1 systemd System and Service Manager
5 * The DHCP implementation of systemd-networkd gained a set of
8 - The DHCP server now supports emitting DNS and NTP
9 information. It may be enabled and configured via
10 EmitDNS=, DNS=, EmitNTP=, and NTP=. If transmission of DNS
11 and NTP information is enabled, but no servers are
12 configured, the corresponding uplink information (if there
13 is any) is propagated.
15 - Server and client now support transmission and reception
16 of timezone information. It can be configured via the
17 newly introduced network options UseTimezone=,
18 EmitTimezone=, and Timezone=. Transmission of timezone
19 information is enabled between host and containers by
20 default now: the container will change its local timezone
21 to what the host has set.
23 - Lease timeouts can now be configured via
24 MaxLeaseTimeSec= and DefaultLeaseTimeSec=.
26 - The DHCP server improved on the stability of
27 leases. Clients are more likely to get the same lease
28 information back, even if the server loses state.
30 - The DHCP server supports two new configuration options to
31 control the lease address pool metrics, PoolOffset= and
34 * The encapsulation limit of tunnels in systemd-networkd may
35 now be configured via 'EncapsulationLimit='. It allows
36 modifying the maximum additional levels of encapsulation
37 that are permitted to be prepended to a packet.
39 * systemd now supports the concept of user buses replacing
40 session buses, if used with dbus-1.10 (and enabled via dbus
41 --enable-user-session). It previously only supported this on
42 kdbus-enabled systems, and this release expands this to
43 'dbus-daemon' systems.
45 * systemd-networkd now supports predictable interface names
48 * systemd now optionally supports the new Linux kernel
49 "unified" control group hierarchy. If enabled via the kernel
50 command-line option 'systemd.unified_cgroup_hierarchy=1',
51 systemd will try to mount the unified cgroup hierarchy
52 directly on /sys/fs/cgroup. If not enabled, or not
53 available, systemd will fall back to the legacy cgroup
54 hierarchy setup, as before. Host system and containers can
55 mix and match legacy and unified hierarchies as they
56 wish. nspawn understands the $UNIFIED_CROUP_HIERARCHY
57 environment variable to individually select the hierarchy to
58 use for executed containers. By default, nspawn will use the
59 unified hierarchy for the containers if the host uses the
60 unified hierarchy, and the legacy hierarchy otherwise.
61 Please note that at this point the unified hierarchy is an
62 experimental kernel feature and is likely to change in one
63 of the next kernel releases. Therefore, it should not be
64 enabled by default in downstream distributions yet. The
65 minimum required kernel version for the unified hierarchy to
66 work is 4.2. Note that when the unified hierarchy is used
67 for the first time delegated access to controllers is
68 safe. Because of this systemd-nspawn containers will get
69 access to controllers now, as will systemd user
70 sessions. This means containers and user sessions may now
71 manage their own resources, partitioning up what the system
74 * A new special scope unit "init.scope" has been introduced
75 that encapsulates PID 1 of the system. It may be used to
76 determine resource usage and enforce resource limits on PID
77 1 itself. PID 1 hence moved out of the root of the control
80 * The cgtop tool gained support for filtering out kernel
81 threads when counting tasks in a control group. Also, the
82 count of processes is now recursively summed up by
83 default. Two options -k and --recursive= have been added to
84 revert to old behaviour. The tool has also been updated to
85 work correctly in containers now.
87 * systemd-nspawn's --bind= and --bind-ro= options have been
88 extended to allow creation of non-recursive bind mounts.
90 * libsystemd gained two new calls sd_pid_get_cgroup() and
91 sd_peer_get_cgroup() which return the control group path of
92 a process or peer of a connected AF_UNIX socket. This
93 function call is particularly useful when implementing
94 delegated subtrees support in the control group hierarchy.
96 * The "sd-event" event loop API of libsystemd now supports
97 correct dequeuing of real-time signals, without losing
100 * When systemd requests a PolicyKit decision when managing
101 units it will now add additional fields to the request,
102 including unit name and desired operation. This enables more
103 powerful PolicyKit policies, that make decisions depending
106 * nspawn learnt support for .nspawn settings files, that may
107 accompany the image files or directories of containers, and
108 may contain additional settings for the container. This is
109 an alternative to configuring container parameters via the
112 Contributions from: Cristian Rodríguez, Daniel Mack, David
113 Herrmann, Eugene Yakubovich, Evgeny Vereshchagin, Filipe
114 Brandenburger, Hans de Goede, Jan Alexander Steffens, Jan
115 Synacek, Kay Sievers, Lennart Poettering, Mangix, Marcel
116 Holtmann, Martin Pitt, Michael Biebl, Michael Chapman, Michal
117 Sekletar, Peter Hutterer, Piotr Drąg, reverendhomer, Robin
118 Hack, Susant Sahani, Sylvain Pasche, Thomas Hindoe Paaboel
119 Andersen, Tom Gundersen, Torstein Husebø
121 -- Berlin, 2015-09-08
125 * machinectl gained a new verb 'shell' which opens a fresh
126 shell on the target container or the host. It is similar to
127 the existing 'login' command of machinectl, but spawns the
128 shell directly without prompting for username or
129 password. The pseudo machine '.host' now refers to the local
130 host and is used by default. Hence, 'machinectl shell' can
131 be used as replacement for 'su -' which spawns a session as
132 a fresh systemd unit in a way that is fully isolated from
133 the originating session.
135 * systemd-networkd learned to cope with private-zone DHCP
136 options and allows other programs to query the values.
138 * SELinux access control when enabling/disabling units is no
139 longer enforced with this release. The previous
140 implementation was incorrect, and a new corrected
141 implementation is not yet available. As unit file operations
142 are still protected via PolicyKit and D-Bus policy this is
143 not a security problem. Yet, distributions which care about
144 optimal SELinux support should probably not stabilize on
147 * sd-bus gained support for matches of type "arg0has=", that
148 test for membership of strings in string arrays sent in bus
151 * systemd-resolved now dumps the contents of its DNS and LLMNR
152 caches to the logs on reception of the SIGUSR1 signal. This
153 is useful to debug DNS behaviour.
155 * The coredumpctl tool gained a new --directory= option to
156 operate on journal files in a specific directory.
158 * "systemctl reboot" and related commands gained a new
159 "--message=" option which may be used to set a free-text
160 wall message when shutting down or rebooting the
161 system. This message is also logged, which is useful for
162 figuring out the reason for a reboot or shutdown a
165 * The "systemd-resolve-host" tool's -i switch now takes
166 network interface numbers as alternative to interface names.
168 * A new unit file setting for services has been introduced:
169 UtmpMode= allows configuration of how precisely systemd
170 handles utmp and wtmp entries for the service if this is
171 enabled. This allows writing services that appear similar to
172 user sessions in the output of the "w", "who", "last" and
175 * systemd-resolved will now locally synthesize DNS resource
176 records for the "localhost" and "gateway" domains as well as
177 the local hostname. This should ensure that clients querying
178 RRs via resolved will get similar results as those going via
179 NSS, if nss-myhostname is enabled.
181 Contributions from: Alastair Hughes, Alex Crawford, Daniel
182 Mack, David Herrmann, Dimitri John Ledkov, Eric Kostrowski,
183 Evgeny Vereshchagin, Felipe Sateler, HATAYAMA Daisuke, Jan
184 Pokorný, Jan Synacek, Johnny Robeson, Karel Zak, Kay Sievers,
185 Kefeng Wang, Lennart Poettering, Major Hayden, Marcel
186 Holtmann, Markus Elfring, Martin Mikkelsen, Martin Pitt, Matt
187 Turner, Maxim Mikityanskiy, Michael Biebl, Namhyung Kim,
188 Nicolas Cornu, Owen W. Taylor, Patrik Flykt, Peter Hutterer,
189 reverendhomer, Richard Maw, Ronny Chevalier, Seth Jennings,
190 Stef Walter, Susant Sahani, Thomas Blume, Thomas Hindoe
191 Paaboel Andersen, Thomas Meyer, Tom Gundersen, Vincent Batts,
192 WaLyong Cho, Zbigniew Jędrzejewski-Szmek
194 -- Berlin, 2015-08-27
198 * The systemd-efi-boot-generator functionality was merged into
199 systemd-gpt-auto-generator.
201 * systemd-networkd now supports Group Policy for vxlan
202 devices. It can be enabled via the new boolean configuration
203 option called 'GroupPolicyExtension='.
205 Contributions from: Andreas Kempf, Christian Hesse, Daniel Mack, David
206 Herrmann, Herman Fries, Johannes Nixdorf, Kay Sievers, Lennart
207 Poettering, Peter Hutterer, Susant Sahani, Tom Gundersen
209 -- Berlin, 2015-07-31
213 * The python-systemd code has been removed from the systemd repository.
214 A new repository has been created which accommodates the code from
215 now on, and we kindly ask distributions to create a separate package
216 for this: https://github.com/systemd/python-systemd
218 * The systemd daemon will now reload its main configuration
219 (/etc/systemd/system.conf) on daemon-reload.
221 * sd-dhcp now exposes vendor specific extensions via
222 sd_dhcp_lease_get_vendor_specific().
224 * systemd-networkd gained a number of new configuration options.
226 - A new boolean configuration option for TAP devices called
227 'VNetHeader='. If set, the IFF_VNET_HDR flag is set for the
228 device, thus allowing to send and receive GSO packets.
230 - A new tunnel configuration option called 'CopyDSCP='.
231 If enabled, the DSCP field of ip6 tunnels is copied into the
234 - A set of boolean bridge configuration options were added.
235 'UseBPDU=', 'HairPin=', 'FastLeave=', 'AllowPortToBeRoot=',
236 and 'UnicastFlood=' are now parsed by networkd and applied to the
237 respective bridge link device via the respective IFLA_BRPORT_*
240 - A new string configuration option to override the hostname sent
241 to a DHCP server, called 'Hostname='. If set and 'SendHostname='
242 is true, networkd will use the configured hostname instead of the
243 system hostname when sending DHCP requests.
245 - A new tunnel configuration option called 'IPv6FlowLabel='. If set,
246 networkd will configure the IPv6 flow-label of the tunnel device
247 according to RFC2460.
249 - The 'macvtap' virtual network devices are now supported, similar to
250 the already supported 'macvlan' devices.
252 * systemd-resolved now implements RFC5452 to improve resilience against
253 cache poisoning. Additionally, source port randomization is enabled
254 by default to further protect against DNS spoofing attacks.
256 * nss-mymachines now supports translating UIDs and GIDs of running
257 containers with user-namespaces enabled. If a container 'foo'
258 translates a host uid 'UID' to the container uid 'TUID', then
259 nss-mymachines will also map uid 'UID' to/from username 'vu-foo-TUID'
260 (with 'foo' and 'TUID' replaced accordingly). Similarly, groups are
261 mapped as 'vg-foo-TGID'.
263 Contributions from: Beniamino Galvani, cee1, Christian Hesse, Daniel
264 Buch, Daniel Mack, daurnimator, David Herrmann, Dimitri John Ledkov,
265 HATAYAMA Daisuke, Ivan Shapovalov, Jan Alexander Steffens (heftig),
266 Johan Ouwerkerk, Jose Carlos Venegas Munoz, Karel Zak, Kay Sievers,
267 Lennart Poettering, Lidong Zhong, Martin Pitt, Michael Biebl, Michael
268 Olbrich, Michal Schmidt, Michal Sekletar, Mike Gilbert, Namhyung Kim,
269 Nick Owens, Peter Hutterer, Richard Maw, Steven Allen, Sungbae Yoo,
270 Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel Andersen, Tom
271 Gundersen, Torstein Husebø, Umut Tezduyar Lindskog, Vito Caputo,
272 Vivenzio Pagliari, Zbigniew Jędrzejewski-Szmek
274 -- Berlin, 2015-07-29
278 * udev does not longer support the WAIT_FOR_SYSFS= key in udev rules.
279 There are no known issues with current sysfs, and udev does not need
280 or should be used to work around such bugs.
282 * udev does no longer enable USB HID power management. Several reports
283 indicate, that some devices cannot handle that setting.
285 * The udev accelerometer helper was removed. The functionality
286 is now fully included in iio-sensor-proxy. But this means,
287 older iio-sensor-proxy versions will no longer provide
288 accelerometer/orientation data with this systemd version.
289 Please upgrade iio-sensor-proxy to version 1.0.
291 * networkd gained a new configuration option IPv6PrivacyExtensions=
292 which enables IPv6 privacy extensions (RFC 4941, "Privacy Extensions
293 for Stateless Address") on selected networks.
295 * For the sake of fewer build-time dependencies and less code in the
296 main repository, the python bindings are about to be removed in the
297 next release. A new repository has been created which accommodates
298 the code from now on, and we kindly ask distributions to create a
299 separate package for this. The removal will take place in v223.
301 https://github.com/systemd/python-systemd
303 Contributions from: Abdo Roig-Maranges, Andrew Eikum, Bastien Nocera,
304 Cédric Delmas, Christian Hesse, Christos Trochalakis, Daniel Mack,
305 daurnimator, David Herrmann, Dimitri John Ledkov, Eric Biggers, Eric
306 Cook, Felipe Sateler, Geert Jansen, Gerd Hoffmann, Gianpaolo Macario,
307 Greg Kroah-Hartman, Iago López Galeiras, Jan Alexander Steffens
308 (heftig), Jan Engelhardt, Jay Strict, Kay Sievers, Lennart Poettering,
309 Markus Knetschke, Martin Pitt, Michael Biebl, Michael Marineau, Michal
310 Sekletar, Miguel Bernal Marin, Peter Hutterer, Richard Maw, rinrinne,
311 Susant Sahani, Thomas Hindoe Paaboel Andersen, Tom Gundersen, Torstein
312 Husebø, Vedran Miletić, WaLyong Cho, Zbigniew Jędrzejewski-Szmek
314 -- Berlin, 2015-07-07
318 * The sd-bus.h and sd-event.h APIs have now been declared
319 stable and have been added to the official interface of
320 libsystemd.so. sd-bus implements an alternative D-Bus client
321 library, that is relatively easy to use, very efficient and
322 supports both classic D-Bus as well as kdbus as transport
323 backend. sd-event is a generic event loop abstraction that
324 is built around Linux epoll, but adds features such as event
325 prioritization or efficient timer handling. Both APIs are good
326 choices for C programs looking for a bus and/or event loop
327 implementation that is minimal and does not have to be
328 portable to other kernels.
330 * kdbus support is no longer compile-time optional. It is now
331 always built-in. However, it can still be disabled at
332 runtime using the kdbus=0 kernel command line setting, and
333 that setting may be changed to default to off, by specifying
334 --disable-kdbus at build-time. Note though that the kernel
335 command line setting has no effect if the kdbus.ko kernel
336 module is not installed, in which case kdbus is (obviously)
337 also disabled. We encourage all downstream distributions to
338 begin testing kdbus by adding it to the kernel images in the
339 development distributions, and leaving kdbus support in
342 * The minimal required util-linux version has been bumped to
345 * Support for chkconfig (--enable-chkconfig) was removed in
346 favor of calling an abstraction tool
347 /lib/systemd/systemd-sysv-install. This needs to be
348 implemented for your distribution. See "SYSV INIT.D SCRIPTS"
349 in README for details.
351 * If there's a systemd unit and a SysV init script for the
352 same service name, and the user executes "systemctl enable"
353 for it (or a related call), then this will now enable both
354 (or execute the related operation on both), not just the
357 * The libudev API documentation has been converted from gtkdoc
360 * gudev has been removed from the systemd tree, it is now an
363 * The systemd-cgtop tool learnt a new --raw switch to generate
364 "raw" (machine parsable) output.
366 * networkd's IPForwarding= .network file setting learnt the
367 new setting "kernel", which ensures that networkd does not
368 change the IP forwarding sysctl from the default kernel
371 * The systemd-logind bus API now exposes a new boolean
372 property "Docked" that reports whether logind considers the
373 system "docked", i.e. connected to a docking station or not.
375 Contributions from: Alex Crawford, Andreas Pokorny, Andrei
376 Borzenkov, Charles Duffy, Colin Guthrie, Cristian Rodríguez,
377 Daniele Medri, Daniel Hahler, Daniel Mack, David Herrmann,
378 David Mohr, Dimitri John Ledkov, Djalal Harouni, dslul, Ed
379 Swierk, Eric Cook, Filipe Brandenburger, Gianpaolo Macario,
380 Harald Hoyer, Iago López Galeiras, Igor Vuk, Jan Synacek,
381 Jason Pleau, Jason S. McMullan, Jean Delvare, Jeff Huang,
382 Jonathan Boulle, Karel Zak, Kay Sievers, kloun, Lennart
383 Poettering, Marc-Antoine Perennou, Marcel Holtmann, Mario
384 Limonciello, Martin Pitt, Michael Biebl, Michael Olbrich,
385 Michal Schmidt, Mike Gilbert, Nick Owens, Pablo Lezaeta Reyes,
386 Patrick Donnelly, Pavel Odvody, Peter Hutterer, Philip
387 Withnall, Ronny Chevalier, Simon McVittie, Susant Sahani,
388 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Torstein
389 Husebø, Umut Tezduyar Lindskog, Viktar Vauchkevich, Werner
390 Fink, Zbigniew Jędrzejewski-Szmek
392 -- Berlin, 2015-06-19
396 * The gudev library has been extracted into a separate repository
397 available at: https://git.gnome.org/browse/libgudev/
398 It is now managed as part of the Gnome project. Distributions
399 are recommended to pass --disable-gudev to systemd and use
400 gudev from the Gnome project instead. gudev is still included
401 in systemd, for now. It will be removed soon, though. Please
402 also see the announcement-thread on systemd-devel:
403 http://lists.freedesktop.org/archives/systemd-devel/2015-May/032070.html
405 * systemd now exposes a CPUUsageNSec= property for each
406 service unit on the bus, that contains the overall consumed
407 CPU time of a service (the sum of what each process of the
408 service consumed). This value is only available if
409 CPUAccounting= is turned on for a service, and is then shown
410 in the "systemctl status" output.
412 * Support for configuring alternative mappings of the old SysV
413 runlevels to systemd targets has been removed. They are now
414 hardcoded in a way that runlevels 2, 3, 4 all map to
415 multi-user.target and 5 to graphical.target (which
416 previously was already the default behaviour).
418 * The auto-mounter logic gained support for mount point
419 expiry, using a new TimeoutIdleSec= setting in .automount
420 units. (Also available as x-systemd.idle-timeout= in /etc/fstab).
422 * The EFI System Partition (ESP) as mounted to /boot by
423 systemd-efi-boot-generator will now be unmounted
424 automatically after 2 minutes of not being used. This should
425 minimize the risk of ESP corruptions.
427 * New /etc/fstab options x-systemd.requires= and
428 x-systemd.requires-mounts-for= are now supported to express
429 additional dependencies for mounts. This is useful for
430 journalling file systems that support external journal
431 devices or overlay file systems that require underlying file
432 systems to be mounted.
434 * systemd does not support direct live-upgrades (via systemctl
435 daemon-reexec) from versions older than v44 anymore. As no
436 distribution we are aware of shipped such old versions in a
437 stable release this should not be problematic.
439 * When systemd forks off a new per-connection service instance
440 it will now set the $REMOTE_ADDR environment variable to the
441 remote IP address, and $REMOTE_PORT environment variable to
442 the remote IP port. This behaviour is similar to the
443 corresponding environment variables defined by CGI.
445 * systemd-networkd gained support for uplink failure
446 detection. The BindCarrier= option allows binding interface
447 configuration dynamically to the link sense of other
448 interfaces. This is useful to achieve behaviour like in
451 * systemd-networkd gained support for configuring the DHCP
452 client identifier to use when requesting leases.
454 * systemd-networkd now has a per-network UseNTP= option to
455 configure whether NTP server information acquired via DHCP
456 is passed on to services like systemd-timesyncd.
458 * systemd-networkd gained support for vti6 tunnels.
460 * Note that systemd-networkd manages the sysctl variable
461 /proc/sys/net/ipv[46]/conf/*/forwarding for each interface
462 it is configured for since v219. The variable controls IP
463 forwarding, and is a per-interface alternative to the global
464 /proc/sys/net/ipv[46]/ip_forward. This setting is
465 configurable in the IPForward= option, which defaults to
466 "no". This means if networkd is used for an interface it is
467 no longer sufficient to set the global sysctl option to turn
468 on IP forwarding! Instead, the .network file option
469 IPForward= needs to be turned on! Note that the
470 implementation of this behaviour was broken in v219 and has
473 * Many bonding and vxlan options are now configurable in
476 * systemd-nspawn gained a new --property= setting to set unit
477 properties for the container scope. This is useful for
478 setting resource parameters (e.g "CPUShares=500") on
479 containers started from the command line.
481 * systemd-nspawn gained a new --private-users= switch to make
482 use of user namespacing available on recent Linux kernels.
484 * systemd-nspawn may now be called as part of a shell pipeline
485 in which case the pipes used for stdin and stdout are passed
486 directly to the process invoked in the container, without
487 indirection via a pseudo tty.
489 * systemd-nspawn gained a new switch to control the UNIX
490 signal to use when killing the init process of the container
493 * systemd-nspawn gained a new --overlay= switch for mounting
494 overlay file systems into the container using the new kernel
497 * When a container image is imported via systemd-importd and
498 the host file system is not btrfs, a loopback block device
499 file is created in /var/lib/machines.raw with a btrfs file
500 system inside. It is then mounted to /var/lib/machines to
501 enable btrfs features for container management. The loopback
502 file and btrfs file system is grown as needed when container
503 images are imported via systemd-importd.
505 * systemd-machined/systemd-importd gained support for btrfs
506 quota, to enforce container disk space limits on disk. This
507 is exposed in "machinectl set-limit".
509 * systemd-importd now can import containers from local .tar,
510 .raw and .qcow2 images, and export them to .tar and .raw. It
511 can also import dkr v2 images now from the network (on top
514 * systemd-importd gained support for verifying downloaded
515 images with gpg2 (previously only gpg1 was supported).
517 * systemd-machined, systemd-logind, systemd: most bus calls
518 are now accessible to unprivileged processes via
519 PolicyKit. Also, systemd-logind will now allow users to kill
520 their own sessions without further privileges or
523 * systemd-shutdownd has been removed. This service was
524 previously responsible for implementing scheduled shutdowns
525 as exposed in /usr/bin/shutdown's time parameter. This
526 functionality has now been moved into systemd-logind and is
527 accessible via a bus interface.
529 * "systemctl reboot" gained a new switch --firmware-setup that
530 can be used to reboot into the EFI firmware setup, if that
531 is available. systemd-logind now exposes an API on the bus
532 to trigger such reboots, in case graphical desktop UIs want
533 to cover this functionality.
535 * "systemctl enable", "systemctl disable" and "systemctl mask"
536 now support a new "--now" switch. If specified the units
537 that are enabled will also be started, and the ones
538 disabled/masked also stopped.
540 * The Gummiboot EFI boot loader tool has been merged into
541 systemd, and renamed to "systemd-boot". The bootctl tool has been
542 updated to support systemd-boot.
544 * An EFI kernel stub has been added that may be used to create
545 kernel EFI binaries that contain not only the actual kernel,
546 but also an initrd, boot splash, command line and OS release
547 information. This combined binary can then be signed as a
548 single image, so that the firmware can verify it all in one
549 step. systemd-boot has special support for EFI binaries created
550 like this and can extract OS release information from them
551 and show them in the boot menu. This functionality is useful
552 to implement cryptographically verified boot schemes.
554 * Optional support has been added to systemd-fsck to pass
555 fsck's progress report to an AF_UNIX socket in the file
558 * udev will no longer create device symlinks for all block
559 devices by default. A blacklist for excluding special block
560 devices from this logic has been turned into a whitelist
561 that requires picking block devices explicitly that require
564 * A new (currently still internal) API sd-device.h has been
565 added to libsystemd. This modernized API is supposed to
566 replace libudev eventually. In fact, already much of libudev
567 is now just a wrapper around sd-device.h.
569 * A new hwdb database for storing metadata about pointing
570 stick devices has been added.
572 * systemd-tmpfiles gained support for setting file attributes
573 similar to the "chattr" tool with new 'h' and 'H' lines.
575 * systemd-journald will no longer unconditionally set the
576 btrfs NOCOW flag on new journal files. This is instead done
577 with tmpfiles snippet using the new 'h' line type. This
578 allows easy disabling of this logic, by masking the
579 journal-nocow.conf tmpfiles file.
581 * systemd-journald will now translate audit message types to
582 human readable identifiers when writing them to the
583 journal. This should improve readability of audit messages.
585 * The LUKS logic gained support for the offset= and skip=
586 options in /etc/crypttab, as previously implemented by
589 * /usr/lib/os-release gained a new optional field VARIANT= for
590 distributions that support multiple variants (such as a
591 desktop edition, a server edition, ...)
593 Contributions from: Aaro Koskinen, Adam Goode, Alban Crequy,
594 Alberto Fanjul Alonso, Alexander Sverdlin, Alex Puchades, Alin
595 Rauta, Alison Chaiken, Andrew Jones, Arend van Spriel,
596 Benedikt Morbach, Benjamin Franzke, Benjamin Tissoires, Blaž
597 Tomažič, Chris Morgan, Chris Morin, Colin Walters, Cristian
598 Rodríguez, Daniel Buch, Daniel Drake, Daniele Medri, Daniel
599 Mack, Daniel Mustieles, daurnimator, Davide Bettio, David
600 Herrmann, David Strauss, Didier Roche, Dimitri John Ledkov,
601 Eric Cook, Gavin Li, Goffredo Baroncelli, Hannes Reinecke,
602 Hans de Goede, Hans-Peter Deifel, Harald Hoyer, Iago López
603 Galeiras, Ivan Shapovalov, Jan Engelhardt, Jan Janssen, Jan
604 Pazdziora, Jan Synacek, Jasper St. Pierre, Jay Faulkner, John
605 Paul Adrian Glaubitz, Jonathon Gilbert, Karel Zak, Kay
606 Sievers, Koen Kooi, Lennart Poettering, Lubomir Rintel, Lucas
607 De Marchi, Lukas Nykryn, Lukas Rusak, Lukasz Skalski, Łukasz
608 Stelmach, Mantas Mikulėnas, Marc-Antoine Perennou, Marcel
609 Holtmann, Martin Pitt, Mathieu Chevrier, Matthew Garrett,
610 Michael Biebl, Michael Marineau, Michael Olbrich, Michal
611 Schmidt, Michal Sekletar, Mirco Tischler, Nir Soffer, Patrik
612 Flykt, Pavel Odvody, Peter Hutterer, Peter Lemenkov, Peter
613 Waller, Piotr Drąg, Raul Gutierrez S, Richard Maw, Ronny
614 Chevalier, Ross Burton, Sebastian Rasmussen, Sergey Ptashnick,
615 Seth Jennings, Shawn Landden, Simon Farnsworth, Stefan Junker,
616 Stephen Gallagher, Susant Sahani, Sylvain Plantefève, Thomas
617 Haller, Thomas Hindoe Paaboel Andersen, Tobias Hunger, Tom
618 Gundersen, Torstein Husebø, Umut Tezduyar Lindskog, Will
619 Woods, Zachary Cook, Zbigniew Jędrzejewski-Szmek
621 -- Berlin, 2015-05-22
625 * Introduce a new API "sd-hwdb.h" for querying the hardware
626 metadata database. With this minimal interface one can query
627 and enumerate the udev hwdb, decoupled from the old libudev
628 library. libudev's interface for this is now only a wrapper
629 around sd-hwdb. A new tool systemd-hwdb has been added to
630 interface with and update the database.
632 * When any of systemd's tools copies files (for example due to
633 tmpfiles' C lines) a btrfs reflink will attempted first,
634 before bytewise copying is done.
636 * systemd-nspawn gained a new --ephemeral switch. When
637 specified a btrfs snapshot is taken of the container's root
638 directory, and immediately removed when the container
639 terminates again. Thus, a container can be started whose
640 changes never alter the container's root directory, and are
641 lost on container termination. This switch can also be used
642 for starting a container off the root file system of the
643 host without affecting the host OS. This switch is only
644 available on btrfs file systems.
646 * systemd-nspawn gained a new --template= switch. It takes the
647 path to a container tree to use as template for the tree
648 specified via --directory=, should that directory be
649 missing. This allows instantiating containers dynamically,
650 on first run. This switch is only available on btrfs file
653 * When a .mount unit refers to a mount point on which multiple
654 mounts are stacked, and the .mount unit is stopped all of
655 the stacked mount points will now be unmounted until no
658 * systemd now has an explicit notion of supported and
659 unsupported unit types. Jobs enqueued for unsupported unit
660 types will now fail with an "unsupported" error code. More
661 specifically .swap, .automount and .device units are not
662 supported in containers, .busname units are not supported on
663 non-kdbus systems. .swap and .automount are also not
664 supported if their respective kernel compile time options
667 * machinectl gained support for two new "copy-from" and
668 "copy-to" commands for copying files from a running
669 container to the host or vice versa.
671 * machinectl gained support for a new "bind" command to bind
672 mount host directories into local containers. This is
673 currently only supported for nspawn containers.
675 * networkd gained support for configuring bridge forwarding
676 database entries (fdb) from .network files.
678 * A new tiny daemon "systemd-importd" has been added that can
679 download container images in tar, raw, qcow2 or dkr formats,
680 and make them available locally in /var/lib/machines, so
681 that they can run as nspawn containers. The daemon can GPG
682 verify the downloads (not supported for dkr, since it has no
683 provisions for verifying downloads). It will transparently
684 decompress bz2, xz, gzip compressed downloads if necessary,
685 and restore sparse files on disk. The daemon uses privilege
686 separation to ensure the actual download logic runs with
687 fewer privileges than the daemon itself. machinectl has
688 gained new commands "pull-tar", "pull-raw" and "pull-dkr" to
689 make the functionality of importd available to the
690 user. With this in place the Fedora and Ubuntu "Cloud"
691 images can be downloaded and booted as containers unmodified
692 (the Fedora images lack the appropriate GPG signature files
693 currently, so they cannot be verified, but this will change
694 soon, hopefully). Note that downloading images is currently
695 only fully supported on btrfs.
697 * machinectl is now able to list container images found in
698 /var/lib/machines, along with some metadata about sizes of
699 disk and similar. If the directory is located on btrfs and
700 quota is enabled, this includes quota display. A new command
701 "image-status" has been added that shows additional
702 information about images.
704 * machinectl is now able to clone container images
705 efficiently, if the underlying file system (btrfs) supports
706 it, with the new "machinectl list-images" command. It also
707 gained commands for renaming and removing images, as well as
708 marking them read-only or read-write (supported also on
709 legacy file systems).
711 * networkd gained support for collecting LLDP network
712 announcements, from hardware that supports this. This is
713 shown in networkctl output.
715 * systemd-run gained support for a new -t (--pty) switch for
716 invoking a binary on a pty whose input and output is
717 connected to the invoking terminal. This allows executing
718 processes as system services while interactively
719 communicating with them via the terminal. Most interestingly
720 this is supported across container boundaries. Invoking
721 "systemd-run -t /bin/bash" is an alternative to running a
722 full login session, the difference being that the former
723 will not register a session, nor go through the PAM session
726 * tmpfiles gained support for a new "v" line type for creating
727 btrfs subvolumes. If the underlying file system is a legacy
728 file system, this automatically degrades to creating a
729 normal directory. Among others /var/lib/machines is now
730 created like this at boot, should it be missing.
732 * The directory /var/lib/containers/ has been deprecated and
733 been replaced by /var/lib/machines. The term "machines" has
734 been used in the systemd context as generic term for both
735 VMs and containers, and hence appears more appropriate for
736 this, as the directory can also contain raw images bootable
739 * systemd-nspawn when invoked with -M but without --directory=
740 or --image= is now capable of searching for the container
741 root directory, subvolume or disk image automatically, in
742 /var/lib/machines. systemd-nspawn@.service has been updated
743 to make use of this, thus allowing it to be used for raw
746 * A new machines.target unit has been introduced that is
747 supposed to group all containers/VMs invoked as services on
748 the system. systemd-nspawn@.service has been updated to
751 * machinectl gained a new "start" command, for invoking a
752 container as a service. "machinectl start foo" is mostly
753 equivalent to "systemctl start systemd-nspawn@foo.service",
754 but handles escaping in a nicer way.
756 * systemd-nspawn will now mount most of the cgroupfs tree
757 read-only into each container, with the exception of the
758 container's own subtree in the name=systemd hierarchy.
760 * journald now sets the special FS_NOCOW file flag for its
761 journal files. This should improve performance on btrfs, by
762 avoiding heavy fragmentation when journald's write-pattern
763 is used on COW file systems. It degrades btrfs' data
764 integrity guarantees for the files to the same levels as for
765 ext3/ext4 however. This should be OK though as journald does
766 its own data integrity checks and all its objects are
767 checksummed on disk. Also, journald should handle btrfs disk
768 full events a lot more gracefully now, by processing SIGBUS
769 errors, and not relying on fallocate() anymore.
771 * When journald detects that journal files it is writing to
772 have been deleted it will immediately start new journal
775 * systemd now provides a way to store file descriptors
776 per-service in PID 1.This is useful for daemons to ensure
777 that fds they require are not lost during a daemon
778 restart. The fds are passed to the daemon on the next
779 invocation in the same way socket activation fds are
780 passed. This is now used by journald to ensure that the
781 various sockets connected to all the system's stdout/stderr
782 are not lost when journald is restarted. File descriptors
783 may be stored in PID 1 via the sd_pid_notify_with_fds() API,
784 an extension to sd_notify(). Note that a limit is enforced
785 on the number of fds a service can store in PID 1, and it
786 defaults to 0, so that no fds may be stored, unless this is
787 explicitly turned on.
789 * The default TERM variable to use for units connected to a
790 terminal, when no other value is explicitly is set is now
791 vt220 rather than vt102. This should be fairly safe still,
792 but allows PgUp/PgDn work.
794 * The /etc/crypttab option header= as known from Debian is now
797 * "loginctl user-status" and "loginctl session-status" will
798 now show the last 10 lines of log messages of the
799 user/session following the status output. Similar,
800 "machinectl status" will show the last 10 log lines
801 associated with a virtual machine or container
802 service. (Note that this is usually not the log messages
803 done in the VM/container itself, but simply what the
804 container manager logs. For nspawn this includes all console
807 * "loginctl session-status" without further argument will now
808 show the status of the session of the caller. Similar,
809 "lock-session", "unlock-session", "activate",
810 "enable-linger", "disable-linger" may now be called without
811 session/user parameter in which case they apply to the
812 caller's session/user.
814 * An X11 session scriptlet is now shipped that uploads
815 $DISPLAY and $XAUTHORITY into the environment of the systemd
816 --user daemon if a session begins. This should improve
817 compatibility with X11 enabled applications run as systemd
820 * Generators are now subject to masking via /etc and /run, the
821 same way as unit files.
823 * networkd .network files gained support for configuring
824 per-link IPv4/IPv6 packet forwarding as well as IPv4
825 masquerading. This is by default turned on for veth links to
826 containers, as registered by systemd-nspawn. This means that
827 nspawn containers run with --network-veth will now get
828 automatic routed access to the host's networks without any
829 further configuration or setup, as long as networkd runs on
832 * systemd-nspawn gained the --port= (-p) switch to expose TCP
833 or UDP posts of a container on the host. With this in place
834 it is possible to run containers with private veth links
835 (--network-veth), and have their functionality exposed on
836 the host as if their services were running directly on the
839 * systemd-nspawn's --network-veth switch now gained a short
840 version "-n", since with the changes above it is now truly
841 useful out-of-the-box. The systemd-nspawn@.service has been
842 updated to make use of it too by default.
844 * systemd-nspawn will now maintain a per-image R/W lock, to
845 ensure that the same image is not started more than once
846 writable. (It's OK to run an image multiple times
847 simultaneously in read-only mode.)
849 * systemd-nspawn's --image= option is now capable of
850 dissecting and booting MBR and GPT disk images that contain
851 only a single active Linux partition. Previously it
852 supported only GPT disk images with proper GPT type
853 IDs. This allows running cloud images from major
854 distributions directly with systemd-nspawn, without
857 * In addition to collecting mouse dpi data in the udev
858 hardware database, there's now support for collecting angle
859 information for mouse scroll wheels. The database is
860 supposed to guarantee similar scrolling behavior on mice
861 that it knows about. There's also support for collecting
862 information about Touchpad types.
864 * udev's input_id built-in will now also collect touch screen
865 dimension data and attach it to probed devices.
867 * /etc/os-release gained support for a Distribution Privacy
870 * networkd gained support for creating "ipvlan", "gretap",
871 "ip6gre", "ip6gretap" and "ip6tnl" network devices.
873 * systemd-tmpfiles gained support for "a" lines for setting
876 * systemd-nspawn will now mount /tmp in the container to
877 tmpfs, automatically.
879 * systemd now exposes the memory.usage_in_bytes cgroup
880 attribute and shows it for each service in the "systemctl
881 status" output, if available.
883 * When the user presses Ctrl-Alt-Del more than 7x within 2s an
884 immediate reboot is triggered. This useful if shutdown is
885 hung and is unable to complete, to expedite the
886 operation. Note that this kind of reboot will still unmount
887 all file systems, and hence should not result in fsck being
890 * A .device unit for an optical block device will now be
891 considered active only when a medium is in the drive. Also,
892 mount units are now bound to their backing devices thus
893 triggering automatic unmounting when devices become
894 unavailable. With this in place systemd will now
895 automatically unmount left-over mounts when a CD-ROM is
896 ejected or an USB stick is yanked from the system.
898 * networkd-wait-online now has support for waiting for
899 specific interfaces only (with globbing), and for giving up
900 after a configurable timeout.
902 * networkd now exits when idle. It will be automatically
903 restarted as soon as interfaces show up, are removed or
904 change state. networkd will stay around as long as there is
905 at least one DHCP state machine or similar around, that keep
908 * networkd may now configure IPv6 link-local addressing in
909 addition to IPv4 link-local addressing.
911 * The IPv6 "token" for use in SLAAC may now be configured for
912 each .network interface in networkd.
914 * Routes configured with networkd may now be assigned a scope
917 * networkd's [Match] sections now support globbing and lists
918 of multiple space-separated matches per item.
920 Contributions from: Alban Crequy, Alin Rauta, Andrey Chaser,
921 Bastien Nocera, Bruno Bottazzini, Carlos Garnacho, Carlos
922 Morata Castillo, Chris Atkinson, Chris J. Arges, Christian
923 Kirbach, Christian Seiler, Christoph Brill, Colin Guthrie,
924 Colin Walters, Cristian Rodríguez, Daniele Medri, Daniel Mack,
925 Dave Reisner, David Herrmann, Djalal Harouni, Erik Auerswald,
926 Filipe Brandenburger, Frank Theile, Gabor Kelemen, Gabriel de
927 Perthuis, Harald Hoyer, Hui Wang, Ivan Shapovalov, Jan
928 Engelhardt, Jan Synacek, Jay Faulkner, Johannes Hölzl, Jonas
929 Ådahl, Jonathan Boulle, Josef Andersson, Kay Sievers, Ken
930 Werner, Lennart Poettering, Lucas De Marchi, Lukas Märdian,
931 Lukas Nykryn, Lukasz Skalski, Luke Shumaker, Mantas Mikulėnas,
932 Manuel Mendez, Marcel Holtmann, Marc Schmitzer, Marko
933 Myllynen, Martin Pitt, Maxim Mikityanskiy, Michael Biebl,
934 Michael Marineau, Michael Olbrich, Michal Schmidt, Mindaugas
935 Baranauskas, Moez Bouhlel, Naveen Kumar, Patrik Flykt, Paul
936 Martin, Peter Hutterer, Peter Mattern, Philippe De Swert,
937 Piotr Drąg, Rafael Ferreira, Rami Rosen, Robert Milasan, Ronny
938 Chevalier, Sangjung Woo, Sebastien Bacher, Sergey Ptashnick,
939 Shawn Landden, Stéphane Graber, Susant Sahani, Sylvain
940 Plantefève, Thomas Hindoe Paaboel Andersen, Tim JP, Tom
941 Gundersen, Topi Miettinen, Torstein Husebø, Umut Tezduyar
942 Lindskog, Veres Lajos, Vincent Batts, WaLyong Cho, Wieland
943 Hoffmann, Zbigniew Jędrzejewski-Szmek
945 -- Berlin, 2015-02-16
949 * When querying unit file enablement status (for example via
950 "systemctl is-enabled"), a new state "indirect" is now known
951 which indicates that a unit might not be enabled itself, but
952 another unit listed in its Also= setting might be.
954 * Similar to the various existing ConditionXYZ= settings for
955 units there are now matching AssertXYZ= settings. While
956 failing conditions cause a unit to be skipped, but its job
957 to succeed, failing assertions declared like this will cause
958 a unit start operation and its job to fail.
960 * hostnamed now knows a new chassis type "embedded".
962 * systemctl gained a new "edit" command. When used on a unit
963 file this allows extending unit files with .d/ drop-in
964 configuration snippets or editing the full file (after
965 copying it from /usr/lib to /etc). This will invoke the
966 user's editor (as configured with $EDITOR), and reload the
967 modified configuration after editing.
969 * "systemctl status" now shows the suggested enablement state
970 for a unit, as declared in the (usually vendor-supplied)
973 * nss-myhostname will now resolve the single-label host name
974 "gateway" to the locally configured default IP routing
975 gateways, ordered by their metrics. This assigns a stable
976 name to the used gateways, regardless which ones are
977 currently configured. Note that the name will only be
978 resolved after all other name sources (if nss-myhostname is
979 configured properly) and should hence not negatively impact
980 systems that use the single-label host name "gateway" in
983 * systemd-inhibit now allows filtering by mode when listing
986 * Scope and service units gained a new "Delegate" boolean
987 property, which when set allows processes running inside the
988 unit to further partition resources. This is primarily
989 useful for systemd user instances as well as container
992 * journald will now pick up audit messages directly from
993 the kernel, and log them like any other log message. The
994 audit fields are split up and fully indexed. This means that
995 journalctl in many ways is now a (nicer!) alternative to
996 ausearch, the traditional audit client. Note that this
997 implements only a minimal audit client, if you want the
998 special audit modes like reboot-on-log-overflow, please use
999 the traditional auditd instead, which can be used in
1000 parallel to journald.
1002 * The ConditionSecurity= unit file option now understands the
1003 special string "audit" to check whether auditing is
1006 * journalctl gained two new commands --vacuum-size= and
1007 --vacuum-time= to delete old journal files until the
1008 remaining ones take up no more the specified size on disk,
1009 or are not older than the specified time.
1011 * A new, native PPPoE library has been added to sd-network,
1012 systemd's library of light-weight networking protocols. This
1013 library will be used in a future version of networkd to
1014 enable PPPoE communication without an external pppd daemon.
1016 * The busctl tool now understands a new "capture" verb that
1017 works similar to "monitor", but writes a packet capture
1018 trace to STDOUT that can be redirected to a file which is
1019 compatible with libcap's capture file format. This can then
1020 be loaded in Wireshark and similar tools to inspect bus
1023 * The busctl tool now understands a new "tree" verb that shows
1024 the object trees of a specific service on the bus, or of all
1027 * The busctl tool now understands a new "introspect" verb that
1028 shows all interfaces and members of objects on the bus,
1029 including their signature and values. This is particularly
1030 useful to get more information about bus objects shown by
1031 the new "busctl tree" command.
1033 * The busctl tool now understands new verbs "call",
1034 "set-property" and "get-property" for invoking bus method
1035 calls, setting and getting bus object properties in a
1038 * busctl gained a new --augment-creds= argument that controls
1039 whether the tool shall augment credential information it
1040 gets from the bus with data from /proc, in a possibly
1043 * nspawn's --link-journal= switch gained two new values
1044 "try-guest" and "try-host" that work like "guest" and
1045 "host", but do not fail if the host has no persistent
1046 journalling enabled. -j is now equivalent to
1047 --link-journal=try-guest.
1049 * macvlan network devices created by nspawn will now have
1050 stable MAC addresses.
1052 * A new SmackProcessLabel= unit setting has been added, which
1053 controls the SMACK security label processes forked off by
1054 the respective unit shall use.
1056 * If compiled with --enable-xkbcommon, systemd-localed will
1057 verify x11 keymap settings by compiling the given keymap. It
1058 will spew out warnings if the compilation fails. This
1059 requires libxkbcommon to be installed.
1061 * When a coredump is collected a larger number of metadata
1062 fields is now collected and included in the journal records
1063 created for it. More specifically control group membership,
1064 environment variables, memory maps, working directory,
1065 chroot directory, /proc/$PID/status, and a list of open file
1066 descriptors is now stored in the log entry.
1068 * The udev hwdb now contains DPI information for mice. For
1071 http://who-t.blogspot.de/2014/12/building-a-dpi-database-for-mice.html
1073 * All systemd programs that read standalone configuration
1074 files in /etc now also support a corresponding series of
1075 .conf.d configuration directories in /etc/, /run/,
1076 /usr/local/lib/, /usr/lib/, and (if configured with
1077 --enable-split-usr) /lib/. In particular, the following
1078 configuration files now have corresponding configuration
1079 directories: system.conf user.conf, logind.conf,
1080 journald.conf, sleep.conf, bootchart.conf, coredump.conf,
1081 resolved.conf, timesyncd.conf, journal-remote.conf, and
1082 journal-upload.conf. Note that distributions should use the
1083 configuration directories in /usr/lib/; the directories in
1084 /etc/ are reserved for the system administrator.
1086 * systemd-rfkill will no longer take the rfkill device name
1087 into account when storing rfkill state on disk, as the name
1088 might be dynamically assigned and not stable. Instead, the
1089 ID_PATH udev variable combined with the rfkill type (wlan,
1090 bluetooth, ...) is used.
1092 * A new service systemd-machine-id-commit.service has been
1093 added. When used on systems where /etc is read-only during
1094 boot, and /etc/machine-id is not initialized (but an empty
1095 file), this service will copy the temporary machine ID
1096 created as replacement into /etc after the system is fully
1097 booted up. This is useful for systems that are freshly
1098 installed with a non-initialized machine ID, but should get
1099 a fixed machine ID for subsequent boots.
1101 * networkd's .netdev files now provide a large set of
1102 configuration parameters for VXLAN devices. Similar, the
1103 bridge port cost parameter is now configurable in .network
1104 files. There's also new support for configuring IP source
1105 routing. networkd .link files gained support for a new
1106 OriginalName= match that is useful to match against the
1107 original interface name the kernel assigned. .network files
1108 may include MTU= and MACAddress= fields for altering the MTU
1109 and MAC address while being connected to a specific network
1112 * The LUKS logic gained supported for configuring
1113 UUID-specific key files. There's also new support for naming
1114 LUKS device from the kernel command line, using the new
1115 luks.name= argument.
1117 * Timer units may now be transiently created via the bus API
1118 (this was previously already available for scope and service
1119 units). In addition it is now possible to create multiple
1120 transient units at the same time with a single bus call. The
1121 "systemd-run" tool has been updated to make use of this for
1122 running commands on a specified time, in at(1)-style.
1124 * tmpfiles gained support for "t" lines, for assigning
1125 extended attributes to files. Among other uses this may be
1126 used to assign SMACK labels to files.
1128 Contributions from: Alin Rauta, Alison Chaiken, Andrej
1129 Manduch, Bastien Nocera, Chris Atkinson, Chris Leech, Chris
1130 Mayo, Colin Guthrie, Colin Walters, Cristian Rodríguez,
1131 Daniele Medri, Daniel Mack, Dan Williams, Dan Winship, Dave
1132 Reisner, David Herrmann, Didier Roche, Felipe Sateler, Gavin
1133 Li, Hans de Goede, Harald Hoyer, Iago López Galeiras, Ivan
1134 Shapovalov, Jakub Filak, Jan Janssen, Jan Synacek, Joe
1135 Lawrence, Josh Triplett, Kay Sievers, Lennart Poettering,
1136 Lukas Nykryn, Łukasz Stelmach, Maciej Wereski, Mantas
1137 Mikulėnas, Marcel Holtmann, Martin Pitt, Maurizio Lombardi,
1138 Michael Biebl, Michael Chapman, Michael Marineau, Michal
1139 Schmidt, Michal Sekletar, Olivier Brunel, Patrik Flykt, Peter
1140 Hutterer, Przemyslaw Kedzierski, Rami Rosen, Ray Strode,
1141 Richard Schütz, Richard W.M. Jones, Ronny Chevalier, Ross
1142 Lagerwall, Sean Young, Stanisław Pitucha, Susant Sahani,
1143 Thomas Haller, Thomas Hindoe Paaboel Andersen, Tom Gundersen,
1144 Torstein Husebø, Umut Tezduyar Lindskog, Vicente Olivert
1145 Riera, WaLyong Cho, Wesley Dawson, Zbigniew Jędrzejewski-Szmek
1147 -- Berlin, 2014-12-10
1151 * journalctl gained the new options -t/--identifier= to match
1152 on the syslog identifier (aka "tag"), as well as --utc to
1153 show log timestamps in the UTC timezone. journalctl now also
1154 accepts -n/--lines=all to disable line capping in a pager.
1156 * journalctl gained a new switch, --flush, that synchronously
1157 flushes logs from /run/log/journal to /var/log/journal if
1158 persistent storage is enabled. systemd-journal-flush.service
1159 now waits until the operation is complete.
1161 * Services can notify the manager before they start a reload
1162 (by sending RELOADING=1) or shutdown (by sending
1163 STOPPING=1). This allows the manager to track and show the
1164 internal state of daemons and closes a race condition when
1165 the process is still running but has closed its D-Bus
1168 * Services with Type=oneshot do not have to have any ExecStart
1171 * User units are now loaded also from
1172 $XDG_RUNTIME_DIR/systemd/user/. This is similar to the
1173 /run/systemd/user directory that was already previously
1174 supported, but is under the control of the user.
1176 * Job timeouts (i.e. time-outs on the time a job that is
1177 queued stays in the run queue) can now optionally result in
1178 immediate reboot or power-off actions (JobTimeoutAction= and
1179 JobTimeoutRebootArgument=). This is useful on ".target"
1180 units, to limit the maximum time a target remains
1181 undispatched in the run queue, and to trigger an emergency
1182 operation in such a case. This is now used by default to
1183 turn off the system if boot-up (as defined by everything in
1184 basic.target) hangs and does not complete for at least
1185 15min. Also, if power-off or reboot hang for at least 30min
1186 an immediate power-off/reboot operation is triggered. This
1187 functionality is particularly useful to increase reliability
1188 on embedded devices, but also on laptops which might
1189 accidentally get powered on when carried in a backpack and
1190 whose boot stays stuck in a hard disk encryption passphrase
1193 * systemd-logind can be configured to also handle lid switch
1194 events even when the machine is docked or multiple displays
1195 are attached (HandleLidSwitchDocked= option).
1197 * A helper binary and a service have been added which can be
1198 used to resume from hibernation in the initramfs. A
1199 generator will parse the resume= option on the kernel
1200 command line to trigger resume.
1202 * A user console daemon systemd-consoled has been
1203 added. Currently, it is a preview, and will so far open a
1204 single terminal on each session of the user marked as
1205 Desktop=systemd-console.
1207 * Route metrics can be specified for DHCP routes added by
1210 * The SELinux context of socket-activated services can be set
1211 from the information provided by the networking stack
1212 (SELinuxContextFromNet= option).
1214 * Userspace firmware loading support has been removed and
1215 the minimum supported kernel version is thus bumped to 3.7.
1217 * Timeout for udev workers has been increased from 1 to 3
1218 minutes, but a warning will be printed after 1 minute to
1219 help diagnose kernel modules that take a long time to load.
1221 * Udev rules can now remove tags on devices with TAG-="foobar".
1223 * systemd's readahead implementation has been removed. In many
1224 circumstances it didn't give expected benefits even for
1225 rotational disk drives and was becoming less relevant in the
1226 age of SSDs. As none of the developers has been using
1227 rotating media anymore, and nobody stepped up to actively
1228 maintain this component of systemd it has now been removed.
1230 * Swap units can use Options= to specify discard options.
1231 Discard options specified for swaps in /etc/fstab are now
1234 * Docker containers are now detected as a separate type of
1237 * The Password Agent protocol gained support for queries where
1238 the user input is shown, useful e.g. for user names.
1239 systemd-ask-password gained a new --echo option to turn that
1242 * The default sysctl.d/ snippets will now set:
1244 net.core.default_qdisc = fq_codel
1246 This selects Fair Queuing Controlled Delay as the default
1247 queuing discipline for network interfaces. fq_codel helps
1248 fight the network bufferbloat problem. It is believed to be
1249 a good default with no tuning required for most workloads.
1250 Downstream distributions may override this choice. On 10Gbit
1251 servers that do not do forwarding, "fq" may perform better.
1252 Systems without a good clocksource should use "pfifo_fast".
1254 * If kdbus is enabled during build a new option BusPolicy= is
1255 available for service units, that allows locking all service
1256 processes into a stricter bus policy, in order to limit
1257 access to various bus services, or even hide most of them
1258 from the service's view entirely.
1260 * networkctl will now show the .network and .link file
1261 networkd has applied to a specific interface.
1263 * sd-login gained a new API call sd_session_get_desktop() to
1264 query which desktop environment has been selected for a
1267 * UNIX utmp support is now compile-time optional to support
1268 legacy-free systems.
1270 * systemctl gained two new commands "add-wants" and
1271 "add-requires" for pulling in units from specific targets
1274 * If the word "rescue" is specified on the kernel command line
1275 the system will now boot into rescue mode (aka
1276 rescue.target), which was previously available only by
1277 specifying "1" or "systemd.unit=rescue.target" on the kernel
1278 command line. This new kernel command line option nicely
1279 mirrors the already existing "emergency" kernel command line
1282 * New kernel command line options mount.usr=, mount.usrflags=,
1283 mount.usrfstype= have been added that match root=, rootflags=,
1284 rootfstype= but allow mounting a specific file system to
1287 * The $NOTIFY_SOCKET is now also passed to control processes of
1288 services, not only the main process.
1290 * This version reenables support for fsck's -l switch. This
1291 means at least version v2.25 of util-linux is required for
1292 operation, otherwise dead-locks on device nodes may
1293 occur. Again: you need to update util-linux to at least
1294 v2.25 when updating systemd to v217.
1296 * The "multi-seat-x" tool has been removed from systemd, as
1297 its functionality has been integrated into X servers 1.16,
1298 and the tool is hence redundant. It is recommended to update
1299 display managers invoking this tool to simply invoke X
1300 directly from now on, again.
1302 * Support for the new ALLOW_INTERACTIVE_AUTHORIZATION D-Bus
1303 message flag has been added for all of systemd's PolicyKit
1304 authenticated method calls has been added. In particular
1305 this now allows optional interactive authorization via
1306 PolicyKit for many of PID1's privileged operations such as
1307 unit file enabling and disabling.
1309 * "udevadm hwdb --update" learnt a new switch "--usr" for
1310 placing the rebuilt hardware database in /usr instead of
1311 /etc. When used only hardware database entries stored in
1312 /usr will be used, and any user database entries in /etc are
1313 ignored. This functionality is useful for vendors to ship a
1314 pre-built database on systems where local configuration is
1315 unnecessary or unlikely.
1317 * Calendar time specifications in .timer units now also
1318 understand the strings "semi-annually", "quarterly" and
1319 "minutely" as shortcuts (in addition to the preexisting
1320 "anually", "hourly", ...).
1322 * systemd-tmpfiles will now correctly create files in /dev
1323 at boot which are marked for creation only at boot. It is
1324 recommended to always create static device nodes with 'c!'
1325 and 'b!', so that they are created only at boot and not
1326 overwritten at runtime.
1328 * When the watchdog logic is used for a service (WatchdogSec=)
1329 and the watchdog timeout is hit the service will now be
1330 terminated with SIGABRT (instead of just SIGTERM), in order
1331 to make sure a proper coredump and backtrace is
1332 generated. This ensures that hanging services will result in
1333 similar coredump/backtrace behaviour as services that hit a
1336 Contributions from: Andreas Henriksson, Andrei Borzenkov,
1337 Angus Gibson, Ansgar Burchardt, Ben Wolsieffer, Brandon L.
1338 Black, Christian Hesse, Cristian Rodríguez, Daniel Buch,
1339 Daniele Medri, Daniel Mack, Dan Williams, Dave Reisner, David
1340 Herrmann, David Sommerseth, David Strauss, Emil Renner
1341 Berthing, Eric Cook, Evangelos Foutras, Filipe Brandenburger,
1342 Gustavo Sverzut Barbieri, Hans de Goede, Harald Hoyer, Hristo
1343 Venev, Hugo Grostabussiat, Ivan Shapovalov, Jan Janssen, Jan
1344 Synacek, Jonathan Liu, Juho Son, Karel Zak, Kay Sievers, Klaus
1345 Purer, Koen Kooi, Lennart Poettering, Lukas Nykryn, Lukasz
1346 Skalski, Łukasz Stelmach, Mantas Mikulėnas, Marcel Holtmann,
1347 Marius Tessmann, Marko Myllynen, Martin Pitt, Michael Biebl,
1348 Michael Marineau, Michael Olbrich, Michael Scherer, Michal
1349 Schmidt, Michal Sekletar, Miroslav Lichvar, Patrik Flykt,
1350 Philippe De Swert, Piotr Drąg, Rahul Sundaram, Richard
1351 Weinberger, Robert Milasan, Ronny Chevalier, Ruben Kerkhof,
1352 Santiago Vila, Sergey Ptashnick, Simon McVittie, Sjoerd
1353 Simons, Stefan Brüns, Steven Allen, Steven Noonan, Susant
1354 Sahani, Sylvain Plantefève, Thomas Hindoe Paaboel Andersen,
1355 Timofey Titovets, Tobias Hunger, Tom Gundersen, Torstein
1356 Husebø, Umut Tezduyar Lindskog, WaLyong Cho, Zbigniew
1359 -- Berlin, 2014-10-28
1363 * timedated no longer reads NTP implementation unit names from
1364 /usr/lib/systemd/ntp-units.d/*.list. Alternative NTP
1365 implementations should add a
1367 Conflicts=systemd-timesyncd.service
1369 to their unit files to take over and replace systemd's NTP
1370 default functionality.
1372 * systemd-sysusers gained a new line type "r" for configuring
1373 which UID/GID ranges to allocate system users/groups
1374 from. Lines of type "u" may now add an additional column
1375 that specifies the home directory for the system user to be
1376 created. Also, systemd-sysusers may now optionally read user
1377 information from STDIN instead of a file. This is useful for
1378 invoking it from RPM preinst scriptlets that need to create
1379 users before the first RPM file is installed since these
1380 files might need to be owned by them. A new
1381 %sysusers_create_inline RPM macro has been introduced to do
1382 just that. systemd-sysusers now updates the shadow files as
1383 well as the user/group databases, which should enhance
1384 compatibility with certain tools like grpck.
1386 * A number of bus APIs of PID 1 now optionally consult
1387 PolicyKit to permit access for otherwise unprivileged
1388 clients under certain conditions. Note that this currently
1389 doesn't support interactive authentication yet, but this is
1390 expected to be added eventually, too.
1392 * /etc/machine-info now has new fields for configuring the
1393 deployment environment of the machine, as well as the
1394 location of the machine. hostnamectl has been updated with
1395 new command to update these fields.
1397 * systemd-timesyncd has been updated to automatically acquire
1398 NTP server information from systemd-networkd, which might
1399 have been discovered via DHCP.
1401 * systemd-resolved now includes a caching DNS stub resolver
1402 and a complete LLMNR name resolution implementation. A new
1403 NSS module "nss-resolve" has been added which can be used
1404 instead of glibc's own "nss-dns" to resolve hostnames via
1405 systemd-resolved. Hostnames, addresses and arbitrary RRs may
1406 be resolved via systemd-resolved D-Bus APIs. In contrast to
1407 the glibc internal resolver systemd-resolved is aware of
1408 multi-homed system, and keeps DNS server and caches separate
1409 and per-interface. Queries are sent simultaneously on all
1410 interfaces that have DNS servers configured, in order to
1411 properly handle VPNs and local LANs which might resolve
1412 separate sets of domain names. systemd-resolved may acquire
1413 DNS server information from systemd-networkd automatically,
1414 which in turn might have discovered them via DHCP. A tool
1415 "systemd-resolve-host" has been added that may be used to
1416 query the DNS logic in resolved. systemd-resolved implements
1417 IDNA and automatically uses IDNA or UTF-8 encoding depending
1418 on whether classic DNS or LLMNR is used as transport. In the
1419 next releases we intend to add a DNSSEC and mDNS/DNS-SD
1420 implementation to systemd-resolved.
1422 * A new NSS module nss-mymachines has been added, that
1423 automatically resolves the names of all local registered
1424 containers to their respective IP addresses.
1426 * A new client tool "networkctl" for systemd-networkd has been
1427 added. It currently is entirely passive and will query
1428 networking configuration from udev, rtnetlink and networkd,
1429 and present it to the user in a very friendly
1430 way. Eventually, we hope to extend it to become a full
1431 control utility for networkd.
1433 * .socket units gained a new DeferAcceptSec= setting that
1434 controls the kernels' TCP_DEFER_ACCEPT sockopt for
1435 TCP. Similar, support for controlling TCP keep-alive
1436 settings has been added (KeepAliveTimeSec=,
1437 KeepAliveIntervalSec=, KeepAliveProbes=). Also, support for
1438 turning off Nagle's algorithm on TCP has been added
1441 * logind learned a new session type "web", for use in projects
1442 like Cockpit which register web clients as PAM sessions.
1444 * timer units with at least one OnCalendar= setting will now
1445 be started only after timer-sync.target has been
1446 reached. This way they will not elapse before the system
1447 clock has been corrected by a local NTP client or
1448 similar. This is particular useful on RTC-less embedded
1449 machines, that come up with an invalid system clock.
1451 * systemd-nspawn's --network-veth= switch should now result in
1452 stable MAC addresses for both the outer and the inner side
1455 * systemd-nspawn gained a new --volatile= switch for running
1456 container instances with /etc or /var unpopulated.
1458 * The kdbus client code has been updated to use the new Linux
1459 3.17 memfd subsystem instead of the old kdbus-specific one.
1461 * systemd-networkd's DHCP client and server now support
1462 FORCERENEW. There are also new configuration options to
1463 configure the vendor client identifier and broadcast mode
1466 * systemd will no longer inform the kernel about the current
1467 timezone, as this is necessarily incorrect and racy as the
1468 kernel has no understanding of DST and similar
1469 concepts. This hence means FAT timestamps will be always
1470 considered UTC, similar to what Android is already
1471 doing. Also, when the RTC is configured to the local time
1472 (rather than UTC) systemd will never synchronize back to it,
1473 as this might confuse Windows at a later boot.
1475 * systemd-analyze gained a new command "verify" for offline
1476 validation of unit files.
1478 * systemd-networkd gained support for a couple of additional
1479 settings for bonding networking setups. Also, the metric for
1480 statically configured routes may now be configured. For
1481 network interfaces where this is appropriate the peer IP
1482 address may now be configured.
1484 * systemd-networkd's DHCP client will no longer request
1485 broadcasting by default, as this tripped up some networks.
1486 For hardware where broadcast is required the feature should
1487 be switched back on using RequestBroadcast=yes.
1489 * systemd-networkd will now set up IPv4LL addresses (when
1490 enabled) even if DHCP is configured successfully.
1492 * udev will now default to respect network device names given
1493 by the kernel when the kernel indicates that these are
1494 predictable. This behavior can be tweaked by changing
1495 NamePolicy= in the relevant .link file.
1497 * A new library systemd-terminal has been added that
1498 implements full TTY stream parsing and rendering. This
1499 library is supposed to be used later on for implementing a
1500 full userspace VT subsystem, replacing the current kernel
1503 * A new tool systemd-journal-upload has been added to push
1504 journal data to a remote system running
1505 systemd-journal-remote.
1507 * journald will no longer forward all local data to another
1508 running syslog daemon. This change has been made because
1509 rsyslog (which appears to be the most commonly used syslog
1510 implementation these days) no longer makes use of this, and
1511 instead pulls the data out of the journal on its own. Since
1512 forwarding the messages to a non-existent syslog server is
1513 more expensive than we assumed we have now turned this
1514 off. If you run a syslog server that is not a recent rsyslog
1515 version, you have to turn this option on again
1516 (ForwardToSyslog= in journald.conf).
1518 * journald now optionally supports the LZ4 compressor for
1519 larger journal fields. This compressor should perform much
1520 better than XZ which was the previous default.
1522 * machinectl now shows the IP addresses of local containers,
1523 if it knows them, plus the interface name of the container.
1525 * A new tool "systemd-escape" has been added that makes it
1526 easy to escape strings to build unit names and similar.
1528 * sd_notify() messages may now include a new ERRNO= field
1529 which is parsed and collected by systemd and shown among the
1530 "systemctl status" output for a service.
1532 * A new component "systemd-firstboot" has been added that
1533 queries the most basic systemd information (timezone,
1534 hostname, root password) interactively on first
1535 boot. Alternatively it may also be used to provision these
1536 things offline on OS images installed into directories.
1538 * The default sysctl.d/ snippets will now set
1540 net.ipv4.conf.default.promote_secondaries=1
1542 This has the benefit of no flushing secondary IP addresses
1543 when primary addresses are removed.
1545 Contributions from: Ansgar Burchardt, Bastien Nocera, Colin
1546 Walters, Dan Dedrick, Daniel Buch, Daniel Korostil, Daniel
1547 Mack, Dan Williams, Dave Reisner, David Herrmann, Denis
1548 Kenzior, Eelco Dolstra, Eric Cook, Hannes Reinecke, Harald
1549 Hoyer, Hong Shick Pak, Hui Wang, Jean-André Santoni, Jóhann
1550 B. Guðmundsson, Jon Severinsson, Karel Zak, Kay Sievers, Kevin
1551 Wells, Lennart Poettering, Lukas Nykryn, Mantas Mikulėnas,
1552 Marc-Antoine Perennou, Martin Pitt, Michael Biebl, Michael
1553 Marineau, Michael Olbrich, Michal Schmidt, Michal Sekletar,
1554 Miguel Angel Ajo, Mike Gilbert, Olivier Brunel, Robert
1555 Schiele, Ronny Chevalier, Simon McVittie, Sjoerd Simons, Stef
1556 Walter, Steven Noonan, Susant Sahani, Tanu Kaskinen, Thomas
1557 Blume, Thomas Hindoe Paaboel Andersen, Timofey Titovets,
1558 Tobias Geerinckx-Rice, Tomasz Torcz, Tom Gundersen, Umut
1559 Tezduyar Lindskog, Zbigniew Jędrzejewski-Szmek
1561 -- Berlin, 2014-08-19
1565 * A new tool systemd-sysusers has been added. This tool
1566 creates system users and groups in /etc/passwd and
1567 /etc/group, based on static declarative system user/group
1568 definitions in /usr/lib/sysusers.d/. This is useful to
1569 enable factory resets and volatile systems that boot up with
1570 an empty /etc directory, and thus need system users and
1571 groups created during early boot. systemd now also ships
1572 with two default sysusers.d/ files for the most basic
1573 users and groups systemd and the core operating system
1576 * A new tmpfiles snippet has been added that rebuilds the
1577 essential files in /etc on boot, should they be missing.
1579 * A directive for ensuring automatic clean-up of
1580 /var/cache/man/ has been removed from the default
1581 configuration. This line should now be shipped by the man
1582 implementation. The necessary change has been made to the
1583 man-db implementation. Note that you need to update your man
1584 implementation to one that ships this line, otherwise no
1585 automatic clean-up of /var/cache/man will take place.
1587 * A new condition ConditionNeedsUpdate= has been added that
1588 may conditionalize services to only run when /etc or /var
1589 are "older" than the vendor operating system resources in
1590 /usr. This is useful for reconstructing or updating /etc
1591 after an offline update of /usr or a factory reset, on the
1592 next reboot. Services that want to run once after such an
1593 update or reset should use this condition and order
1594 themselves before the new systemd-update-done.service, which
1595 will mark the two directories as fully updated. A number of
1596 service files have been added making use of this, to rebuild
1597 the udev hardware database, the journald message catalog and
1598 dynamic loader cache (ldconfig). The systemd-sysusers tool
1599 described above also makes use of this now. With this in
1600 place it is now possible to start up a minimal operating
1601 system with /etc empty cleanly. For more information on the
1602 concepts involved see this recent blog story:
1604 http://0pointer.de/blog/projects/stateless.html
1606 * A new system group "input" has been introduced, and all
1607 input device nodes get this group assigned. This is useful
1608 for system-level software to get access to input devices. It
1609 complements what is already done for "audio" and "video".
1611 * systemd-networkd learnt minimal DHCPv4 server support in
1612 addition to the existing DHCPv4 client support. It also
1613 learnt DHCPv6 client and IPv6 Router Solicitation client
1614 support. The DHCPv4 client gained support for static routes
1615 passed in from the server. Note that the [DHCPv4] section
1616 known in older systemd-networkd versions has been renamed to
1617 [DHCP] and is now also used by the DHCPv6 client. Existing
1618 .network files using settings of this section should be
1619 updated, though compatibility is maintained. Optionally, the
1620 client hostname may now be sent to the DHCP server.
1622 * networkd gained support for vxlan virtual networks as well
1623 as tun/tap and dummy devices.
1625 * networkd gained support for automatic allocation of address
1626 ranges for interfaces from a system-wide pool of
1627 addresses. This is useful for dynamically managing a large
1628 number of interfaces with a single network configuration
1629 file. In particular this is useful to easily assign
1630 appropriate IP addresses to the veth links of a large number
1631 of nspawn instances.
1633 * RPM macros for processing sysusers, sysctl and binfmt
1634 drop-in snippets at package installation time have been
1637 * The /etc/os-release file should now be placed in
1638 /usr/lib/os-release. The old location is automatically
1639 created as symlink. /usr/lib is the more appropriate
1640 location of this file, since it shall actually describe the
1641 vendor operating system shipped in /usr, and not the
1642 configuration stored in /etc.
1644 * .mount units gained a new boolean SloppyOptions= setting
1645 that maps to mount(8)'s -s option which enables permissive
1646 parsing of unknown mount options.
1648 * tmpfiles learnt a new "L+" directive which creates a symlink
1649 but (unlike "L") deletes a pre-existing file first, should
1650 it already exist and not already be the correct
1651 symlink. Similar, "b+", "c+" and "p+" directives have been
1652 added as well, which create block and character devices, as
1653 well as fifos in the filesystem, possibly removing any
1654 pre-existing files of different types.
1656 * For tmpfiles' "L", "L+", "C" and "C+" directives the final
1657 'argument' field (which so far specified the source to
1658 symlink/copy the files from) is now optional. If omitted the
1659 same file os copied from /usr/share/factory/ suffixed by the
1660 full destination path. This is useful for populating /etc
1661 with essential files, by copying them from vendor defaults
1662 shipped in /usr/share/factory/etc.
1664 * A new command "systemctl preset-all" has been added that
1665 applies the service preset settings to all installed unit
1666 files. A new switch --preset-mode= has been added that
1667 controls whether only enable or only disable operations
1670 * A new command "systemctl is-system-running" has been added
1671 that allows checking the overall state of the system, for
1672 example whether it is fully up and running.
1674 * When the system boots up with an empty /etc, the equivalent
1675 to "systemctl preset-all" is executed during early boot, to
1676 make sure all default services are enabled after a factory
1679 * systemd now contains a minimal preset file that enables the
1680 most basic services systemd ships by default.
1682 * Unit files' [Install] section gained a new DefaultInstance=
1683 field for defining the default instance to create if a
1684 template unit is enabled with no instance specified.
1686 * A new passive target cryptsetup-pre.target has been added
1687 that may be used by services that need to make they run and
1688 finish before the first LUKS cryptographic device is set up.
1690 * The /dev/loop-control and /dev/btrfs-control device nodes
1691 are now owned by the "disk" group by default, opening up
1692 access to this group.
1694 * systemd-coredump will now automatically generate a
1695 stack trace of all core dumps taking place on the system,
1696 based on elfutils' libdw library. This stack trace is logged
1699 * systemd-coredump may now optionally store coredumps directly
1700 on disk (in /var/lib/systemd/coredump, possibly compressed),
1701 instead of storing them unconditionally in the journal. This
1702 mode is the new default. A new configuration file
1703 /etc/systemd/coredump.conf has been added to configure this
1704 and other parameters of systemd-coredump.
1706 * coredumpctl gained a new "info" verb to show details about a
1707 specific coredump. A new switch "-1" has also been added
1708 that makes sure to only show information about the most
1709 recent entry instead of all entries. Also, as the tool is
1710 generally useful now the "systemd-" prefix of the binary
1711 name has been removed. Distributions that want to maintain
1712 compatibility with the old name should add a symlink from
1713 the old name to the new name.
1715 * journald's SplitMode= now defaults to "uid". This makes sure
1716 that unprivileged users can access their own coredumps with
1717 coredumpctl without restrictions.
1719 * New kernel command line options "systemd.wants=" (for
1720 pulling an additional unit during boot), "systemd.mask="
1721 (for masking a specific unit for the boot), and
1722 "systemd.debug-shell" (for enabling the debug shell on tty9)
1723 have been added. This is implemented in the new generator
1724 "systemd-debug-generator".
1726 * systemd-nspawn will now by default filter a couple of
1727 syscalls for containers, among them those required for
1728 kernel module loading, direct x86 IO port access, swap
1729 management, and kexec. Most importantly though
1730 open_by_handle_at() is now prohibited for containers,
1731 closing a hole similar to a recently discussed vulnerability
1732 in docker regarding access to files on file hierarchies the
1733 container should normally not have access to. Note that for
1734 nspawn we generally make no security claims anyway (and
1735 this is explicitly documented in the man page), so this is
1736 just a fix for one of the most obvious problems.
1738 * A new man page file-hierarchy(7) has been added that
1739 contains a minimized, modernized version of the file system
1740 layout systemd expects, similar in style to the FHS
1741 specification or hier(5). A new tool systemd-path(1) has
1742 been added to query many of these paths for the local
1745 * Automatic time-based clean-up of $XDG_RUNTIME_DIR is no
1746 longer done. Since the directory now has a per-user size
1747 limit, and is cleaned on logout this appears unnecessary,
1748 in particular since this now brings the lifecycle of this
1749 directory closer in line with how IPC objects are handled.
1751 * systemd.pc now exports a number of additional directories,
1752 including $libdir (which is useful to identify the library
1753 path for the primary architecture of the system), and a
1754 couple of drop-in directories.
1756 * udev's predictable network interface names now use the dev_port
1757 sysfs attribute, introduced in linux 3.15 instead of dev_id to
1758 distinguish between ports of the same PCI function. dev_id should
1759 only be used for ports using the same HW address, hence the need
1762 * machined has been updated to export the OS version of a
1763 container (read from /etc/os-release and
1764 /usr/lib/os-release) on the bus. This is now shown in
1765 "machinectl status" for a machine.
1767 * A new service setting RestartForceExitStatus= has been
1768 added. If configured to a set of exit signals or process
1769 return values, the service will be restarted when the main
1770 daemon process exits with any of them, regardless of the
1773 * systemctl's -H switch for connecting to remote systemd
1774 machines has been extended so that it may be used to
1775 directly connect to a specific container on the
1776 host. "systemctl -H root@foobar:waldi" will now connect as
1777 user "root" to host "foobar", and then proceed directly to
1778 the container named "waldi". Note that currently you have to
1779 authenticate as user "root" for this to work, as entering
1780 containers is a privileged operation.
1782 Contributions from: Andreas Henriksson, Benjamin Steinwender,
1783 Carl Schaefer, Christian Hesse, Colin Ian King, Cristian
1784 Rodríguez, Daniel Mack, Dave Reisner, David Herrmann, Eugene
1785 Yakubovich, Filipe Brandenburger, Frederic Crozat, Hristo
1786 Venev, Jan Engelhardt, Jonathan Boulle, Kay Sievers, Lennart
1787 Poettering, Luke Shumaker, Mantas Mikulėnas, Marc-Antoine
1788 Perennou, Marcel Holtmann, Michael Marineau, Michael Olbrich,
1789 Michał Bartoszkiewicz, Michal Sekletar, Patrik Flykt, Ronan Le
1790 Martret, Ronny Chevalier, Ruediger Oertel, Steven Noonan,
1791 Susant Sahani, Thadeu Lima de Souza Cascardo, Thomas Hindoe
1792 Paaboel Andersen, Tom Gundersen, Tom Hirst, Umut Tezduyar
1793 Lindskog, Uoti Urpala, Zbigniew Jędrzejewski-Szmek
1795 -- Berlin, 2014-07-03
1799 * As an experimental feature, udev now tries to lock the
1800 disk device node (flock(LOCK_SH|LOCK_NB)) while it
1801 executes events for the disk or any of its partitions.
1802 Applications like partitioning programs can lock the
1803 disk device node (flock(LOCK_EX)) and claim temporary
1804 device ownership that way; udev will entirely skip all event
1805 handling for this disk and its partitions. If the disk
1806 was opened for writing, the close will trigger a partition
1807 table rescan in udev's "watch" facility, and if needed
1808 synthesize "change" events for the disk and all its partitions.
1809 This is now unconditionally enabled, and if it turns out to
1810 cause major problems, we might turn it on only for specific
1811 devices, or might need to disable it entirely. Device Mapper
1812 devices are excluded from this logic.
1814 * We temporarily dropped the "-l" switch for fsck invocations,
1815 since they collide with the flock() logic above. util-linux
1816 upstream has been changed already to avoid this conflict,
1817 and we will readd "-l" as soon as util-linux with this
1818 change has been released.
1820 * The dependency on libattr has been removed. Since a long
1821 time, the extended attribute calls have moved to glibc, and
1822 libattr is thus unnecessary.
1824 * Virtualization detection works without priviliges now. This
1825 means the systemd-detect-virt binary no longer requires
1826 CAP_SYS_PTRACE file capabilities, and our daemons can run
1827 with fewer privileges.
1829 * systemd-networkd now runs under its own "systemd-network"
1830 user. It retains the CAP_NET_ADMIN, CAP_NET_BIND_SERVICE,
1831 CAP_NET_BROADCAST, CAP_NET_RAW capabilities though, but
1832 loses the ability to write to files owned by root this way.
1834 * Similar, systemd-resolved now runs under its own
1835 "systemd-resolve" user with no capabilities remaining.
1837 * Similar, systemd-bus-proxyd now runs under its own
1838 "systemd-bus-proxy" user with only CAP_IPC_OWNER remaining.
1840 * systemd-networkd gained support for setting up "veth"
1841 virtual ethernet devices for container connectivity, as well
1842 as GRE and VTI tunnels.
1844 * systemd-networkd will no longer automatically attempt to
1845 manually load kernel modules necessary for certain tunnel
1846 transports. Instead, it is assumed the kernel loads them
1847 automatically when required. This only works correctly on
1848 very new kernels. On older kernels, please consider adding
1849 the kernel modules to /etc/modules-load.d/ as a work-around.
1851 * The resolv.conf file systemd-resolved generates has been
1852 moved to /run/systemd/resolve/. If you have a symlink from
1853 /etc/resolv.conf, it might be necessary to correct it.
1855 * Two new service settings, ProtectHome= and ProtectSystem=,
1856 have been added. When enabled, they will make the user data
1857 (such as /home) inaccessible or read-only and the system
1858 (such as /usr) read-only, for specific services. This allows
1859 very light-weight per-service sandboxing to avoid
1860 modifications of user data or system files from
1861 services. These two new switches have been enabled for all
1862 of systemd's long-running services, where appropriate.
1864 * Socket units gained new SocketUser= and SocketGroup=
1865 settings to set the owner user and group of AF_UNIX sockets
1866 and FIFOs in the file system.
1868 * Socket units gained a new RemoveOnStop= setting. If enabled,
1869 all FIFOS and sockets in the file system will be removed
1870 when the specific socket unit is stopped.
1872 * Socket units gained a new Symlinks= setting. It takes a list
1873 of symlinks to create to file system sockets or FIFOs
1874 created by the specific Unix sockets. This is useful to
1875 manage symlinks to socket nodes with the same life-cycle as
1878 * The /dev/log socket and /dev/initctl FIFO have been moved to
1879 /run, and have been replaced by symlinks. This allows
1880 connecting to these facilities even if PrivateDevices=yes is
1881 used for a service (which makes /dev/log itself unavailable,
1882 but /run is left). This also has the benefit of ensuring
1883 that /dev only contains device nodes, directories and
1884 symlinks, and nothing else.
1886 * sd-daemon gained two new calls sd_pid_notify() and
1887 sd_pid_notifyf(). They are similar to sd_notify() and
1888 sd_notifyf(), but allow overriding of the source PID of
1889 notification messages if permissions permit this. This is
1890 useful to send notify messages on behalf of a different
1891 process (for example, the parent process). The
1892 systemd-notify tool has been updated to make use of this
1893 when sending messages (so that notification messages now
1894 originate from the shell script invoking systemd-notify and
1895 not the systemd-notify process itself. This should minimize
1896 a race where systemd fails to associate notification
1897 messages to services when the originating process already
1900 * A new "on-abnormal" setting for Restart= has been added. If
1901 set, it will result in automatic restarts on all "abnormal"
1902 reasons for a process to exit, which includes unclean
1903 signals, core dumps, timeouts and watchdog timeouts, but
1904 does not include clean and unclean exit codes or clean
1905 signals. Restart=on-abnormal is an alternative for
1906 Restart=on-failure for services that shall be able to
1907 terminate and avoid restarts on certain errors, by
1908 indicating so with an unclean exit code. Restart=on-failure
1909 or Restart=on-abnormal is now the recommended setting for
1910 all long-running services.
1912 * If the InaccessibleDirectories= service setting points to a
1913 mount point (or if there are any submounts contained within
1914 it), it is now attempted to completely unmount it, to make
1915 the file systems truly unavailable for the respective
1918 * The ReadOnlyDirectories= service setting and
1919 systemd-nspawn's --read-only parameter are now recursively
1920 applied to all submounts, too.
1922 * Mount units may now be created transiently via the bus APIs.
1924 * The support for SysV and LSB init scripts has been removed
1925 from the systemd daemon itself. Instead, it is now
1926 implemented as a generator that creates native systemd units
1927 from these scripts when needed. This enables us to remove a
1928 substantial amount of legacy code from PID 1, following the
1929 fact that many distributions only ship a very small number
1930 of LSB/SysV init scripts nowadays.
1932 * Privileged Xen (dom0) domains are not considered
1933 virtualization anymore by the virtualization detection
1934 logic. After all, they generally have unrestricted access to
1935 the hardware and usually are used to manage the unprivileged
1938 * systemd-tmpfiles gained a new "C" line type, for copying
1939 files or entire directories.
1941 * systemd-tmpfiles "m" lines are now fully equivalent to "z"
1942 lines. So far, they have been non-globbing versions of the
1943 latter, and have thus been redundant. In future, it is
1944 recommended to only use "z". "m" has hence been removed
1945 from the documentation, even though it stays supported.
1947 * A tmpfiles snippet to recreate the most basic structure in
1948 /var has been added. This is enough to create the /var/run →
1949 /run symlink and create a couple of structural
1950 directories. This allows systems to boot up with an empty or
1951 volatile /var. Of course, while with this change, the core OS
1952 now is capable with dealing with a volatile /var, not all
1953 user services are ready for it. However, we hope that sooner
1954 or later, many service daemons will be changed upstream so
1955 that they are able to automatically create their necessary
1956 directories in /var at boot, should they be missing. This is
1957 the first step to allow state-less systems that only require
1958 the vendor image for /usr to boot.
1960 * systemd-nspawn has gained a new --tmpfs= switch to mount an
1961 empty tmpfs instance to a specific directory. This is
1962 particularly useful for making use of the automatic
1963 reconstruction of /var (see above), by passing --tmpfs=/var.
1965 * Access modes specified in tmpfiles snippets may now be
1966 prefixed with "~", which indicates that they shall be masked
1967 by whether the existing file or directory is currently
1968 writable, readable or executable at all. Also, if specified,
1969 the sgid/suid/sticky bits will be masked for all
1972 * A new passive target unit "network-pre.target" has been
1973 added which is useful for services that shall run before any
1974 network is configured, for example firewall scripts.
1976 * The "floppy" group that previously owned the /dev/fd*
1977 devices is no longer used. The "disk" group is now used
1978 instead. Distributions should probably deprecate usage of
1981 Contributions from: Camilo Aguilar, Christian Hesse, Colin Ian
1982 King, Cristian Rodríguez, Daniel Buch, Dave Reisner, David
1983 Strauss, Denis Tikhomirov, John, Jonathan Liu, Kay Sievers,
1984 Lennart Poettering, Mantas Mikulėnas, Mark Eichin, Ronny
1985 Chevalier, Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel
1986 Andersen, Tom Gundersen, Umut Tezduyar Lindskog, Zbigniew
1989 -- Berlin, 2014-06-11
1993 * A new "systemd-timesyncd" daemon has been added for
1994 synchronizing the system clock across the network. It
1995 implements an SNTP client. In contrast to NTP
1996 implementations such as chrony or the NTP reference server,
1997 this only implements a client side, and does not bother with
1998 the full NTP complexity, focusing only on querying time from
1999 one remote server and synchronizing the local clock to
2000 it. Unless you intend to serve NTP to networked clients or
2001 want to connect to local hardware clocks, this simple NTP
2002 client should be more than appropriate for most
2003 installations. The daemon runs with minimal privileges, and
2004 has been hooked up with networkd to only operate when
2005 network connectivity is available. The daemon saves the
2006 current clock to disk every time a new NTP sync has been
2007 acquired, and uses this to possibly correct the system clock
2008 early at bootup, in order to accommodate for systems that
2009 lack an RTC such as the Raspberry Pi and embedded devices,
2010 and to make sure that time monotonically progresses on these
2011 systems, even if it is not always correct. To make use of
2012 this daemon, a new system user and group "systemd-timesync"
2013 needs to be created on installation of systemd.
2015 * The queue "seqnum" interface of libudev has been disabled, as
2016 it was generally incompatible with device namespacing as
2017 sequence numbers of devices go "missing" if the devices are
2018 part of a different namespace.
2020 * "systemctl list-timers" and "systemctl list-sockets" gained
2021 a --recursive switch for showing units of these types also
2022 for all local containers, similar in style to the already
2023 supported --recursive switch for "systemctl list-units".
2025 * A new RebootArgument= setting has been added for service
2026 units, which may be used to specify a kernel reboot argument
2027 to use when triggering reboots with StartLimitAction=.
2029 * A new FailureAction= setting has been added for service
2030 units which may be used to specify an operation to trigger
2031 when a service fails. This works similarly to
2032 StartLimitAction=, but unlike it, controls what is done
2033 immediately rather than only after several attempts to
2034 restart the service in question.
2036 * hostnamed got updated to also expose the kernel name,
2037 release, and version on the bus. This is useful for
2038 executing commands like hostnamectl with the -H switch.
2039 systemd-analyze makes use of this to properly display
2040 details when running non-locally.
2042 * The bootchart tool can now show cgroup information in the
2043 graphs it generates.
2045 * The CFS CPU quota cgroup attribute is now exposed for
2046 services. The new CPUQuota= switch has been added for this
2047 which takes a percentage value. Setting this will have the
2048 result that a service may never get more CPU time than the
2049 specified percentage, even if the machine is otherwise idle.
2051 * systemd-networkd learned IPIP and SIT tunnel support.
2053 * LSB init scripts exposing a dependency on $network will now
2054 get a dependency on network-online.target rather than simply
2055 network.target. This should bring LSB handling closer to
2056 what it was on SysV systems.
2058 * A new fsck.repair= kernel option has been added to control
2059 how fsck shall deal with unclean file systems at boot.
2061 * The (.ini) configuration file parser will now silently
2062 ignore sections whose name begins with "X-". This may be
2063 used to maintain application-specific extension sections in unit
2066 * machined gained a new API to query the IP addresses of
2067 registered containers. "machinectl status" has been updated
2068 to show these addresses in its output.
2070 * A new call sd_uid_get_display() has been added to the
2071 sd-login APIs for querying the "primary" session of a
2072 user. The "primary" session of the user is elected from the
2073 user's sessions and generally a graphical session is
2074 preferred over a text one.
2076 * A minimal systemd-resolved daemon has been added. It
2077 currently simply acts as a companion to systemd-networkd and
2078 manages resolv.conf based on per-interface DNS
2079 configuration, possibly supplied via DHCP. In the long run
2080 we hope to extend this into a local DNSSEC enabled DNS and
2083 * The systemd-networkd-wait-online tool is now enabled by
2084 default. It will delay network-online.target until a network
2085 connection has been configured. The tool primarily integrates
2086 with networkd, but will also make a best effort to make sense
2087 of network configuration performed in some other way.
2089 * Two new service options StartupCPUShares= and
2090 StartupBlockIOWeight= have been added that work similarly to
2091 CPUShares= and BlockIOWeight= however only apply during
2092 system startup. This is useful to prioritize certain services
2093 differently during bootup than during normal runtime.
2095 * hostnamed has been changed to prefer the statically
2096 configured hostname in /etc/hostname (unless set to
2097 'localhost' or empty) over any dynamic one supplied by
2098 dhcp. With this change, the rules for picking the hostname
2099 match more closely the rules of other configuration settings
2100 where the local administrator's configuration in /etc always
2101 overrides any other settings.
2103 Contributions fron: Ali H. Caliskan, Alison Chaiken, Bas van
2104 den Berg, Brandon Philips, Cristian Rodríguez, Daniel Buch,
2105 Dan Kilman, Dave Reisner, David Härdeman, David Herrmann,
2106 David Strauss, Dimitris Spingos, Djalal Harouni, Eelco
2107 Dolstra, Evan Nemerson, Florian Albrechtskirchinger, Greg
2108 Kroah-Hartman, Harald Hoyer, Holger Hans Peter Freyther, Jan
2109 Engelhardt, Jani Nikula, Jason St. John, Jeffrey Clark,
2110 Jonathan Boulle, Kay Sievers, Lennart Poettering, Lukas
2111 Nykryn, Lukasz Skalski, Łukasz Stelmach, Mantas Mikulėnas,
2112 Marcel Holtmann, Martin Pitt, Matthew Monaco, Michael
2113 Marineau, Michael Olbrich, Michal Sekletar, Mike Gilbert, Nis
2114 Martensen, Patrik Flykt, Philip Lorenz, poma, Ray Strode,
2115 Reyad Attiyat, Robert Milasan, Scott Thrasher, Stef Walter,
2116 Steven Siloti, Susant Sahani, Tanu Kaskinen, Thomas Bächler,
2117 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar
2118 Lindskog, WaLyong Cho, Will Woods, Zbigniew
2121 -- Beijing, 2014-05-28
2125 * When restoring the screen brightness at boot, stay away from
2126 the darkest setting or from the lowest 5% of the available
2127 range, depending on which is the larger value of both. This
2128 should effectively protect the user from rebooting into a
2129 black screen, should the brightness have been set to minimum
2132 * sd-login gained a new sd_machine_get_class() call to
2133 determine the class ("vm" or "container") of a machine
2134 registered with machined.
2136 * sd-login gained new calls
2137 sd_peer_get_{session,owner_uid,unit,user_unit,slice,machine_name}(),
2138 to query the identity of the peer of a local AF_UNIX
2139 connection. They operate similarly to their sd_pid_get_xyz()
2142 * PID 1 will now maintain a system-wide system state engine
2143 with the states "starting", "running", "degraded",
2144 "maintenance", "stopping". These states are bound to system
2145 startup, normal runtime, runtime with at least one failed
2146 service, rescue/emergency mode and system shutdown. This
2147 state is shown in the "systemctl status" output when no unit
2148 name is passed. It is useful to determine system state, in
2149 particularly when doing so for many systems or containers at
2152 * A new command "list-machines" has been added to "systemctl"
2153 that lists all local OS containers and shows their system
2154 state (see above), if systemd runs inside of them.
2156 * systemctl gained a new "-r" switch to recursively enumerate
2157 units on all local containers, when used with the
2158 "list-unit" command (which is the default one that is
2159 executed when no parameters are specified).
2161 * The GPT automatic partition discovery logic will now honour
2162 two GPT partition flags: one may be set on a partition to
2163 cause it to be mounted read-only, and the other may be set
2164 on a partition to ignore it during automatic discovery.
2166 * Two new GPT type UUIDs have been added for automatic root
2167 partition discovery, for 32-bit and 64-bit ARM. This is not
2168 particularly useful for discovering the root directory on
2169 these architectures during bare-metal boots (since UEFI is
2170 not common there), but still very useful to allow booting of
2171 ARM disk images in nspawn with the -i option.
2173 * MAC addresses of interfaces created with nspawn's
2174 --network-interface= switch will now be generated from the
2175 machine name, and thus be stable between multiple invocations
2178 * logind will now automatically remove all IPC objects owned
2179 by a user if she or he fully logs out. This makes sure that
2180 users who are logged out cannot continue to consume IPC
2181 resources. This covers SysV memory, semaphores and message
2182 queues as well as POSIX shared memory and message
2183 queues. Traditionally, SysV and POSIX IPC had no life-cycle
2184 limits. With this functionality, that is corrected. This may
2185 be turned off by using the RemoveIPC= switch of logind.conf.
2187 * The systemd-machine-id-setup and tmpfiles tools gained a
2188 --root= switch to operate on a specific root directory,
2191 * journald can now forward logged messages to the TTYs of all
2192 logged in users ("wall"). This is the default for all
2193 emergency messages now.
2195 * A new tool systemd-journal-remote has been added to stream
2196 journal log messages across the network.
2198 * /sys/fs/cgroup/ is now mounted read-only after all cgroup
2199 controller trees are mounted into it. Note that the
2200 directories mounted beneath it are not read-only. This is a
2201 security measure and is particularly useful because glibc
2202 actually includes a search logic to pick any tmpfs it can
2203 find to implement shm_open() if /dev/shm is not available
2204 (which it might very well be in namespaced setups).
2206 * machinectl gained a new "poweroff" command to cleanly power
2207 down a local OS container.
2209 * The PrivateDevices= unit file setting will now also drop the
2210 CAP_MKNOD capability from the capability bound set, and
2211 imply DevicePolicy=closed.
2213 * PrivateDevices=, PrivateNetwork= and PrivateTmp= is now used
2214 comprehensively on all long-running systemd services where
2215 this is appropriate.
2217 * systemd-udevd will now run in a disassociated mount
2218 namespace. To mount directories from udev rules, make sure to
2219 pull in mount units via SYSTEMD_WANTS properties.
2221 * The kdbus support gained support for uploading policy into
2222 the kernel. sd-bus gained support for creating "monitoring"
2223 connections that can eavesdrop into all bus communication
2224 for debugging purposes.
2226 * Timestamps may now be specified in seconds since the UNIX
2227 epoch Jan 1st, 1970 by specifying "@" followed by the value
2230 * Native tcpwrap support in systemd has been removed. tcpwrap
2231 is old code, not really maintained anymore and has serious
2232 shortcomings, and better options such as firewalls
2233 exist. For setups that require tcpwrap usage, please
2234 consider invoking your socket-activated service via tcpd,
2235 like on traditional inetd.
2237 * A new system.conf configuration option
2238 DefaultTimerAccuracySec= has been added that controls the
2239 default AccuracySec= setting of .timer units.
2241 * Timer units gained a new WakeSystem= switch. If enabled,
2242 timers configured this way will cause the system to resume
2243 from system suspend (if the system supports that, which most
2246 * Timer units gained a new Persistent= switch. If enabled,
2247 timers configured this way will save to disk when they have
2248 been last triggered. This information is then used on next
2249 reboot to possible execute overdue timer events, that
2250 could not take place because the system was powered off.
2251 This enables simple anacron-like behaviour for timer units.
2253 * systemctl's "list-timers" will now also list the time a
2254 timer unit was last triggered in addition to the next time
2255 it will be triggered.
2257 * systemd-networkd will now assign predictable IPv4LL
2258 addresses to its local interfaces.
2260 Contributions from: Brandon Philips, Daniel Buch, Daniel Mack,
2261 Dave Reisner, David Herrmann, Gerd Hoffmann, Greg
2262 Kroah-Hartman, Hendrik Brueckner, Jason St. John, Josh
2263 Triplett, Kay Sievers, Lennart Poettering, Marc-Antoine
2264 Perennou, Michael Marineau, Michael Olbrich, Miklos Vajna,
2265 Patrik Flykt, poma, Sebastian Thorarensen, Thomas Bächler,
2266 Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom Gundersen,
2267 Umut Tezduyar Lindskog, Wieland Hoffmann, Zbigniew
2270 -- Berlin, 2014-03-25
2274 * A new unit file setting RestrictAddressFamilies= has been
2275 added to restrict which socket address families unit
2276 processes gain access to. This takes address family names
2277 like "AF_INET" or "AF_UNIX", and is useful to minimize the
2278 attack surface of services via exotic protocol stacks. This
2279 is built on seccomp system call filters.
2281 * Two new unit file settings RuntimeDirectory= and
2282 RuntimeDirectoryMode= have been added that may be used to
2283 manage a per-daemon runtime directories below /run. This is
2284 an alternative for setting up directory permissions with
2285 tmpfiles snippets, and has the advantage that the runtime
2286 directory's lifetime is bound to the daemon runtime and that
2287 the daemon starts up with an empty directory each time. This
2288 is particularly useful when writing services that drop
2289 privileges using the User= or Group= setting.
2291 * The DeviceAllow= unit setting now supports globbing for
2292 matching against device group names.
2294 * The systemd configuration file system.conf gained new
2295 settings DefaultCPUAccounting=, DefaultBlockIOAccounting=,
2296 DefaultMemoryAccounting= to globally turn on/off accounting
2297 for specific resources (cgroups) for all units. These
2298 settings may still be overridden individually in each unit
2301 * systemd-gpt-auto-generator is now able to discover /srv and
2302 root partitions in addition to /home and swap partitions. It
2303 also supports LUKS-encrypted partitions now. With this in
2304 place, automatic discovery of partitions to mount following
2305 the Discoverable Partitions Specification
2306 (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec)
2307 is now a lot more complete. This allows booting without
2308 /etc/fstab and without root= on the kernel command line on
2309 systems prepared appropriately.
2311 * systemd-nspawn gained a new --image= switch which allows
2312 booting up disk images and Linux installations on any block
2313 device that follow the Discoverable Partitions Specification
2314 (see above). This means that installations made with
2315 appropriately updated installers may now be started and
2316 deployed using container managers, completely
2317 unmodified. (We hope that libvirt-lxc will add support for
2318 this feature soon, too.)
2320 * systemd-nspawn gained a new --network-macvlan= setting to
2321 set up a private macvlan interface for the
2322 container. Similarly, systemd-networkd gained a new
2323 Kind=macvlan setting in .netdev files.
2325 * systemd-networkd now supports configuring local addresses
2328 * A new tool systemd-network-wait-online has been added to
2329 synchronously wait for network connectivity using
2332 * The sd-bus.h bus API gained a new sd_bus_track object for
2333 tracking the life-cycle of bus peers. Note that sd-bus.h is
2334 still not a public API though (unless you specify
2335 --enable-kdbus on the configure command line, which however
2336 voids your warranty and you get no API stability guarantee).
2338 * The $XDG_RUNTIME_DIR runtime directories for each user are
2339 now individual tmpfs instances, which has the benefit of
2340 introducing separate pools for each user, with individual
2341 size limits, and thus making sure that unprivileged clients
2342 can no longer negatively impact the system or other users by
2343 filling up their $XDG_RUNTIME_DIR. A new logind.conf setting
2344 RuntimeDirectorySize= has been introduced that allows
2345 controlling the default size limit for all users. It
2346 defaults to 10% of the available physical memory. This is no
2347 replacement for quotas on tmpfs though (which the kernel
2348 still does not support), as /dev/shm and /tmp are still
2349 shared resources used by both the system and unprivileged
2352 * logind will now automatically turn off automatic suspending
2353 on laptop lid close when more than one display is
2354 connected. This was previously expected to be implemented
2355 individually in desktop environments (such as GNOME),
2356 however has been added to logind now, in order to fix a
2357 boot-time race where a desktop environment might not have
2358 been started yet and thus not been able to take an inhibitor
2359 lock at the time where logind already suspends the system
2360 due to a closed lid.
2362 * logind will now wait at least 30s after each system
2363 suspend/resume cycle, and 3min after system boot before
2364 suspending the system due to a closed laptop lid. This
2365 should give USB docking stations and similar enough time to
2366 be probed and configured after system resume and boot in
2367 order to then act as suspend blocker.
2369 * systemd-run gained a new --property= setting which allows
2370 initialization of resource control properties (and others)
2371 for the created scope or service unit. Example: "systemd-run
2372 --property=BlockIOWeight=10 updatedb" may be used to run
2373 updatedb at a low block IO scheduling weight.
2375 * systemd-run's --uid=, --gid=, --setenv=, --setenv= switches
2376 now also work in --scope mode.
2378 * When systemd is compiled with kdbus support, basic support
2379 for enforced policies is now in place. (Note that enabling
2380 kdbus still voids your warranty and no API compatibility
2383 Contributions from: Andrey Borzenkov, Ansgar Burchardt, Armin
2384 K., Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
2385 Harald Hoyer, Henrik Grindal Bakken, Jasper St. Pierre, Kay
2386 Sievers, Kieran Clancy, Lennart Poettering, Lukas Nykryn,
2387 Mantas Mikulėnas, Marcel Holtmann, Mark Oteiza, Martin Pitt,
2388 Mike Gilbert, Peter Rajnoha, poma, Samuli Suominen, Stef
2389 Walter, Susant Sahani, Tero Roponen, Thomas Andersen, Thomas
2390 Bächler, Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom
2391 Gundersen, Umut Tezduyar Lindskog, Uoti Urpala, Zachary Cook,
2392 Zbigniew Jędrzejewski-Szmek
2394 -- Berlin, 2014-03-12
2398 * systemd will now relabel /dev after loading the SMACK policy
2399 according to SMACK rules.
2401 * A new unit file option AppArmorProfile= has been added to
2402 set the AppArmor profile for the processes of a unit.
2404 * A new condition check ConditionArchitecture= has been added
2405 to conditionalize units based on the system architecture, as
2406 reported by uname()'s "machine" field.
2408 * systemd-networkd now supports matching on the system
2409 virtualization, architecture, kernel command line, host name
2412 * logind is now a lot more aggressive when suspending the
2413 machine due to a closed laptop lid. Instead of acting only
2414 on the lid close action, it will continuously watch the lid
2415 status and act on it. This is useful for laptops where the
2416 power button is on the outside of the chassis so that it can
2417 be reached without opening the lid (such as the Lenovo
2418 Yoga). On those machines, logind will now immediately
2419 re-suspend the machine if the power button has been
2420 accidentally pressed while the laptop was suspended and in a
2421 backpack or similar.
2423 * logind will now watch SW_DOCK switches and inhibit reaction
2424 to the lid switch if it is pressed. This means that logind
2425 will not suspend the machine anymore if the lid is closed
2426 and the system is docked, if the laptop supports SW_DOCK
2427 notifications via the input layer. Note that ACPI docking
2428 stations do not generate this currently. Also note that this
2429 logic is usually not fully sufficient and Desktop
2430 Environments should take a lid switch inhibitor lock when an
2431 external display is connected, as systemd will not watch
2434 * nspawn will now make use of the devices cgroup controller by
2435 default, and only permit creation of and access to the usual
2436 API device nodes like /dev/null or /dev/random, as well as
2437 access to (but not creation of) the pty devices.
2439 * We will now ship a default .network file for
2440 systemd-networkd that automatically configures DHCP for
2441 network interfaces created by nspawn's --network-veth or
2442 --network-bridge= switches.
2444 * systemd will now understand the usual M, K, G, T suffixes
2445 according to SI conventions (i.e. to the base 1000) when
2446 referring to throughput and hardware metrics. It will stay
2447 with IEC conventions (i.e. to the base 1024) for software
2448 metrics, according to what is customary according to
2449 Wikipedia. We explicitly document which base applies for
2450 each configuration option.
2452 * The DeviceAllow= setting in unit files now supports a syntax
2453 to whitelist an entire group of devices node majors at once,
2454 based on the /proc/devices listing. For example, with the
2455 string "char-pts", it is now possible to whitelist all
2456 current and future pseudo-TTYs at once.
2458 * sd-event learned a new "post" event source. Event sources of
2459 this type are triggered by the dispatching of any event
2460 source of a type that is not "post". This is useful for
2461 implementing clean-up and check event sources that are
2462 triggered by other work being done in the program.
2464 * systemd-networkd is no longer statically enabled, but uses
2465 the usual [Install] sections so that it can be
2466 enabled/disabled using systemctl. It still is enabled by
2469 * When creating a veth interface pair with systemd-nspawn, the
2470 host side will now be prefixed with "vb-" if
2471 --network-bridge= is used, and with "ve-" if --network-veth
2472 is used. This way, it is easy to distinguish these cases on
2473 the host, for example to apply different configuration to
2474 them with systemd-networkd.
2476 * The compatibility libraries for libsystemd-journal.so,
2477 libsystem-id128.so, libsystemd-login.so and
2478 libsystemd-daemon.so do not make use of IFUNC
2479 anymore. Instead, we now build libsystemd.so multiple times
2480 under these alternative names. This means that the footprint
2481 is drastically increased, but given that these are
2482 transitional compatibility libraries, this should not matter
2483 much. This change has been made necessary to support the ARM
2484 platform for these compatibility libraries, as the ARM
2485 toolchain is not really at the same level as the toolchain
2486 for other architectures like x86 and does not support
2487 IFUNC. Please make sure to use --enable-compat-libs only
2488 during a transitional period!
2490 Contributions from: Andreas Fuchs, Armin K., Colin Walters,
2491 Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
2492 Holger Schurig, Jason A. Donenfeld, Jason St. John, Jasper
2493 St. Pierre, Kay Sievers, Lennart Poettering, Łukasz Stelmach,
2494 Marcel Holtmann, Michael Scherer, Michal Sekletar, Mike
2495 Gilbert, Samuli Suominen, Thomas Bächler, Thomas Hindoe
2496 Paaboel Andersen, Tom Gundersen, Umut Tezduyar Lindskog,
2497 Zbigniew Jędrzejewski-Szmek
2499 -- Berlin, 2014-02-24
2503 * A new component "systemd-networkd" has been added that can
2504 be used to configure local network interfaces statically or
2505 via DHCP. It is capable of bringing up bridges, VLANs, and
2506 bonding. Currently, no hook-ups for interactive network
2507 configuration are provided. Use this for your initrd,
2508 container, embedded, or server setup if you need a simple,
2509 yet powerful, network configuration solution. This
2510 configuration subsystem is quite nifty, as it allows wildcard
2511 hotplug matching in interfaces. For example, with a single
2512 configuration snippet, you can configure that all Ethernet
2513 interfaces showing up are automatically added to a bridge,
2514 or similar. It supports link-sensing and more.
2516 * A new tool "systemd-socket-proxyd" has been added which can
2517 act as a bidirectional proxy for TCP sockets. This is
2518 useful for adding socket activation support to services that
2519 do not actually support socket activation, including virtual
2520 machines and the like.
2522 * Add a new tool to save/restore rfkill state on
2525 * Save/restore state of keyboard backlights in addition to
2526 display backlights on shutdown/boot.
2528 * udev learned a new SECLABEL{} construct to label device
2529 nodes with a specific security label when they appear. For
2530 now, only SECLABEL{selinux} is supported, but the syntax is
2531 prepared for additional security frameworks.
2533 * udev gained a new scheme to configure link-level attributes
2534 from files in /etc/systemd/network/*.link. These files can
2535 match against MAC address, device path, driver name and type,
2536 and will apply attributes like the naming policy, link speed,
2537 MTU, duplex settings, Wake-on-LAN settings, MAC address, MAC
2538 address assignment policy (randomized, ...).
2540 * The configuration of network interface naming rules for
2541 "permanent interface names" has changed: a new NamePolicy=
2542 setting in the [Link] section of .link files determines the
2543 priority of possible naming schemes (onboard, slot, mac,
2544 path). The default value of this setting is determined by
2545 /usr/lib/net/links/99-default.link. Old
2546 80-net-name-slot.rules udev configuration file has been
2547 removed, so local configuration overriding this file should
2548 be adapated to override 99-default.link instead.
2550 * When the User= switch is used in a unit file, also
2551 initialize $SHELL= based on the user database entry.
2553 * systemd no longer depends on libdbus. All communication is
2554 now done with sd-bus, systemd's low-level bus library
2557 * kdbus support has been added to PID 1 itself. When kdbus is
2558 enabled, this causes PID 1 to set up the system bus and
2559 enable support for a new ".busname" unit type that
2560 encapsulates bus name activation on kdbus. It works a little
2561 bit like ".socket" units, except for bus names. A new
2562 generator has been added that converts classic dbus1 service
2563 activation files automatically into native systemd .busname
2566 * sd-bus: add a light-weight vtable implementation that allows
2567 defining objects on the bus with a simple static const
2568 vtable array of its methods, signals and properties.
2570 * systemd will not generate or install static dbus
2571 introspection data anymore to /usr/share/dbus-1/interfaces,
2572 as the precise format of these files is unclear, and
2573 nothing makes use of it.
2575 * A proxy daemon is now provided to proxy clients connecting
2576 via classic D-Bus AF_UNIX sockets to kdbus, to provide full
2577 compatibility with classic D-Bus.
2579 * A bus driver implementation has been added that supports the
2580 classic D-Bus bus driver calls on kdbus, also for
2581 compatibility purposes.
2583 * A new API "sd-event.h" has been added that implements a
2584 minimal event loop API built around epoll. It provides a
2585 couple of features that direct epoll usage is lacking:
2586 prioritization of events, scales to large numbers of timer
2587 events, per-event timer slack (accuracy), system-wide
2588 coalescing of timer events, exit handlers, watchdog
2589 supervision support using systemd's sd_notify() API, child
2592 * A new API "sd-rntl.h" has been added that provides an API
2593 around the route netlink interface of the kernel, similar in
2594 style to "sd-bus.h".
2596 * A new API "sd-dhcp-client.h" has been added that provides a
2597 small DHCPv4 client-side implementation. This is used by
2600 * There is a new kernel command line option
2601 "systemd.restore_state=0|1". When set to "0", none of the
2602 systemd tools will restore saved runtime state to hardware
2603 devices. More specifically, the rfkill and backlight states
2606 * The FsckPassNo= compatibility option in mount/service units
2607 has been removed. The fstab generator will now add the
2608 necessary dependencies automatically, and does not require
2609 PID1's support for that anymore.
2611 * journalctl gained a new switch, --list-boots, that lists
2612 recent boots with their times and boot IDs.
2614 * The various tools like systemctl, loginctl, timedatectl,
2615 busctl, systemd-run, ... have gained a new switch "-M" to
2616 connect to a specific, local OS container (as direct
2617 connection, without requiring SSH). This works on any
2618 container that is registered with machined, such as those
2619 created by libvirt-lxc or nspawn.
2621 * systemd-run and systemd-analyze also gained support for "-H"
2622 to connect to remote hosts via SSH. This is particularly
2623 useful for systemd-run because it enables queuing of jobs
2624 onto remote systems.
2626 * machinectl gained a new command "login" to open a getty
2627 login in any local container. This works with any container
2628 that is registered with machined (such as those created by
2629 libvirt-lxc or nspawn), and which runs systemd inside.
2631 * machinectl gained a new "reboot" command that may be used to
2632 trigger a reboot on a specific container that is registered
2633 with machined. This works on any container that runs an init
2634 system of some kind.
2636 * systemctl gained a new "list-timers" command to print a nice
2637 listing of installed timer units with the times they elapse
2640 * Alternative reboot() parameters may now be specified on the
2641 "systemctl reboot" command line and are passed to the
2642 reboot() system call.
2644 * systemctl gained a new --job-mode= switch to configure the
2645 mode to queue a job with. This is a more generic version of
2646 --fail, --irreversible, and --ignore-dependencies, which are
2647 still available but not advertised anymore.
2649 * /etc/systemd/system.conf gained new settings to configure
2650 various default timeouts of units, as well as the default
2651 start limit interval and burst. These may still be overridden
2654 * PID1 will now export on the bus profile data of the security
2655 policy upload process (such as the SELinux policy upload to
2658 * journald: when forwarding logs to the console, include
2659 timestamps (following the setting in
2660 /sys/module/printk/parameters/time).
2662 * OnCalendar= in timer units now understands the special
2663 strings "yearly" and "annually". (Both are equivalent)
2665 * The accuracy of timer units is now configurable with the new
2666 AccuracySec= setting. It defaults to 1min.
2668 * A new dependency type JoinsNamespaceOf= has been added that
2669 allows running two services within the same /tmp and network
2670 namespace, if PrivateNetwork= or PrivateTmp= are used.
2672 * A new command "cat" has been added to systemctl. It outputs
2673 the original unit file of a unit, and concatenates the
2674 contents of additional "drop-in" unit file snippets, so that
2675 the full configuration is shown.
2677 * systemctl now supports globbing on the various "list-xyz"
2678 commands, like "list-units" or "list-sockets", as well as on
2679 those commands which take multiple unit names.
2681 * journalctl's --unit= switch gained support for globbing.
2683 * All systemd daemons now make use of the watchdog logic so
2684 that systemd automatically notices when they hang.
2686 * If the $container_ttys environment variable is set,
2687 getty-generator will automatically spawn a getty for each
2688 listed tty. This is useful for container managers to request
2689 login gettys to be spawned on as many ttys as needed.
2691 * %h, %s, %U specifier support is not available anymore when
2692 used in unit files for PID 1. This is because NSS calls are
2693 not safe from PID 1. They stay available for --user
2694 instances of systemd, and as special case for the root user.
2696 * loginctl gained a new "--no-legend" switch to turn off output
2699 * The "sd-login.h" API gained three new calls:
2700 sd_session_is_remote(), sd_session_get_remote_user(),
2701 sd_session_get_remote_host() to query information about
2704 * The udev hardware database now also carries vendor/product
2705 information of SDIO devices.
2707 * The "sd-daemon.h" API gained a new sd_watchdog_enabled() to
2708 determine whether watchdog notifications are requested by
2711 * Socket-activated per-connection services now include a
2712 short description of the connection parameters in the
2715 * tmpfiles gained a new "--boot" option. When this is not used,
2716 only lines where the command character is not suffixed with
2717 "!" are executed. When this option is specified, those
2718 options are executed too. This partitions tmpfiles
2719 directives into those that can be safely executed at any
2720 time, and those which should be run only at boot (for
2721 example, a line that creates /run/nologin).
2723 * A new API "sd-resolve.h" has been added which provides a simple
2724 asynchronous wrapper around glibc NSS host name resolution
2725 calls, such as getaddrinfo(). In contrast to glibc's
2726 getaddrinfo_a(), it does not use signals. In contrast to most
2727 other asynchronous name resolution libraries, this one does
2728 not reimplement DNS, but reuses NSS, so that alternate
2729 host name resolution systems continue to work, such as mDNS,
2730 LDAP, etc. This API is based on libasyncns, but it has been
2731 cleaned up for inclusion in systemd.
2733 * The APIs "sd-journal.h", "sd-login.h", "sd-id128.h",
2734 "sd-daemon.h" are no longer found in individual libraries
2735 libsystemd-journal.so, libsystemd-login.so,
2736 libsystemd-id128.so, libsystemd-daemon.so. Instead, we have
2737 merged them into a single library, libsystemd.so, which
2738 provides all symbols. The reason for this is cyclic
2739 dependencies, as these libraries tend to use each other's
2740 symbols. So far, we have managed to workaround that by linking
2741 a copy of a good part of our code into each of these
2742 libraries again and again, which, however, makes certain
2743 things hard to do, like sharing static variables. Also, it
2744 substantially increases footprint. With this change, there
2745 is only one library for the basic APIs systemd
2746 provides. Also, "sd-bus.h", "sd-memfd.h", "sd-event.h",
2747 "sd-rtnl.h", "sd-resolve.h", "sd-utf8.h" are found in this
2748 library as well, however are subject to the --enable-kdbus
2749 switch (see below). Note that "sd-dhcp-client.h" is not part
2750 of this library (this is because it only consumes, never
2751 provides, services of/to other APIs). To make the transition
2752 easy from the separate libraries to the unified one, we
2753 provide the --enable-compat-libs compile-time switch which
2754 will generate stub libraries that are compatible with the
2755 old ones but redirect all calls to the new one.
2757 * All of the kdbus logic and the new APIs "sd-bus.h",
2758 "sd-memfd.h", "sd-event.h", "sd-rtnl.h", "sd-resolve.h",
2759 and "sd-utf8.h" are compile-time optional via the
2760 "--enable-kdbus" switch, and they are not compiled in by
2761 default. To make use of kdbus, you have to explicitly enable
2762 the switch. Note however, that neither the kernel nor the
2763 userspace API for all of this is considered stable yet. We
2764 want to maintain the freedom to still change the APIs for
2765 now. By specifying this build-time switch, you acknowledge
2766 that you are aware of the instability of the current
2769 * Also, note that while kdbus is pretty much complete,
2770 it lacks one thing: proper policy support. This means you
2771 can build a fully working system with all features; however,
2772 it will be highly insecure. Policy support will be added in
2773 one of the next releases, at the same time that we will
2774 declare the APIs stable.
2776 * When the kernel command line argument "kdbus" is specified,
2777 systemd will automatically load the kdbus.ko kernel module. At
2778 this stage of development, it is only useful for testing kdbus
2779 and should not be used in production. Note: if "--enable-kdbus"
2780 is specified, and the kdbus.ko kernel module is available, and
2781 "kdbus" is added to the kernel command line, the entire system
2782 runs with kdbus instead of dbus-daemon, with the above mentioned
2783 problem of missing the system policy enforcement. Also a future
2784 version of kdbus.ko or a newer systemd will not be compatible with
2785 each other, and will unlikely be able to boot the machine if only
2786 one of them is updated.
2788 * systemctl gained a new "import-environment" command which
2789 uploads the caller's environment (or parts thereof) into the
2790 service manager so that it is inherited by services started
2791 by the manager. This is useful to upload variables like
2792 $DISPLAY into the user service manager.
2794 * A new PrivateDevices= switch has been added to service units
2795 which allows running a service with a namespaced /dev
2796 directory that does not contain any device nodes for
2797 physical devices. More specifically, it only includes devices
2798 such as /dev/null, /dev/urandom, and /dev/zero which are API
2801 * logind has been extended to support behaviour like VT
2802 switching on seats that do not support a VT. This makes
2803 multi-session available on seats that are not the first seat
2804 (seat0), and on systems where kernel support for VTs has
2805 been disabled at compile-time.
2807 * If a process holds a delay lock for system sleep or shutdown
2808 and fails to release it in time, we will now log its
2809 identity. This makes it easier to identify processes that
2810 cause slow suspends or power-offs.
2812 * When parsing /etc/crypttab, support for a new key-slot=
2813 option as supported by Debian is added. It allows indicating
2814 which LUKS slot to use on disk, speeding up key loading.
2816 * The sd_journald_sendv() API call has been checked and
2817 officially declared to be async-signal-safe so that it may
2818 be invoked from signal handlers for logging purposes.
2820 * Boot-time status output is now enabled automatically after a
2821 short timeout if boot does not progress, in order to give
2822 the user an indication what she or he is waiting for.
2824 * The boot-time output has been improved to show how much time
2825 remains until jobs expire.
2827 * The KillMode= switch in service units gained a new possible
2828 value "mixed". If set, and the unit is shut down, then the
2829 initial SIGTERM signal is sent only to the main daemon
2830 process, while the following SIGKILL signal is sent to
2831 all remaining processes of the service.
2833 * When a scope unit is registered, a new property "Controller"
2834 may be set. If set to a valid bus name, systemd will send a
2835 RequestStop() signal to this name when it would like to shut
2836 down the scope. This may be used to hook manager logic into
2837 the shutdown logic of scope units. Also, scope units may now
2838 be put in a special "abandoned" state, in which case the
2839 manager process which created them takes no further
2840 responsibilities for it.
2842 * When reading unit files, systemd will now verify
2843 the access mode of these files, and warn about certain
2844 suspicious combinations. This has been added to make it
2845 easier to track down packaging bugs where unit files are
2846 marked executable or world-writable.
2848 * systemd-nspawn gained a new "--setenv=" switch to set
2849 container-wide environment variables. The similar option in
2850 systemd-activate was renamed from "--environment=" to
2851 "--setenv=" for consistency.
2853 * systemd-nspawn has been updated to create a new kdbus domain
2854 for each container that is invoked, thus allowing each
2855 container to have its own set of system and user buses,
2856 independent of the host.
2858 * systemd-nspawn gained a new --drop-capability= switch to run
2859 the container with less capabilities than the default. Both
2860 --drop-capability= and --capability= now take the special
2861 string "all" for dropping or keeping all capabilities.
2863 * systemd-nspawn gained new switches for executing containers
2864 with specific SELinux labels set.
2866 * systemd-nspawn gained a new --quiet switch to not generate
2867 any additional output but the container's own console
2870 * systemd-nspawn gained a new --share-system switch to run a
2871 container without PID namespacing enabled.
2873 * systemd-nspawn gained a new --register= switch to control
2874 whether the container is registered with systemd-machined or
2875 not. This is useful for containers that do not run full
2876 OS images, but only specific apps.
2878 * systemd-nspawn gained a new --keep-unit which may be used
2879 when invoked as the only program from a service unit, and
2880 results in registration of the unit service itself in
2881 systemd-machined, instead of a newly opened scope unit.
2883 * systemd-nspawn gained a new --network-interface= switch for
2884 moving arbitrary interfaces to the container. The new
2885 --network-veth switch creates a virtual Ethernet connection
2886 between host and container. The new --network-bridge=
2887 switch then allows assigning the host side of this virtual
2888 Ethernet connection to a bridge device.
2890 * systemd-nspawn gained a new --personality= switch for
2891 setting the kernel personality for the container. This is
2892 useful when running a 32-bit container on a 64-bit host. A
2893 similar option Personality= is now also available for service
2896 * logind will now also track a "Desktop" identifier for each
2897 session which encodes the desktop environment of it. This is
2898 useful for desktop environments that want to identify
2899 multiple running sessions of itself easily.
2901 * A new SELinuxContext= setting for service units has been
2902 added that allows setting a specific SELinux execution
2903 context for a service.
2905 * Most systemd client tools will now honour $SYSTEMD_LESS for
2906 settings of the "less" pager. By default, these tools will
2907 override $LESS to allow certain operations to work, such as
2908 jump-to-the-end. With $SYSTEMD_LESS, it is possible to
2909 influence this logic.
2911 * systemd's "seccomp" hook-up has been changed to make use of
2912 the libseccomp library instead of using its own
2913 implementation. This has benefits for portability among
2916 * For usage together with SystemCallFilter=, a new
2917 SystemCallErrorNumber= setting has been introduced that
2918 allows configuration of a system error number to be returned
2919 on filtered system calls, instead of immediately killing the
2920 process. Also, SystemCallArchitectures= has been added to
2921 limit access to system calls of a particular architecture
2922 (in order to turn off support for unused secondary
2923 architectures). There is also a global
2924 SystemCallArchitectures= setting in system.conf now to turn
2925 off support for non-native system calls system-wide.
2927 * systemd requires a kernel with a working name_to_handle_at(),
2928 please see the kernel config requirements in the README file.
2930 Contributions from: Adam Williamson, Alex Jia, Anatol Pomozov,
2931 Ansgar Burchardt, AppleBloom, Auke Kok, Bastien Nocera,
2932 Chengwei Yang, Christian Seiler, Colin Guthrie, Colin Walters,
2933 Cristian Rodríguez, Daniel Buch, Daniele Medri, Daniel J
2934 Walsh, Daniel Mack, Dan McGee, Dave Reisner, David Coppa,
2935 David Herrmann, David Strauss, Djalal Harouni, Dmitry Pisklov,
2936 Elia Pinto, Florian Weimer, George McCollister, Goffredo
2937 Baroncelli, Greg Kroah-Hartman, Hendrik Brueckner, Igor
2938 Zhbanov, Jan Engelhardt, Jan Janssen, Jason A. Donenfeld,
2939 Jason St. John, Jasper St. Pierre, Jóhann B. Guðmundsson, Jose
2940 Ignacio Naranjo, Karel Zak, Kay Sievers, Kristian Høgsberg,
2941 Lennart Poettering, Lubomir Rintel, Lukas Nykryn, Lukasz
2942 Skalski, Łukasz Stelmach, Luke Shumaker, Mantas Mikulėnas,
2943 Marc-Antoine Perennou, Marcel Holtmann, Marcos Felipe Rasia de
2944 Mello, Marko Myllynen, Martin Pitt, Matthew Monaco, Michael
2945 Marineau, Michael Scherer, Michał Górny, Michal Sekletar,
2946 Michele Curti, Oleksii Shevchuk, Olivier Brunel, Patrik Flykt,
2947 Pavel Holica, Raudi, Richard Marko, Ronny Chevalier, Sébastien
2948 Luttringer, Sergey Ptashnick, Shawn Landden, Simon Peeters,
2949 Stefan Beller, Susant Sahani, Sylvain Plantefeve, Sylvia Else,
2950 Tero Roponen, Thomas Bächler, Thomas Hindoe Paaboel Andersen,
2951 Tom Gundersen, Umut Tezduyar Lindskog, Unai Uribarri, Václav
2952 Pavlín, Vincent Batts, WaLyong Cho, William Giokas, Yang
2953 Zhiyong, Yin Kangkai, Yuxuan Shui, Zbigniew Jędrzejewski-Szmek
2955 -- Berlin, 2014-02-20
2959 * logind has gained support for facilitating privileged input
2960 and drm device access for unprivileged clients. This work is
2961 useful to allow Wayland display servers (and similar
2962 programs, such as kmscon) to run under the user's ID and
2963 access input and drm devices which are normally
2964 protected. When this is used (and the kernel is new enough)
2965 logind will "mute" IO on the file descriptors passed to
2966 Wayland as long as it is in the background and "unmute" it
2967 if it returns into the foreground. This allows secure
2968 session switching without allowing background sessions to
2969 eavesdrop on input and display data. This also introduces
2970 session switching support if VT support is turned off in the
2971 kernel, and on seats that are not seat0.
2973 * A new kernel command line option luks.options= is understood
2974 now which allows specifying LUKS options for usage for LUKS
2975 encrypted partitions specified with luks.uuid=.
2977 * tmpfiles.d(5) snippets may now use specifier expansion in
2978 path names. More specifically %m, %b, %H, %v, are now
2979 replaced by the local machine id, boot id, hostname, and
2980 kernel version number.
2982 * A new tmpfiles.d(5) command "m" has been introduced which
2983 may be used to change the owner/group/access mode of a file
2984 or directory if it exists, but do nothing if it does not.
2986 * This release removes high-level support for the
2987 MemorySoftLimit= cgroup setting. The underlying kernel
2988 cgroup attribute memory.soft_limit= is currently badly
2989 designed and likely to be removed from the kernel API in its
2990 current form, hence we should not expose it for now.
2992 * The memory.use_hierarchy cgroup attribute is now enabled for
2993 all cgroups systemd creates in the memory cgroup
2994 hierarchy. This option is likely to be come the built-in
2995 default in the kernel anyway, and the non-hierarchical mode
2996 never made much sense in the intrinsically hierarchical
2999 * A new field _SYSTEMD_SLICE= is logged along with all journal
3000 messages containing the slice a message was generated
3001 from. This is useful to allow easy per-customer filtering of
3002 logs among other things.
3004 * systemd-journald will no longer adjust the group of journal
3005 files it creates to the "systemd-journal" group. Instead we
3006 rely on the journal directory to be owned by the
3007 "systemd-journal" group, and its setgid bit set, so that the
3008 kernel file system layer will automatically enforce that
3009 journal files inherit this group assignment. The reason for
3010 this change is that we cannot allow NSS look-ups from
3011 journald which would be necessary to resolve
3012 "systemd-journal" to a numeric GID, because this might
3013 create deadlocks if NSS involves synchronous queries to
3014 other daemons (such as nscd, or sssd) which in turn are
3015 logging clients of journald and might block on it, which
3016 would then dead lock. A tmpfiles.d(5) snippet included in
3017 systemd will make sure the setgid bit and group are
3018 properly set on the journal directory if it exists on every
3019 boot. However, we recommend adjusting it manually after
3020 upgrades too (or from RPM scriptlets), so that the change is
3021 not delayed until next reboot.
3023 * Backlight and random seed files in /var/lib/ have moved into
3024 the /var/lib/systemd/ directory, in order to centralize all
3025 systemd generated files in one directory.
3027 * Boot time performance measurements (as displayed by
3028 "systemd-analyze" for example) will now read ACPI 5.0 FPDT
3029 performance information if that's available to determine how
3030 much time BIOS and boot loader initialization required. With
3031 a sufficiently new BIOS you hence no longer need to boot
3032 with Gummiboot to get access to such information.
3034 Contributions from: Andrey Borzenkov, Chen Jie, Colin Walters,
3035 Cristian Rodríguez, Dave Reisner, David Herrmann, David
3036 Mackey, David Strauss, Eelco Dolstra, Evan Callicoat, Gao
3037 feng, Harald Hoyer, Jimmie Tauriainen, Kay Sievers, Lennart
3038 Poettering, Lukas Nykryn, Mantas Mikulėnas, Martin Pitt,
3039 Michael Scherer, Michał Górny, Mike Gilbert, Patrick McCarty,
3040 Sebastian Ott, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
3042 -- Berlin, 2013-10-02
3046 * The Restart= option for services now understands a new
3047 on-watchdog setting, which will restart the service
3048 automatically if the service stops sending out watchdog keep
3049 alive messages (as configured with WatchdogSec=).
3051 * The getty generator (which is responsible for bringing up a
3052 getty on configured serial consoles) will no longer only
3053 start a getty on the primary kernel console but on all
3054 others, too. This makes the order in which console= is
3055 specified on the kernel command line less important.
3057 * libsystemd-logind gained a new sd_session_get_vt() call to
3058 retrieve the VT number of a session.
3060 * If the option "tries=0" is set for an entry of /etc/crypttab
3061 its passphrase is queried indefinitely instead of any
3062 maximum number of tries.
3064 * If a service with a configure PID file terminates its PID
3065 file will now be removed automatically if it still exists
3066 afterwards. This should put an end to stale PID files.
3068 * systemd-run will now also take relative binary path names
3069 for execution and no longer insists on absolute paths.
3071 * InaccessibleDirectories= and ReadOnlyDirectories= now take
3072 paths that are optionally prefixed with "-" to indicate that
3073 it should not be considered a failure if they do not exist.
3075 * journalctl -o (and similar commands) now understands a new
3076 output mode "short-precise", it is similar to "short" but
3077 shows timestamps with usec accuracy.
3079 * The option "discard" (as known from Debian) is now
3080 synonymous to "allow-discards" in /etc/crypttab. In fact,
3081 "discard" is preferred now (since it is easier to remember
3084 * Some licensing clean-ups were made, so that more code is now
3085 LGPL-2.1 licensed than before.
3087 * A minimal tool to save/restore the display backlight
3088 brightness across reboots has been added. It will store the
3089 backlight setting as late as possible at shutdown, and
3090 restore it as early as possible during reboot.
3092 * A logic to automatically discover and enable home and swap
3093 partitions on GPT disks has been added. With this in place
3094 /etc/fstab becomes optional for many setups as systemd can
3095 discover certain partitions located on the root disk
3096 automatically. Home partitions are recognized under their
3097 GPT type ID 933ac7e12eb44f13b8440e14e2aef915. Swap
3098 partitions are recognized under their GPT type ID
3099 0657fd6da4ab43c484e50933c84b4f4f.
3101 * systemd will no longer pass any environment from the kernel
3102 or initrd to system services. If you want to set an
3103 environment for all services, do so via the kernel command
3104 line systemd.setenv= assignment.
3106 * The systemd-sysctl tool no longer natively reads the file
3107 /etc/sysctl.conf. If desired, the file should be symlinked
3108 from /etc/sysctl.d/99-sysctl.conf. Apart from providing
3109 legacy support by a symlink rather than built-in code, it
3110 also makes the otherwise hidden order of application of the
3111 different files visible. (Note that this partly reverts to a
3112 pre-198 application order of sysctl knobs!)
3114 * The "systemctl set-log-level" and "systemctl dump" commands
3115 have been moved to systemd-analyze.
3117 * systemd-run learned the new --remain-after-exit switch,
3118 which causes the scope unit not to be cleaned up
3119 automatically after the process terminated.
3121 * tmpfiles learned a new --exclude-prefix= switch to exclude
3122 certain paths from operation.
3124 * journald will now automatically flush all messages to disk
3125 as soon as a message at the log level CRIT, ALERT or EMERG
3128 Contributions from: Andrew Cook, Brandon Philips, Christian
3129 Hesse, Christoph Junghans, Colin Walters, Daniel Schaal,
3130 Daniel Wallace, Dave Reisner, David Herrmann, Gao feng, George
3131 McCollister, Giovanni Campagna, Hannes Reinecke, Harald Hoyer,
3132 Herczeg Zsolt, Holger Hans Peter Freyther, Jan Engelhardt,
3133 Jesper Larsen, Kay Sievers, Khem Raj, Lennart Poettering,
3134 Lukas Nykryn, Maciej Wereski, Mantas Mikulėnas, Marcel
3135 Holtmann, Martin Pitt, Michael Biebl, Michael Marineau,
3136 Michael Scherer, Michael Stapelberg, Michal Sekletar, Michał
3137 Górny, Olivier Brunel, Ondrej Balaz, Ronny Chevalier, Shawn
3138 Landden, Steven Hiscocks, Thomas Bächler, Thomas Hindoe
3139 Paaboel Andersen, Tom Gundersen, Umut Tezduyar, WANG Chao,
3140 William Giokas, Zbigniew Jędrzejewski-Szmek
3142 -- Berlin, 2013-09-13
3146 * The documentation has been updated to cover the various new
3147 concepts introduced with 205.
3149 * Unit files now understand the new %v specifier which
3150 resolves to the kernel version string as returned by "uname
3153 * systemctl now supports filtering the unit list output by
3154 load state, active state and sub state, using the new
3157 * "systemctl status" will now show the results of the
3158 condition checks (like ConditionPathExists= and similar) of
3159 the last start attempts of the unit. They are also logged to
3162 * "journalctl -b" may now be used to look for boot output of a
3163 specific boot. Try "journalctl -b -1" for the previous boot,
3164 but the syntax is substantially more powerful.
3166 * "journalctl --show-cursor" has been added which prints the
3167 cursor string the last shown log line. This may then be used
3168 with the new "journalctl --after-cursor=" switch to continue
3169 browsing logs from that point on.
3171 * "journalctl --force" may now be used to force regeneration
3174 * Creation of "dead" device nodes has been moved from udev
3175 into kmod and tmpfiles. Previously, udev would read the kmod
3176 databases to pre-generate dead device nodes based on meta
3177 information contained in kernel modules, so that these would
3178 be auto-loaded on access rather then at boot. As this
~ianmdlvl / elogind.git / blob