1 systemd System and Service Manager
5 * As an experimental feature, udev now tries to lock the
6 disk device node (flock(LOCK_SH|LOCK_NB)) while it
7 executes events for the disk or any of its partitions.
8 Applications like partitioning programs can lock the
9 disk device node (flock(LOCK_EX)) and claim temporary
10 device ownership that way; udev will entirely skip all event
11 handling for this disk and its partitions. If the disk
12 was opened for writing, the close will trigger a partition
13 table rescan in udev's "watch" facility, and if needed
14 synthesize "change" events for the disk and all its partitions.
15 This is now unconditionally enabled, if it turns out to
16 cause major problems, we might turn it on only for specific
17 devices, or might need to disable it entirely. Device-mapper
18 devices are excluded from this logic.
20 * We temporarily dropped the "-l" switch for fsck invocations,
21 since they collide with the flock() logic above. util-linux
22 upstream has been changed already to avoid this conflict,
23 and we will readd "-l" as soon as util-linux with this
24 change has been released.
26 * The dependency on libattr has been removed. Since a long
27 time the extended attribute calls have moved to glibc, and
28 libattr is thus unnecessary.
30 * Virtualization detection works without priviliges now. This
31 means the systemd-detect-virt binary no longer requires
32 CAP_SYS_PTRACE file capabilities, and our daemons can run
33 with fewer privileges.
35 * systemd-networkd now runs under its own "systemd-network"
36 user. It retains the CAP_NET_ADMIN, CAP_NET_BIND_SERVICE,
37 CAP_NET_BROADCAST, CAP_NET_RAW capabilities though, but
38 loses the ability to write to files owned by root this way.
40 * Similar, systemd-resolved now runs under its own
41 "systemd-resolve" user with no capabilities remaining.
43 * Similar, systemd-bus-proxyd now runs under its own
44 "systemd-bus-proxy" user with only CAP_IPC_OWNER remaining.
46 * systemd-networkd gained support for setting up "veth"
47 virtual ethernet devices for container connectivity, as well
48 as GRE and VTI tunnels.
50 * systemd-networkd will no longer automatically attempt to
51 manually load kernel modules necessary for certain tunnel
52 transports. Instead it is assumed the kernel loads them
53 automatically when required. This only works correctly on
54 very new kernels. On older kernels, please consider adding
55 the kernel modules to /etc/modules-load.d/ as a work-around.
57 * The resolv.conf file systemd-resolved generates has been
58 moved to /run/systemd/resolve/, if you have a symlink from
59 /etc/resolv.conf it might be necessary to correct it.
61 * Two new service settings ProtectedHome= and ProtectedSystem=
62 have been added. When enabled they will make the user data
63 (such as /home) inaccessible or read-only and the system
64 (such as /usr) read-only, for specific services. This allows
65 very light-weight per-service sandboxing to avoid
66 modifications of user data or system files from
67 services. These two new switches have been enabled for all
68 of systemd's long-running services, where appropriate.
70 * Socket units gained new SocketUser= and SocketGroup=
71 settings to set the owner user and group of AF_UNIX sockets
72 and FIFOs in the file system.
74 * Socket units gained a new RemoveOnStop= setting. If enabled
75 all FIFOS and sockets in the file system will be removed
76 when the specific socket unit is stopped.
78 * Socket units gained a new Symlinks= setting. It takes a list
79 of symlinks to create to file system sockets or FIFOs
80 created by the specific unix sockets. This is useful to
81 manage symlinks to socket nodes with the same life-cycle as
84 * The /dev/log socket and /dev/initctl FIFO have been moved to
85 /run, and have been replaced by symlinks. This allows
86 connecting to these facilities even if PrivateDevices=yes is
87 used for a service (which makes /dev/log itself unavailable,
88 but /run is left). This also has the benefit of ensuring
89 that /dev only contains device nodes, directories and
90 symlinks, and nothing else.
92 * sd-daemon gained two new calls sd_pid_notify() and
93 sd_pid_notifyf(). They are similar to sd_notify() and
94 sd_notifyf(), but allow overriding of the source PID of
95 notification messages if permissions permit this. This is
96 useful to send notify messages on behalf of a different
97 process (for example, the parent process). The
98 systemd-notify tool has been updated to make use of this
99 when sending messages (so that notification messages now
100 originate from the shell script invoking systemd-notify and
101 not the systemd-notify process itself. This should minimize
102 a race where systemd fails to associate notification
103 messages to services when the originating process already
106 * A new "on-abnormal" setting for Restart= has been added. If
107 set it will result in automatic restarts on all "abnormal"
108 reasons for a process to exit, which includes unclean
109 signals, core dumps, timeouts and watchdog timeouts, but
110 does not include clean and unclean exit codes or clean
111 signals. Restart=on-abnormal is an alternative for
112 Restart=on-failure for services that shall be able to
113 terminate and avoid restarts on certain errors, by
114 indicating so with an unclean exit code. Restart=on-failure
115 or Restart=on-abnormal is now the recommended setting for
116 all long-running services.
118 * If the InaccessibleDirectories= service setting points to a
119 mount point (or if there are any submounts contained within
120 it), it is now attempted to completely unmount it, to make
121 the file systems truly unavailable for the respective
124 * The ReadOnlyDirectories= service setting and
125 systemd-nspawn's --read-only parameter are now recursively
126 applied to all submounts, too.
128 * Mount units may now be created transiently via the bus APIs.
130 * The support for SysV and LSB init scripts has been removed
131 from the systemd daemon itself. Instead, it is now
132 implemented as a generator that creates native systemd units
133 from these scripts when needed. This enables us to remove a
134 substantial amount of legacy code from PID 1, following the
135 fact that many distributions only ship a very small number
136 of LSB/SysV init scripts nowadays.
138 * Priviliged Xen (dom0) domains are not considered
139 virtualization anymore by the virtualization detection
140 logic. After all, they generally have unrestricted access to
141 the hardware and usually are used to manage the unprivileged
144 * systemd-tmpfiles gained a new "C" line type, for copying
145 files or entire directories.
147 * systemd-tmpfiles "m" lines are now fully equivalent to "z"
148 lines. So far they have been non-globbing versions of the
149 latter, and have thus been redundant. In future it is
150 recommended to only use "z"; and "m" has hence been removed
151 from the documentation, even though it stays supported.
153 * A tmpfiles snippet to recreate the most basic structure in
154 /var has been added. This is enough to create the /var/run →
155 /run symlink and create a couple of structural
156 directories. This allows systems to boot up with an empty or
157 volatile /var. Of course, while with this change the core OS
158 now is capable with dealing with a volatile /var not all
159 user services are ready for it. However, we hope that sooner
160 or later many service daemons will be changed upstream so
161 that they are able to automatically create their necessary
162 directories in /var at boot, should they be missing. This is
163 the first step to allow state-less systems that only require
164 the vendor image for /usr to boot.
166 * systemd-nspawn has gained a new --tmpfs= switch to mount an
167 empty tmpfs instance to a specific directory. This is
168 particularly useful for making use of the automatic
169 reconstruction of /var (see above), by passing --tmpfs=/var.
171 * Access modes specified in tmpfiles snippets may now be
172 prefixed with "~", which indicates that they shall be masked
173 by whether the existing file or directly is currently
174 writable, readable or executable at all. Also, if specified
175 the sgid/suid/sticky bits will be masked for all
178 * A new passive target unit "network-pre.target" has been
179 added which is useful for services that shall run before any
180 network is configured, for example firewall scripts.
182 * The "floppy" group that previously owned the /dev/fd*
183 devices is no longer used. The "disk" group is now used
184 instead. Distributions should probably deprecate usage of
187 Contributions from: Camilo Aguilar, Christian Hesse, Colin Ian
188 King, Cristian Rodríguez, Daniel Buch, Dave Reisner, David
189 Strauss, Denis Tikhomirov, John, Jonathan Liu, Kay Sievers,
190 Lennart Poettering, Mantas Mikulėnas, Mark Eichin, Ronny
191 Chevalier, Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel
192 Andersen, Tom Gundersen, Umut Tezduyar Lindskog, Zbigniew
195 -- Berlin, 2014-06-11
199 * A new "systemd-timesyncd" daemon has been added for
200 synchronizing the system clock across the network. It
201 implements an SNTP client. In contrast to NTP
202 implementations such as chrony or the NTP reference server
203 this only implements a client side, and does not bother with
204 the full NTP complexity, focusing only on querying time from
205 one remote server and synchronizing the local clock to
206 it. Unless you intend to serve NTP to networked clients or
207 want to connect to local hardware clocks this simple NTP
208 client should be more than appropriate for most
209 installations. The daemon runs with minimal privileges, and
210 has been hooked up with networkd to only operate when
211 network connectivity is available. The daemon saves the
212 current clock to disk every time a new NTP sync has been
213 acquired, and uses this to possibly correct the system clock
214 early at bootup, in order to accommodate for systems that
215 lack an RTC such as the Raspberry Pi and embedded devices,
216 and make sure that time monotonically progresses on these
217 systems, even if it is not always correct. To make use of
218 this daemon a new system user and group "systemd-timesync"
219 needs to be created on installation of systemd.
221 * The queue "seqnum" interface of libudev has been disabled, as
222 it was generally incompatible with device namespacing as
223 sequence numbers of devices go "missing" if the devices are
224 part of a different namespace.
226 * "systemctl list-timers" and "systemctl list-sockets" gained
227 a --recursive switch for showing units of these types also
228 for all local containers, similar in style to the already
229 supported --recursive switch for "systemctl list-units".
231 * A new RebootArgument= setting has been added for service
232 units, which may be used to specify a kernel reboot argument
233 to use when triggering reboots with StartLimitAction=.
235 * A new FailureAction= setting has been added for service
236 units which may be used to specify an operation to trigger
237 when a service fails. This works similarly to
238 StartLimitAction=, but unlike it controls what is done
239 immediately rather than only after several attempts to
240 restart the service in question.
242 * hostnamed got updated to also expose the kernel name,
243 release, and version on the bus. This is useful for
244 executing commands like hostnamectl with the -H switch.
245 systemd-analyze makes use of this to properly display
246 details when running non-locally.
248 * The bootchart tool can now show cgroup information in the
251 * The CFS CPU quota cgroup attribute is now exposed for
252 services. The new CPUQuota= switch has been added for this
253 which takes a percentage value. Setting this will have the
254 result that a service may never get more CPU time than the
255 specified percentage, even if the machine is otherwise idle.
257 * systemd-networkd learned IPIP and SIT tunnel support.
259 * LSB init scripts exposing a dependency on $network will now
260 get a dependency on network-online.target rather than simply
261 network.target. This should bring LSB handling closer to
262 what it was on SysV systems.
264 * A new fsck.repair= kernel option has been added to control
265 how fsck shall deal with unclean file systems at boot.
267 * The (.ini) configuration file parser will now silently
268 ignore sections whose name begins with "X-". This may be
269 used to maintain application-specific extension sections in unit
272 * machined gained a new API to query the IP addresses of
273 registered containers. "machinectl status" has been updated
274 to show these addresses in its output.
276 * A new call sd_uid_get_display() has been added to the
277 sd-login APIs for querying the "primary" session of a
278 user. The "primary" session of the user is elected from the
279 user's sessions and generally a graphical session is
280 preferred over a text one.
282 * A minimal systemd-resolved daemon has been added. It
283 currently simply acts as a companion to systemd-networkd and
284 manages resolv.conf based on per-interface DNS
285 configuration, possibly supplied via DHCP. In the long run
286 we hope to extend this into a local DNSSEC enabled DNS and
289 * The systemd-networkd-wait-online tool is now enabled by
290 default. It will delay network-online.target until a network
291 connection has been configured. The tool primarily integrates
292 with networkd, but will also make a best effort to make sense
293 of network configuration performed in some other way.
295 * Two new service options StartupCPUShares= and
296 StartupBlockIOWeight= have been added that work similarly to
297 CPUShares= and BlockIOWeight= however only apply during
298 system startup. This is useful to prioritize certain services
299 differently during bootup than during normal runtime.
301 * hostnamed has been changed to prefer the statically
302 configured hostname in /etc/hostname (unless set to
303 'localhost' or empty) over any dynamic one supplied by
304 dhcp. With this change the rules for picking the hostname
305 match more closely the rules of other configuration settings
306 where the local administrator's configuration in /etc always
307 overrides any other settings.
309 Contributions fron: Ali H. Caliskan, Alison Chaiken, Bas van
310 den Berg, Brandon Philips, Cristian Rodríguez, Daniel Buch,
311 Dan Kilman, Dave Reisner, David Härdeman, David Herrmann,
312 David Strauss, Dimitris Spingos, Djalal Harouni, Eelco
313 Dolstra, Evan Nemerson, Florian Albrechtskirchinger, Greg
314 Kroah-Hartman, Harald Hoyer, Holger Hans Peter Freyther, Jan
315 Engelhardt, Jani Nikula, Jason St. John, Jeffrey Clark,
316 Jonathan Boulle, Kay Sievers, Lennart Poettering, Lukas
317 Nykryn, Lukasz Skalski, Łukasz Stelmach, Mantas Mikulėnas,
318 Marcel Holtmann, Martin Pitt, Matthew Monaco, Michael
319 Marineau, Michael Olbrich, Michal Sekletar, Mike Gilbert, Nis
320 Martensen, Patrik Flykt, Philip Lorenz, poma, Ray Strode,
321 Reyad Attiyat, Robert Milasan, Scott Thrasher, Stef Walter,
322 Steven Siloti, Susant Sahani, Tanu Kaskinen, Thomas Bächler,
323 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar
324 Lindskog, WaLyong Cho, Will Woods, Zbigniew
327 -- Beijing, 2014-05-28
331 * When restoring the screen brightness at boot, stay away from
332 the darkest setting or from the lowest 5% of the available
333 range, depending on which is the larger value of both. This
334 should effectively protect the user from rebooting into a
335 black screen, should the brightness have been set to minimum
338 * sd-login gained a new sd_machine_get_class() call to
339 determine the class ("vm" or "container") of a machine
340 registered with machined.
342 * sd-login gained new calls
343 sd_peer_get_{session,owner_uid,unit,user_unit,slice,machine_name}(),
344 to query the identity of the peer of a local AF_UNIX
345 connection. They operate similarly to their sd_pid_get_xyz()
348 * PID 1 will now maintain a system-wide system state engine
349 with the states "starting", "running", "degraded",
350 "maintenance", "stopping". These states are bound to system
351 startup, normal runtime, runtime with at least one failed
352 service, rescue/emergency mode and system shutdown. This
353 state is shown in the "systemctl status" output when no unit
354 name is passed. It is useful to determine system state, in
355 particularly when doing so for many systems or containers at
358 * A new command "list-machines" has been added to "systemctl"
359 that lists all local OS containers and shows their system
360 state (see above), if systemd runs inside of them.
362 * systemctl gained a new "-r" switch to recursively enumerate
363 units on all local containers, when used with the
364 "list-unit" command (which is the default one that is
365 executed when no parameters are specified).
367 * The GPT automatic partition discovery logic will now honour
368 two GPT partition flags: one may be set on a partition to
369 cause it to be mounted read-only, and the other may be set
370 on a partition to ignore it during automatic discovery.
372 * Two new GPT type UUIDs have been added for automatic root
373 partition discovery, for 32-bit and 64-bit ARM. This is not
374 particularly useful for discovering the root directory on
375 these architectures during bare-metal boots (since UEFI is
376 not common there), but still very useful to allow booting of
377 ARM disk images in nspawn with the -i option.
379 * MAC addresses of interfaces created with nspawn's
380 --network-interface= switch will now be generated from the
381 machine name, and thus be stable between multiple invocations
384 * logind will now automatically remove all IPC objects owned
385 by a user if she or he fully logs out. This makes sure that
386 users who are logged out cannot continue to consume IPC
387 resources. This covers SysV memory, semaphores and message
388 queues as well as POSIX shared memory and message
389 queues. Traditionally, SysV and POSIX IPC had no life-cycle
390 limits. With this functionality, that is corrected. This may
391 be turned off by using the RemoveIPC= switch of logind.conf.
393 * The systemd-machine-id-setup and tmpfiles tools gained a
394 --root= switch to operate on a specific root directory,
397 * journald can now forward logged messages to the TTYs of all
398 logged in users ("wall"). This is the default for all
399 emergency messages now.
401 * A new tool systemd-journal-remote has been added to stream
402 journal log messages across the network.
404 * /sys/fs/cgroup/ is now mounted read-only after all cgroup
405 controller trees are mounted into it. Note that the
406 directories mounted beneath it are not read-only. This is a
407 security measure and is particularly useful because glibc
408 actually includes a search logic to pick any tmpfs it can
409 find to implement shm_open() if /dev/shm is not available
410 (which it might very well be in namespaced setups).
412 * machinectl gained a new "poweroff" command to cleanly power
413 down a local OS container.
415 * The PrivateDevices= unit file setting will now also drop the
416 CAP_MKNOD capability from the capability bound set, and
417 imply DevicePolicy=closed.
419 * PrivateDevices=, PrivateNetwork= and PrivateTmp= is now used
420 comprehensively on all long-running systemd services where
423 * systemd-udevd will now run in a disassociated mount
424 namespace. To mount directories from udev rules, make sure to
425 pull in mount units via SYSTEMD_WANTS properties.
427 * The kdbus support gained support for uploading policy into
428 the kernel. sd-bus gained support for creating "monitoring"
429 connections that can eavesdrop into all bus communication
430 for debugging purposes.
432 * Timestamps may now be specified in seconds since the UNIX
433 epoch Jan 1st, 1970 by specifying "@" followed by the value
436 * Native tcpwrap support in systemd has been removed. tcpwrap
437 is old code, not really maintained anymore and has serious
438 shortcomings, and better options such as firewalls
439 exist. For setups that require tcpwrap usage, please
440 consider invoking your socket-activated service via tcpd,
441 like on traditional inetd.
443 * A new system.conf configuration option
444 DefaultTimerAccuracySec= has been added that controls the
445 default AccuracySec= setting of .timer units.
447 * Timer units gained a new WakeSystem= switch. If enabled,
448 timers configured this way will cause the system to resume
449 from system suspend (if the system supports that, which most
452 * Timer units gained a new Persistent= switch. If enabled,
453 timers configured this way will save to disk when they have
454 been last triggered. This information is then used on next
455 reboot to possible execute overdue timer events, that
456 could not take place because the system was powered off.
457 This enables simple anacron-like behaviour for timer units.
459 * systemctl's "list-timers" will now also list the time a
460 timer unit was last triggered in addition to the next time
461 it will be triggered.
463 * systemd-networkd will now assign predictable IPv4LL
464 addresses to its local interfaces.
466 Contributions from: Brandon Philips, Daniel Buch, Daniel Mack,
467 Dave Reisner, David Herrmann, Gerd Hoffmann, Greg
468 Kroah-Hartman, Hendrik Brueckner, Jason St. John, Josh
469 Triplett, Kay Sievers, Lennart Poettering, Marc-Antoine
470 Perennou, Michael Marineau, Michael Olbrich, Miklos Vajna,
471 Patrik Flykt, poma, Sebastian Thorarensen, Thomas Bächler,
472 Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom Gundersen,
473 Umut Tezduyar Lindskog, Wieland Hoffmann, Zbigniew
476 -- Berlin, 2014-03-25
480 * A new unit file setting RestrictAddressFamilies= has been
481 added to restrict which socket address families unit
482 processes gain access to. This takes address family names
483 like "AF_INET" or "AF_UNIX", and is useful to minimize the
484 attack surface of services via exotic protocol stacks. This
485 is built on seccomp system call filters.
487 * Two new unit file settings RuntimeDirectory= and
488 RuntimeDirectoryMode= have been added that may be used to
489 manage a per-daemon runtime directories below /run. This is
490 an alternative for setting up directory permissions with
491 tmpfiles snippets, and has the advantage that the runtime
492 directory's lifetime is bound to the daemon runtime and that
493 the daemon starts up with an empty directory each time. This
494 is particularly useful when writing services that drop
495 privileges using the User= or Group= setting.
497 * The DeviceAllow= unit setting now supports globbing for
498 matching against device group names.
500 * The systemd configuration file system.conf gained new
501 settings DefaultCPUAccounting=, DefaultBlockIOAccounting=,
502 DefaultMemoryAccounting= to globally turn on/off accounting
503 for specific resources (cgroups) for all units. These
504 settings may still be overridden individually in each unit
507 * systemd-gpt-auto-generator is now able to discover /srv and
508 root partitions in addition to /home and swap partitions. It
509 also supports LUKS-encrypted partitions now. With this in
510 place, automatic discovery of partitions to mount following
511 the Discoverable Partitions Specification
512 (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec)
513 is now a lot more complete. This allows booting without
514 /etc/fstab and without root= on the kernel command line on
515 systems prepared appropriately.
517 * systemd-nspawn gained a new --image= switch which allows
518 booting up disk images and Linux installations on any block
519 device that follow the Discoverable Partitions Specification
520 (see above). This means that installations made with
521 appropriately updated installers may now be started and
522 deployed using container managers, completely
523 unmodified. (We hope that libvirt-lxc will add support for
524 this feature soon, too.)
526 * systemd-nspawn gained a new --network-macvlan= setting to
527 set up a private macvlan interface for the
528 container. Similarly, systemd-networkd gained a new
529 Kind=macvlan setting in .netdev files.
531 * systemd-networkd now supports configuring local addresses
534 * A new tool systemd-network-wait-online has been added to
535 synchronously wait for network connectivity using
538 * The sd-bus.h bus API gained a new sd_bus_track object for
539 tracking the life-cycle of bus peers. Note that sd-bus.h is
540 still not a public API though (unless you specify
541 --enable-kdbus on the configure command line, which however
542 voids your warranty and you get no API stability guarantee).
544 * The $XDG_RUNTIME_DIR runtime directories for each user are
545 now individual tmpfs instances, which has the benefit of
546 introducing separate pools for each user, with individual
547 size limits, and thus making sure that unprivileged clients
548 can no longer negatively impact the system or other users by
549 filling up their $XDG_RUNTIME_DIR. A new logind.conf setting
550 RuntimeDirectorySize= has been introduced that allows
551 controlling the default size limit for all users. It
552 defaults to 10% of the available physical memory. This is no
553 replacement for quotas on tmpfs though (which the kernel
554 still does not support), as /dev/shm and /tmp are still
555 shared resources used by both the system and unprivileged
558 * logind will now automatically turn off automatic suspending
559 on laptop lid close when more than one display is
560 connected. This was previously expected to be implemented
561 individually in desktop environments (such as GNOME),
562 however has been added to logind now, in order to fix a
563 boot-time race where a desktop environment might not have
564 been started yet and thus not been able to take an inhibitor
565 lock at the time where logind already suspends the system
568 * logind will now wait at least 30s after each system
569 suspend/resume cycle, and 3min after system boot before
570 suspending the system due to a closed laptop lid. This
571 should give USB docking stations and similar enough time to
572 be probed and configured after system resume and boot in
573 order to then act as suspend blocker.
575 * systemd-run gained a new --property= setting which allows
576 initialization of resource control properties (and others)
577 for the created scope or service unit. Example: "systemd-run
578 --property=BlockIOWeight=10 updatedb" may be used to run
579 updatedb at a low block IO scheduling weight.
581 * systemd-run's --uid=, --gid=, --setenv=, --setenv= switches
582 now also work in --scope mode.
584 * When systemd is compiled with kdbus support, basic support
585 for enforced policies is now in place. (Note that enabling
586 kdbus still voids your warranty and no API compatibility
589 Contributions from: Andrey Borzenkov, Ansgar Burchardt, Armin
590 K., Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
591 Harald Hoyer, Henrik Grindal Bakken, Jasper St. Pierre, Kay
592 Sievers, Kieran Clancy, Lennart Poettering, Lukas Nykryn,
593 Mantas Mikulėnas, Marcel Holtmann, Mark Oteiza, Martin Pitt,
594 Mike Gilbert, Peter Rajnoha, poma, Samuli Suominen, Stef
595 Walter, Susant Sahani, Tero Roponen, Thomas Andersen, Thomas
596 Bächler, Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom
597 Gundersen, Umut Tezduyar Lindskog, Uoti Urpala, Zachary Cook,
598 Zbigniew Jędrzejewski-Szmek
600 -- Berlin, 2014-03-12
604 * systemd will now relabel /dev after loading the SMACK policy
605 according to SMACK rules.
607 * A new unit file option AppArmorProfile= has been added to
608 set the AppArmor profile for the processes of a unit.
610 * A new condition check ConditionArchitecture= has been added
611 to conditionalize units based on the system architecture, as
612 reported by uname()'s "machine" field.
614 * systemd-networkd now supports matching on the system
615 virtualization, architecture, kernel command line, host name
618 * logind is now a lot more aggressive when suspending the
619 machine due to a closed laptop lid. Instead of acting only
620 on the lid close action, it will continuously watch the lid
621 status and act on it. This is useful for laptops where the
622 power button is on the outside of the chassis so that it can
623 be reached without opening the lid (such as the Lenovo
624 Yoga). On those machines, logind will now immediately
625 re-suspend the machine if the power button has been
626 accidentally pressed while the laptop was suspended and in a
629 * logind will now watch SW_DOCK switches and inhibit reaction
630 to the lid switch if it is pressed. This means that logind
631 will not suspend the machine anymore if the lid is closed
632 and the system is docked, if the laptop supports SW_DOCK
633 notifications via the input layer. Note that ACPI docking
634 stations do not generate this currently. Also note that this
635 logic is usually not fully sufficient and Desktop
636 Environments should take a lid switch inhibitor lock when an
637 external display is connected, as systemd will not watch
640 * nspawn will now make use of the devices cgroup controller by
641 default, and only permit creation of and access to the usual
642 API device nodes like /dev/null or /dev/random, as well as
643 access to (but not creation of) the pty devices.
645 * We will now ship a default .network file for
646 systemd-networkd that automatically configures DHCP for
647 network interfaces created by nspawn's --network-veth or
648 --network-bridge= switches.
650 * systemd will now understand the usual M, K, G, T suffixes
651 according to SI conventions (i.e. to the base 1000) when
652 referring to throughput and hardware metrics. It will stay
653 with IEC conventions (i.e. to the base 1024) for software
654 metrics, according to what is customary according to
655 Wikipedia. We explicitly document which base applies for
656 each configuration option.
658 * The DeviceAllow= setting in unit files now supports a syntax
659 to whitelist an entire group of devices node majors at once,
660 based on the /proc/devices listing. For example, with the
661 string "char-pts", it is now possible to whitelist all
662 current and future pseudo-TTYs at once.
664 * sd-event learned a new "post" event source. Event sources of
665 this type are triggered by the dispatching of any event
666 source of a type that is not "post". This is useful for
667 implementing clean-up and check event sources that are
668 triggered by other work being done in the program.
670 * systemd-networkd is no longer statically enabled, but uses
671 the usual [Install] sections so that it can be
672 enabled/disabled using systemctl. It still is enabled by
675 * When creating a veth interface pair with systemd-nspawn, the
676 host side will now be prefixed with "vb-" if
677 --network-bridge= is used, and with "ve-" if --network-veth
678 is used. This way, it is easy to distinguish these cases on
679 the host, for example to apply different configuration to
680 them with systemd-networkd.
682 * The compatibility libraries for libsystemd-journal.so,
683 libsystem-id128.so, libsystemd-login.so and
684 libsystemd-daemon.so do not make use of IFUNC
685 anymore. Instead, we now build libsystemd.so multiple times
686 under these alternative names. This means that the footprint
687 is drastically increased, but given that these are
688 transitional compatibility libraries, this should not matter
689 much. This change has been made necessary to support the ARM
690 platform for these compatibility libraries, as the ARM
691 toolchain is not really at the same level as the toolchain
692 for other architectures like x86 and does not support
693 IFUNC. Please make sure to use --enable-compat-libs only
694 during a transitional period!
696 Contributions from: Andreas Fuchs, Armin K., Colin Walters,
697 Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
698 Holger Schurig, Jason A. Donenfeld, Jason St. John, Jasper
699 St. Pierre, Kay Sievers, Lennart Poettering, Łukasz Stelmach,
700 Marcel Holtmann, Michael Scherer, Michal Sekletar, Mike
701 Gilbert, Samuli Suominen, Thomas Bächler, Thomas Hindoe
702 Paaboel Andersen, Tom Gundersen, Umut Tezduyar Lindskog,
703 Zbigniew Jędrzejewski-Szmek
705 -- Berlin, 2014-02-24
709 * A new component "systemd-networkd" has been added that can
710 be used to configure local network interfaces statically or
711 via DHCP. It is capable of bringing up bridges, VLANs, and
712 bonding. Currently, no hook-ups for interactive network
713 configuration are provided. Use this for your initrd,
714 container, embedded, or server setup if you need a simple,
715 yet powerful, network configuration solution. This
716 configuration subsystem is quite nifty, as it allows wildcard
717 hotplug matching in interfaces. For example, with a single
718 configuration snippet, you can configure that all Ethernet
719 interfaces showing up are automatically added to a bridge,
720 or similar. It supports link-sensing and more.
722 * A new tool "systemd-socket-proxyd" has been added which can
723 act as a bidirectional proxy for TCP sockets. This is
724 useful for adding socket activation support to services that
725 do not actually support socket activation, including virtual
726 machines and the like.
728 * Add a new tool to save/restore rfkill state on
731 * Save/restore state of keyboard backlights in addition to
732 display backlights on shutdown/boot.
734 * udev learned a new SECLABEL{} construct to label device
735 nodes with a specific security label when they appear. For
736 now, only SECLABEL{selinux} is supported, but the syntax is
737 prepared for additional security frameworks.
739 * udev gained a new scheme to configure link-level attributes
740 from files in /etc/systemd/network/*.link. These files can
741 match against MAC address, device path, driver name and type,
742 and will apply attributes like the naming policy, link speed,
743 MTU, duplex settings, Wake-on-LAN settings, MAC address, MAC
744 address assignment policy (randomized, ...).
746 * The configuration of network interface naming rules for
747 "permanent interface names" has changed: a new NamePolicy=
748 setting in the [Link] section of .link files determines the
749 priority of possible naming schemes (onboard, slot, mac,
750 path). The default value of this setting is determined by
751 /usr/lib/net/links/99-default.link. Old
752 80-net-name-slot.rules udev configuration file has been
753 removed, so local configuration overriding this file should
754 be adapated to override 99-default.link instead.
756 * When the User= switch is used in a unit file, also
757 initialize $SHELL= based on the user database entry.
759 * systemd no longer depends on libdbus. All communication is
760 now done with sd-bus, systemd's low-level bus library
763 * kdbus support has been added to PID 1 itself. When kdbus is
764 enabled, this causes PID 1 to set up the system bus and
765 enable support for a new ".busname" unit type that
766 encapsulates bus name activation on kdbus. It works a little
767 bit like ".socket" units, except for bus names. A new
768 generator has been added that converts classic dbus1 service
769 activation files automatically into native systemd .busname
772 * sd-bus: add a light-weight vtable implementation that allows
773 defining objects on the bus with a simple static const
774 vtable array of its methods, signals and properties.
776 * systemd will not generate or install static dbus
777 introspection data anymore to /usr/share/dbus-1/interfaces,
778 as the precise format of these files is unclear, and
779 nothing makes use of it.
781 * A proxy daemon is now provided to proxy clients connecting
782 via classic D-Bus AF_UNIX sockets to kdbus, to provide full
783 compatibility with classic D-Bus.
785 * A bus driver implementation has been added that supports the
786 classic D-Bus bus driver calls on kdbus, also for
787 compatibility purposes.
789 * A new API "sd-event.h" has been added that implements a
790 minimal event loop API built around epoll. It provides a
791 couple of features that direct epoll usage is lacking:
792 prioritization of events, scales to large numbers of timer
793 events, per-event timer slack (accuracy), system-wide
794 coalescing of timer events, exit handlers, watchdog
795 supervision support using systemd's sd_notify() API, child
798 * A new API "sd-rntl.h" has been added that provides an API
799 around the route netlink interface of the kernel, similar in
802 * A new API "sd-dhcp-client.h" has been added that provides a
803 small DHCPv4 client-side implementation. This is used by
806 * There is a new kernel command line option
807 "systemd.restore_state=0|1". When set to "0", none of the
808 systemd tools will restore saved runtime state to hardware
809 devices. More specifically, the rfkill and backlight states
812 * The FsckPassNo= compatibility option in mount/service units
813 has been removed. The fstab generator will now add the
814 necessary dependencies automatically, and does not require
815 PID1's support for that anymore.
817 * journalctl gained a new switch, --list-boots, that lists
818 recent boots with their times and boot IDs.
820 * The various tools like systemctl, loginctl, timedatectl,
821 busctl, systemd-run, ... have gained a new switch "-M" to
822 connect to a specific, local OS container (as direct
823 connection, without requiring SSH). This works on any
824 container that is registered with machined, such as those
825 created by libvirt-lxc or nspawn.
827 * systemd-run and systemd-analyze also gained support for "-H"
828 to connect to remote hosts via SSH. This is particularly
829 useful for systemd-run because it enables queuing of jobs
832 * machinectl gained a new command "login" to open a getty
833 login in any local container. This works with any container
834 that is registered with machined (such as those created by
835 libvirt-lxc or nspawn), and which runs systemd inside.
837 * machinectl gained a new "reboot" command that may be used to
838 trigger a reboot on a specific container that is registered
839 with machined. This works on any container that runs an init
842 * systemctl gained a new "list-timers" command to print a nice
843 listing of installed timer units with the times they elapse
846 * Alternative reboot() parameters may now be specified on the
847 "systemctl reboot" command line and are passed to the
848 reboot() system call.
850 * systemctl gained a new --job-mode= switch to configure the
851 mode to queue a job with. This is a more generic version of
852 --fail, --irreversible, and --ignore-dependencies, which are
853 still available but not advertised anymore.
855 * /etc/systemd/system.conf gained new settings to configure
856 various default timeouts of units, as well as the default
857 start limit interval and burst. These may still be overridden
860 * PID1 will now export on the bus profile data of the security
861 policy upload process (such as the SELinux policy upload to
864 * journald: when forwarding logs to the console, include
865 timestamps (following the setting in
866 /sys/module/printk/parameters/time).
868 * OnCalendar= in timer units now understands the special
869 strings "yearly" and "annually". (Both are equivalent)
871 * The accuracy of timer units is now configurable with the new
872 AccuracySec= setting. It defaults to 1min.
874 * A new dependency type JoinsNamespaceOf= has been added that
875 allows running two services within the same /tmp and network
876 namespace, if PrivateNetwork= or PrivateTmp= are used.
878 * A new command "cat" has been added to systemctl. It outputs
879 the original unit file of a unit, and concatenates the
880 contents of additional "drop-in" unit file snippets, so that
881 the full configuration is shown.
883 * systemctl now supports globbing on the various "list-xyz"
884 commands, like "list-units" or "list-sockets", as well as on
885 those commands which take multiple unit names.
887 * journalctl's --unit= switch gained support for globbing.
889 * All systemd daemons now make use of the watchdog logic so
890 that systemd automatically notices when they hang.
892 * If the $container_ttys environment variable is set,
893 getty-generator will automatically spawn a getty for each
894 listed tty. This is useful for container managers to request
895 login gettys to be spawned on as many ttys as needed.
897 * %h, %s, %U specifier support is not available anymore when
898 used in unit files for PID 1. This is because NSS calls are
899 not safe from PID 1. They stay available for --user
900 instances of systemd, and as special case for the root user.
902 * loginctl gained a new "--no-legend" switch to turn off output
905 * The "sd-login.h" API gained three new calls:
906 sd_session_is_remote(), sd_session_get_remote_user(),
907 sd_session_get_remote_host() to query information about
910 * The udev hardware database now also carries vendor/product
911 information of SDIO devices.
913 * The "sd-daemon.h" API gained a new sd_watchdog_enabled() to
914 determine whether watchdog notifications are requested by
917 * Socket-activated per-connection services now include a
918 short description of the connection parameters in the
921 * tmpfiles gained a new "--boot" option. When this is not used,
922 only lines where the command character is not suffixed with
923 "!" are executed. When this option is specified, those
924 options are executed too. This partitions tmpfiles
925 directives into those that can be safely executed at any
926 time, and those which should be run only at boot (for
927 example, a line that creates /run/nologin).
929 * A new API "sd-resolve.h" has been added which provides a simple
930 asynchronous wrapper around glibc NSS host name resolution
931 calls, such as getaddrinfo(). In contrast to glibc's
932 getaddrinfo_a(), it does not use signals. In contrast to most
933 other asynchronous name resolution libraries, this one does
934 not reimplement DNS, but reuses NSS, so that alternate
935 host name resolution systems continue to work, such as mDNS,
936 LDAP, etc. This API is based on libasyncns, but it has been
937 cleaned up for inclusion in systemd.
939 * The APIs "sd-journal.h", "sd-login.h", "sd-id128.h",
940 "sd-daemon.h" are no longer found in individual libraries
941 libsystemd-journal.so, libsystemd-login.so,
942 libsystemd-id128.so, libsystemd-daemon.so. Instead, we have
943 merged them into a single library, libsystemd.so, which
944 provides all symbols. The reason for this is cyclic
945 dependencies, as these libraries tend to use each other's
946 symbols. So far, we have managed to workaround that by linking
947 a copy of a good part of our code into each of these
948 libraries again and again, which, however, makes certain
949 things hard to do, like sharing static variables. Also, it
950 substantially increases footprint. With this change, there
951 is only one library for the basic APIs systemd
952 provides. Also, "sd-bus.h", "sd-memfd.h", "sd-event.h",
953 "sd-rtnl.h", "sd-resolve.h", "sd-utf8.h" are found in this
954 library as well, however are subject to the --enable-kdbus
955 switch (see below). Note that "sd-dhcp-client.h" is not part
956 of this library (this is because it only consumes, never
957 provides, services of/to other APIs). To make the transition
958 easy from the separate libraries to the unified one, we
959 provide the --enable-compat-libs compile-time switch which
960 will generate stub libraries that are compatible with the
961 old ones but redirect all calls to the new one.
963 * All of the kdbus logic and the new APIs "sd-bus.h",
964 "sd-memfd.h", "sd-event.h", "sd-rtnl.h", "sd-resolve.h",
965 and "sd-utf8.h" are compile-time optional via the
966 "--enable-kdbus" switch, and they are not compiled in by
967 default. To make use of kdbus, you have to explicitly enable
968 the switch. Note however, that neither the kernel nor the
969 userspace API for all of this is considered stable yet. We
970 want to maintain the freedom to still change the APIs for
971 now. By specifying this build-time switch, you acknowledge
972 that you are aware of the instability of the current
975 * Also, note that while kdbus is pretty much complete,
976 it lacks one thing: proper policy support. This means you
977 can build a fully working system with all features; however,
978 it will be highly insecure. Policy support will be added in
979 one of the next releases, at the same time that we will
980 declare the APIs stable.
982 * When the kernel command-line argument "kdbus" is specified,
983 systemd will automatically load the kdbus.ko kernel module. At
984 this stage of development, it is only useful for testing kdbus
985 and should not be used in production. Note: if "--enable-kdbus"
986 is specified, and the kdbus.ko kernel module is available, and
987 "kdbus" is added to the kernel command line, the entire system
988 runs with kdbus instead of dbus-daemon, with the above mentioned
989 problem of missing the system policy enforcement. Also a future
990 version of kdbus.ko or a newer systemd will not be compatible with
991 each other, and will unlikely be able to boot the machine if only
992 one of them is updated.
994 * systemctl gained a new "import-environment" command which
995 uploads the caller's environment (or parts thereof) into the
996 service manager so that it is inherited by services started
997 by the manager. This is useful to upload variables like
998 $DISPLAY into the user service manager.
1000 * A new PrivateDevices= switch has been added to service units
1001 which allows running a service with a namespaced /dev
1002 directory that does not contain any device nodes for
1003 physical devices. More specifically, it only includes devices
1004 such as /dev/null, /dev/urandom, and /dev/zero which are API
1007 * logind has been extended to support behaviour like VT
1008 switching on seats that do not support a VT. This makes
1009 multi-session available on seats that are not the first seat
1010 (seat0), and on systems where kernel support for VTs has
1011 been disabled at compile-time.
1013 * If a process holds a delay lock for system sleep or shutdown
1014 and fails to release it in time, we will now log its
1015 identity. This makes it easier to identify processes that
1016 cause slow suspends or power-offs.
1018 * When parsing /etc/crypttab, support for a new key-slot=
1019 option as supported by Debian is added. It allows indicating
1020 which LUKS slot to use on disk, speeding up key loading.
1022 * The sd_journald_sendv() API call has been checked and
1023 officially declared to be async-signal-safe so that it may
1024 be invoked from signal handlers for logging purposes.
1026 * Boot-time status output is now enabled automatically after a
1027 short timeout if boot does not progress, in order to give
1028 the user an indication what she or he is waiting for.
1030 * The boot-time output has been improved to show how much time
1031 remains until jobs expire.
1033 * The KillMode= switch in service units gained a new possible
1034 value "mixed". If set, and the unit is shut down, then the
1035 initial SIGTERM signal is sent only to the main daemon
1036 process, while the following SIGKILL signal is sent to
1037 all remaining processes of the service.
1039 * When a scope unit is registered, a new property "Controller"
1040 may be set. If set to a valid bus name, systemd will send a
1041 RequestStop() signal to this name when it would like to shut
1042 down the scope. This may be used to hook manager logic into
1043 the shutdown logic of scope units. Also, scope units may now
1044 be put in a special "abandoned" state, in which case the
1045 manager process which created them takes no further
1046 responsibilities for it.
1048 * When reading unit files, systemd will now verify
1049 the access mode of these files, and warn about certain
1050 suspicious combinations. This has been added to make it
1051 easier to track down packaging bugs where unit files are
1052 marked executable or world-writable.
1054 * systemd-nspawn gained a new "--setenv=" switch to set
1055 container-wide environment variables. The similar option in
1056 systemd-activate was renamed from "--environment=" to
1057 "--setenv=" for consistency.
1059 * systemd-nspawn has been updated to create a new kdbus domain
1060 for each container that is invoked, thus allowing each
1061 container to have its own set of system and user buses,
1062 independent of the host.
1064 * systemd-nspawn gained a new --drop-capability= switch to run
1065 the container with less capabilities than the default. Both
1066 --drop-capability= and --capability= now take the special
1067 string "all" for dropping or keeping all capabilities.
1069 * systemd-nspawn gained new switches for executing containers
1070 with specific SELinux labels set.
1072 * systemd-nspawn gained a new --quiet switch to not generate
1073 any additional output but the container's own console
1076 * systemd-nspawn gained a new --share-system switch to run a
1077 container without PID namespacing enabled.
1079 * systemd-nspawn gained a new --register= switch to control
1080 whether the container is registered with systemd-machined or
1081 not. This is useful for containers that do not run full
1082 OS images, but only specific apps.
1084 * systemd-nspawn gained a new --keep-unit which may be used
1085 when invoked as the only program from a service unit, and
1086 results in registration of the unit service itself in
1087 systemd-machined, instead of a newly opened scope unit.
1089 * systemd-nspawn gained a new --network-interface= switch for
1090 moving arbitrary interfaces to the container. The new
1091 --network-veth switch creates a virtual Ethernet connection
1092 between host and container. The new --network-bridge=
1093 switch then allows assigning the host side of this virtual
1094 Ethernet connection to a bridge device.
1096 * systemd-nspawn gained a new --personality= switch for
1097 setting the kernel personality for the container. This is
1098 useful when running a 32-bit container on a 64-bit host. A
1099 similar option Personality= is now also available for service
1102 * logind will now also track a "Desktop" identifier for each
1103 session which encodes the desktop environment of it. This is
1104 useful for desktop environments that want to identify
1105 multiple running sessions of itself easily.
1107 * A new SELinuxContext= setting for service units has been
1108 added that allows setting a specific SELinux execution
1109 context for a service.
1111 * Most systemd client tools will now honour $SYSTEMD_LESS for
1112 settings of the "less" pager. By default, these tools will
1113 override $LESS to allow certain operations to work, such as
1114 jump-to-the-end. With $SYSTEMD_LESS, it is possible to
1115 influence this logic.
1117 * systemd's "seccomp" hook-up has been changed to make use of
1118 the libseccomp library instead of using its own
1119 implementation. This has benefits for portability among
1122 * For usage together with SystemCallFilter=, a new
1123 SystemCallErrorNumber= setting has been introduced that
1124 allows configuration of a system error number to be returned
1125 on filtered system calls, instead of immediately killing the
1126 process. Also, SystemCallArchitectures= has been added to
1127 limit access to system calls of a particular architecture
1128 (in order to turn off support for unused secondary
1129 architectures). There is also a global
1130 SystemCallArchitectures= setting in system.conf now to turn
1131 off support for non-native system calls system-wide.
1133 * systemd requires a kernel with a working name_to_handle_at(),
1134 please see the kernel config requirements in the README file.
1136 Contributions from: Adam Williamson, Alex Jia, Anatol Pomozov,
1137 Ansgar Burchardt, AppleBloom, Auke Kok, Bastien Nocera,
1138 Chengwei Yang, Christian Seiler, Colin Guthrie, Colin Walters,
1139 Cristian Rodríguez, Daniel Buch, Daniele Medri, Daniel J
1140 Walsh, Daniel Mack, Dan McGee, Dave Reisner, David Coppa,
1141 David Herrmann, David Strauss, Djalal Harouni, Dmitry Pisklov,
1142 Elia Pinto, Florian Weimer, George McCollister, Goffredo
1143 Baroncelli, Greg Kroah-Hartman, Hendrik Brueckner, Igor
1144 Zhbanov, Jan Engelhardt, Jan Janssen, Jason A. Donenfeld,
1145 Jason St. John, Jasper St. Pierre, Jóhann B. Guðmundsson, Jose
1146 Ignacio Naranjo, Karel Zak, Kay Sievers, Kristian Høgsberg,
1147 Lennart Poettering, Lubomir Rintel, Lukas Nykryn, Lukasz
1148 Skalski, Łukasz Stelmach, Luke Shumaker, Mantas Mikulėnas,
1149 Marc-Antoine Perennou, Marcel Holtmann, Marcos Felipe Rasia de
1150 Mello, Marko Myllynen, Martin Pitt, Matthew Monaco, Michael
1151 Marineau, Michael Scherer, Michał Górny, Michal Sekletar,
1152 Michele Curti, Oleksii Shevchuk, Olivier Brunel, Patrik Flykt,
1153 Pavel Holica, Raudi, Richard Marko, Ronny Chevalier, Sébastien
1154 Luttringer, Sergey Ptashnick, Shawn Landden, Simon Peeters,
1155 Stefan Beller, Susant Sahani, Sylvain Plantefeve, Sylvia Else,
1156 Tero Roponen, Thomas Bächler, Thomas Hindoe Paaboel Andersen,
1157 Tom Gundersen, Umut Tezduyar Lindskog, Unai Uribarri, Václav
1158 Pavlín, Vincent Batts, WaLyong Cho, William Giokas, Yang
1159 Zhiyong, Yin Kangkai, Yuxuan Shui, Zbigniew Jędrzejewski-Szmek
1161 -- Berlin, 2014-02-20
1165 * logind has gained support for facilitating privileged input
1166 and drm device access for unprivileged clients. This work is
1167 useful to allow Wayland display servers (and similar
1168 programs, such as kmscon) to run under the user's ID and
1169 access input and drm devices which are normally
1170 protected. When this is used (and the kernel is new enough)
1171 logind will "mute" IO on the file descriptors passed to
1172 Wayland as long as it is in the background and "unmute" it
1173 if it returns into the foreground. This allows secure
1174 session switching without allowing background sessions to
1175 eavesdrop on input and display data. This also introduces
1176 session switching support if VT support is turned off in the
1177 kernel, and on seats that are not seat0.
1179 * A new kernel command line option luks.options= is understood
1180 now which allows specifiying LUKS options for usage for LUKS
1181 encrypted partitions specified with luks.uuid=.
1183 * tmpfiles.d(5) snippets may now use specifier expansion in
1184 path names. More specifically %m, %b, %H, %v, are now
1185 replaced by the local machine id, boot id, hostname, and
1186 kernel version number.
1188 * A new tmpfiles.d(5) command "m" has been introduced which
1189 may be used to change the owner/group/access mode of a file
1190 or directory if it exists, but do nothing if it does not.
1192 * This release removes high-level support for the
1193 MemorySoftLimit= cgroup setting. The underlying kernel
1194 cgroup attribute memory.soft_limit= is currently badly
1195 designed and likely to be removed from the kernel API in its
1196 current form, hence we should not expose it for now.
1198 * The memory.use_hierarchy cgroup attribute is now enabled for
1199 all cgroups systemd creates in the memory cgroup
1200 hierarchy. This option is likely to be come the built-in
1201 default in the kernel anyway, and the non-hierarchial mode
1202 never made much sense in the intrinsically hierarchial
1205 * A new field _SYSTEMD_SLICE= is logged along with all journal
1206 messages containing the slice a message was generated
1207 from. This is useful to allow easy per-customer filtering of
1208 logs among other things.
1210 * systemd-journald will no longer adjust the group of journal
1211 files it creates to the "systemd-journal" group. Instead we
1212 rely on the journal directory to be owned by the
1213 "systemd-journal" group, and its setgid bit set, so that the
1214 kernel file system layer will automatically enforce that
1215 journal files inherit this group assignment. The reason for
1216 this change is that we cannot allow NSS look-ups from
1217 journald which would be necessary to resolve
1218 "systemd-journal" to a numeric GID, because this might
1219 create deadlocks if NSS involves synchronous queries to
1220 other daemons (such as nscd, or sssd) which in turn are
1221 logging clients of journald and might block on it, which
1222 would then dead lock. A tmpfiles.d(5) snippet included in
1223 systemd will make sure the setgid bit and group are
1224 properly set on the journal directory if it exists on every
1225 boot. However, we recommend adjusting it manually after
1226 upgrades too (or from RPM scriptlets), so that the change is
1227 not delayed until next reboot.
1229 * Backlight and random seed files in /var/lib/ have moved into
1230 the /var/lib/systemd/ directory, in order to centralize all
1231 systemd generated files in one directory.
1233 * Boot time performance measurements (as displayed by
1234 "systemd-analyze" for example) will now read ACPI 5.0 FPDT
1235 performance information if that's available to determine how
1236 much time BIOS and boot loader initialization required. With
1237 a sufficiently new BIOS you hence no longer need to boot
1238 with Gummiboot to get access to such information.
1240 Contributions from: Andrey Borzenkov, Chen Jie, Colin Walters,
1241 Cristian Rodríguez, Dave Reisner, David Herrmann, David
1242 Mackey, David Strauss, Eelco Dolstra, Evan Callicoat, Gao
1243 feng, Harald Hoyer, Jimmie Tauriainen, Kay Sievers, Lennart
1244 Poettering, Lukas Nykryn, Mantas Mikulėnas, Martin Pitt,
1245 Michael Scherer, Michał Górny, Mike Gilbert, Patrick McCarty,
1246 Sebastian Ott, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
1248 -- Berlin, 2013-10-02
1252 * The Restart= option for services now understands a new
1253 on-watchdog setting, which will restart the service
1254 automatically if the service stops sending out watchdog keep
1255 alive messages (as configured with WatchdogSec=).
1257 * The getty generator (which is responsible for bringing up a
1258 getty on configured serial consoles) will no longer only
1259 start a getty on the primary kernel console but on all
1260 others, too. This makes the order in which console= is
1261 specified on the kernel command line less important.
1263 * libsystemd-logind gained a new sd_session_get_vt() call to
1264 retrieve the VT number of a session.
1266 * If the option "tries=0" is set for an entry of /etc/crypttab
1267 its passphrase is queried indefinitely instead of any
1268 maximum number of tries.
1270 * If a service with a configure PID file terminates its PID
1271 file will now be removed automatically if it still exists
1272 afterwards. This should put an end to stale PID files.
1274 * systemd-run will now also take relative binary path names
1275 for execution and no longer insists on absolute paths.
1277 * InaccessibleDirectories= and ReadOnlyDirectories= now take
1278 paths that are optionally prefixed with "-" to indicate that
1279 it should not be considered a failure if they do not exist.
1281 * journalctl -o (and similar commands) now understands a new
1282 output mode "short-precise", it is similar to "short" but
1283 shows timestamps with usec accuracy.
1285 * The option "discard" (as known from Debian) is now
1286 synonymous to "allow-discards" in /etc/crypttab. In fact,
1287 "discard" is preferred now (since it is easier to remember
1290 * Some licensing clean-ups were made, so that more code is now
1291 LGPL-2.1 licensed than before.
1293 * A minimal tool to save/restore the display backlight
1294 brightness across reboots has been added. It will store the
1295 backlight setting as late as possible at shutdown, and
1296 restore it as early as possible during reboot.
1298 * A logic to automatically discover and enable home and swap
1299 partitions on GPT disks has been added. With this in place
1300 /etc/fstab becomes optional for many setups as systemd can
1301 discover certain partitions located on the root disk
1302 automatically. Home partitions are recognized under their
1303 GPT type ID 933ac7e12eb44f13b8440e14e2aef915. Swap
1304 partitions are recognized under their GPT type ID
1305 0657fd6da4ab43c484e50933c84b4f4f.
1307 * systemd will no longer pass any environment from the kernel
1308 or initrd to system services. If you want to set an
1309 environment for all services, do so via the kernel command
1310 line systemd.setenv= assignment.
1312 * The systemd-sysctl tool no longer natively reads the file
1313 /etc/sysctl.conf. If desired, the file should be symlinked
1314 from /etc/sysctl.d/99-sysctl.conf. Apart from providing
1315 legacy support by a symlink rather than built-in code, it
1316 also makes the otherwise hidden order of application of the
1317 different files visible. (Note that this partly reverts to a
1318 pre-198 application order of sysctl knobs!)
1320 * The "systemctl set-log-level" and "systemctl dump" commands
1321 have been moved to systemd-analyze.
1323 * systemd-run learned the new --remain-after-exit switch,
1324 which causes the scope unit not to be cleaned up
1325 automatically after the process terminated.
1327 * tmpfiles learned a new --exclude-prefix= switch to exclude
1328 certain paths from operation.
1330 * journald will now automatically flush all messages to disk
1331 as soon as a message of the log priorities CRIT, ALERT or
1334 Contributions from: Andrew Cook, Brandon Philips, Christian
1335 Hesse, Christoph Junghans, Colin Walters, Daniel Schaal,
1336 Daniel Wallace, Dave Reisner, David Herrmann, Gao feng, George
1337 McCollister, Giovanni Campagna, Hannes Reinecke, Harald Hoyer,
1338 Herczeg Zsolt, Holger Hans Peter Freyther, Jan Engelhardt,
1339 Jesper Larsen, Kay Sievers, Khem Raj, Lennart Poettering,
1340 Lukas Nykryn, Maciej Wereski, Mantas Mikulėnas, Marcel
1341 Holtmann, Martin Pitt, Michael Biebl, Michael Marineau,
1342 Michael Scherer, Michael Stapelberg, Michal Sekletar, Michał
1343 Górny, Olivier Brunel, Ondrej Balaz, Ronny Chevalier, Shawn
1344 Landden, Steven Hiscocks, Thomas Bächler, Thomas Hindoe
1345 Paaboel Andersen, Tom Gundersen, Umut Tezduyar, WANG Chao,
1346 William Giokas, Zbigniew Jędrzejewski-Szmek
1348 -- Berlin, 2013-09-13
1352 * The documentation has been updated to cover the various new
1353 concepts introduced with 205.
1355 * Unit files now understand the new %v specifier which
1356 resolves to the kernel version string as returned by "uname
1359 * systemctl now supports filtering the unit list output by
1360 load state, active state and sub state, using the new
1363 * "systemctl status" will now show the results of the
1364 condition checks (like ConditionPathExists= and similar) of
1365 the last start attempts of the unit. They are also logged to
1368 * "journalctl -b" may now be used to look for boot output of a
1369 specific boot. Try "journalctl -b -1" for the previous boot,
1370 but the syntax is substantially more powerful.
1372 * "journalctl --show-cursor" has been added which prints the
1373 cursor string the last shown log line. This may then be used
1374 with the new "journalctl --after-cursor=" switch to continue
1375 browsing logs from that point on.
1377 * "journalctl --force" may now be used to force regeneration
1380 * Creation of "dead" device nodes has been moved from udev
1381 into kmod and tmpfiles. Previously, udev would read the kmod
1382 databases to pre-generate dead device nodes based on meta
1383 information contained in kernel modules, so that these would
1384 be auto-loaded on access rather then at boot. As this
1385 does not really have much to do with the exposing actual
1386 kernel devices to userspace this has always been slightly
1387 alien in the udev codebase. Following the new scheme kmod
1388 will now generate a runtime snippet for tmpfiles from the
1389 module meta information and it now is tmpfiles' job to the
1390 create the nodes. This also allows overriding access and
1391 other parameters for the nodes using the usual tmpfiles
1392 facilities. As side effect this allows us to remove the
1393 CAP_SYS_MKNOD capability bit from udevd entirely.
1395 * logind's device ACLs may now be applied to these "dead"
1396 devices nodes too, thus finally allowing managed access to
1397 devices such as /dev/snd/sequencer whithout loading the
1398 backing module right-away.
1400 * A new RPM macro has been added that may be used to apply
1401 tmpfiles configuration during package installation.
1403 * systemd-detect-virt and ConditionVirtualization= now can
1404 detect User-Mode-Linux machines (UML).
1406 * journald will now implicitly log the effective capabilities
1407 set of processes in the message metadata.
1409 * systemd-cryptsetup has gained support for TrueCrypt volumes.
1411 * The initrd interface has been simplified (more specifically,
1412 support for passing performance data via environment
1413 variables and fsck results via files in /run has been
1414 removed). These features were non-essential, and are
1415 nowadays available in a much nicer way by having systemd in
1416 the initrd serialize its state and have the hosts systemd
1417 deserialize it again.
1419 * The udev "keymap" data files and tools to apply keyboard
1420 specific mappings of scan to key codes, and force-release
1421 scan code lists have been entirely replaced by a udev
1422 "keyboard" builtin and a hwdb data file.
1424 * systemd will now honour the kernel's "quiet" command line
1425 argument also during late shutdown, resulting in a
1426 completely silent shutdown when used.
1428 * There's now an option to control the SO_REUSEPORT socket
1429 option in .socket units.
1431 * Instance units will now automatically get a per-template
1432 subslice of system.slice unless something else is explicitly
1433 configured. For example, instances of sshd@.service will now
1434 implicitly be placed in system-sshd.slice rather than
1435 system.slice as before.
1437 * Test coverage support may now be enabled at build time.
1439 Contributions from: Dave Reisner, Frederic Crozat, Harald
1440 Hoyer, Holger Hans Peter Freyther, Jan Engelhardt, Jan
1441 Janssen, Jason St. John, Jesper Larsen, Kay Sievers, Lennart
1442 Poettering, Lukas Nykryn, Maciej Wereski, Martin Pitt, Michael
1443 Olbrich, Ramkumar Ramachandra, Ross Lagerwall, Shawn Landden,
1444 Thomas H.P. Andersen, Tom Gundersen, Tomasz Torcz, William
1445 Giokas, Zbigniew Jędrzejewski-Szmek
1447 -- Berlin, 2013-07-23
1451 * Two new unit types have been introduced:
1453 Scope units are very similar to service units, however, are
1454 created out of pre-existing processes -- instead of PID 1
1455 forking off the processes. By using scope units it is
1456 possible for system services and applications to group their
1457 own child processes (worker processes) in a powerful way
1458 which then maybe used to organize them, or kill them
1459 together, or apply resource limits on them.
1461 Slice units may be used to partition system resources in an
1462 hierarchial fashion and then assign other units to them. By
1463 default there are now three slices: system.slice (for all
1464 system services), user.slice (for all user sessions),
1465 machine.slice (for VMs and containers).
1467 Slices and scopes have been introduced primarily in
1468 context of the work to move cgroup handling to a
1469 single-writer scheme, where only PID 1
1470 creates/removes/manages cgroups.
1472 * There's a new concept of "transient" units. In contrast to
1473 normal units these units are created via an API at runtime,
1474 not from configuration from disk. More specifically this
1475 means it is now possible to run arbitrary programs as
1476 independent services, with all execution parameters passed
1477 in via bus APIs rather than read from disk. Transient units
1478 make systemd substantially more dynamic then it ever was,
1479 and useful as a general batch manager.
1481 * logind has been updated to make use of scope and slice units
1482 for managing user sessions. As a user logs in he will get
1483 his own private slice unit, to which all sessions are added
1484 as scope units. We also added support for automatically
1485 adding an instance of user@.service for the user into the
1486 slice. Effectively logind will no longer create cgroup
1487 hierarchies on its own now, it will defer entirely to PID 1
1488 for this by means of scope, service and slice units. Since
1489 user sessions this way become entities managed by PID 1
1490 the output of "systemctl" is now a lot more comprehensive.
1492 * A new mini-daemon "systemd-machined" has been added which
1493 may be used by virtualization managers to register local
1494 VMs/containers. nspawn has been updated accordingly, and
1495 libvirt will be updated shortly. machined will collect a bit
1496 of meta information about the VMs/containers, and assign
1497 them their own scope unit (see above). The collected
1498 meta-data is then made available via the "machinectl" tool,
1499 and exposed in "ps" and similar tools. machined/machinectl
1500 is compile-time optional.
1502 * As discussed earlier, the low-level cgroup configuration
1503 options ControlGroup=, ControlGroupModify=,
1504 ControlGroupPersistent=, ControlGroupAttribute= have been
1505 removed. Please use high-level attribute settings instead as
1506 well as slice units.
1508 * A new bus call SetUnitProperties() has been added to alter
1509 various runtime parameters of a unit. This is primarily
1510 useful to alter cgroup parameters dynamically in a nice way,
1511 but will be extended later on to make more properties
1512 modifiable at runtime. systemctl gained a new set-properties
1513 command that wraps this call.
1515 * A new tool "systemd-run" has been added which can be used to
1516 run arbitrary command lines as transient services or scopes,
1517 while configuring a number of settings via the command
1518 line. This tool is currently very basic, however already
1519 very useful. We plan to extend this tool to even allow
1520 queuing of execution jobs with time triggers from the
1521 command line, similar in fashion to "at".
1523 * nspawn will now inform the user explicitly that kernels with
1524 audit enabled break containers, and suggest the user to turn
1527 * Support for detecting the IMA and AppArmor security
1528 frameworks with ConditionSecurity= has been added.
1530 * journalctl gained a new "-k" switch for showing only kernel
1531 messages, mimicking dmesg output; in addition to "--user"
1532 and "--system" switches for showing only user's own logs
1535 * systemd-delta can now show information about drop-in
1536 snippets extending unit files.
1538 * libsystemd-bus has been substantially updated but is still
1539 not available as public API.
1541 * systemd will now look for the "debug" argument on the kernel
1542 command line and enable debug logging, similar to what
1543 "systemd.log_level=debug" already did before.
1545 * "systemctl set-default", "systemctl get-default" has been
1546 added to configure the default.target symlink, which
1547 controls what to boot into by default.
1549 * "systemctl set-log-level" has been added as a convenient
1550 way to raise and lower systemd logging threshold.
1552 * "systemd-analyze plot" will now show the time the various
1553 generators needed for execution, as well as information
1554 about the unit file loading.
1556 * libsystemd-journal gained a new sd_journal_open_files() call
1557 for opening specific journal files. journactl also gained a
1558 new switch to expose this new functionality. Previously we
1559 only supported opening all files from a directory, or all
1560 files from the system, as opening individual files only is
1561 racy due to journal file rotation.
1563 * systemd gained the new DefaultEnvironment= setting in
1564 /etc/systemd/system.conf to set environment variables for
1567 * If a privileged process logs a journal message with the
1568 OBJECT_PID= field set, then journald will automatically
1569 augment this with additional OBJECT_UID=, OBJECT_GID=,
1570 OBJECT_COMM=, OBJECT_EXE=, ... fields. This is useful if
1571 system services want to log events about specific client
1572 processes. journactl/systemctl has been updated to make use
1573 of this information if all log messages regarding a specific
1576 Contributions from: Auke Kok, Chengwei Yang, Colin Walters,
1577 Cristian Rodríguez, Daniel Albers, Daniel Wallace, Dave
1578 Reisner, David Coppa, David King, David Strauss, Eelco
1579 Dolstra, Gabriel de Perthuis, Harald Hoyer, Jan Alexander
1580 Steffens, Jan Engelhardt, Jan Janssen, Jason St. John, Johan
1581 Heikkilä, Karel Zak, Karol Lewandowski, Kay Sievers, Lennart
1582 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marius Vollmer,
1583 Martin Pitt, Michael Biebl, Michael Olbrich, Michael Tremer,
1584 Michal Schmidt, Michał Bartoszkiewicz, Nirbheek Chauhan,
1585 Pierre Neidhardt, Ross Burton, Ross Lagerwall, Sean McGovern,
1586 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar,
1587 Václav Pavlín, Zachary Cook, Zbigniew Jędrzejewski-Szmek,
1588 Łukasz Stelmach, 장동준
1592 * The Python bindings gained some minimal support for the APIs
1593 exposed by libsystemd-logind.
1595 * ConditionSecurity= gained support for detecting SMACK. Since
1596 this condition already supports SELinux and AppArmor we only
1597 miss IMA for this. Patches welcome!
1599 Contributions from: Karol Lewandowski, Lennart Poettering,
1600 Zbigniew Jędrzejewski-Szmek
1604 * systemd-nspawn will now create /etc/resolv.conf if
1605 necessary, before bind-mounting the host's file onto it.
1607 * systemd-nspawn will now store meta information about a
1608 container on the container's cgroup as extended attribute
1609 fields, including the root directory.
1611 * The cgroup hierarchy has been reworked in many ways. All
1612 objects any of the components systemd creates in the cgroup
1613 tree are now suffixed. More specifically, user sessions are
1614 now placed in cgroups suffixed with ".session", users in
1615 cgroups suffixed with ".user", and nspawn containers in
1616 cgroups suffixed with ".nspawn". Furthermore, all cgroup
1617 names are now escaped in a simple scheme to avoid collision
1618 of userspace object names with kernel filenames. This work
1619 is preparation for making these objects relocatable in the
1620 cgroup tree, in order to allow easy resource partitioning of
1621 these objects without causing naming conflicts.
1623 * systemctl list-dependencies gained the new switches
1624 --plain, --reverse, --after and --before.
1626 * systemd-inhibit now shows the process name of processes that
1627 have taken an inhibitor lock.
1629 * nss-myhostname will now also resolve "localhost"
1630 implicitly. This makes /etc/hosts an optional file and
1631 nicely handles that on IPv6 ::1 maps to both "localhost" and
1634 * libsystemd-logind.so gained a new call
1635 sd_get_machine_names() to enumerate running containers and
1636 VMs (currently only supported by very new libvirt and
1637 nspawn). sd_login_monitor can now be used to watch
1638 VMs/containers coming and going.
1640 * .include is not allowed recursively anymore, and only in
1641 unit files. Usually it is better to use drop-in snippets in
1642 .d/*.conf anyway, as introduced with systemd 198.
1644 * systemd-analyze gained a new "critical-chain" command that
1645 determines the slowest chain of units run during system
1646 boot-up. It is very useful for tracking down where
1647 optimizing boot time is the most beneficial.
1649 * systemd will no longer allow manipulating service paths in
1650 the name=systemd:/system cgroup tree using ControlGroup= in
1651 units. (But is still fine with it in all other dirs.)
1653 * There's a new systemd-nspawn@.service service file that may
1654 be used to easily run nspawn containers as system
1655 services. With the container's root directory in
1656 /var/lib/container/foobar it is now sufficient to run
1657 "systemctl start systemd-nspawn@foobar.service" to boot it.
1659 * systemd-cgls gained a new parameter "--machine" to list only
1660 the processes within a certain container.
1662 * ConditionSecurity= now can check for "apparmor". We still
1663 are lacking checks for SMACK and IMA for this condition
1664 check though. Patches welcome!
1666 * A new configuration file /etc/systemd/sleep.conf has been
1667 added that may be used to configure which kernel operation
1668 systemd is supposed to execute when "suspend", "hibernate"
1669 or "hybrid-sleep" is requested. This makes the new kernel
1670 "freeze" state accessible to the user.
1672 * ENV{SYSTEMD_WANTS} in udev rules will now implicitly escape
1673 the passed argument if applicable.
1675 Contributions from: Auke Kok, Colin Guthrie, Colin Walters,
1676 Cristian Rodríguez, Daniel Buch, Daniel Wallace, Dave Reisner,
1677 Evangelos Foutras, Greg Kroah-Hartman, Harald Hoyer, Josh
1678 Triplett, Kay Sievers, Lennart Poettering, Lukas Nykryn,
1679 MUNEDA Takahiro, Mantas Mikulėnas, Mirco Tischler, Nathaniel
1680 Chen, Nirbheek Chauhan, Ronny Chevalier, Ross Lagerwall, Tom
1681 Gundersen, Umut Tezduyar, Ville Skyttä, Zbigniew
1686 * The output of 'systemctl list-jobs' got some polishing. The
1687 '--type=' argument may now be passed more than once. A new
1688 command 'systemctl list-sockets' has been added which shows
1689 a list of kernel sockets systemd is listening on with the
1690 socket units they belong to, plus the units these socket
1693 * The experimental libsystemd-bus library got substantial
1694 updates to work in conjunction with the (also experimental)
1695 kdbus kernel project. It works well enough to exchange
1696 messages with some sophistication. Note that kdbus is not
1697 ready yet, and the library is mostly an elaborate test case
1698 for now, and not installable.
1700 * systemd gained a new unit 'systemd-static-nodes.service'
1701 that generates static device nodes earlier during boot, and
1702 can run in conjunction with udev.
1704 * libsystemd-login gained a new call sd_pid_get_user_unit()
1705 to retrieve the user systemd unit a process is running
1706 in. This is useful for systems where systemd is used as
1709 * systemd-nspawn now places all containers in the new /machine
1710 top-level cgroup directory in the name=systemd
1711 hierarchy. libvirt will soon do the same, so that we get a
1712 uniform separation of /system, /user and /machine for system
1713 services, user processes and containers/virtual
1714 machines. This new cgroup hierarchy is also useful to stick
1715 stable names to specific container instances, which can be
1716 recognized later this way (this name may be controlled
1717 via systemd-nspawn's new -M switch). libsystemd-login also
1718 gained a new call sd_pid_get_machine_name() to retrieve the
1719 name of the container/VM a specific process belongs to.
1721 * bootchart can now store its data in the journal.
1723 * libsystemd-journal gained a new call
1724 sd_journal_add_conjunction() for AND expressions to the
1725 matching logic. This can be used to express more complex
1726 logical expressions.
1728 * journactl can now take multiple --unit= and --user-unit=
1731 * The cryptsetup logic now understands the "luks.key=" kernel
1732 command line switch for specifying a file to read the
1733 decryption key from. Also, if a configured key file is not
1734 found the tool will now automatically fall back to prompting
1737 * Python systemd.journal module was updated to wrap recently
1738 added functions from libsystemd-journal. The interface was
1739 changed to bring the low level interface in s.j._Reader
1740 closer to the C API, and the high level interface in
1741 s.j.Reader was updated to wrap and convert all data about
1744 Contributions from: Anatol Pomozov, Auke Kok, Harald Hoyer,
1745 Henrik Grindal Bakken, Josh Triplett, Kay Sievers, Lennart
1746 Poettering, Lukas Nykryn, Mantas Mikulėnas Marius Vollmer,
1747 Martin Jansa, Martin Pitt, Michael Biebl, Michal Schmidt,
1748 Mirco Tischler, Pali Rohar, Simon Peeters, Steven Hiscocks,
1749 Tom Gundersen, Zbigniew Jędrzejewski-Szmek
1753 * journalctl --update-catalog now understands a new --root=
1754 option to operate on catalogs found in a different root
1757 * During shutdown after systemd has terminated all running
1758 services a final killing loop kills all remaining left-over
1759 processes. We will now print the name of these processes
1760 when we send SIGKILL to them, since this usually indicates a
1763 * If /etc/crypttab refers to password files stored on
1764 configured mount points automatic dependencies will now be
1765 generated to ensure the specific mount is established first
1766 before the key file is attempted to be read.
1768 * 'systemctl status' will now show information about the
1769 network sockets a socket unit is listening on.
1771 * 'systemctl status' will also shown information about any
1772 drop-in configuration file for units. (Drop-In configuration
1773 files in this context are files such as
1774 /etc/systemd/systemd/foobar.service.d/*.conf)
1776 * systemd-cgtop now optionally shows summed up CPU times of
1777 cgroups. Press '%' while running cgtop to switch between
1778 percentage and absolute mode. This is useful to determine
1779 which cgroups use up the most CPU time over the entire
1780 runtime of the system. systemd-cgtop has also been updated
1781 to be 'pipeable' for processing with further shell tools.
1783 * 'hostnamectl set-hostname' will now allow setting of FQDN
1786 * The formatting and parsing of time span values has been
1787 changed. The parser now understands fractional expressions
1788 such as "5.5h". The formatter will now output fractional
1789 expressions for all time spans under 1min, i.e. "5.123456s"
1790 rather than "5s 123ms 456us". For time spans under 1s
1791 millisecond values are shown, for those under 1ms
1792 microsecond values are shown. This should greatly improve
1793 all time-related output of systemd.
1795 * libsystemd-login and libsystemd-journal gained new
1796 functions for querying the poll() events mask and poll()
1797 timeout value for integration into arbitrary event
1800 * localectl gained the ability to list available X11 keymaps
1801 (models, layouts, variants, options).
1803 * 'systemd-analyze dot' gained the ability to filter for
1804 specific units via shell-style globs, to create smaller,
1805 more useful graphs. I.e. it is now possible to create simple
1806 graphs of all the dependencies between only target units, or
1807 of all units that Avahi has dependencies with.
1809 Contributions from: Cristian Rodríguez, Dr. Tilmann Bubeck,
1810 Harald Hoyer, Holger Hans Peter Freyther, Kay Sievers, Kelly
1811 Anderson, Koen Kooi, Lennart Poettering, Maksim Melnikau,
1812 Marc-Antoine Perennou, Marius Vollmer, Martin Pitt, Michal
1813 Schmidt, Oleksii Shevchuk, Ronny Chevalier, Simon McVittie,
1814 Steven Hiscocks, Thomas Weißschuh, Umut Tezduyar, Václav
1815 Pavlín, Zbigniew Jędrzejewski-Szmek, Łukasz Stelmach
1819 * The boot-time readahead implementation for rotating media
1820 will now read the read-ahead data in multiple passes which
1821 consist of all read requests made in equidistant time
1822 intervals. This means instead of strictly reading read-ahead
1823 data in its physical order on disk we now try to find a
1824 middle ground between physical and access time order.
1826 * /etc/os-release files gained a new BUILD_ID= field for usage
1827 on operating systems that provide continuous builds of OS
1830 Contributions from: Auke Kok, Eelco Dolstra, Kay Sievers,
1831 Lennart Poettering, Lukas Nykryn, Martin Pitt, Václav Pavlín
1832 William Douglas, Zbigniew Jędrzejewski-Szmek
1836 * systemd-python gained an API exposing libsystemd-daemon.
1838 * The SMACK setup logic gained support for uploading CIPSO
1841 * Behaviour of PrivateTmp=, ReadWriteDirectories=,
1842 ReadOnlyDirectories= and InaccessibleDirectories= has
1843 changed. The private /tmp and /var/tmp directories are now
1844 shared by all processes of a service (which means
1845 ExecStartPre= may now leave data in /tmp that ExecStart= of
1846 the same service can still access). When a service is
1847 stopped its temporary directories are immediately deleted
1848 (normal clean-up with tmpfiles is still done in addition to
1851 * By default, systemd will now set a couple of sysctl
1852 variables in the kernel: the safe sysrq options are turned
1853 on, IP route verification is turned on, and source routing
1854 disabled. The recently added hardlink and softlink
1855 protection of the kernel is turned on. These settings should
1856 be reasonably safe, and good defaults for all new systems.
1858 * The predictable network naming logic may now be turned off
1859 with a new kernel command line switch: net.ifnames=0.
1861 * A new libsystemd-bus module has been added that implements a
1862 pretty complete D-Bus client library. For details see:
1864 http://lists.freedesktop.org/archives/systemd-devel/2013-March/009797.html
1866 * journald will now explicitly flush the journal files to disk
1867 at the latest 5min after each write. The file will then also
1868 be marked offline until the next write. This should increase
1869 reliability in case of a crash. The synchronization delay
1870 can be configured via SyncIntervalSec= in journald.conf.
1872 * There's a new remote-fs-setup.target unit that can be used
1873 to pull in specific services when at least one remote file
1874 system is to be mounted.
1876 * There are new targets timers.target and paths.target as
1877 canonical targets to pull user timer and path units in
1878 from. This complements sockets.target with a similar
1879 purpose for socket units.
1881 * libudev gained a new call udev_device_set_attribute_value()
1882 to set sysfs attributes of a device.
1884 * The udev daemon now sets the default number of worker
1885 processes executed in parallel based on the number of available
1886 CPUs instead of the amount of available RAM. This is supposed
1887 to provide a more reliable default and limit a too aggressive
1888 paralellism for setups with 1000s of devices connected.
1890 Contributions from: Auke Kok, Colin Walters, Cristian
1891 Rodríguez, Daniel Buch, Dave Reisner, Frederic Crozat, Hannes
1892 Reinecke, Harald Hoyer, Jan Alexander Steffens, Jan
1893 Engelhardt, Josh Triplett, Kay Sievers, Lennart Poettering,
1894 Mantas Mikulėnas, Martin Pitt, Mathieu Bridon, Michael Biebl,
1895 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nathaniel Chen,
1896 Oleksii Shevchuk, Ozan Çağlayan, Thomas Hindoe Paaboel
1897 Andersen, Tollef Fog Heen, Tom Gundersen, Umut Tezduyar,
1898 Zbigniew Jędrzejewski-Szmek
1902 * Configuration of unit files may now be extended via drop-in
1903 files without having to edit/override the unit files
1904 themselves. More specifically, if the administrator wants to
1905 change one value for a service file foobar.service he can
1906 now do so by dropping in a configuration snippet into
1907 /etc/systemd/system/foobar.service.d/*.conf. The unit logic
1908 will load all these snippets and apply them on top of the
1909 main unit configuration file, possibly extending or
1910 overriding its settings. Using these drop-in snippets is
1911 generally nicer than the two earlier options for changing
1912 unit files locally: copying the files from
1913 /usr/lib/systemd/system/ to /etc/systemd/system/ and editing
1914 them there; or creating a new file in /etc/systemd/system/
1915 that incorporates the original one via ".include". Drop-in
1916 snippets into these .d/ directories can be placed in any
1917 directory systemd looks for units in, and the usual
1918 overriding semantics between /usr/lib, /etc and /run apply
1921 * Most unit file settings which take lists of items can now be
1922 reset by assigning the empty string to them. For example,
1923 normally, settings such as Environment=FOO=BAR append a new
1924 environment variable assignment to the environment block,
1925 each time they are used. By assigning Environment= the empty
1926 string the environment block can be reset to empty. This is
1927 particularly useful with the .d/*.conf drop-in snippets
1928 mentioned above, since this adds the ability to reset list
1929 settings from vendor unit files via these drop-ins.
1931 * systemctl gained a new "list-dependencies" command for
1932 listing the dependencies of a unit recursively.
1934 * Inhibitors are now honored and listed by "systemctl
1935 suspend", "systemctl poweroff" (and similar) too, not only
1936 GNOME. These commands will also list active sessions by
1939 * Resource limits (as exposed by the various control group
1940 controllers) can now be controlled dynamically at runtime
1941 for all units. More specifically, you can now use a command
1942 like "systemctl set-cgroup-attr foobar.service cpu.shares
1943 2000" to alter the CPU shares a specific service gets. These
1944 settings are stored persistently on disk, and thus allow the
1945 administrator to easily adjust the resource usage of
1946 services with a few simple commands. This dynamic resource
1947 management logic is also available to other programs via the
1948 bus. Almost any kernel cgroup attribute and controller is
1951 * systemd-vconsole-setup will now copy all font settings to
1952 all allocated VTs, where it previously applied them only to
1955 * libsystemd-login gained the new sd_session_get_tty() API
1958 * This release drops support for a few legacy or
1959 distribution-specific LSB facility names when parsing init
1960 scripts: $x-display-manager, $mail-transfer-agent,
1961 $mail-transport-agent, $mail-transfer-agent, $smtp,
1962 $null. Also, the mail-transfer-agent.target unit backing
1963 this has been removed. Distributions which want to retain
1964 compatibility with this should carry the burden for
1965 supporting this themselves and patch support for these back
1966 in, if they really need to. Also, the facilities $syslog and
1967 $local_fs are now ignored, since systemd does not support
1968 early-boot LSB init scripts anymore, and these facilities
1969 are implied anyway for normal services. syslog.target has
1972 * There are new bus calls on PID1's Manager object for
1973 cancelling jobs, and removing snapshot units. Previously,
1974 both calls were only available on the Job and Snapshot
1977 * systemd-journal-gatewayd gained SSL support.
1979 * The various "environment" files, such as /etc/locale.conf
1980 now support continuation lines with a backslash ("\") as
1981 last character in the line, similarly in style (but different)
1982 to how this is supported in shells.
1984 * For normal user processes the _SYSTEMD_USER_UNIT= field is
1985 now implicitly appended to every log entry logged. systemctl
1986 has been updated to filter by this field when operating on a
1987 user systemd instance.
1989 * nspawn will now implicitly add the CAP_AUDIT_WRITE and
1990 CAP_AUDIT_CONTROL capabilities to the capabilities set for
1991 the container. This makes it easier to boot unmodified
1992 Fedora systems in a container, which however still requires
1993 audit=0 to be passed on the kernel command line. Auditing in
1994 kernel and userspace is unfortunately still too broken in
1995 context of containers, hence we recommend compiling it out
1996 of the kernel or using audit=0. Hopefully this will be fixed
1997 one day for good in the kernel.
1999 * nspawn gained the new --bind= and --bind-ro= parameters to
2000 bind mount specific directories from the host into the
2003 * nspawn will now mount its own devpts file system instance
2004 into the container, in order not to leak pty devices from
2005 the host into the container.
2007 * systemd will now read the firmware boot time performance
2008 information from the EFI variables, if the used boot loader
2009 supports this, and takes it into account for boot performance
2010 analysis via "systemd-analyze". This is currently supported
2011 only in conjunction with Gummiboot, but could be supported
2012 by other boot loaders too. For details see:
2014 http://www.freedesktop.org/wiki/Software/systemd/BootLoaderInterface
2016 * A new generator has been added that automatically mounts the
2017 EFI System Partition (ESP) to /boot, if that directory
2018 exists, is empty, and no other file system has been
2019 configured to be mounted there.
2021 * logind will now send out PrepareForSleep(false) out
2022 unconditionally, after coming back from suspend. This may be
2023 used by applications as asynchronous notification for
2024 system resume events.
2026 * "systemctl unlock-sessions" has been added, that allows
2027 unlocking the screens of all user sessions at once, similar
2028 to how "systemctl lock-sessions" already locked all users
2029 sessions. This is backed by a new D-Bus call UnlockSessions().
2031 * "loginctl seat-status" will now show the master device of a
2032 seat. (i.e. the device of a seat that needs to be around for
2033 the seat to be considered available, usually the graphics
2036 * tmpfiles gained a new "X" line type, that allows
2037 configuration of files and directories (with wildcards) that
2038 shall be excluded from automatic cleanup ("aging").
2040 * udev default rules set the device node permissions now only
2041 at "add" events, and do not change them any longer with a
2042 later "change" event.
2044 * The log messages for lid events and power/sleep keypresses
2045 now carry a message ID.
2047 * We now have a substantially larger unit test suite, but this
2048 continues to be work in progress.
2050 * udevadm hwdb gained a new --root= parameter to change the
2051 root directory to operate relative to.
2053 * logind will now issue a background sync() request to the kernel
2054 early at shutdown, so that dirty buffers are flushed to disk early
2055 instead of at the last moment, in order to optimize shutdown
2058 * A new bootctl tool has been added that is an interface for
2059 certain boot loader operations. This is currently a preview
2060 and is likely to be extended into a small mechanism daemon
2061 like timedated, localed, hostnamed, and can be used by
2062 graphical UIs to enumerate available boot options, and
2063 request boot into firmware operations.
2065 * systemd-bootchart has been relicensed to LGPLv2.1+ to match
2066 the rest of the package. It also has been updated to work
2067 correctly in initrds.
2069 * Policykit previously has been runtime optional, and is now
2070 also compile time optional via a configure switch.
2072 * systemd-analyze has been reimplemented in C. Also "systemctl
2073 dot" has moved into systemd-analyze.
2075 * "systemctl status" with no further parameters will now print
2076 the status of all active or failed units.
2078 * Operations such as "systemctl start" can now be executed
2079 with a new mode "--irreversible" which may be used to queue
2080 operations that cannot accidentally be reversed by a later
2081 job queuing. This is by default used to make shutdown
2082 requests more robust.
2084 * The Python API of systemd now gained a new module for
2085 reading journal files.
2087 * A new tool kernel-install has been added that can install
2088 kernel images according to the Boot Loader Specification:
2090 http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec
2092 * Boot time console output has been improved to provide
2093 animated boot time output for hanging jobs.
2095 * A new tool systemd-activate has been added which can be used
2096 to test socket activation with, directly from the command
2097 line. This should make it much easier to test and debug
2098 socket activation in daemons.
2100 * journalctl gained a new "--reverse" (or -r) option to show
2101 journal output in reverse order (i.e. newest line first).
2103 * journalctl gained a new "--pager-end" (or -e) option to jump
2104 to immediately jump to the end of the journal in the
2105 pager. This is only supported in conjunction with "less".
2107 * journalctl gained a new "--user-unit=" option, that works
2108 similarly to "--unit=" but filters for user units rather than
2111 * A number of unit files to ease adoption of systemd in
2112 initrds has been added. This moves some minimal logic from
2113 the various initrd implementations into systemd proper.
2115 * The journal files are now owned by a new group
2116 "systemd-journal", which exists specifically to allow access
2117 to the journal, and nothing else. Previously, we used the
2118 "adm" group for that, which however possibly covers more
2119 than just journal/log file access. This new group is now
2120 already used by systemd-journal-gatewayd to ensure this
2121 daemon gets access to the journal files and as little else
2122 as possible. Note that "make install" will also set FS ACLs
2123 up for /var/log/journal to give "adm" and "wheel" read
2124 access to it, in addition to "systemd-journal" which owns
2125 the journal files. We recommend that packaging scripts also
2126 add read access to "adm" + "wheel" to /var/log/journal, and
2127 all existing/future journal files. To normal users and
2128 administrators little changes, however packagers need to
2129 ensure to create the "systemd-journal" system group at
2130 package installation time.
2132 * The systemd-journal-gatewayd now runs as unprivileged user
2133 systemd-journal-gateway:systemd-journal-gateway. Packaging
2134 scripts need to create these system user/group at
2137 * timedated now exposes a new boolean property CanNTP that
2138 indicates whether a local NTP service is available or not.
2140 * systemd-detect-virt will now also detect xen PVs
2142 * The pstore file system is now mounted by default, if it is
2145 * In addition to the SELinux and IMA policies we will now also
2146 load SMACK policies at early boot.
2148 Contributions from: Adel Gadllah, Aleksander Morgado, Auke
2149 Kok, Ayan George, Bastien Nocera, Colin Walters, Daniel Buch,
2150 Daniel Wallace, Dave Reisner, David Herrmann, David Strauss,
2151 Eelco Dolstra, Enrico Scholz, Frederic Crozat, Harald Hoyer,
2152 Jan Janssen, Jonathan Callen, Kay Sievers, Lennart Poettering,
2153 Lukas Nykryn, Mantas Mikulėnas, Marc-Antoine Perennou, Martin
2154 Pitt, Mauro Dreissig, Max F. Albrecht, Michael Biebl, Michael
2155 Olbrich, Michal Schmidt, Michal Sekletar, Michal Vyskocil,
2156 Michał Bartoszkiewicz, Mirco Tischler, Nathaniel Chen, Nestor
2157 Ovroy, Oleksii Shevchuk, Paul W. Frields, Piotr Drąg, Rob
2158 Clark, Ryan Lortie, Simon McVittie, Simon Peeters, Steven
2159 Hiscocks, Thomas Hindoe Paaboel Andersen, Tollef Fog Heen, Tom
2160 Gundersen, Umut Tezduyar, William Giokas, Zbigniew
2161 Jędrzejewski-Szmek, Zeeshan Ali (Khattak)
2165 * Timer units now support calendar time events in addition to
2166 monotonic time events. That means you can now trigger a unit
2167 based on a calendar time specification such as "Thu,Fri
2168 2013-*-1,5 11:12:13" which refers to 11:12:13 of the first
2169 or fifth day of any month of the year 2013, given that it is
2170 a thursday or friday. This brings timer event support
2171 considerably closer to cron's capabilities. For details on
2172 the supported calendar time specification language see
2175 * udev now supports a number of different naming policies for
2176 network interfaces for predictable names, and a combination
2177 of these policies is now the default. Please see this wiki
2178 document for details:
2180 http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames
2182 * Auke Kok's bootchart implementation has been added to the
2183 systemd tree. It is an optional component that can graph the
2184 boot in quite some detail. It is one of the best bootchart
2185 implementations around and minimal in its code and
2188 * nss-myhostname has been integrated into the systemd source
2189 tree. nss-myhostname guarantees that the local hostname
2190 always stays resolvable via NSS. It has been a weak
2191 requirement of systemd-hostnamed since a long time, and
2192 since its code is actually trivial we decided to just
2193 include it in systemd's source tree. It can be turned off
2194 with a configure switch.
2196 * The read-ahead logic is now capable of properly detecting
2197 whether a btrfs file system is on SSD or rotating media, in
2198 order to optimize the read-ahead scheme. Previously, it was
2199 only capable of detecting this on traditional file systems
2202 * In udev, additional device properties are now read from the
2203 IAB in addition to the OUI database. Also, Bluetooth company
2204 identities are attached to the devices as well.
2206 * In service files %U may be used as specifier that is
2207 replaced by the configured user name of the service.
2209 * nspawn may now be invoked without a controlling TTY. This
2210 makes it suitable for invocation as its own service. This
2211 may be used to set up a simple containerized server system
2212 using only core OS tools.
2214 * systemd and nspawn can now accept socket file descriptors
2215 when they are started for socket activation. This enables
2216 implementation of socket activated nspawn
2217 containers. i.e. think about autospawning an entire OS image
2218 when the first SSH or HTTP connection is received. We expect
2219 that similar functionality will also be added to libvirt-lxc
2222 * journalctl will now suppress ANSI color codes when
2223 presenting log data.
2225 * systemctl will no longer show control group information for
2226 a unit if a the control group is empty anyway.
2228 * logind can now automatically suspend/hibernate/shutdown the
2231 * /etc/machine-info and hostnamed now also expose the chassis
2232 type of the system. This can be used to determine whether
2233 the local system is a laptop, desktop, handset or
2234 tablet. This information may either be configured by the
2235 user/vendor or is automatically determined from ACPI and DMI
2236 information if possible.
2238 * A number of PolicyKit actions are now bound together with
2239 "imply" rules. This should simplify creating UIs because
2240 many actions will now authenticate similar ones as well.
2242 * Unit files learnt a new condition ConditionACPower= which
2243 may be used to conditionalize a unit depending on whether an
2244 AC power source is connected or not, of whether the system
2245 is running on battery power.
2247 * systemctl gained a new "is-failed" verb that may be used in
2248 shell scripts and suchlike to check whether a specific unit
2249 is in the "failed" state.
2251 * The EnvironmentFile= setting in unit files now supports file
2252 globbing, and can hence be used to easily read a number of
2253 environment files at once.
2255 * systemd will no longer detect and recognize specific
2256 distributions. All distribution-specific #ifdeffery has been
2257 removed, systemd is now fully generic and
2258 distribution-agnostic. Effectively, not too much is lost as
2259 a lot of the code is still accessible via explicit configure
2260 switches. However, support for some distribution specific
2261 legacy configuration file formats has been dropped. We
2262 recommend distributions to simply adopt the configuration
2263 files everybody else uses now and convert the old
2264 configuration from packaging scripts. Most distributions
2265 already did that. If that's not possible or desirable,
2266 distributions are welcome to forward port the specific
2267 pieces of code locally from the git history.
2269 * When logging a message about a unit systemd will now always
2270 log the unit name in the message meta data.
2272 * localectl will now also discover system locale data that is
2273 not stored in locale archives, but directly unpacked.
2275 * logind will no longer unconditionally use framebuffer
2276 devices as seat masters, i.e. as devices that are required
2277 to be existing before a seat is considered preset. Instead,
2278 it will now look for all devices that are tagged as
2279 "seat-master" in udev. By default framebuffer devices will
2280 be marked as such, but depending on local systems other
2281 devices might be marked as well. This may be used to
2282 integrate graphics cards using closed source drivers (such
2283 as NVidia ones) more nicely into logind. Note however, that
2284 we recommend using the open source NVidia drivers instead,
2285 and no udev rules for the closed-source drivers will be
2286 shipped from us upstream.
2288 Contributions from: Adam Williamson, Alessandro Crismani, Auke
2289 Kok, Colin Walters, Daniel Wallace, Dave Reisner, David
2290 Herrmann, David Strauss, Dimitrios Apostolou, Eelco Dolstra,
2291 Eric Benoit, Giovanni Campagna, Hannes Reinecke, Henrik
2292 Grindal Bakken, Hermann Gausterer, Kay Sievers, Lennart
2293 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marcel Holtmann,
2294 Martin Pitt, Matthew Monaco, Michael Biebl, Michael Terry,
2295 Michal Schmidt, Michal Sekletar, Michał Bartoszkiewicz, Oleg
2296 Samarin, Pekka Lundstrom, Philip Nilsson, Ramkumar
2297 Ramachandra, Richard Yao, Robert Millan, Sami Kerola, Shawn
2298 Landden, Thomas Hindoe Paaboel Andersen, Thomas Jarosch,
2299 Tollef Fog Heen, Tom Gundersen, Umut Tezduyar, Zbigniew
2304 * udev gained support for loading additional device properties
2305 from an indexed database that is keyed by vendor/product IDs
2306 and similar device identifiers. For the beginning this
2307 "hwdb" is populated with data from the well-known PCI and
2308 USB database, but also includes PNP, ACPI and OID data. In
2309 the longer run this indexed database shall grow into
2310 becoming the one central database for non-essential
2311 userspace device metadata. Previously, data from the PCI/USB
2312 database was only attached to select devices, since the
2313 lookup was a relatively expensive operation due to O(n) time
2314 complexity (with n being the number of entries in the
2315 database). Since this is now O(1), we decided to add in this
2316 data for all devices where this is available, by
2317 default. Note that the indexed database needs to be rebuilt
2318 when new data files are installed. To achieve this you need
2319 to update your packaging scripts to invoke "udevadm hwdb
2320 --update" after installation of hwdb data files. For
2321 RPM-based distributions we introduced the new
2322 %udev_hwdb_update macro for this purpose.
2324 * The Journal gained support for the "Message Catalog", an
2325 indexed database to link up additional information with
2326 journal entries. For further details please check:
2328 http://www.freedesktop.org/wiki/Software/systemd/catalog
2330 The indexed message catalog database also needs to be
2331 rebuilt after installation of message catalog files. Use
2332 "journalctl --update-catalog" for this. For RPM-based
2333 distributions we introduced the %journal_catalog_update
2334 macro for this purpose.
2336 * The Python Journal bindings gained support for the standard
2337 Python logging framework.
2339 * The Journal API gained new functions for checking whether
2340 the underlying file system of a journal file is capable of
2341 properly reporting file change notifications, or whether
2342 applications that want to reflect journal changes "live"
2343 need to recheck journal files continuously in appropriate
2346 * It is now possible to set the "age" field for tmpfiles
2347 entries to 0, indicating that files matching this entry
2348 shall always be removed when the directories are cleaned up.
2350 * coredumpctl gained a new "gdb" verb which invokes gdb
2351 right-away on the selected coredump.
2353 * There's now support for "hybrid sleep" on kernels that
2354 support this, in addition to "suspend" and "hibernate". Use
2355 "systemctl hybrid-sleep" to make use of this.
2357 * logind's HandleSuspendKey= setting (and related settings)
2358 now gained support for a new "lock" setting to simply
2359 request the screen lock on all local sessions, instead of
2360 actually executing a suspend or hibernation.
2362 * systemd will now mount the EFI variables file system by
2365 * Socket units now gained support for configuration of the
2366 SMACK security label.
2368 * timedatectl will now output the time of the last and next
2369 daylight saving change.
2371 * We dropped support for various legacy and distro-specific
2372 concepts, such as insserv, early-boot SysV services
2373 (i.e. those for non-standard runlevels such as 'b' or 'S')
2374 or ArchLinux /etc/rc.conf support. We recommend the
2375 distributions who still need support this to either continue
2376 to maintain the necessary patches downstream, or find a
2377 different solution. (Talk to us if you have questions!)
2379 * Various systemd components will now bypass PolicyKit checks
2380 for root and otherwise handle properly if PolicyKit is not
2381 found to be around. This should fix most issues for
2382 PolicyKit-less systems. Quite frankly this should have been
2383 this way since day one. It is absolutely our intention to
2384 make systemd work fine on PolicyKit-less systems, and we
2385 consider it a bug if something does not work as it should if
2386 PolicyKit is not around.
2388 * For embedded systems it is now possible to build udev and
2389 systemd without blkid and/or kmod support.
2391 * "systemctl switch-root" is now capable of switching root
2392 more than once. I.e. in addition to transitions from the
2393 initrd to the host OS it is now possible to transition to
2394 further OS images from the host. This is useful to implement
2395 offline updating tools.
2397 * Various other additions have been made to the RPM macros
2398 shipped with systemd. Use %udev_rules_update() after
2399 installing new udev rules files. %_udevhwdbdir,
2400 %_udevrulesdir, %_journalcatalogdir, %_tmpfilesdir,
2401 %_sysctldir are now available which resolve to the right
2402 directories for packages to place various data files in.
2404 * journalctl gained the new --full switch (in addition to
2405 --all, to disable ellipsation for long messages.
2407 Contributions from: Anders Olofsson, Auke Kok, Ben Boeckel,
2408 Colin Walters, Cosimo Cecchi, Daniel Wallace, Dave Reisner,
2409 Eelco Dolstra, Holger Hans Peter Freyther, Kay Sievers,
2410 Chun-Yi Lee, Lekensteyn, Lennart Poettering, Mantas Mikulėnas,
2411 Marti Raudsepp, Martin Pitt, Mauro Dreissig, Michael Biebl,
2412 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nis Martensen,
2413 Oleksii Shevchuk, Olivier Brunel, Ramkumar Ramachandra, Thomas
2414 Bächler, Thomas Hindoe Paaboel Andersen, Tom Gundersen, Tony
2415 Camuso, Umut Tezduyar, Zbigniew Jędrzejewski-Szmek
2419 * journalctl gained new --since= and --until= switches to
2420 filter by time. It also now supports nice filtering for
2421 units via --unit=/-u.
2423 * Type=oneshot services may use ExecReload= and do the
2426 * The journal daemon now supports time-based rotation and
2427 vacuuming, in addition to the usual disk-space based
2430 * The journal will now index the available field values for
2431 each field name. This enables clients to show pretty drop
2432 downs of available match values when filtering. The bash
2433 completion of journalctl has been updated
2434 accordingly. journalctl gained a new switch -F to list all
2435 values a certain field takes in the journal database.
2437 * More service events are now written as structured messages
2438 to the journal, and made recognizable via message IDs.
2440 * The timedated, localed and hostnamed mini-services which
2441 previously only provided support for changing time, locale
2442 and hostname settings from graphical DEs such as GNOME now
2443 also have a minimal (but very useful) text-based client
2444 utility each. This is probably the nicest way to changing
2445 these settings from the command line now, especially since
2446 it lists available options and is fully integrated with bash
2449 * There's now a new tool "systemd-coredumpctl" to list and
2450 extract coredumps from the journal.
2452 * We now install a README each in /var/log/ and
2453 /etc/rc.d/init.d explaining where the system logs and init
2454 scripts went. This hopefully should help folks who go to
2455 that dirs and look into the otherwise now empty void and
2456 scratch their heads.
2458 * When user-services are invoked (by systemd --user) the
2459 $MANAGERPID env var is set to the PID of systemd.
2461 * SIGRTMIN+24 when sent to a --user instance will now result
2462 in immediate termination of systemd.
2464 * gatewayd received numerous feature additions such as a
2465 "follow" mode, for live syncing and filtering.
2467 * browse.html now allows filtering and showing detailed
2468 information on specific entries. Keyboard navigation and
2469 mouse screen support has been added.
2471 * gatewayd/journalctl now supports HTML5/JSON
2472 Server-Sent-Events as output.
2474 * The SysV init script compatibility logic will now
2475 heuristically determine whether a script supports the
2476 "reload" verb, and only then make this available as
2479 * "systemctl status --follow" has been removed, use "journalctl
2482 * journald.conf's RuntimeMinSize=, PersistentMinSize= settings
2483 have been removed since they are hardly useful to be
2486 * And I'd like to take the opportunity to specifically mention
2487 Zbigniew for his great contributions. Zbigniew, you rock!
2489 Contributions from: Andrew Eikum, Christian Hesse, Colin
2490 Guthrie, Daniel J Walsh, Dave Reisner, Eelco Dolstra, Ferenc
2491 Wágner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Mantas
2492 Mikulėnas, Martin Mikkelsen, Martin Pitt, Michael Olbrich,
2493 Michael Stapelberg, Michal Schmidt, Sebastian Ott, Thomas
2494 Bächler, Umut Tezduyar, Will Woods, Wulf C. Krueger, Zbigniew
2495 Jędrzejewski-Szmek, Сковорода Никита Андреевич
2499 * If /etc/vconsole.conf is non-existent or empty we will no
2500 longer load any console font or key map at boot by
2501 default. Instead the kernel defaults will be left
2502 intact. This is definitely the right thing to do, as no
2503 configuration should mean no configuration, and hard-coding
2504 font names that are different on all archs is probably a bad
2505 idea. Also, the kernel default key map and font should be
2506 good enough for most cases anyway, and mostly identical to
2507 the userspace fonts/key maps we previously overloaded them
2508 with. If distributions want to continue to default to a
2509 non-kernel font or key map they should ship a default
2510 /etc/vconsole.conf with the appropriate contents.
2512 Contributions from: Colin Walters, Daniel J Walsh, Dave
2513 Reisner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Tollef
2514 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
2518 * journalctl gained a new --cursor= switch to show entries
2519 starting from the specified location in the journal.
2521 * We now enforce a size limit on journal entry fields exported
2522 with "-o json" in journalctl. Fields larger than 4K will be
2523 assigned null. This can be turned off with --all.
2525 * An (optional) journal gateway daemon is now available as
2526 "systemd-journal-gatewayd.service". This service provides
2527 access to the journal via HTTP and JSON. This functionality
2528 will be used to implement live log synchronization in both
2529 pull and push modes, but has various other users too, such
2530 as easy log access for debugging of embedded devices. Right
2531 now it is already useful to retrieve the journal via HTTP:
2533 # systemctl start systemd-journal-gatewayd.service
2534 # wget http://localhost:19531/entries
2536 This will download the journal contents in a
2537 /var/log/messages compatible format. The same as JSON:
2539 # curl -H"Accept: application/json" http://localhost:19531/entries
2541 This service is also accessible via a web browser where a
2542 single static HTML5 app is served that uses the JSON logic
2543 to enable the user to do some basic browsing of the
2544 journal. This will be extended later on. Here's an example
2545 screenshot of this app in its current state:
2547 http://0pointer.de/public/journal-gatewayd
2549 Contributions from: Kay Sievers, Lennart Poettering, Robert
2550 Milasan, Tom Gundersen
2554 * The bash completion logic is now available for journalctl
2557 * We do not mount the "cpuset" controller anymore together with
2558 "cpu" and "cpuacct", as "cpuset" groups generally cannot be
2559 started if no parameters are assigned to it. "cpuset" hence
2560 broke code that assumed it it could create "cpu" groups and
2563 * journalctl -f will now subscribe to terminal size changes,
2564 and line break accordingly.
2566 Contributions from: Dave Reisner, Kay Sievers, Lennart
2567 Poettering, Lukas Nykrynm, Mirco Tischler, Václav Pavlín
2571 * nspawn will now create a symlink /etc/localtime in the
2572 container environment, copying the host's timezone
2573 setting. Previously this has been done via a bind mount, but
2574 since symlinks cannot be bind mounted this has now been
2575 changed to create/update the appropriate symlink.
2577 * journalctl -n's line number argument is now optional, and
2578 will default to 10 if omitted.
2580 * journald will now log the maximum size the journal files may
2581 take up on disk. This is particularly useful if the default
2582 built-in logic of determining this parameter from the file
2583 system size is used. Use "systemctl status
2584 systemd-journald.service" to see this information.
2586 * The multi-seat X wrapper tool has been stripped down. As X
2587 is now capable of enumerating graphics devices via udev in a
2588 seat-aware way the wrapper is not strictly necessary
2589 anymore. A stripped down temporary stop-gap is still shipped
2590 until the upstream display managers have been updated to
2591 fully support the new X logic. Expect this wrapper to be
2592 removed entirely in one of the next releases.
2594 * HandleSleepKey= in logind.conf has been split up into
2595 HandleSuspendKey= and HandleHibernateKey=. The old setting
2596 is not available anymore. X11 and the kernel are
2597 distuingishing between these keys and we should too. This
2598 also means the inhibition lock for these keys has been split
2601 Contributions from: Dave Airlie, Eelco Dolstra, Lennart
2602 Poettering, Lukas Nykryn, Václav Pavlín
2606 * Whenever a unit changes state we will now log this to the
2607 journal and show along the unit's own log output in
2610 * ConditionPathIsMountPoint= can now properly detect bind
2611 mount points too. (Previously, a bind mount of one file
2612 system to another place in the same file system could not be
2613 detected as mount, since they shared struct stat's st_dev
2616 * We will now mount the cgroup controllers cpu, cpuacct,
2617 cpuset and the controllers net_cls, net_prio together by
2620 * nspawn containers will now have a virtualized boot
2621 ID. (i.e. /proc/sys/kernel/random/boot_id is now mounted
2622 over with a randomized ID at container initialization). This
2623 has the effect of making "journalctl -b" do the right thing
2626 * The JSON output journal serialization has been updated not
2627 to generate "endless" list objects anymore, but rather one
2628 JSON object per line. This is more in line how most JSON
2629 parsers expect JSON objects. The new output mode
2630 "json-pretty" has been added to provide similar output, but
2631 neatly aligned for readability by humans.
2633 * We dropped all explicit sync() invocations in the shutdown
2634 code. The kernel does this implicitly anyway in the kernel
2635 reboot() syscall. halt(8)'s -n option is now a compatibility
2638 * We now support virtualized reboot() in containers, as
2639 supported by newer kernels. We will fall back to exit() if
2640 CAP_SYS_REBOOT is not available to the container. Also,
2641 nspawn makes use of this now and will actually reboot the
2642 container if the containerized OS asks for that.
2644 * journalctl will only show local log output by default
2645 now. Use --merge (-m) to show remote log output, too.
2647 * libsystemd-journal gained the new sd_journal_get_usage()
2648 call to determine the current disk usage of all journal
2649 files. This is exposed in the new "journalctl --disk-usage"
2652 * journald gained a new configuration setting SplitMode= in
2653 journald.conf which may be used to control how user journals
2654 are split off. See journald.conf(5) for details.
2656 * A new condition type ConditionFileNotEmpty= has been added.
2658 * tmpfiles' "w" lines now support file globbing, to write
2659 multiple files at once.
2661 * We added Python bindings for the journal submission
2662 APIs. More Python APIs for a number of selected APIs will
2663 likely follow. Note that we intend to add native bindings
2664 only for the Python language, as we consider it common
2665 enough to deserve bindings shipped within systemd. There are
2666 various projects outside of systemd that provide bindings
2667 for languages such as PHP or Lua.
2669 * Many conditions will now resolve specifiers such as %i. In
2670 addition, PathChanged= and related directives of .path units
2671 now support specifiers as well.
2673 * There's now a new RPM macro definition for the system preset
2676 * journald will now warn if it ca not forward a message to the
2677 syslog daemon because its socket is full.
2679 * timedated will no longer write or process /etc/timezone,
2680 except on Debian. As we do not support late mounted /usr
2681 anymore /etc/localtime always being a symlink is now safe,
2682 and hence the information in /etc/timezone is not necessary
2685 * logind will now always reserve one VT for a text getty (VT6
2686 by default). Previously if more than 6 X sessions where
2687 started they took up all the VTs with auto-spawned gettys,
2688 so that no text gettys were available anymore.
2690 * udev will now automatically inform the btrfs kernel logic
2691 about btrfs RAID components showing up. This should make
2692 simple hotplug based btrfs RAID assembly work.
2694 * PID 1 will now increase its RLIMIT_NOFILE to 64K by default
2695 (but not for its children which will stay at the kernel
2696 default). This should allow setups with a lot more listening
2699 * systemd will now always pass the configured timezone to the
2700 kernel at boot. timedated will do the same when the timezone
2703 * logind's inhibition logic has been updated. By default,
2704 logind will now handle the lid switch, the power and sleep
2705 keys all the time, even in graphical sessions. If DEs want
2706 to handle these events on their own they should take the new
2707 handle-power-key, handle-sleep-key and handle-lid-switch
2708 inhibitors during their runtime. A simple way to achiveve
2709 that is to invoke the DE wrapped in an invocation of:
2711 systemd-inhibit --what=handle-power-key:handle-sleep-key:handle-lid-switch ...
2713 * Access to unit operations is now checked via SELinux taking
2714 the unit file label and client process label into account.
2716 * systemd will now notify the administrator in the journal
2717 when he over-mounts a non-empty directory.
2719 * There are new specifiers that are resolved in unit files,
2720 for the host name (%H), the machine ID (%m) and the boot ID
2723 Contributions from: Allin Cottrell, Auke Kok, Brandon Philips,
2724 Colin Guthrie, Colin Walters, Daniel J Walsh, Dave Reisner,
2725 Eelco Dolstra, Jan Engelhardt, Kay Sievers, Lennart
2726 Poettering, Lucas De Marchi, Lukas Nykryn, Mantas Mikulėnas,
2727 Martin Pitt, Matthias Clasen, Michael Olbrich, Pierre Schmitz,
2728 Shawn Landden, Thomas Hindoe Paaboel Andersen, Tom Gundersen,
2729 Václav Pavlín, Yin Kangkai, Zbigniew Jędrzejewski-Szmek
2733 * Support for reading structured kernel messages from
2734 /dev/kmsg has now been added and is enabled by default.
2736 * Support for reading kernel messages from /proc/kmsg has now
2737 been removed. If you want kernel messages in the journal
2738 make sure to run a recent kernel (>= 3.5) that supports
2739 reading structured messages from /dev/kmsg (see
2740 above). /proc/kmsg is now exclusive property of classic
2741 syslog daemons again.
2743 * The libudev API gained the new
2744 udev_device_new_from_device_id() call.
2746 * The logic for file system namespace (ReadOnlyDirectory=,
2747 ReadWriteDirectoy=, PrivateTmp=) has been reworked not to
2748 require pivot_root() anymore. This means fewer temporary
2749 directories are created below /tmp for this feature.
2751 * nspawn containers will now see and receive all submounts
2752 made on the host OS below the root file system of the
2755 * Forward Secure Sealing is now supported for Journal files,
2756 which provide cryptographical sealing of journal files so
2757 that attackers cannot alter log history anymore without this
2758 being detectable. Lennart will soon post a blog story about
2759 this explaining it in more detail.
2761 * There are two new service settings RestartPreventExitStatus=
2762 and SuccessExitStatus= which allow configuration of exit
2763 status (exit code or signal) which will be excepted from the
2764 restart logic, resp. consider successful.
2766 * journalctl gained the new --verify switch that can be used
2767 to check the integrity of the structure of journal files and
2768 (if Forward Secure Sealing is enabled) the contents of
2771 * nspawn containers will now be run with /dev/stdin, /dev/fd/
2772 and similar symlinks pre-created. This makes running shells
2773 as container init process a lot more fun.
2775 * The fstab support can now handle PARTUUID= and PARTLABEL=
2778 * A new ConditionHost= condition has been added to match
2779 against the hostname (with globs) and machine ID. This is
2780 useful for clusters where a single OS image is used to
2781 provision a large number of hosts which shall run slightly
2782 different sets of services.
2784 * Services which hit the restart limit will now be placed in a
2787 Contributions from: Bertram Poettering, Dave Reisner, Huang
2788 Hang, Kay Sievers, Lennart Poettering, Lukas Nykryn, Martin
2789 Pitt, Simon Peeters, Zbigniew Jędrzejewski-Szmek
2793 * When running in --user mode systemd will now become a
2794 subreaper (PR_SET_CHILD_SUBREAPER). This should make the ps
2795 tree a lot more organized.
2797 * A new PartOf= unit dependency type has been introduced that
2798 may be used to group services in a natural way.
2800 * "systemctl enable" may now be used to enable instances of
2803 * journalctl now prints error log levels in red, and
2804 warning/notice log levels in bright white. It also supports
2805 filtering by log level now.
2807 * cgtop gained a new -n switch (similar to top), to configure
2808 the maximum number of iterations to run for. It also gained
2809 -b, to run in batch mode (accepting no input).
2811 * The suffix ".service" may now be omitted on most systemctl
2812 command lines involving service unit names.
2814 * There's a new bus call in logind to lock all sessions, as
2815 well as a loginctl verb for it "lock-sessions".
2817 * libsystemd-logind.so gained a new call sd_journal_perror()
2818 that works similar to libc perror() but logs to the journal
2819 and encodes structured information about the error number.
2821 * /etc/crypttab entries now understand the new keyfile-size=
2824 * shutdown(8) now can send a (configurable) wall message when
2825 a shutdown is cancelled.
2827 * The mount propagation mode for the root file system will now
2828 default to "shared", which is useful to make containers work
2829 nicely out-of-the-box so that they receive new mounts from
2830 the host. This can be undone locally by running "mount
2831 --make-rprivate /" if needed.
2833 * The prefdm.service file has been removed. Distributions
2834 should maintain this unit downstream if they intend to keep
2835 it around. However, we recommend writing normal unit files
2836 for display managers instead.
2838 * Since systemd is a crucial part of the OS we will now
2839 default to a number of compiler switches that improve
2840 security (hardening) such as read-only relocations, stack
2841 protection, and suchlike.
2843 * The TimeoutSec= setting for services is now split into
2844 TimeoutStartSec= and TimeoutStopSec= to allow configuration
2845 of individual time outs for the start and the stop phase of
2848 Contributions from: Artur Zaprzala, Arvydas Sidorenko, Auke
2849 Kok, Bryan Kadzban, Dave Reisner, David Strauss, Harald Hoyer,
2850 Jim Meyering, Kay Sievers, Lennart Poettering, Mantas
2851 Mikulėnas, Martin Pitt, Michal Schmidt, Michal Sekletar, Peter
2852 Alfredsen, Shawn Landden, Simon Peeters, Terence Honles, Tom
2853 Gundersen, Zbigniew Jędrzejewski-Szmek
2857 * The journal and id128 C APIs are now fully documented as man
2860 * Extra safety checks have been added when transitioning from
2861 the initial RAM disk to the main system to avoid accidental
2864 * /etc/crypttab entries now understand the new keyfile-offset=
2867 * systemctl -t can now be used to filter by unit load state.
2869 * The journal C API gained the new sd_journal_wait() call to
2870 make writing synchronous journal clients easier.
2872 * journalctl gained the new -D switch to show journals from a
2875 * journalctl now displays a special marker between log
2876 messages of two different boots.
2878 * The journal is now explicitly flushed to /var via a service
2879 systemd-journal-flush.service, rather than implicitly simply
2880 by seeing /var/log/journal to be writable.
2882 * journalctl (and the journal C APIs) can now match for much
2883 more complex expressions, with alternatives and
2886 * When transitioning from the initial RAM disk to the main
2887 system we will now kill all processes in a killing spree to
2888 ensure no processes stay around by accident.
2890 * Three new specifiers may be used in unit files: %u, %h, %s
2891 resolve to the user name, user home directory resp. user
2892 shell. This is useful for running systemd user instances.
2894 * We now automatically rotate journal files if their data
2895 object hash table gets a fill level > 75%. We also size the
2896 hash table based on the configured maximum file size. This
2897 together should lower hash collisions drastically and thus
2898 speed things up a bit.
2900 * journalctl gained the new "--header" switch to introspect
2901 header data of journal files.
2903 * A new setting SystemCallFilters= has been added to services
2904 which may be used to apply blacklists or whitelists to
2905 system calls. This is based on SECCOMP Mode 2 of Linux 3.5.
2907 * nspawn gained a new --link-journal= switch (and quicker: -j)
2908 to link the container journal with the host. This makes it
2909 very easy to centralize log viewing on the host for all
2910 guests while still keeping the journal files separated.
2912 * Many bugfixes and optimizations
2914 Contributions from: Auke Kok, Eelco Dolstra, Harald Hoyer, Kay
2915 Sievers, Lennart Poettering, Malte Starostik, Paul Menzel, Rex
2916 Tsai, Shawn Landden, Tom Gundersen, Ville Skyttä, Zbigniew
2921 * Several tools now understand kernel command line arguments,
2922 which are only read when run in an initial RAM disk. They
2923 usually follow closely their normal counterparts, but are
2926 * There's a new tool to analyze the readahead files that are
2927 automatically generated at boot. Use:
2929 /usr/lib/systemd/systemd-readahead analyze /.readahead
2931 * We now provide an early debug shell on tty9 if this enabled. Use:
2933 systemctl enable debug-shell.service
2935 * All plymouth related units have been moved into the Plymouth
2936 package. Please make sure to upgrade your Plymouth version
2939 * systemd-tmpfiles now supports getting passed the basename of
2940 a configuration file only, in which case it will look for it
2941 in all appropriate directories automatically.
2943 * udevadm info now takes a /dev or /sys path as argument, and
2944 does the right thing. Example:
2946 udevadm info /dev/sda
2947 udevadm info /sys/class/block/sda
2949 * systemctl now prints a warning if a unit is stopped but a
2950 unit that might trigger it continues to run. Example: a
2951 service is stopped but the socket that activates it is left
2954 * "systemctl status" will now mention if the log output was
2955 shortened due to rotation since a service has been started.
2957 * The journal API now exposes functions to determine the
2958 "cutoff" times due to rotation.
2960 * journald now understands SIGUSR1 and SIGUSR2 for triggering
2961 immediately flushing of runtime logs to /var if possible,
2962 resp. for triggering immediate rotation of the journal
2965 * It is now considered an error if a service is attempted to
2966 be stopped that is not loaded.
2968 * XDG_RUNTIME_DIR now uses numeric UIDs instead of usernames.
2970 * systemd-analyze now supports Python 3
2972 * tmpfiles now supports cleaning up directories via aging
2973 where the first level dirs are always kept around but
2974 directories beneath it automatically aged. This is enabled
2975 by prefixing the age field with '~'.
2977 * Seat objects now expose CanGraphical, CanTTY properties
2978 which is required to deal with very fast bootups where the
2979 display manager might be running before the graphics drivers
2980 completed initialization.
2982 * Seat objects now expose a State property.
2984 * We now include RPM macros for service enabling/disabling
2985 based on the preset logic. We recommend RPM based
2986 distributions to make use of these macros if possible. This
2987 makes it simpler to reuse RPM spec files across
2990 * We now make sure that the collected systemd unit name is
2991 always valid when services log to the journal via
2994 * There's a new man page kernel-command-line(7) detailing all
2995 command line options we understand.
2997 * The fstab generator may now be disabled at boot by passing
2998 fstab=0 on the kernel command line.
3000 * A new kernel command line option modules-load= is now understood
3001 to load a specific kernel module statically, early at boot.
3003 * Unit names specified on the systemctl command line are now
3004 automatically escaped as needed. Also, if file system or
3005 device paths are specified they are automatically turned
3006 into the appropriate mount or device unit names. Example:
3008 systemctl status /home
3009 systemctl status /dev/sda
3011 * The SysVConsole= configuration option has been removed from
3012 system.conf parsing.
3014 * The SysV search path is no longer exported on the D-Bus
3017 * The Names= option is been removed from unit file parsing.
3019 * There's a new man page bootup(7) detailing the boot process.
3021 * Every unit and every generator we ship with systemd now
3022 comes with full documentation. The self-explanatory boot is
3025 * A couple of services gained "systemd-" prefixes in their
3026 name if they wrap systemd code, rather than only external
3027 code. Among them fsck@.service which is now
3028 systemd-fsck@.service.
3030 * The HaveWatchdog property has been removed from the D-Bus
3033 * systemd.confirm_spawn= on the kernel command line should now
3036 * There's a new man page crypttab(5) which details all options
3037 we actually understand.
3039 * systemd-nspawn gained a new --capability= switch to pass
3040 additional capabilities to the container.
3042 * timedated will now read known NTP implementation unit names
3043 from /usr/lib/systemd/ntp-units.d/*.list,
3044 systemd-timedated-ntp.target has been removed.
3046 * journalctl gained a new switch "-b" that lists log data of
3047 the current boot only.
3049 * The notify socket is in the abstract namespace again, in
3050 order to support daemons which chroot() at start-up.
3052 * There is a new Storage= configuration option for journald
3053 which allows configuration of where log data should go. This
3054 also provides a way to disable journal logging entirely, so
3055 that data collected is only forwarded to the console, the
3056 kernel log buffer or another syslog implementation.
3058 * Many bugfixes and optimizations
3060 Contributions from: Auke Kok, Colin Guthrie, Dave Reisner,
3061 David Strauss, Eelco Dolstra, Kay Sievers, Lennart Poettering,
3062 Lukas Nykryn, Michal Schmidt, Michal Sekletar, Paul Menzel,
3063 Shawn Landden, Tom Gundersen
3067 * "systemctl help <unit>" now shows the man page if one is
3070 * Several new man pages have been added.
3072 * MaxLevelStore=, MaxLevelSyslog=, MaxLevelKMsg=,
3073 MaxLevelConsole= can now be specified in
3074 journald.conf. These options allow reducing the amount of
3075 data stored on disk or forwarded by the log level.
3077 * TimerSlackNSec= can now be specified in system.conf for
3078 PID1. This allows system-wide power savings.
3080 Contributions from: Dave Reisner, Kay Sievers, Lauri Kasanen,
3081 Lennart Poettering, Malte Starostik, Marc-Antoine Perennou,
3086 * logind is now capable of (optionally) handling power and
3087 sleep keys as well as the lid switch.
3089 * journalctl now understands the syntax "journalctl
3090 /usr/bin/avahi-daemon" to get all log output of a specific
3093 * CapabilityBoundingSet= in system.conf now also influences
3094 the capability bound set of usermode helpers of the kernel.
3096 Contributions from: Daniel Drake, Daniel J. Walsh, Gert
3097 Michael Kulyk, Harald Hoyer, Jean Delvare, Kay Sievers,
3098 Lennart Poettering, Matthew Garrett, Matthias Clasen, Paul
3099 Menzel, Shawn Landden, Tero Roponen, Tom Gundersen
3103 * Note that we skipped 139 releases here in order to set the
3104 new version to something that is greater than both udev's
3105 and systemd's most recent version number.
3107 * udev: all udev sources are merged into the systemd source tree now.
3108 All future udev development will happen in the systemd tree. It
3109 is still fully supported to use the udev daemon and tools without
3110 systemd running, like in initramfs or other init systems. Building
3111 udev though, will require the *build* of the systemd tree, but
3112 udev can be properly *run* without systemd.
3114 * udev: /lib/udev/devices/ are not read anymore; systemd-tmpfiles
3115 should be used to create dead device nodes as workarounds for broken
3118 * udev: RUN+="socket:..." and udev_monitor_new_from_socket() is
3119 no longer supported. udev_monitor_new_from_netlink() needs to be
3120 used to subscribe to events.
3122 * udev: when udevd is started by systemd, processes which are left
3123 behind by forking them off of udev rules, are unconditionally cleaned
3124 up and killed now after the event handling has finished. Services or
3125 daemons must be started as systemd services. Services can be
3126 pulled-in by udev to get started, but they can no longer be directly
3127 forked by udev rules.
3129 * udev: the daemon binary is called systemd-udevd now and installed
3130 in /usr/lib/systemd/. Standalone builds or non-systemd systems need
3131 to adapt to that, create symlink, or rename the binary after building
3134 * libudev no longer provides these symbols:
3135 udev_monitor_from_socket()
3136 udev_queue_get_failed_list_entry()
3137 udev_get_{dev,sys,run}_path()
3138 The versions number was bumped and symbol versioning introduced.
3140 * systemd-loginctl and systemd-journalctl have been renamed
3141 to loginctl and journalctl to match systemctl.
3143 * The config files: /etc/systemd/systemd-logind.conf and
3144 /etc/systemd/systemd-journald.conf have been renamed to
3145 logind.conf and journald.conf. Package updates should rename
3146 the files to the new names on upgrade.
3148 * For almost all files the license is now LGPL2.1+, changed
3149 from the previous GPL2.0+. Exceptions are some minor stuff
3150 of udev (which will be changed to LGPL2.1 eventually, too),
3151 and the MIT licensed sd-daemon.[ch] library that is suitable
3152 to be used as drop-in files.
3154 * systemd and logind now handle system sleep states, in
3155 particular suspending and hibernating.
3157 * logind now implements a sleep/shutdown/idle inhibiting logic
3158 suitable for a variety of uses. Soonishly Lennart will blog
3159 about this in more detail.
3161 * var-run.mount and var-lock.mount are no longer provided
3162 (which prevously bind mounted these directories to their new
3163 places). Distributions which have not converted these
3164 directories to symlinks should consider stealing these files
3165 from git history and add them downstream.
3167 * We introduced the Documentation= field for units and added
3168 this to all our shipped units. This is useful to make it
3169 easier to explore the boot and the purpose of the various
3172 * All smaller setup units (such as
3173 systemd-vconsole-setup.service) now detect properly if they
3174 are run in a container and are skipped when
3175 appropriate. This guarantees an entirely noise-free boot in
3176 Linux container environments such as systemd-nspawn.
3178 * A framework for implementing offline system updates is now
3179 integrated, for details see:
3180 http://freedesktop.org/wiki/Software/systemd/SystemUpdates
3182 * A new service type Type=idle is available now which helps us
3183 avoiding ugly interleaving of getty output and boot status
3186 * There's now a system-wide CapabilityBoundingSet= option to
3187 globally reduce the set of capabilities for the
3188 system. This is useful to drop CAP_SYS_MKNOD, CAP_SYS_RAWIO,
3189 CAP_NET_RAW, CAP_SYS_MODULE, CAP_SYS_TIME, CAP_SYS_PTRACE or
3190 even CAP_NET_ADMIN system-wide for secure systems.
3192 * There are now system-wide DefaultLimitXXX= options to
3193 globally change the defaults of the various resource limits
3194 for all units started by PID 1.
3196 * Harald Hoyer's systemd test suite has been integrated into
3197 systemd which allows easy testing of systemd builds in qemu
3198 and nspawn. (This is really awesome! Ask us for details!)
3200 * The fstab parser is now implemented as generator, not inside
3203 * systemctl will now warn you if .mount units generated from
3204 /etc/fstab are out of date due to changes in fstab that
3205 have not been read by systemd yet.
3207 * systemd is now suitable for usage in initrds. Dracut has
3208 already been updated to make use of this. With this in place
3209 initrds get a slight bit faster but primarily are much
3210 easier to introspect and debug since "systemctl status" in
3211 the host system can be used to introspect initrd services,
3212 and the journal from the initrd is kept around too.
3214 * systemd-delta has been added, a tool to explore differences
3215 between user/admin configuration and vendor defaults.
3217 * PrivateTmp= now affects both /tmp and /var/tmp.
3219 * Boot time status messages are now much prettier and feature
3220 proper english language. Booting up systemd has never been
3223 * Read-ahead pack files now include the inode number of all
3224 files to pre-cache. When the inode changes the pre-caching
3225 is not attempted. This should be nicer to deal with updated
3226 packages which might result in changes of read-ahead
3229 * We now temporaritly lower the kernel's read_ahead_kb variable
3230 when collecting read-ahead data to ensure the kernel's
3231 built-in read-ahead does not add noise to our measurements
3232 of necessary blocks to pre-cache.
3234 * There's now RequiresMountsFor= to add automatic dependencies
3235 for all mounts necessary for a specific file system path.
3237 * MountAuto= and SwapAuto= have been removed from
3238 system.conf. Mounting file systems at boot has to take place
3241 * nspawn now learned a new switch --uuid= to set the machine
3242 ID on the command line.
3244 * nspawn now learned the -b switch to automatically search
3247 * vt102 is now the default TERM for serial TTYs, upgraded from
3250 * systemd-logind now works on VT-less systems.
3252 * The build tree has been reorganized. The individual
3253 components now have directories of their own.
3255 * A new condition type ConditionPathIsReadWrite= is now available.
3257 * nspawn learned the new -C switch to create cgroups for the
3258 container in other hierarchies.
3260 * We now have support for hardware watchdogs, configurable in
3263 * The scheduled shutdown logic now has a public API.
3265 * We now mount /tmp as tmpfs by default, but this can be
3266 masked and /etc/fstab can override it.
3268 * Since udisks does not make use of /media anymore we are not
3269 mounting a tmpfs on it anymore.
3271 * journalctl gained a new --local switch to only interleave
3272 locally generated journal files.
3274 * We can now load the IMA policy at boot automatically.
3276 * The GTK tools have been split off into a systemd-ui.
3278 Contributions from: Andreas Schwab, Auke Kok, Ayan George,
3279 Colin Guthrie, Daniel Mack, Dave Reisner, David Ward, Elan
3280 Ruusamäe, Frederic Crozat, Gergely Nagy, Guillermo Vidal,
3281 Hannes Reinecke, Harald Hoyer, Javier Jardón, Kay Sievers,
3282 Lennart Poettering, Lucas De Marchi, Léo Gillot-Lamure,
3283 Marc-Antoine Perennou, Martin Pitt, Matthew Monaco, Maxim
3284 A. Mikityanskiy, Michael Biebl, Michael Olbrich, Michal
3285 Schmidt, Nis Martensen, Patrick McCarty, Roberto Sassu, Shawn
3286 Landden, Sjoerd Simons, Sven Anders, Tollef Fog Heen, Tom
3291 * This is mostly a bugfix release
3293 * Support optional initialization of the machine ID from the
3294 KVM or container configured UUID.
3296 * Support immediate reboots with "systemctl reboot -ff"
3298 * Show /etc/os-release data in systemd-analyze output
3300 * Many bugfixes for the journal, including endianness fixes and
3301 ensuring that disk space enforcement works
3303 * sd-login.h is C++ comptaible again
3305 * Extend the /etc/os-release format on request of the Debian
3308 * We now refuse non-UTF8 strings used in various configuration
3309 and unit files. This is done to ensure we do not pass invalid
3310 data over D-Bus or expose it elsewhere.
3312 * Register Mimo USB Screens as suitable for automatic seat
3315 * Read SELinux client context from journal clients in a race
3318 * Reorder configuration file lookup order. /etc now always
3319 overrides /run in order to allow the administrator to always
3320 and unconditionally override vendor supplied or
3321 automatically generated data.
3323 * The various user visible bits of the journal now have man
3324 pages. We still lack man pages for the journal API calls
3327 * We now ship all man pages in HTML format again in the
3330 Contributions from: Dave Reisner, Dirk Eibach, Frederic
3331 Crozat, Harald Hoyer, Kay Sievers, Lennart Poettering, Marti
3332 Raudsepp, Michal Schmidt, Shawn Landden, Tero Roponen, Thierry
3337 * This is mostly a bugfix release
3339 * systems lacking /etc/os-release are no longer supported.
3341 * Various functionality updates to libsystemd-login.so
3343 * Track class of PAM logins to distuingish greeters from
3346 Contributions from: Kay Sievers, Lennart Poettering, Michael
3351 * This is an important bugfix release for v41.
3353 * Building man pages is now optional which should be useful
3354 for those building systemd from git but unwilling to install
3357 * Watchdog support for supervising services is now usable. In
3358 a future release support for hardware watchdogs
3359 (i.e. /dev/watchdog) will be added building on this.
3361 * Service start rate limiting is now configurable and can be
3362 turned off per service. When a start rate limit is hit a
3363 reboot can automatically be triggered.
3365 * New CanReboot(), CanPowerOff() bus calls in systemd-logind.
3367 Contributions from: Benjamin Franzke, Bill Nottingham,
3368 Frederic Crozat, Lennart Poettering, Michael Olbrich, Michal
3369 Schmidt, Michał Górny, Piotr Drąg
3373 * The systemd binary is installed /usr/lib/systemd/systemd now;
3374 An existing /sbin/init symlink needs to be adapted with the
3377 * The code that loads kernel modules has been ported to invoke
3378 libkmod directly, instead of modprobe. This means we do not
3379 support systems with module-init-tools anymore.
3381 * Watchdog support is now already useful, but still not
3384 * A new kernel command line option systemd.setenv= is
3385 understood to set system wide environment variables
3386 dynamically at boot.
3388 * We now limit the set of capabilities of systemd-journald.
3390 * We now set SIGPIPE to ignore by default, since it only is
3391 useful in shell pipelines, and has little use in general
3392 code. This can be disabled with IgnoreSIPIPE=no in unit
3395 Contributions from: Benjamin Franzke, Kay Sievers, Lennart
3396 Poettering, Michael Olbrich, Michal Schmidt, Tom Gundersen,
3401 * This is mostly a bugfix release
3403 * We now expose the reason why a service failed in the
3404 "Result" D-Bus property.
3406 * Rudimentary service watchdog support (will be completed over
3407 the next few releases.)
3409 * When systemd forks off in order execute some service we will
3410 now immediately changes its argv[0] to reflect which process
3411 it will execute. This is useful to minimize the time window
3412 with a generic argv[0], which makes bootcharts more useful
3414 Contributions from: Alvaro Soliverez, Chris Paulson-Ellis, Kay
3415 Sievers, Lennart Poettering, Michael Olbrich, Michal Schmidt,
3416 Mike Kazantsev, Ray Strode
3420 * This is mostly a test release, but incorporates many
3423 * New systemd-cgtop tool to show control groups by their
3426 * Linking against libacl for ACLs is optional again. If
3427 disabled, support tracking device access for active logins
3428 goes becomes unavailable, and so does access to the user
3429 journals by the respective users.
3431 * If a group "adm" exists, journal files are automatically
3432 owned by them, thus allow members of this group full access
3433 to the system journal as well as all user journals.
3435 * The journal now stores the SELinux context of the logging
3436 client for all entries.
3438 * Add C++ inclusion guards to all public headers
3440 * New output mode "cat" in the journal to print only text
3441 messages, without any meta data like date or time.
3443 * Include tiny X server wrapper as a temporary stop-gap to
3444 teach XOrg udev display enumeration. This is used by display
3445 managers such as gdm, and will go away as soon as XOrg
3446 learned native udev hotplugging for display devices.
3448 * Add new systemd-cat tool for executing arbitrary programs
3449 with STDERR/STDOUT connected to the journal. Can also act as
3450 BSD logger replacement, and does so by default.
3452 * Optionally store all locally generated coredumps in the
3453 journal along with meta data.
3455 * systemd-tmpfiles learnt four new commands: n, L, c, b, for
3456 writing short strings to files (for usage for /sys), and for
3457 creating symlinks, character and block device nodes.
3459 * New unit file option ControlGroupPersistent= to make cgroups
3460 persistent, following the mechanisms outlined in
3461 http://www.freedesktop.org/wiki/Software/systemd/PaxControlGroups
3463 * Support multiple local RTCs in a sane way
3465 * No longer monopolize IO when replaying readahead data on
3466 rotating disks, since we might starve non-file-system IO to
3467 death, since fanotify() will not see accesses done by blkid,
3470 * Do not show kernel threads in systemd-cgls anymore, unless
3471 requested with new -k switch.
3473 Contributions from: Dan Horák, Kay Sievers, Lennart
3474 Poettering, Michal Schmidt
3478 * This is mostly a test release, but incorporates many
3481 * The git repository moved to:
3482 git://anongit.freedesktop.org/systemd/systemd
3483 ssh://git.freedesktop.org/git/systemd/systemd
3485 * First release with the journal
3486 http://0pointer.de/blog/projects/the-journal.html
3488 * The journal replaces both systemd-kmsg-syslogd and
3489 systemd-stdout-bridge.
3491 * New sd_pid_get_unit() API call in libsystemd-logind
3493 * Many systemadm clean-ups
3495 * Introduce remote-fs-pre.target which is ordered before all
3496 remote mounts and may be used to start services before all
3499 * Added Mageia support
3501 * Add bash completion for systemd-loginctl
3503 * Actively monitor PID file creation for daemons which exit in
3504 the parent process before having finished writing the PID
3505 file in the daemon process. Daemons which do this need to be
3506 fixed (i.e. PID file creation must have finished before the
3507 parent exits), but we now react a bit more gracefully to them.
3509 * Add colourful boot output, mimicking the well-known output
3510 of existing distributions.
3512 * New option PassCredentials= for socket units, for
3513 compatibility with a recent kernel ABI breakage.
3515 * /etc/rc.local is now hooked in via a generator binary, and
3516 thus will no longer act as synchronization point during
3519 * systemctl list-unit-files now supports --root=.
3521 * systemd-tmpfiles now understands two new commands: z, Z for
3522 relabelling files according to the SELinux database. This is
3523 useful to apply SELinux labels to specific files in /sys,
3526 * Output of SysV services is now forwarded to both the console
3527 and the journal by default, not only just the console.
3529 * New man pages for all APIs from libsystemd-login.
3531 * The build tree got reorganized and a the build system is a
3532 lot more modular allowing embedded setups to specifically
3533 select the components of systemd they are interested in.
3535 * Support for Linux systems lacking the kernel VT subsystem is
3538 * configure's --with-rootdir= got renamed to
3539 --with-rootprefix= to follow the naming used by udev and
3542 * Unless specified otherwise we will now install to /usr instead
3543 of /usr/local by default.
3545 * Processes with '@' in argv[0][0] are now excluded from the
3546 final shut-down killing spree, following the logic explained
3548 http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons
3550 * All processes remaining in a service cgroup when we enter
3551 the START or START_PRE states are now killed with
3552 SIGKILL. That means it is no longer possible to spawn
3553 background processes from ExecStart= lines (which was never
3554 supported anyway, and bad style).
3556 * New PropagateReloadTo=/PropagateReloadFrom= options to bind
3557 reloading of units together.
3559 Contributions from: Bill Nottingham, Daniel J. Walsh, Dave
3560 Reisner, Dexter Morgan, Gregs Gregs, Jonathan Nieder, Kay
3561 Sievers, Lennart Poettering, Michael Biebl, Michal Schmidt,
3562 Michał Górny, Ran Benita, Thomas Jarosch, Tim Waugh, Tollef
3563 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek