1 systemd System and Service Manager
5 * journalctl gained the new options -t/--identifier= to match
6 on the syslog identifier (aka "tag"), as well as --utc to
7 show log timestamps in the UTC timezone. journalctl now also
8 accepts -n/--lines=all to disable line capping in a pager.
10 * Services can notify the manager before they start a reload
11 (by sending RELOADING=1) or shutdown (by sending
12 STOPPING=1). This allows the manager to track and show the
13 internal state of daemons and closes a race condition when
14 the process is still running but has closed its D-Bus
17 * Services with Type=oneshot do not have to have any ExecStart
20 * User units are now loaded also from
21 $XDG_RUNTIME_DIR/systemd/user/. This is similar to the
22 /run/systemd/user directory that was already previously
23 supported, but is under the control of the user.
25 * A timeout for the bootup of the whole system can now be
26 configured. The system can be configured to reboot or
27 poweroff if the basic system default target is not reached
28 before the timeout (new StartTimeoutSec=,
29 StartTimeoutAction=, StartTimeoutRebootArgument= options).
31 * systemd-logind can be configured to also handle lid switch
32 events even when the machine is docked or multiple displays
33 are attached (HandleLidSwitchDocked= option).
35 * A helper binary and a service have been added which can be
36 used to resume from hibernation in the initramfs. A
37 generator will parse the resume= option on the kernel
38 command-line to trigger resume.
40 * A user console daemon systemd-consoled has been
41 added. Currently, it is a preview, and will so far open a
42 single terminal on each session of the user marked as
43 Desktop=SYSTEMD-CONSOLE.
45 * Route metrics can be specified for DHCP routes added by
48 * The SELinux context of socket-actived services can be set
49 from the information provided by the networking stack
50 (SELinuxContextFromNet= option).
52 * Userspace firmware loading support has been removed and
53 the minimum supported kernel version is thus bumped to 3.7.
55 * Timeout for udev workers has been increased from 1 to 3
56 minutes, but a warning will be printed after 1 minute to
57 help diagnose kernel modules that take a long time to load.
59 * Udev rules can now remove tags on devices with TAG-="foobar".
61 * systemd's readahead implementation has been removed. In many
62 circumstatances it didn't give expected benefits even for
63 rotational disk drives and was becoming less relevant in the
64 age of SSDs. As none of the developers has been using
65 rotating media anymore, and nobody stepped up to actively
66 maintain this component of systemd it has now been removed.
68 * Swap units can use Discard= to specify discard options.
69 Discard options specified for swaps in /etc/fstab are now
72 * Docker containers are now detected as a separate type of
75 * The Password Agent protocol gained support for queries where
76 the user input is shown, useful e.g. for usernames.
77 systemd-ask-password gained a new --echo option to turn that
80 * The default sysctl.d/ snippets will now set:
82 net.core.default_qdisc = fq_codel
84 This selects Fair Queueing Controlled Delay as the default
85 queueing discipline for network interfaces. fq_codel helps
86 fight the network bufferbloat problem. It is believed to be
87 a good default with no tuning required for most workloads.
88 Downstream distributions may override this choice. On 10Gbit
89 servers that do not do forwarding, "fq" may perform better.
90 Systems without a good clocksource should use "pfifo_fast".
92 * If kdbus is enabled during build a new option BusPolicy= is
93 available for service units, that allows locking all service
94 processes into a stricter bus policy, in order to limit
95 access to various bus services, or even hide most of them
96 from the service's view entirely.
98 * networkctl will now show the .network and .link file
99 networkd has applied to a specific interface.
101 * sd-login gained a new API call sd_session_get_desktop() to
102 query which desktop environment has been selected for a
105 * UNIX utmp support is now compile-time optional to support
108 * systemctl gained two new commands "add-wants" and
109 "add-requires" for pulling in units from specific targets
112 * If the word "rescue" is specified on the kernel command line
113 the system will now boot into rescue mode (aka
114 rescue.target), which was previously available only by
115 specifying "1" or "systemd.unit=rescue.target" on the kernel
116 command line. This new kernel command line option nicely
117 mirrors the already existing "emergency" kernel command line
120 * New kernel command line options mount.usr=, mount.usrflags=,
121 usrfstype= have been added that match root=, rootflags=,
122 rootfstype= but allow mounting a specific file system to
125 * The $NOTIFY_SOCKET is now also passed to control processesof
126 services, not only the main process.
128 * This version reenables support for fsck's -l switch. This
129 means at least version v2.25 of util-linux is required for
130 operation, otherwise dead-locks on device nodes may
131 occur. Again: you need to update util-linux to at least
132 v2.25 when updating systemd to v217.
136 * timedated no longer reads NTP implementation unit names from
137 /usr/lib/systemd/ntp-units.d/*.list. Alternative NTP
138 implementations should add a
140 Conflicts=systemd-timesyncd.service
142 to their unit files to take over and replace systemd's NTP
143 default functionality.
145 * systemd-sysusers gained a new line type "r" for configuring
146 which UID/GID ranges to allocate system users/groups
147 from. Lines of type "u" may now add an additional column
148 that specifies the home directory for the system user to be
149 created. Also, systemd-sysusers may now optionally read user
150 information from STDIN instead of a file. This is useful for
151 invoking it from RPM preinst scriptlets that need to create
152 users before the first RPM file is installed since these
153 files might need to be owned by them. A new
154 %sysusers_create_inline RPM macro has been introduced to do
155 just that. systemd-sysusers now updates the shadow files as
156 well as the user/group databases, which should enhance
157 compatibility with certain tools like grpck.
159 * A number of bus APIs of PID 1 now optionally consult
160 PolicyKit to permit access for otherwise unprivileged
161 clients under certain conditions. Note that this currently
162 doesn't support interactive authentication yet, but this is
163 expected to be added eventually, too.
165 * /etc/machine-info now has new fields for configuring the
166 deployment environment of the machine, as well as the
167 location of the machine. hostnamectl has been updated with
168 new command to update these fields.
170 * systemd-timesyncd has been updated to automatically acquire
171 NTP server information from systemd-networkd, which might
172 have been discovered via DHCP.
174 * systemd-resolved now includes a caching DNS stub resolver
175 and a complete LLMNR name resolution implementation. A new
176 NSS module "nss-resolve" has been added which can be used
177 instead of glibc's own "nss-dns" to resolve hostnames via
178 systemd-resolved. Hostnames, addresses and arbitrary RRs may
179 be resolved via systemd-resolved D-Bus APIs. In contrast to
180 the glibc internal resolver systemd-resolved is aware of
181 multi-homed system, and keeps DNS server and caches separate
182 and per-interface. Queries are sent simultaneously on all
183 interfaces that have DNS servers configured, in order to
184 properly handle VPNs and local LANs which might resolve
185 separate sets of domain names. systemd-resolved may acquire
186 DNS server information from systemd-networkd automatically,
187 which in turn might have discovered them via DHCP. A tool
188 "systemd-resolve-host" has been added that may be used to
189 query the DNS logic in resolved. systemd-resolved implements
190 IDNA and automatically uses IDNA or UTF-8 encoding depending
191 on whether classic DNS or LLMNR is used as transport. In the
192 next releases we intend to add a DNSSEC and mDNS/DNS-SD
193 implementation to systemd-resolved.
195 * A new NSS module nss-mymachines has been added, that
196 automatically resolves the names of all local registered
197 containers to their respective IP addresses.
199 * A new client tool "networkctl" for systemd-networkd has been
200 added. It currently is entirely passive and will query
201 networking configuration from udev, rtnetlink and networkd,
202 and present it to the user in a very friendly
203 way. Eventually, we hope to extend it to become a full
204 control utility for networkd.
206 * .socket units gained a new DeferAcceptSec= setting that
207 controls the kernels' TCP_DEFER_ACCEPT sockopt for
208 TCP. Similar, support for controlling TCP keep-alive
209 settings has been added (KeepAliveTimeSec=,
210 KeepAliveIntervalSec=, KeepAliveProbes=). Also, support for
211 turning off Nagle's algorithm on TCP has been added
214 * logind learned a new session type "web", for use in projects
215 like Cockpit which register web clients as PAM sessions.
217 * timer units with at least one OnCalendar= setting will now
218 be started only after timer-sync.target has been
219 reached. This way they will not elapse before the system
220 clock has been corrected by a local NTP client or
221 similar. This is particular useful on RTC-less embedded
222 machines, that come up with an invalid system clock.
224 * systemd-nspawn's --network-veth= switch should now result in
225 stable MAC addresses for both the outer and the inner side
228 * systemd-nspawn gained a new --volatile= switch for running
229 container instances with /etc or /var unpopulated.
231 * The kdbus client code has been updated to use the new Linux
232 3.17 memfd subsystem instead of the old kdbus-specific one.
234 * systemd-networkd's DHCP client and server now support
235 FORCERENEW. There are also new configuration options to
236 configure the vendor client identifier and broadcast mode
239 * systemd will no longer inform the kernel about the current
240 timezone, as this is necessarily incorrect and racy as the
241 kernel has no understanding of DST and similar
242 concepts. This hence means FAT timestamps will be always
243 considered UTC, similar to what Android is already
244 doing. Also, when the RTC is configured to the local time
245 (rather than UTC) systemd will never synchronize back to it,
246 as this might confuse Windows at a later boot.
248 * systemd-analyze gained a new command "verify" for offline
249 validation of unit files.
251 * systemd-networkd gained support for a couple of additional
252 settings for bonding networking setups. Also, the metric for
253 statically configured routes may now be configured. For
254 network interfaces where this is appropriate the peer IP
255 address may now be configured.
257 * systemd-networkd's DHCP client will no longer request
258 broadcasting by default, as this tripped up some networks.
259 For hardware where broadcast is required the feature should
260 be switched back on using RequestBroadcast=yes.
262 * systemd-networkd will now set up IPv4LL addresses (when
263 enabled) even if DHCP is configured successfully.
265 * udev will now default to respect network device names given
266 by the kernel when the kernel indicates that these are
267 predictable. This behavior can be tweaked by changing
268 NamePolicy= in the relevant .link file.
270 * A new library systemd-terminal has been added that
271 implements full TTY stream parsing and rendering. This
272 library is supposed to be used later on for implementing a
273 full userspace VT subsystem, replacing the current kernel
276 * A new tool systemd-journal-upload has been added to push
277 journal data to a remote system running
278 systemd-journal-remote.
280 * journald will no longer forward all local data to another
281 running syslog daemon. This change has been made because
282 rsyslog (which appears to be the most commonly used syslog
283 implementation these days) no longer makes use of this, and
284 instead pulls the data out of the journal on its own. Since
285 forwarding the messages to a non-existent syslog server is
286 more expensive than we assumed we have now turned this
287 off. If you run a syslog server that is not a recent rsyslog
288 version, you have to turn this option on again
289 (ForwardToSyslog= in journald.conf).
291 * journald now optionally supports the LZ4 compressor for
292 larger journal fields. This compressor should perform much
293 better than XZ which was the previous default.
295 * machinectl now shows the IP addresses of local containers,
296 if it knows them, plus the interface name of the container.
298 * A new tool "systemd-escape" has been added that makes it
299 easy to escape strings to build unit names and similar.
301 * sd_notify() messages may now include a new ERRNO= field
302 which is parsed and collected by systemd and shown among the
303 "systemctl status" output for a service.
305 * A new component "systemd-firstboot" has been added that
306 queries the most basic systemd information (timezone,
307 hostname, root password) interactively on first
308 boot. Alternatively it may also be used to provision these
309 things offline on OS images installed into directories.
311 * The default sysctl.d/ snippets will now set
313 net.ipv4.conf.default.promote_secondaries=1
315 This has the benefit of no flushing secondary IP addresses
316 when primary addresses are removed.
318 Contributions from: Ansgar Burchardt, Bastien Nocera, Colin
319 Walters, Dan Dedrick, Daniel Buch, Daniel Korostil, Daniel
320 Mack, Dan Williams, Dave Reisner, David Herrmann, Denis
321 Kenzior, Eelco Dolstra, Eric Cook, Hannes Reinecke, Harald
322 Hoyer, Hong Shick Pak, Hui Wang, Jean-André Santoni, Jóhann
323 B. Guðmundsson, Jon Severinsson, Karel Zak, Kay Sievers, Kevin
324 Wells, Lennart Poettering, Lukas Nykryn, Mantas Mikulėnas,
325 Marc-Antoine Perennou, Martin Pitt, Michael Biebl, Michael
326 Marineau, Michael Olbrich, Michal Schmidt, Michal Sekletar,
327 Miguel Angel Ajo, Mike Gilbert, Olivier Brunel, Robert
328 Schiele, Ronny Chevalier, Simon McVittie, Sjoerd Simons, Stef
329 Walter, Steven Noonan, Susant Sahani, Tanu Kaskinen, Thomas
330 Blume, Thomas Hindoe Paaboel Andersen, Timofey Titovets,
331 Tobias Geerinckx-Rice, Tomasz Torcz, Tom Gundersen, Umut
332 Tezduyar Lindskog, Zbigniew Jędrzejewski-Szmek
334 -- Berlin, 2014-08-19
338 * A new tool systemd-sysusers has been added. This tool
339 creates system users and groups in /etc/passwd and
340 /etc/group, based on static declarative system user/group
341 definitions in /usr/lib/sysusers.d/. This is useful to
342 enable factory resets and volatile systems that boot up with
343 an empty /etc directory, and thus need system users and
344 groups created during early boot. systemd now also ships
345 with two default sysusers.d/ files for the most basic
346 users and groups systemd and the core operating system
349 * A new tmpfiles snippet has been added that rebuilds the
350 essential files in /etc on boot, should they be missing.
352 * A directive for ensuring automatic clean-up of
353 /var/cache/man/ has been removed from the default
354 configuration. This line should now be shipped by the man
355 implementation. The necessary change has been made to the
356 man-db implementation. Note that you need to update your man
357 implementation to one that ships this line, otherwise no
358 automatic clean-up of /var/cache/man will take place.
360 * A new condition ConditionNeedsUpdate= has been added that
361 may conditionalize services to only run when /etc or /var
362 are "older" than the vendor operating system resources in
363 /usr. This is useful for reconstructing or updating /etc
364 after an offline update of /usr or a factory reset, on the
365 next reboot. Services that want to run once after such an
366 update or reset should use this condition and order
367 themselves before the new systemd-update-done.service, which
368 will mark the two directories as fully updated. A number of
369 service files have been added making use of this, to rebuild
370 the udev hardware database, the journald message catalog and
371 dynamic loader cache (ldconfig). The systemd-sysusers tool
372 described above also makes use of this now. With this in
373 place it is now possible to start up a minimal operating
374 system with /etc empty cleanly. For more information on the
375 concepts involved see this recent blog story:
377 http://0pointer.de/blog/projects/stateless.html
379 * A new system group "input" has been introduced, and all
380 input device nodes get this group assigned. This is useful
381 for system-level software to get access to input devices. It
382 complements what is already done for "audio" and "video".
384 * systemd-networkd learnt minimal DHCPv4 server support in
385 addition to the existing DHCPv4 client support. It also
386 learnt DHCPv6 client and IPv6 Router Solicitation client
387 support. The DHCPv4 client gained support for static routes
388 passed in from the server. Note that the [DHCPv4] section
389 known in older systemd-networkd versions has been renamed to
390 [DHCP] and is now also used by the DHCPv6 client. Existing
391 .network files using settings of this section should be
392 updated, though compatibility is maintained. Optionally, the
393 client hostname may now be sent to the DHCP server.
395 * networkd gained support for vxlan virtual networks as well
396 as tun/tap and dummy devices.
398 * networkd gained support for automatic allocation of address
399 ranges for interfaces from a system-wide pool of
400 addresses. This is useful for dynamically managing a large
401 number of interfaces with a single network configuration
402 file. In particular this is useful to easily assign
403 appropriate IP addresses to the veth links of a large number
406 * RPM macros for processing sysusers, sysctl and binfmt
407 drop-in snippets at package installation time have been
410 * The /etc/os-release file should now be placed in
411 /usr/lib/os-release. The old location is automatically
412 created as symlink. /usr/lib is the more appropriate
413 location of this file, since it shall actually describe the
414 vendor operating system shipped in /usr, and not the
415 configuration stored in /etc.
417 * .mount units gained a new boolean SloppyOptions= setting
418 that maps to mount(8)'s -s option which enables permissive
419 parsing of unknown mount options.
421 * tmpfiles learnt a new "L+" directive which creates a symlink
422 but (unlike "L") deletes a pre-existing file first, should
423 it already exist and not already be the correct
424 symlink. Similar, "b+", "c+" and "p+" directives have been
425 added as well, which create block and character devices, as
426 well as fifos in the filesystem, possibly removing any
427 pre-existing files of different types.
429 * For tmpfiles' "L", "L+", "C" and "C+" directives the final
430 'argument' field (which so far specified the source to
431 symlink/copy the files from) is now optional. If omitted the
432 same file os copied from /usr/share/factory/ suffixed by the
433 full destination path. This is useful for populating /etc
434 with essential files, by copying them from vendor defaults
435 shipped in /usr/share/factory/etc.
437 * A new command "systemctl preset-all" has been added that
438 applies the service preset settings to all installed unit
439 files. A new switch --preset-mode= has been added that
440 controls whether only enable or only disable operations
443 * A new command "systemctl is-system-running" has been added
444 that allows checking the overall state of the system, for
445 example whether it is fully up and running.
447 * When the system boots up with an empty /etc, the equivalent
448 to "systemctl preset-all" is executed during early boot, to
449 make sure all default services are enabled after a factory
452 * systemd now contains a minimal preset file that enables the
453 most basic services systemd ships by default.
455 * Unit files' [Install] section gained a new DefaultInstance=
456 field for defining the default instance to create if a
457 template unit is enabled with no instance specified.
459 * A new passive target cryptsetup-pre.target has been added
460 that may be used by services that need to make they run and
461 finish before the first LUKS cryptographic device is set up.
463 * The /dev/loop-control and /dev/btrfs-control device nodes
464 are now owned by the "disk" group by default, opening up
465 access to this group.
467 * systemd-coredump will now automatically generate a
468 stack trace of all core dumps taking place on the system,
469 based on elfutils' libdw library. This stack trace is logged
472 * systemd-coredump may now optionally store coredumps directly
473 on disk (in /var/lib/systemd/coredump, possibly compressed),
474 instead of storing them unconditionally in the journal. This
475 mode is the new default. A new configuration file
476 /etc/systemd/coredump.conf has been added to configure this
477 and other parameters of systemd-coredump.
479 * coredumpctl gained a new "info" verb to show details about a
480 specific coredump. A new switch "-1" has also been added
481 that makes sure to only show information about the most
482 recent entry instead of all entries. Also, as the tool is
483 generally useful now the "systemd-" prefix of the binary
484 name has been removed. Distributions that want to maintain
485 compatibility with the old name should add a symlink from
486 the old name to the new name.
488 * journald's SplitMode= now defaults to "uid". This makes sure
489 that unprivileged users can access their own coredumps with
490 coredumpctl without restrictions.
492 * New kernel command line options "systemd.wants=" (for
493 pulling an additional unit during boot), "systemd.mask="
494 (for masking a specific unit for the boot), and
495 "systemd.debug-shell" (for enabling the debug shell on tty9)
496 have been added. This is implemented in the new generator
497 "systemd-debug-generator".
499 * systemd-nspawn will now by default filter a couple of
500 syscalls for containers, among them those required for
501 kernel module loading, direct x86 IO port access, swap
502 management, and kexec. Most importantly though
503 open_by_handle_at() is now prohibited for containers,
504 closing a hole similar to a recently discussed vulnerability
505 in docker regarding access to files on file hierarchies the
506 container should normally not have access to. Note that for
507 nspawn we generally make no security claims anyway (and
508 this is explicitly documented in the man page), so this is
509 just a fix for one of the most obvious problems.
511 * A new man page file-hierarchy(7) has been added that
512 contains a minimized, modernized version of the file system
513 layout systemd expects, similar in style to the FHS
514 specification or hier(5). A new tool systemd-path(1) has
515 been added to query many of these paths for the local
518 * Automatic time-based clean-up of $XDG_RUNTIME_DIR is no
519 longer done. Since the directory now has a per-user size
520 limit, and is cleaned on logout this appears unnecessary,
521 in particular since this now brings the lifecycle of this
522 directory closer in line with how IPC objects are handled.
524 * systemd.pc now exports a number of additional directories,
525 including $libdir (which is useful to identify the library
526 path for the primary architecture of the system), and a
527 couple of drop-in directories.
529 * udev's predictable network interface names now use the dev_port
530 sysfs attribute, introduced in linux 3.15 instead of dev_id to
531 distinguish between ports of the same PCI function. dev_id should
532 only be used for ports using the same HW address, hence the need
535 * machined has been updated to export the OS version of a
536 container (read from /etc/os-release and
537 /usr/lib/os-release) on the bus. This is now shown in
538 "machinectl status" for a machine.
540 * A new service setting RestartForceExitStatus= has been
541 added. If configured to a set of exit signals or process
542 return values, the service will be restarted when the main
543 daemon process exits with any of them, regardless of the
546 * systemctl's -H switch for connecting to remote systemd
547 machines has been extended so that it may be used to
548 directly connect to a specific container on the
549 host. "systemctl -H root@foobar:waldi" will now connect as
550 user "root" to host "foobar", and then proceed directly to
551 the container named "waldi". Note that currently you have to
552 authenticate as user "root" for this to work, as entering
553 containers is a privileged operation.
555 Contributions from: Andreas Henriksson, Benjamin Steinwender,
556 Carl Schaefer, Christian Hesse, Colin Ian King, Cristian
557 Rodríguez, Daniel Mack, Dave Reisner, David Herrmann, Eugene
558 Yakubovich, Filipe Brandenburger, Frederic Crozat, Hristo
559 Venev, Jan Engelhardt, Jonathan Boulle, Kay Sievers, Lennart
560 Poettering, Luke Shumaker, Mantas Mikulėnas, Marc-Antoine
561 Perennou, Marcel Holtmann, Michael Marineau, Michael Olbrich,
562 Michał Bartoszkiewicz, Michal Sekletar, Patrik Flykt, Ronan Le
563 Martret, Ronny Chevalier, Ruediger Oertel, Steven Noonan,
564 Susant Sahani, Thadeu Lima de Souza Cascardo, Thomas Hindoe
565 Paaboel Andersen, Tom Gundersen, Tom Hirst, Umut Tezduyar
566 Lindskog, Uoti Urpala, Zbigniew Jędrzejewski-Szmek
568 -- Berlin, 2014-07-03
572 * As an experimental feature, udev now tries to lock the
573 disk device node (flock(LOCK_SH|LOCK_NB)) while it
574 executes events for the disk or any of its partitions.
575 Applications like partitioning programs can lock the
576 disk device node (flock(LOCK_EX)) and claim temporary
577 device ownership that way; udev will entirely skip all event
578 handling for this disk and its partitions. If the disk
579 was opened for writing, the close will trigger a partition
580 table rescan in udev's "watch" facility, and if needed
581 synthesize "change" events for the disk and all its partitions.
582 This is now unconditionally enabled, and if it turns out to
583 cause major problems, we might turn it on only for specific
584 devices, or might need to disable it entirely. Device Mapper
585 devices are excluded from this logic.
587 * We temporarily dropped the "-l" switch for fsck invocations,
588 since they collide with the flock() logic above. util-linux
589 upstream has been changed already to avoid this conflict,
590 and we will readd "-l" as soon as util-linux with this
591 change has been released.
593 * The dependency on libattr has been removed. Since a long
594 time, the extended attribute calls have moved to glibc, and
595 libattr is thus unnecessary.
597 * Virtualization detection works without priviliges now. This
598 means the systemd-detect-virt binary no longer requires
599 CAP_SYS_PTRACE file capabilities, and our daemons can run
600 with fewer privileges.
602 * systemd-networkd now runs under its own "systemd-network"
603 user. It retains the CAP_NET_ADMIN, CAP_NET_BIND_SERVICE,
604 CAP_NET_BROADCAST, CAP_NET_RAW capabilities though, but
605 loses the ability to write to files owned by root this way.
607 * Similar, systemd-resolved now runs under its own
608 "systemd-resolve" user with no capabilities remaining.
610 * Similar, systemd-bus-proxyd now runs under its own
611 "systemd-bus-proxy" user with only CAP_IPC_OWNER remaining.
613 * systemd-networkd gained support for setting up "veth"
614 virtual ethernet devices for container connectivity, as well
615 as GRE and VTI tunnels.
617 * systemd-networkd will no longer automatically attempt to
618 manually load kernel modules necessary for certain tunnel
619 transports. Instead, it is assumed the kernel loads them
620 automatically when required. This only works correctly on
621 very new kernels. On older kernels, please consider adding
622 the kernel modules to /etc/modules-load.d/ as a work-around.
624 * The resolv.conf file systemd-resolved generates has been
625 moved to /run/systemd/resolve/. If you have a symlink from
626 /etc/resolv.conf, it might be necessary to correct it.
628 * Two new service settings, ProtectHome= and ProtectSystem=,
629 have been added. When enabled, they will make the user data
630 (such as /home) inaccessible or read-only and the system
631 (such as /usr) read-only, for specific services. This allows
632 very light-weight per-service sandboxing to avoid
633 modifications of user data or system files from
634 services. These two new switches have been enabled for all
635 of systemd's long-running services, where appropriate.
637 * Socket units gained new SocketUser= and SocketGroup=
638 settings to set the owner user and group of AF_UNIX sockets
639 and FIFOs in the file system.
641 * Socket units gained a new RemoveOnStop= setting. If enabled,
642 all FIFOS and sockets in the file system will be removed
643 when the specific socket unit is stopped.
645 * Socket units gained a new Symlinks= setting. It takes a list
646 of symlinks to create to file system sockets or FIFOs
647 created by the specific Unix sockets. This is useful to
648 manage symlinks to socket nodes with the same life-cycle as
651 * The /dev/log socket and /dev/initctl FIFO have been moved to
652 /run, and have been replaced by symlinks. This allows
653 connecting to these facilities even if PrivateDevices=yes is
654 used for a service (which makes /dev/log itself unavailable,
655 but /run is left). This also has the benefit of ensuring
656 that /dev only contains device nodes, directories and
657 symlinks, and nothing else.
659 * sd-daemon gained two new calls sd_pid_notify() and
660 sd_pid_notifyf(). They are similar to sd_notify() and
661 sd_notifyf(), but allow overriding of the source PID of
662 notification messages if permissions permit this. This is
663 useful to send notify messages on behalf of a different
664 process (for example, the parent process). The
665 systemd-notify tool has been updated to make use of this
666 when sending messages (so that notification messages now
667 originate from the shell script invoking systemd-notify and
668 not the systemd-notify process itself. This should minimize
669 a race where systemd fails to associate notification
670 messages to services when the originating process already
673 * A new "on-abnormal" setting for Restart= has been added. If
674 set, it will result in automatic restarts on all "abnormal"
675 reasons for a process to exit, which includes unclean
676 signals, core dumps, timeouts and watchdog timeouts, but
677 does not include clean and unclean exit codes or clean
678 signals. Restart=on-abnormal is an alternative for
679 Restart=on-failure for services that shall be able to
680 terminate and avoid restarts on certain errors, by
681 indicating so with an unclean exit code. Restart=on-failure
682 or Restart=on-abnormal is now the recommended setting for
683 all long-running services.
685 * If the InaccessibleDirectories= service setting points to a
686 mount point (or if there are any submounts contained within
687 it), it is now attempted to completely unmount it, to make
688 the file systems truly unavailable for the respective
691 * The ReadOnlyDirectories= service setting and
692 systemd-nspawn's --read-only parameter are now recursively
693 applied to all submounts, too.
695 * Mount units may now be created transiently via the bus APIs.
697 * The support for SysV and LSB init scripts has been removed
698 from the systemd daemon itself. Instead, it is now
699 implemented as a generator that creates native systemd units
700 from these scripts when needed. This enables us to remove a
701 substantial amount of legacy code from PID 1, following the
702 fact that many distributions only ship a very small number
703 of LSB/SysV init scripts nowadays.
705 * Priviliged Xen (dom0) domains are not considered
706 virtualization anymore by the virtualization detection
707 logic. After all, they generally have unrestricted access to
708 the hardware and usually are used to manage the unprivileged
711 * systemd-tmpfiles gained a new "C" line type, for copying
712 files or entire directories.
714 * systemd-tmpfiles "m" lines are now fully equivalent to "z"
715 lines. So far, they have been non-globbing versions of the
716 latter, and have thus been redundant. In future, it is
717 recommended to only use "z". "m" has hence been removed
718 from the documentation, even though it stays supported.
720 * A tmpfiles snippet to recreate the most basic structure in
721 /var has been added. This is enough to create the /var/run →
722 /run symlink and create a couple of structural
723 directories. This allows systems to boot up with an empty or
724 volatile /var. Of course, while with this change, the core OS
725 now is capable with dealing with a volatile /var, not all
726 user services are ready for it. However, we hope that sooner
727 or later, many service daemons will be changed upstream so
728 that they are able to automatically create their necessary
729 directories in /var at boot, should they be missing. This is
730 the first step to allow state-less systems that only require
731 the vendor image for /usr to boot.
733 * systemd-nspawn has gained a new --tmpfs= switch to mount an
734 empty tmpfs instance to a specific directory. This is
735 particularly useful for making use of the automatic
736 reconstruction of /var (see above), by passing --tmpfs=/var.
738 * Access modes specified in tmpfiles snippets may now be
739 prefixed with "~", which indicates that they shall be masked
740 by whether the existing file or directory is currently
741 writable, readable or executable at all. Also, if specified,
742 the sgid/suid/sticky bits will be masked for all
745 * A new passive target unit "network-pre.target" has been
746 added which is useful for services that shall run before any
747 network is configured, for example firewall scripts.
749 * The "floppy" group that previously owned the /dev/fd*
750 devices is no longer used. The "disk" group is now used
751 instead. Distributions should probably deprecate usage of
754 Contributions from: Camilo Aguilar, Christian Hesse, Colin Ian
755 King, Cristian Rodríguez, Daniel Buch, Dave Reisner, David
756 Strauss, Denis Tikhomirov, John, Jonathan Liu, Kay Sievers,
757 Lennart Poettering, Mantas Mikulėnas, Mark Eichin, Ronny
758 Chevalier, Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel
759 Andersen, Tom Gundersen, Umut Tezduyar Lindskog, Zbigniew
762 -- Berlin, 2014-06-11
766 * A new "systemd-timesyncd" daemon has been added for
767 synchronizing the system clock across the network. It
768 implements an SNTP client. In contrast to NTP
769 implementations such as chrony or the NTP reference server,
770 this only implements a client side, and does not bother with
771 the full NTP complexity, focusing only on querying time from
772 one remote server and synchronizing the local clock to
773 it. Unless you intend to serve NTP to networked clients or
774 want to connect to local hardware clocks, this simple NTP
775 client should be more than appropriate for most
776 installations. The daemon runs with minimal privileges, and
777 has been hooked up with networkd to only operate when
778 network connectivity is available. The daemon saves the
779 current clock to disk every time a new NTP sync has been
780 acquired, and uses this to possibly correct the system clock
781 early at bootup, in order to accommodate for systems that
782 lack an RTC such as the Raspberry Pi and embedded devices,
783 and to make sure that time monotonically progresses on these
784 systems, even if it is not always correct. To make use of
785 this daemon, a new system user and group "systemd-timesync"
786 needs to be created on installation of systemd.
788 * The queue "seqnum" interface of libudev has been disabled, as
789 it was generally incompatible with device namespacing as
790 sequence numbers of devices go "missing" if the devices are
791 part of a different namespace.
793 * "systemctl list-timers" and "systemctl list-sockets" gained
794 a --recursive switch for showing units of these types also
795 for all local containers, similar in style to the already
796 supported --recursive switch for "systemctl list-units".
798 * A new RebootArgument= setting has been added for service
799 units, which may be used to specify a kernel reboot argument
800 to use when triggering reboots with StartLimitAction=.
802 * A new FailureAction= setting has been added for service
803 units which may be used to specify an operation to trigger
804 when a service fails. This works similarly to
805 StartLimitAction=, but unlike it, controls what is done
806 immediately rather than only after several attempts to
807 restart the service in question.
809 * hostnamed got updated to also expose the kernel name,
810 release, and version on the bus. This is useful for
811 executing commands like hostnamectl with the -H switch.
812 systemd-analyze makes use of this to properly display
813 details when running non-locally.
815 * The bootchart tool can now show cgroup information in the
818 * The CFS CPU quota cgroup attribute is now exposed for
819 services. The new CPUQuota= switch has been added for this
820 which takes a percentage value. Setting this will have the
821 result that a service may never get more CPU time than the
822 specified percentage, even if the machine is otherwise idle.
824 * systemd-networkd learned IPIP and SIT tunnel support.
826 * LSB init scripts exposing a dependency on $network will now
827 get a dependency on network-online.target rather than simply
828 network.target. This should bring LSB handling closer to
829 what it was on SysV systems.
831 * A new fsck.repair= kernel option has been added to control
832 how fsck shall deal with unclean file systems at boot.
834 * The (.ini) configuration file parser will now silently
835 ignore sections whose name begins with "X-". This may be
836 used to maintain application-specific extension sections in unit
839 * machined gained a new API to query the IP addresses of
840 registered containers. "machinectl status" has been updated
841 to show these addresses in its output.
843 * A new call sd_uid_get_display() has been added to the
844 sd-login APIs for querying the "primary" session of a
845 user. The "primary" session of the user is elected from the
846 user's sessions and generally a graphical session is
847 preferred over a text one.
849 * A minimal systemd-resolved daemon has been added. It
850 currently simply acts as a companion to systemd-networkd and
851 manages resolv.conf based on per-interface DNS
852 configuration, possibly supplied via DHCP. In the long run
853 we hope to extend this into a local DNSSEC enabled DNS and
856 * The systemd-networkd-wait-online tool is now enabled by
857 default. It will delay network-online.target until a network
858 connection has been configured. The tool primarily integrates
859 with networkd, but will also make a best effort to make sense
860 of network configuration performed in some other way.
862 * Two new service options StartupCPUShares= and
863 StartupBlockIOWeight= have been added that work similarly to
864 CPUShares= and BlockIOWeight= however only apply during
865 system startup. This is useful to prioritize certain services
866 differently during bootup than during normal runtime.
868 * hostnamed has been changed to prefer the statically
869 configured hostname in /etc/hostname (unless set to
870 'localhost' or empty) over any dynamic one supplied by
871 dhcp. With this change, the rules for picking the hostname
872 match more closely the rules of other configuration settings
873 where the local administrator's configuration in /etc always
874 overrides any other settings.
876 Contributions fron: Ali H. Caliskan, Alison Chaiken, Bas van
877 den Berg, Brandon Philips, Cristian Rodríguez, Daniel Buch,
878 Dan Kilman, Dave Reisner, David Härdeman, David Herrmann,
879 David Strauss, Dimitris Spingos, Djalal Harouni, Eelco
880 Dolstra, Evan Nemerson, Florian Albrechtskirchinger, Greg
881 Kroah-Hartman, Harald Hoyer, Holger Hans Peter Freyther, Jan
882 Engelhardt, Jani Nikula, Jason St. John, Jeffrey Clark,
883 Jonathan Boulle, Kay Sievers, Lennart Poettering, Lukas
884 Nykryn, Lukasz Skalski, Łukasz Stelmach, Mantas Mikulėnas,
885 Marcel Holtmann, Martin Pitt, Matthew Monaco, Michael
886 Marineau, Michael Olbrich, Michal Sekletar, Mike Gilbert, Nis
887 Martensen, Patrik Flykt, Philip Lorenz, poma, Ray Strode,
888 Reyad Attiyat, Robert Milasan, Scott Thrasher, Stef Walter,
889 Steven Siloti, Susant Sahani, Tanu Kaskinen, Thomas Bächler,
890 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar
891 Lindskog, WaLyong Cho, Will Woods, Zbigniew
894 -- Beijing, 2014-05-28
898 * When restoring the screen brightness at boot, stay away from
899 the darkest setting or from the lowest 5% of the available
900 range, depending on which is the larger value of both. This
901 should effectively protect the user from rebooting into a
902 black screen, should the brightness have been set to minimum
905 * sd-login gained a new sd_machine_get_class() call to
906 determine the class ("vm" or "container") of a machine
907 registered with machined.
909 * sd-login gained new calls
910 sd_peer_get_{session,owner_uid,unit,user_unit,slice,machine_name}(),
911 to query the identity of the peer of a local AF_UNIX
912 connection. They operate similarly to their sd_pid_get_xyz()
915 * PID 1 will now maintain a system-wide system state engine
916 with the states "starting", "running", "degraded",
917 "maintenance", "stopping". These states are bound to system
918 startup, normal runtime, runtime with at least one failed
919 service, rescue/emergency mode and system shutdown. This
920 state is shown in the "systemctl status" output when no unit
921 name is passed. It is useful to determine system state, in
922 particularly when doing so for many systems or containers at
925 * A new command "list-machines" has been added to "systemctl"
926 that lists all local OS containers and shows their system
927 state (see above), if systemd runs inside of them.
929 * systemctl gained a new "-r" switch to recursively enumerate
930 units on all local containers, when used with the
931 "list-unit" command (which is the default one that is
932 executed when no parameters are specified).
934 * The GPT automatic partition discovery logic will now honour
935 two GPT partition flags: one may be set on a partition to
936 cause it to be mounted read-only, and the other may be set
937 on a partition to ignore it during automatic discovery.
939 * Two new GPT type UUIDs have been added for automatic root
940 partition discovery, for 32-bit and 64-bit ARM. This is not
941 particularly useful for discovering the root directory on
942 these architectures during bare-metal boots (since UEFI is
943 not common there), but still very useful to allow booting of
944 ARM disk images in nspawn with the -i option.
946 * MAC addresses of interfaces created with nspawn's
947 --network-interface= switch will now be generated from the
948 machine name, and thus be stable between multiple invocations
951 * logind will now automatically remove all IPC objects owned
952 by a user if she or he fully logs out. This makes sure that
953 users who are logged out cannot continue to consume IPC
954 resources. This covers SysV memory, semaphores and message
955 queues as well as POSIX shared memory and message
956 queues. Traditionally, SysV and POSIX IPC had no life-cycle
957 limits. With this functionality, that is corrected. This may
958 be turned off by using the RemoveIPC= switch of logind.conf.
960 * The systemd-machine-id-setup and tmpfiles tools gained a
961 --root= switch to operate on a specific root directory,
964 * journald can now forward logged messages to the TTYs of all
965 logged in users ("wall"). This is the default for all
966 emergency messages now.
968 * A new tool systemd-journal-remote has been added to stream
969 journal log messages across the network.
971 * /sys/fs/cgroup/ is now mounted read-only after all cgroup
972 controller trees are mounted into it. Note that the
973 directories mounted beneath it are not read-only. This is a
974 security measure and is particularly useful because glibc
975 actually includes a search logic to pick any tmpfs it can
976 find to implement shm_open() if /dev/shm is not available
977 (which it might very well be in namespaced setups).
979 * machinectl gained a new "poweroff" command to cleanly power
980 down a local OS container.
982 * The PrivateDevices= unit file setting will now also drop the
983 CAP_MKNOD capability from the capability bound set, and
984 imply DevicePolicy=closed.
986 * PrivateDevices=, PrivateNetwork= and PrivateTmp= is now used
987 comprehensively on all long-running systemd services where
990 * systemd-udevd will now run in a disassociated mount
991 namespace. To mount directories from udev rules, make sure to
992 pull in mount units via SYSTEMD_WANTS properties.
994 * The kdbus support gained support for uploading policy into
995 the kernel. sd-bus gained support for creating "monitoring"
996 connections that can eavesdrop into all bus communication
997 for debugging purposes.
999 * Timestamps may now be specified in seconds since the UNIX
1000 epoch Jan 1st, 1970 by specifying "@" followed by the value
1003 * Native tcpwrap support in systemd has been removed. tcpwrap
1004 is old code, not really maintained anymore and has serious
1005 shortcomings, and better options such as firewalls
1006 exist. For setups that require tcpwrap usage, please
1007 consider invoking your socket-activated service via tcpd,
1008 like on traditional inetd.
1010 * A new system.conf configuration option
1011 DefaultTimerAccuracySec= has been added that controls the
1012 default AccuracySec= setting of .timer units.
1014 * Timer units gained a new WakeSystem= switch. If enabled,
1015 timers configured this way will cause the system to resume
1016 from system suspend (if the system supports that, which most
1019 * Timer units gained a new Persistent= switch. If enabled,
1020 timers configured this way will save to disk when they have
1021 been last triggered. This information is then used on next
1022 reboot to possible execute overdue timer events, that
1023 could not take place because the system was powered off.
1024 This enables simple anacron-like behaviour for timer units.
1026 * systemctl's "list-timers" will now also list the time a
1027 timer unit was last triggered in addition to the next time
1028 it will be triggered.
1030 * systemd-networkd will now assign predictable IPv4LL
1031 addresses to its local interfaces.
1033 Contributions from: Brandon Philips, Daniel Buch, Daniel Mack,
1034 Dave Reisner, David Herrmann, Gerd Hoffmann, Greg
1035 Kroah-Hartman, Hendrik Brueckner, Jason St. John, Josh
1036 Triplett, Kay Sievers, Lennart Poettering, Marc-Antoine
1037 Perennou, Michael Marineau, Michael Olbrich, Miklos Vajna,
1038 Patrik Flykt, poma, Sebastian Thorarensen, Thomas Bächler,
1039 Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom Gundersen,
1040 Umut Tezduyar Lindskog, Wieland Hoffmann, Zbigniew
1043 -- Berlin, 2014-03-25
1047 * A new unit file setting RestrictAddressFamilies= has been
1048 added to restrict which socket address families unit
1049 processes gain access to. This takes address family names
1050 like "AF_INET" or "AF_UNIX", and is useful to minimize the
1051 attack surface of services via exotic protocol stacks. This
1052 is built on seccomp system call filters.
1054 * Two new unit file settings RuntimeDirectory= and
1055 RuntimeDirectoryMode= have been added that may be used to
1056 manage a per-daemon runtime directories below /run. This is
1057 an alternative for setting up directory permissions with
1058 tmpfiles snippets, and has the advantage that the runtime
1059 directory's lifetime is bound to the daemon runtime and that
1060 the daemon starts up with an empty directory each time. This
1061 is particularly useful when writing services that drop
1062 privileges using the User= or Group= setting.
1064 * The DeviceAllow= unit setting now supports globbing for
1065 matching against device group names.
1067 * The systemd configuration file system.conf gained new
1068 settings DefaultCPUAccounting=, DefaultBlockIOAccounting=,
1069 DefaultMemoryAccounting= to globally turn on/off accounting
1070 for specific resources (cgroups) for all units. These
1071 settings may still be overridden individually in each unit
1074 * systemd-gpt-auto-generator is now able to discover /srv and
1075 root partitions in addition to /home and swap partitions. It
1076 also supports LUKS-encrypted partitions now. With this in
1077 place, automatic discovery of partitions to mount following
1078 the Discoverable Partitions Specification
1079 (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec)
1080 is now a lot more complete. This allows booting without
1081 /etc/fstab and without root= on the kernel command line on
1082 systems prepared appropriately.
1084 * systemd-nspawn gained a new --image= switch which allows
1085 booting up disk images and Linux installations on any block
1086 device that follow the Discoverable Partitions Specification
1087 (see above). This means that installations made with
1088 appropriately updated installers may now be started and
1089 deployed using container managers, completely
1090 unmodified. (We hope that libvirt-lxc will add support for
1091 this feature soon, too.)
1093 * systemd-nspawn gained a new --network-macvlan= setting to
1094 set up a private macvlan interface for the
1095 container. Similarly, systemd-networkd gained a new
1096 Kind=macvlan setting in .netdev files.
1098 * systemd-networkd now supports configuring local addresses
1101 * A new tool systemd-network-wait-online has been added to
1102 synchronously wait for network connectivity using
1105 * The sd-bus.h bus API gained a new sd_bus_track object for
1106 tracking the life-cycle of bus peers. Note that sd-bus.h is
1107 still not a public API though (unless you specify
1108 --enable-kdbus on the configure command line, which however
1109 voids your warranty and you get no API stability guarantee).
1111 * The $XDG_RUNTIME_DIR runtime directories for each user are
1112 now individual tmpfs instances, which has the benefit of
1113 introducing separate pools for each user, with individual
1114 size limits, and thus making sure that unprivileged clients
1115 can no longer negatively impact the system or other users by
1116 filling up their $XDG_RUNTIME_DIR. A new logind.conf setting
1117 RuntimeDirectorySize= has been introduced that allows
1118 controlling the default size limit for all users. It
1119 defaults to 10% of the available physical memory. This is no
1120 replacement for quotas on tmpfs though (which the kernel
1121 still does not support), as /dev/shm and /tmp are still
1122 shared resources used by both the system and unprivileged
1125 * logind will now automatically turn off automatic suspending
1126 on laptop lid close when more than one display is
1127 connected. This was previously expected to be implemented
1128 individually in desktop environments (such as GNOME),
1129 however has been added to logind now, in order to fix a
1130 boot-time race where a desktop environment might not have
1131 been started yet and thus not been able to take an inhibitor
1132 lock at the time where logind already suspends the system
1133 due to a closed lid.
1135 * logind will now wait at least 30s after each system
1136 suspend/resume cycle, and 3min after system boot before
1137 suspending the system due to a closed laptop lid. This
1138 should give USB docking stations and similar enough time to
1139 be probed and configured after system resume and boot in
1140 order to then act as suspend blocker.
1142 * systemd-run gained a new --property= setting which allows
1143 initialization of resource control properties (and others)
1144 for the created scope or service unit. Example: "systemd-run
1145 --property=BlockIOWeight=10 updatedb" may be used to run
1146 updatedb at a low block IO scheduling weight.
1148 * systemd-run's --uid=, --gid=, --setenv=, --setenv= switches
1149 now also work in --scope mode.
1151 * When systemd is compiled with kdbus support, basic support
1152 for enforced policies is now in place. (Note that enabling
1153 kdbus still voids your warranty and no API compatibility
1156 Contributions from: Andrey Borzenkov, Ansgar Burchardt, Armin
1157 K., Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
1158 Harald Hoyer, Henrik Grindal Bakken, Jasper St. Pierre, Kay
1159 Sievers, Kieran Clancy, Lennart Poettering, Lukas Nykryn,
1160 Mantas Mikulėnas, Marcel Holtmann, Mark Oteiza, Martin Pitt,
1161 Mike Gilbert, Peter Rajnoha, poma, Samuli Suominen, Stef
1162 Walter, Susant Sahani, Tero Roponen, Thomas Andersen, Thomas
1163 Bächler, Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom
1164 Gundersen, Umut Tezduyar Lindskog, Uoti Urpala, Zachary Cook,
1165 Zbigniew Jędrzejewski-Szmek
1167 -- Berlin, 2014-03-12
1171 * systemd will now relabel /dev after loading the SMACK policy
1172 according to SMACK rules.
1174 * A new unit file option AppArmorProfile= has been added to
1175 set the AppArmor profile for the processes of a unit.
1177 * A new condition check ConditionArchitecture= has been added
1178 to conditionalize units based on the system architecture, as
1179 reported by uname()'s "machine" field.
1181 * systemd-networkd now supports matching on the system
1182 virtualization, architecture, kernel command line, host name
1185 * logind is now a lot more aggressive when suspending the
1186 machine due to a closed laptop lid. Instead of acting only
1187 on the lid close action, it will continuously watch the lid
1188 status and act on it. This is useful for laptops where the
1189 power button is on the outside of the chassis so that it can
1190 be reached without opening the lid (such as the Lenovo
1191 Yoga). On those machines, logind will now immediately
1192 re-suspend the machine if the power button has been
1193 accidentally pressed while the laptop was suspended and in a
1194 backpack or similar.
1196 * logind will now watch SW_DOCK switches and inhibit reaction
1197 to the lid switch if it is pressed. This means that logind
1198 will not suspend the machine anymore if the lid is closed
1199 and the system is docked, if the laptop supports SW_DOCK
1200 notifications via the input layer. Note that ACPI docking
1201 stations do not generate this currently. Also note that this
1202 logic is usually not fully sufficient and Desktop
1203 Environments should take a lid switch inhibitor lock when an
1204 external display is connected, as systemd will not watch
1207 * nspawn will now make use of the devices cgroup controller by
1208 default, and only permit creation of and access to the usual
1209 API device nodes like /dev/null or /dev/random, as well as
1210 access to (but not creation of) the pty devices.
1212 * We will now ship a default .network file for
1213 systemd-networkd that automatically configures DHCP for
1214 network interfaces created by nspawn's --network-veth or
1215 --network-bridge= switches.
1217 * systemd will now understand the usual M, K, G, T suffixes
1218 according to SI conventions (i.e. to the base 1000) when
1219 referring to throughput and hardware metrics. It will stay
1220 with IEC conventions (i.e. to the base 1024) for software
1221 metrics, according to what is customary according to
1222 Wikipedia. We explicitly document which base applies for
1223 each configuration option.
1225 * The DeviceAllow= setting in unit files now supports a syntax
1226 to whitelist an entire group of devices node majors at once,
1227 based on the /proc/devices listing. For example, with the
1228 string "char-pts", it is now possible to whitelist all
1229 current and future pseudo-TTYs at once.
1231 * sd-event learned a new "post" event source. Event sources of
1232 this type are triggered by the dispatching of any event
1233 source of a type that is not "post". This is useful for
1234 implementing clean-up and check event sources that are
1235 triggered by other work being done in the program.
1237 * systemd-networkd is no longer statically enabled, but uses
1238 the usual [Install] sections so that it can be
1239 enabled/disabled using systemctl. It still is enabled by
1242 * When creating a veth interface pair with systemd-nspawn, the
1243 host side will now be prefixed with "vb-" if
1244 --network-bridge= is used, and with "ve-" if --network-veth
1245 is used. This way, it is easy to distinguish these cases on
1246 the host, for example to apply different configuration to
1247 them with systemd-networkd.
1249 * The compatibility libraries for libsystemd-journal.so,
1250 libsystem-id128.so, libsystemd-login.so and
1251 libsystemd-daemon.so do not make use of IFUNC
1252 anymore. Instead, we now build libsystemd.so multiple times
1253 under these alternative names. This means that the footprint
1254 is drastically increased, but given that these are
1255 transitional compatibility libraries, this should not matter
1256 much. This change has been made necessary to support the ARM
1257 platform for these compatibility libraries, as the ARM
1258 toolchain is not really at the same level as the toolchain
1259 for other architectures like x86 and does not support
1260 IFUNC. Please make sure to use --enable-compat-libs only
1261 during a transitional period!
1263 Contributions from: Andreas Fuchs, Armin K., Colin Walters,
1264 Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
1265 Holger Schurig, Jason A. Donenfeld, Jason St. John, Jasper
1266 St. Pierre, Kay Sievers, Lennart Poettering, Łukasz Stelmach,
1267 Marcel Holtmann, Michael Scherer, Michal Sekletar, Mike
1268 Gilbert, Samuli Suominen, Thomas Bächler, Thomas Hindoe
1269 Paaboel Andersen, Tom Gundersen, Umut Tezduyar Lindskog,
1270 Zbigniew Jędrzejewski-Szmek
1272 -- Berlin, 2014-02-24
1276 * A new component "systemd-networkd" has been added that can
1277 be used to configure local network interfaces statically or
1278 via DHCP. It is capable of bringing up bridges, VLANs, and
1279 bonding. Currently, no hook-ups for interactive network
1280 configuration are provided. Use this for your initrd,
1281 container, embedded, or server setup if you need a simple,
1282 yet powerful, network configuration solution. This
1283 configuration subsystem is quite nifty, as it allows wildcard
1284 hotplug matching in interfaces. For example, with a single
1285 configuration snippet, you can configure that all Ethernet
1286 interfaces showing up are automatically added to a bridge,
1287 or similar. It supports link-sensing and more.
1289 * A new tool "systemd-socket-proxyd" has been added which can
1290 act as a bidirectional proxy for TCP sockets. This is
1291 useful for adding socket activation support to services that
1292 do not actually support socket activation, including virtual
1293 machines and the like.
1295 * Add a new tool to save/restore rfkill state on
1298 * Save/restore state of keyboard backlights in addition to
1299 display backlights on shutdown/boot.
1301 * udev learned a new SECLABEL{} construct to label device
1302 nodes with a specific security label when they appear. For
1303 now, only SECLABEL{selinux} is supported, but the syntax is
1304 prepared for additional security frameworks.
1306 * udev gained a new scheme to configure link-level attributes
1307 from files in /etc/systemd/network/*.link. These files can
1308 match against MAC address, device path, driver name and type,
1309 and will apply attributes like the naming policy, link speed,
1310 MTU, duplex settings, Wake-on-LAN settings, MAC address, MAC
1311 address assignment policy (randomized, ...).
1313 * The configuration of network interface naming rules for
1314 "permanent interface names" has changed: a new NamePolicy=
1315 setting in the [Link] section of .link files determines the
1316 priority of possible naming schemes (onboard, slot, mac,
1317 path). The default value of this setting is determined by
1318 /usr/lib/net/links/99-default.link. Old
1319 80-net-name-slot.rules udev configuration file has been
1320 removed, so local configuration overriding this file should
1321 be adapated to override 99-default.link instead.
1323 * When the User= switch is used in a unit file, also
1324 initialize $SHELL= based on the user database entry.
1326 * systemd no longer depends on libdbus. All communication is
1327 now done with sd-bus, systemd's low-level bus library
1330 * kdbus support has been added to PID 1 itself. When kdbus is
1331 enabled, this causes PID 1 to set up the system bus and
1332 enable support for a new ".busname" unit type that
1333 encapsulates bus name activation on kdbus. It works a little
1334 bit like ".socket" units, except for bus names. A new
1335 generator has been added that converts classic dbus1 service
1336 activation files automatically into native systemd .busname
1339 * sd-bus: add a light-weight vtable implementation that allows
1340 defining objects on the bus with a simple static const
1341 vtable array of its methods, signals and properties.
1343 * systemd will not generate or install static dbus
1344 introspection data anymore to /usr/share/dbus-1/interfaces,
1345 as the precise format of these files is unclear, and
1346 nothing makes use of it.
1348 * A proxy daemon is now provided to proxy clients connecting
1349 via classic D-Bus AF_UNIX sockets to kdbus, to provide full
1350 compatibility with classic D-Bus.
1352 * A bus driver implementation has been added that supports the
1353 classic D-Bus bus driver calls on kdbus, also for
1354 compatibility purposes.
1356 * A new API "sd-event.h" has been added that implements a
1357 minimal event loop API built around epoll. It provides a
1358 couple of features that direct epoll usage is lacking:
1359 prioritization of events, scales to large numbers of timer
1360 events, per-event timer slack (accuracy), system-wide
1361 coalescing of timer events, exit handlers, watchdog
1362 supervision support using systemd's sd_notify() API, child
1365 * A new API "sd-rntl.h" has been added that provides an API
1366 around the route netlink interface of the kernel, similar in
1367 style to "sd-bus.h".
1369 * A new API "sd-dhcp-client.h" has been added that provides a
1370 small DHCPv4 client-side implementation. This is used by
1373 * There is a new kernel command line option
1374 "systemd.restore_state=0|1". When set to "0", none of the
1375 systemd tools will restore saved runtime state to hardware
1376 devices. More specifically, the rfkill and backlight states
1379 * The FsckPassNo= compatibility option in mount/service units
1380 has been removed. The fstab generator will now add the
1381 necessary dependencies automatically, and does not require
1382 PID1's support for that anymore.
1384 * journalctl gained a new switch, --list-boots, that lists
1385 recent boots with their times and boot IDs.
1387 * The various tools like systemctl, loginctl, timedatectl,
1388 busctl, systemd-run, ... have gained a new switch "-M" to
1389 connect to a specific, local OS container (as direct
1390 connection, without requiring SSH). This works on any
1391 container that is registered with machined, such as those
1392 created by libvirt-lxc or nspawn.
1394 * systemd-run and systemd-analyze also gained support for "-H"
1395 to connect to remote hosts via SSH. This is particularly
1396 useful for systemd-run because it enables queuing of jobs
1397 onto remote systems.
1399 * machinectl gained a new command "login" to open a getty
1400 login in any local container. This works with any container
1401 that is registered with machined (such as those created by
1402 libvirt-lxc or nspawn), and which runs systemd inside.
1404 * machinectl gained a new "reboot" command that may be used to
1405 trigger a reboot on a specific container that is registered
1406 with machined. This works on any container that runs an init
1407 system of some kind.
1409 * systemctl gained a new "list-timers" command to print a nice
1410 listing of installed timer units with the times they elapse
1413 * Alternative reboot() parameters may now be specified on the
1414 "systemctl reboot" command line and are passed to the
1415 reboot() system call.
1417 * systemctl gained a new --job-mode= switch to configure the
1418 mode to queue a job with. This is a more generic version of
1419 --fail, --irreversible, and --ignore-dependencies, which are
1420 still available but not advertised anymore.
1422 * /etc/systemd/system.conf gained new settings to configure
1423 various default timeouts of units, as well as the default
1424 start limit interval and burst. These may still be overridden
1427 * PID1 will now export on the bus profile data of the security
1428 policy upload process (such as the SELinux policy upload to
1431 * journald: when forwarding logs to the console, include
1432 timestamps (following the setting in
1433 /sys/module/printk/parameters/time).
1435 * OnCalendar= in timer units now understands the special
1436 strings "yearly" and "annually". (Both are equivalent)
1438 * The accuracy of timer units is now configurable with the new
1439 AccuracySec= setting. It defaults to 1min.
1441 * A new dependency type JoinsNamespaceOf= has been added that
1442 allows running two services within the same /tmp and network
1443 namespace, if PrivateNetwork= or PrivateTmp= are used.
1445 * A new command "cat" has been added to systemctl. It outputs
1446 the original unit file of a unit, and concatenates the
1447 contents of additional "drop-in" unit file snippets, so that
1448 the full configuration is shown.
1450 * systemctl now supports globbing on the various "list-xyz"
1451 commands, like "list-units" or "list-sockets", as well as on
1452 those commands which take multiple unit names.
1454 * journalctl's --unit= switch gained support for globbing.
1456 * All systemd daemons now make use of the watchdog logic so
1457 that systemd automatically notices when they hang.
1459 * If the $container_ttys environment variable is set,
1460 getty-generator will automatically spawn a getty for each
1461 listed tty. This is useful for container managers to request
1462 login gettys to be spawned on as many ttys as needed.
1464 * %h, %s, %U specifier support is not available anymore when
1465 used in unit files for PID 1. This is because NSS calls are
1466 not safe from PID 1. They stay available for --user
1467 instances of systemd, and as special case for the root user.
1469 * loginctl gained a new "--no-legend" switch to turn off output
1472 * The "sd-login.h" API gained three new calls:
1473 sd_session_is_remote(), sd_session_get_remote_user(),
1474 sd_session_get_remote_host() to query information about
1477 * The udev hardware database now also carries vendor/product
1478 information of SDIO devices.
1480 * The "sd-daemon.h" API gained a new sd_watchdog_enabled() to
1481 determine whether watchdog notifications are requested by
1484 * Socket-activated per-connection services now include a
1485 short description of the connection parameters in the
1488 * tmpfiles gained a new "--boot" option. When this is not used,
1489 only lines where the command character is not suffixed with
1490 "!" are executed. When this option is specified, those
1491 options are executed too. This partitions tmpfiles
1492 directives into those that can be safely executed at any
1493 time, and those which should be run only at boot (for
1494 example, a line that creates /run/nologin).
1496 * A new API "sd-resolve.h" has been added which provides a simple
1497 asynchronous wrapper around glibc NSS host name resolution
1498 calls, such as getaddrinfo(). In contrast to glibc's
1499 getaddrinfo_a(), it does not use signals. In contrast to most
1500 other asynchronous name resolution libraries, this one does
1501 not reimplement DNS, but reuses NSS, so that alternate
1502 host name resolution systems continue to work, such as mDNS,
1503 LDAP, etc. This API is based on libasyncns, but it has been
1504 cleaned up for inclusion in systemd.
1506 * The APIs "sd-journal.h", "sd-login.h", "sd-id128.h",
1507 "sd-daemon.h" are no longer found in individual libraries
1508 libsystemd-journal.so, libsystemd-login.so,
1509 libsystemd-id128.so, libsystemd-daemon.so. Instead, we have
1510 merged them into a single library, libsystemd.so, which
1511 provides all symbols. The reason for this is cyclic
1512 dependencies, as these libraries tend to use each other's
1513 symbols. So far, we have managed to workaround that by linking
1514 a copy of a good part of our code into each of these
1515 libraries again and again, which, however, makes certain
1516 things hard to do, like sharing static variables. Also, it
1517 substantially increases footprint. With this change, there
1518 is only one library for the basic APIs systemd
1519 provides. Also, "sd-bus.h", "sd-memfd.h", "sd-event.h",
1520 "sd-rtnl.h", "sd-resolve.h", "sd-utf8.h" are found in this
1521 library as well, however are subject to the --enable-kdbus
1522 switch (see below). Note that "sd-dhcp-client.h" is not part
1523 of this library (this is because it only consumes, never
1524 provides, services of/to other APIs). To make the transition
1525 easy from the separate libraries to the unified one, we
1526 provide the --enable-compat-libs compile-time switch which
1527 will generate stub libraries that are compatible with the
1528 old ones but redirect all calls to the new one.
1530 * All of the kdbus logic and the new APIs "sd-bus.h",
1531 "sd-memfd.h", "sd-event.h", "sd-rtnl.h", "sd-resolve.h",
1532 and "sd-utf8.h" are compile-time optional via the
1533 "--enable-kdbus" switch, and they are not compiled in by
1534 default. To make use of kdbus, you have to explicitly enable
1535 the switch. Note however, that neither the kernel nor the
1536 userspace API for all of this is considered stable yet. We
1537 want to maintain the freedom to still change the APIs for
1538 now. By specifying this build-time switch, you acknowledge
1539 that you are aware of the instability of the current
1542 * Also, note that while kdbus is pretty much complete,
1543 it lacks one thing: proper policy support. This means you
1544 can build a fully working system with all features; however,
1545 it will be highly insecure. Policy support will be added in
1546 one of the next releases, at the same time that we will
1547 declare the APIs stable.
1549 * When the kernel command-line argument "kdbus" is specified,
1550 systemd will automatically load the kdbus.ko kernel module. At
1551 this stage of development, it is only useful for testing kdbus
1552 and should not be used in production. Note: if "--enable-kdbus"
1553 is specified, and the kdbus.ko kernel module is available, and
1554 "kdbus" is added to the kernel command line, the entire system
1555 runs with kdbus instead of dbus-daemon, with the above mentioned
1556 problem of missing the system policy enforcement. Also a future
1557 version of kdbus.ko or a newer systemd will not be compatible with
1558 each other, and will unlikely be able to boot the machine if only
1559 one of them is updated.
1561 * systemctl gained a new "import-environment" command which
1562 uploads the caller's environment (or parts thereof) into the
1563 service manager so that it is inherited by services started
1564 by the manager. This is useful to upload variables like
1565 $DISPLAY into the user service manager.
1567 * A new PrivateDevices= switch has been added to service units
1568 which allows running a service with a namespaced /dev
1569 directory that does not contain any device nodes for
1570 physical devices. More specifically, it only includes devices
1571 such as /dev/null, /dev/urandom, and /dev/zero which are API
1574 * logind has been extended to support behaviour like VT
1575 switching on seats that do not support a VT. This makes
1576 multi-session available on seats that are not the first seat
1577 (seat0), and on systems where kernel support for VTs has
1578 been disabled at compile-time.
1580 * If a process holds a delay lock for system sleep or shutdown
1581 and fails to release it in time, we will now log its
1582 identity. This makes it easier to identify processes that
1583 cause slow suspends or power-offs.
1585 * When parsing /etc/crypttab, support for a new key-slot=
1586 option as supported by Debian is added. It allows indicating
1587 which LUKS slot to use on disk, speeding up key loading.
1589 * The sd_journald_sendv() API call has been checked and
1590 officially declared to be async-signal-safe so that it may
1591 be invoked from signal handlers for logging purposes.
1593 * Boot-time status output is now enabled automatically after a
1594 short timeout if boot does not progress, in order to give
1595 the user an indication what she or he is waiting for.
1597 * The boot-time output has been improved to show how much time
1598 remains until jobs expire.
1600 * The KillMode= switch in service units gained a new possible
1601 value "mixed". If set, and the unit is shut down, then the
1602 initial SIGTERM signal is sent only to the main daemon
1603 process, while the following SIGKILL signal is sent to
1604 all remaining processes of the service.
1606 * When a scope unit is registered, a new property "Controller"
1607 may be set. If set to a valid bus name, systemd will send a
1608 RequestStop() signal to this name when it would like to shut
1609 down the scope. This may be used to hook manager logic into
1610 the shutdown logic of scope units. Also, scope units may now
1611 be put in a special "abandoned" state, in which case the
1612 manager process which created them takes no further
1613 responsibilities for it.
1615 * When reading unit files, systemd will now verify
1616 the access mode of these files, and warn about certain
1617 suspicious combinations. This has been added to make it
1618 easier to track down packaging bugs where unit files are
1619 marked executable or world-writable.
1621 * systemd-nspawn gained a new "--setenv=" switch to set
1622 container-wide environment variables. The similar option in
1623 systemd-activate was renamed from "--environment=" to
1624 "--setenv=" for consistency.
1626 * systemd-nspawn has been updated to create a new kdbus domain
1627 for each container that is invoked, thus allowing each
1628 container to have its own set of system and user buses,
1629 independent of the host.
1631 * systemd-nspawn gained a new --drop-capability= switch to run
1632 the container with less capabilities than the default. Both
1633 --drop-capability= and --capability= now take the special
1634 string "all" for dropping or keeping all capabilities.
1636 * systemd-nspawn gained new switches for executing containers
1637 with specific SELinux labels set.
1639 * systemd-nspawn gained a new --quiet switch to not generate
1640 any additional output but the container's own console
1643 * systemd-nspawn gained a new --share-system switch to run a
1644 container without PID namespacing enabled.
1646 * systemd-nspawn gained a new --register= switch to control
1647 whether the container is registered with systemd-machined or
1648 not. This is useful for containers that do not run full
1649 OS images, but only specific apps.
1651 * systemd-nspawn gained a new --keep-unit which may be used
1652 when invoked as the only program from a service unit, and
1653 results in registration of the unit service itself in
1654 systemd-machined, instead of a newly opened scope unit.
1656 * systemd-nspawn gained a new --network-interface= switch for
1657 moving arbitrary interfaces to the container. The new
1658 --network-veth switch creates a virtual Ethernet connection
1659 between host and container. The new --network-bridge=
1660 switch then allows assigning the host side of this virtual
1661 Ethernet connection to a bridge device.
1663 * systemd-nspawn gained a new --personality= switch for
1664 setting the kernel personality for the container. This is
1665 useful when running a 32-bit container on a 64-bit host. A
1666 similar option Personality= is now also available for service
1669 * logind will now also track a "Desktop" identifier for each
1670 session which encodes the desktop environment of it. This is
1671 useful for desktop environments that want to identify
1672 multiple running sessions of itself easily.
1674 * A new SELinuxContext= setting for service units has been
1675 added that allows setting a specific SELinux execution
1676 context for a service.
1678 * Most systemd client tools will now honour $SYSTEMD_LESS for
1679 settings of the "less" pager. By default, these tools will
1680 override $LESS to allow certain operations to work, such as
1681 jump-to-the-end. With $SYSTEMD_LESS, it is possible to
1682 influence this logic.
1684 * systemd's "seccomp" hook-up has been changed to make use of
1685 the libseccomp library instead of using its own
1686 implementation. This has benefits for portability among
1689 * For usage together with SystemCallFilter=, a new
1690 SystemCallErrorNumber= setting has been introduced that
1691 allows configuration of a system error number to be returned
1692 on filtered system calls, instead of immediately killing the
1693 process. Also, SystemCallArchitectures= has been added to
1694 limit access to system calls of a particular architecture
1695 (in order to turn off support for unused secondary
1696 architectures). There is also a global
1697 SystemCallArchitectures= setting in system.conf now to turn
1698 off support for non-native system calls system-wide.
1700 * systemd requires a kernel with a working name_to_handle_at(),
1701 please see the kernel config requirements in the README file.
1703 Contributions from: Adam Williamson, Alex Jia, Anatol Pomozov,
1704 Ansgar Burchardt, AppleBloom, Auke Kok, Bastien Nocera,
1705 Chengwei Yang, Christian Seiler, Colin Guthrie, Colin Walters,
1706 Cristian Rodríguez, Daniel Buch, Daniele Medri, Daniel J
1707 Walsh, Daniel Mack, Dan McGee, Dave Reisner, David Coppa,
1708 David Herrmann, David Strauss, Djalal Harouni, Dmitry Pisklov,
1709 Elia Pinto, Florian Weimer, George McCollister, Goffredo
1710 Baroncelli, Greg Kroah-Hartman, Hendrik Brueckner, Igor
1711 Zhbanov, Jan Engelhardt, Jan Janssen, Jason A. Donenfeld,
1712 Jason St. John, Jasper St. Pierre, Jóhann B. Guðmundsson, Jose
1713 Ignacio Naranjo, Karel Zak, Kay Sievers, Kristian Høgsberg,
1714 Lennart Poettering, Lubomir Rintel, Lukas Nykryn, Lukasz
1715 Skalski, Łukasz Stelmach, Luke Shumaker, Mantas Mikulėnas,
1716 Marc-Antoine Perennou, Marcel Holtmann, Marcos Felipe Rasia de
1717 Mello, Marko Myllynen, Martin Pitt, Matthew Monaco, Michael
1718 Marineau, Michael Scherer, Michał Górny, Michal Sekletar,
1719 Michele Curti, Oleksii Shevchuk, Olivier Brunel, Patrik Flykt,
1720 Pavel Holica, Raudi, Richard Marko, Ronny Chevalier, Sébastien
1721 Luttringer, Sergey Ptashnick, Shawn Landden, Simon Peeters,
1722 Stefan Beller, Susant Sahani, Sylvain Plantefeve, Sylvia Else,
1723 Tero Roponen, Thomas Bächler, Thomas Hindoe Paaboel Andersen,
1724 Tom Gundersen, Umut Tezduyar Lindskog, Unai Uribarri, Václav
1725 Pavlín, Vincent Batts, WaLyong Cho, William Giokas, Yang
1726 Zhiyong, Yin Kangkai, Yuxuan Shui, Zbigniew Jędrzejewski-Szmek
1728 -- Berlin, 2014-02-20
1732 * logind has gained support for facilitating privileged input
1733 and drm device access for unprivileged clients. This work is
1734 useful to allow Wayland display servers (and similar
1735 programs, such as kmscon) to run under the user's ID and
1736 access input and drm devices which are normally
1737 protected. When this is used (and the kernel is new enough)
1738 logind will "mute" IO on the file descriptors passed to
1739 Wayland as long as it is in the background and "unmute" it
1740 if it returns into the foreground. This allows secure
1741 session switching without allowing background sessions to
1742 eavesdrop on input and display data. This also introduces
1743 session switching support if VT support is turned off in the
1744 kernel, and on seats that are not seat0.
1746 * A new kernel command line option luks.options= is understood
1747 now which allows specifying LUKS options for usage for LUKS
1748 encrypted partitions specified with luks.uuid=.
1750 * tmpfiles.d(5) snippets may now use specifier expansion in
1751 path names. More specifically %m, %b, %H, %v, are now
1752 replaced by the local machine id, boot id, hostname, and
1753 kernel version number.
1755 * A new tmpfiles.d(5) command "m" has been introduced which
1756 may be used to change the owner/group/access mode of a file
1757 or directory if it exists, but do nothing if it does not.
1759 * This release removes high-level support for the
1760 MemorySoftLimit= cgroup setting. The underlying kernel
1761 cgroup attribute memory.soft_limit= is currently badly
1762 designed and likely to be removed from the kernel API in its
1763 current form, hence we should not expose it for now.
1765 * The memory.use_hierarchy cgroup attribute is now enabled for
1766 all cgroups systemd creates in the memory cgroup
1767 hierarchy. This option is likely to be come the built-in
1768 default in the kernel anyway, and the non-hierarchial mode
1769 never made much sense in the intrinsically hierarchial
1772 * A new field _SYSTEMD_SLICE= is logged along with all journal
1773 messages containing the slice a message was generated
1774 from. This is useful to allow easy per-customer filtering of
1775 logs among other things.
1777 * systemd-journald will no longer adjust the group of journal
1778 files it creates to the "systemd-journal" group. Instead we
1779 rely on the journal directory to be owned by the
1780 "systemd-journal" group, and its setgid bit set, so that the
1781 kernel file system layer will automatically enforce that
1782 journal files inherit this group assignment. The reason for
1783 this change is that we cannot allow NSS look-ups from
1784 journald which would be necessary to resolve
1785 "systemd-journal" to a numeric GID, because this might
1786 create deadlocks if NSS involves synchronous queries to
1787 other daemons (such as nscd, or sssd) which in turn are
1788 logging clients of journald and might block on it, which
1789 would then dead lock. A tmpfiles.d(5) snippet included in
1790 systemd will make sure the setgid bit and group are
1791 properly set on the journal directory if it exists on every
1792 boot. However, we recommend adjusting it manually after
1793 upgrades too (or from RPM scriptlets), so that the change is
1794 not delayed until next reboot.
1796 * Backlight and random seed files in /var/lib/ have moved into
1797 the /var/lib/systemd/ directory, in order to centralize all
1798 systemd generated files in one directory.
1800 * Boot time performance measurements (as displayed by
1801 "systemd-analyze" for example) will now read ACPI 5.0 FPDT
1802 performance information if that's available to determine how
1803 much time BIOS and boot loader initialization required. With
1804 a sufficiently new BIOS you hence no longer need to boot
1805 with Gummiboot to get access to such information.
1807 Contributions from: Andrey Borzenkov, Chen Jie, Colin Walters,
1808 Cristian Rodríguez, Dave Reisner, David Herrmann, David
1809 Mackey, David Strauss, Eelco Dolstra, Evan Callicoat, Gao
1810 feng, Harald Hoyer, Jimmie Tauriainen, Kay Sievers, Lennart
1811 Poettering, Lukas Nykryn, Mantas Mikulėnas, Martin Pitt,
1812 Michael Scherer, Michał Górny, Mike Gilbert, Patrick McCarty,
1813 Sebastian Ott, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
1815 -- Berlin, 2013-10-02
1819 * The Restart= option for services now understands a new
1820 on-watchdog setting, which will restart the service
1821 automatically if the service stops sending out watchdog keep
1822 alive messages (as configured with WatchdogSec=).
1824 * The getty generator (which is responsible for bringing up a
1825 getty on configured serial consoles) will no longer only
1826 start a getty on the primary kernel console but on all
1827 others, too. This makes the order in which console= is
1828 specified on the kernel command line less important.
1830 * libsystemd-logind gained a new sd_session_get_vt() call to
1831 retrieve the VT number of a session.
1833 * If the option "tries=0" is set for an entry of /etc/crypttab
1834 its passphrase is queried indefinitely instead of any
1835 maximum number of tries.
1837 * If a service with a configure PID file terminates its PID
1838 file will now be removed automatically if it still exists
1839 afterwards. This should put an end to stale PID files.
1841 * systemd-run will now also take relative binary path names
1842 for execution and no longer insists on absolute paths.
1844 * InaccessibleDirectories= and ReadOnlyDirectories= now take
1845 paths that are optionally prefixed with "-" to indicate that
1846 it should not be considered a failure if they do not exist.
1848 * journalctl -o (and similar commands) now understands a new
1849 output mode "short-precise", it is similar to "short" but
1850 shows timestamps with usec accuracy.
1852 * The option "discard" (as known from Debian) is now
1853 synonymous to "allow-discards" in /etc/crypttab. In fact,
1854 "discard" is preferred now (since it is easier to remember
1857 * Some licensing clean-ups were made, so that more code is now
1858 LGPL-2.1 licensed than before.
1860 * A minimal tool to save/restore the display backlight
1861 brightness across reboots has been added. It will store the
1862 backlight setting as late as possible at shutdown, and
1863 restore it as early as possible during reboot.
1865 * A logic to automatically discover and enable home and swap
1866 partitions on GPT disks has been added. With this in place
1867 /etc/fstab becomes optional for many setups as systemd can
1868 discover certain partitions located on the root disk
1869 automatically. Home partitions are recognized under their
1870 GPT type ID 933ac7e12eb44f13b8440e14e2aef915. Swap
1871 partitions are recognized under their GPT type ID
1872 0657fd6da4ab43c484e50933c84b4f4f.
1874 * systemd will no longer pass any environment from the kernel
1875 or initrd to system services. If you want to set an
1876 environment for all services, do so via the kernel command
1877 line systemd.setenv= assignment.
1879 * The systemd-sysctl tool no longer natively reads the file
1880 /etc/sysctl.conf. If desired, the file should be symlinked
1881 from /etc/sysctl.d/99-sysctl.conf. Apart from providing
1882 legacy support by a symlink rather than built-in code, it
1883 also makes the otherwise hidden order of application of the
1884 different files visible. (Note that this partly reverts to a
1885 pre-198 application order of sysctl knobs!)
1887 * The "systemctl set-log-level" and "systemctl dump" commands
1888 have been moved to systemd-analyze.
1890 * systemd-run learned the new --remain-after-exit switch,
1891 which causes the scope unit not to be cleaned up
1892 automatically after the process terminated.
1894 * tmpfiles learned a new --exclude-prefix= switch to exclude
1895 certain paths from operation.
1897 * journald will now automatically flush all messages to disk
1898 as soon as a message of the log priorities CRIT, ALERT or
1901 Contributions from: Andrew Cook, Brandon Philips, Christian
1902 Hesse, Christoph Junghans, Colin Walters, Daniel Schaal,
1903 Daniel Wallace, Dave Reisner, David Herrmann, Gao feng, George
1904 McCollister, Giovanni Campagna, Hannes Reinecke, Harald Hoyer,
1905 Herczeg Zsolt, Holger Hans Peter Freyther, Jan Engelhardt,
1906 Jesper Larsen, Kay Sievers, Khem Raj, Lennart Poettering,
1907 Lukas Nykryn, Maciej Wereski, Mantas Mikulėnas, Marcel
1908 Holtmann, Martin Pitt, Michael Biebl, Michael Marineau,
1909 Michael Scherer, Michael Stapelberg, Michal Sekletar, Michał
1910 Górny, Olivier Brunel, Ondrej Balaz, Ronny Chevalier, Shawn
1911 Landden, Steven Hiscocks, Thomas Bächler, Thomas Hindoe
1912 Paaboel Andersen, Tom Gundersen, Umut Tezduyar, WANG Chao,
1913 William Giokas, Zbigniew Jędrzejewski-Szmek
1915 -- Berlin, 2013-09-13
1919 * The documentation has been updated to cover the various new
1920 concepts introduced with 205.
1922 * Unit files now understand the new %v specifier which
1923 resolves to the kernel version string as returned by "uname
1926 * systemctl now supports filtering the unit list output by
1927 load state, active state and sub state, using the new
1930 * "systemctl status" will now show the results of the
1931 condition checks (like ConditionPathExists= and similar) of
1932 the last start attempts of the unit. They are also logged to
1935 * "journalctl -b" may now be used to look for boot output of a
1936 specific boot. Try "journalctl -b -1" for the previous boot,
1937 but the syntax is substantially more powerful.
1939 * "journalctl --show-cursor" has been added which prints the
1940 cursor string the last shown log line. This may then be used
1941 with the new "journalctl --after-cursor=" switch to continue
1942 browsing logs from that point on.
1944 * "journalctl --force" may now be used to force regeneration
1947 * Creation of "dead" device nodes has been moved from udev
1948 into kmod and tmpfiles. Previously, udev would read the kmod
1949 databases to pre-generate dead device nodes based on meta
1950 information contained in kernel modules, so that these would
1951 be auto-loaded on access rather then at boot. As this
1952 does not really have much to do with the exposing actual
1953 kernel devices to userspace this has always been slightly
1954 alien in the udev codebase. Following the new scheme kmod
1955 will now generate a runtime snippet for tmpfiles from the
1956 module meta information and it now is tmpfiles' job to the
1957 create the nodes. This also allows overriding access and
1958 other parameters for the nodes using the usual tmpfiles
1959 facilities. As side effect this allows us to remove the
1960 CAP_SYS_MKNOD capability bit from udevd entirely.
1962 * logind's device ACLs may now be applied to these "dead"
1963 devices nodes too, thus finally allowing managed access to
1964 devices such as /dev/snd/sequencer whithout loading the
1965 backing module right-away.
1967 * A new RPM macro has been added that may be used to apply
1968 tmpfiles configuration during package installation.
1970 * systemd-detect-virt and ConditionVirtualization= now can
1971 detect User-Mode-Linux machines (UML).
1973 * journald will now implicitly log the effective capabilities
1974 set of processes in the message metadata.
1976 * systemd-cryptsetup has gained support for TrueCrypt volumes.
1978 * The initrd interface has been simplified (more specifically,
1979 support for passing performance data via environment
1980 variables and fsck results via files in /run has been
1981 removed). These features were non-essential, and are
1982 nowadays available in a much nicer way by having systemd in
1983 the initrd serialize its state and have the hosts systemd
1984 deserialize it again.
1986 * The udev "keymap" data files and tools to apply keyboard
1987 specific mappings of scan to key codes, and force-release
1988 scan code lists have been entirely replaced by a udev
1989 "keyboard" builtin and a hwdb data file.
1991 * systemd will now honour the kernel's "quiet" command line
1992 argument also during late shutdown, resulting in a
1993 completely silent shutdown when used.
1995 * There's now an option to control the SO_REUSEPORT socket
1996 option in .socket units.
1998 * Instance units will now automatically get a per-template
1999 subslice of system.slice unless something else is explicitly
2000 configured. For example, instances of sshd@.service will now
2001 implicitly be placed in system-sshd.slice rather than
2002 system.slice as before.
2004 * Test coverage support may now be enabled at build time.
2006 Contributions from: Dave Reisner, Frederic Crozat, Harald
2007 Hoyer, Holger Hans Peter Freyther, Jan Engelhardt, Jan
2008 Janssen, Jason St. John, Jesper Larsen, Kay Sievers, Lennart
2009 Poettering, Lukas Nykryn, Maciej Wereski, Martin Pitt, Michael
2010 Olbrich, Ramkumar Ramachandra, Ross Lagerwall, Shawn Landden,
2011 Thomas H.P. Andersen, Tom Gundersen, Tomasz Torcz, William
2012 Giokas, Zbigniew Jędrzejewski-Szmek
2014 -- Berlin, 2013-07-23
2018 * Two new unit types have been introduced:
2020 Scope units are very similar to service units, however, are
2021 created out of pre-existing processes -- instead of PID 1
2022 forking off the processes. By using scope units it is
2023 possible for system services and applications to group their
2024 own child processes (worker processes) in a powerful way
2025 which then maybe used to organize them, or kill them
2026 together, or apply resource limits on them.
2028 Slice units may be used to partition system resources in an
2029 hierarchial fashion and then assign other units to them. By
2030 default there are now three slices: system.slice (for all
2031 system services), user.slice (for all user sessions),
2032 machine.slice (for VMs and containers).
2034 Slices and scopes have been introduced primarily in
2035 context of the work to move cgroup handling to a
2036 single-writer scheme, where only PID 1
2037 creates/removes/manages cgroups.
2039 * There's a new concept of "transient" units. In contrast to
2040 normal units these units are created via an API at runtime,
2041 not from configuration from disk. More specifically this
2042 means it is now possible to run arbitrary programs as
2043 independent services, with all execution parameters passed
2044 in via bus APIs rather than read from disk. Transient units
2045 make systemd substantially more dynamic then it ever was,
2046 and useful as a general batch manager.
2048 * logind has been updated to make use of scope and slice units
2049 for managing user sessions. As a user logs in he will get
2050 his own private slice unit, to which all sessions are added
2051 as scope units. We also added support for automatically
2052 adding an instance of user@.service for the user into the
2053 slice. Effectively logind will no longer create cgroup
2054 hierarchies on its own now, it will defer entirely to PID 1
2055 for this by means of scope, service and slice units. Since
2056 user sessions this way become entities managed by PID 1
2057 the output of "systemctl" is now a lot more comprehensive.
2059 * A new mini-daemon "systemd-machined" has been added which
2060 may be used by virtualization managers to register local
2061 VMs/containers. nspawn has been updated accordingly, and
2062 libvirt will be updated shortly. machined will collect a bit
2063 of meta information about the VMs/containers, and assign
2064 them their own scope unit (see above). The collected
2065 meta-data is then made available via the "machinectl" tool,
2066 and exposed in "ps" and similar tools. machined/machinectl
2067 is compile-time optional.
2069 * As discussed earlier, the low-level cgroup configuration
2070 options ControlGroup=, ControlGroupModify=,
2071 ControlGroupPersistent=, ControlGroupAttribute= have been
2072 removed. Please use high-level attribute settings instead as
2073 well as slice units.
2075 * A new bus call SetUnitProperties() has been added to alter
2076 various runtime parameters of a unit. This is primarily
2077 useful to alter cgroup parameters dynamically in a nice way,
2078 but will be extended later on to make more properties
2079 modifiable at runtime. systemctl gained a new set-properties
2080 command that wraps this call.
2082 * A new tool "systemd-run" has been added which can be used to
2083 run arbitrary command lines as transient services or scopes,
2084 while configuring a number of settings via the command
2085 line. This tool is currently very basic, however already
2086 very useful. We plan to extend this tool to even allow
2087 queuing of execution jobs with time triggers from the
2088 command line, similar in fashion to "at".
2090 * nspawn will now inform the user explicitly that kernels with
2091 audit enabled break containers, and suggest the user to turn
2094 * Support for detecting the IMA and AppArmor security
2095 frameworks with ConditionSecurity= has been added.
2097 * journalctl gained a new "-k" switch for showing only kernel
2098 messages, mimicking dmesg output; in addition to "--user"
2099 and "--system" switches for showing only user's own logs
2102 * systemd-delta can now show information about drop-in
2103 snippets extending unit files.
2105 * libsystemd-bus has been substantially updated but is still
2106 not available as public API.
2108 * systemd will now look for the "debug" argument on the kernel
2109 command line and enable debug logging, similar to what
2110 "systemd.log_level=debug" already did before.
2112 * "systemctl set-default", "systemctl get-default" has been
2113 added to configure the default.target symlink, which
2114 controls what to boot into by default.
2116 * "systemctl set-log-level" has been added as a convenient
2117 way to raise and lower systemd logging threshold.
2119 * "systemd-analyze plot" will now show the time the various
2120 generators needed for execution, as well as information
2121 about the unit file loading.
2123 * libsystemd-journal gained a new sd_journal_open_files() call
2124 for opening specific journal files. journactl also gained a
2125 new switch to expose this new functionality. Previously we
2126 only supported opening all files from a directory, or all
2127 files from the system, as opening individual files only is
2128 racy due to journal file rotation.
2130 * systemd gained the new DefaultEnvironment= setting in
2131 /etc/systemd/system.conf to set environment variables for
2134 * If a privileged process logs a journal message with the
2135 OBJECT_PID= field set, then journald will automatically
2136 augment this with additional OBJECT_UID=, OBJECT_GID=,
2137 OBJECT_COMM=, OBJECT_EXE=, ... fields. This is useful if
2138 system services want to log events about specific client
2139 processes. journactl/systemctl has been updated to make use
2140 of this information if all log messages regarding a specific
2143 Contributions from: Auke Kok, Chengwei Yang, Colin Walters,
2144 Cristian Rodríguez, Daniel Albers, Daniel Wallace, Dave
2145 Reisner, David Coppa, David King, David Strauss, Eelco
2146 Dolstra, Gabriel de Perthuis, Harald Hoyer, Jan Alexander
2147 Steffens, Jan Engelhardt, Jan Janssen, Jason St. John, Johan
2148 Heikkilä, Karel Zak, Karol Lewandowski, Kay Sievers, Lennart
2149 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marius Vollmer,
2150 Martin Pitt, Michael Biebl, Michael Olbrich, Michael Tremer,
2151 Michal Schmidt, Michał Bartoszkiewicz, Nirbheek Chauhan,
2152 Pierre Neidhardt, Ross Burton, Ross Lagerwall, Sean McGovern,
2153 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar,
2154 Václav Pavlín, Zachary Cook, Zbigniew Jędrzejewski-Szmek,
2155 Łukasz Stelmach, 장동준
2159 * The Python bindings gained some minimal support for the APIs
2160 exposed by libsystemd-logind.
2162 * ConditionSecurity= gained support for detecting SMACK. Since
2163 this condition already supports SELinux and AppArmor we only
2164 miss IMA for this. Patches welcome!
2166 Contributions from: Karol Lewandowski, Lennart Poettering,
2167 Zbigniew Jędrzejewski-Szmek
2171 * systemd-nspawn will now create /etc/resolv.conf if
2172 necessary, before bind-mounting the host's file onto it.
2174 * systemd-nspawn will now store meta information about a
2175 container on the container's cgroup as extended attribute
2176 fields, including the root directory.
2178 * The cgroup hierarchy has been reworked in many ways. All
2179 objects any of the components systemd creates in the cgroup
2180 tree are now suffixed. More specifically, user sessions are
2181 now placed in cgroups suffixed with ".session", users in
2182 cgroups suffixed with ".user", and nspawn containers in
2183 cgroups suffixed with ".nspawn". Furthermore, all cgroup
2184 names are now escaped in a simple scheme to avoid collision
2185 of userspace object names with kernel filenames. This work
2186 is preparation for making these objects relocatable in the
2187 cgroup tree, in order to allow easy resource partitioning of
2188 these objects without causing naming conflicts.
2190 * systemctl list-dependencies gained the new switches
2191 --plain, --reverse, --after and --before.
2193 * systemd-inhibit now shows the process name of processes that
2194 have taken an inhibitor lock.
2196 * nss-myhostname will now also resolve "localhost"
2197 implicitly. This makes /etc/hosts an optional file and
2198 nicely handles that on IPv6 ::1 maps to both "localhost" and
2201 * libsystemd-logind.so gained a new call
2202 sd_get_machine_names() to enumerate running containers and
2203 VMs (currently only supported by very new libvirt and
2204 nspawn). sd_login_monitor can now be used to watch
2205 VMs/containers coming and going.
2207 * .include is not allowed recursively anymore, and only in
2208 unit files. Usually it is better to use drop-in snippets in
2209 .d/*.conf anyway, as introduced with systemd 198.
2211 * systemd-analyze gained a new "critical-chain" command that
2212 determines the slowest chain of units run during system
2213 boot-up. It is very useful for tracking down where
2214 optimizing boot time is the most beneficial.
2216 * systemd will no longer allow manipulating service paths in
2217 the name=systemd:/system cgroup tree using ControlGroup= in
2218 units. (But is still fine with it in all other dirs.)
2220 * There's a new systemd-nspawn@.service service file that may
2221 be used to easily run nspawn containers as system
2222 services. With the container's root directory in
2223 /var/lib/container/foobar it is now sufficient to run
2224 "systemctl start systemd-nspawn@foobar.service" to boot it.
2226 * systemd-cgls gained a new parameter "--machine" to list only
2227 the processes within a certain container.
2229 * ConditionSecurity= now can check for "apparmor". We still
2230 are lacking checks for SMACK and IMA for this condition
2231 check though. Patches welcome!
2233 * A new configuration file /etc/systemd/sleep.conf has been
2234 added that may be used to configure which kernel operation
2235 systemd is supposed to execute when "suspend", "hibernate"
2236 or "hybrid-sleep" is requested. This makes the new kernel
2237 "freeze" state accessible to the user.
2239 * ENV{SYSTEMD_WANTS} in udev rules will now implicitly escape
2240 the passed argument if applicable.
2242 Contributions from: Auke Kok, Colin Guthrie, Colin Walters,
2243 Cristian Rodríguez, Daniel Buch, Daniel Wallace, Dave Reisner,
2244 Evangelos Foutras, Greg Kroah-Hartman, Harald Hoyer, Josh
2245 Triplett, Kay Sievers, Lennart Poettering, Lukas Nykryn,
2246 MUNEDA Takahiro, Mantas Mikulėnas, Mirco Tischler, Nathaniel
2247 Chen, Nirbheek Chauhan, Ronny Chevalier, Ross Lagerwall, Tom
2248 Gundersen, Umut Tezduyar, Ville Skyttä, Zbigniew
2253 * The output of 'systemctl list-jobs' got some polishing. The
2254 '--type=' argument may now be passed more than once. A new
2255 command 'systemctl list-sockets' has been added which shows
2256 a list of kernel sockets systemd is listening on with the
2257 socket units they belong to, plus the units these socket
2260 * The experimental libsystemd-bus library got substantial
2261 updates to work in conjunction with the (also experimental)
2262 kdbus kernel project. It works well enough to exchange
2263 messages with some sophistication. Note that kdbus is not
2264 ready yet, and the library is mostly an elaborate test case
2265 for now, and not installable.
2267 * systemd gained a new unit 'systemd-static-nodes.service'
2268 that generates static device nodes earlier during boot, and
2269 can run in conjunction with udev.
2271 * libsystemd-login gained a new call sd_pid_get_user_unit()
2272 to retrieve the user systemd unit a process is running
2273 in. This is useful for systems where systemd is used as
2276 * systemd-nspawn now places all containers in the new /machine
2277 top-level cgroup directory in the name=systemd
2278 hierarchy. libvirt will soon do the same, so that we get a
2279 uniform separation of /system, /user and /machine for system
2280 services, user processes and containers/virtual
2281 machines. This new cgroup hierarchy is also useful to stick
2282 stable names to specific container instances, which can be
2283 recognized later this way (this name may be controlled
2284 via systemd-nspawn's new -M switch). libsystemd-login also
2285 gained a new call sd_pid_get_machine_name() to retrieve the
2286 name of the container/VM a specific process belongs to.
2288 * bootchart can now store its data in the journal.
2290 * libsystemd-journal gained a new call
2291 sd_journal_add_conjunction() for AND expressions to the
2292 matching logic. This can be used to express more complex
2293 logical expressions.
2295 * journactl can now take multiple --unit= and --user-unit=
2298 * The cryptsetup logic now understands the "luks.key=" kernel
2299 command line switch for specifying a file to read the
2300 decryption key from. Also, if a configured key file is not
2301 found the tool will now automatically fall back to prompting
2304 * Python systemd.journal module was updated to wrap recently
2305 added functions from libsystemd-journal. The interface was
2306 changed to bring the low level interface in s.j._Reader
2307 closer to the C API, and the high level interface in
2308 s.j.Reader was updated to wrap and convert all data about
2311 Contributions from: Anatol Pomozov, Auke Kok, Harald Hoyer,
2312 Henrik Grindal Bakken, Josh Triplett, Kay Sievers, Lennart
2313 Poettering, Lukas Nykryn, Mantas Mikulėnas Marius Vollmer,
2314 Martin Jansa, Martin Pitt, Michael Biebl, Michal Schmidt,
2315 Mirco Tischler, Pali Rohar, Simon Peeters, Steven Hiscocks,
2316 Tom Gundersen, Zbigniew Jędrzejewski-Szmek
2320 * journalctl --update-catalog now understands a new --root=
2321 option to operate on catalogs found in a different root
2324 * During shutdown after systemd has terminated all running
2325 services a final killing loop kills all remaining left-over
2326 processes. We will now print the name of these processes
2327 when we send SIGKILL to them, since this usually indicates a
2330 * If /etc/crypttab refers to password files stored on
2331 configured mount points automatic dependencies will now be
2332 generated to ensure the specific mount is established first
2333 before the key file is attempted to be read.
2335 * 'systemctl status' will now show information about the
2336 network sockets a socket unit is listening on.
2338 * 'systemctl status' will also shown information about any
2339 drop-in configuration file for units. (Drop-In configuration
2340 files in this context are files such as
2341 /etc/systemd/systemd/foobar.service.d/*.conf)
2343 * systemd-cgtop now optionally shows summed up CPU times of
2344 cgroups. Press '%' while running cgtop to switch between
2345 percentage and absolute mode. This is useful to determine
2346 which cgroups use up the most CPU time over the entire
2347 runtime of the system. systemd-cgtop has also been updated
2348 to be 'pipeable' for processing with further shell tools.
2350 * 'hostnamectl set-hostname' will now allow setting of FQDN
2353 * The formatting and parsing of time span values has been
2354 changed. The parser now understands fractional expressions
2355 such as "5.5h". The formatter will now output fractional
2356 expressions for all time spans under 1min, i.e. "5.123456s"
2357 rather than "5s 123ms 456us". For time spans under 1s
2358 millisecond values are shown, for those under 1ms
2359 microsecond values are shown. This should greatly improve
2360 all time-related output of systemd.
2362 * libsystemd-login and libsystemd-journal gained new
2363 functions for querying the poll() events mask and poll()
2364 timeout value for integration into arbitrary event
2367 * localectl gained the ability to list available X11 keymaps
2368 (models, layouts, variants, options).
2370 * 'systemd-analyze dot' gained the ability to filter for
2371 specific units via shell-style globs, to create smaller,
2372 more useful graphs. I.e. it is now possible to create simple
2373 graphs of all the dependencies between only target units, or
2374 of all units that Avahi has dependencies with.
2376 Contributions from: Cristian Rodríguez, Dr. Tilmann Bubeck,
2377 Harald Hoyer, Holger Hans Peter Freyther, Kay Sievers, Kelly
2378 Anderson, Koen Kooi, Lennart Poettering, Maksim Melnikau,
2379 Marc-Antoine Perennou, Marius Vollmer, Martin Pitt, Michal
2380 Schmidt, Oleksii Shevchuk, Ronny Chevalier, Simon McVittie,
2381 Steven Hiscocks, Thomas Weißschuh, Umut Tezduyar, Václav
2382 Pavlín, Zbigniew Jędrzejewski-Szmek, Łukasz Stelmach
2386 * The boot-time readahead implementation for rotating media
2387 will now read the read-ahead data in multiple passes which
2388 consist of all read requests made in equidistant time
2389 intervals. This means instead of strictly reading read-ahead
2390 data in its physical order on disk we now try to find a
2391 middle ground between physical and access time order.
2393 * /etc/os-release files gained a new BUILD_ID= field for usage
2394 on operating systems that provide continuous builds of OS
2397 Contributions from: Auke Kok, Eelco Dolstra, Kay Sievers,
2398 Lennart Poettering, Lukas Nykryn, Martin Pitt, Václav Pavlín
2399 William Douglas, Zbigniew Jędrzejewski-Szmek
2403 * systemd-python gained an API exposing libsystemd-daemon.
2405 * The SMACK setup logic gained support for uploading CIPSO
2408 * Behaviour of PrivateTmp=, ReadWriteDirectories=,
2409 ReadOnlyDirectories= and InaccessibleDirectories= has
2410 changed. The private /tmp and /var/tmp directories are now
2411 shared by all processes of a service (which means
2412 ExecStartPre= may now leave data in /tmp that ExecStart= of
2413 the same service can still access). When a service is
2414 stopped its temporary directories are immediately deleted
2415 (normal clean-up with tmpfiles is still done in addition to
2418 * By default, systemd will now set a couple of sysctl
2419 variables in the kernel: the safe sysrq options are turned
2420 on, IP route verification is turned on, and source routing
2421 disabled. The recently added hardlink and softlink
2422 protection of the kernel is turned on. These settings should
2423 be reasonably safe, and good defaults for all new systems.
2425 * The predictable network naming logic may now be turned off
2426 with a new kernel command line switch: net.ifnames=0.
2428 * A new libsystemd-bus module has been added that implements a
2429 pretty complete D-Bus client library. For details see:
2431 http://lists.freedesktop.org/archives/systemd-devel/2013-March/009797.html
2433 * journald will now explicitly flush the journal files to disk
2434 at the latest 5min after each write. The file will then also
2435 be marked offline until the next write. This should increase
2436 reliability in case of a crash. The synchronization delay
2437 can be configured via SyncIntervalSec= in journald.conf.
2439 * There's a new remote-fs-setup.target unit that can be used
2440 to pull in specific services when at least one remote file
2441 system is to be mounted.
2443 * There are new targets timers.target and paths.target as
2444 canonical targets to pull user timer and path units in
2445 from. This complements sockets.target with a similar
2446 purpose for socket units.
2448 * libudev gained a new call udev_device_set_attribute_value()
2449 to set sysfs attributes of a device.
2451 * The udev daemon now sets the default number of worker
2452 processes executed in parallel based on the number of available
2453 CPUs instead of the amount of available RAM. This is supposed
2454 to provide a more reliable default and limit a too aggressive
2455 paralellism for setups with 1000s of devices connected.
2457 Contributions from: Auke Kok, Colin Walters, Cristian
2458 Rodríguez, Daniel Buch, Dave Reisner, Frederic Crozat, Hannes
2459 Reinecke, Harald Hoyer, Jan Alexander Steffens, Jan
2460 Engelhardt, Josh Triplett, Kay Sievers, Lennart Poettering,
2461 Mantas Mikulėnas, Martin Pitt, Mathieu Bridon, Michael Biebl,
2462 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nathaniel Chen,
2463 Oleksii Shevchuk, Ozan Çağlayan, Thomas Hindoe Paaboel
2464 Andersen, Tollef Fog Heen, Tom Gundersen, Umut Tezduyar,
2465 Zbigniew Jędrzejewski-Szmek
2469 * Configuration of unit files may now be extended via drop-in
2470 files without having to edit/override the unit files
2471 themselves. More specifically, if the administrator wants to
2472 change one value for a service file foobar.service he can
2473 now do so by dropping in a configuration snippet into
2474 /etc/systemd/system/foobar.service.d/*.conf. The unit logic
2475 will load all these snippets and apply them on top of the
2476 main unit configuration file, possibly extending or
2477 overriding its settings. Using these drop-in snippets is
2478 generally nicer than the two earlier options for changing
2479 unit files locally: copying the files from
2480 /usr/lib/systemd/system/ to /etc/systemd/system/ and editing
2481 them there; or creating a new file in /etc/systemd/system/
2482 that incorporates the original one via ".include". Drop-in
2483 snippets into these .d/ directories can be placed in any
2484 directory systemd looks for units in, and the usual
2485 overriding semantics between /usr/lib, /etc and /run apply
2488 * Most unit file settings which take lists of items can now be
2489 reset by assigning the empty string to them. For example,
2490 normally, settings such as Environment=FOO=BAR append a new
2491 environment variable assignment to the environment block,
2492 each time they are used. By assigning Environment= the empty
2493 string the environment block can be reset to empty. This is
2494 particularly useful with the .d/*.conf drop-in snippets
2495 mentioned above, since this adds the ability to reset list
2496 settings from vendor unit files via these drop-ins.
2498 * systemctl gained a new "list-dependencies" command for
2499 listing the dependencies of a unit recursively.
2501 * Inhibitors are now honored and listed by "systemctl
2502 suspend", "systemctl poweroff" (and similar) too, not only
2503 GNOME. These commands will also list active sessions by
2506 * Resource limits (as exposed by the various control group
2507 controllers) can now be controlled dynamically at runtime
2508 for all units. More specifically, you can now use a command
2509 like "systemctl set-cgroup-attr foobar.service cpu.shares
2510 2000" to alter the CPU shares a specific service gets. These
2511 settings are stored persistently on disk, and thus allow the
2512 administrator to easily adjust the resource usage of
2513 services with a few simple commands. This dynamic resource
2514 management logic is also available to other programs via the
2515 bus. Almost any kernel cgroup attribute and controller is
2518 * systemd-vconsole-setup will now copy all font settings to
2519 all allocated VTs, where it previously applied them only to
2522 * libsystemd-login gained the new sd_session_get_tty() API
2525 * This release drops support for a few legacy or
2526 distribution-specific LSB facility names when parsing init
2527 scripts: $x-display-manager, $mail-transfer-agent,
2528 $mail-transport-agent, $mail-transfer-agent, $smtp,
2529 $null. Also, the mail-transfer-agent.target unit backing
2530 this has been removed. Distributions which want to retain
2531 compatibility with this should carry the burden for
2532 supporting this themselves and patch support for these back
2533 in, if they really need to. Also, the facilities $syslog and
2534 $local_fs are now ignored, since systemd does not support
2535 early-boot LSB init scripts anymore, and these facilities
2536 are implied anyway for normal services. syslog.target has
2539 * There are new bus calls on PID1's Manager object for
2540 cancelling jobs, and removing snapshot units. Previously,
2541 both calls were only available on the Job and Snapshot
2544 * systemd-journal-gatewayd gained SSL support.
2546 * The various "environment" files, such as /etc/locale.conf
2547 now support continuation lines with a backslash ("\") as
2548 last character in the line, similarly in style (but different)
2549 to how this is supported in shells.
2551 * For normal user processes the _SYSTEMD_USER_UNIT= field is
2552 now implicitly appended to every log entry logged. systemctl
2553 has been updated to filter by this field when operating on a
2554 user systemd instance.
2556 * nspawn will now implicitly add the CAP_AUDIT_WRITE and
2557 CAP_AUDIT_CONTROL capabilities to the capabilities set for
2558 the container. This makes it easier to boot unmodified
2559 Fedora systems in a container, which however still requires
2560 audit=0 to be passed on the kernel command line. Auditing in
2561 kernel and userspace is unfortunately still too broken in
2562 context of containers, hence we recommend compiling it out
2563 of the kernel or using audit=0. Hopefully this will be fixed
2564 one day for good in the kernel.
2566 * nspawn gained the new --bind= and --bind-ro= parameters to
2567 bind mount specific directories from the host into the
2570 * nspawn will now mount its own devpts file system instance
2571 into the container, in order not to leak pty devices from
2572 the host into the container.
2574 * systemd will now read the firmware boot time performance
2575 information from the EFI variables, if the used boot loader
2576 supports this, and takes it into account for boot performance
2577 analysis via "systemd-analyze". This is currently supported
2578 only in conjunction with Gummiboot, but could be supported
2579 by other boot loaders too. For details see:
2581 http://www.freedesktop.org/wiki/Software/systemd/BootLoaderInterface
2583 * A new generator has been added that automatically mounts the
2584 EFI System Partition (ESP) to /boot, if that directory
2585 exists, is empty, and no other file system has been
2586 configured to be mounted there.
2588 * logind will now send out PrepareForSleep(false) out
2589 unconditionally, after coming back from suspend. This may be
2590 used by applications as asynchronous notification for
2591 system resume events.
2593 * "systemctl unlock-sessions" has been added, that allows
2594 unlocking the screens of all user sessions at once, similar
2595 to how "systemctl lock-sessions" already locked all users
2596 sessions. This is backed by a new D-Bus call UnlockSessions().
2598 * "loginctl seat-status" will now show the master device of a
2599 seat. (i.e. the device of a seat that needs to be around for
2600 the seat to be considered available, usually the graphics
2603 * tmpfiles gained a new "X" line type, that allows
2604 configuration of files and directories (with wildcards) that
2605 shall be excluded from automatic cleanup ("aging").
2607 * udev default rules set the device node permissions now only
2608 at "add" events, and do not change them any longer with a
2609 later "change" event.
2611 * The log messages for lid events and power/sleep keypresses
2612 now carry a message ID.
2614 * We now have a substantially larger unit test suite, but this
2615 continues to be work in progress.
2617 * udevadm hwdb gained a new --root= parameter to change the
2618 root directory to operate relative to.
2620 * logind will now issue a background sync() request to the kernel
2621 early at shutdown, so that dirty buffers are flushed to disk early
2622 instead of at the last moment, in order to optimize shutdown
2625 * A new bootctl tool has been added that is an interface for
2626 certain boot loader operations. This is currently a preview
2627 and is likely to be extended into a small mechanism daemon
2628 like timedated, localed, hostnamed, and can be used by
2629 graphical UIs to enumerate available boot options, and
2630 request boot into firmware operations.
2632 * systemd-bootchart has been relicensed to LGPLv2.1+ to match
2633 the rest of the package. It also has been updated to work
2634 correctly in initrds.
2636 * Policykit previously has been runtime optional, and is now
2637 also compile time optional via a configure switch.
2639 * systemd-analyze has been reimplemented in C. Also "systemctl
2640 dot" has moved into systemd-analyze.
2642 * "systemctl status" with no further parameters will now print
2643 the status of all active or failed units.
2645 * Operations such as "systemctl start" can now be executed
2646 with a new mode "--irreversible" which may be used to queue
2647 operations that cannot accidentally be reversed by a later
2648 job queuing. This is by default used to make shutdown
2649 requests more robust.
2651 * The Python API of systemd now gained a new module for
2652 reading journal files.
2654 * A new tool kernel-install has been added that can install
2655 kernel images according to the Boot Loader Specification:
2657 http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec
2659 * Boot time console output has been improved to provide
2660 animated boot time output for hanging jobs.
2662 * A new tool systemd-activate has been added which can be used
2663 to test socket activation with, directly from the command
2664 line. This should make it much easier to test and debug
2665 socket activation in daemons.
2667 * journalctl gained a new "--reverse" (or -r) option to show
2668 journal output in reverse order (i.e. newest line first).
2670 * journalctl gained a new "--pager-end" (or -e) option to jump
2671 to immediately jump to the end of the journal in the
2672 pager. This is only supported in conjunction with "less".
2674 * journalctl gained a new "--user-unit=" option, that works
2675 similarly to "--unit=" but filters for user units rather than
2678 * A number of unit files to ease adoption of systemd in
2679 initrds has been added. This moves some minimal logic from
2680 the various initrd implementations into systemd proper.
2682 * The journal files are now owned by a new group
2683 "systemd-journal", which exists specifically to allow access
2684 to the journal, and nothing else. Previously, we used the
2685 "adm" group for that, which however possibly covers more
2686 than just journal/log file access. This new group is now
2687 already used by systemd-journal-gatewayd to ensure this
2688 daemon gets access to the journal files and as little else
2689 as possible. Note that "make install" will also set FS ACLs
2690 up for /var/log/journal to give "adm" and "wheel" read
2691 access to it, in addition to "systemd-journal" which owns
2692 the journal files. We recommend that packaging scripts also
2693 add read access to "adm" + "wheel" to /var/log/journal, and
2694 all existing/future journal files. To normal users and
2695 administrators little changes, however packagers need to
2696 ensure to create the "systemd-journal" system group at
2697 package installation time.
2699 * The systemd-journal-gatewayd now runs as unprivileged user
2700 systemd-journal-gateway:systemd-journal-gateway. Packaging
2701 scripts need to create these system user/group at
2704 * timedated now exposes a new boolean property CanNTP that
2705 indicates whether a local NTP service is available or not.
2707 * systemd-detect-virt will now also detect xen PVs
2709 * The pstore file system is now mounted by default, if it is
2712 * In addition to the SELinux and IMA policies we will now also
2713 load SMACK policies at early boot.
2715 Contributions from: Adel Gadllah, Aleksander Morgado, Auke
2716 Kok, Ayan George, Bastien Nocera, Colin Walters, Daniel Buch,
2717 Daniel Wallace, Dave Reisner, David Herrmann, David Strauss,
2718 Eelco Dolstra, Enrico Scholz, Frederic Crozat, Harald Hoyer,
2719 Jan Janssen, Jonathan Callen, Kay Sievers, Lennart Poettering,
2720 Lukas Nykryn, Mantas Mikulėnas, Marc-Antoine Perennou, Martin
2721 Pitt, Mauro Dreissig, Max F. Albrecht, Michael Biebl, Michael
2722 Olbrich, Michal Schmidt, Michal Sekletar, Michal Vyskocil,
2723 Michał Bartoszkiewicz, Mirco Tischler, Nathaniel Chen, Nestor
2724 Ovroy, Oleksii Shevchuk, Paul W. Frields, Piotr Drąg, Rob
2725 Clark, Ryan Lortie, Simon McVittie, Simon Peeters, Steven
2726 Hiscocks, Thomas Hindoe Paaboel Andersen, Tollef Fog Heen, Tom
2727 Gundersen, Umut Tezduyar, William Giokas, Zbigniew
2728 Jędrzejewski-Szmek, Zeeshan Ali (Khattak)
2732 * Timer units now support calendar time events in addition to
2733 monotonic time events. That means you can now trigger a unit
2734 based on a calendar time specification such as "Thu,Fri
2735 2013-*-1,5 11:12:13" which refers to 11:12:13 of the first
2736 or fifth day of any month of the year 2013, given that it is
2737 a thursday or friday. This brings timer event support
2738 considerably closer to cron's capabilities. For details on
2739 the supported calendar time specification language see
2742 * udev now supports a number of different naming policies for
2743 network interfaces for predictable names, and a combination
2744 of these policies is now the default. Please see this wiki
2745 document for details:
2747 http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames
2749 * Auke Kok's bootchart implementation has been added to the
2750 systemd tree. It is an optional component that can graph the
2751 boot in quite some detail. It is one of the best bootchart
2752 implementations around and minimal in its code and
2755 * nss-myhostname has been integrated into the systemd source
2756 tree. nss-myhostname guarantees that the local hostname
2757 always stays resolvable via NSS. It has been a weak
2758 requirement of systemd-hostnamed since a long time, and
2759 since its code is actually trivial we decided to just
2760 include it in systemd's source tree. It can be turned off
2761 with a configure switch.
2763 * The read-ahead logic is now capable of properly detecting
2764 whether a btrfs file system is on SSD or rotating media, in
2765 order to optimize the read-ahead scheme. Previously, it was
2766 only capable of detecting this on traditional file systems
2769 * In udev, additional device properties are now read from the
2770 IAB in addition to the OUI database. Also, Bluetooth company
2771 identities are attached to the devices as well.
2773 * In service files %U may be used as specifier that is
2774 replaced by the configured user name of the service.
2776 * nspawn may now be invoked without a controlling TTY. This
2777 makes it suitable for invocation as its own service. This
2778 may be used to set up a simple containerized server system
2779 using only core OS tools.
2781 * systemd and nspawn can now accept socket file descriptors
2782 when they are started for socket activation. This enables
2783 implementation of socket activated nspawn
2784 containers. i.e. think about autospawning an entire OS image
2785 when the first SSH or HTTP connection is received. We expect
2786 that similar functionality will also be added to libvirt-lxc
2789 * journalctl will now suppress ANSI color codes when
2790 presenting log data.
2792 * systemctl will no longer show control group information for
2793 a unit if a the control group is empty anyway.
2795 * logind can now automatically suspend/hibernate/shutdown the
2798 * /etc/machine-info and hostnamed now also expose the chassis
2799 type of the system. This can be used to determine whether
2800 the local system is a laptop, desktop, handset or
2801 tablet. This information may either be configured by the
2802 user/vendor or is automatically determined from ACPI and DMI
2803 information if possible.
2805 * A number of PolicyKit actions are now bound together with
2806 "imply" rules. This should simplify creating UIs because
2807 many actions will now authenticate similar ones as well.
2809 * Unit files learnt a new condition ConditionACPower= which
2810 may be used to conditionalize a unit depending on whether an
2811 AC power source is connected or not, of whether the system
2812 is running on battery power.
2814 * systemctl gained a new "is-failed" verb that may be used in
2815 shell scripts and suchlike to check whether a specific unit
2816 is in the "failed" state.
2818 * The EnvironmentFile= setting in unit files now supports file
2819 globbing, and can hence be used to easily read a number of
2820 environment files at once.
2822 * systemd will no longer detect and recognize specific
2823 distributions. All distribution-specific #ifdeffery has been
2824 removed, systemd is now fully generic and
2825 distribution-agnostic. Effectively, not too much is lost as
2826 a lot of the code is still accessible via explicit configure
2827 switches. However, support for some distribution specific
2828 legacy configuration file formats has been dropped. We
2829 recommend distributions to simply adopt the configuration
2830 files everybody else uses now and convert the old
2831 configuration from packaging scripts. Most distributions
2832 already did that. If that's not possible or desirable,
2833 distributions are welcome to forward port the specific
2834 pieces of code locally from the git history.
2836 * When logging a message about a unit systemd will now always
2837 log the unit name in the message meta data.
2839 * localectl will now also discover system locale data that is
2840 not stored in locale archives, but directly unpacked.
2842 * logind will no longer unconditionally use framebuffer
2843 devices as seat masters, i.e. as devices that are required
2844 to be existing before a seat is considered preset. Instead,
2845 it will now look for all devices that are tagged as
2846 "seat-master" in udev. By default framebuffer devices will
2847 be marked as such, but depending on local systems other
2848 devices might be marked as well. This may be used to
2849 integrate graphics cards using closed source drivers (such
2850 as NVidia ones) more nicely into logind. Note however, that
2851 we recommend using the open source NVidia drivers instead,
2852 and no udev rules for the closed-source drivers will be
2853 shipped from us upstream.
2855 Contributions from: Adam Williamson, Alessandro Crismani, Auke
2856 Kok, Colin Walters, Daniel Wallace, Dave Reisner, David
2857 Herrmann, David Strauss, Dimitrios Apostolou, Eelco Dolstra,
2858 Eric Benoit, Giovanni Campagna, Hannes Reinecke, Henrik
2859 Grindal Bakken, Hermann Gausterer, Kay Sievers, Lennart
2860 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marcel Holtmann,
2861 Martin Pitt, Matthew Monaco, Michael Biebl, Michael Terry,
2862 Michal Schmidt, Michal Sekletar, Michał Bartoszkiewicz, Oleg
2863 Samarin, Pekka Lundstrom, Philip Nilsson, Ramkumar
2864 Ramachandra, Richard Yao, Robert Millan, Sami Kerola, Shawn
2865 Landden, Thomas Hindoe Paaboel Andersen, Thomas Jarosch,
2866 Tollef Fog Heen, Tom Gundersen, Umut Tezduyar, Zbigniew
2871 * udev gained support for loading additional device properties
2872 from an indexed database that is keyed by vendor/product IDs
2873 and similar device identifiers. For the beginning this
2874 "hwdb" is populated with data from the well-known PCI and
2875 USB database, but also includes PNP, ACPI and OID data. In
2876 the longer run this indexed database shall grow into
2877 becoming the one central database for non-essential
2878 userspace device metadata. Previously, data from the PCI/USB
2879 database was only attached to select devices, since the
2880 lookup was a relatively expensive operation due to O(n) time
2881 complexity (with n being the number of entries in the
2882 database). Since this is now O(1), we decided to add in this
2883 data for all devices where this is available, by
2884 default. Note that the indexed database needs to be rebuilt
2885 when new data files are installed. To achieve this you need
2886 to update your packaging scripts to invoke "udevadm hwdb
2887 --update" after installation of hwdb data files. For
2888 RPM-based distributions we introduced the new
2889 %udev_hwdb_update macro for this purpose.
2891 * The Journal gained support for the "Message Catalog", an
2892 indexed database to link up additional information with
2893 journal entries. For further details please check:
2895 http://www.freedesktop.org/wiki/Software/systemd/catalog
2897 The indexed message catalog database also needs to be
2898 rebuilt after installation of message catalog files. Use
2899 "journalctl --update-catalog" for this. For RPM-based
2900 distributions we introduced the %journal_catalog_update
2901 macro for this purpose.
2903 * The Python Journal bindings gained support for the standard
2904 Python logging framework.
2906 * The Journal API gained new functions for checking whether
2907 the underlying file system of a journal file is capable of
2908 properly reporting file change notifications, or whether
2909 applications that want to reflect journal changes "live"
2910 need to recheck journal files continuously in appropriate
2913 * It is now possible to set the "age" field for tmpfiles
2914 entries to 0, indicating that files matching this entry
2915 shall always be removed when the directories are cleaned up.
2917 * coredumpctl gained a new "gdb" verb which invokes gdb
2918 right-away on the selected coredump.
2920 * There's now support for "hybrid sleep" on kernels that
2921 support this, in addition to "suspend" and "hibernate". Use
2922 "systemctl hybrid-sleep" to make use of this.
2924 * logind's HandleSuspendKey= setting (and related settings)
2925 now gained support for a new "lock" setting to simply
2926 request the screen lock on all local sessions, instead of
2927 actually executing a suspend or hibernation.
2929 * systemd will now mount the EFI variables file system by
2932 * Socket units now gained support for configuration of the
2933 SMACK security label.
2935 * timedatectl will now output the time of the last and next
2936 daylight saving change.
2938 * We dropped support for various legacy and distro-specific
2939 concepts, such as insserv, early-boot SysV services
2940 (i.e. those for non-standard runlevels such as 'b' or 'S')
2941 or ArchLinux /etc/rc.conf support. We recommend the
2942 distributions who still need support this to either continue
2943 to maintain the necessary patches downstream, or find a
2944 different solution. (Talk to us if you have questions!)
2946 * Various systemd components will now bypass PolicyKit checks
2947 for root and otherwise handle properly if PolicyKit is not
2948 found to be around. This should fix most issues for
2949 PolicyKit-less systems. Quite frankly this should have been
2950 this way since day one. It is absolutely our intention to
2951 make systemd work fine on PolicyKit-less systems, and we
2952 consider it a bug if something does not work as it should if
2953 PolicyKit is not around.
2955 * For embedded systems it is now possible to build udev and
2956 systemd without blkid and/or kmod support.
2958 * "systemctl switch-root" is now capable of switching root
2959 more than once. I.e. in addition to transitions from the
2960 initrd to the host OS it is now possible to transition to
2961 further OS images from the host. This is useful to implement
2962 offline updating tools.
2964 * Various other additions have been made to the RPM macros
2965 shipped with systemd. Use %udev_rules_update() after
2966 installing new udev rules files. %_udevhwdbdir,
2967 %_udevrulesdir, %_journalcatalogdir, %_tmpfilesdir,
2968 %_sysctldir are now available which resolve to the right
2969 directories for packages to place various data files in.
2971 * journalctl gained the new --full switch (in addition to
2972 --all, to disable ellipsation for long messages.
2974 Contributions from: Anders Olofsson, Auke Kok, Ben Boeckel,
2975 Colin Walters, Cosimo Cecchi, Daniel Wallace, Dave Reisner,
2976 Eelco Dolstra, Holger Hans Peter Freyther, Kay Sievers,
2977 Chun-Yi Lee, Lekensteyn, Lennart Poettering, Mantas Mikulėnas,
2978 Marti Raudsepp, Martin Pitt, Mauro Dreissig, Michael Biebl,
2979 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nis Martensen,
2980 Oleksii Shevchuk, Olivier Brunel, Ramkumar Ramachandra, Thomas
2981 Bächler, Thomas Hindoe Paaboel Andersen, Tom Gundersen, Tony
2982 Camuso, Umut Tezduyar, Zbigniew Jędrzejewski-Szmek
2986 * journalctl gained new --since= and --until= switches to
2987 filter by time. It also now supports nice filtering for
2988 units via --unit=/-u.
2990 * Type=oneshot services may use ExecReload= and do the
2993 * The journal daemon now supports time-based rotation and
2994 vacuuming, in addition to the usual disk-space based
2997 * The journal will now index the available field values for
2998 each field name. This enables clients to show pretty drop
2999 downs of available match values when filtering. The bash
3000 completion of journalctl has been updated
3001 accordingly. journalctl gained a new switch -F to list all
3002 values a certain field takes in the journal database.
3004 * More service events are now written as structured messages
3005 to the journal, and made recognizable via message IDs.
3007 * The timedated, localed and hostnamed mini-services which
3008 previously only provided support for changing time, locale
3009 and hostname settings from graphical DEs such as GNOME now
3010 also have a minimal (but very useful) text-based client
3011 utility each. This is probably the nicest way to changing
3012 these settings from the command line now, especially since
3013 it lists available options and is fully integrated with bash
3016 * There's now a new tool "systemd-coredumpctl" to list and
3017 extract coredumps from the journal.
3019 * We now install a README each in /var/log/ and
3020 /etc/rc.d/init.d explaining where the system logs and init
3021 scripts went. This hopefully should help folks who go to
3022 that dirs and look into the otherwise now empty void and
3023 scratch their heads.
3025 * When user-services are invoked (by systemd --user) the
3026 $MANAGERPID env var is set to the PID of systemd.
3028 * SIGRTMIN+24 when sent to a --user instance will now result
3029 in immediate termination of systemd.
3031 * gatewayd received numerous feature additions such as a
3032 "follow" mode, for live syncing and filtering.
3034 * browse.html now allows filtering and showing detailed
3035 information on specific entries. Keyboard navigation and
3036 mouse screen support has been added.
3038 * gatewayd/journalctl now supports HTML5/JSON
3039 Server-Sent-Events as output.
3041 * The SysV init script compatibility logic will now
3042 heuristically determine whether a script supports the
3043 "reload" verb, and only then make this available as
3046 * "systemctl status --follow" has been removed, use "journalctl
3049 * journald.conf's RuntimeMinSize=, PersistentMinSize= settings
3050 have been removed since they are hardly useful to be
3053 * And I'd like to take the opportunity to specifically mention
3054 Zbigniew for his great contributions. Zbigniew, you rock!
3056 Contributions from: Andrew Eikum, Christian Hesse, Colin
3057 Guthrie, Daniel J Walsh, Dave Reisner, Eelco Dolstra, Ferenc
3058 Wágner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Mantas
3059 Mikulėnas, Martin Mikkelsen, Martin Pitt, Michael Olbrich,
3060 Michael Stapelberg, Michal Schmidt, Sebastian Ott, Thomas
3061 Bächler, Umut Tezduyar, Will Woods, Wulf C. Krueger, Zbigniew
3062 Jędrzejewski-Szmek, Сковорода Никита Андреевич
3066 * If /etc/vconsole.conf is non-existent or empty we will no
3067 longer load any console font or key map at boot by
3068 default. Instead the kernel defaults will be left
3069 intact. This is definitely the right thing to do, as no
3070 configuration should mean no configuration, and hard-coding
3071 font names that are different on all archs is probably a bad
3072 idea. Also, the kernel default key map and font should be
3073 good enough for most cases anyway, and mostly identical to
3074 the userspace fonts/key maps we previously overloaded them
3075 with. If distributions want to continue to default to a
3076 non-kernel font or key map they should ship a default
3077 /etc/vconsole.conf with the appropriate contents.
3079 Contributions from: Colin Walters, Daniel J Walsh, Dave
3080 Reisner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Tollef
3081 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
3085 * journalctl gained a new --cursor= switch to show entries
3086 starting from the specified location in the journal.
3088 * We now enforce a size limit on journal entry fields exported
3089 with "-o json" in journalctl. Fields larger than 4K will be
3090 assigned null. This can be turned off with --all.
3092 * An (optional) journal gateway daemon is now available as
3093 "systemd-journal-gatewayd.service". This service provides
3094 access to the journal via HTTP and JSON. This functionality
3095 will be used to implement live log synchronization in both
3096 pull and push modes, but has various other users too, such
3097 as easy log access for debugging of embedded devices. Right
3098 now it is already useful to retrieve the journal via HTTP:
3100 # systemctl start systemd-journal-gatewayd.service
3101 # wget http://localhost:19531/entries
3103 This will download the journal contents in a
3104 /var/log/messages compatible format. The same as JSON:
3106 # curl -H"Accept: application/json" http://localhost:19531/entries
3108 This service is also accessible via a web browser where a
3109 single static HTML5 app is served that uses the JSON logic
3110 to enable the user to do some basic browsing of the
3111 journal. This will be extended later on. Here's an example
3112 screenshot of this app in its current state:
3114 http://0pointer.de/public/journal-gatewayd
3116 Contributions from: Kay Sievers, Lennart Poettering, Robert
3117 Milasan, Tom Gundersen
3121 * The bash completion logic is now available for journalctl
3124 * We do not mount the "cpuset" controller anymore together with
3125 "cpu" and "cpuacct", as "cpuset" groups generally cannot be
3126 started if no parameters are assigned to it. "cpuset" hence
3127 broke code that assumed it it could create "cpu" groups and
3130 * journalctl -f will now subscribe to terminal size changes,
3131 and line break accordingly.
3133 Contributions from: Dave Reisner, Kay Sievers, Lennart
3134 Poettering, Lukas Nykrynm, Mirco Tischler, Václav Pavlín
3138 * nspawn will now create a symlink /etc/localtime in the
3139 container environment, copying the host's timezone
3140 setting. Previously this has been done via a bind mount, but
3141 since symlinks cannot be bind mounted this has now been
3142 changed to create/update the appropriate symlink.
3144 * journalctl -n's line number argument is now optional, and
3145 will default to 10 if omitted.
3147 * journald will now log the maximum size the journal files may
3148 take up on disk. This is particularly useful if the default
3149 built-in logic of determining this parameter from the file
3150 system size is used. Use "systemctl status
3151 systemd-journald.service" to see this information.
3153 * The multi-seat X wrapper tool has been stripped down. As X
3154 is now capable of enumerating graphics devices via udev in a
3155 seat-aware way the wrapper is not strictly necessary
3156 anymore. A stripped down temporary stop-gap is still shipped
3157 until the upstream display managers have been updated to
3158 fully support the new X logic. Expect this wrapper to be
3159 removed entirely in one of the next releases.
3161 * HandleSleepKey= in logind.conf has been split up into
3162 HandleSuspendKey= and HandleHibernateKey=. The old setting
3163 is not available anymore. X11 and the kernel are
3164 distuingishing between these keys and we should too. This
3165 also means the inhibition lock for these keys has been split
3168 Contributions from: Dave Airlie, Eelco Dolstra, Lennart
3169 Poettering, Lukas Nykryn, Václav Pavlín
3173 * Whenever a unit changes state we will now log this to the
3174 journal and show along the unit's own log output in
3177 * ConditionPathIsMountPoint= can now properly detect bind
3178 mount points too. (Previously, a bind mount of one file
3179 system to another place in the same file system could not be
3180 detected as mount, since they shared struct stat's st_dev
3183 * We will now mount the cgroup controllers cpu, cpuacct,
3184 cpuset and the controllers net_cls, net_prio together by
3187 * nspawn containers will now have a virtualized boot
3188 ID. (i.e. /proc/sys/kernel/random/boot_id is now mounted
3189 over with a randomized ID at container initialization). This
3190 has the effect of making "journalctl -b" do the right thing
3193 * The JSON output journal serialization has been updated not
3194 to generate "endless" list objects anymore, but rather one
3195 JSON object per line. This is more in line how most JSON
3196 parsers expect JSON objects. The new output mode
3197 "json-pretty" has been added to provide similar output, but
3198 neatly aligned for readability by humans.
3200 * We dropped all explicit sync() invocations in the shutdown
3201 code. The kernel does this implicitly anyway in the kernel
3202 reboot() syscall. halt(8)'s -n option is now a compatibility
3205 * We now support virtualized reboot() in containers, as
3206 supported by newer kernels. We will fall back to exit() if
3207 CAP_SYS_REBOOT is not available to the container. Also,
3208 nspawn makes use of this now and will actually reboot the
3209 container if the containerized OS asks for that.
3211 * journalctl will only show local log output by default
3212 now. Use --merge (-m) to show remote log output, too.
3214 * libsystemd-journal gained the new sd_journal_get_usage()
3215 call to determine the current disk usage of all journal
3216 files. This is exposed in the new "journalctl --disk-usage"
3219 * journald gained a new configuration setting SplitMode= in
3220 journald.conf which may be used to control how user journals
3221 are split off. See journald.conf(5) for details.
3223 * A new condition type ConditionFileNotEmpty= has been added.
3225 * tmpfiles' "w" lines now support file globbing, to write
3226 multiple files at once.
3228 * We added Python bindings for the journal submission
3229 APIs. More Python APIs for a number of selected APIs will
3230 likely follow. Note that we intend to add native bindings
3231 only for the Python language, as we consider it common
3232 enough to deserve bindings shipped within systemd. There are
3233 various projects outside of systemd that provide bindings
3234 for languages such as PHP or Lua.
3236 * Many conditions will now resolve specifiers such as %i. In
3237 addition, PathChanged= and related directives of .path units
3238 now support specifiers as well.
3240 * There's now a new RPM macro definition for the system preset
3243 * journald will now warn if it ca not forward a message to the
3244 syslog daemon because its socket is full.
3246 * timedated will no longer write or process /etc/timezone,
3247 except on Debian. As we do not support late mounted /usr
3248 anymore /etc/localtime always being a symlink is now safe,
3249 and hence the information in /etc/timezone is not necessary
3252 * logind will now always reserve one VT for a text getty (VT6
3253 by default). Previously if more than 6 X sessions where
3254 started they took up all the VTs with auto-spawned gettys,
3255 so that no text gettys were available anymore.
3257 * udev will now automatically inform the btrfs kernel logic
3258 about btrfs RAID components showing up. This should make
3259 simple hotplug based btrfs RAID assembly work.
3261 * PID 1 will now increase its RLIMIT_NOFILE to 64K by default
3262 (but not for its children which will stay at the kernel
3263 default). This should allow setups with a lot more listening
3266 * systemd will now always pass the configured timezone to the
3267 kernel at boot. timedated will do the same when the timezone
3270 * logind's inhibition logic has been updated. By default,
3271 logind will now handle the lid switch, the power and sleep
3272 keys all the time, even in graphical sessions. If DEs want
3273 to handle these events on their own they should take the new
3274 handle-power-key, handle-sleep-key and handle-lid-switch
3275 inhibitors during their runtime. A simple way to achiveve
3276 that is to invoke the DE wrapped in an invocation of:
3278 systemd-inhibit --what=handle-power-key:handle-sleep-key:handle-lid-switch ...
3280 * Access to unit operations is now checked via SELinux taking
3281 the unit file label and client process label into account.
3283 * systemd will now notify the administrator in the journal
3284 when he over-mounts a non-empty directory.
3286 * There are new specifiers that are resolved in unit files,
3287 for the host name (%H), the machine ID (%m) and the boot ID
3290 Contributions from: Allin Cottrell, Auke Kok, Brandon Philips,
3291 Colin Guthrie, Colin Walters, Daniel J Walsh, Dave Reisner,
3292 Eelco Dolstra, Jan Engelhardt, Kay Sievers, Lennart
3293 Poettering, Lucas De Marchi, Lukas Nykryn, Mantas Mikulėnas,
3294 Martin Pitt, Matthias Clasen, Michael Olbrich, Pierre Schmitz,
3295 Shawn Landden, Thomas Hindoe Paaboel Andersen, Tom Gundersen,
3296 Václav Pavlín, Yin Kangkai, Zbigniew Jędrzejewski-Szmek
3300 * Support for reading structured kernel messages from
3301 /dev/kmsg has now been added and is enabled by default.
3303 * Support for reading kernel messages from /proc/kmsg has now
3304 been removed. If you want kernel messages in the journal
3305 make sure to run a recent kernel (>= 3.5) that supports
3306 reading structured messages from /dev/kmsg (see
3307 above). /proc/kmsg is now exclusive property of classic
3308 syslog daemons again.
3310 * The libudev API gained the new
3311 udev_device_new_from_device_id() call.
3313 * The logic for file system namespace (ReadOnlyDirectory=,
3314 ReadWriteDirectoy=, PrivateTmp=) has been reworked not to
3315 require pivot_root() anymore. This means fewer temporary
3316 directories are created below /tmp for this feature.
3318 * nspawn containers will now see and receive all submounts
3319 made on the host OS below the root file system of the
3322 * Forward Secure Sealing is now supported for Journal files,
3323 which provide cryptographical sealing of journal files so
3324 that attackers cannot alter log history anymore without this
3325 being detectable. Lennart will soon post a blog story about
3326 this explaining it in more detail.
3328 * There are two new service settings RestartPreventExitStatus=
3329 and SuccessExitStatus= which allow configuration of exit
3330 status (exit code or signal) which will be excepted from the
3331 restart logic, resp. consider successful.
3333 * journalctl gained the new --verify switch that can be used
3334 to check the integrity of the structure of journal files and
3335 (if Forward Secure Sealing is enabled) the contents of
3338 * nspawn containers will now be run with /dev/stdin, /dev/fd/
3339 and similar symlinks pre-created. This makes running shells
3340 as container init process a lot more fun.
3342 * The fstab support can now handle PARTUUID= and PARTLABEL=
3345 * A new ConditionHost= condition has been added to match
3346 against the hostname (with globs) and machine ID. This is
3347 useful for clusters where a single OS image is used to
3348 provision a large number of hosts which shall run slightly
3349 different sets of services.
3351 * Services which hit the restart limit will now be placed in a
3354 Contributions from: Bertram Poettering, Dave Reisner, Huang
3355 Hang, Kay Sievers, Lennart Poettering, Lukas Nykryn, Martin
3356 Pitt, Simon Peeters, Zbigniew Jędrzejewski-Szmek
3360 * When running in --user mode systemd will now become a
3361 subreaper (PR_SET_CHILD_SUBREAPER). This should make the ps
3362 tree a lot more organized.
3364 * A new PartOf= unit dependency type has been introduced that
3365 may be used to group services in a natural way.
3367 * "systemctl enable" may now be used to enable instances of
3370 * journalctl now prints error log levels in red, and
3371 warning/notice log levels in bright white. It also supports
3372 filtering by log level now.
3374 * cgtop gained a new -n switch (similar to top), to configure
3375 the maximum number of iterations to run for. It also gained
3376 -b, to run in batch mode (accepting no input).
3378 * The suffix ".service" may now be omitted on most systemctl
3379 command lines involving service unit names.
3381 * There's a new bus call in logind to lock all sessions, as
3382 well as a loginctl verb for it "lock-sessions".
3384 * libsystemd-logind.so gained a new call sd_journal_perror()
3385 that works similar to libc perror() but logs to the journal
3386 and encodes structured information about the error number.
3388 * /etc/crypttab entries now understand the new keyfile-size=
3391 * shutdown(8) now can send a (configurable) wall message when
3392 a shutdown is cancelled.
3394 * The mount propagation mode for the root file system will now
3395 default to "shared", which is useful to make containers work
3396 nicely out-of-the-box so that they receive new mounts from
3397 the host. This can be undone locally by running "mount
3398 --make-rprivate /" if needed.
3400 * The prefdm.service file has been removed. Distributions
3401 should maintain this unit downstream if they intend to keep
3402 it around. However, we recommend writing normal unit files
3403 for display managers instead.
3405 * Since systemd is a crucial part of the OS we will now
3406 default to a number of compiler switches that improve
3407 security (hardening) such as read-only relocations, stack
3408 protection, and suchlike.
3410 * The TimeoutSec= setting for services is now split into
3411 TimeoutStartSec= and TimeoutStopSec= to allow configuration
3412 of individual time outs for the start and the stop phase of
3415 Contributions from: Artur Zaprzala, Arvydas Sidorenko, Auke
3416 Kok, Bryan Kadzban, Dave Reisner, David Strauss, Harald Hoyer,
3417 Jim Meyering, Kay Sievers, Lennart Poettering, Mantas
3418 Mikulėnas, Martin Pitt, Michal Schmidt, Michal Sekletar, Peter
3419 Alfredsen, Shawn Landden, Simon Peeters, Terence Honles, Tom
3420 Gundersen, Zbigniew Jędrzejewski-Szmek
3424 * The journal and id128 C APIs are now fully documented as man
3427 * Extra safety checks have been added when transitioning from
3428 the initial RAM disk to the main system to avoid accidental
3431 * /etc/crypttab entries now understand the new keyfile-offset=
3434 * systemctl -t can now be used to filter by unit load state.
3436 * The journal C API gained the new sd_journal_wait() call to
3437 make writing synchronous journal clients easier.
3439 * journalctl gained the new -D switch to show journals from a
3442 * journalctl now displays a special marker between log
3443 messages of two different boots.
3445 * The journal is now explicitly flushed to /var via a service
3446 systemd-journal-flush.service, rather than implicitly simply
3447 by seeing /var/log/journal to be writable.
3449 * journalctl (and the journal C APIs) can now match for much
3450 more complex expressions, with alternatives and
3453 * When transitioning from the initial RAM disk to the main
3454 system we will now kill all processes in a killing spree to
3455 ensure no processes stay around by accident.
3457 * Three new specifiers may be used in unit files: %u, %h, %s
3458 resolve to the user name, user home directory resp. user
3459 shell. This is useful for running systemd user instances.
3461 * We now automatically rotate journal files if their data
3462 object hash table gets a fill level > 75%. We also size the
3463 hash table based on the configured maximum file size. This
3464 together should lower hash collisions drastically and thus
3465 speed things up a bit.
3467 * journalctl gained the new "--header" switch to introspect
3468 header data of journal files.
3470 * A new setting SystemCallFilters= has been added to services
3471 which may be used to apply blacklists or whitelists to
3472 system calls. This is based on SECCOMP Mode 2 of Linux 3.5.
3474 * nspawn gained a new --link-journal= switch (and quicker: -j)
3475 to link the container journal with the host. This makes it
3476 very easy to centralize log viewing on the host for all
3477 guests while still keeping the journal files separated.
3479 * Many bugfixes and optimizations
3481 Contributions from: Auke Kok, Eelco Dolstra, Harald Hoyer, Kay
3482 Sievers, Lennart Poettering, Malte Starostik, Paul Menzel, Rex
3483 Tsai, Shawn Landden, Tom Gundersen, Ville Skyttä, Zbigniew
3488 * Several tools now understand kernel command line arguments,
3489 which are only read when run in an initial RAM disk. They
3490 usually follow closely their normal counterparts, but are
3493 * There's a new tool to analyze the readahead files that are
3494 automatically generated at boot. Use:
3496 /usr/lib/systemd/systemd-readahead analyze /.readahead
3498 * We now provide an early debug shell on tty9 if this enabled. Use:
3500 systemctl enable debug-shell.service
3502 * All plymouth related units have been moved into the Plymouth
3503 package. Please make sure to upgrade your Plymouth version
3506 * systemd-tmpfiles now supports getting passed the basename of
3507 a configuration file only, in which case it will look for it
3508 in all appropriate directories automatically.
3510 * udevadm info now takes a /dev or /sys path as argument, and
3511 does the right thing. Example:
3513 udevadm info /dev/sda
3514 udevadm info /sys/class/block/sda
3516 * systemctl now prints a warning if a unit is stopped but a
3517 unit that might trigger it continues to run. Example: a
3518 service is stopped but the socket that activates it is left
3521 * "systemctl status" will now mention if the log output was
3522 shortened due to rotation since a service has been started.
3524 * The journal API now exposes functions to determine the
3525 "cutoff" times due to rotation.
3527 * journald now understands SIGUSR1 and SIGUSR2 for triggering
3528 immediately flushing of runtime logs to /var if possible,
3529 resp. for triggering immediate rotation of the journal
3532 * It is now considered an error if a service is attempted to
3533 be stopped that is not loaded.
3535 * XDG_RUNTIME_DIR now uses numeric UIDs instead of usernames.
3537 * systemd-analyze now supports Python 3
3539 * tmpfiles now supports cleaning up directories via aging
3540 where the first level dirs are always kept around but
3541 directories beneath it automatically aged. This is enabled
3542 by prefixing the age field with '~'.
3544 * Seat objects now expose CanGraphical, CanTTY properties
3545 which is required to deal with very fast bootups where the
3546 display manager might be running before the graphics drivers
3547 completed initialization.
3549 * Seat objects now expose a State property.
3551 * We now include RPM macros for service enabling/disabling
3552 based on the preset logic. We recommend RPM based
3553 distributions to make use of these macros if possible. This
3554 makes it simpler to reuse RPM spec files across
3557 * We now make sure that the collected systemd unit name is
3558 always valid when services log to the journal via
3561 * There's a new man page kernel-command-line(7) detailing all
3562 command line options we understand.
3564 * The fstab generator may now be disabled at boot by passing
3565 fstab=0 on the kernel command line.
3567 * A new kernel command line option modules-load= is now understood
3568 to load a specific kernel module statically, early at boot.
3570 * Unit names specified on the systemctl command line are now
3571 automatically escaped as needed. Also, if file system or
3572 device paths are specified they are automatically turned
3573 into the appropriate mount or device unit names. Example:
3575 systemctl status /home
3576 systemctl status /dev/sda
3578 * The SysVConsole= configuration option has been removed from
3579 system.conf parsing.
3581 * The SysV search path is no longer exported on the D-Bus
3584 * The Names= option is been removed from unit file parsing.
3586 * There's a new man page bootup(7) detailing the boot process.
3588 * Every unit and every generator we ship with systemd now
3589 comes with full documentation. The self-explanatory boot is
3592 * A couple of services gained "systemd-" prefixes in their
3593 name if they wrap systemd code, rather than only external
3594 code. Among them fsck@.service which is now
3595 systemd-fsck@.service.
3597 * The HaveWatchdog property has been removed from the D-Bus
3600 * systemd.confirm_spawn= on the kernel command line should now
3603 * There's a new man page crypttab(5) which details all options
3604 we actually understand.
3606 * systemd-nspawn gained a new --capability= switch to pass
3607 additional capabilities to the container.
3609 * timedated will now read known NTP implementation unit names
3610 from /usr/lib/systemd/ntp-units.d/*.list,
3611 systemd-timedated-ntp.target has been removed.
3613 * journalctl gained a new switch "-b" that lists log data of
3614 the current boot only.
3616 * The notify socket is in the abstract namespace again, in
3617 order to support daemons which chroot() at start-up.
3619 * There is a new Storage= configuration option for journald
3620 which allows configuration of where log data should go. This
3621 also provides a way to disable journal logging entirely, so
3622 that data collected is only forwarded to the console, the
3623 kernel log buffer or another syslog implementation.
3625 * Many bugfixes and optimizations
3627 Contributions from: Auke Kok, Colin Guthrie, Dave Reisner,
3628 David Strauss, Eelco Dolstra, Kay Sievers, Lennart Poettering,
3629 Lukas Nykryn, Michal Schmidt, Michal Sekletar, Paul Menzel,
3630 Shawn Landden, Tom Gundersen
3634 * "systemctl help <unit>" now shows the man page if one is
3637 * Several new man pages have been added.
3639 * MaxLevelStore=, MaxLevelSyslog=, MaxLevelKMsg=,
3640 MaxLevelConsole= can now be specified in
3641 journald.conf. These options allow reducing the amount of
3642 data stored on disk or forwarded by the log level.
3644 * TimerSlackNSec= can now be specified in system.conf for
3645 PID1. This allows system-wide power savings.
3647 Contributions from: Dave Reisner, Kay Sievers, Lauri Kasanen,
3648 Lennart Poettering, Malte Starostik, Marc-Antoine Perennou,
3653 * logind is now capable of (optionally) handling power and
3654 sleep keys as well as the lid switch.
3656 * journalctl now understands the syntax "journalctl
3657 /usr/bin/avahi-daemon" to get all log output of a specific
3660 * CapabilityBoundingSet= in system.conf now also influences
3661 the capability bound set of usermode helpers of the kernel.
3663 Contributions from: Daniel Drake, Daniel J. Walsh, Gert
3664 Michael Kulyk, Harald Hoyer, Jean Delvare, Kay Sievers,
3665 Lennart Poettering, Matthew Garrett, Matthias Clasen, Paul
3666 Menzel, Shawn Landden, Tero Roponen, Tom Gundersen
3670 * Note that we skipped 139 releases here in order to set the
3671 new version to something that is greater than both udev's
3672 and systemd's most recent version number.
3674 * udev: all udev sources are merged into the systemd source tree now.
3675 All future udev development will happen in the systemd tree. It
3676 is still fully supported to use the udev daemon and tools without
3677 systemd running, like in initramfs or other init systems. Building
3678 udev though, will require the *build* of the systemd tree, but
3679 udev can be properly *run* without systemd.
3681 * udev: /lib/udev/devices/ are not read anymore; systemd-tmpfiles
3682 should be used to create dead device nodes as workarounds for broken
3685 * udev: RUN+="socket:..." and udev_monitor_new_from_socket() is
3686 no longer supported. udev_monitor_new_from_netlink() needs to be
3687 used to subscribe to events.
3689 * udev: when udevd is started by systemd, processes which are left
3690 behind by forking them off of udev rules, are unconditionally cleaned
3691 up and killed now after the event handling has finished. Services or
3692 daemons must be started as systemd services. Services can be
3693 pulled-in by udev to get started, but they can no longer be directly
3694 forked by udev rules.
3696 * udev: the daemon binary is called systemd-udevd now and installed
3697 in /usr/lib/systemd/. Standalone builds or non-systemd systems need
3698 to adapt to that, create symlink, or rename the binary after building
3701 * libudev no longer provides these symbols:
3702 udev_monitor_from_socket()
3703 udev_queue_get_failed_list_entry()
3704 udev_get_{dev,sys,run}_path()
3705 The versions number was bumped and symbol versioning introduced.
3707 * systemd-loginctl and systemd-journalctl have been renamed
3708 to loginctl and journalctl to match systemctl.
3710 * The config files: /etc/systemd/systemd-logind.conf and
3711 /etc/systemd/systemd-journald.conf have been renamed to
3712 logind.conf and journald.conf. Package updates should rename
3713 the files to the new names on upgrade.
3715 * For almost all files the license is now LGPL2.1+, changed
3716 from the previous GPL2.0+. Exceptions are some minor stuff
3717 of udev (which will be changed to LGPL2.1 eventually, too),
3718 and the MIT licensed sd-daemon.[ch] library that is suitable
3719 to be used as drop-in files.
3721 * systemd and logind now handle system sleep states, in
3722 particular suspending and hibernating.
3724 * logind now implements a sleep/shutdown/idle inhibiting logic
3725 suitable for a variety of uses. Soonishly Lennart will blog
3726 about this in more detail.
3728 * var-run.mount and var-lock.mount are no longer provided
3729 (which prevously bind mounted these directories to their new
3730 places). Distributions which have not converted these
3731 directories to symlinks should consider stealing these files
3732 from git history and add them downstream.
3734 * We introduced the Documentation= field for units and added
3735 this to all our shipped units. This is useful to make it
3736 easier to explore the boot and the purpose of the various
3739 * All smaller setup units (such as
3740 systemd-vconsole-setup.service) now detect properly if they
3741 are run in a container and are skipped when
3742 appropriate. This guarantees an entirely noise-free boot in
3743 Linux container environments such as systemd-nspawn.
3745 * A framework for implementing offline system updates is now
3746 integrated, for details see:
3747 http://freedesktop.org/wiki/Software/systemd/SystemUpdates
3749 * A new service type Type=idle is available now which helps us
3750 avoiding ugly interleaving of getty output and boot status
3753 * There's now a system-wide CapabilityBoundingSet= option to
3754 globally reduce the set of capabilities for the
3755 system. This is useful to drop CAP_SYS_MKNOD, CAP_SYS_RAWIO,
3756 CAP_NET_RAW, CAP_SYS_MODULE, CAP_SYS_TIME, CAP_SYS_PTRACE or
3757 even CAP_NET_ADMIN system-wide for secure systems.
3759 * There are now system-wide DefaultLimitXXX= options to
3760 globally change the defaults of the various resource limits
3761 for all units started by PID 1.
3763 * Harald Hoyer's systemd test suite has been integrated into
3764 systemd which allows easy testing of systemd builds in qemu
3765 and nspawn. (This is really awesome! Ask us for details!)
3767 * The fstab parser is now implemented as generator, not inside
3770 * systemctl will now warn you if .mount units generated from
3771 /etc/fstab are out of date due to changes in fstab that
3772 have not been read by systemd yet.
3774 * systemd is now suitable for usage in initrds. Dracut has
3775 already been updated to make use of this. With this in place
3776 initrds get a slight bit faster but primarily are much
3777 easier to introspect and debug since "systemctl status" in
3778 the host system can be used to introspect initrd services,
3779 and the journal from the initrd is kept around too.
3781 * systemd-delta has been added, a tool to explore differences
3782 between user/admin configuration and vendor defaults.
3784 * PrivateTmp= now affects both /tmp and /var/tmp.
3786 * Boot time status messages are now much prettier and feature
3787 proper english language. Booting up systemd has never been
3790 * Read-ahead pack files now include the inode number of all
3791 files to pre-cache. When the inode changes the pre-caching
3792 is not attempted. This should be nicer to deal with updated
3793 packages which might result in changes of read-ahead
3796 * We now temporaritly lower the kernel's read_ahead_kb variable
3797 when collecting read-ahead data to ensure the kernel's
3798 built-in read-ahead does not add noise to our measurements
3799 of necessary blocks to pre-cache.
3801 * There's now RequiresMountsFor= to add automatic dependencies
3802 for all mounts necessary for a specific file system path.
3804 * MountAuto= and SwapAuto= have been removed from
3805 system.conf. Mounting file systems at boot has to take place
3808 * nspawn now learned a new switch --uuid= to set the machine
3809 ID on the command line.
3811 * nspawn now learned the -b switch to automatically search
3814 * vt102 is now the default TERM for serial TTYs, upgraded from
3817 * systemd-logind now works on VT-less systems.
3819 * The build tree has been reorganized. The individual
3820 components now have directories of their own.
3822 * A new condition type ConditionPathIsReadWrite= is now available.
3824 * nspawn learned the new -C switch to create cgroups for the
3825 container in other hierarchies.
3827 * We now have support for hardware watchdogs, configurable in
3830 * The scheduled shutdown logic now has a public API.
3832 * We now mount /tmp as tmpfs by default, but this can be
3833 masked and /etc/fstab can override it.
3835 * Since udisks does not make use of /media anymore we are not
3836 mounting a tmpfs on it anymore.
3838 * journalctl gained a new --local switch to only interleave
3839 locally generated journal files.
3841 * We can now load the IMA policy at boot automatically.
3843 * The GTK tools have been split off into a systemd-ui.
3845 Contributions from: Andreas Schwab, Auke Kok, Ayan George,
3846 Colin Guthrie, Daniel Mack, Dave Reisner, David Ward, Elan
3847 Ruusamäe, Frederic Crozat, Gergely Nagy, Guillermo Vidal,
3848 Hannes Reinecke, Harald Hoyer, Javier Jardón, Kay Sievers,
3849 Lennart Poettering, Lucas De Marchi, Léo Gillot-Lamure,
3850 Marc-Antoine Perennou, Martin Pitt, Matthew Monaco, Maxim
3851 A. Mikityanskiy, Michael Biebl, Michael Olbrich, Michal
3852 Schmidt, Nis Martensen, Patrick McCarty, Roberto Sassu, Shawn
3853 Landden, Sjoerd Simons, Sven Anders, Tollef Fog Heen, Tom
3858 * This is mostly a bugfix release
3860 * Support optional initialization of the machine ID from the
3861 KVM or container configured UUID.
3863 * Support immediate reboots with "systemctl reboot -ff"
3865 * Show /etc/os-release data in systemd-analyze output
3867 * Many bugfixes for the journal, including endianness fixes and
3868 ensuring that disk space enforcement works
3870 * sd-login.h is C++ comptaible again
3872 * Extend the /etc/os-release format on request of the Debian
3875 * We now refuse non-UTF8 strings used in various configuration
3876 and unit files. This is done to ensure we do not pass invalid
3877 data over D-Bus or expose it elsewhere.
3879 * Register Mimo USB Screens as suitable for automatic seat
3882 * Read SELinux client context from journal clients in a race
3885 * Reorder configuration file lookup order. /etc now always
3886 overrides /run in order to allow the administrator to always
3887 and unconditionally override vendor supplied or
3888 automatically generated data.
3890 * The various user visible bits of the journal now have man
3891 pages. We still lack man pages for the journal API calls
3894 * We now ship all man pages in HTML format again in the
3897 Contributions from: Dave Reisner, Dirk Eibach, Frederic
3898 Crozat, Harald Hoyer, Kay Sievers, Lennart Poettering, Marti
3899 Raudsepp, Michal Schmidt, Shawn Landden, Tero Roponen, Thierry
3904 * This is mostly a bugfix release
3906 * systems lacking /etc/os-release are no longer supported.
3908 * Various functionality updates to libsystemd-login.so
3910 * Track class of PAM logins to distuingish greeters from
3913 Contributions from: Kay Sievers, Lennart Poettering, Michael
3918 * This is an important bugfix release for v41.
3920 * Building man pages is now optional which should be useful
3921 for those building systemd from git but unwilling to install
3924 * Watchdog support for supervising services is now usable. In
3925 a future release support for hardware watchdogs
3926 (i.e. /dev/watchdog) will be added building on this.
3928 * Service start rate limiting is now configurable and can be
3929 turned off per service. When a start rate limit is hit a
3930 reboot can automatically be triggered.
3932 * New CanReboot(), CanPowerOff() bus calls in systemd-logind.
3934 Contributions from: Benjamin Franzke, Bill Nottingham,
3935 Frederic Crozat, Lennart Poettering, Michael Olbrich, Michal
3936 Schmidt, Michał Górny, Piotr Drąg
3940 * The systemd binary is installed /usr/lib/systemd/systemd now;
3941 An existing /sbin/init symlink needs to be adapted with the
3944 * The code that loads kernel modules has been ported to invoke
3945 libkmod directly, instead of modprobe. This means we do not
3946 support systems with module-init-tools anymore.
3948 * Watchdog support is now already useful, but still not
3951 * A new kernel command line option systemd.setenv= is
3952 understood to set system wide environment variables
3953 dynamically at boot.
3955 * We now limit the set of capabilities of systemd-journald.
3957 * We now set SIGPIPE to ignore by default, since it only is
3958 useful in shell pipelines, and has little use in general
3959 code. This can be disabled with IgnoreSIPIPE=no in unit
3962 Contributions from: Benjamin Franzke, Kay Sievers, Lennart
3963 Poettering, Michael Olbrich, Michal Schmidt, Tom Gundersen,
3968 * This is mostly a bugfix release
3970 * We now expose the reason why a service failed in the
3971 "Result" D-Bus property.
3973 * Rudimentary service watchdog support (will be completed over
3974 the next few releases.)
3976 * When systemd forks off in order execute some service we will
3977 now immediately changes its argv[0] to reflect which process
3978 it will execute. This is useful to minimize the time window
3979 with a generic argv[0], which makes bootcharts more useful
3981 Contributions from: Alvaro Soliverez, Chris Paulson-Ellis, Kay
3982 Sievers, Lennart Poettering, Michael Olbrich, Michal Schmidt,
3983 Mike Kazantsev, Ray Strode
3987 * This is mostly a test release, but incorporates many
3990 * New systemd-cgtop tool to show control groups by their
3993 * Linking against libacl for ACLs is optional again. If
3994 disabled, support tracking device access for active logins
3995 goes becomes unavailable, and so does access to the user
3996 journals by the respective users.
3998 * If a group "adm" exists, journal files are automatically
3999 owned by them, thus allow members of this group full access
4000 to the system journal as well as all user journals.
4002 * The journal now stores the SELinux context of the logging
4003 client for all entries.
4005 * Add C++ inclusion guards to all public headers
4007 * New output mode "cat" in the journal to print only text
4008 messages, without any meta data like date or time.
4010 * Include tiny X server wrapper as a temporary stop-gap to
4011 teach XOrg udev display enumeration. This is used by display
4012 managers such as gdm, and will go away as soon as XOrg
4013 learned native udev hotplugging for display devices.
4015 * Add new systemd-cat tool for executing arbitrary programs
4016 with STDERR/STDOUT connected to the journal. Can also act as
4017 BSD logger replacement, and does so by default.
4019 * Optionally store all locally generated coredumps in the
4020 journal along with meta data.
4022 * systemd-tmpfiles learnt four new commands: n, L, c, b, for
4023 writing short strings to files (for usage for /sys), and for
4024 creating symlinks, character and block device nodes.
4026 * New unit file option ControlGroupPersistent= to make cgroups
4027 persistent, following the mechanisms outlined in
4028 http://www.freedesktop.org/wiki/Software/systemd/PaxControlGroups
4030 * Support multiple local RTCs in a sane way
4032 * No longer monopolize IO when replaying readahead data on
4033 rotating disks, since we might starve non-file-system IO to
4034 death, since fanotify() will not see accesses done by blkid,
4037 * Do not show kernel threads in systemd-cgls anymore, unless
4038 requested with new -k switch.
4040 Contributions from: Dan Horák, Kay Sievers, Lennart
4041 Poettering, Michal Schmidt
4045 * This is mostly a test release, but incorporates many
4048 * The git repository moved to:
4049 git://anongit.freedesktop.org/systemd/systemd
4050 ssh://git.freedesktop.org/git/systemd/systemd
4052 * First release with the journal
4053 http://0pointer.de/blog/projects/the-journal.html
4055 * The journal replaces both systemd-kmsg-syslogd and
4056 systemd-stdout-bridge.
4058 * New sd_pid_get_unit() API call in libsystemd-logind
4060 * Many systemadm clean-ups
4062 * Introduce remote-fs-pre.target which is ordered before all
4063 remote mounts and may be used to start services before all
4066 * Added Mageia support
4068 * Add bash completion for systemd-loginctl
4070 * Actively monitor PID file creation for daemons which exit in
4071 the parent process before having finished writing the PID
4072 file in the daemon process. Daemons which do this need to be
4073 fixed (i.e. PID file creation must have finished before the
4074 parent exits), but we now react a bit more gracefully to them.
4076 * Add colourful boot output, mimicking the well-known output
4077 of existing distributions.
4079 * New option PassCredentials= for socket units, for
4080 compatibility with a recent kernel ABI breakage.
4082 * /etc/rc.local is now hooked in via a generator binary, and
4083 thus will no longer act as synchronization point during
4086 * systemctl list-unit-files now supports --root=.
4088 * systemd-tmpfiles now understands two new commands: z, Z for
4089 relabelling files according to the SELinux database. This is
4090 useful to apply SELinux labels to specific files in /sys,
4093 * Output of SysV services is now forwarded to both the console
4094 and the journal by default, not only just the console.
4096 * New man pages for all APIs from libsystemd-login.
4098 * The build tree got reorganized and a the build system is a
4099 lot more modular allowing embedded setups to specifically
4100 select the components of systemd they are interested in.
4102 * Support for Linux systems lacking the kernel VT subsystem is
4105 * configure's --with-rootdir= got renamed to
4106 --with-rootprefix= to follow the naming used by udev and
4109 * Unless specified otherwise we will now install to /usr instead
4110 of /usr/local by default.
4112 * Processes with '@' in argv[0][0] are now excluded from the
4113 final shut-down killing spree, following the logic explained
4115 http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons
4117 * All processes remaining in a service cgroup when we enter
4118 the START or START_PRE states are now killed with
4119 SIGKILL. That means it is no longer possible to spawn
4120 background processes from ExecStart= lines (which was never
4121 supported anyway, and bad style).
4123 * New PropagateReloadTo=/PropagateReloadFrom= options to bind
4124 reloading of units together.
4126 Contributions from: Bill Nottingham, Daniel J. Walsh, Dave
4127 Reisner, Dexter Morgan, Gregs Gregs, Jonathan Nieder, Kay
4128 Sievers, Lennart Poettering, Michael Biebl, Michal Schmidt,
4129 Michał Górny, Ran Benita, Thomas Jarosch, Tim Waugh, Tollef
4130 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek