systemd-networkd that automatically configures DHCP for
842 network interfaces created by nspawn's --network-veth or
843 --network-bridge= switches.
845 * systemd will now understand the usual M, K, G, T suffixes
846 according to SI conventions (i.e. to the base 1000) when
847 referring to throughput and hardware metrics. It will stay
848 with IEC conventions (i.e. to the base 1024) for software
849 metrics, according to what is customary according to
850 Wikipedia. We explicitly document which base applies for
851 each configuration option.
853 * The DeviceAllow= setting in unit files now supports a syntax
854 to whitelist an entire group of devices node majors at once,
855 based on the /proc/devices listing. For example, with the
856 string "char-pts", it is now possible to whitelist all
857 current and future pseudo-TTYs at once.
859 * sd-event learned a new "post" event source. Event sources of
860 this type are triggered by the dispatching of any event
861 source of a type that is not "post". This is useful for
862 implementing clean-up and check event sources that are
863 triggered by other work being done in the program.
865 * systemd-networkd is no longer statically enabled, but uses
866 the usual [Install] sections so that it can be
867 enabled/disabled using systemctl. It still is enabled by
870 * When creating a veth interface pair with systemd-nspawn, the
871 host side will now be prefixed with "vb-" if
872 --network-bridge= is used, and with "ve-" if --network-veth
873 is used. This way, it is easy to distinguish these cases on
874 the host, for example to apply different configuration to
875 them with systemd-networkd.
877 * The compatibility libraries for libsystemd-journal.so,
878 libsystem-id128.so, libsystemd-login.so and
879 libsystemd-daemon.so do not make use of IFUNC
880 anymore. Instead, we now build libsystemd.so multiple times
881 under these alternative names. This means that the footprint
882 is drastically increased, but given that these are
883 transitional compatibility libraries, this should not matter
884 much. This change has been made necessary to support the ARM
885 platform for these compatibility libraries, as the ARM
886 toolchain is not really at the same level as the toolchain
887 for other architectures like x86 and does not support
888 IFUNC. Please make sure to use --enable-compat-libs only
889 during a transitional period!
891 Contributions from: Andreas Fuchs, Armin K., Colin Walters,
892 Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
893 Holger Schurig, Jason A. Donenfeld, Jason St. John, Jasper
894 St. Pierre, Kay Sievers, Lennart Poettering, Łukasz Stelmach,
895 Marcel Holtmann, Michael Scherer, Michal Sekletar, Mike
896 Gilbert, Samuli Suominen, Thomas Bächler, Thomas Hindoe
897 Paaboel Andersen, Tom Gundersen, Umut Tezduyar Lindskog,
898 Zbigniew Jędrzejewski-Szmek
900 -- Berlin, 2014-02-24
904 * A new component "systemd-networkd" has been added that can
905 be used to configure local network interfaces statically or
906 via DHCP. It is capable of bringing up bridges, VLANs, and
907 bonding. Currently, no hook-ups for interactive network
908 configuration are provided. Use this for your initrd,
909 container, embedded, or server setup if you need a simple,
910 yet powerful, network configuration solution. This
911 configuration subsystem is quite nifty, as it allows wildcard
912 hotplug matching in interfaces. For example, with a single
913 configuration snippet, you can configure that all Ethernet
914 interfaces showing up are automatically added to a bridge,
915 or similar. It supports link-sensing and more.
917 * A new tool "systemd-socket-proxyd" has been added which can
918 act as a bidirectional proxy for TCP sockets. This is
919 useful for adding socket activation support to services that
920 do not actually support socket activation, including virtual
921 machines and the like.
923 * Add a new tool to save/restore rfkill state on
926 * Save/restore state of keyboard backlights in addition to
927 display backlights on shutdown/boot.
929 * udev learned a new SECLABEL{} construct to label device
930 nodes with a specific security label when they appear. For
931 now, only SECLABEL{selinux} is supported, but the syntax is
932 prepared for additional security frameworks.
934 * udev gained a new scheme to configure link-level attributes
935 from files in /etc/systemd/network/*.link. These files can
936 match against MAC address, device path, driver name and type,
937 and will apply attributes like the naming policy, link speed,
938 MTU, duplex settings, Wake-on-LAN settings, MAC address, MAC
939 address assignment policy (randomized, ...).
941 * The configuration of network interface naming rules for
942 "permanent interface names" has changed: a new NamePolicy=
943 setting in the [Link] section of .link files determines the
944 priority of possible naming schemes (onboard, slot, mac,
945 path). The default value of this setting is determined by
946 /usr/lib/net/links/99-default.link. Old
947 80-net-name-slot.rules udev configuration file has been
948 removed, so local configuration overriding this file should
949 be adapated to override 99-default.link instead.
951 * When the User= switch is used in a unit file, also
952 initialize $SHELL= based on the user database entry.
954 * systemd no longer depends on libdbus. All communication is
955 now done with sd-bus, systemd's low-level bus library
958 * kdbus support has been added to PID 1 itself. When kdbus is
959 enabled, this causes PID 1 to set up the system bus and
960 enable support for a new ".busname" unit type that
961 encapsulates bus name activation on kdbus. It works a little
962 bit like ".socket" units, except for bus names. A new
963 generator has been added that converts classic dbus1 service
964 activation files automatically into native systemd .busname
967 * sd-bus: add a light-weight vtable implementation that allows
968 defining objects on the bus with a simple static const
969 vtable array of its methods, signals and properties.
971 * systemd will not generate or install static dbus
972 introspection data anymore to /usr/share/dbus-1/interfaces,
973 as the precise format of these files is unclear, and
974 nothing makes use of it.
976 * A proxy daemon is now provided to proxy clients connecting
977 via classic D-Bus AF_UNIX sockets to kdbus, to provide full
978 compatibility with classic D-Bus.
980 * A bus driver implementation has been added that supports the
981 classic D-Bus bus driver calls on kdbus, also for
982 compatibility purposes.
984 * A new API "sd-event.h" has been added that implements a
985 minimal event loop API built around epoll. It provides a
986 couple of features that direct epoll usage is lacking:
987 prioritization of events, scales to large numbers of timer
988 events, per-event timer slack (accuracy), system-wide
989 coalescing of timer events, exit handlers, watchdog
990 supervision support using systemd's sd_notify() API, child
993 * A new API "sd-rntl.h" has been added that provides an API
994 around the route netlink interface of the kernel, similar in
997 * A new API "sd-dhcp-client.h" has been added that provides a
998 small DHCPv4 client-side implementation. This is used by
1001 * There is a new kernel command line option
1002 "systemd.restore_state=0|1". When set to "0", none of the
1003 systemd tools will restore saved runtime state to hardware
1004 devices. More specifically, the rfkill and backlight states
1007 * The FsckPassNo= compatibility option in mount/service units
1008 has been removed. The fstab generator will now add the
1009 necessary dependencies automatically, and does not require
1010 PID1's support for that anymore.
1012 * journalctl gained a new switch, --list-boots, that lists
1013 recent boots with their times and boot IDs.
1015 * The various tools like systemctl, loginctl, timedatectl,
1016 busctl, systemd-run, ... have gained a new switch "-M" to
1017 connect to a specific, local OS container (as direct
1018 connection, without requiring SSH). This works on any
1019 container that is registered with machined, such as those
1020 created by libvirt-lxc or nspawn.
1022 * systemd-run and systemd-analyze also gained support for "-H"
1023 to connect to remote hosts via SSH. This is particularly
1024 useful for systemd-run because it enables queuing of jobs
1025 onto remote systems.
1027 * machinectl gained a new command "login" to open a getty
1028 login in any local container. This works with any container
1029 that is registered with machined (such as those created by
1030 libvirt-lxc or nspawn), and which runs systemd inside.
1032 * machinectl gained a new "reboot" command that may be used to
1033 trigger a reboot on a specific container that is registered
1034 with machined. This works on any container that runs an init
1035 system of some kind.
1037 * systemctl gained a new "list-timers" command to print a nice
1038 listing of installed timer units with the times they elapse
1041 * Alternative reboot() parameters may now be specified on the
1042 "systemctl reboot" command line and are passed to the
1043 reboot() system call.
1045 * systemctl gained a new --job-mode= switch to configure the
1046 mode to queue a job with. This is a more generic version of
1047 --fail, --irreversible, and --ignore-dependencies, which are
1048 still available but not advertised anymore.
1050 * /etc/systemd/system.conf gained new settings to configure
1051 various default timeouts of units, as well as the default
1052 start limit interval and burst. These may still be overridden
1055 * PID1 will now export on the bus profile data of the security
1056 policy upload process (such as the SELinux policy upload to
1059 * journald: when forwarding logs to the console, include
1060 timestamps (following the setting in
1061 /sys/module/printk/parameters/time).
1063 * OnCalendar= in timer units now understands the special
1064 strings "yearly" and "annually". (Both are equivalent)
1066 * The accuracy of timer units is now configurable with the new
1067 AccuracySec= setting. It defaults to 1min.
1069 * A new dependency type JoinsNamespaceOf= has been added that
1070 allows running two services within the same /tmp and network
1071 namespace, if PrivateNetwork= or PrivateTmp= are used.
1073 * A new command "cat" has been added to systemctl. It outputs
1074 the original unit file of a unit, and concatenates the
1075 contents of additional "drop-in" unit file snippets, so that
1076 the full configuration is shown.
1078 * systemctl now supports globbing on the various "list-xyz"
1079 commands, like "list-units" or "list-sockets", as well as on
1080 those commands which take multiple unit names.
1082 * journalctl's --unit= switch gained support for globbing.
1084 * All systemd daemons now make use of the watchdog logic so
1085 that systemd automatically notices when they hang.
1087 * If the $container_ttys environment variable is set,
1088 getty-generator will automatically spawn a getty for each
1089 listed tty. This is useful for container managers to request
1090 login gettys to be spawned on as many ttys as needed.
1092 * %h, %s, %U specifier support is not available anymore when
1093 used in unit files for PID 1. This is because NSS calls are
1094 not safe from PID 1. They stay available for --user
1095 instances of systemd, and as special case for the root user.
1097 * loginctl gained a new "--no-legend" switch to turn off output
1100 * The "sd-login.h" API gained three new calls:
1101 sd_session_is_remote(), sd_session_get_remote_user(),
1102 sd_session_get_remote_host() to query information about
1105 * The udev hardware database now also carries vendor/product
1106 information of SDIO devices.
1108 * The "sd-daemon.h" API gained a new sd_watchdog_enabled() to
1109 determine whether watchdog notifications are requested by
1112 * Socket-activated per-connection services now include a
1113 short description of the connection parameters in the
1116 * tmpfiles gained a new "--boot" option. When this is not used,
1117 only lines where the command character is not suffixed with
1118 "!" are executed. When this option is specified, those
1119 options are executed too. This partitions tmpfiles
1120 directives into those that can be safely executed at any
1121 time, and those which should be run only at boot (for
1122 example, a line that creates /run/nologin).
1124 * A new API "sd-resolve.h" has been added which provides a simple
1125 asynchronous wrapper around glibc NSS host name resolution
1126 calls, such as getaddrinfo(). In contrast to glibc's
1127 getaddrinfo_a(), it does not use signals. In contrast to most
1128 other asynchronous name resolution libraries, this one does
1129 not reimplement DNS, but reuses NSS, so that alternate
1130 host name resolution systems continue to work, such as mDNS,
1131 LDAP, etc. This API is based on libasyncns, but it has been
1132 cleaned up for inclusion in systemd.
1134 * The APIs "sd-journal.h", "sd-login.h", "sd-id128.h",
1135 "sd-daemon.h" are no longer found in individual libraries
1136 libsystemd-journal.so, libsystemd-login.so,
1137 libsystemd-id128.so, libsystemd-daemon.so. Instead, we have
1138 merged them into a single library, libsystemd.so, which
1139 provides all symbols. The reason for this is cyclic
1140 dependencies, as these libraries tend to use each other's
1141 symbols. So far, we have managed to workaround that by linking
1142 a copy of a good part of our code into each of these
1143 libraries again and again, which, however, makes certain
1144 things hard to do, like sharing static variables. Also, it
1145 substantially increases footprint. With this change, there
1146 is only one library for the basic APIs systemd
1147 provides. Also, "sd-bus.h", "sd-memfd.h", "sd-event.h",
1148 "sd-rtnl.h", "sd-resolve.h", "sd-utf8.h" are found in this
1149 library as well, however are subject to the --enable-kdbus
1150 switch (see below). Note that "sd-dhcp-client.h" is not part
1151 of this library (this is because it only consumes, never
1152 provides, services of/to other APIs). To make the transition
1153 easy from the separate libraries to the unified one, we
1154 provide the --enable-compat-libs compile-time switch which
1155 will generate stub libraries that are compatible with the
1156 old ones but redirect all calls to the new one.
1158 * All of the kdbus logic and the new APIs "sd-bus.h",
1159 "sd-memfd.h", "sd-event.h", "sd-rtnl.h", "sd-resolve.h",
1160 and "sd-utf8.h" are compile-time optional via the
1161 "--enable-kdbus" switch, and they are not compiled in by
1162 default. To make use of kdbus, you have to explicitly enable
1163 the switch. Note however, that neither the kernel nor the
1164 userspace API for all of this is considered stable yet. We
1165 want to maintain the freedom to still change the APIs for
1166 now. By specifying this build-time switch, you acknowledge
1167 that you are aware of the instability of the current
1170 * Also, note that while kdbus is pretty much complete,
1171 it lacks one thing: proper policy support. This means you
1172 can build a fully working system with all features; however,
1173 it will be highly insecure. Policy support will be added in
1174 one of the next releases, at the same time that we will
1175 declare the APIs stable.
1177 * When the kernel command-line argument "kdbus" is specified,
1178 systemd will automatically load the kdbus.ko kernel module. At
1179 this stage of development, it is only useful for testing kdbus
1180 and should not be used in production. Note: if "--enable-kdbus"
1181 is specified, and the kdbus.ko kernel module is available, and
1182 "kdbus" is added to the kernel command line, the entire system
1183 runs with kdbus instead of dbus-daemon, with the above mentioned
1184 problem of missing the system policy enforcement. Also a future
1185 version of kdbus.ko or a newer systemd will not be compatible with
1186 each other, and will unlikely be able to boot the machine if only
1187 one of them is updated.
1189 * systemctl gained a new "import-environment" command which
1190 uploads the caller's environment (or parts thereof) into the
1191 service manager so that it is inherited by services started
1192 by the manager. This is useful to upload variables like
1193 $DISPLAY into the user service manager.
1195 * A new PrivateDevices= switch has been added to service units
1196 which allows running a service with a namespaced /dev
1197 directory that does not contain any device nodes for
1198 physical devices. More specifically, it only includes devices
1199 such as /dev/null, /dev/urandom, and /dev/zero which are API
1202 * logind has been extended to support behaviour like VT
1203 switching on seats that do not support a VT. This makes
1204 multi-session available on seats that are not the first seat
1205 (seat0), and on systems where kernel support for VTs has
1206 been disabled at compile-time.
1208 * If a process holds a delay lock for system sleep or shutdown
1209 and fails to release it in time, we will now log its
1210 identity. This makes it easier to identify processes that
1211 cause slow suspends or power-offs.
1213 * When parsing /etc/crypttab, support for a new key-slot=
1214 option as supported by Debian is added. It allows indicating
1215 which LUKS slot to use on disk, speeding up key loading.
1217 * The sd_journald_sendv() API call has been checked and
1218 officially declared to be async-signal-safe so that it may
1219 be invoked from signal handlers for logging purposes.
1221 * Boot-time status output is now enabled automatically after a
1222 short timeout if boot does not progress, in order to give
1223 the user an indication what she or he is waiting for.
1225 * The boot-time output has been improved to show how much time
1226 remains until jobs expire.
1228 * The KillMode= switch in service units gained a new possible
1229 value "mixed". If set, and the unit is shut down, then the
1230 initial SIGTERM signal is sent only to the main daemon
1231 process, while the following SIGKILL signal is sent to
1232 all remaining processes of the service.
1234 * When a scope unit is registered, a new property "Controller"
1235 may be set. If set to a valid bus name, systemd will send a
1236 RequestStop() signal to this name when it would like to shut
1237 down the scope. This may be used to hook manager logic into
1238 the shutdown logic of scope units. Also, scope units may now
1239 be put in a special "abandoned" state, in which case the
1240 manager process which created them takes no further
1241 responsibilities for it.
1243 * When reading unit files, systemd will now verify
1244 the access mode of these files, and warn about certain
1245 suspicious combinations. This has been added to make it
1246 easier to track down packaging bugs where unit files are
1247 marked executable or world-writable.
1249 * systemd-nspawn gained a new "--setenv=" switch to set
1250 container-wide environment variables. The similar option in
1251 systemd-activate was renamed from "--environment=" to
1252 "--setenv=" for consistency.
1254 * systemd-nspawn has been updated to create a new kdbus domain
1255 for each container that is invoked, thus allowing each
1256 container to have its own set of system and user buses,
1257 independent of the host.
1259 * systemd-nspawn gained a new --drop-capability= switch to run
1260 the container with less capabilities than the default. Both
1261 --drop-capability= and --capability= now take the special
1262 string "all" for dropping or keeping all capabilities.
1264 * systemd-nspawn gained new switches for executing containers
1265 with specific SELinux labels set.
1267 * systemd-nspawn gained a new --quiet switch to not generate
1268 any additional output but the container's own console
1271 * systemd-nspawn gained a new --share-system switch to run a
1272 container without PID namespacing enabled.
1274 * systemd-nspawn gained a new --register= switch to control
1275 whether the container is registered with systemd-machined or
1276 not. This is useful for containers that do not run full
1277 OS images, but only specific apps.
1279 * systemd-nspawn gained a new --keep-unit which may be used
1280 when invoked as the only program from a service unit, and
1281 results in registration of the unit service itself in
1282 systemd-machined, instead of a newly opened scope unit.
1284 * systemd-nspawn gained a new --network-interface= switch for
1285 moving arbitrary interfaces to the container. The new
1286 --network-veth switch creates a virtual Ethernet connection
1287 between host and container. The new --network-bridge=
1288 switch then allows assigning the host side of this virtual
1289 Ethernet connection to a bridge device.
1291 * systemd-nspawn gained a new --personality= switch for
1292 setting the kernel personality for the container. This is
1293 useful when running a 32-bit container on a 64-bit host. A
1294 similar option Personality= is now also available for service
1297 * logind will now also track a "Desktop" identifier for each
1298 session which encodes the desktop environment of it. This is
1299 useful for desktop environments that want to identify
1300 multiple running sessions of itself easily.
1302 * A new SELinuxContext= setting for service units has been
1303 added that allows setting a specific SELinux execution
1304 context for a service.
1306 * Most systemd client tools will now honour $SYSTEMD_LESS for
1307 settings of the "less" pager. By default, these tools will
1308 override $LESS to allow certain operations to work, such as
1309 jump-to-the-end. With $SYSTEMD_LESS, it is possible to
1310 influence this logic.
1312 * systemd's "seccomp" hook-up has been changed to make use of
1313 the libseccomp library instead of using its own
1314 implementation. This has benefits for portability among
1317 * For usage together with SystemCallFilter=, a new
1318 SystemCallErrorNumber= setting has been introduced that
1319 allows configuration of a system error number to be returned
1320 on filtered system calls, instead of immediately killing the
1321 process. Also, SystemCallArchitectures= has been added to
1322 limit access to system calls of a particular architecture
1323 (in order to turn off support for unused secondary
1324 architectures). There is also a global
1325 SystemCallArchitectures= setting in system.conf now to turn
1326 off support for non-native system calls system-wide.
1328 * systemd requires a kernel with a working name_to_handle_at(),
1329 please see the kernel config requirements in the README file.
1331 Contributions from: Adam Williamson, Alex Jia, Anatol Pomozov,
1332 Ansgar Burchardt, AppleBloom, Auke Kok, Bastien Nocera,
1333 Chengwei Yang, Christian Seiler, Colin Guthrie, Colin Walters,
1334 Cristian Rodríguez, Daniel Buch, Daniele Medri, Daniel J
1335 Walsh, Daniel Mack, Dan McGee, Dave Reisner, David Coppa,
1336 David Herrmann, David Strauss, Djalal Harouni, Dmitry Pisklov,
1337 Elia Pinto, Florian Weimer, George McCollister, Goffredo
1338 Baroncelli, Greg Kroah-Hartman, Hendrik Brueckner, Igor
1339 Zhbanov, Jan Engelhardt, Jan Janssen, Jason A. Donenfeld,
1340 Jason St. John, Jasper St. Pierre, Jóhann B. Guðmundsson, Jose
1341 Ignacio Naranjo, Karel Zak, Kay Sievers, Kristian Høgsberg,
1342 Lennart Poettering, Lubomir Rintel, Lukas Nykryn, Lukasz
1343 Skalski, Łukasz Stelmach, Luke Shumaker, Mantas Mikulėnas,
1344 Marc-Antoine Perennou, Marcel Holtmann, Marcos Felipe Rasia de
1345 Mello, Marko Myllynen, Martin Pitt, Matthew Monaco, Michael
1346 Marineau, Michael Scherer, Michał Górny, Michal Sekletar,
1347 Michele Curti, Oleksii Shevchuk, Olivier Brunel, Patrik Flykt,
1348 Pavel Holica, Raudi, Richard Marko, Ronny Chevalier, Sébastien
1349 Luttringer, Sergey Ptashnick, Shawn Landden, Simon Peeters,
1350 Stefan Beller, Susant Sahani, Sylvain Plantefeve, Sylvia Else,
1351 Tero Roponen, Thomas Bächler, Thomas Hindoe Paaboel Andersen,
1352 Tom Gundersen, Umut Tezduyar Lindskog, Unai Uribarri, Václav
1353 Pavlín, Vincent Batts, WaLyong Cho, William Giokas, Yang
1354 Zhiyong, Yin Kangkai, Yuxuan Shui, Zbigniew Jędrzejewski-Szmek
1356 -- Berlin, 2014-02-20
1360 * logind has gained support for facilitating privileged input
1361 and drm device access for unprivileged clients. This work is
1362 useful to allow Wayland display servers (and similar
1363 programs, such as kmscon) to run under the user's ID and
1364 access input and drm devices which are normally
1365 protected. When this is used (and the kernel is new enough)
1366 logind will "mute" IO on the file descriptors passed to
1367 Wayland as long as it is in the background and "unmute" it
1368 if it returns into the foreground. This allows secure
1369 session switching without allowing background sessions to
1370 eavesdrop on input and display data. This also introduces
1371 session switching support if VT support is turned off in the
1372 kernel, and on seats that are not seat0.
1374 * A new kernel command line option luks.options= is understood
1375 now which allows specifiying LUKS options for usage for LUKS
1376 encrypted partitions specified with luks.uuid=.
1378 * tmpfiles.d(5) snippets may now use specifier expansion in
1379 path names. More specifically %m, %b, %H, %v, are now
1380 replaced by the local machine id, boot id, hostname, and
1381 kernel version number.
1383 * A new tmpfiles.d(5) command "m" has been introduced which
1384 may be used to change the owner/group/access mode of a file
1385 or directory if it exists, but do nothing if it does not.
1387 * This release removes high-level support for the
1388 MemorySoftLimit= cgroup setting. The underlying kernel
1389 cgroup attribute memory.soft_limit= is currently badly
1390 designed and likely to be removed from the kernel API in its
1391 current form, hence we should not expose it for now.
1393 * The memory.use_hierarchy cgroup attribute is now enabled for
1394 all cgroups systemd creates in the memory cgroup
1395 hierarchy. This option is likely to be come the built-in
1396 default in the kernel anyway, and the non-hierarchial mode
1397 never made much sense in the intrinsically hierarchial
1400 * A new field _SYSTEMD_SLICE= is logged along with all journal
1401 messages containing the slice a message was generated
1402 from. This is useful to allow easy per-customer filtering of
1403 logs among other things.
1405 * systemd-journald will no longer adjust the group of journal
1406 files it creates to the "systemd-journal" group. Instead we
1407 rely on the journal directory to be owned by the
1408 "systemd-journal" group, and its setgid bit set, so that the
1409 kernel file system layer will automatically enforce that
1410 journal files inherit this group assignment. The reason for
1411 this change is that we cannot allow NSS look-ups from
1412 journald which would be necessary to resolve
1413 "systemd-journal" to a numeric GID, because this might
1414 create deadlocks if NSS involves synchronous queries to
1415 other daemons (such as nscd, or sssd) which in turn are
1416 logging clients of journald and might block on it, which
1417 would then dead lock. A tmpfiles.d(5) snippet included in
1418 systemd will make sure the setgid bit and group are
1419 properly set on the journal directory if it exists on every
1420 boot. However, we recommend adjusting it manually after
1421 upgrades too (or from RPM scriptlets), so that the change is
1422 not delayed until next reboot.
1424 * Backlight and random seed files in /var/lib/ have moved into
1425 the /var/lib/systemd/ directory, in order to centralize all
1426 systemd generated files in one directory.
1428 * Boot time performance measurements (as displayed by
1429 "systemd-analyze" for example) will now read ACPI 5.0 FPDT
1430 performance information if that's available to determine how
1431 much time BIOS and boot loader initialization required. With
1432 a sufficiently new BIOS you hence no longer need to boot
1433 with Gummiboot to get access to such information.
1435 Contributions from: Andrey Borzenkov, Chen Jie, Colin Walters,
1436 Cristian Rodríguez, Dave Reisner, David Herrmann, David
1437 Mackey, David Strauss, Eelco Dolstra, Evan Callicoat, Gao
1438 feng, Harald Hoyer, Jimmie Tauriainen, Kay Sievers, Lennart
1439 Poettering, Lukas Nykryn, Mantas Mikulėnas, Martin Pitt,
1440 Michael Scherer, Michał Górny, Mike Gilbert, Patrick McCarty,
1441 Sebastian Ott, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
1443 -- Berlin, 2013-10-02
1447 * The Restart= option for services now understands a new
1448 on-watchdog setting, which will restart the service
1449 automatically if the service stops sending out watchdog keep
1450 alive messages (as configured with WatchdogSec=).
1452 * The getty generator (which is responsible for bringing up a
1453 getty on configured serial consoles) will no longer only
1454 start a getty on the primary kernel console but on all
1455 others, too. This makes the order in which console= is
1456 specified on the kernel command line less important.
1458 * libsystemd-logind gained a new sd_session_get_vt() call to
1459 retrieve the VT number of a session.
1461 * If the option "tries=0" is set for an entry of /etc/crypttab
1462 its passphrase is queried indefinitely instead of any
1463 maximum number of tries.
1465 * If a service with a configure PID file terminates its PID
1466 file will now be removed automatically if it still exists
1467 afterwards. This should put an end to stale PID files.
1469 * systemd-run will now also take relative binary path names
1470 for execution and no longer insists on absolute paths.
1472 * InaccessibleDirectories= and ReadOnlyDirectories= now take
1473 paths that are optionally prefixed with "-" to indicate that
1474 it should not be considered a failure if they do not exist.
1476 * journalctl -o (and similar commands) now understands a new
1477 output mode "short-precise", it is similar to "short" but
1478 shows timestamps with usec accuracy.
1480 * The option "discard" (as known from Debian) is now
1481 synonymous to "allow-discards" in /etc/crypttab. In fact,
1482 "discard" is preferred now (since it is easier to remember
1485 * Some licensing clean-ups were made, so that more code is now
1486 LGPL-2.1 licensed than before.
1488 * A minimal tool to save/restore the display backlight
1489 brightness across reboots has been added. It will store the
1490 backlight setting as late as possible at shutdown, and
1491 restore it as early as possible during reboot.
1493 * A logic to automatically discover and enable home and swap
1494 partitions on GPT disks has been added. With this in place
1495 /etc/fstab becomes optional for many setups as systemd can
1496 discover certain partitions located on the root disk
1497 automatically. Home partitions are recognized under their
1498 GPT type ID 933ac7e12eb44f13b8440e14e2aef915. Swap
1499 partitions are recognized under their GPT type ID
1500 0657fd6da4ab43c484e50933c84b4f4f.
1502 * systemd will no longer pass any environment from the kernel
1503 or initrd to system services. If you want to set an
1504 environment for all services, do so via the kernel command
1505 line systemd.setenv= assignment.
1507 * The systemd-sysctl tool no longer natively reads the file
1508 /etc/sysctl.conf. If desired, the file should be symlinked
1509 from /etc/sysctl.d/99-sysctl.conf. Apart from providing
1510 legacy support by a symlink rather than built-in code, it
1511 also makes the otherwise hidden order of application of the
1512 different files visible. (Note that this partly reverts to a
1513 pre-198 application order of sysctl knobs!)
1515 * The "systemctl set-log-level" and "systemctl dump" commands
1516 have been moved to systemd-analyze.
1518 * systemd-run learned the new --remain-after-exit switch,
1519 which causes the scope unit not to be cleaned up
1520 automatically after the process terminated.
1522 * tmpfiles learned a new --exclude-prefix= switch to exclude
1523 certain paths from operation.
1525 * journald will now automatically flush all messages to disk
1526 as soon as a message of the log priorities CRIT, ALERT or
1529 Contributions from: Andrew Cook, Brandon Philips, Christian
1530 Hesse, Christoph Junghans, Colin Walters, Daniel Schaal,
1531 Daniel Wallace, Dave Reisner, David Herrmann, Gao feng, George
1532 McCollister, Giovanni Campagna, Hannes Reinecke, Harald Hoyer,
1533 Herczeg Zsolt, Holger Hans Peter Freyther, Jan Engelhardt,
1534 Jesper Larsen, Kay Sievers, Khem Raj, Lennart Poettering,
1535 Lukas Nykryn, Maciej Wereski, Mantas Mikulėnas, Marcel
1536 Holtmann, Martin Pitt, Michael Biebl, Michael Marineau,
1537 Michael Scherer, Michael Stapelberg, Michal Sekletar, Michał
1538 Górny, Olivier Brunel, Ondrej Balaz, Ronny Chevalier, Shawn
1539 Landden, Steven Hiscocks, Thomas Bächler, Thomas Hindoe
1540 Paaboel Andersen, Tom Gundersen, Umut Tezduyar, WANG Chao,
1541 William Giokas, Zbigniew Jędrzejewski-Szmek
1543 -- Berlin, 2013-09-13
1547 * The documentation has been updated to cover the various new
1548 concepts introduced with 205.
1550 * Unit files now understand the new %v specifier which
1551 resolves to the kernel version string as returned by "uname
1554 * systemctl now supports filtering the unit list output by
1555 load state, active state and sub state, using the new
1558 * "systemctl status" will now show the results of the
1559 condition checks (like ConditionPathExists= and similar) of
1560 the last start attempts of the unit. They are also logged to
1563 * "journalctl -b" may now be used to look for boot output of a
1564 specific boot. Try "journalctl -b -1" for the previous boot,
1565 but the syntax is substantially more powerful.
1567 * "journalctl --show-cursor" has been added which prints the
1568 cursor string the last shown log line. This may then be used
1569 with the new "journalctl --after-cursor=" switch to continue
1570 browsing logs from that point on.
1572 * "journalctl --force" may now be used to force regeneration
1575 * Creation of "dead" device nodes has been moved from udev
1576 into kmod and tmpfiles. Previously, udev would read the kmod
1577 databases to pre-generate dead device nodes based on meta
1578 information contained in kernel modules, so that these would
1579 be auto-loaded on access rather then at boot. As this
1580 does not really have much to do with the exposing actual
1581 kernel devices to userspace this has always been slightly
1582 alien in the udev codebase. Following the new scheme kmod
1583 will now generate a runtime snippet for tmpfiles from the
1584 module meta information and it now is tmpfiles' job to the
1585 create the nodes. This also allows overriding access and
1586 other parameters for the nodes using the usual tmpfiles
1587 facilities. As side effect this allows us to remove the
1588 CAP_SYS_MKNOD capability bit from udevd entirely.
1590 * logind's device ACLs may now be applied to these "dead"
1591 devices nodes too, thus finally allowing managed access to
1592 devices such as /dev/snd/sequencer whithout loading the
1593 backing module right-away.
1595 * A new RPM macro has been added that may be used to apply
1596 tmpfiles configuration during package installation.
1598 * systemd-detect-virt and ConditionVirtualization= now can
1599 detect User-Mode-Linux machines (UML).
1601 * journald will now implicitly log the effective capabilities
1602 set of processes in the message metadata.
1604 * systemd-cryptsetup has gained support for TrueCrypt volumes.
1606 * The initrd interface has been simplified (more specifically,
1607 support for passing performance data via environment
1608 variables and fsck results via files in /run has been
1609 removed). These features were non-essential, and are
1610 nowadays available in a much nicer way by having systemd in
1611 the initrd serialize its state and have the hosts systemd
1612 deserialize it again.
1614 * The udev "keymap" data files and tools to apply keyboard
1615 specific mappings of scan to key codes, and force-release
1616 scan code lists have been entirely replaced by a udev
1617 "keyboard" builtin and a hwdb data file.
1619 * systemd will now honour the kernel's "quiet" command line
1620 argument also during late shutdown, resulting in a
1621 completely silent shutdown when used.
1623 * There's now an option to control the SO_REUSEPORT socket
1624 option in .socket units.
1626 * Instance units will now automatically get a per-template
1627 subslice of system.slice unless something else is explicitly
1628 configured. For example, instances of sshd@.service will now
1629 implicitly be placed in system-sshd.slice rather than
1630 system.slice as before.
1632 * Test coverage support may now be enabled at build time.
1634 Contributions from: Dave Reisner, Frederic Crozat, Harald
1635 Hoyer, Holger Hans Peter Freyther, Jan Engelhardt, Jan
1636 Janssen, Jason St. John, Jesper Larsen, Kay Sievers, Lennart
1637 Poettering, Lukas Nykryn, Maciej Wereski, Martin Pitt, Michael
1638 Olbrich, Ramkumar Ramachandra, Ross Lagerwall, Shawn Landden,
1639 Thomas H.P. Andersen, Tom Gundersen, Tomasz Torcz, William
1640 Giokas, Zbigniew Jędrzejewski-Szmek
1642 -- Berlin, 2013-07-23
1646 * Two new unit types have been introduced:
1648 Scope units are very similar to service units, however, are
1649 created out of pre-existing processes -- instead of PID 1
1650 forking off the processes. By using scope units it is
1651 possible for system services and applications to group their
1652 own child processes (worker processes) in a powerful way
1653 which then maybe used to organize them, or kill them
1654 together, or apply resource limits on them.
1656 Slice units may be used to partition system resources in an
1657 hierarchial fashion and then assign other units to them. By
1658 default there are now three slices: system.slice (for all
1659 system services), user.slice (for all user sessions),
1660 machine.slice (for VMs and containers).
1662 Slices and scopes have been introduced primarily in
1663 context of the work to move cgroup handling to a
1664 single-writer scheme, where only PID 1
1665 creates/removes/manages cgroups.
1667 * There's a new concept of "transient" units. In contrast to
1668 normal units these units are created via an API at runtime,
1669 not from configuration from disk. More specifically this
1670 means it is now possible to run arbitrary programs as
1671 independent services, with all execution parameters passed
1672 in via bus APIs rather than read from disk. Transient units
1673 make systemd substantially more dynamic then it ever was,
1674 and useful as a general batch manager.
1676 * logind has been updated to make use of scope and slice units
1677 for managing user sessions. As a user logs in he will get
1678 his own private slice unit, to which all sessions are added
1679 as scope units. We also added support for automatically
1680 adding an instance of user@.service for the user into the
1681 slice. Effectively logind will no longer create cgroup
1682 hierarchies on its own now, it will defer entirely to PID 1
1683 for this by means of scope, service and slice units. Since
1684 user sessions this way become entities managed by PID 1
1685 the output of "systemctl" is now a lot more comprehensive.
1687 * A new mini-daemon "systemd-machined" has been added which
1688 may be used by virtualization managers to register local
1689 VMs/containers. nspawn has been updated accordingly, and
1690 libvirt will be updated shortly. machined will collect a bit
1691 of meta information about the VMs/containers, and assign
1692 them their own scope unit (see above). The collected
1693 meta-data is then made available via the "machinectl" tool,
1694 and exposed in "ps" and similar tools. machined/machinectl
1695 is compile-time optional.
1697 * As discussed earlier, the low-level cgroup configuration
1698 options ControlGroup=, ControlGroupModify=,
1699 ControlGroupPersistent=, ControlGroupAttribute= have been
1700 removed. Please use high-level attribute settings instead as
1701 well as slice units.
1703 * A new bus call SetUnitProperties() has been added to alter
1704 various runtime parameters of a unit. This is primarily
1705 useful to alter cgroup parameters dynamically in a nice way,
1706 but will be extended later on to make more properties
1707 modifiable at runtime. systemctl gained a new set-properties
1708 command that wraps this call.
1710 * A new tool "systemd-run" has been added which can be used to
1711 run arbitrary command lines as transient services or scopes,
1712 while configuring a number of settings via the command
1713 line. This tool is currently very basic, however already
1714 very useful. We plan to extend this tool to even allow
1715 queuing of execution jobs with time triggers from the
1716 command line, similar in fashion to "at".
1718 * nspawn will now inform the user explicitly that kernels with
1719 audit enabled break containers, and suggest the user to turn
1722 * Support for detecting the IMA and AppArmor security
1723 frameworks with ConditionSecurity= has been added.
1725 * journalctl gained a new "-k" switch for showing only kernel
1726 messages, mimicking dmesg output; in addition to "--user"
1727 and "--system" switches for showing only user's own logs
1730 * systemd-delta can now show information about drop-in
1731 snippets extending unit files.
1733 * libsystemd-bus has been substantially updated but is still
1734 not available as public API.
1736 * systemd will now look for the "debug" argument on the kernel
1737 command line and enable debug logging, similar to what
1738 "systemd.log_level=debug" already did before.
1740 * "systemctl set-default", "systemctl get-default" has been
1741 added to configure the default.target symlink, which
1742 controls what to boot into by default.
1744 * "systemctl set-log-level" has been added as a convenient
1745 way to raise and lower systemd logging threshold.
1747 * "systemd-analyze plot" will now show the time the various
1748 generators needed for execution, as well as information
1749 about the unit file loading.
1751 * libsystemd-journal gained a new sd_journal_open_files() call
1752 for opening specific journal files. journactl also gained a
1753 new switch to expose this new functionality. Previously we
1754 only supported opening all files from a directory, or all
1755 files from the system, as opening individual files only is
1756 racy due to journal file rotation.
1758 * systemd gained the new DefaultEnvironment= setting in
1759 /etc/systemd/system.conf to set environment variables for
1762 * If a privileged process logs a journal message with the
1763 OBJECT_PID= field set, then journald will automatically
1764 augment this with additional OBJECT_UID=, OBJECT_GID=,
1765 OBJECT_COMM=, OBJECT_EXE=, ... fields. This is useful if
1766 system services want to log events about specific client
1767 processes. journactl/systemctl has been updated to make use
1768 of this information if all log messages regarding a specific
1771 Contributions from: Auke Kok, Chengwei Yang, Colin Walters,
1772 Cristian Rodríguez, Daniel Albers, Daniel Wallace, Dave
1773 Reisner, David Coppa, David King, David Strauss, Eelco
1774 Dolstra, Gabriel de Perthuis, Harald Hoyer, Jan Alexander
1775 Steffens, Jan Engelhardt, Jan Janssen, Jason St. John, Johan
1776 Heikkilä, Karel Zak, Karol Lewandowski, Kay Sievers, Lennart
1777 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marius Vollmer,
1778 Martin Pitt, Michael Biebl, Michael Olbrich, Michael Tremer,
1779 Michal Schmidt, Michał Bartoszkiewicz, Nirbheek Chauhan,
1780 Pierre Neidhardt, Ross Burton, Ross Lagerwall, Sean McGovern,
1781 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar,
1782 Václav Pavlín, Zachary Cook, Zbigniew Jędrzejewski-Szmek,
1783 Łukasz Stelmach, 장동준
1787 * The Python bindings gained some minimal support for the APIs
1788 exposed by libsystemd-logind.
1790 * ConditionSecurity= gained support for detecting SMACK. Since
1791 this condition already supports SELinux and AppArmor we only
1792 miss IMA for this. Patches welcome!
1794 Contributions from: Karol Lewandowski, Lennart Poettering,
1795 Zbigniew Jędrzejewski-Szmek
1799 * systemd-nspawn will now create /etc/resolv.conf if
1800 necessary, before bind-mounting the host's file onto it.
1802 * systemd-nspawn will now store meta information about a
1803 container on the container's cgroup as extended attribute
1804 fields, including the root directory.
1806 * The cgroup hierarchy has been reworked in many ways. All
1807 objects any of the components systemd creates in the cgroup
1808 tree are now suffixed. More specifically, user sessions are
1809 now placed in cgroups suffixed with ".session", users in
1810 cgroups suffixed with ".user", and nspawn containers in
1811 cgroups suffixed with ".nspawn". Furthermore, all cgroup
1812 names are now escaped in a simple scheme to avoid collision
1813 of userspace object names with kernel filenames. This work
1814 is preparation for making these objects relocatable in the
1815 cgroup tree, in order to allow easy resource partitioning of
1816 these objects without causing naming conflicts.
1818 * systemctl list-dependencies gained the new switches
1819 --plain, --reverse, --after and --before.
1821 * systemd-inhibit now shows the process name of processes that
1822 have taken an inhibitor lock.
1824 * nss-myhostname will now also resolve "localhost"
1825 implicitly. This makes /etc/hosts an optional file and
1826 nicely handles that on IPv6 ::1 maps to both "localhost" and
1829 * libsystemd-logind.so gained a new call
1830 sd_get_machine_names() to enumerate running containers and
1831 VMs (currently only supported by very new libvirt and
1832 nspawn). sd_login_monitor can now be used to watch
1833 VMs/containers coming and going.
1835 * .include is not allowed recursively anymore, and only in
1836 unit files. Usually it is better to use drop-in snippets in
1837 .d/*.conf anyway, as introduced with systemd 198.
1839 * systemd-analyze gained a new "critical-chain" command that
1840 determines the slowest chain of units run during system
1841 boot-up. It is very useful for tracking down where
1842 optimizing boot time is the most beneficial.
1844 * systemd will no longer allow manipulating service paths in
1845 the name=systemd:/system cgroup tree using ControlGroup= in
1846 units. (But is still fine with it in all other dirs.)
1848 * There's a new systemd-nspawn@.service service file that may
1849 be used to easily run nspawn containers as system
1850 services. With the container's root directory in
1851 /var/lib/container/foobar it is now sufficient to run
1852 "systemctl start systemd-nspawn@foobar.service" to boot it.
1854 * systemd-cgls gained a new parameter "--machine" to list only
1855 the processes within a certain container.
1857 * ConditionSecurity= now can check for "apparmor". We still
1858 are lacking checks for SMACK and IMA for this condition
1859 check though. Patches welcome!
1861 * A new configuration file /etc/systemd/sleep.conf has been
1862 added that may be used to configure which kernel operation
1863 systemd is supposed to execute when "suspend", "hibernate"
1864 or "hybrid-sleep" is requested. This makes the new kernel
1865 "freeze" state accessible to the user.
1867 * ENV{SYSTEMD_WANTS} in udev rules will now implicitly escape
1868 the passed argument if applicable.
1870 Contributions from: Auke Kok, Colin Guthrie, Colin Walters,
1871 Cristian Rodríguez, Daniel Buch, Daniel Wallace, Dave Reisner,
1872 Evangelos Foutras, Greg Kroah-Hartman, Harald Hoyer, Josh
1873 Triplett, Kay Sievers, Lennart Poettering, Lukas Nykryn,
1874 MUNEDA Takahiro, Mantas Mikulėnas, Mirco Tischler, Nathaniel
1875 Chen, Nirbheek Chauhan, Ronny Chevalier, Ross Lagerwall, Tom
1876 Gundersen, Umut Tezduyar, Ville Skyttä, Zbigniew
1881 * The output of 'systemctl list-jobs' got some polishing. The
1882 '--type=' argument may now be passed more than once. A new
1883 command 'systemctl list-sockets' has been added which shows
1884 a list of kernel sockets systemd is listening on with the
1885 socket units they belong to, plus the units these socket
1888 * The experimental libsystemd-bus library got substantial
1889 updates to work in conjunction with the (also experimental)
1890 kdbus kernel project. It works well enough to exchange
1891 messages with some sophistication. Note that kdbus is not
1892 ready yet, and the library is mostly an elaborate test case
1893 for now, and not installable.
1895 * systemd gained a new unit 'systemd-static-nodes.service'
1896 that generates static device nodes earlier during boot, and
1897 can run in conjunction with udev.
1899 * libsystemd-login gained a new call sd_pid_get_user_unit()
1900 to retrieve the user systemd unit a process is running
1901 in. This is useful for systems where systemd is used as
1904 * systemd-nspawn now places all containers in the new /machine
1905 top-level cgroup directory in the name=systemd
1906 hierarchy. libvirt will soon do the same, so that we get a
1907 uniform separation of /system, /user and /machine for system
1908 services, user processes and containers/virtual
1909 machines. This new cgroup hierarchy is also useful to stick
1910 stable names to specific container instances, which can be
1911 recognized later this way (this name may be controlled
1912 via systemd-nspawn's new -M switch). libsystemd-login also
1913 gained a new call sd_pid_get_machine_name() to retrieve the
1914 name of the container/VM a specific process belongs to.
1916 * bootchart can now store its data in the journal.
1918 * libsystemd-journal gained a new call
1919 sd_journal_add_conjunction() for AND expressions to the
1920 matching logic. This can be used to express more complex
1921 logical expressions.
1923 * journactl can now take multiple --unit= and --user-unit=
1926 * The cryptsetup logic now understands the "luks.key=" kernel
1927 command line switch for specifying a file to read the
1928 decryption key from. Also, if a configured key file is not
1929 found the tool will now automatically fall back to prompting
1932 * Python systemd.journal module was updated to wrap recently
1933 added functions from libsystemd-journal. The interface was
1934 changed to bring the low level interface in s.j._Reader
1935 closer to the C API, and the high level interface in
1936 s.j.Reader was updated to wrap and convert all data about
1939 Contributions from: Anatol Pomozov, Auke Kok, Harald Hoyer,
1940 Henrik Grindal Bakken, Josh Triplett, Kay Sievers, Lennart
1941 Poettering, Lukas Nykryn, Mantas Mikulėnas Marius Vollmer,
1942 Martin Jansa, Martin Pitt, Michael Biebl, Michal Schmidt,
1943 Mirco Tischler, Pali Rohar, Simon Peeters, Steven Hiscocks,
1944 Tom Gundersen, Zbigniew Jędrzejewski-Szmek
1948 * journalctl --update-catalog now understands a new --root=
1949 option to operate on catalogs found in a different root
1952 * During shutdown after systemd has terminated all running
1953 services a final killing loop kills all remaining left-over
1954 processes. We will now print the name of these processes
1955 when we send SIGKILL to them, since this usually indicates a
1958 * If /etc/crypttab refers to password files stored on
1959 configured mount points automatic dependencies will now be
1960 generated to ensure the specific mount is established first
1961 before the key file is attempted to be read.
1963 * 'systemctl status' will now show information about the
1964 network sockets a socket unit is listening on.
1966 * 'systemctl status' will also shown information about any
1967 drop-in configuration file for units. (Drop-In configuration
1968 files in this context are files such as
1969 /etc/systemd/systemd/foobar.service.d/*.conf)
1971 * systemd-cgtop now optionally shows summed up CPU times of
1972 cgroups. Press '%' while running cgtop to switch between
1973 percentage and absolute mode. This is useful to determine
1974 which cgroups use up the most CPU time over the entire
1975 runtime of the system. systemd-cgtop has also been updated
1976 to be 'pipeable' for processing with further shell tools.
1978 * 'hostnamectl set-hostname' will now allow setting of FQDN
1981 * The formatting and parsing of time span values has been
1982 changed. The parser now understands fractional expressions
1983 such as "5.5h". The formatter will now output fractional
1984 expressions for all time spans under 1min, i.e. "5.123456s"
1985 rather than "5s 123ms 456us". For time spans under 1s
1986 millisecond values are shown, for those under 1ms
1987 microsecond values are shown. This should greatly improve
1988 all time-related output of systemd.
1990 * libsystemd-login and libsystemd-journal gained new
1991 functions for querying the poll() events mask and poll()
1992 timeout value for integration into arbitrary event
1995 * localectl gained the ability to list available X11 keymaps
1996 (models, layouts, variants, options).
1998 * 'systemd-analyze dot' gained the ability to filter for
1999 specific units via shell-style globs, to create smaller,
2000 more useful graphs. I.e. it is now possible to create simple
2001 graphs of all the dependencies between only target units, or
2002 of all units that Avahi has dependencies with.
2004 Contributions from: Cristian Rodríguez, Dr. Tilmann Bubeck,
2005 Harald Hoyer, Holger Hans Peter Freyther, Kay Sievers, Kelly
2006 Anderson, Koen Kooi, Lennart Poettering, Maksim Melnikau,
2007 Marc-Antoine Perennou, Marius Vollmer, Martin Pitt, Michal
2008 Schmidt, Oleksii Shevchuk, Ronny Chevalier, Simon McVittie,
2009 Steven Hiscocks, Thomas Weißschuh, Umut Tezduyar, Václav
2010 Pavlín, Zbigniew Jędrzejewski-Szmek, Łukasz Stelmach
2014 * The boot-time readahead implementation for rotating media
2015 will now read the read-ahead data in multiple passes which
2016 consist of all read requests made in equidistant time
2017 intervals. This means instead of strictly reading read-ahead
2018 data in its physical order on disk we now try to find a
2019 middle ground between physical and access time order.
2021 * /etc/os-release files gained a new BUILD_ID= field for usage
2022 on operating systems that provide continuous builds of OS
2025 Contributions from: Auke Kok, Eelco Dolstra, Kay Sievers,
2026 Lennart Poettering, Lukas Nykryn, Martin Pitt, Václav Pavlín
2027 William Douglas, Zbigniew Jędrzejewski-Szmek
2031 * systemd-python gained an API exposing libsystemd-daemon.
2033 * The SMACK setup logic gained support for uploading CIPSO
2036 * Behaviour of PrivateTmp=, ReadWriteDirectories=,
2037 ReadOnlyDirectories= and InaccessibleDirectories= has
2038 changed. The private /tmp and /var/tmp directories are now
2039 shared by all processes of a service (which means
2040 ExecStartPre= may now leave data in /tmp that ExecStart= of
2041 the same service can still access). When a service is
2042 stopped its temporary directories are immediately deleted
2043 (normal clean-up with tmpfiles is still done in addition to
2046 * By default, systemd will now set a couple of sysctl
2047 variables in the kernel: the safe sysrq options are turned
2048 on, IP route verification is turned on, and source routing
2049 disabled. The recently added hardlink and softlink
2050 protection of the kernel is turned on. These settings should
2051 be reasonably safe, and good defaults for all new systems.
2053 * The predictable network naming logic may now be turned off
2054 with a new kernel command line switch: net.ifnames=0.
2056 * A new libsystemd-bus module has been added that implements a
2057 pretty complete D-Bus client library. For details see:
2059 http://lists.freedesktop.org/archives/systemd-devel/2013-March/009797.html
2061 * journald will now explicitly flush the journal files to disk
2062 at the latest 5min after each write. The file will then also
2063 be marked offline until the next write. This should increase
2064 reliability in case of a crash. The synchronization delay
2065 can be configured via SyncIntervalSec= in journald.conf.
2067 * There's a new remote-fs-setup.target unit that can be used
2068 to pull in specific services when at least one remote file
2069 system is to be mounted.
2071 * There are new targets timers.target and paths.target as
2072 canonical targets to pull user timer and path units in
2073 from. This complements sockets.target with a similar
2074 purpose for socket units.
2076 * libudev gained a new call udev_device_set_attribute_value()
2077 to set sysfs attributes of a device.
2079 * The udev daemon now sets the default number of worker
2080 processes executed in parallel based on the number of available
2081 CPUs instead of the amount of available RAM. This is supposed
2082 to provide a more reliable default and limit a too aggressive
2083 paralellism for setups with 1000s of devices connected.
2085 Contributions from: Auke Kok, Colin Walters, Cristian
2086 Rodríguez, Daniel Buch, Dave Reisner, Frederic Crozat, Hannes
2087 Reinecke, Harald Hoyer, Jan Alexander Steffens, Jan
2088 Engelhardt, Josh Triplett, Kay Sievers, Lennart Poettering,
2089 Mantas Mikulėnas, Martin Pitt, Mathieu Bridon, Michael Biebl,
2090 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nathaniel Chen,
2091 Oleksii Shevchuk, Ozan Çağlayan, Thomas Hindoe Paaboel
2092 Andersen, Tollef Fog Heen, Tom Gundersen, Umut Tezduyar,
2093 Zbigniew Jędrzejewski-Szmek
2097 * Configuration of unit files may now be extended via drop-in
2098 files without having to edit/override the unit files
2099 themselves. More specifically, if the administrator wants to
2100 change one value for a service file foobar.service he can
2101 now do so by dropping in a configuration snippet into
2102 /etc/systemd/system/foobar.service.d/*.conf. The unit logic
2103 will load all these snippets and apply them on top of the
2104 main unit configuration file, possibly extending or
2105 overriding its settings. Using these drop-in snippets is
2106 generally nicer than the two earlier options for changing
2107 unit files locally: copying the files from
2108 /usr/lib/systemd/system/ to /etc/systemd/system/ and editing
2109 them there; or creating a new file in /etc/systemd/system/
2110 that incorporates the original one via ".include". Drop-in
2111 snippets into these .d/ directories can be placed in any
2112 directory systemd looks for units in, and the usual
2113 overriding semantics between /usr/lib, /etc and /run apply
2116 * Most unit file settings which take lists of items can now be
2117 reset by assigning the empty string to them. For example,
2118 normally, settings such as Environment=FOO=BAR append a new
2119 environment variable assignment to the environment block,
2120 each time they are used. By assigning Environment= the empty
2121 string the environment block can be reset to empty. This is
2122 particularly useful with the .d/*.conf drop-in snippets
2123 mentioned above, since this adds the ability to reset list
2124 settings from vendor unit files via these drop-ins.
2126 * systemctl gained a new "list-dependencies" command for
2127 listing the dependencies of a unit recursively.
2129 * Inhibitors are now honored and listed by "systemctl
2130 suspend", "systemctl poweroff" (and similar) too, not only
2131 GNOME. These commands will also list active sessions by
2134 * Resource limits (as exposed by the various control group
2135 controllers) can now be controlled dynamically at runtime
2136 for all units. More specifically, you can now use a command
2137 like "systemctl set-cgroup-attr foobar.service cpu.shares
2138 2000" to alter the CPU shares a specific service gets. These
2139 settings are stored persistently on disk, and thus allow the
2140 administrator to easily adjust the resource usage of
2141 services with a few simple commands. This dynamic resource
2142 management logic is also available to other programs via the
2143 bus. Almost any kernel cgroup attribute and controller is
2146 * systemd-vconsole-setup will now copy all font settings to
2147 all allocated VTs, where it previously applied them only to
2150 * libsystemd-login gained the new sd_session_get_tty() API
2153 * This release drops support for a few legacy or
2154 distribution-specific LSB facility names when parsing init
2155 scripts: $x-display-manager, $mail-transfer-agent,
2156 $mail-transport-agent, $mail-transfer-agent, $smtp,
2157 $null. Also, the mail-transfer-agent.target unit backing
2158 this has been removed. Distributions which want to retain
2159 compatibility with this should carry the burden for
2160 supporting this themselves and patch support for these back
2161 in, if they really need to. Also, the facilities $syslog and
2162 $local_fs are now ignored, since systemd does not support
2163 early-boot LSB init scripts anymore, and these facilities
2164 are implied anyway for normal services. syslog.target has
2167 * There are new bus calls on PID1's Manager object for
2168 cancelling jobs, and removing snapshot units. Previously,
2169 both calls were only available on the Job and Snapshot
2172 * systemd-journal-gatewayd gained SSL support.
2174 * The various "environment" files, such as /etc/locale.conf
2175 now support continuation lines with a backslash ("\") as
2176 last character in the line, similarly in style (but different)
2177 to how this is supported in shells.
2179 * For normal user processes the _SYSTEMD_USER_UNIT= field is
2180 now implicitly appended to every log entry logged. systemctl
2181 has been updated to filter by this field when operating on a
2182 user systemd instance.
2184 * nspawn will now implicitly add the CAP_AUDIT_WRITE and
2185 CAP_AUDIT_CONTROL capabilities to the capabilities set for
2186 the container. This makes it easier to boot unmodified
2187 Fedora systems in a container, which however still requires
2188 audit=0 to be passed on the kernel command line. Auditing in
2189 kernel and userspace is unfortunately still too broken in
2190 context of containers, hence we recommend compiling it out
2191 of the kernel or using audit=0. Hopefully this will be fixed
2192 one day for good in the kernel.
2194 * nspawn gained the new --bind= and --bind-ro= parameters to
2195 bind mount specific directories from the host into the
2198 * nspawn will now mount its own devpts file system instance
2199 into the container, in order not to leak pty devices from
2200 the host into the container.
2202 * systemd will now read the firmware boot time performance
2203 information from the EFI variables, if the used boot loader
2204 supports this, and takes it into account for boot performance
2205 analysis via "systemd-analyze". This is currently supported
2206 only in conjunction with Gummiboot, but could be supported
2207 by other boot loaders too. For details see:
2209 http://www.freedesktop.org/wiki/Software/systemd/BootLoaderInterface
2211 * A new generator has been added that automatically mounts the
2212 EFI System Partition (ESP) to /boot, if that directory
2213 exists, is empty, and no other file system has been
2214 configured to be mounted there.
2216 * logind will now send out PrepareForSleep(false) out
2217 unconditionally, after coming back from suspend. This may be
2218 used by applications as asynchronous notification for
2219 system resume events.
2221 * "systemctl unlock-sessions" has been added, that allows
2222 unlocking the screens of all user sessions at once, similar
2223 to how "systemctl lock-sessions" already locked all users
2224 sessions. This is backed by a new D-Bus call UnlockSessions().
2226 * "loginctl seat-status" will now show the master device of a
2227 seat. (i.e. the device of a seat that needs to be around for
2228 the seat to be considered available, usually the graphics
2231 * tmpfiles gained a new "X" line type, that allows
2232 configuration of files and directories (with wildcards) that
2233 shall be excluded from automatic cleanup ("aging").
2235 * udev default rules set the device node permissions now only
2236 at "add" events, and do not change them any longer with a
2237 later "change" event.
2239 * The log messages for lid events and power/sleep keypresses
2240 now carry a message ID.
2242 * We now have a substantially larger unit test suite, but this
2243 continues to be work in progress.
2245 * udevadm hwdb gained a new --root= parameter to change the
2246 root directory to operate relative to.
2248 * logind will now issue a background sync() request to the kernel
2249 early at shutdown, so that dirty buffers are flushed to disk early
2250 instead of at the last moment, in order to optimize shutdown
2253 * A new bootctl tool has been added that is an interface for
2254 certain boot loader operations. This is currently a preview
2255 and is likely to be extended into a small mechanism daemon
2256 like timedated, localed, hostnamed, and can be used by
2257 graphical UIs to enumerate available boot options, and
2258 request boot into firmware operations.
2260 * systemd-bootchart has been relicensed to LGPLv2.1+ to match
2261 the rest of the package. It also has been updated to work
2262 correctly in initrds.
2264 * Policykit previously has been runtime optional, and is now
2265 also compile time optional via a configure switch.
2267 * systemd-analyze has been reimplemented in C. Also "systemctl
2268 dot" has moved into systemd-analyze.
2270 * "systemctl status" with no further parameters will now print
2271 the status of all active or failed units.
2273 * Operations such as "systemctl start" can now be executed
2274 with a new mode "--irreversible" which may be used to queue
2275 operations that cannot accidentally be reversed by a later
2276 job queuing. This is by default used to make shutdown
2277 requests more robust.
2279 * The Python API of systemd now gained a new module for
2280 reading journal files.
2282 * A new tool kernel-install has been added that can install
2283 kernel images according to the Boot Loader Specification:
2285 http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec
2287 * Boot time console output has been improved to provide
2288 animated boot time output for hanging jobs.
2290 * A new tool systemd-activate has been added which can be used
2291 to test socket activation with, directly from the command
2292 line. This should make it much easier to test and debug
2293 socket activation in daemons.
2295 * journalctl gained a new "--reverse" (or -r) option to show
2296 journal output in reverse order (i.e. newest line first).
2298 * journalctl gained a new "--pager-end" (or -e) option to jump
2299 to immediately jump to the end of the journal in the
2300 pager. This is only supported in conjunction with "less".
2302 * journalctl gained a new "--user-unit=" option, that works
2303 similarly to "--unit=" but filters for user units rather than
2306 * A number of unit files to ease adoption of systemd in
2307 initrds has been added. This moves some minimal logic from
2308 the various initrd implementations into systemd proper.
2310 * The journal files are now owned by a new group
2311 "systemd-journal", which exists specifically to allow access
2312 to the journal, and nothing else. Previously, we used the
2313 "adm" group for that, which however possibly covers more
2314 than just journal/log file access. This new group is now
2315 already used by systemd-journal-gatewayd to ensure this
2316 daemon gets access to the journal files and as little else
2317 as possible. Note that "make install" will also set FS ACLs
2318 up for /var/log/journal to give "adm" and "wheel" read
2319 access to it, in addition to "systemd-journal" which owns
2320 the journal files. We recommend that packaging scripts also
2321 add read access to "adm" + "wheel" to /var/log/journal, and
2322 all existing/future journal files. To normal users and
2323 administrators little changes, however packagers need to
2324 ensure to create the "systemd-journal" system group at
2325 package installation time.
2327 * The systemd-journal-gatewayd now runs as unprivileged user
2328 systemd-journal-gateway:systemd-journal-gateway. Packaging
2329 scripts need to create these system user/group at
2332 * timedated now exposes a new boolean property CanNTP that
2333 indicates whether a local NTP service is available or not.
2335 * systemd-detect-virt will now also detect xen PVs
2337 * The pstore file system is now mounted by default, if it is
2340 * In addition to the SELinux and IMA policies we will now also
2341 load SMACK policies at early boot.
2343 Contributions from: Adel Gadllah, Aleksander Morgado, Auke
2344 Kok, Ayan George, Bastien Nocera, Colin Walters, Daniel Buch,
2345 Daniel Wallace, Dave Reisner, David Herrmann, David Strauss,
2346 Eelco Dolstra, Enrico Scholz, Frederic Crozat, Harald Hoyer,
2347 Jan Janssen, Jonathan Callen, Kay Sievers, Lennart Poettering,
2348 Lukas Nykryn, Mantas Mikulėnas, Marc-Antoine Perennou, Martin
2349 Pitt, Mauro Dreissig, Max F. Albrecht, Michael Biebl, Michael
2350 Olbrich, Michal Schmidt, Michal Sekletar, Michal Vyskocil,
2351 Michał Bartoszkiewicz, Mirco Tischler, Nathaniel Chen, Nestor
2352 Ovroy, Oleksii Shevchuk, Paul W. Frields, Piotr Drąg, Rob
2353 Clark, Ryan Lortie, Simon McVittie, Simon Peeters, Steven
2354 Hiscocks, Thomas Hindoe Paaboel Andersen, Tollef Fog Heen, Tom
2355 Gundersen, Umut Tezduyar, William Giokas, Zbigniew
2356 Jędrzejewski-Szmek, Zeeshan Ali (Khattak)
2360 * Timer units now support calendar time events in addition to
2361 monotonic time events. That means you can now trigger a unit
2362 based on a calendar time specification such as "Thu,Fri
2363 2013-*-1,5 11:12:13" which refers to 11:12:13 of the first
2364 or fifth day of any month of the year 2013, given that it is
2365 a thursday or friday. This brings timer event support
2366 considerably closer to cron's capabilities. For details on
2367 the supported calendar time specification language see
2370 * udev now supports a number of different naming policies for
2371 network interfaces for predictable names, and a combination
2372 of these policies is now the default. Please see this wiki
2373 document for details:
2375 http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames
2377 * Auke Kok's bootchart implementation has been added to the
2378 systemd tree. It is an optional component that can graph the
2379 boot in quite some detail. It is one of the best bootchart
2380 implementations around and minimal in its code and
2383 * nss-myhostname has been integrated into the systemd source
2384 tree. nss-myhostname guarantees that the local hostname
2385 always stays resolvable via NSS. It has been a weak
2386 requirement of systemd-hostnamed since a long time, and
2387 since its code is actually trivial we decided to just
2388 include it in systemd's source tree. It can be turned off
2389 with a configure switch.
2391 * The read-ahead logic is now capable of properly detecting
2392 whether a btrfs file system is on SSD or rotating media, in
2393 order to optimize the read-ahead scheme. Previously, it was
2394 only capable of detecting this on traditional file systems
2397 * In udev, additional device properties are now read from the
2398 IAB in addition to the OUI database. Also, Bluetooth company
2399 identities are attached to the devices as well.
2401 * In service files %U may be used as specifier that is
2402 replaced by the configured user name of the service.
2404 * nspawn may now be invoked without a controlling TTY. This
2405 makes it suitable for invocation as its own service. This
2406 may be used to set up a simple containerized server system
2407 using only core OS tools.
2409 * systemd and nspawn can now accept socket file descriptors
2410 when they are started for socket activation. This enables
2411 implementation of socket activated nspawn
2412 containers. i.e. think about autospawning an entire OS image
2413 when the first SSH or HTTP connection is received. We expect
2414 that similar functionality will also be added to libvirt-lxc
2417 * journalctl will now suppress ANSI color codes when
2418 presenting log data.
2420 * systemctl will no longer show control group information for
2421 a unit if a the control group is empty anyway.
2423 * logind can now automatically suspend/hibernate/shutdown the
2426 * /etc/machine-info and hostnamed now also expose the chassis
2427 type of the system. This can be used to determine whether
2428 the local system is a laptop, desktop, handset or
2429 tablet. This information may either be configured by the
2430 user/vendor or is automatically determined from ACPI and DMI
2431 information if possible.
2433 * A number of PolicyKit actions are now bound together with
2434 "imply" rules. This should simplify creating UIs because
2435 many actions will now authenticate similar ones as well.
2437 * Unit files learnt a new condition ConditionACPower= which
2438 may be used to conditionalize a unit depending on whether an
2439 AC power source is connected or not, of whether the system
2440 is running on battery power.
2442 * systemctl gained a new "is-failed" verb that may be used in
2443 shell scripts and suchlike to check whether a specific unit
2444 is in the "failed" state.
2446 * The EnvironmentFile= setting in unit files now supports file
2447 globbing, and can hence be used to easily read a number of
2448 environment files at once.
2450 * systemd will no longer detect and recognize specific
2451 distributions. All distribution-specific #ifdeffery has been
2452 removed, systemd is now fully generic and
2453 distribution-agnostic. Effectively, not too much is lost as
2454 a lot of the code is still accessible via explicit configure
2455 switches. However, support for some distribution specific
2456 legacy configuration file formats has been dropped. We
2457 recommend distributions to simply adopt the configuration
2458 files everybody else uses now and convert the old
2459 configuration from packaging scripts. Most distributions
2460 already did that. If that's not possible or desirable,
2461 distributions are welcome to forward port the specific
2462 pieces of code locally from the git history.
2464 * When logging a message about a unit systemd will now always
2465 log the unit name in the message meta data.
2467 * localectl will now also discover system locale data that is
2468 not stored in locale archives, but directly unpacked.
2470 * logind will no longer unconditionally use framebuffer
2471 devices as seat masters, i.e. as devices that are required
2472 to be existing before a seat is considered preset. Instead,
2473 it will now look for all devices that are tagged as
2474 "seat-master" in udev. By default framebuffer devices will
2475 be marked as such, but depending on local systems other
2476 devices might be marked as well. This may be used to
2477 integrate graphics cards using closed source drivers (such
2478 as NVidia ones) more nicely into logind. Note however, that
2479 we recommend using the open source NVidia drivers instead,
2480 and no udev rules for the closed-source drivers will be
2481 shipped from us upstream.
2483 Contributions from: Adam Williamson, Alessandro Crismani, Auke
2484 Kok, Colin Walters, Daniel Wallace, Dave Reisner, David
2485 Herrmann, David Strauss, Dimitrios Apostolou, Eelco Dolstra,
2486 Eric Benoit, Giovanni Campagna, Hannes Reinecke, Henrik
2487 Grindal Bakken, Hermann Gausterer, Kay Sievers, Lennart
2488 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marcel Holtmann,
2489 Martin Pitt, Matthew Monaco, Michael Biebl, Michael Terry,
2490 Michal Schmidt, Michal Sekletar, Michał Bartoszkiewicz, Oleg
2491 Samarin, Pekka Lundstrom, Philip Nilsson, Ramkumar
2492 Ramachandra, Richard Yao, Robert Millan, Sami Kerola, Shawn
2493 Landden, Thomas Hindoe Paaboel Andersen, Thomas Jarosch,
2494 Tollef Fog Heen, Tom Gundersen, Umut Tezduyar, Zbigniew
2499 * udev gained support for loading additional device properties
2500 from an indexed database that is keyed by vendor/product IDs
2501 and similar device identifiers. For the beginning this
2502 "hwdb" is populated with data from the well-known PCI and
2503 USB database, but also includes PNP, ACPI and OID data. In
2504 the longer run this indexed database shall grow into
2505 becoming the one central database for non-essential
2506 userspace device metadata. Previously, data from the PCI/USB
2507 database was only attached to select devices, since the
2508 lookup was a relatively expensive operation due to O(n) time
2509 complexity (with n being the number of entries in the
2510 database). Since this is now O(1), we decided to add in this
2511 data for all devices where this is available, by
2512 default. Note that the indexed database needs to be rebuilt
2513 when new data files are installed. To achieve this you need
2514 to update your packaging scripts to invoke "udevadm hwdb
2515 --update" after installation of hwdb data files. For
2516 RPM-based distributions we introduced the new
2517 %udev_hwdb_update macro for this purpose.
2519 * The Journal gained support for the "Message Catalog", an
2520 indexed database to link up additional information with
2521 journal entries. For further details please check:
2523 http://www.freedesktop.org/wiki/Software/systemd/catalog
2525 The indexed message catalog database also needs to be
2526 rebuilt after installation of message catalog files. Use
2527 "journalctl --update-catalog" for this. For RPM-based
2528 distributions we introduced the %journal_catalog_update
2529 macro for this purpose.
2531 * The Python Journal bindings gained support for the standard
2532 Python logging framework.
2534 * The Journal API gained new functions for checking whether
2535 the underlying file system of a journal file is capable of
2536 properly reporting file change notifications, or whether
2537 applications that want to reflect journal changes "live"
2538 need to recheck journal files continuously in appropriate
2541 * It is now possible to set the "age" field for tmpfiles
2542 entries to 0, indicating that files matching this entry
2543 shall always be removed when the directories are cleaned up.
2545 * coredumpctl gained a new "gdb" verb which invokes gdb
2546 right-away on the selected coredump.
2548 * There's now support for "hybrid sleep" on kernels that
2549 support this, in addition to "suspend" and "hibernate". Use
2550 "systemctl hybrid-sleep" to make use of this.
2552 * logind's HandleSuspendKey= setting (and related settings)
2553 now gained support for a new "lock" setting to simply
2554 request the screen lock on all local sessions, instead of
2555 actually executing a suspend or hibernation.
2557 * systemd will now mount the EFI variables file system by
2560 * Socket units now gained support for configuration of the
2561 SMACK security label.
2563 * timedatectl will now output the time of the last and next
2564 daylight saving change.
2566 * We dropped support for various legacy and distro-specific
2567 concepts, such as insserv, early-boot SysV services
2568 (i.e. those for non-standard runlevels such as 'b' or 'S')
2569 or ArchLinux /etc/rc.conf support. We recommend the
2570 distributions who still need support this to either continue
2571 to maintain the necessary patches downstream, or find a
2572 different solution. (Talk to us if you have questions!)
2574 * Various systemd components will now bypass PolicyKit checks
2575 for root and otherwise handle properly if PolicyKit is not
2576 found to be around. This should fix most issues for
2577 PolicyKit-less systems. Quite frankly this should have been
2578 this way since day one. It is absolutely our intention to
2579 make systemd work fine on PolicyKit-less systems, and we
2580 consider it a bug if something does not work as it should if
2581 PolicyKit is not around.
2583 * For embedded systems it is now possible to build udev and
2584 systemd without blkid and/or kmod support.
2586 * "systemctl switch-root" is now capable of switching root
2587 more than once. I.e. in addition to transitions from the
2588 initrd to the host OS it is now possible to transition to
2589 further OS images from the host. This is useful to implement
2590 offline updating tools.
2592 * Various other additions have been made to the RPM macros
2593 shipped with systemd. Use %udev_rules_update() after
2594 installing new udev rules files. %_udevhwdbdir,
2595 %_udevrulesdir, %_journalcatalogdir, %_tmpfilesdir,
2596 %_sysctldir are now available which resolve to the right
2597 directories for packages to place various data files in.
2599 * journalctl gained the new --full switch (in addition to
2600 --all, to disable ellipsation for long messages.
2602 Contributions from: Anders Olofsson, Auke Kok, Ben Boeckel,
2603 Colin Walters, Cosimo Cecchi, Daniel Wallace, Dave Reisner,
2604 Eelco Dolstra, Holger Hans Peter Freyther, Kay Sievers,
2605 Chun-Yi Lee, Lekensteyn, Lennart Poettering, Mantas Mikulėnas,
2606 Marti Raudsepp, Martin Pitt, Mauro Dreissig, Michael Biebl,
2607 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nis Martensen,
2608 Oleksii Shevchuk, Olivier Brunel, Ramkumar Ramachandra, Thomas
2609 Bächler, Thomas Hindoe Paaboel Andersen, Tom Gundersen, Tony
2610 Camuso, Umut Tezduyar, Zbigniew Jędrzejewski-Szmek
2614 * journalctl gained new --since= and --until= switches to
2615 filter by time. It also now supports nice filtering for
2616 units via --unit=/-u.
2618 * Type=oneshot services may use ExecReload= and do the
2621 * The journal daemon now supports time-based rotation and
2622 vacuuming, in addition to the usual disk-space based
2625 * The journal will now index the available field values for
2626 each field name. This enables clients to show pretty drop
2627 downs of available match values when filtering. The bash
2628 completion of journalctl has been updated
2629 accordingly. journalctl gained a new switch -F to list all
2630 values a certain field takes in the journal database.
2632 * More service events are now written as structured messages
2633 to the journal, and made recognizable via message IDs.
2635 * The timedated, localed and hostnamed mini-services which
2636 previously only provided support for changing time, locale
2637 and hostname settings from graphical DEs such as GNOME now
2638 also have a minimal (but very useful) text-based client
2639 utility each. This is probably the nicest way to changing
2640 these settings from the command line now, especially since
2641 it lists available options and is fully integrated with bash
2644 * There's now a new tool "systemd-coredumpctl" to list and
2645 extract coredumps from the journal.
2647 * We now install a README each in /var/log/ and
2648 /etc/rc.d/init.d explaining where the system logs and init
2649 scripts went. This hopefully should help folks who go to
2650 that dirs and look into the otherwise now empty void and
2651 scratch their heads.
2653 * When user-services are invoked (by systemd --user) the
2654 $MANAGERPID env var is set to the PID of systemd.
2656 * SIGRTMIN+24 when sent to a --user instance will now result
2657 in immediate termination of systemd.
2659 * gatewayd received numerous feature additions such as a
2660 "follow" mode, for live syncing and filtering.
2662 * browse.html now allows filtering and showing detailed
2663 information on specific entries. Keyboard navigation and
2664 mouse screen support has been added.
2666 * gatewayd/journalctl now supports HTML5/JSON
2667 Server-Sent-Events as output.
2669 * The SysV init script compatibility logic will now
2670 heuristically determine whether a script supports the
2671 "reload" verb, and only then make this available as
2674 * "systemctl status --follow" has been removed, use "journalctl
2677 * journald.conf's RuntimeMinSize=, PersistentMinSize= settings
2678 have been removed since they are hardly useful to be
2681 * And I'd like to take the opportunity to specifically mention
2682 Zbigniew for his great contributions. Zbigniew, you rock!
2684 Contributions from: Andrew Eikum, Christian Hesse, Colin
2685 Guthrie, Daniel J Walsh, Dave Reisner, Eelco Dolstra, Ferenc
2686 Wágner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Mantas
2687 Mikulėnas, Martin Mikkelsen, Martin Pitt, Michael Olbrich,
2688 Michael Stapelberg, Michal Schmidt, Sebastian Ott, Thomas
2689 Bächler, Umut Tezduyar, Will Woods, Wulf C. Krueger, Zbigniew
2690 Jędrzejewski-Szmek, Сковорода Никита Андреевич
2694 * If /etc/vconsole.conf is non-existent or empty we will no
2695 longer load any console font or key map at boot by
2696 default. Instead the kernel defaults will be left
2697 intact. This is definitely the right thing to do, as no
2698 configuration should mean no configuration, and hard-coding
2699 font names that are different on all archs is probably a bad
2700 idea. Also, the kernel default key map and font should be
2701 good enough for most cases anyway, and mostly identical to
2702 the userspace fonts/key maps we previously overloaded them
2703 with. If distributions want to continue to default to a
2704 non-kernel font or key map they should ship a default
2705 /etc/vconsole.conf with the appropriate contents.
2707 Contributions from: Colin Walters, Daniel J Walsh, Dave
2708 Reisner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Tollef
2709 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
2713 * journalctl gained a new --cursor= switch to show entries
2714 starting from the specified location in the journal.
2716 * We now enforce a size limit on journal entry fields exported
2717 with "-o json" in journalctl. Fields larger than 4K will be
2718 assigned null. This can be turned off with --all.
2720 * An (optional) journal gateway daemon is now available as
2721 "systemd-journal-gatewayd.service". This service provides
2722 access to the journal via HTTP and JSON. This functionality
2723 will be used to implement live log synchronization in both
2724 pull and push modes, but has various other users too, such
2725 as easy log access for debugging of embedded devices. Right
2726 now it is already useful to retrieve the journal via HTTP:
2728 # systemctl start systemd-journal-gatewayd.service
2729 # wget http://localhost:19531/entries
2731 This will download the journal contents in a
2732 /var/log/messages compatible format. The same as JSON:
2734 # curl -H"Accept: application/json" http://localhost:19531/entries
2736 This service is also accessible via a web browser where a
2737 single static HTML5 app is served that uses the JSON logic
2738 to enable the user to do some basic browsing of the
2739 journal. This will be extended later on. Here's an example
2740 screenshot of this app in its current state:
2742 http://0pointer.de/public/journal-gatewayd
2744 Contributions from: Kay Sievers, Lennart Poettering, Robert
2745 Milasan, Tom Gundersen
2749 * The bash completion logic is now available for journalctl
2752 * We do not mount the "cpuset" controller anymore together with
2753 "cpu" and "cpuacct", as "cpuset" groups generally cannot be
2754 started if no parameters are assigned to it. "cpuset" hence
2755 broke code that assumed it it could create "cpu" groups and
2758 * journalctl -f will now subscribe to terminal size changes,
2759 and line break accordingly.
2761 Contributions from: Dave Reisner, Kay Sievers, Lennart
2762 Poettering, Lukas Nykrynm, Mirco Tischler, Václav Pavlín
2766 * nspawn will now create a symlink /etc/localtime in the
2767 container environment, copying the host's timezone
2768 setting. Previously this has been done via a bind mount, but
2769 since symlinks cannot be bind mounted this has now been
2770 changed to create/update the appropriate symlink.
2772 * journalctl -n's line number argument is now optional, and
2773 will default to 10 if omitted.
2775 * journald will now log the maximum size the journal files may
2776 take up on disk. This is particularly useful if the default
2777 built-in logic of determining this parameter from the file
2778 system size is used. Use "systemctl status
2779 systemd-journald.service" to see this information.
2781 * The multi-seat X wrapper tool has been stripped down. As X
2782 is now capable of enumerating graphics devices via udev in a
2783 seat-aware way the wrapper is not strictly necessary
2784 anymore. A stripped down temporary stop-gap is still shipped
2785 until the upstream display managers have been updated to
2786 fully support the new X logic. Expect this wrapper to be
2787 removed entirely in one of the next releases.
2789 * HandleSleepKey= in logind.conf has been split up into
2790 HandleSuspendKey= and HandleHibernateKey=. The old setting
2791 is not available anymore. X11 and the kernel are
2792 distuingishing between these keys and we should too. This
2793 also means the inhibition lock for these keys has been split
2796 Contributions from: Dave Airlie, Eelco Dolstra, Lennart
2797 Poettering, Lukas Nykryn, Václav Pavlín
2801 * Whenever a unit changes state we will now log this to the
2802 journal and show along the unit's own log output in
2805 * ConditionPathIsMountPoint= can now properly detect bind
2806 mount points too. (Previously, a bind mount of one file
2807 system to another place in the same file system could not be
2808 detected as mount, since they shared struct stat's st_dev
2811 * We will now mount the cgroup controllers cpu, cpuacct,
2812 cpuset and the controllers net_cls, net_prio together by
2815 * nspawn containers will now have a virtualized boot
2816 ID. (i.e. /proc/sys/kernel/random/boot_id is now mounted
2817 over with a randomized ID at container initialization). This
2818 has the effect of making "journalctl -b" do the right thing
2821 * The JSON output journal serialization has been updated not
2822 to generate "endless" list objects anymore, but rather one
2823 JSON object per line. This is more in line how most JSON
2824 parsers expect JSON objects. The new output mode
2825 "json-pretty" has been added to provide similar output, but
2826 neatly aligned for readability by humans.
2828 * We dropped all explicit sync() invocations in the shutdown
2829 code. The kernel does this implicitly anyway in the kernel
2830 reboot() syscall. halt(8)'s -n option is now a compatibility
2833 * We now support virtualized reboot() in containers, as
2834 supported by newer kernels. We will fall back to exit() if
2835 CAP_SYS_REBOOT is not available to the container. Also,
2836 nspawn makes use of this now and will actually reboot the
2837 container if the containerized OS asks for that.
2839 * journalctl will only show local log output by default
2840 now. Use --merge (-m) to show remote log output, too.
2842 * libsystemd-journal gained the new sd_journal_get_usage()
2843 call to determine the current disk usage of all journal
2844 files. This is exposed in the new "journalctl --disk-usage"
2847 * journald gained a new configuration setting SplitMode= in
2848 journald.conf which may be used to control how user journals
2849 are split off. See journald.conf(5) for details.
2851 * A new condition type ConditionFileNotEmpty= has been added.
2853 * tmpfiles' "w" lines now support file globbing, to write
2854 multiple files at once.
2856 * We added Python bindings for the journal submission
2857 APIs. More Python APIs for a number of selected APIs will
2858 likely follow. Note that we intend to add native bindings
2859 only for the Python language, as we consider it common
2860 enough to deserve bindings shipped within systemd. There are
2861 various projects outside of systemd that provide bindings
2862 for languages such as PHP or Lua.
2864 * Many conditions will now resolve specifiers such as %i. In
2865 addition, PathChanged= and related directives of .path units
2866 now support specifiers as well.
2868 * There's now a new RPM macro definition for the system preset
2871 * journald will now warn if it ca not forward a message to the
2872 syslog daemon because its socket is full.
2874 * timedated will no longer write or process /etc/timezone,
2875 except on Debian. As we do not support late mounted /usr
2876 anymore /etc/localtime always being a symlink is now safe,
2877 and hence the information in /etc/timezone is not necessary
2880 * logind will now always reserve one VT for a text getty (VT6
2881 by default). Previously if more than 6 X sessions where
2882 started they took up all the VTs with auto-spawned gettys,
2883 so that no text gettys were available anymore.
2885 * udev will now automatically inform the btrfs kernel logic
2886 about btrfs RAID components showing up. This should make
2887 simple hotplug based btrfs RAID assembly work.
2889 * PID 1 will now increase its RLIMIT_NOFILE to 64K by default
2890 (but not for its children which will stay at the kernel
2891 default). This should allow setups with a lot more listening
2894 * systemd will now always pass the configured timezone to the
2895 kernel at boot. timedated will do the same when the timezone
2898 * logind's inhibition logic has been updated. By default,
2899 logind will now handle the lid switch, the power and sleep
2900 keys all the time, even in graphical sessions. If DEs want
2901 to handle these events on their own they should take the new
2902 handle-power-key, handle-sleep-key and handle-lid-switch
2903 inhibitors during their runtime. A simple way to achiveve
2904 that is to invoke the DE wrapped in an invocation of:
2906 systemd-inhibit --what=handle-power-key:handle-sleep-key:handle-lid-switch ...
2908 * Access to unit operations is now checked via SELinux taking
2909 the unit file label and client process label into account.
2911 * systemd will now notify the administrator in the journal
2912 when he over-mounts a non-empty directory.
2914 * There are new specifiers that are resolved in unit files,
2915 for the host name (%H), the machine ID (%m) and the boot ID
2918 Contributions from: Allin Cottrell, Auke Kok, Brandon Philips,
2919 Colin Guthrie, Colin Walters, Daniel J Walsh, Dave Reisner,
2920 Eelco Dolstra, Jan Engelhardt, Kay Sievers, Lennart
2921 Poettering, Lucas De Marchi, Lukas Nykryn, Mantas Mikulėnas,
2922 Martin Pitt, Matthias Clasen, Michael Olbrich, Pierre Schmitz,
2923 Shawn Landden, Thomas Hindoe Paaboel Andersen, Tom Gundersen,
2924 Václav Pavlín, Yin Kangkai, Zbigniew Jędrzejewski-Szmek
2928 * Support for reading structured kernel messages from
2929 /dev/kmsg has now been added and is enabled by default.
2931 * Support for reading kernel messages from /proc/kmsg has now
2932 been removed. If you want kernel messages in the journal
2933 make sure to run a recent kernel (>= 3.5) that supports
2934 reading structured messages from /dev/kmsg (see
2935 above). /proc/kmsg is now exclusive property of classic
2936 syslog daemons again.
2938 * The libudev API gained the new
2939 udev_device_new_from_device_id() call.
2941 * The logic for file system namespace (ReadOnlyDirectory=,
2942 ReadWriteDirectoy=, PrivateTmp=) has been reworked not to
2943 require pivot_root() anymore. This means fewer temporary
2944 directories are created below /tmp for this feature.
2946 * nspawn containers will now see and receive all submounts
2947 made on the host OS below the root file system of the
2950 * Forward Secure Sealing is now supported for Journal files,
2951 which provide cryptographical sealing of journal files so
2952 that attackers cannot alter log history anymore without this
2953 being detectable. Lennart will soon post a blog story about
2954 this explaining it in more detail.
2956 * There are two new service settings RestartPreventExitStatus=
2957 and SuccessExitStatus= which allow configuration of exit
2958 status (exit code or signal) which will be excepted from the
2959 restart logic, resp. consider successful.
2961 * journalctl gained the new --verify switch that can be used
2962 to check the integrity of the structure of journal files and
2963 (if Forward Secure Sealing is enabled) the contents of
2966 * nspawn containers will now be run with /dev/stdin, /dev/fd/
2967 and similar symlinks pre-created. This makes running shells
2968 as container init process a lot more fun.
2970 * The fstab support can now handle PARTUUID= and PARTLABEL=
2973 * A new ConditionHost= condition has been added to match
2974 against the hostname (with globs) and machine ID. This is
2975 useful for clusters where a single OS image is used to
2976 provision a large number of hosts which shall run slightly
2977 different sets of services.
2979 * Services which hit the restart limit will now be placed in a
2982 Contributions from: Bertram Poettering, Dave Reisner, Huang
2983 Hang, Kay Sievers, Lennart Poettering, Lukas Nykryn, Martin
2984 Pitt, Simon Peeters, Zbigniew Jędrzejewski-Szmek
2988 * When running in --user mode systemd will now become a
2989 subreaper (PR_SET_CHILD_SUBREAPER). This should make the ps
2990 tree a lot more organized.
2992 * A new PartOf= unit dependency type has been introduced that
2993 may be used to group services in a natural way.
2995 * "systemctl enable" may now be used to enable instances of
2998 * journalctl now prints error log levels in red, and
2999 warning/notice log levels in bright white. It also supports
3000 filtering by log level now.
3002 * cgtop gained a new -n switch (similar to top), to configure
3003 the maximum number of iterations to run for. It also gained
3004 -b, to run in batch mode (accepting no input).
3006 * The suffix ".service" may now be omitted on most systemctl
3007 command lines involving service unit names.
3009 * There's a new bus call in logind to lock all sessions, as
3010 well as a loginctl verb for it "lock-sessions".
3012 * libsystemd-logind.so gained a new call sd_journal_perror()
3013 that works similar to libc perror() but logs to the journal
3014 and encodes structured information about the error number.
3016 * /etc/crypttab entries now understand the new keyfile-size=
3019 * shutdown(8) now can send a (configurable) wall message when
3020 a shutdown is cancelled.
3022 * The mount propagation mode for the root file system will now
3023 default to "shared", which is useful to make containers work
3024 nicely out-of-the-box so that they receive new mounts from
3025 the host. This can be undone locally by running "mount
3026 --make-rprivate /" if needed.
3028 * The prefdm.service file has been removed. Distributions
3029 should maintain this unit downstream if they intend to keep
3030 it around. However, we recommend writing normal unit files
3031 for display managers instead.
3033 * Since systemd is a crucial part of the OS we will now
3034 default to a number of compiler switches that improve
3035 security (hardening) such as read-only relocations, stack
3036 protection, and suchlike.
3038 * The TimeoutSec= setting for services is now split into
3039 TimeoutStartSec= and TimeoutStopSec= to allow configuration
3040 of individual time outs for the start and the stop phase of
3043 Contributions from: Artur Zaprzala, Arvydas Sidorenko, Auke
3044 Kok, Bryan Kadzban, Dave Reisner, David Strauss, Harald Hoyer,
3045 Jim Meyering, Kay Sievers, Lennart Poettering, Mantas
3046 Mikulėnas, Martin Pitt, Michal Schmidt, Michal Sekletar, Peter
3047 Alfredsen, Shawn Landden, Simon Peeters, Terence Honles, Tom
3048 Gundersen, Zbigniew Jędrzejewski-Szmek
3052 * The journal and id128 C APIs are now fully documented as man
3055 * Extra safety checks have been added when transitioning from
3056 the initial RAM disk to the main system to avoid accidental
3059 * /etc/crypttab entries now understand the new keyfile-offset=
3062 * systemctl -t can now be used to filter by unit load state.
3064 * The journal C API gained the new sd_journal_wait() call to
3065 make writing synchronous journal clients easier.
3067 * journalctl gained the new -D switch to show journals from a
3070 * journalctl now displays a special marker between log
3071 messages of two different boots.
3073 * The journal is now explicitly flushed to /var via a service
3074 systemd-journal-flush.service, rather than implicitly simply
3075 by seeing /var/log/journal to be writable.
3077 * journalctl (and the journal C APIs) can now match for much
3078 more complex expressions, with alternatives and
3081 * When transitioning from the initial RAM disk to the main
3082 system we will now kill all processes in a killing spree to
3083 ensure no processes stay around by accident.
3085 * Three new specifiers may be used in unit files: %u, %h, %s
3086 resolve to the user name, user home directory resp. user
3087 shell. This is useful for running systemd user instances.
3089 * We now automatically rotate journal files if their data
3090 object hash table gets a fill level > 75%. We also size the
3091 hash table based on the configured maximum file size. This
3092 together should lower hash collisions drastically and thus
3093 speed things up a bit.
3095 * journalctl gained the new "--header" switch to introspect
3096 header data of journal files.
3098 * A new setting SystemCallFilters= has been added to services
3099 which may be used to apply blacklists or whitelists to
3100 system calls. This is based on SECCOMP Mode 2 of Linux 3.5.
3102 * nspawn gained a new --link-journal= switch (and quicker: -j)
3103 to link the container journal with the host. This makes it
3104 very easy to centralize log viewing on the host for all
3105 guests while still keeping the journal files separated.
3107 * Many bugfixes and optimizations
3109 Contributions from: Auke Kok, Eelco Dolstra, Harald Hoyer, Kay
3110 Sievers, Lennart Poettering, Malte Starostik, Paul Menzel, Rex
3111 Tsai, Shawn Landden, Tom Gundersen, Ville Skyttä, Zbigniew
3116 * Several tools now understand kernel command line arguments,
3117 which are only read when run in an initial RAM disk. They
3118 usually follow closely their normal counterparts, but are
3121 * There's a new tool to analyze the readahead files that are
3122 automatically generated at boot. Use:
3124 /usr/lib/systemd/systemd-readahead analyze /.readahead
3126 * We now provide an early debug shell on tty9 if this enabled. Use:
3128 systemctl enable debug-shell.service
3130 * All plymouth related units have been moved into the Plymouth
3131 package. Please make sure to upgrade your Plymouth version
3134 * systemd-tmpfiles now supports getting passed the basename of
3135 a configuration file only, in which case it will look for it
3136 in all appropriate directories automatically.
3138 * udevadm info now takes a /dev or /sys path as argument, and
3139 does the right thing. Example:
3141 udevadm info /dev/sda
3142 udevadm info /sys/class/block/sda
3144 * systemctl now prints a warning if a unit is stopped but a
3145 unit that might trigger it continues to run. Example: a
3146 service is stopped but the socket that activates it is left
3149 * "systemctl status" will now mention if the log output was
3150 shortened due to rotation since a service has been started.
3152 * The journal API now exposes functions to determine the
3153 "cutoff" times due to rotation.
3155 * journald now understands SIGUSR1 and SIGUSR2 for triggering
3156 immediately flushing of runtime logs to /var if possible,
3157 resp. for triggering immediate rotation of the journal
3160 * It is now considered an error if a service is attempted to
3161 be stopped that is not loaded.
3163 * XDG_RUNTIME_DIR now uses numeric UIDs instead of usernames.
3165 * systemd-analyze now supports Python 3
3167 * tmpfiles now supports cleaning up directories via aging
3168 where the first level dirs are always kept around but
3169 directories beneath it automatically aged. This is enabled
3170 by prefixing the age field with '~'.
3172 * Seat objects now expose CanGraphical, CanTTY properties
3173 which is required to deal with very fast bootups where the
3174 display manager might be running before the graphics drivers
3175 completed initialization.
3177 * Seat objects now expose a State property.
3179 * We now include RPM macros for service enabling/disabling
3180 based on the preset logic. We recommend RPM based
3181 distributions to make use of these macros if possible. This
3182 makes it simpler to reuse RPM spec files across
3185 * We now make sure that the collected systemd unit name is
3186 always valid when services log to the journal via
3189 * There's a new man page kernel-command-line(7) detailing all
3190 command line options we understand.
3192 * The fstab generator may now be disabled at boot by passing
3193 fstab=0 on the kernel command line.
3195 * A new kernel command line option modules-load= is now understood
3196 to load a specific kernel module statically, early at boot.
3198 * Unit names specified on the systemctl command line are now
3199 automatically escaped as needed. Also, if file system or
3200 device paths are specified they are automatically turned
3201 into the appropriate mount or device unit names. Example:
3203 systemctl status /home
3204 systemctl status /dev/sda
3206 * The SysVConsole= configuration option has been removed from
3207 system.conf parsing.
3209 * The SysV search path is no longer exported on the D-Bus
3212 * The Names= option is been removed from unit file parsing.
3214 * There's a new man page bootup(7) detailing the boot process.
3216 * Every unit and every generator we ship with systemd now
3217 comes with full documentation. The self-explanatory boot is
3220 * A couple of services gained "systemd-" prefixes in their
3221 name if they wrap systemd code, rather than only external
3222 code. Among them fsck@.service which is now
3223 systemd-fsck@.service.
3225 * The HaveWatchdog property has been removed from the D-Bus
3228 * systemd.confirm_spawn= on the kernel command line should now
3231 * There's a new man page crypttab(5) which details all options
3232 we actually understand.
3234 * systemd-nspawn gained a new --capability= switch to pass
3235 additional capabilities to the container.
3237 * timedated will now read known NTP implementation unit names
3238 from /usr/lib/systemd/ntp-units.d/*.list,
3239 systemd-timedated-ntp.target has been removed.
3241 * journalctl gained a new switch "-b" that lists log data of
3242 the current boot only.
3244 * The notify socket is in the abstract namespace again, in
3245 order to support daemons which chroot() at start-up.
3247 * There is a new Storage= configuration option for journald
3248 which allows configuration of where log data should go. This
3249 also provides a way to disable journal logging entirely, so
3250 that data collected is only forwarded to the console, the
3251 kernel log buffer or another syslog implementation.
3253 * Many bugfixes and optimizations
3255 Contributions from: Auke Kok, Colin Guthrie, Dave Reisner,
3256 David Strauss, Eelco Dolstra, Kay Sievers, Lennart Poettering,
3257 Lukas Nykryn, Michal Schmidt, Michal Sekletar, Paul Menzel,
3258 Shawn Landden, Tom Gundersen
3262 * "systemctl help <unit>" now shows the man page if one is
3265 * Several new man pages have been added.
3267 * MaxLevelStore=, MaxLevelSyslog=, MaxLevelKMsg=,
3268 MaxLevelConsole= can now be specified in
3269 journald.conf. These options allow reducing the amount of
3270 data stored on disk or forwarded by the log level.
3272 * TimerSlackNSec= can now be specified in system.conf for
3273 PID1. This allows system-wide power savings.
3275 Contributions from: Dave Reisner, Kay Sievers, Lauri Kasanen,
3276 Lennart Poettering, Malte Starostik, Marc-Antoine Perennou,
3281 * logind is now capable of (optionally) handling power and
3282 sleep keys as well as the lid switch.
3284 * journalctl now understands the syntax "journalctl
3285 /usr/bin/avahi-daemon" to get all log output of a specific
3288 * CapabilityBoundingSet= in system.conf now also influences
3289 the capability bound set of usermode helpers of the kernel.
3291 Contributions from: Daniel Drake, Daniel J. Walsh, Gert
3292 Michael Kulyk, Harald Hoyer, Jean Delvare, Kay Sievers,
3293 Lennart Poettering, Matthew Garrett, Matthias Clasen, Paul
3294 Menzel, Shawn Landden, Tero Roponen, Tom Gundersen
3298 * Note that we skipped 139 releases here in order to set the
3299 new version to something that is greater than both udev's
3300 and systemd's most recent version number.
3302 * udev: all udev sources are merged into the systemd source tree now.
3303 All future udev development will happen in the systemd tree. It
3304 is still fully supported to use the udev daemon and tools without
3305 systemd running, like in initramfs or other init systems. Building
3306 udev though, will require the *build* of the systemd tree, but
3307 udev can be properly *run* without systemd.
3309 * udev: /lib/udev/devices/ are not read anymore; systemd-tmpfiles
3310 should be used to create dead device nodes as workarounds for broken
3313 * udev: RUN+="socket:..." and udev_monitor_new_from_socket() is
3314 no longer supported. udev_monitor_new_from_netlink() needs to be
3315 used to subscribe to events.
3317 * udev: when udevd is started by systemd, processes which are left
3318 behind by forking them off of udev rules, are unconditionally cleaned
3319 up and killed now after the event handling has finished. Services or
3320 daemons must be started as systemd services. Services can be
3321 pulled-in by udev to get started, but they can no longer be directly
3322 forked by udev rules.
3324 * udev: the daemon binary is called systemd-udevd now and installed
3325 in /usr/lib/systemd/. Standalone builds or non-systemd systems need
3326 to adapt to that, create symlink, or rename the binary after building
3329 * libudev no longer provides these symbols:
3330 udev_monitor_from_socket()
3331 udev_queue_get_failed_list_entry()
3332 udev_get_{dev,sys,run}_path()
3333 The versions number was bumped and symbol versioning introduced.
3335 * systemd-loginctl and systemd-journalctl have been renamed
3336 to loginctl and journalctl to match systemctl.
3338 * The config files: /etc/systemd/systemd-logind.conf and
3339 /etc/systemd/systemd-journald.conf have been renamed to
3340 logind.conf and journald.conf. Package updates should rename
3341 the files to the new names on upgrade.
3343 * For almost all files the license is now LGPL2.1+, changed
3344 from the previous GPL2.0+. Exceptions are some minor stuff
3345 of udev (which will be changed to LGPL2.1 eventually, too),
3346 and the MIT licensed sd-daemon.[ch] library that is suitable
3347 to be used as drop-in files.
3349 * systemd and logind now handle system sleep states, in
3350 particular suspending and hibernating.
3352 * logind now implements a sleep/shutdown/idle inhibiting logic
3353 suitable for a variety of uses. Soonishly Lennart will blog
3354 about this in more detail.
3356 * var-run.mount and var-lock.mount are no longer provided
3357 (which prevously bind mounted these directories to their new
3358 places). Distributions which have not converted these
3359 directories to symlinks should consider stealing these files
3360 from git history and add them downstream.
3362 * We introduced the Documentation= field for units and added
3363 this to all our shipped units. This is useful to make it
3364 easier to explore the boot and the purpose of the various
3367 * All smaller setup units (such as
3368 systemd-vconsole-setup.service) now detect properly if they
3369 are run in a container and are skipped when
3370 appropriate. This guarantees an entirely noise-free boot in
3371 Linux container environments such as systemd-nspawn.
3373 * A framework for implementing offline system updates is now
3374 integrated, for details see:
3375 http://freedesktop.org/wiki/Software/systemd/SystemUpdates
3377 * A new service type Type=idle is available now which helps us
3378 avoiding ugly interleaving of getty output and boot status
3381 * There's now a system-wide CapabilityBoundingSet= option to
3382 globally reduce the set of capabilities for the
3383 system. This is useful to drop CAP_SYS_MKNOD, CAP_SYS_RAWIO,
3384 CAP_NET_RAW, CAP_SYS_MODULE, CAP_SYS_TIME, CAP_SYS_PTRACE or
3385 even CAP_NET_ADMIN system-wide for secure systems.
3387 * There are now system-wide DefaultLimitXXX= options to
3388 globally change the defaults of the various resource limits
3389 for all units started by PID 1.
3391 * Harald Hoyer's systemd test suite has been integrated into
3392 systemd which allows easy testing of systemd builds in qemu
3393 and nspawn. (This is really awesome! Ask us for details!)
3395 * The fstab parser is now implemented as generator, not inside
3398 * systemctl will now warn you if .mount units generated from
3399 /etc/fstab are out of date due to changes in fstab that
3400 have not been read by systemd yet.
3402 * systemd is now suitable for usage in initrds. Dracut has
3403 already been updated to make use of this. With this in place
3404 initrds get a slight bit faster but primarily are much
3405 easier to introspect and debug since "systemctl status" in
3406 the host system can be used to introspect initrd services,
3407 and the journal from the initrd is kept around too.
3409 * systemd-delta has been added, a tool to explore differences
3410 between user/admin configuration and vendor defaults.
3412 * PrivateTmp= now affects both /tmp and /var/tmp.
3414 * Boot time status messages are now much prettier and feature
3415 proper english language. Booting up systemd has never been
3418 * Read-ahead pack files now include the inode number of all
3419 files to pre-cache. When the inode changes the pre-caching
3420 is not attempted. This should be nicer to deal with updated
3421 packages which might result in changes of read-ahead
3424 * We now temporaritly lower the kernel's read_ahead_kb variable
3425 when collecting read-ahead data to ensure the kernel's
3426 built-in read-ahead does not add noise to our measurements
3427 of necessary blocks to pre-cache.
3429 * There's now RequiresMountsFor= to add automatic dependencies
3430 for all mounts necessary for a specific file system path.
3432 * MountAuto= and SwapAuto= have been removed from
3433 system.conf. Mounting file systems at boot has to take place
3436 * nspawn now learned a new switch --uuid= to set the machine
3437 ID on the command line.
3439 * nspawn now learned the -b switch to automatically search
3442 * vt102 is now the default TERM for serial TTYs, upgraded from
3445 * systemd-logind now works on VT-less systems.
3447 * The build tree has been reorganized. The individual
3448 components now have directories of their own.
3450 * A new condition type ConditionPathIsReadWrite= is now available.
3452 * nspawn learned the new -C switch to create cgroups for the
3453 container in other hierarchies.
3455 * We now have support for hardware watchdogs, configurable in
3458 * The scheduled shutdown logic now has a public API.
3460 * We now mount /tmp as tmpfs by default, but this can be
3461 masked and /etc/fstab can override it.
3463 * Since udisks does not make use of /media anymore we are not
3464 mounting a tmpfs on it anymore.
3466 * journalctl gained a new --local switch to only interleave
3467 locally generated journal files.
3469 * We can now load the IMA policy at boot automatically.
3471 * The GTK tools have been split off into a systemd-ui.
3473 Contributions from: Andreas Schwab, Auke Kok, Ayan George,
3474 Colin Guthrie, Daniel Mack, Dave Reisner, David Ward, Elan
3475 Ruusamäe, Frederic Crozat, Gergely Nagy, Guillermo Vidal,
3476 Hannes Reinecke, Harald Hoyer, Javier Jardón, Kay Sievers,
3477 Lennart Poettering, Lucas De Marchi, Léo Gillot-Lamure,
3478 Marc-Antoine Perennou, Martin Pitt, Matthew Monaco, Maxim
3479 A. Mikityanskiy, Michael Biebl, Michael Olbrich, Michal
3480 Schmidt, Nis Martensen, Patrick McCarty, Roberto Sassu, Shawn
3481 Landden, Sjoerd Simons, Sven Anders, Tollef Fog Heen, Tom
3486 * This is mostly a bugfix release
3488 * Support optional initialization of the machine ID from the
3489 KVM or container configured UUID.
3491 * Support immediate reboots with "systemctl reboot -ff"
3493 * Show /etc/os-release data in systemd-analyze output
3495 * Many bugfixes for the journal, including endianness fixes and
3496 ensuring that disk space enforcement works
3498 * sd-login.h is C++ comptaible again
3500 * Extend the /etc/os-release format on request of the Debian
3503 * We now refuse non-UTF8 strings used in various configuration
3504 and unit files. This is done to ensure we do not pass invalid
3505 data over D-Bus or expose it elsewhere.
3507 * Register Mimo USB Screens as suitable for automatic seat
3510 * Read SELinux client context from journal clients in a race
3513 * Reorder configuration file lookup order. /etc now always
3514 overrides /run in order to allow the administrator to always
3515 and unconditionally override vendor supplied or
3516 automatically generated data.
3518 * The various user visible bits of the journal now have man
3519 pages. We still lack man pages for the journal API calls
3522 * We now ship all man pages in HTML format again in the
3525 Contributions from: Dave Reisner, Dirk Eibach, Frederic
3526 Crozat, Harald Hoyer, Kay Sievers, Lennart Poettering, Marti
3527 Raudsepp, Michal Schmidt, Shawn Landden, Tero Roponen, Thierry
3532 * This is mostly a bugfix release
3534 * systems lacking /etc/os-release are no longer supported.
3536 * Various functionality updates to libsystemd-login.so
3538 * Track class of PAM logins to distuingish greeters from
3541 Contributions from: Kay Sievers, Lennart Poettering, Michael
3546 * This is an important bugfix release for v41.
3548 * Building man pages is now optional which should be useful
3549 for those building systemd from git but unwilling to install
3552 * Watchdog support for supervising services is now usable. In
3553 a future release support for hardware watchdogs
3554 (i.e. /dev/watchdog) will be added building on this.
3556 * Service start rate limiting is now configurable and can be
3557 turned off per service. When a start rate limit is hit a
3558 reboot can automatically be triggered.
3560 * New CanReboot(), CanPowerOff() bus calls in systemd-logind.
3562 Contributions from: Benjamin Franzke, Bill Nottingham,
3563 Frederic Crozat, Lennart Poettering, Michael Olbrich, Michal
3564 Schmidt, Michał Górny, Piotr Drąg
3568 * The systemd binary is installed /usr/lib/systemd/systemd now;
3569 An existing /sbin/init symlink needs to be adapted with the
3572 * The code that loads kernel modules has been ported to invoke
3573 libkmod directly, instead of modprobe. This means we do not
3574 support systems with module-init-tools anymore.
3576 * Watchdog support is now already useful, but still not
3579 * A new kernel command line option systemd.setenv= is
3580 understood to set system wide environment variables
3581 dynamically at boot.
3583 * We now limit the set of capabilities of systemd-journald.
3585 * We now set SIGPIPE to ignore by default, since it only is
3586 useful in shell pipelines, and has little use in general
3587 code. This can be disabled with IgnoreSIPIPE=no in unit
3590 Contributions from: Benjamin Franzke, Kay Sievers, Lennart
3591 Poettering, Michael Olbrich, Michal Schmidt, Tom Gundersen,
3596 * This is mostly a bugfix release
3598 * We now expose the reason why a service failed in the
3599 "Result" D-Bus property.
3601 * Rudimentary service watchdog support (will be completed over
3602 the next few releases.)
3604 * When systemd forks off in order execute some service we will
3605 now immediately changes its argv[0] to reflect which process
3606 it will execute. This is useful to minimize the time window
3607 with a generic argv[0], which makes bootcharts more useful
3609 Contributions from: Alvaro Soliverez, Chris Paulson-Ellis, Kay
3610 Sievers, Lennart Poettering, Michael Olbrich, Michal Schmidt,
3611 Mike Kazantsev, Ray Strode
3615 * This is mostly a test release, but incorporates many
3618 * New systemd-cgtop tool to show control groups by their
3621 * Linking against libacl for ACLs is optional again. If
3622 disabled, support tracking device access for active logins
3623 goes becomes unavailable, and so does access to the user
3624 journals by the respective users.
3626 * If a group "adm" exists, journal files are automatically
3627 owned by them, thus allow members of this group full access
3628 to the system journal as well as all user journals.
3630 * The journal now stores the SELinux context of the logging
3631 client for all entries.
3633 * Add C++ inclusion guards to all public headers
3635 * New output mode "cat" in the journal to print only text
3636 messages, without any meta data like date or time.
3638 * Include tiny X server wrapper as a temporary stop-gap to
3639 teach XOrg udev display enumeration. This is used by display
3640 managers such as gdm, and will go away as soon as XOrg
3641 learned native udev hotplugging for display devices.
3643 * Add new systemd-cat tool for executing arbitrary programs
3644 with STDERR/STDOUT connected to the journal. Can also act as
3645 BSD logger replacement, and does so by default.
3647 * Optionally store all locally generated coredumps in the
3648 journal along with meta data.
3650 * systemd-tmpfiles learnt four new commands: n, L, c, b, for
3651 writing short strings to files (for usage for /sys), and for
3652 creating symlinks, character and block device nodes.
3654 * New unit file option ControlGroupPersistent= to make cgroups
3655 persistent, following the mechanisms outlined in
3656 http://www.freedesktop.org/wiki/Software/systemd/PaxControlGroups
3658 * Support multiple local RTCs in a sane way
3660 * No longer monopolize IO when replaying readahead data on
3661 rotating disks, since we might starve non-file-system IO to
3662 death, since fanotify() will not see accesses done by blkid,
3665 * Do not show kernel threads in systemd-cgls anymore, unless
3666 requested with new -k switch.
3668 Contributions from: Dan Horák, Kay Sievers, Lennart
3669 Poettering, Michal Schmidt
3673 * This is mostly a test release, but incorporates many
3676 * The git repository moved to:
3677 git://anongit.freedesktop.org/systemd/systemd
3678 ssh://git.freedesktop.org/git/systemd/systemd
3680 * First release with the journal
3681 http://0pointer.de/blog/projects/the-journal.html
3683 * The journal replaces both systemd-kmsg-syslogd and
3684 systemd-stdout-bridge.
3686 * New sd_pid_get_unit() API call in libsystemd-logind
3688 * Many systemadm clean-ups
3690 * Introduce remote-fs-pre.target which is ordered before all
3691 remote mounts and may be used to start services before all
3694 * Added Mageia support
3696 * Add bash completion for systemd-loginctl
3698 * Actively monitor PID file creation for daemons which exit in
3699 the parent process before having finished writing the PID
3700 file in the daemon process. Daemons which do this need to be
3701 fixed (i.e. PID file creation must have finished before the
3702 parent exits), but we now react a bit more gracefully to them.
3704 * Add colourful boot output, mimicking the well-known output
3705 of existing distributions.
3707 * New option PassCredentials= for socket units, for
3708 compatibility with a recent kernel ABI breakage.
3710 * /etc/rc.local is now hooked in via a generator binary, and
3711 thus will no longer act as synchronization point during
3714 * systemctl list-unit-files now supports --root=.
3716 * systemd-tmpfiles now understands two new commands: z, Z for
3717 relabelling files according to the SELinux database. This is
3718 useful to apply SELinux labels to specific files in /sys,
3721 * Output of SysV services is now forwarded to both the console
3722 and the journal by default, not only just the console.
3724 * New man pages for all APIs from libsystemd-login.
3726 * The build tree got reorganized and a the build system is a
3727 lot more modular allowing embedded setups to specifically
3728 select the components of systemd they are interested in.
3730 * Support for Linux systems lacking the kernel VT subsystem is
3733 * configure's --with-rootdir= got renamed to
3734 --with-rootprefix= to follow the naming used by udev and
3737 * Unless specified otherwise we will now install to /usr instead
3738 of /usr/local by default.
3740 * Processes with '@' in argv[0][0] are now excluded from the
3741 final shut-down killing spree, following the logic explained
3743 http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons
3745 * All processes remaining in a service cgroup when we enter
3746 the START or START_PRE states are now killed with
3747 SIGKILL. That means it is no longer possible to spawn
3748 background processes from ExecStart= lines (which was never
3749 supported anyway, and bad style).
3751 * New PropagateReloadTo=/PropagateReloadFrom= options to bind
3752 reloading of units together.
3754 Contributions from: Bill Nottingham, Daniel J. Walsh, Dave
3755 Reisner, Dexter Morgan, Gregs Gregs, Jonathan Nieder, Kay
3756 Sievers, Lennart Poettering, Michael Biebl, Michal Schmidt,
3757 Michał Górny, Ran Benita, Thomas Jarosch, Tim Waugh, Tollef
3758 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek