1 systemd System and Service Manager
5 * If compiled with --enable-xkbcommon, systemd-localed will
6 verify x11 keymap settings by compiling the given keymap. It
7 will spew out warnings if the compilation fails. This
8 requires libxkbcommon to be installed.
12 * journalctl gained the new options -t/--identifier= to match
13 on the syslog identifier (aka "tag"), as well as --utc to
14 show log timestamps in the UTC timezone. journalctl now also
15 accepts -n/--lines=all to disable line capping in a pager.
17 * journalctl gained a new switch, --flush, that synchronously
18 flushes logs from /run/log/journal to /var/log/journal if
19 persistent storage is enabled. systemd-journal-flush.service
20 now waits until the operation is complete.
22 * Services can notify the manager before they start a reload
23 (by sending RELOADING=1) or shutdown (by sending
24 STOPPING=1). This allows the manager to track and show the
25 internal state of daemons and closes a race condition when
26 the process is still running but has closed its D-Bus
29 * Services with Type=oneshot do not have to have any ExecStart
32 * User units are now loaded also from
33 $XDG_RUNTIME_DIR/systemd/user/. This is similar to the
34 /run/systemd/user directory that was already previously
35 supported, but is under the control of the user.
37 * Job timeouts (i.e. time-outs on the time a job that is
38 queued stays in the run queue) can now optionally result in
39 immediate reboot or power-off actions (JobTimeoutAction= and
40 JobTimeoutRebootArgument=). This is useful on ".target"
41 units, to limit the maximum time a target remains
42 undispatched in the run queue, and to trigger an emergency
43 operation in such a case. This is now used by default to
44 turn off the system if boot-up (as defined by everything in
45 basic.target) hangs and does not complete for at least
46 15min. Also, if power-off or reboot hang for at least 30min
47 an immediate power-off/reboot operation is triggered. This
48 functionality is particularly useful to increase reliability
49 on embedded devices, but also on laptops which might
50 accidentally get powered on when carried in a backpack and
51 whose boot stays stuck in a hard disk encryption passphrase
54 * systemd-logind can be configured to also handle lid switch
55 events even when the machine is docked or multiple displays
56 are attached (HandleLidSwitchDocked= option).
58 * A helper binary and a service have been added which can be
59 used to resume from hibernation in the initramfs. A
60 generator will parse the resume= option on the kernel
61 command line to trigger resume.
63 * A user console daemon systemd-consoled has been
64 added. Currently, it is a preview, and will so far open a
65 single terminal on each session of the user marked as
66 Desktop=systemd-console.
68 * Route metrics can be specified for DHCP routes added by
71 * The SELinux context of socket-activated services can be set
72 from the information provided by the networking stack
73 (SELinuxContextFromNet= option).
75 * Userspace firmware loading support has been removed and
76 the minimum supported kernel version is thus bumped to 3.7.
78 * Timeout for udev workers has been increased from 1 to 3
79 minutes, but a warning will be printed after 1 minute to
80 help diagnose kernel modules that take a long time to load.
82 * Udev rules can now remove tags on devices with TAG-="foobar".
84 * systemd's readahead implementation has been removed. In many
85 circumstances it didn't give expected benefits even for
86 rotational disk drives and was becoming less relevant in the
87 age of SSDs. As none of the developers has been using
88 rotating media anymore, and nobody stepped up to actively
89 maintain this component of systemd it has now been removed.
91 * Swap units can use Options= to specify discard options.
92 Discard options specified for swaps in /etc/fstab are now
95 * Docker containers are now detected as a separate type of
98 * The Password Agent protocol gained support for queries where
99 the user input is shown, useful e.g. for user names.
100 systemd-ask-password gained a new --echo option to turn that
103 * The default sysctl.d/ snippets will now set:
105 net.core.default_qdisc = fq_codel
107 This selects Fair Queuing Controlled Delay as the default
108 queuing discipline for network interfaces. fq_codel helps
109 fight the network bufferbloat problem. It is believed to be
110 a good default with no tuning required for most workloads.
111 Downstream distributions may override this choice. On 10Gbit
112 servers that do not do forwarding, "fq" may perform better.
113 Systems without a good clocksource should use "pfifo_fast".
115 * If kdbus is enabled during build a new option BusPolicy= is
116 available for service units, that allows locking all service
117 processes into a stricter bus policy, in order to limit
118 access to various bus services, or even hide most of them
119 from the service's view entirely.
121 * networkctl will now show the .network and .link file
122 networkd has applied to a specific interface.
124 * sd-login gained a new API call sd_session_get_desktop() to
125 query which desktop environment has been selected for a
128 * UNIX utmp support is now compile-time optional to support
131 * systemctl gained two new commands "add-wants" and
132 "add-requires" for pulling in units from specific targets
135 * If the word "rescue" is specified on the kernel command line
136 the system will now boot into rescue mode (aka
137 rescue.target), which was previously available only by
138 specifying "1" or "systemd.unit=rescue.target" on the kernel
139 command line. This new kernel command line option nicely
140 mirrors the already existing "emergency" kernel command line
143 * New kernel command line options mount.usr=, mount.usrflags=,
144 mount.usrfstype= have been added that match root=, rootflags=,
145 rootfstype= but allow mounting a specific file system to
148 * The $NOTIFY_SOCKET is now also passed to control processes of
149 services, not only the main process.
151 * This version reenables support for fsck's -l switch. This
152 means at least version v2.25 of util-linux is required for
153 operation, otherwise dead-locks on device nodes may
154 occur. Again: you need to update util-linux to at least
155 v2.25 when updating systemd to v217.
157 * The "multi-seat-x" tool has been removed from systemd, as
158 its functionality has been integrated into X servers 1.16,
159 and the tool is hence redundant. It is recommended to update
160 display managers invoking this tool to simply invoke X
161 directly from now on, again.
163 * Support for the new ALLOW_INTERACTIVE_AUTHORIZATION D-Bus
164 message flag has been added for all of systemd's PolicyKit
165 authenticated method calls has been added. In particular
166 this now allows optional interactive authorization via
167 PolicyKit for many of PID1's privileged operations such as
168 unit file enabling and disabling.
170 * "udevadm hwdb --update" learnt a new switch "--usr" for
171 placing the rebuilt hardware database in /usr instead of
172 /etc. When used only hardware database entries stored in
173 /usr will be used, and any user database entries in /etc are
174 ignored. This functionality is useful for vendors to ship a
175 pre-built database on systems where local configuration is
176 unnecessary or unlikely.
178 * Calendar time specifications in .timer units now also
179 understand the strings "semi-annually", "quarterly" and
180 "minutely" as shortcuts (in addition to the preexisting
181 "anually", "hourly", ...).
183 * systemd-tmpfiles will now correctly create files in /dev
184 at boot which are marked for creation only at boot. It is
185 recommended to always create static device nodes with 'c!'
186 and 'b!', so that they are created only at boot and not
187 overwritten at runtime.
189 * When the watchdog logic is used for a service (WatchdogSec=)
190 and the watchdog timeout is hit the service will now be
191 terminated with SIGABRT (instead of just SIGTERM), in order
192 to make sure a proper coredump and backtrace is
193 generated. This ensures that hanging services will result in
194 similar coredump/backtrace behaviour as services that hit a
197 Contributions from: Andreas Henriksson, Andrei Borzenkov,
198 Angus Gibson, Ansgar Burchardt, Ben Wolsieffer, Brandon L.
199 Black, Christian Hesse, Cristian Rodríguez, Daniel Buch,
200 Daniele Medri, Daniel Mack, Dan Williams, Dave Reisner, David
201 Herrmann, David Sommerseth, David Strauss, Emil Renner
202 Berthing, Eric Cook, Evangelos Foutras, Filipe Brandenburger,
203 Gustavo Sverzut Barbieri, Hans de Goede, Harald Hoyer, Hristo
204 Venev, Hugo Grostabussiat, Ivan Shapovalov, Jan Janssen, Jan
205 Synacek, Jonathan Liu, Juho Son, Karel Zak, Kay Sievers, Klaus
206 Purer, Koen Kooi, Lennart Poettering, Lukas Nykryn, Lukasz
207 Skalski, Łukasz Stelmach, Mantas Mikulėnas, Marcel Holtmann,
208 Marius Tessmann, Marko Myllynen, Martin Pitt, Michael Biebl,
209 Michael Marineau, Michael Olbrich, Michael Scherer, Michal
210 Schmidt, Michal Sekletar, Miroslav Lichvar, Patrik Flykt,
211 Philippe De Swert, Piotr Drąg, Rahul Sundaram, Richard
212 Weinberger, Robert Milasan, Ronny Chevalier, Ruben Kerkhof,
213 Santiago Vila, Sergey Ptashnick, Simon McVittie, Sjoerd
214 Simons, Stefan Brüns, Steven Allen, Steven Noonan, Susant
215 Sahani, Sylvain Plantefève, Thomas Hindoe Paaboel Andersen,
216 Timofey Titovets, Tobias Hunger, Tom Gundersen, Torstein
217 Husebø, Umut Tezduyar Lindskog, WaLyong Cho, Zbigniew
220 -- Berlin, 2014-10-28
224 * timedated no longer reads NTP implementation unit names from
225 /usr/lib/systemd/ntp-units.d/*.list. Alternative NTP
226 implementations should add a
228 Conflicts=systemd-timesyncd.service
230 to their unit files to take over and replace systemd's NTP
231 default functionality.
233 * systemd-sysusers gained a new line type "r" for configuring
234 which UID/GID ranges to allocate system users/groups
235 from. Lines of type "u" may now add an additional column
236 that specifies the home directory for the system user to be
237 created. Also, systemd-sysusers may now optionally read user
238 information from STDIN instead of a file. This is useful for
239 invoking it from RPM preinst scriptlets that need to create
240 users before the first RPM file is installed since these
241 files might need to be owned by them. A new
242 %sysusers_create_inline RPM macro has been introduced to do
243 just that. systemd-sysusers now updates the shadow files as
244 well as the user/group databases, which should enhance
245 compatibility with certain tools like grpck.
247 * A number of bus APIs of PID 1 now optionally consult
248 PolicyKit to permit access for otherwise unprivileged
249 clients under certain conditions. Note that this currently
250 doesn't support interactive authentication yet, but this is
251 expected to be added eventually, too.
253 * /etc/machine-info now has new fields for configuring the
254 deployment environment of the machine, as well as the
255 location of the machine. hostnamectl has been updated with
256 new command to update these fields.
258 * systemd-timesyncd has been updated to automatically acquire
259 NTP server information from systemd-networkd, which might
260 have been discovered via DHCP.
262 * systemd-resolved now includes a caching DNS stub resolver
263 and a complete LLMNR name resolution implementation. A new
264 NSS module "nss-resolve" has been added which can be used
265 instead of glibc's own "nss-dns" to resolve hostnames via
266 systemd-resolved. Hostnames, addresses and arbitrary RRs may
267 be resolved via systemd-resolved D-Bus APIs. In contrast to
268 the glibc internal resolver systemd-resolved is aware of
269 multi-homed system, and keeps DNS server and caches separate
270 and per-interface. Queries are sent simultaneously on all
271 interfaces that have DNS servers configured, in order to
272 properly handle VPNs and local LANs which might resolve
273 separate sets of domain names. systemd-resolved may acquire
274 DNS server information from systemd-networkd automatically,
275 which in turn might have discovered them via DHCP. A tool
276 "systemd-resolve-host" has been added that may be used to
277 query the DNS logic in resolved. systemd-resolved implements
278 IDNA and automatically uses IDNA or UTF-8 encoding depending
279 on whether classic DNS or LLMNR is used as transport. In the
280 next releases we intend to add a DNSSEC and mDNS/DNS-SD
281 implementation to systemd-resolved.
283 * A new NSS module nss-mymachines has been added, that
284 automatically resolves the names of all local registered
285 containers to their respective IP addresses.
287 * A new client tool "networkctl" for systemd-networkd has been
288 added. It currently is entirely passive and will query
289 networking configuration from udev, rtnetlink and networkd,
290 and present it to the user in a very friendly
291 way. Eventually, we hope to extend it to become a full
292 control utility for networkd.
294 * .socket units gained a new DeferAcceptSec= setting that
295 controls the kernels' TCP_DEFER_ACCEPT sockopt for
296 TCP. Similar, support for controlling TCP keep-alive
297 settings has been added (KeepAliveTimeSec=,
298 KeepAliveIntervalSec=, KeepAliveProbes=). Also, support for
299 turning off Nagle's algorithm on TCP has been added
302 * logind learned a new session type "web", for use in projects
303 like Cockpit which register web clients as PAM sessions.
305 * timer units with at least one OnCalendar= setting will now
306 be started only after timer-sync.target has been
307 reached. This way they will not elapse before the system
308 clock has been corrected by a local NTP client or
309 similar. This is particular useful on RTC-less embedded
310 machines, that come up with an invalid system clock.
312 * systemd-nspawn's --network-veth= switch should now result in
313 stable MAC addresses for both the outer and the inner side
316 * systemd-nspawn gained a new --volatile= switch for running
317 container instances with /etc or /var unpopulated.
319 * The kdbus client code has been updated to use the new Linux
320 3.17 memfd subsystem instead of the old kdbus-specific one.
322 * systemd-networkd's DHCP client and server now support
323 FORCERENEW. There are also new configuration options to
324 configure the vendor client identifier and broadcast mode
327 * systemd will no longer inform the kernel about the current
328 timezone, as this is necessarily incorrect and racy as the
329 kernel has no understanding of DST and similar
330 concepts. This hence means FAT timestamps will be always
331 considered UTC, similar to what Android is already
332 doing. Also, when the RTC is configured to the local time
333 (rather than UTC) systemd will never synchronize back to it,
334 as this might confuse Windows at a later boot.
336 * systemd-analyze gained a new command "verify" for offline
337 validation of unit files.
339 * systemd-networkd gained support for a couple of additional
340 settings for bonding networking setups. Also, the metric for
341 statically configured routes may now be configured. For
342 network interfaces where this is appropriate the peer IP
343 address may now be configured.
345 * systemd-networkd's DHCP client will no longer request
346 broadcasting by default, as this tripped up some networks.
347 For hardware where broadcast is required the feature should
348 be switched back on using RequestBroadcast=yes.
350 * systemd-networkd will now set up IPv4LL addresses (when
351 enabled) even if DHCP is configured successfully.
353 * udev will now default to respect network device names given
354 by the kernel when the kernel indicates that these are
355 predictable. This behavior can be tweaked by changing
356 NamePolicy= in the relevant .link file.
358 * A new library systemd-terminal has been added that
359 implements full TTY stream parsing and rendering. This
360 library is supposed to be used later on for implementing a
361 full userspace VT subsystem, replacing the current kernel
364 * A new tool systemd-journal-upload has been added to push
365 journal data to a remote system running
366 systemd-journal-remote.
368 * journald will no longer forward all local data to another
369 running syslog daemon. This change has been made because
370 rsyslog (which appears to be the most commonly used syslog
371 implementation these days) no longer makes use of this, and
372 instead pulls the data out of the journal on its own. Since
373 forwarding the messages to a non-existent syslog server is
374 more expensive than we assumed we have now turned this
375 off. If you run a syslog server that is not a recent rsyslog
376 version, you have to turn this option on again
377 (ForwardToSyslog= in journald.conf).
379 * journald now optionally supports the LZ4 compressor for
380 larger journal fields. This compressor should perform much
381 better than XZ which was the previous default.
383 * machinectl now shows the IP addresses of local containers,
384 if it knows them, plus the interface name of the container.
386 * A new tool "systemd-escape" has been added that makes it
387 easy to escape strings to build unit names and similar.
389 * sd_notify() messages may now include a new ERRNO= field
390 which is parsed and collected by systemd and shown among the
391 "systemctl status" output for a service.
393 * A new component "systemd-firstboot" has been added that
394 queries the most basic systemd information (timezone,
395 hostname, root password) interactively on first
396 boot. Alternatively it may also be used to provision these
397 things offline on OS images installed into directories.
399 * The default sysctl.d/ snippets will now set
401 net.ipv4.conf.default.promote_secondaries=1
403 This has the benefit of no flushing secondary IP addresses
404 when primary addresses are removed.
406 Contributions from: Ansgar Burchardt, Bastien Nocera, Colin
407 Walters, Dan Dedrick, Daniel Buch, Daniel Korostil, Daniel
408 Mack, Dan Williams, Dave Reisner, David Herrmann, Denis
409 Kenzior, Eelco Dolstra, Eric Cook, Hannes Reinecke, Harald
410 Hoyer, Hong Shick Pak, Hui Wang, Jean-André Santoni, Jóhann
411 B. Guðmundsson, Jon Severinsson, Karel Zak, Kay Sievers, Kevin
412 Wells, Lennart Poettering, Lukas Nykryn, Mantas Mikulėnas,
413 Marc-Antoine Perennou, Martin Pitt, Michael Biebl, Michael
414 Marineau, Michael Olbrich, Michal Schmidt, Michal Sekletar,
415 Miguel Angel Ajo, Mike Gilbert, Olivier Brunel, Robert
416 Schiele, Ronny Chevalier, Simon McVittie, Sjoerd Simons, Stef
417 Walter, Steven Noonan, Susant Sahani, Tanu Kaskinen, Thomas
418 Blume, Thomas Hindoe Paaboel Andersen, Timofey Titovets,
419 Tobias Geerinckx-Rice, Tomasz Torcz, Tom Gundersen, Umut
420 Tezduyar Lindskog, Zbigniew Jędrzejewski-Szmek
422 -- Berlin, 2014-08-19
426 * A new tool systemd-sysusers has been added. This tool
427 creates system users and groups in /etc/passwd and
428 /etc/group, based on static declarative system user/group
429 definitions in /usr/lib/sysusers.d/. This is useful to
430 enable factory resets and volatile systems that boot up with
431 an empty /etc directory, and thus need system users and
432 groups created during early boot. systemd now also ships
433 with two default sysusers.d/ files for the most basic
434 users and groups systemd and the core operating system
437 * A new tmpfiles snippet has been added that rebuilds the
438 essential files in /etc on boot, should they be missing.
440 * A directive for ensuring automatic clean-up of
441 /var/cache/man/ has been removed from the default
442 configuration. This line should now be shipped by the man
443 implementation. The necessary change has been made to the
444 man-db implementation. Note that you need to update your man
445 implementation to one that ships this line, otherwise no
446 automatic clean-up of /var/cache/man will take place.
448 * A new condition ConditionNeedsUpdate= has been added that
449 may conditionalize services to only run when /etc or /var
450 are "older" than the vendor operating system resources in
451 /usr. This is useful for reconstructing or updating /etc
452 after an offline update of /usr or a factory reset, on the
453 next reboot. Services that want to run once after such an
454 update or reset should use this condition and order
455 themselves before the new systemd-update-done.service, which
456 will mark the two directories as fully updated. A number of
457 service files have been added making use of this, to rebuild
458 the udev hardware database, the journald message catalog and
459 dynamic loader cache (ldconfig). The systemd-sysusers tool
460 described above also makes use of this now. With this in
461 place it is now possible to start up a minimal operating
462 system with /etc empty cleanly. For more information on the
463 concepts involved see this recent blog story:
465 http://0pointer.de/blog/projects/stateless.html
467 * A new system group "input" has been introduced, and all
468 input device nodes get this group assigned. This is useful
469 for system-level software to get access to input devices. It
470 complements what is already done for "audio" and "video".
472 * systemd-networkd learnt minimal DHCPv4 server support in
473 addition to the existing DHCPv4 client support. It also
474 learnt DHCPv6 client and IPv6 Router Solicitation client
475 support. The DHCPv4 client gained support for static routes
476 passed in from the server. Note that the [DHCPv4] section
477 known in older systemd-networkd versions has been renamed to
478 [DHCP] and is now also used by the DHCPv6 client. Existing
479 .network files using settings of this section should be
480 updated, though compatibility is maintained. Optionally, the
481 client hostname may now be sent to the DHCP server.
483 * networkd gained support for vxlan virtual networks as well
484 as tun/tap and dummy devices.
486 * networkd gained support for automatic allocation of address
487 ranges for interfaces from a system-wide pool of
488 addresses. This is useful for dynamically managing a large
489 number of interfaces with a single network configuration
490 file. In particular this is useful to easily assign
491 appropriate IP addresses to the veth links of a large number
494 * RPM macros for processing sysusers, sysctl and binfmt
495 drop-in snippets at package installation time have been
498 * The /etc/os-release file should now be placed in
499 /usr/lib/os-release. The old location is automatically
500 created as symlink. /usr/lib is the more appropriate
501 location of this file, since it shall actually describe the
502 vendor operating system shipped in /usr, and not the
503 configuration stored in /etc.
505 * .mount units gained a new boolean SloppyOptions= setting
506 that maps to mount(8)'s -s option which enables permissive
507 parsing of unknown mount options.
509 * tmpfiles learnt a new "L+" directive which creates a symlink
510 but (unlike "L") deletes a pre-existing file first, should
511 it already exist and not already be the correct
512 symlink. Similar, "b+", "c+" and "p+" directives have been
513 added as well, which create block and character devices, as
514 well as fifos in the filesystem, possibly removing any
515 pre-existing files of different types.
517 * For tmpfiles' "L", "L+", "C" and "C+" directives the final
518 'argument' field (which so far specified the source to
519 symlink/copy the files from) is now optional. If omitted the
520 same file os copied from /usr/share/factory/ suffixed by the
521 full destination path. This is useful for populating /etc
522 with essential files, by copying them from vendor defaults
523 shipped in /usr/share/factory/etc.
525 * A new command "systemctl preset-all" has been added that
526 applies the service preset settings to all installed unit
527 files. A new switch --preset-mode= has been added that
528 controls whether only enable or only disable operations
531 * A new command "systemctl is-system-running" has been added
532 that allows checking the overall state of the system, for
533 example whether it is fully up and running.
535 * When the system boots up with an empty /etc, the equivalent
536 to "systemctl preset-all" is executed during early boot, to
537 make sure all default services are enabled after a factory
540 * systemd now contains a minimal preset file that enables the
541 most basic services systemd ships by default.
543 * Unit files' [Install] section gained a new DefaultInstance=
544 field for defining the default instance to create if a
545 template unit is enabled with no instance specified.
547 * A new passive target cryptsetup-pre.target has been added
548 that may be used by services that need to make they run and
549 finish before the first LUKS cryptographic device is set up.
551 * The /dev/loop-control and /dev/btrfs-control device nodes
552 are now owned by the "disk" group by default, opening up
553 access to this group.
555 * systemd-coredump will now automatically generate a
556 stack trace of all core dumps taking place on the system,
557 based on elfutils' libdw library. This stack trace is logged
560 * systemd-coredump may now optionally store coredumps directly
561 on disk (in /var/lib/systemd/coredump, possibly compressed),
562 instead of storing them unconditionally in the journal. This
563 mode is the new default. A new configuration file
564 /etc/systemd/coredump.conf has been added to configure this
565 and other parameters of systemd-coredump.
567 * coredumpctl gained a new "info" verb to show details about a
568 specific coredump. A new switch "-1" has also been added
569 that makes sure to only show information about the most
570 recent entry instead of all entries. Also, as the tool is
571 generally useful now the "systemd-" prefix of the binary
572 name has been removed. Distributions that want to maintain
573 compatibility with the old name should add a symlink from
574 the old name to the new name.
576 * journald's SplitMode= now defaults to "uid". This makes sure
577 that unprivileged users can access their own coredumps with
578 coredumpctl without restrictions.
580 * New kernel command line options "systemd.wants=" (for
581 pulling an additional unit during boot), "systemd.mask="
582 (for masking a specific unit for the boot), and
583 "systemd.debug-shell" (for enabling the debug shell on tty9)
584 have been added. This is implemented in the new generator
585 "systemd-debug-generator".
587 * systemd-nspawn will now by default filter a couple of
588 syscalls for containers, among them those required for
589 kernel module loading, direct x86 IO port access, swap
590 management, and kexec. Most importantly though
591 open_by_handle_at() is now prohibited for containers,
592 closing a hole similar to a recently discussed vulnerability
593 in docker regarding access to files on file hierarchies the
594 container should normally not have access to. Note that for
595 nspawn we generally make no security claims anyway (and
596 this is explicitly documented in the man page), so this is
597 just a fix for one of the most obvious problems.
599 * A new man page file-hierarchy(7) has been added that
600 contains a minimized, modernized version of the file system
601 layout systemd expects, similar in style to the FHS
602 specification or hier(5). A new tool systemd-path(1) has
603 been added to query many of these paths for the local
606 * Automatic time-based clean-up of $XDG_RUNTIME_DIR is no
607 longer done. Since the directory now has a per-user size
608 limit, and is cleaned on logout this appears unnecessary,
609 in particular since this now brings the lifecycle of this
610 directory closer in line with how IPC objects are handled.
612 * systemd.pc now exports a number of additional directories,
613 including $libdir (which is useful to identify the library
614 path for the primary architecture of the system), and a
615 couple of drop-in directories.
617 * udev's predictable network interface names now use the dev_port
618 sysfs attribute, introduced in linux 3.15 instead of dev_id to
619 distinguish between ports of the same PCI function. dev_id should
620 only be used for ports using the same HW address, hence the need
623 * machined has been updated to export the OS version of a
624 container (read from /etc/os-release and
625 /usr/lib/os-release) on the bus. This is now shown in
626 "machinectl status" for a machine.
628 * A new service setting RestartForceExitStatus= has been
629 added. If configured to a set of exit signals or process
630 return values, the service will be restarted when the main
631 daemon process exits with any of them, regardless of the
634 * systemctl's -H switch for connecting to remote systemd
635 machines has been extended so that it may be used to
636 directly connect to a specific container on the
637 host. "systemctl -H root@foobar:waldi" will now connect as
638 user "root" to host "foobar", and then proceed directly to
639 the container named "waldi". Note that currently you have to
640 authenticate as user "root" for this to work, as entering
641 containers is a privileged operation.
643 Contributions from: Andreas Henriksson, Benjamin Steinwender,
644 Carl Schaefer, Christian Hesse, Colin Ian King, Cristian
645 Rodríguez, Daniel Mack, Dave Reisner, David Herrmann, Eugene
646 Yakubovich, Filipe Brandenburger, Frederic Crozat, Hristo
647 Venev, Jan Engelhardt, Jonathan Boulle, Kay Sievers, Lennart
648 Poettering, Luke Shumaker, Mantas Mikulėnas, Marc-Antoine
649 Perennou, Marcel Holtmann, Michael Marineau, Michael Olbrich,
650 Michał Bartoszkiewicz, Michal Sekletar, Patrik Flykt, Ronan Le
651 Martret, Ronny Chevalier, Ruediger Oertel, Steven Noonan,
652 Susant Sahani, Thadeu Lima de Souza Cascardo, Thomas Hindoe
653 Paaboel Andersen, Tom Gundersen, Tom Hirst, Umut Tezduyar
654 Lindskog, Uoti Urpala, Zbigniew Jędrzejewski-Szmek
656 -- Berlin, 2014-07-03
660 * As an experimental feature, udev now tries to lock the
661 disk device node (flock(LOCK_SH|LOCK_NB)) while it
662 executes events for the disk or any of its partitions.
663 Applications like partitioning programs can lock the
664 disk device node (flock(LOCK_EX)) and claim temporary
665 device ownership that way; udev will entirely skip all event
666 handling for this disk and its partitions. If the disk
667 was opened for writing, the close will trigger a partition
668 table rescan in udev's "watch" facility, and if needed
669 synthesize "change" events for the disk and all its partitions.
670 This is now unconditionally enabled, and if it turns out to
671 cause major problems, we might turn it on only for specific
672 devices, or might need to disable it entirely. Device Mapper
673 devices are excluded from this logic.
675 * We temporarily dropped the "-l" switch for fsck invocations,
676 since they collide with the flock() logic above. util-linux
677 upstream has been changed already to avoid this conflict,
678 and we will readd "-l" as soon as util-linux with this
679 change has been released.
681 * The dependency on libattr has been removed. Since a long
682 time, the extended attribute calls have moved to glibc, and
683 libattr is thus unnecessary.
685 * Virtualization detection works without priviliges now. This
686 means the systemd-detect-virt binary no longer requires
687 CAP_SYS_PTRACE file capabilities, and our daemons can run
688 with fewer privileges.
690 * systemd-networkd now runs under its own "systemd-network"
691 user. It retains the CAP_NET_ADMIN, CAP_NET_BIND_SERVICE,
692 CAP_NET_BROADCAST, CAP_NET_RAW capabilities though, but
693 loses the ability to write to files owned by root this way.
695 * Similar, systemd-resolved now runs under its own
696 "systemd-resolve" user with no capabilities remaining.
698 * Similar, systemd-bus-proxyd now runs under its own
699 "systemd-bus-proxy" user with only CAP_IPC_OWNER remaining.
701 * systemd-networkd gained support for setting up "veth"
702 virtual ethernet devices for container connectivity, as well
703 as GRE and VTI tunnels.
705 * systemd-networkd will no longer automatically attempt to
706 manually load kernel modules necessary for certain tunnel
707 transports. Instead, it is assumed the kernel loads them
708 automatically when required. This only works correctly on
709 very new kernels. On older kernels, please consider adding
710 the kernel modules to /etc/modules-load.d/ as a work-around.
712 * The resolv.conf file systemd-resolved generates has been
713 moved to /run/systemd/resolve/. If you have a symlink from
714 /etc/resolv.conf, it might be necessary to correct it.
716 * Two new service settings, ProtectHome= and ProtectSystem=,
717 have been added. When enabled, they will make the user data
718 (such as /home) inaccessible or read-only and the system
719 (such as /usr) read-only, for specific services. This allows
720 very light-weight per-service sandboxing to avoid
721 modifications of user data or system files from
722 services. These two new switches have been enabled for all
723 of systemd's long-running services, where appropriate.
725 * Socket units gained new SocketUser= and SocketGroup=
726 settings to set the owner user and group of AF_UNIX sockets
727 and FIFOs in the file system.
729 * Socket units gained a new RemoveOnStop= setting. If enabled,
730 all FIFOS and sockets in the file system will be removed
731 when the specific socket unit is stopped.
733 * Socket units gained a new Symlinks= setting. It takes a list
734 of symlinks to create to file system sockets or FIFOs
735 created by the specific Unix sockets. This is useful to
736 manage symlinks to socket nodes with the same life-cycle as
739 * The /dev/log socket and /dev/initctl FIFO have been moved to
740 /run, and have been replaced by symlinks. This allows
741 connecting to these facilities even if PrivateDevices=yes is
742 used for a service (which makes /dev/log itself unavailable,
743 but /run is left). This also has the benefit of ensuring
744 that /dev only contains device nodes, directories and
745 symlinks, and nothing else.
747 * sd-daemon gained two new calls sd_pid_notify() and
748 sd_pid_notifyf(). They are similar to sd_notify() and
749 sd_notifyf(), but allow overriding of the source PID of
750 notification messages if permissions permit this. This is
751 useful to send notify messages on behalf of a different
752 process (for example, the parent process). The
753 systemd-notify tool has been updated to make use of this
754 when sending messages (so that notification messages now
755 originate from the shell script invoking systemd-notify and
756 not the systemd-notify process itself. This should minimize
757 a race where systemd fails to associate notification
758 messages to services when the originating process already
761 * A new "on-abnormal" setting for Restart= has been added. If
762 set, it will result in automatic restarts on all "abnormal"
763 reasons for a process to exit, which includes unclean
764 signals, core dumps, timeouts and watchdog timeouts, but
765 does not include clean and unclean exit codes or clean
766 signals. Restart=on-abnormal is an alternative for
767 Restart=on-failure for services that shall be able to
768 terminate and avoid restarts on certain errors, by
769 indicating so with an unclean exit code. Restart=on-failure
770 or Restart=on-abnormal is now the recommended setting for
771 all long-running services.
773 * If the InaccessibleDirectories= service setting points to a
774 mount point (or if there are any submounts contained within
775 it), it is now attempted to completely unmount it, to make
776 the file systems truly unavailable for the respective
779 * The ReadOnlyDirectories= service setting and
780 systemd-nspawn's --read-only parameter are now recursively
781 applied to all submounts, too.
783 * Mount units may now be created transiently via the bus APIs.
785 * The support for SysV and LSB init scripts has been removed
786 from the systemd daemon itself. Instead, it is now
787 implemented as a generator that creates native systemd units
788 from these scripts when needed. This enables us to remove a
789 substantial amount of legacy code from PID 1, following the
790 fact that many distributions only ship a very small number
791 of LSB/SysV init scripts nowadays.
793 * Priviliged Xen (dom0) domains are not considered
794 virtualization anymore by the virtualization detection
795 logic. After all, they generally have unrestricted access to
796 the hardware and usually are used to manage the unprivileged
799 * systemd-tmpfiles gained a new "C" line type, for copying
800 files or entire directories.
802 * systemd-tmpfiles "m" lines are now fully equivalent to "z"
803 lines. So far, they have been non-globbing versions of the
804 latter, and have thus been redundant. In future, it is
805 recommended to only use "z". "m" has hence been removed
806 from the documentation, even though it stays supported.
808 * A tmpfiles snippet to recreate the most basic structure in
809 /var has been added. This is enough to create the /var/run →
810 /run symlink and create a couple of structural
811 directories. This allows systems to boot up with an empty or
812 volatile /var. Of course, while with this change, the core OS
813 now is capable with dealing with a volatile /var, not all
814 user services are ready for it. However, we hope that sooner
815 or later, many service daemons will be changed upstream so
816 that they are able to automatically create their necessary
817 directories in /var at boot, should they be missing. This is
818 the first step to allow state-less systems that only require
819 the vendor image for /usr to boot.
821 * systemd-nspawn has gained a new --tmpfs= switch to mount an
822 empty tmpfs instance to a specific directory. This is
823 particularly useful for making use of the automatic
824 reconstruction of /var (see above), by passing --tmpfs=/var.
826 * Access modes specified in tmpfiles snippets may now be
827 prefixed with "~", which indicates that they shall be masked
828 by whether the existing file or directory is currently
829 writable, readable or executable at all. Also, if specified,
830 the sgid/suid/sticky bits will be masked for all
833 * A new passive target unit "network-pre.target" has been
834 added which is useful for services that shall run before any
835 network is configured, for example firewall scripts.
837 * The "floppy" group that previously owned the /dev/fd*
838 devices is no longer used. The "disk" group is now used
839 instead. Distributions should probably deprecate usage of
842 Contributions from: Camilo Aguilar, Christian Hesse, Colin Ian
843 King, Cristian Rodríguez, Daniel Buch, Dave Reisner, David
844 Strauss, Denis Tikhomirov, John, Jonathan Liu, Kay Sievers,
845 Lennart Poettering, Mantas Mikulėnas, Mark Eichin, Ronny
846 Chevalier, Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel
847 Andersen, Tom Gundersen, Umut Tezduyar Lindskog, Zbigniew
850 -- Berlin, 2014-06-11
854 * A new "systemd-timesyncd" daemon has been added for
855 synchronizing the system clock across the network. It
856 implements an SNTP client. In contrast to NTP
857 implementations such as chrony or the NTP reference server,
858 this only implements a client side, and does not bother with
859 the full NTP complexity, focusing only on querying time from
860 one remote server and synchronizing the local clock to
861 it. Unless you intend to serve NTP to networked clients or
862 want to connect to local hardware clocks, this simple NTP
863 client should be more than appropriate for most
864 installations. The daemon runs with minimal privileges, and
865 has been hooked up with networkd to only operate when
866 network connectivity is available. The daemon saves the
867 current clock to disk every time a new NTP sync has been
868 acquired, and uses this to possibly correct the system clock
869 early at bootup, in order to accommodate for systems that
870 lack an RTC such as the Raspberry Pi and embedded devices,
871 and to make sure that time monotonically progresses on these
872 systems, even if it is not always correct. To make use of
873 this daemon, a new system user and group "systemd-timesync"
874 needs to be created on installation of systemd.
876 * The queue "seqnum" interface of libudev has been disabled, as
877 it was generally incompatible with device namespacing as
878 sequence numbers of devices go "missing" if the devices are
879 part of a different namespace.
881 * "systemctl list-timers" and "systemctl list-sockets" gained
882 a --recursive switch for showing units of these types also
883 for all local containers, similar in style to the already
884 supported --recursive switch for "systemctl list-units".
886 * A new RebootArgument= setting has been added for service
887 units, which may be used to specify a kernel reboot argument
888 to use when triggering reboots with StartLimitAction=.
890 * A new FailureAction= setting has been added for service
891 units which may be used to specify an operation to trigger
892 when a service fails. This works similarly to
893 StartLimitAction=, but unlike it, controls what is done
894 immediately rather than only after several attempts to
895 restart the service in question.
897 * hostnamed got updated to also expose the kernel name,
898 release, and version on the bus. This is useful for
899 executing commands like hostnamectl with the -H switch.
900 systemd-analyze makes use of this to properly display
901 details when running non-locally.
903 * The bootchart tool can now show cgroup information in the
906 * The CFS CPU quota cgroup attribute is now exposed for
907 services. The new CPUQuota= switch has been added for this
908 which takes a percentage value. Setting this will have the
909 result that a service may never get more CPU time than the
910 specified percentage, even if the machine is otherwise idle.
912 * systemd-networkd learned IPIP and SIT tunnel support.
914 * LSB init scripts exposing a dependency on $network will now
915 get a dependency on network-online.target rather than simply
916 network.target. This should bring LSB handling closer to
917 what it was on SysV systems.
919 * A new fsck.repair= kernel option has been added to control
920 how fsck shall deal with unclean file systems at boot.
922 * The (.ini) configuration file parser will now silently
923 ignore sections whose name begins with "X-". This may be
924 used to maintain application-specific extension sections in unit
927 * machined gained a new API to query the IP addresses of
928 registered containers. "machinectl status" has been updated
929 to show these addresses in its output.
931 * A new call sd_uid_get_display() has been added to the
932 sd-login APIs for querying the "primary" session of a
933 user. The "primary" session of the user is elected from the
934 user's sessions and generally a graphical session is
935 preferred over a text one.
937 * A minimal systemd-resolved daemon has been added. It
938 currently simply acts as a companion to systemd-networkd and
939 manages resolv.conf based on per-interface DNS
940 configuration, possibly supplied via DHCP. In the long run
941 we hope to extend this into a local DNSSEC enabled DNS and
944 * The systemd-networkd-wait-online tool is now enabled by
945 default. It will delay network-online.target until a network
946 connection has been configured. The tool primarily integrates
947 with networkd, but will also make a best effort to make sense
948 of network configuration performed in some other way.
950 * Two new service options StartupCPUShares= and
951 StartupBlockIOWeight= have been added that work similarly to
952 CPUShares= and BlockIOWeight= however only apply during
953 system startup. This is useful to prioritize certain services
954 differently during bootup than during normal runtime.
956 * hostnamed has been changed to prefer the statically
957 configured hostname in /etc/hostname (unless set to
958 'localhost' or empty) over any dynamic one supplied by
959 dhcp. With this change, the rules for picking the hostname
960 match more closely the rules of other configuration settings
961 where the local administrator's configuration in /etc always
962 overrides any other settings.
964 Contributions fron: Ali H. Caliskan, Alison Chaiken, Bas van
965 den Berg, Brandon Philips, Cristian Rodríguez, Daniel Buch,
966 Dan Kilman, Dave Reisner, David Härdeman, David Herrmann,
967 David Strauss, Dimitris Spingos, Djalal Harouni, Eelco
968 Dolstra, Evan Nemerson, Florian Albrechtskirchinger, Greg
969 Kroah-Hartman, Harald Hoyer, Holger Hans Peter Freyther, Jan
970 Engelhardt, Jani Nikula, Jason St. John, Jeffrey Clark,
971 Jonathan Boulle, Kay Sievers, Lennart Poettering, Lukas
972 Nykryn, Lukasz Skalski, Łukasz Stelmach, Mantas Mikulėnas,
973 Marcel Holtmann, Martin Pitt, Matthew Monaco, Michael
974 Marineau, Michael Olbrich, Michal Sekletar, Mike Gilbert, Nis
975 Martensen, Patrik Flykt, Philip Lorenz, poma, Ray Strode,
976 Reyad Attiyat, Robert Milasan, Scott Thrasher, Stef Walter,
977 Steven Siloti, Susant Sahani, Tanu Kaskinen, Thomas Bächler,
978 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar
979 Lindskog, WaLyong Cho, Will Woods, Zbigniew
982 -- Beijing, 2014-05-28
986 * When restoring the screen brightness at boot, stay away from
987 the darkest setting or from the lowest 5% of the available
988 range, depending on which is the larger value of both. This
989 should effectively protect the user from rebooting into a
990 black screen, should the brightness have been set to minimum
993 * sd-login gained a new sd_machine_get_class() call to
994 determine the class ("vm" or "container") of a machine
995 registered with machined.
997 * sd-login gained new calls
998 sd_peer_get_{session,owner_uid,unit,user_unit,slice,machine_name}(),
999 to query the identity of the peer of a local AF_UNIX
1000 connection. They operate similarly to their sd_pid_get_xyz()
1003 * PID 1 will now maintain a system-wide system state engine
1004 with the states "starting", "running", "degraded",
1005 "maintenance", "stopping". These states are bound to system
1006 startup, normal runtime, runtime with at least one failed
1007 service, rescue/emergency mode and system shutdown. This
1008 state is shown in the "systemctl status" output when no unit
1009 name is passed. It is useful to determine system state, in
1010 particularly when doing so for many systems or containers at
1013 * A new command "list-machines" has been added to "systemctl"
1014 that lists all local OS containers and shows their system
1015 state (see above), if systemd runs inside of them.
1017 * systemctl gained a new "-r" switch to recursively enumerate
1018 units on all local containers, when used with the
1019 "list-unit" command (which is the default one that is
1020 executed when no parameters are specified).
1022 * The GPT automatic partition discovery logic will now honour
1023 two GPT partition flags: one may be set on a partition to
1024 cause it to be mounted read-only, and the other may be set
1025 on a partition to ignore it during automatic discovery.
1027 * Two new GPT type UUIDs have been added for automatic root
1028 partition discovery, for 32-bit and 64-bit ARM. This is not
1029 particularly useful for discovering the root directory on
1030 these architectures during bare-metal boots (since UEFI is
1031 not common there), but still very useful to allow booting of
1032 ARM disk images in nspawn with the -i option.
1034 * MAC addresses of interfaces created with nspawn's
1035 --network-interface= switch will now be generated from the
1036 machine name, and thus be stable between multiple invocations
1039 * logind will now automatically remove all IPC objects owned
1040 by a user if she or he fully logs out. This makes sure that
1041 users who are logged out cannot continue to consume IPC
1042 resources. This covers SysV memory, semaphores and message
1043 queues as well as POSIX shared memory and message
1044 queues. Traditionally, SysV and POSIX IPC had no life-cycle
1045 limits. With this functionality, that is corrected. This may
1046 be turned off by using the RemoveIPC= switch of logind.conf.
1048 * The systemd-machine-id-setup and tmpfiles tools gained a
1049 --root= switch to operate on a specific root directory,
1052 * journald can now forward logged messages to the TTYs of all
1053 logged in users ("wall"). This is the default for all
1054 emergency messages now.
1056 * A new tool systemd-journal-remote has been added to stream
1057 journal log messages across the network.
1059 * /sys/fs/cgroup/ is now mounted read-only after all cgroup
1060 controller trees are mounted into it. Note that the
1061 directories mounted beneath it are not read-only. This is a
1062 security measure and is particularly useful because glibc
1063 actually includes a search logic to pick any tmpfs it can
1064 find to implement shm_open() if /dev/shm is not available
1065 (which it might very well be in namespaced setups).
1067 * machinectl gained a new "poweroff" command to cleanly power
1068 down a local OS container.
1070 * The PrivateDevices= unit file setting will now also drop the
1071 CAP_MKNOD capability from the capability bound set, and
1072 imply DevicePolicy=closed.
1074 * PrivateDevices=, PrivateNetwork= and PrivateTmp= is now used
1075 comprehensively on all long-running systemd services where
1076 this is appropriate.
1078 * systemd-udevd will now run in a disassociated mount
1079 namespace. To mount directories from udev rules, make sure to
1080 pull in mount units via SYSTEMD_WANTS properties.
1082 * The kdbus support gained support for uploading policy into
1083 the kernel. sd-bus gained support for creating "monitoring"
1084 connections that can eavesdrop into all bus communication
1085 for debugging purposes.
1087 * Timestamps may now be specified in seconds since the UNIX
1088 epoch Jan 1st, 1970 by specifying "@" followed by the value
1091 * Native tcpwrap support in systemd has been removed. tcpwrap
1092 is old code, not really maintained anymore and has serious
1093 shortcomings, and better options such as firewalls
1094 exist. For setups that require tcpwrap usage, please
1095 consider invoking your socket-activated service via tcpd,
1096 like on traditional inetd.
1098 * A new system.conf configuration option
1099 DefaultTimerAccuracySec= has been added that controls the
1100 default AccuracySec= setting of .timer units.
1102 * Timer units gained a new WakeSystem= switch. If enabled,
1103 timers configured this way will cause the system to resume
1104 from system suspend (if the system supports that, which most
1107 * Timer units gained a new Persistent= switch. If enabled,
1108 timers configured this way will save to disk when they have
1109 been last triggered. This information is then used on next
1110 reboot to possible execute overdue timer events, that
1111 could not take place because the system was powered off.
1112 This enables simple anacron-like behaviour for timer units.
1114 * systemctl's "list-timers" will now also list the time a
1115 timer unit was last triggered in addition to the next time
1116 it will be triggered.
1118 * systemd-networkd will now assign predictable IPv4LL
1119 addresses to its local interfaces.
1121 Contributions from: Brandon Philips, Daniel Buch, Daniel Mack,
1122 Dave Reisner, David Herrmann, Gerd Hoffmann, Greg
1123 Kroah-Hartman, Hendrik Brueckner, Jason St. John, Josh
1124 Triplett, Kay Sievers, Lennart Poettering, Marc-Antoine
1125 Perennou, Michael Marineau, Michael Olbrich, Miklos Vajna,
1126 Patrik Flykt, poma, Sebastian Thorarensen, Thomas Bächler,
1127 Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom Gundersen,
1128 Umut Tezduyar Lindskog, Wieland Hoffmann, Zbigniew
1131 -- Berlin, 2014-03-25
1135 * A new unit file setting RestrictAddressFamilies= has been
1136 added to restrict which socket address families unit
1137 processes gain access to. This takes address family names
1138 like "AF_INET" or "AF_UNIX", and is useful to minimize the
1139 attack surface of services via exotic protocol stacks. This
1140 is built on seccomp system call filters.
1142 * Two new unit file settings RuntimeDirectory= and
1143 RuntimeDirectoryMode= have been added that may be used to
1144 manage a per-daemon runtime directories below /run. This is
1145 an alternative for setting up directory permissions with
1146 tmpfiles snippets, and has the advantage that the runtime
1147 directory's lifetime is bound to the daemon runtime and that
1148 the daemon starts up with an empty directory each time. This
1149 is particularly useful when writing services that drop
1150 privileges using the User= or Group= setting.
1152 * The DeviceAllow= unit setting now supports globbing for
1153 matching against device group names.
1155 * The systemd configuration file system.conf gained new
1156 settings DefaultCPUAccounting=, DefaultBlockIOAccounting=,
1157 DefaultMemoryAccounting= to globally turn on/off accounting
1158 for specific resources (cgroups) for all units. These
1159 settings may still be overridden individually in each unit
1162 * systemd-gpt-auto-generator is now able to discover /srv and
1163 root partitions in addition to /home and swap partitions. It
1164 also supports LUKS-encrypted partitions now. With this in
1165 place, automatic discovery of partitions to mount following
1166 the Discoverable Partitions Specification
1167 (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec)
1168 is now a lot more complete. This allows booting without
1169 /etc/fstab and without root= on the kernel command line on
1170 systems prepared appropriately.
1172 * systemd-nspawn gained a new --image= switch which allows
1173 booting up disk images and Linux installations on any block
1174 device that follow the Discoverable Partitions Specification
1175 (see above). This means that installations made with
1176 appropriately updated installers may now be started and
1177 deployed using container managers, completely
1178 unmodified. (We hope that libvirt-lxc will add support for
1179 this feature soon, too.)
1181 * systemd-nspawn gained a new --network-macvlan= setting to
1182 set up a private macvlan interface for the
1183 container. Similarly, systemd-networkd gained a new
1184 Kind=macvlan setting in .netdev files.
1186 * systemd-networkd now supports configuring local addresses
1189 * A new tool systemd-network-wait-online has been added to
1190 synchronously wait for network connectivity using
1193 * The sd-bus.h bus API gained a new sd_bus_track object for
1194 tracking the life-cycle of bus peers. Note that sd-bus.h is
1195 still not a public API though (unless you specify
1196 --enable-kdbus on the configure command line, which however
1197 voids your warranty and you get no API stability guarantee).
1199 * The $XDG_RUNTIME_DIR runtime directories for each user are
1200 now individual tmpfs instances, which has the benefit of
1201 introducing separate pools for each user, with individual
1202 size limits, and thus making sure that unprivileged clients
1203 can no longer negatively impact the system or other users by
1204 filling up their $XDG_RUNTIME_DIR. A new logind.conf setting
1205 RuntimeDirectorySize= has been introduced that allows
1206 controlling the default size limit for all users. It
1207 defaults to 10% of the available physical memory. This is no
1208 replacement for quotas on tmpfs though (which the kernel
1209 still does not support), as /dev/shm and /tmp are still
1210 shared resources used by both the system and unprivileged
1213 * logind will now automatically turn off automatic suspending
1214 on laptop lid close when more than one display is
1215 connected. This was previously expected to be implemented
1216 individually in desktop environments (such as GNOME),
1217 however has been added to logind now, in order to fix a
1218 boot-time race where a desktop environment might not have
1219 been started yet and thus not been able to take an inhibitor
1220 lock at the time where logind already suspends the system
1221 due to a closed lid.
1223 * logind will now wait at least 30s after each system
1224 suspend/resume cycle, and 3min after system boot before
1225 suspending the system due to a closed laptop lid. This
1226 should give USB docking stations and similar enough time to
1227 be probed and configured after system resume and boot in
1228 order to then act as suspend blocker.
1230 * systemd-run gained a new --property= setting which allows
1231 initialization of resource control properties (and others)
1232 for the created scope or service unit. Example: "systemd-run
1233 --property=BlockIOWeight=10 updatedb" may be used to run
1234 updatedb at a low block IO scheduling weight.
1236 * systemd-run's --uid=, --gid=, --setenv=, --setenv= switches
1237 now also work in --scope mode.
1239 * When systemd is compiled with kdbus support, basic support
1240 for enforced policies is now in place. (Note that enabling
1241 kdbus still voids your warranty and no API compatibility
1244 Contributions from: Andrey Borzenkov, Ansgar Burchardt, Armin
1245 K., Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
1246 Harald Hoyer, Henrik Grindal Bakken, Jasper St. Pierre, Kay
1247 Sievers, Kieran Clancy, Lennart Poettering, Lukas Nykryn,
1248 Mantas Mikulėnas, Marcel Holtmann, Mark Oteiza, Martin Pitt,
1249 Mike Gilbert, Peter Rajnoha, poma, Samuli Suominen, Stef
1250 Walter, Susant Sahani, Tero Roponen, Thomas Andersen, Thomas
1251 Bächler, Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom
1252 Gundersen, Umut Tezduyar Lindskog, Uoti Urpala, Zachary Cook,
1253 Zbigniew Jędrzejewski-Szmek
1255 -- Berlin, 2014-03-12
1259 * systemd will now relabel /dev after loading the SMACK policy
1260 according to SMACK rules.
1262 * A new unit file option AppArmorProfile= has been added to
1263 set the AppArmor profile for the processes of a unit.
1265 * A new condition check ConditionArchitecture= has been added
1266 to conditionalize units based on the system architecture, as
1267 reported by uname()'s "machine" field.
1269 * systemd-networkd now supports matching on the system
1270 virtualization, architecture, kernel command line, host name
1273 * logind is now a lot more aggressive when suspending the
1274 machine due to a closed laptop lid. Instead of acting only
1275 on the lid close action, it will continuously watch the lid
1276 status and act on it. This is useful for laptops where the
1277 power button is on the outside of the chassis so that it can
1278 be reached without opening the lid (such as the Lenovo
1279 Yoga). On those machines, logind will now immediately
1280 re-suspend the machine if the power button has been
1281 accidentally pressed while the laptop was suspended and in a
1282 backpack or similar.
1284 * logind will now watch SW_DOCK switches and inhibit reaction
1285 to the lid switch if it is pressed. This means that logind
1286 will not suspend the machine anymore if the lid is closed
1287 and the system is docked, if the laptop supports SW_DOCK
1288 notifications via the input layer. Note that ACPI docking
1289 stations do not generate this currently. Also note that this
1290 logic is usually not fully sufficient and Desktop
1291 Environments should take a lid switch inhibitor lock when an
1292 external display is connected, as systemd will not watch
1295 * nspawn will now make use of the devices cgroup controller by
1296 default, and only permit creation of and access to the usual
1297 API device nodes like /dev/null or /dev/random, as well as
1298 access to (but not creation of) the pty devices.
1300 * We will now ship a default .network file for
1301 systemd-networkd that automatically configures DHCP for
1302 network interfaces created by nspawn's --network-veth or
1303 --network-bridge= switches.
1305 * systemd will now understand the usual M, K, G, T suffixes
1306 according to SI conventions (i.e. to the base 1000) when
1307 referring to throughput and hardware metrics. It will stay
1308 with IEC conventions (i.e. to the base 1024) for software
1309 metrics, according to what is customary according to
1310 Wikipedia. We explicitly document which base applies for
1311 each configuration option.
1313 * The DeviceAllow= setting in unit files now supports a syntax
1314 to whitelist an entire group of devices node majors at once,
1315 based on the /proc/devices listing. For example, with the
1316 string "char-pts", it is now possible to whitelist all
1317 current and future pseudo-TTYs at once.
1319 * sd-event learned a new "post" event source. Event sources of
1320 this type are triggered by the dispatching of any event
1321 source of a type that is not "post". This is useful for
1322 implementing clean-up and check event sources that are
1323 triggered by other work being done in the program.
1325 * systemd-networkd is no longer statically enabled, but uses
1326 the usual [Install] sections so that it can be
1327 enabled/disabled using systemctl. It still is enabled by
1330 * When creating a veth interface pair with systemd-nspawn, the
1331 host side will now be prefixed with "vb-" if
1332 --network-bridge= is used, and with "ve-" if --network-veth
1333 is used. This way, it is easy to distinguish these cases on
1334 the host, for example to apply different configuration to
1335 them with systemd-networkd.
1337 * The compatibility libraries for libsystemd-journal.so,
1338 libsystem-id128.so, libsystemd-login.so and
1339 libsystemd-daemon.so do not make use of IFUNC
1340 anymore. Instead, we now build libsystemd.so multiple times
1341 under these alternative names. This means that the footprint
1342 is drastically increased, but given that these are
1343 transitional compatibility libraries, this should not matter
1344 much. This change has been made necessary to support the ARM
1345 platform for these compatibility libraries, as the ARM
1346 toolchain is not really at the same level as the toolchain
1347 for other architectures like x86 and does not support
1348 IFUNC. Please make sure to use --enable-compat-libs only
1349 during a transitional period!
1351 Contributions from: Andreas Fuchs, Armin K., Colin Walters,
1352 Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
1353 Holger Schurig, Jason A. Donenfeld, Jason St. John, Jasper
1354 St. Pierre, Kay Sievers, Lennart Poettering, Łukasz Stelmach,
1355 Marcel Holtmann, Michael Scherer, Michal Sekletar, Mike
1356 Gilbert, Samuli Suominen, Thomas Bächler, Thomas Hindoe
1357 Paaboel Andersen, Tom Gundersen, Umut Tezduyar Lindskog,
1358 Zbigniew Jędrzejewski-Szmek
1360 -- Berlin, 2014-02-24
1364 * A new component "systemd-networkd" has been added that can
1365 be used to configure local network interfaces statically or
1366 via DHCP. It is capable of bringing up bridges, VLANs, and
1367 bonding. Currently, no hook-ups for interactive network
1368 configuration are provided. Use this for your initrd,
1369 container, embedded, or server setup if you need a simple,
1370 yet powerful, network configuration solution. This
1371 configuration subsystem is quite nifty, as it allows wildcard
1372 hotplug matching in interfaces. For example, with a single
1373 configuration snippet, you can configure that all Ethernet
1374 interfaces showing up are automatically added to a bridge,
1375 or similar. It supports link-sensing and more.
1377 * A new tool "systemd-socket-proxyd" has been added which can
1378 act as a bidirectional proxy for TCP sockets. This is
1379 useful for adding socket activation support to services that
1380 do not actually support socket activation, including virtual
1381 machines and the like.
1383 * Add a new tool to save/restore rfkill state on
1386 * Save/restore state of keyboard backlights in addition to
1387 display backlights on shutdown/boot.
1389 * udev learned a new SECLABEL{} construct to label device
1390 nodes with a specific security label when they appear. For
1391 now, only SECLABEL{selinux} is supported, but the syntax is
1392 prepared for additional security frameworks.
1394 * udev gained a new scheme to configure link-level attributes
1395 from files in /etc/systemd/network/*.link. These files can
1396 match against MAC address, device path, driver name and type,
1397 and will apply attributes like the naming policy, link speed,
1398 MTU, duplex settings, Wake-on-LAN settings, MAC address, MAC
1399 address assignment policy (randomized, ...).
1401 * The configuration of network interface naming rules for
1402 "permanent interface names" has changed: a new NamePolicy=
1403 setting in the [Link] section of .link files determines the
1404 priority of possible naming schemes (onboard, slot, mac,
1405 path). The default value of this setting is determined by
1406 /usr/lib/net/links/99-default.link. Old
1407 80-net-name-slot.rules udev configuration file has been
1408 removed, so local configuration overriding this file should
1409 be adapated to override 99-default.link instead.
1411 * When the User= switch is used in a unit file, also
1412 initialize $SHELL= based on the user database entry.
1414 * systemd no longer depends on libdbus. All communication is
1415 now done with sd-bus, systemd's low-level bus library
1418 * kdbus support has been added to PID 1 itself. When kdbus is
1419 enabled, this causes PID 1 to set up the system bus and
1420 enable support for a new ".busname" unit type that
1421 encapsulates bus name activation on kdbus. It works a little
1422 bit like ".socket" units, except for bus names. A new
1423 generator has been added that converts classic dbus1 service
1424 activation files automatically into native systemd .busname
1427 * sd-bus: add a light-weight vtable implementation that allows
1428 defining objects on the bus with a simple static const
1429 vtable array of its methods, signals and properties.
1431 * systemd will not generate or install static dbus
1432 introspection data anymore to /usr/share/dbus-1/interfaces,
1433 as the precise format of these files is unclear, and
1434 nothing makes use of it.
1436 * A proxy daemon is now provided to proxy clients connecting
1437 via classic D-Bus AF_UNIX sockets to kdbus, to provide full
1438 compatibility with classic D-Bus.
1440 * A bus driver implementation has been added that supports the
1441 classic D-Bus bus driver calls on kdbus, also for
1442 compatibility purposes.
1444 * A new API "sd-event.h" has been added that implements a
1445 minimal event loop API built around epoll. It provides a
1446 couple of features that direct epoll usage is lacking:
1447 prioritization of events, scales to large numbers of timer
1448 events, per-event timer slack (accuracy), system-wide
1449 coalescing of timer events, exit handlers, watchdog
1450 supervision support using systemd's sd_notify() API, child
1453 * A new API "sd-rntl.h" has been added that provides an API
1454 around the route netlink interface of the kernel, similar in
1455 style to "sd-bus.h".
1457 * A new API "sd-dhcp-client.h" has been added that provides a
1458 small DHCPv4 client-side implementation. This is used by
1461 * There is a new kernel command line option
1462 "systemd.restore_state=0|1". When set to "0", none of the
1463 systemd tools will restore saved runtime state to hardware
1464 devices. More specifically, the rfkill and backlight states
1467 * The FsckPassNo= compatibility option in mount/service units
1468 has been removed. The fstab generator will now add the
1469 necessary dependencies automatically, and does not require
1470 PID1's support for that anymore.
1472 * journalctl gained a new switch, --list-boots, that lists
1473 recent boots with their times and boot IDs.
1475 * The various tools like systemctl, loginctl, timedatectl,
1476 busctl, systemd-run, ... have gained a new switch "-M" to
1477 connect to a specific, local OS container (as direct
1478 connection, without requiring SSH). This works on any
1479 container that is registered with machined, such as those
1480 created by libvirt-lxc or nspawn.
1482 * systemd-run and systemd-analyze also gained support for "-H"
1483 to connect to remote hosts via SSH. This is particularly
1484 useful for systemd-run because it enables queuing of jobs
1485 onto remote systems.
1487 * machinectl gained a new command "login" to open a getty
1488 login in any local container. This works with any container
1489 that is registered with machined (such as those created by
1490 libvirt-lxc or nspawn), and which runs systemd inside.
1492 * machinectl gained a new "reboot" command that may be used to
1493 trigger a reboot on a specific container that is registered
1494 with machined. This works on any container that runs an init
1495 system of some kind.
1497 * systemctl gained a new "list-timers" command to print a nice
1498 listing of installed timer units with the times they elapse
1501 * Alternative reboot() parameters may now be specified on the
1502 "systemctl reboot" command line and are passed to the
1503 reboot() system call.
1505 * systemctl gained a new --job-mode= switch to configure the
1506 mode to queue a job with. This is a more generic version of
1507 --fail, --irreversible, and --ignore-dependencies, which are
1508 still available but not advertised anymore.
1510 * /etc/systemd/system.conf gained new settings to configure
1511 various default timeouts of units, as well as the default
1512 start limit interval and burst. These may still be overridden
1515 * PID1 will now export on the bus profile data of the security
1516 policy upload process (such as the SELinux policy upload to
1519 * journald: when forwarding logs to the console, include
1520 timestamps (following the setting in
1521 /sys/module/printk/parameters/time).
1523 * OnCalendar= in timer units now understands the special
1524 strings "yearly" and "annually". (Both are equivalent)
1526 * The accuracy of timer units is now configurable with the new
1527 AccuracySec= setting. It defaults to 1min.
1529 * A new dependency type JoinsNamespaceOf= has been added that
1530 allows running two services within the same /tmp and network
1531 namespace, if PrivateNetwork= or PrivateTmp= are used.
1533 * A new command "cat" has been added to systemctl. It outputs
1534 the original unit file of a unit, and concatenates the
1535 contents of additional "drop-in" unit file snippets, so that
1536 the full configuration is shown.
1538 * systemctl now supports globbing on the various "list-xyz"
1539 commands, like "list-units" or "list-sockets", as well as on
1540 those commands which take multiple unit names.
1542 * journalctl's --unit= switch gained support for globbing.
1544 * All systemd daemons now make use of the watchdog logic so
1545 that systemd automatically notices when they hang.
1547 * If the $container_ttys environment variable is set,
1548 getty-generator will automatically spawn a getty for each
1549 listed tty. This is useful for container managers to request
1550 login gettys to be spawned on as many ttys as needed.
1552 * %h, %s, %U specifier support is not available anymore when
1553 used in unit files for PID 1. This is because NSS calls are
1554 not safe from PID 1. They stay available for --user
1555 instances of systemd, and as special case for the root user.
1557 * loginctl gained a new "--no-legend" switch to turn off output
1560 * The "sd-login.h" API gained three new calls:
1561 sd_session_is_remote(), sd_session_get_remote_user(),
1562 sd_session_get_remote_host() to query information about
1565 * The udev hardware database now also carries vendor/product
1566 information of SDIO devices.
1568 * The "sd-daemon.h" API gained a new sd_watchdog_enabled() to
1569 determine whether watchdog notifications are requested by
1572 * Socket-activated per-connection services now include a
1573 short description of the connection parameters in the
1576 * tmpfiles gained a new "--boot" option. When this is not used,
1577 only lines where the command character is not suffixed with
1578 "!" are executed. When this option is specified, those
1579 options are executed too. This partitions tmpfiles
1580 directives into those that can be safely executed at any
1581 time, and those which should be run only at boot (for
1582 example, a line that creates /run/nologin).
1584 * A new API "sd-resolve.h" has been added which provides a simple
1585 asynchronous wrapper around glibc NSS host name resolution
1586 calls, such as getaddrinfo(). In contrast to glibc's
1587 getaddrinfo_a(), it does not use signals. In contrast to most
1588 other asynchronous name resolution libraries, this one does
1589 not reimplement DNS, but reuses NSS, so that alternate
1590 host name resolution systems continue to work, such as mDNS,
1591 LDAP, etc. This API is based on libasyncns, but it has been
1592 cleaned up for inclusion in systemd.
1594 * The APIs "sd-journal.h", "sd-login.h", "sd-id128.h",
1595 "sd-daemon.h" are no longer found in individual libraries
1596 libsystemd-journal.so, libsystemd-login.so,
1597 libsystemd-id128.so, libsystemd-daemon.so. Instead, we have
1598 merged them into a single library, libsystemd.so, which
1599 provides all symbols. The reason for this is cyclic
1600 dependencies, as these libraries tend to use each other's
1601 symbols. So far, we have managed to workaround that by linking
1602 a copy of a good part of our code into each of these
1603 libraries again and again, which, however, makes certain
1604 things hard to do, like sharing static variables. Also, it
1605 substantially increases footprint. With this change, there
1606 is only one library for the basic APIs systemd
1607 provides. Also, "sd-bus.h", "sd-memfd.h", "sd-event.h",
1608 "sd-rtnl.h", "sd-resolve.h", "sd-utf8.h" are found in this
1609 library as well, however are subject to the --enable-kdbus
1610 switch (see below). Note that "sd-dhcp-client.h" is not part
1611 of this library (this is because it only consumes, never
1612 provides, services of/to other APIs). To make the transition
1613 easy from the separate libraries to the unified one, we
1614 provide the --enable-compat-libs compile-time switch which
1615 will generate stub libraries that are compatible with the
1616 old ones but redirect all calls to the new one.
1618 * All of the kdbus logic and the new APIs "sd-bus.h",
1619 "sd-memfd.h", "sd-event.h", "sd-rtnl.h", "sd-resolve.h",
1620 and "sd-utf8.h" are compile-time optional via the
1621 "--enable-kdbus" switch, and they are not compiled in by
1622 default. To make use of kdbus, you have to explicitly enable
1623 the switch. Note however, that neither the kernel nor the
1624 userspace API for all of this is considered stable yet. We
1625 want to maintain the freedom to still change the APIs for
1626 now. By specifying this build-time switch, you acknowledge
1627 that you are aware of the instability of the current
1630 * Also, note that while kdbus is pretty much complete,
1631 it lacks one thing: proper policy support. This means you
1632 can build a fully working system with all features; however,
1633 it will be highly insecure. Policy support will be added in
1634 one of the next releases, at the same time that we will
1635 declare the APIs stable.
1637 * When the kernel command line argument "kdbus" is specified,
1638 systemd will automatically load the kdbus.ko kernel module. At
1639 this stage of development, it is only useful for testing kdbus
1640 and should not be used in production. Note: if "--enable-kdbus"
1641 is specified, and the kdbus.ko kernel module is available, and
1642 "kdbus" is added to the kernel command line, the entire system
1643 runs with kdbus instead of dbus-daemon, with the above mentioned
1644 problem of missing the system policy enforcement. Also a future
1645 version of kdbus.ko or a newer systemd will not be compatible with
1646 each other, and will unlikely be able to boot the machine if only
1647 one of them is updated.
1649 * systemctl gained a new "import-environment" command which
1650 uploads the caller's environment (or parts thereof) into the
1651 service manager so that it is inherited by services started
1652 by the manager. This is useful to upload variables like
1653 $DISPLAY into the user service manager.
1655 * A new PrivateDevices= switch has been added to service units
1656 which allows running a service with a namespaced /dev
1657 directory that does not contain any device nodes for
1658 physical devices. More specifically, it only includes devices
1659 such as /dev/null, /dev/urandom, and /dev/zero which are API
1662 * logind has been extended to support behaviour like VT
1663 switching on seats that do not support a VT. This makes
1664 multi-session available on seats that are not the first seat
1665 (seat0), and on systems where kernel support for VTs has
1666 been disabled at compile-time.
1668 * If a process holds a delay lock for system sleep or shutdown
1669 and fails to release it in time, we will now log its
1670 identity. This makes it easier to identify processes that
1671 cause slow suspends or power-offs.
1673 * When parsing /etc/crypttab, support for a new key-slot=
1674 option as supported by Debian is added. It allows indicating
1675 which LUKS slot to use on disk, speeding up key loading.
1677 * The sd_journald_sendv() API call has been checked and
1678 officially declared to be async-signal-safe so that it may
1679 be invoked from signal handlers for logging purposes.
1681 * Boot-time status output is now enabled automatically after a
1682 short timeout if boot does not progress, in order to give
1683 the user an indication what she or he is waiting for.
1685 * The boot-time output has been improved to show how much time
1686 remains until jobs expire.
1688 * The KillMode= switch in service units gained a new possible
1689 value "mixed". If set, and the unit is shut down, then the
1690 initial SIGTERM signal is sent only to the main daemon
1691 process, while the following SIGKILL signal is sent to
1692 all remaining processes of the service.
1694 * When a scope unit is registered, a new property "Controller"
1695 may be set. If set to a valid bus name, systemd will send a
1696 RequestStop() signal to this name when it would like to shut
1697 down the scope. This may be used to hook manager logic into
1698 the shutdown logic of scope units. Also, scope units may now
1699 be put in a special "abandoned" state, in which case the
1700 manager process which created them takes no further
1701 responsibilities for it.
1703 * When reading unit files, systemd will now verify
1704 the access mode of these files, and warn about certain
1705 suspicious combinations. This has been added to make it
1706 easier to track down packaging bugs where unit files are
1707 marked executable or world-writable.
1709 * systemd-nspawn gained a new "--setenv=" switch to set
1710 container-wide environment variables. The similar option in
1711 systemd-activate was renamed from "--environment=" to
1712 "--setenv=" for consistency.
1714 * systemd-nspawn has been updated to create a new kdbus domain
1715 for each container that is invoked, thus allowing each
1716 container to have its own set of system and user buses,
1717 independent of the host.
1719 * systemd-nspawn gained a new --drop-capability= switch to run
1720 the container with less capabilities than the default. Both
1721 --drop-capability= and --capability= now take the special
1722 string "all" for dropping or keeping all capabilities.
1724 * systemd-nspawn gained new switches for executing containers
1725 with specific SELinux labels set.
1727 * systemd-nspawn gained a new --quiet switch to not generate
1728 any additional output but the container's own console
1731 * systemd-nspawn gained a new --share-system switch to run a
1732 container without PID namespacing enabled.
1734 * systemd-nspawn gained a new --register= switch to control
1735 whether the container is registered with systemd-machined or
1736 not. This is useful for containers that do not run full
1737 OS images, but only specific apps.
1739 * systemd-nspawn gained a new --keep-unit which may be used
1740 when invoked as the only program from a service unit, and
1741 results in registration of the unit service itself in
1742 systemd-machined, instead of a newly opened scope unit.
1744 * systemd-nspawn gained a new --network-interface= switch for
1745 moving arbitrary interfaces to the container. The new
1746 --network-veth switch creates a virtual Ethernet connection
1747 between host and container. The new --network-bridge=
1748 switch then allows assigning the host side of this virtual
1749 Ethernet connection to a bridge device.
1751 * systemd-nspawn gained a new --personality= switch for
1752 setting the kernel personality for the container. This is
1753 useful when running a 32-bit container on a 64-bit host. A
1754 similar option Personality= is now also available for service
1757 * logind will now also track a "Desktop" identifier for each
1758 session which encodes the desktop environment of it. This is
1759 useful for desktop environments that want to identify
1760 multiple running sessions of itself easily.
1762 * A new SELinuxContext= setting for service units has been
1763 added that allows setting a specific SELinux execution
1764 context for a service.
1766 * Most systemd client tools will now honour $SYSTEMD_LESS for
1767 settings of the "less" pager. By default, these tools will
1768 override $LESS to allow certain operations to work, such as
1769 jump-to-the-end. With $SYSTEMD_LESS, it is possible to
1770 influence this logic.
1772 * systemd's "seccomp" hook-up has been changed to make use of
1773 the libseccomp library instead of using its own
1774 implementation. This has benefits for portability among
1777 * For usage together with SystemCallFilter=, a new
1778 SystemCallErrorNumber= setting has been introduced that
1779 allows configuration of a system error number to be returned
1780 on filtered system calls, instead of immediately killing the
1781 process. Also, SystemCallArchitectures= has been added to
1782 limit access to system calls of a particular architecture
1783 (in order to turn off support for unused secondary
1784 architectures). There is also a global
1785 SystemCallArchitectures= setting in system.conf now to turn
1786 off support for non-native system calls system-wide.
1788 * systemd requires a kernel with a working name_to_handle_at(),
1789 please see the kernel config requirements in the README file.
1791 Contributions from: Adam Williamson, Alex Jia, Anatol Pomozov,
1792 Ansgar Burchardt, AppleBloom, Auke Kok, Bastien Nocera,
1793 Chengwei Yang, Christian Seiler, Colin Guthrie, Colin Walters,
1794 Cristian Rodríguez, Daniel Buch, Daniele Medri, Daniel J
1795 Walsh, Daniel Mack, Dan McGee, Dave Reisner, David Coppa,
1796 David Herrmann, David Strauss, Djalal Harouni, Dmitry Pisklov,
1797 Elia Pinto, Florian Weimer, George McCollister, Goffredo
1798 Baroncelli, Greg Kroah-Hartman, Hendrik Brueckner, Igor
1799 Zhbanov, Jan Engelhardt, Jan Janssen, Jason A. Donenfeld,
1800 Jason St. John, Jasper St. Pierre, Jóhann B. Guðmundsson, Jose
1801 Ignacio Naranjo, Karel Zak, Kay Sievers, Kristian Høgsberg,
1802 Lennart Poettering, Lubomir Rintel, Lukas Nykryn, Lukasz
1803 Skalski, Łukasz Stelmach, Luke Shumaker, Mantas Mikulėnas,
1804 Marc-Antoine Perennou, Marcel Holtmann, Marcos Felipe Rasia de
1805 Mello, Marko Myllynen, Martin Pitt, Matthew Monaco, Michael
1806 Marineau, Michael Scherer, Michał Górny, Michal Sekletar,
1807 Michele Curti, Oleksii Shevchuk, Olivier Brunel, Patrik Flykt,
1808 Pavel Holica, Raudi, Richard Marko, Ronny Chevalier, Sébastien
1809 Luttringer, Sergey Ptashnick, Shawn Landden, Simon Peeters,
1810 Stefan Beller, Susant Sahani, Sylvain Plantefeve, Sylvia Else,
1811 Tero Roponen, Thomas Bächler, Thomas Hindoe Paaboel Andersen,
1812 Tom Gundersen, Umut Tezduyar Lindskog, Unai Uribarri, Václav
1813 Pavlín, Vincent Batts, WaLyong Cho, William Giokas, Yang
1814 Zhiyong, Yin Kangkai, Yuxuan Shui, Zbigniew Jędrzejewski-Szmek
1816 -- Berlin, 2014-02-20
1820 * logind has gained support for facilitating privileged input
1821 and drm device access for unprivileged clients. This work is
1822 useful to allow Wayland display servers (and similar
1823 programs, such as kmscon) to run under the user's ID and
1824 access input and drm devices which are normally
1825 protected. When this is used (and the kernel is new enough)
1826 logind will "mute" IO on the file descriptors passed to
1827 Wayland as long as it is in the background and "unmute" it
1828 if it returns into the foreground. This allows secure
1829 session switching without allowing background sessions to
1830 eavesdrop on input and display data. This also introduces
1831 session switching support if VT support is turned off in the
1832 kernel, and on seats that are not seat0.
1834 * A new kernel command line option luks.options= is understood
1835 now which allows specifying LUKS options for usage for LUKS
1836 encrypted partitions specified with luks.uuid=.
1838 * tmpfiles.d(5) snippets may now use specifier expansion in
1839 path names. More specifically %m, %b, %H, %v, are now
1840 replaced by the local machine id, boot id, hostname, and
1841 kernel version number.
1843 * A new tmpfiles.d(5) command "m" has been introduced which
1844 may be used to change the owner/group/access mode of a file
1845 or directory if it exists, but do nothing if it does not.
1847 * This release removes high-level support for the
1848 MemorySoftLimit= cgroup setting. The underlying kernel
1849 cgroup attribute memory.soft_limit= is currently badly
1850 designed and likely to be removed from the kernel API in its
1851 current form, hence we should not expose it for now.
1853 * The memory.use_hierarchy cgroup attribute is now enabled for
1854 all cgroups systemd creates in the memory cgroup
1855 hierarchy. This option is likely to be come the built-in
1856 default in the kernel anyway, and the non-hierarchial mode
1857 never made much sense in the intrinsically hierarchial
1860 * A new field _SYSTEMD_SLICE= is logged along with all journal
1861 messages containing the slice a message was generated
1862 from. This is useful to allow easy per-customer filtering of
1863 logs among other things.
1865 * systemd-journald will no longer adjust the group of journal
1866 files it creates to the "systemd-journal" group. Instead we
1867 rely on the journal directory to be owned by the
1868 "systemd-journal" group, and its setgid bit set, so that the
1869 kernel file system layer will automatically enforce that
1870 journal files inherit this group assignment. The reason for
1871 this change is that we cannot allow NSS look-ups from
1872 journald which would be necessary to resolve
1873 "systemd-journal" to a numeric GID, because this might
1874 create deadlocks if NSS involves synchronous queries to
1875 other daemons (such as nscd, or sssd) which in turn are
1876 logging clients of journald and might block on it, which
1877 would then dead lock. A tmpfiles.d(5) snippet included in
1878 systemd will make sure the setgid bit and group are
1879 properly set on the journal directory if it exists on every
1880 boot. However, we recommend adjusting it manually after
1881 upgrades too (or from RPM scriptlets), so that the change is
1882 not delayed until next reboot.
1884 * Backlight and random seed files in /var/lib/ have moved into
1885 the /var/lib/systemd/ directory, in order to centralize all
1886 systemd generated files in one directory.
1888 * Boot time performance measurements (as displayed by
1889 "systemd-analyze" for example) will now read ACPI 5.0 FPDT
1890 performance information if that's available to determine how
1891 much time BIOS and boot loader initialization required. With
1892 a sufficiently new BIOS you hence no longer need to boot
1893 with Gummiboot to get access to such information.
1895 Contributions from: Andrey Borzenkov, Chen Jie, Colin Walters,
1896 Cristian Rodríguez, Dave Reisner, David Herrmann, David
1897 Mackey, David Strauss, Eelco Dolstra, Evan Callicoat, Gao
1898 feng, Harald Hoyer, Jimmie Tauriainen, Kay Sievers, Lennart
1899 Poettering, Lukas Nykryn, Mantas Mikulėnas, Martin Pitt,
1900 Michael Scherer, Michał Górny, Mike Gilbert, Patrick McCarty,
1901 Sebastian Ott, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
1903 -- Berlin, 2013-10-02
1907 * The Restart= option for services now understands a new
1908 on-watchdog setting, which will restart the service
1909 automatically if the service stops sending out watchdog keep
1910 alive messages (as configured with WatchdogSec=).
1912 * The getty generator (which is responsible for bringing up a
1913 getty on configured serial consoles) will no longer only
1914 start a getty on the primary kernel console but on all
1915 others, too. This makes the order in which console= is
1916 specified on the kernel command line less important.
1918 * libsystemd-logind gained a new sd_session_get_vt() call to
1919 retrieve the VT number of a session.
1921 * If the option "tries=0" is set for an entry of /etc/crypttab
1922 its passphrase is queried indefinitely instead of any
1923 maximum number of tries.
1925 * If a service with a configure PID file terminates its PID
1926 file will now be removed automatically if it still exists
1927 afterwards. This should put an end to stale PID files.
1929 * systemd-run will now also take relative binary path names
1930 for execution and no longer insists on absolute paths.
1932 * InaccessibleDirectories= and ReadOnlyDirectories= now take
1933 paths that are optionally prefixed with "-" to indicate that
1934 it should not be considered a failure if they do not exist.
1936 * journalctl -o (and similar commands) now understands a new
1937 output mode "short-precise", it is similar to "short" but
1938 shows timestamps with usec accuracy.
1940 * The option "discard" (as known from Debian) is now
1941 synonymous to "allow-discards" in /etc/crypttab. In fact,
1942 "discard" is preferred now (since it is easier to remember
1945 * Some licensing clean-ups were made, so that more code is now
1946 LGPL-2.1 licensed than before.
1948 * A minimal tool to save/restore the display backlight
1949 brightness across reboots has been added. It will store the
1950 backlight setting as late as possible at shutdown, and
1951 restore it as early as possible during reboot.
1953 * A logic to automatically discover and enable home and swap
1954 partitions on GPT disks has been added. With this in place
1955 /etc/fstab becomes optional for many setups as systemd can
1956 discover certain partitions located on the root disk
1957 automatically. Home partitions are recognized under their
1958 GPT type ID 933ac7e12eb44f13b8440e14e2aef915. Swap
1959 partitions are recognized under their GPT type ID
1960 0657fd6da4ab43c484e50933c84b4f4f.
1962 * systemd will no longer pass any environment from the kernel
1963 or initrd to system services. If you want to set an
1964 environment for all services, do so via the kernel command
1965 line systemd.setenv= assignment.
1967 * The systemd-sysctl tool no longer natively reads the file
1968 /etc/sysctl.conf. If desired, the file should be symlinked
1969 from /etc/sysctl.d/99-sysctl.conf. Apart from providing
1970 legacy support by a symlink rather than built-in code, it
1971 also makes the otherwise hidden order of application of the
1972 different files visible. (Note that this partly reverts to a
1973 pre-198 application order of sysctl knobs!)
1975 * The "systemctl set-log-level" and "systemctl dump" commands
1976 have been moved to systemd-analyze.
1978 * systemd-run learned the new --remain-after-exit switch,
1979 which causes the scope unit not to be cleaned up
1980 automatically after the process terminated.
1982 * tmpfiles learned a new --exclude-prefix= switch to exclude
1983 certain paths from operation.
1985 * journald will now automatically flush all messages to disk
1986 as soon as a message at the log level CRIT, ALERT or EMERG
1989 Contributions from: Andrew Cook, Brandon Philips, Christian
1990 Hesse, Christoph Junghans, Colin Walters, Daniel Schaal,
1991 Daniel Wallace, Dave Reisner, David Herrmann, Gao feng, George
1992 McCollister, Giovanni Campagna, Hannes Reinecke, Harald Hoyer,
1993 Herczeg Zsolt, Holger Hans Peter Freyther, Jan Engelhardt,
1994 Jesper Larsen, Kay Sievers, Khem Raj, Lennart Poettering,
1995 Lukas Nykryn, Maciej Wereski, Mantas Mikulėnas, Marcel
1996 Holtmann, Martin Pitt, Michael Biebl, Michael Marineau,
1997 Michael Scherer, Michael Stapelberg, Michal Sekletar, Michał
1998 Górny, Olivier Brunel, Ondrej Balaz, Ronny Chevalier, Shawn
1999 Landden, Steven Hiscocks, Thomas Bächler, Thomas Hindoe
2000 Paaboel Andersen, Tom Gundersen, Umut Tezduyar, WANG Chao,
2001 William Giokas, Zbigniew Jędrzejewski-Szmek
2003 -- Berlin, 2013-09-13
2007 * The documentation has been updated to cover the various new
2008 concepts introduced with 205.
2010 * Unit files now understand the new %v specifier which
2011 resolves to the kernel version string as returned by "uname
2014 * systemctl now supports filtering the unit list output by
2015 load state, active state and sub state, using the new
2018 * "systemctl status" will now show the results of the
2019 condition checks (like ConditionPathExists= and similar) of
2020 the last start attempts of the unit. They are also logged to
2023 * "journalctl -b" may now be used to look for boot output of a
2024 specific boot. Try "journalctl -b -1" for the previous boot,
2025 but the syntax is substantially more powerful.
2027 * "journalctl --show-cursor" has been added which prints the
2028 cursor string the last shown log line. This may then be used
2029 with the new "journalctl --after-cursor=" switch to continue
2030 browsing logs from that point on.
2032 * "journalctl --force" may now be used to force regeneration
2035 * Creation of "dead" device nodes has been moved from udev
2036 into kmod and tmpfiles. Previously, udev would read the kmod
2037 databases to pre-generate dead device nodes based on meta
2038 information contained in kernel modules, so that these would
2039 be auto-loaded on access rather then at boot. As this
2040 does not really have much to do with the exposing actual
2041 kernel devices to userspace this has always been slightly
2042 alien in the udev codebase. Following the new scheme kmod
2043 will now generate a runtime snippet for tmpfiles from the
2044 module meta information and it now is tmpfiles' job to the
2045 create the nodes. This also allows overriding access and
2046 other parameters for the nodes using the usual tmpfiles
2047 facilities. As side effect this allows us to remove the
2048 CAP_SYS_MKNOD capability bit from udevd entirely.
2050 * logind's device ACLs may now be applied to these "dead"
2051 devices nodes too, thus finally allowing managed access to
2052 devices such as /dev/snd/sequencer whithout loading the
2053 backing module right-away.
2055 * A new RPM macro has been added that may be used to apply
2056 tmpfiles configuration during package installation.
2058 * systemd-detect-virt and ConditionVirtualization= now can
2059 detect User-Mode-Linux machines (UML).
2061 * journald will now implicitly log the effective capabilities
2062 set of processes in the message metadata.
2064 * systemd-cryptsetup has gained support for TrueCrypt volumes.
2066 * The initrd interface has been simplified (more specifically,
2067 support for passing performance data via environment
2068 variables and fsck results via files in /run has been
2069 removed). These features were non-essential, and are
2070 nowadays available in a much nicer way by having systemd in
2071 the initrd serialize its state and have the hosts systemd
2072 deserialize it again.
2074 * The udev "keymap" data files and tools to apply keyboard
2075 specific mappings of scan to key codes, and force-release
2076 scan code lists have been entirely replaced by a udev
2077 "keyboard" builtin and a hwdb data file.
2079 * systemd will now honour the kernel's "quiet" command line
2080 argument also during late shutdown, resulting in a
2081 completely silent shutdown when used.
2083 * There's now an option to control the SO_REUSEPORT socket
2084 option in .socket units.
2086 * Instance units will now automatically get a per-template
2087 subslice of system.slice unless something else is explicitly
2088 configured. For example, instances of sshd@.service will now
2089 implicitly be placed in system-sshd.slice rather than
2090 system.slice as before.
2092 * Test coverage support may now be enabled at build time.
2094 Contributions from: Dave Reisner, Frederic Crozat, Harald
2095 Hoyer, Holger Hans Peter Freyther, Jan Engelhardt, Jan
2096 Janssen, Jason St. John, Jesper Larsen, Kay Sievers, Lennart
2097 Poettering, Lukas Nykryn, Maciej Wereski, Martin Pitt, Michael
2098 Olbrich, Ramkumar Ramachandra, Ross Lagerwall, Shawn Landden,
2099 Thomas H.P. Andersen, Tom Gundersen, Tomasz Torcz, William
2100 Giokas, Zbigniew Jędrzejewski-Szmek
2102 -- Berlin, 2013-07-23
2106 * Two new unit types have been introduced:
2108 Scope units are very similar to service units, however, are
2109 created out of pre-existing processes -- instead of PID 1
2110 forking off the processes. By using scope units it is
2111 possible for system services and applications to group their
2112 own child processes (worker processes) in a powerful way
2113 which then maybe used to organize them, or kill them
2114 together, or apply resource limits on them.
2116 Slice units may be used to partition system resources in an
2117 hierarchial fashion and then assign other units to them. By
2118 default there are now three slices: system.slice (for all
2119 system services), user.slice (for all user sessions),
2120 machine.slice (for VMs and containers).
2122 Slices and scopes have been introduced primarily in
2123 context of the work to move cgroup handling to a
2124 single-writer scheme, where only PID 1
2125 creates/removes/manages cgroups.
2127 * There's a new concept of "transient" units. In contrast to
2128 normal units these units are created via an API at runtime,
2129 not from configuration from disk. More specifically this
2130 means it is now possible to run arbitrary programs as
2131 independent services, with all execution parameters passed
2132 in via bus APIs rather than read from disk. Transient units
2133 make systemd substantially more dynamic then it ever was,
2134 and useful as a general batch manager.
2136 * logind has been updated to make use of scope and slice units
2137 for managing user sessions. As a user logs in he will get
2138 his own private slice unit, to which all sessions are added
2139 as scope units. We also added support for automatically
2140 adding an instance of user@.service for the user into the
2141 slice. Effectively logind will no longer create cgroup
2142 hierarchies on its own now, it will defer entirely to PID 1
2143 for this by means of scope, service and slice units. Since
2144 user sessions this way become entities managed by PID 1
2145 the output of "systemctl" is now a lot more comprehensive.
2147 * A new mini-daemon "systemd-machined" has been added which
2148 may be used by virtualization managers to register local
2149 VMs/containers. nspawn has been updated accordingly, and
2150 libvirt will be updated shortly. machined will collect a bit
2151 of meta information about the VMs/containers, and assign
2152 them their own scope unit (see above). The collected
2153 meta-data is then made available via the "machinectl" tool,
2154 and exposed in "ps" and similar tools. machined/machinectl
2155 is compile-time optional.
2157 * As discussed earlier, the low-level cgroup configuration
2158 options ControlGroup=, ControlGroupModify=,
2159 ControlGroupPersistent=, ControlGroupAttribute= have been
2160 removed. Please use high-level attribute settings instead as
2161 well as slice units.
2163 * A new bus call SetUnitProperties() has been added to alter
2164 various runtime parameters of a unit. This is primarily
2165 useful to alter cgroup parameters dynamically in a nice way,
2166 but will be extended later on to make more properties
2167 modifiable at runtime. systemctl gained a new set-properties
2168 command that wraps this call.
2170 * A new tool "systemd-run" has been added which can be used to
2171 run arbitrary command lines as transient services or scopes,
2172 while configuring a number of settings via the command
2173 line. This tool is currently very basic, however already
2174 very useful. We plan to extend this tool to even allow
2175 queuing of execution jobs with time triggers from the
2176 command line, similar in fashion to "at".
2178 * nspawn will now inform the user explicitly that kernels with
2179 audit enabled break containers, and suggest the user to turn
2182 * Support for detecting the IMA and AppArmor security
2183 frameworks with ConditionSecurity= has been added.
2185 * journalctl gained a new "-k" switch for showing only kernel
2186 messages, mimicking dmesg output; in addition to "--user"
2187 and "--system" switches for showing only user's own logs
2190 * systemd-delta can now show information about drop-in
2191 snippets extending unit files.
2193 * libsystemd-bus has been substantially updated but is still
2194 not available as public API.
2196 * systemd will now look for the "debug" argument on the kernel
2197 command line and enable debug logging, similar to what
2198 "systemd.log_level=debug" already did before.
2200 * "systemctl set-default", "systemctl get-default" has been
2201 added to configure the default.target symlink, which
2202 controls what to boot into by default.
2204 * "systemctl set-log-level" has been added as a convenient
2205 way to raise and lower systemd logging threshold.
2207 * "systemd-analyze plot" will now show the time the various
2208 generators needed for execution, as well as information
2209 about the unit file loading.
2211 * libsystemd-journal gained a new sd_journal_open_files() call
2212 for opening specific journal files. journactl also gained a
2213 new switch to expose this new functionality. Previously we
2214 only supported opening all files from a directory, or all
2215 files from the system, as opening individual files only is
2216 racy due to journal file rotation.
2218 * systemd gained the new DefaultEnvironment= setting in
2219 /etc/systemd/system.conf to set environment variables for
2222 * If a privileged process logs a journal message with the
2223 OBJECT_PID= field set, then journald will automatically
2224 augment this with additional OBJECT_UID=, OBJECT_GID=,
2225 OBJECT_COMM=, OBJECT_EXE=, ... fields. This is useful if
2226 system services want to log events about specific client
2227 processes. journactl/systemctl has been updated to make use
2228 of this information if all log messages regarding a specific
2231 Contributions from: Auke Kok, Chengwei Yang, Colin Walters,
2232 Cristian Rodríguez, Daniel Albers, Daniel Wallace, Dave
2233 Reisner, David Coppa, David King, David Strauss, Eelco
2234 Dolstra, Gabriel de Perthuis, Harald Hoyer, Jan Alexander
2235 Steffens, Jan Engelhardt, Jan Janssen, Jason St. John, Johan
2236 Heikkilä, Karel Zak, Karol Lewandowski, Kay Sievers, Lennart
2237 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marius Vollmer,
2238 Martin Pitt, Michael Biebl, Michael Olbrich, Michael Tremer,
2239 Michal Schmidt, Michał Bartoszkiewicz, Nirbheek Chauhan,
2240 Pierre Neidhardt, Ross Burton, Ross Lagerwall, Sean McGovern,
2241 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar,
2242 Václav Pavlín, Zachary Cook, Zbigniew Jędrzejewski-Szmek,
2243 Łukasz Stelmach, 장동준
2247 * The Python bindings gained some minimal support for the APIs
2248 exposed by libsystemd-logind.
2250 * ConditionSecurity= gained support for detecting SMACK. Since
2251 this condition already supports SELinux and AppArmor we only
2252 miss IMA for this. Patches welcome!
2254 Contributions from: Karol Lewandowski, Lennart Poettering,
2255 Zbigniew Jędrzejewski-Szmek
2259 * systemd-nspawn will now create /etc/resolv.conf if
2260 necessary, before bind-mounting the host's file onto it.
2262 * systemd-nspawn will now store meta information about a
2263 container on the container's cgroup as extended attribute
2264 fields, including the root directory.
2266 * The cgroup hierarchy has been reworked in many ways. All
2267 objects any of the components systemd creates in the cgroup
2268 tree are now suffixed. More specifically, user sessions are
2269 now placed in cgroups suffixed with ".session", users in
2270 cgroups suffixed with ".user", and nspawn containers in
2271 cgroups suffixed with ".nspawn". Furthermore, all cgroup
2272 names are now escaped in a simple scheme to avoid collision
2273 of userspace object names with kernel filenames. This work
2274 is preparation for making these objects relocatable in the
2275 cgroup tree, in order to allow easy resource partitioning of
2276 these objects without causing naming conflicts.
2278 * systemctl list-dependencies gained the new switches
2279 --plain, --reverse, --after and --before.
2281 * systemd-inhibit now shows the process name of processes that
2282 have taken an inhibitor lock.
2284 * nss-myhostname will now also resolve "localhost"
2285 implicitly. This makes /etc/hosts an optional file and
2286 nicely handles that on IPv6 ::1 maps to both "localhost" and
2289 * libsystemd-logind.so gained a new call
2290 sd_get_machine_names() to enumerate running containers and
2291 VMs (currently only supported by very new libvirt and
2292 nspawn). sd_login_monitor can now be used to watch
2293 VMs/containers coming and going.
2295 * .include is not allowed recursively anymore, and only in
2296 unit files. Usually it is better to use drop-in snippets in
2297 .d/*.conf anyway, as introduced with systemd 198.
2299 * systemd-analyze gained a new "critical-chain" command that
2300 determines the slowest chain of units run during system
2301 boot-up. It is very useful for tracking down where
2302 optimizing boot time is the most beneficial.
2304 * systemd will no longer allow manipulating service paths in
2305 the name=systemd:/system cgroup tree using ControlGroup= in
2306 units. (But is still fine with it in all other dirs.)
2308 * There's a new systemd-nspawn@.service service file that may
2309 be used to easily run nspawn containers as system
2310 services. With the container's root directory in
2311 /var/lib/container/foobar it is now sufficient to run
2312 "systemctl start systemd-nspawn@foobar.service" to boot it.
2314 * systemd-cgls gained a new parameter "--machine" to list only
2315 the processes within a certain container.
2317 * ConditionSecurity= now can check for "apparmor". We still
2318 are lacking checks for SMACK and IMA for this condition
2319 check though. Patches welcome!
2321 * A new configuration file /etc/systemd/sleep.conf has been
2322 added that may be used to configure which kernel operation
2323 systemd is supposed to execute when "suspend", "hibernate"
2324 or "hybrid-sleep" is requested. This makes the new kernel
2325 "freeze" state accessible to the user.
2327 * ENV{SYSTEMD_WANTS} in udev rules will now implicitly escape
2328 the passed argument if applicable.
2330 Contributions from: Auke Kok, Colin Guthrie, Colin Walters,
2331 Cristian Rodríguez, Daniel Buch, Daniel Wallace, Dave Reisner,
2332 Evangelos Foutras, Greg Kroah-Hartman, Harald Hoyer, Josh
2333 Triplett, Kay Sievers, Lennart Poettering, Lukas Nykryn,
2334 MUNEDA Takahiro, Mantas Mikulėnas, Mirco Tischler, Nathaniel
2335 Chen, Nirbheek Chauhan, Ronny Chevalier, Ross Lagerwall, Tom
2336 Gundersen, Umut Tezduyar, Ville Skyttä, Zbigniew
2341 * The output of 'systemctl list-jobs' got some polishing. The
2342 '--type=' argument may now be passed more than once. A new
2343 command 'systemctl list-sockets' has been added which shows
2344 a list of kernel sockets systemd is listening on with the
2345 socket units they belong to, plus the units these socket
2348 * The experimental libsystemd-bus library got substantial
2349 updates to work in conjunction with the (also experimental)
2350 kdbus kernel project. It works well enough to exchange
2351 messages with some sophistication. Note that kdbus is not
2352 ready yet, and the library is mostly an elaborate test case
2353 for now, and not installable.
2355 * systemd gained a new unit 'systemd-static-nodes.service'
2356 that generates static device nodes earlier during boot, and
2357 can run in conjunction with udev.
2359 * libsystemd-login gained a new call sd_pid_get_user_unit()
2360 to retrieve the user systemd unit a process is running
2361 in. This is useful for systems where systemd is used as
2364 * systemd-nspawn now places all containers in the new /machine
2365 top-level cgroup directory in the name=systemd
2366 hierarchy. libvirt will soon do the same, so that we get a
2367 uniform separation of /system, /user and /machine for system
2368 services, user processes and containers/virtual
2369 machines. This new cgroup hierarchy is also useful to stick
2370 stable names to specific container instances, which can be
2371 recognized later this way (this name may be controlled
2372 via systemd-nspawn's new -M switch). libsystemd-login also
2373 gained a new call sd_pid_get_machine_name() to retrieve the
2374 name of the container/VM a specific process belongs to.
2376 * bootchart can now store its data in the journal.
2378 * libsystemd-journal gained a new call
2379 sd_journal_add_conjunction() for AND expressions to the
2380 matching logic. This can be used to express more complex
2381 logical expressions.
2383 * journactl can now take multiple --unit= and --user-unit=
2386 * The cryptsetup logic now understands the "luks.key=" kernel
2387 command line switch for specifying a file to read the
2388 decryption key from. Also, if a configured key file is not
2389 found the tool will now automatically fall back to prompting
2392 * Python systemd.journal module was updated to wrap recently
2393 added functions from libsystemd-journal. The interface was
2394 changed to bring the low level interface in s.j._Reader
2395 closer to the C API, and the high level interface in
2396 s.j.Reader was updated to wrap and convert all data about
2399 Contributions from: Anatol Pomozov, Auke Kok, Harald Hoyer,
2400 Henrik Grindal Bakken, Josh Triplett, Kay Sievers, Lennart
2401 Poettering, Lukas Nykryn, Mantas Mikulėnas Marius Vollmer,
2402 Martin Jansa, Martin Pitt, Michael Biebl, Michal Schmidt,
2403 Mirco Tischler, Pali Rohar, Simon Peeters, Steven Hiscocks,
2404 Tom Gundersen, Zbigniew Jędrzejewski-Szmek
2408 * journalctl --update-catalog now understands a new --root=
2409 option to operate on catalogs found in a different root
2412 * During shutdown after systemd has terminated all running
2413 services a final killing loop kills all remaining left-over
2414 processes. We will now print the name of these processes
2415 when we send SIGKILL to them, since this usually indicates a
2418 * If /etc/crypttab refers to password files stored on
2419 configured mount points automatic dependencies will now be
2420 generated to ensure the specific mount is established first
2421 before the key file is attempted to be read.
2423 * 'systemctl status' will now show information about the
2424 network sockets a socket unit is listening on.
2426 * 'systemctl status' will also shown information about any
2427 drop-in configuration file for units. (Drop-In configuration
2428 files in this context are files such as
2429 /etc/systemd/systemd/foobar.service.d/*.conf)
2431 * systemd-cgtop now optionally shows summed up CPU times of
2432 cgroups. Press '%' while running cgtop to switch between
2433 percentage and absolute mode. This is useful to determine
2434 which cgroups use up the most CPU time over the entire
2435 runtime of the system. systemd-cgtop has also been updated
2436 to be 'pipeable' for processing with further shell tools.
2438 * 'hostnamectl set-hostname' will now allow setting of FQDN
2441 * The formatting and parsing of time span values has been
2442 changed. The parser now understands fractional expressions
2443 such as "5.5h". The formatter will now output fractional
2444 expressions for all time spans under 1min, i.e. "5.123456s"
2445 rather than "5s 123ms 456us". For time spans under 1s
2446 millisecond values are shown, for those under 1ms
2447 microsecond values are shown. This should greatly improve
2448 all time-related output of systemd.
2450 * libsystemd-login and libsystemd-journal gained new
2451 functions for querying the poll() events mask and poll()
2452 timeout value for integration into arbitrary event
2455 * localectl gained the ability to list available X11 keymaps
2456 (models, layouts, variants, options).
2458 * 'systemd-analyze dot' gained the ability to filter for
2459 specific units via shell-style globs, to create smaller,
2460 more useful graphs. I.e. it is now possible to create simple
2461 graphs of all the dependencies between only target units, or
2462 of all units that Avahi has dependencies with.
2464 Contributions from: Cristian Rodríguez, Dr. Tilmann Bubeck,
2465 Harald Hoyer, Holger Hans Peter Freyther, Kay Sievers, Kelly
2466 Anderson, Koen Kooi, Lennart Poettering, Maksim Melnikau,
2467 Marc-Antoine Perennou, Marius Vollmer, Martin Pitt, Michal
2468 Schmidt, Oleksii Shevchuk, Ronny Chevalier, Simon McVittie,
2469 Steven Hiscocks, Thomas Weißschuh, Umut Tezduyar, Václav
2470 Pavlín, Zbigniew Jędrzejewski-Szmek, Łukasz Stelmach
2474 * The boot-time readahead implementation for rotating media
2475 will now read the read-ahead data in multiple passes which
2476 consist of all read requests made in equidistant time
2477 intervals. This means instead of strictly reading read-ahead
2478 data in its physical order on disk we now try to find a
2479 middle ground between physical and access time order.
2481 * /etc/os-release files gained a new BUILD_ID= field for usage
2482 on operating systems that provide continuous builds of OS
2485 Contributions from: Auke Kok, Eelco Dolstra, Kay Sievers,
2486 Lennart Poettering, Lukas Nykryn, Martin Pitt, Václav Pavlín
2487 William Douglas, Zbigniew Jędrzejewski-Szmek
2491 * systemd-python gained an API exposing libsystemd-daemon.
2493 * The SMACK setup logic gained support for uploading CIPSO
2496 * Behaviour of PrivateTmp=, ReadWriteDirectories=,
2497 ReadOnlyDirectories= and InaccessibleDirectories= has
2498 changed. The private /tmp and /var/tmp directories are now
2499 shared by all processes of a service (which means
2500 ExecStartPre= may now leave data in /tmp that ExecStart= of
2501 the same service can still access). When a service is
2502 stopped its temporary directories are immediately deleted
2503 (normal clean-up with tmpfiles is still done in addition to
2506 * By default, systemd will now set a couple of sysctl
2507 variables in the kernel: the safe sysrq options are turned
2508 on, IP route verification is turned on, and source routing
2509 disabled. The recently added hardlink and softlink
2510 protection of the kernel is turned on. These settings should
2511 be reasonably safe, and good defaults for all new systems.
2513 * The predictable network naming logic may now be turned off
2514 with a new kernel command line switch: net.ifnames=0.
2516 * A new libsystemd-bus module has been added that implements a
2517 pretty complete D-Bus client library. For details see:
2519 http://lists.freedesktop.org/archives/systemd-devel/2013-March/009797.html
2521 * journald will now explicitly flush the journal files to disk
2522 at the latest 5min after each write. The file will then also
2523 be marked offline until the next write. This should increase
2524 reliability in case of a crash. The synchronization delay
2525 can be configured via SyncIntervalSec= in journald.conf.
2527 * There's a new remote-fs-setup.target unit that can be used
2528 to pull in specific services when at least one remote file
2529 system is to be mounted.
2531 * There are new targets timers.target and paths.target as
2532 canonical targets to pull user timer and path units in
2533 from. This complements sockets.target with a similar
2534 purpose for socket units.
2536 * libudev gained a new call udev_device_set_attribute_value()
2537 to set sysfs attributes of a device.
2539 * The udev daemon now sets the default number of worker
2540 processes executed in parallel based on the number of available
2541 CPUs instead of the amount of available RAM. This is supposed
2542 to provide a more reliable default and limit a too aggressive
2543 paralellism for setups with 1000s of devices connected.
2545 Contributions from: Auke Kok, Colin Walters, Cristian
2546 Rodríguez, Daniel Buch, Dave Reisner, Frederic Crozat, Hannes
2547 Reinecke, Harald Hoyer, Jan Alexander Steffens, Jan
2548 Engelhardt, Josh Triplett, Kay Sievers, Lennart Poettering,
2549 Mantas Mikulėnas, Martin Pitt, Mathieu Bridon, Michael Biebl,
2550 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nathaniel Chen,
2551 Oleksii Shevchuk, Ozan Çağlayan, Thomas Hindoe Paaboel
2552 Andersen, Tollef Fog Heen, Tom Gundersen, Umut Tezduyar,
2553 Zbigniew Jędrzejewski-Szmek
2557 * Configuration of unit files may now be extended via drop-in
2558 files without having to edit/override the unit files
2559 themselves. More specifically, if the administrator wants to
2560 change one value for a service file foobar.service he can
2561 now do so by dropping in a configuration snippet into
2562 /etc/systemd/system/foobar.service.d/*.conf. The unit logic
2563 will load all these snippets and apply them on top of the
2564 main unit configuration file, possibly extending or
2565 overriding its settings. Using these drop-in snippets is
2566 generally nicer than the two earlier options for changing
2567 unit files locally: copying the files from
2568 /usr/lib/systemd/system/ to /etc/systemd/system/ and editing
2569 them there; or creating a new file in /etc/systemd/system/
2570 that incorporates the original one via ".include". Drop-in
2571 snippets into these .d/ directories can be placed in any
2572 directory systemd looks for units in, and the usual
2573 overriding semantics between /usr/lib, /etc and /run apply
2576 * Most unit file settings which take lists of items can now be
2577 reset by assigning the empty string to them. For example,
2578 normally, settings such as Environment=FOO=BAR append a new
2579 environment variable assignment to the environment block,
2580 each time they are used. By assigning Environment= the empty
2581 string the environment block can be reset to empty. This is
2582 particularly useful with the .d/*.conf drop-in snippets
2583 mentioned above, since this adds the ability to reset list
2584 settings from vendor unit files via these drop-ins.
2586 * systemctl gained a new "list-dependencies" command for
2587 listing the dependencies of a unit recursively.
2589 * Inhibitors are now honored and listed by "systemctl
2590 suspend", "systemctl poweroff" (and similar) too, not only
2591 GNOME. These commands will also list active sessions by
2594 * Resource limits (as exposed by the various control group
2595 controllers) can now be controlled dynamically at runtime
2596 for all units. More specifically, you can now use a command
2597 like "systemctl set-cgroup-attr foobar.service cpu.shares
2598 2000" to alter the CPU shares a specific service gets. These
2599 settings are stored persistently on disk, and thus allow the
2600 administrator to easily adjust the resource usage of
2601 services with a few simple commands. This dynamic resource
2602 management logic is also available to other programs via the
2603 bus. Almost any kernel cgroup attribute and controller is
2606 * systemd-vconsole-setup will now copy all font settings to
2607 all allocated VTs, where it previously applied them only to
2610 * libsystemd-login gained the new sd_session_get_tty() API
2613 * This release drops support for a few legacy or
2614 distribution-specific LSB facility names when parsing init
2615 scripts: $x-display-manager, $mail-transfer-agent,
2616 $mail-transport-agent, $mail-transfer-agent, $smtp,
2617 $null. Also, the mail-transfer-agent.target unit backing
2618 this has been removed. Distributions which want to retain
2619 compatibility with this should carry the burden for
2620 supporting this themselves and patch support for these back
2621 in, if they really need to. Also, the facilities $syslog and
2622 $local_fs are now ignored, since systemd does not support
2623 early-boot LSB init scripts anymore, and these facilities
2624 are implied anyway for normal services. syslog.target has
2627 * There are new bus calls on PID1's Manager object for
2628 cancelling jobs, and removing snapshot units. Previously,
2629 both calls were only available on the Job and Snapshot
2632 * systemd-journal-gatewayd gained SSL support.
2634 * The various "environment" files, such as /etc/locale.conf
2635 now support continuation lines with a backslash ("\") as
2636 last character in the line, similarly in style (but different)
2637 to how this is supported in shells.
2639 * For normal user processes the _SYSTEMD_USER_UNIT= field is
2640 now implicitly appended to every log entry logged. systemctl
2641 has been updated to filter by this field when operating on a
2642 user systemd instance.
2644 * nspawn will now implicitly add the CAP_AUDIT_WRITE and
2645 CAP_AUDIT_CONTROL capabilities to the capabilities set for
2646 the container. This makes it easier to boot unmodified
2647 Fedora systems in a container, which however still requires
2648 audit=0 to be passed on the kernel command line. Auditing in
2649 kernel and userspace is unfortunately still too broken in
2650 context of containers, hence we recommend compiling it out
2651 of the kernel or using audit=0. Hopefully this will be fixed
2652 one day for good in the kernel.
2654 * nspawn gained the new --bind= and --bind-ro= parameters to
2655 bind mount specific directories from the host into the
2658 * nspawn will now mount its own devpts file system instance
2659 into the container, in order not to leak pty devices from
2660 the host into the container.
2662 * systemd will now read the firmware boot time performance
2663 information from the EFI variables, if the used boot loader
2664 supports this, and takes it into account for boot performance
2665 analysis via "systemd-analyze". This is currently supported
2666 only in conjunction with Gummiboot, but could be supported
2667 by other boot loaders too. For details see:
2669 http://www.freedesktop.org/wiki/Software/systemd/BootLoaderInterface
2671 * A new generator has been added that automatically mounts the
2672 EFI System Partition (ESP) to /boot, if that directory
2673 exists, is empty, and no other file system has been
2674 configured to be mounted there.
2676 * logind will now send out PrepareForSleep(false) out
2677 unconditionally, after coming back from suspend. This may be
2678 used by applications as asynchronous notification for
2679 system resume events.
2681 * "systemctl unlock-sessions" has been added, that allows
2682 unlocking the screens of all user sessions at once, similar
2683 to how "systemctl lock-sessions" already locked all users
2684 sessions. This is backed by a new D-Bus call UnlockSessions().
2686 * "loginctl seat-status" will now show the master device of a
2687 seat. (i.e. the device of a seat that needs to be around for
2688 the seat to be considered available, usually the graphics
2691 * tmpfiles gained a new "X" line type, that allows
2692 configuration of files and directories (with wildcards) that
2693 shall be excluded from automatic cleanup ("aging").
2695 * udev default rules set the device node permissions now only
2696 at "add" events, and do not change them any longer with a
2697 later "change" event.
2699 * The log messages for lid events and power/sleep keypresses
2700 now carry a message ID.
2702 * We now have a substantially larger unit test suite, but this
2703 continues to be work in progress.
2705 * udevadm hwdb gained a new --root= parameter to change the
2706 root directory to operate relative to.
2708 * logind will now issue a background sync() request to the kernel
2709 early at shutdown, so that dirty buffers are flushed to disk early
2710 instead of at the last moment, in order to optimize shutdown
2713 * A new bootctl tool has been added that is an interface for
2714 certain boot loader operations. This is currently a preview
2715 and is likely to be extended into a small mechanism daemon
2716 like timedated, localed, hostnamed, and can be used by
2717 graphical UIs to enumerate available boot options, and
2718 request boot into firmware operations.
2720 * systemd-bootchart has been relicensed to LGPLv2.1+ to match
2721 the rest of the package. It also has been updated to work
2722 correctly in initrds.
2724 * Policykit previously has been runtime optional, and is now
2725 also compile time optional via a configure switch.
2727 * systemd-analyze has been reimplemented in C. Also "systemctl
2728 dot" has moved into systemd-analyze.
2730 * "systemctl status" with no further parameters will now print
2731 the status of all active or failed units.
2733 * Operations such as "systemctl start" can now be executed
2734 with a new mode "--irreversible" which may be used to queue
2735 operations that cannot accidentally be reversed by a later
2736 job queuing. This is by default used to make shutdown
2737 requests more robust.
2739 * The Python API of systemd now gained a new module for
2740 reading journal files.
2742 * A new tool kernel-install has been added that can install
2743 kernel images according to the Boot Loader Specification:
2745 http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec
2747 * Boot time console output has been improved to provide
2748 animated boot time output for hanging jobs.
2750 * A new tool systemd-activate has been added which can be used
2751 to test socket activation with, directly from the command
2752 line. This should make it much easier to test and debug
2753 socket activation in daemons.
2755 * journalctl gained a new "--reverse" (or -r) option to show
2756 journal output in reverse order (i.e. newest line first).
2758 * journalctl gained a new "--pager-end" (or -e) option to jump
2759 to immediately jump to the end of the journal in the
2760 pager. This is only supported in conjunction with "less".
2762 * journalctl gained a new "--user-unit=" option, that works
2763 similarly to "--unit=" but filters for user units rather than
2766 * A number of unit files to ease adoption of systemd in
2767 initrds has been added. This moves some minimal logic from
2768 the various initrd implementations into systemd proper.
2770 * The journal files are now owned by a new group
2771 "systemd-journal", which exists specifically to allow access
2772 to the journal, and nothing else. Previously, we used the
2773 "adm" group for that, which however possibly covers more
2774 than just journal/log file access. This new group is now
2775 already used by systemd-journal-gatewayd to ensure this
2776 daemon gets access to the journal files and as little else
2777 as possible. Note that "make install" will also set FS ACLs
2778 up for /var/log/journal to give "adm" and "wheel" read
2779 access to it, in addition to "systemd-journal" which owns
2780 the journal files. We recommend that packaging scripts also
2781 add read access to "adm" + "wheel" to /var/log/journal, and
2782 all existing/future journal files. To normal users and
2783 administrators little changes, however packagers need to
2784 ensure to create the "systemd-journal" system group at
2785 package installation time.
2787 * The systemd-journal-gatewayd now runs as unprivileged user
2788 systemd-journal-gateway:systemd-journal-gateway. Packaging
2789 scripts need to create these system user/group at
2792 * timedated now exposes a new boolean property CanNTP that
2793 indicates whether a local NTP service is available or not.
2795 * systemd-detect-virt will now also detect xen PVs
2797 * The pstore file system is now mounted by default, if it is
2800 * In addition to the SELinux and IMA policies we will now also
2801 load SMACK policies at early boot.
2803 Contributions from: Adel Gadllah, Aleksander Morgado, Auke
2804 Kok, Ayan George, Bastien Nocera, Colin Walters, Daniel Buch,
2805 Daniel Wallace, Dave Reisner, David Herrmann, David Strauss,
2806 Eelco Dolstra, Enrico Scholz, Frederic Crozat, Harald Hoyer,
2807 Jan Janssen, Jonathan Callen, Kay Sievers, Lennart Poettering,
2808 Lukas Nykryn, Mantas Mikulėnas, Marc-Antoine Perennou, Martin
2809 Pitt, Mauro Dreissig, Max F. Albrecht, Michael Biebl, Michael
2810 Olbrich, Michal Schmidt, Michal Sekletar, Michal Vyskocil,
2811 Michał Bartoszkiewicz, Mirco Tischler, Nathaniel Chen, Nestor
2812 Ovroy, Oleksii Shevchuk, Paul W. Frields, Piotr Drąg, Rob
2813 Clark, Ryan Lortie, Simon McVittie, Simon Peeters, Steven
2814 Hiscocks, Thomas Hindoe Paaboel Andersen, Tollef Fog Heen, Tom
2815 Gundersen, Umut Tezduyar, William Giokas, Zbigniew
2816 Jędrzejewski-Szmek, Zeeshan Ali (Khattak)
2820 * Timer units now support calendar time events in addition to
2821 monotonic time events. That means you can now trigger a unit
2822 based on a calendar time specification such as "Thu,Fri
2823 2013-*-1,5 11:12:13" which refers to 11:12:13 of the first
2824 or fifth day of any month of the year 2013, given that it is
2825 a thursday or friday. This brings timer event support
2826 considerably closer to cron's capabilities. For details on
2827 the supported calendar time specification language see
2830 * udev now supports a number of different naming policies for
2831 network interfaces for predictable names, and a combination
2832 of these policies is now the default. Please see this wiki
2833 document for details:
2835 http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames
2837 * Auke Kok's bootchart implementation has been added to the
2838 systemd tree. It is an optional component that can graph the
2839 boot in quite some detail. It is one of the best bootchart
2840 implementations around and minimal in its code and
2843 * nss-myhostname has been integrated into the systemd source
2844 tree. nss-myhostname guarantees that the local hostname
2845 always stays resolvable via NSS. It has been a weak
2846 requirement of systemd-hostnamed since a long time, and
2847 since its code is actually trivial we decided to just
2848 include it in systemd's source tree. It can be turned off
2849 with a configure switch.
2851 * The read-ahead logic is now capable of properly detecting
2852 whether a btrfs file system is on SSD or rotating media, in
2853 order to optimize the read-ahead scheme. Previously, it was
2854 only capable of detecting this on traditional file systems
2857 * In udev, additional device properties are now read from the
2858 IAB in addition to the OUI database. Also, Bluetooth company
2859 identities are attached to the devices as well.
2861 * In service files %U may be used as specifier that is
2862 replaced by the configured user name of the service.
2864 * nspawn may now be invoked without a controlling TTY. This
2865 makes it suitable for invocation as its own service. This
2866 may be used to set up a simple containerized server system
2867 using only core OS tools.
2869 * systemd and nspawn can now accept socket file descriptors
2870 when they are started for socket activation. This enables
2871 implementation of socket activated nspawn
2872 containers. i.e. think about autospawning an entire OS image
2873 when the first SSH or HTTP connection is received. We expect
2874 that similar functionality will also be added to libvirt-lxc
2877 * journalctl will now suppress ANSI color codes when
2878 presenting log data.
2880 * systemctl will no longer show control group information for
2881 a unit if a the control group is empty anyway.
2883 * logind can now automatically suspend/hibernate/shutdown the
2886 * /etc/machine-info and hostnamed now also expose the chassis
2887 type of the system. This can be used to determine whether
2888 the local system is a laptop, desktop, handset or
2889 tablet. This information may either be configured by the
2890 user/vendor or is automatically determined from ACPI and DMI
2891 information if possible.
2893 * A number of PolicyKit actions are now bound together with
2894 "imply" rules. This should simplify creating UIs because
2895 many actions will now authenticate similar ones as well.
2897 * Unit files learnt a new condition ConditionACPower= which
2898 may be used to conditionalize a unit depending on whether an
2899 AC power source is connected or not, of whether the system
2900 is running on battery power.
2902 * systemctl gained a new "is-failed" verb that may be used in
2903 shell scripts and suchlike to check whether a specific unit
2904 is in the "failed" state.
2906 * The EnvironmentFile= setting in unit files now supports file
2907 globbing, and can hence be used to easily read a number of
2908 environment files at once.
2910 * systemd will no longer detect and recognize specific
2911 distributions. All distribution-specific #ifdeffery has been
2912 removed, systemd is now fully generic and
2913 distribution-agnostic. Effectively, not too much is lost as
2914 a lot of the code is still accessible via explicit configure
2915 switches. However, support for some distribution specific
2916 legacy configuration file formats has been dropped. We
2917 recommend distributions to simply adopt the configuration
2918 files everybody else uses now and convert the old
2919 configuration from packaging scripts. Most distributions
2920 already did that. If that's not possible or desirable,
2921 distributions are welcome to forward port the specific
2922 pieces of code locally from the git history.
2924 * When logging a message about a unit systemd will now always
2925 log the unit name in the message meta data.
2927 * localectl will now also discover system locale data that is
2928 not stored in locale archives, but directly unpacked.
2930 * logind will no longer unconditionally use framebuffer
2931 devices as seat masters, i.e. as devices that are required
2932 to be existing before a seat is considered preset. Instead,
2933 it will now look for all devices that are tagged as
2934 "seat-master" in udev. By default framebuffer devices will
2935 be marked as such, but depending on local systems other
2936 devices might be marked as well. This may be used to
2937 integrate graphics cards using closed source drivers (such
2938 as NVidia ones) more nicely into logind. Note however, that
2939 we recommend using the open source NVidia drivers instead,
2940 and no udev rules for the closed-source drivers will be
2941 shipped from us upstream.
2943 Contributions from: Adam Williamson, Alessandro Crismani, Auke
2944 Kok, Colin Walters, Daniel Wallace, Dave Reisner, David
2945 Herrmann, David Strauss, Dimitrios Apostolou, Eelco Dolstra,
2946 Eric Benoit, Giovanni Campagna, Hannes Reinecke, Henrik
2947 Grindal Bakken, Hermann Gausterer, Kay Sievers, Lennart
2948 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marcel Holtmann,
2949 Martin Pitt, Matthew Monaco, Michael Biebl, Michael Terry,
2950 Michal Schmidt, Michal Sekletar, Michał Bartoszkiewicz, Oleg
2951 Samarin, Pekka Lundstrom, Philip Nilsson, Ramkumar
2952 Ramachandra, Richard Yao, Robert Millan, Sami Kerola, Shawn
2953 Landden, Thomas Hindoe Paaboel Andersen, Thomas Jarosch,
2954 Tollef Fog Heen, Tom Gundersen, Umut Tezduyar, Zbigniew
2959 * udev gained support for loading additional device properties
2960 from an indexed database that is keyed by vendor/product IDs
2961 and similar device identifiers. For the beginning this
2962 "hwdb" is populated with data from the well-known PCI and
2963 USB database, but also includes PNP, ACPI and OID data. In
2964 the longer run this indexed database shall grow into
2965 becoming the one central database for non-essential
2966 userspace device metadata. Previously, data from the PCI/USB
2967 database was only attached to select devices, since the
2968 lookup was a relatively expensive operation due to O(n) time
2969 complexity (with n being the number of entries in the
2970 database). Since this is now O(1), we decided to add in this
2971 data for all devices where this is available, by
2972 default. Note that the indexed database needs to be rebuilt
2973 when new data files are installed. To achieve this you need
2974 to update your packaging scripts to invoke "udevadm hwdb
2975 --update" after installation of hwdb data files. For
2976 RPM-based distributions we introduced the new
2977 %udev_hwdb_update macro for this purpose.
2979 * The Journal gained support for the "Message Catalog", an
2980 indexed database to link up additional information with
2981 journal entries. For further details please check:
2983 http://www.freedesktop.org/wiki/Software/systemd/catalog
2985 The indexed message catalog database also needs to be
2986 rebuilt after installation of message catalog files. Use
2987 "journalctl --update-catalog" for this. For RPM-based
2988 distributions we introduced the %journal_catalog_update
2989 macro for this purpose.
2991 * The Python Journal bindings gained support for the standard
2992 Python logging framework.
2994 * The Journal API gained new functions for checking whether
2995 the underlying file system of a journal file is capable of
2996 properly reporting file change notifications, or whether
2997 applications that want to reflect journal changes "live"
2998 need to recheck journal files continuously in appropriate
3001 * It is now possible to set the "age" field for tmpfiles
3002 entries to 0, indicating that files matching this entry
3003 shall always be removed when the directories are cleaned up.
3005 * coredumpctl gained a new "gdb" verb which invokes gdb
3006 right-away on the selected coredump.
3008 * There's now support for "hybrid sleep" on kernels that
3009 support this, in addition to "suspend" and "hibernate". Use
3010 "systemctl hybrid-sleep" to make use of this.
3012 * logind's HandleSuspendKey= setting (and related settings)
3013 now gained support for a new "lock" setting to simply
3014 request the screen lock on all local sessions, instead of
3015 actually executing a suspend or hibernation.
3017 * systemd will now mount the EFI variables file system by
3020 * Socket units now gained support for configuration of the
3021 SMACK security label.
3023 * timedatectl will now output the time of the last and next
3024 daylight saving change.
3026 * We dropped support for various legacy and distro-specific
3027 concepts, such as insserv, early-boot SysV services
3028 (i.e. those for non-standard runlevels such as 'b' or 'S')
3029 or ArchLinux /etc/rc.conf support. We recommend the
3030 distributions who still need support this to either continue
3031 to maintain the necessary patches downstream, or find a
3032 different solution. (Talk to us if you have questions!)
3034 * Various systemd components will now bypass PolicyKit checks
3035 for root and otherwise handle properly if PolicyKit is not
3036 found to be around. This should fix most issues for
3037 PolicyKit-less systems. Quite frankly this should have been
3038 this way since day one. It is absolutely our intention to
3039 make systemd work fine on PolicyKit-less systems, and we
3040 consider it a bug if something does not work as it should if
3041 PolicyKit is not around.
3043 * For embedded systems it is now possible to build udev and
3044 systemd without blkid and/or kmod support.
3046 * "systemctl switch-root" is now capable of switching root
3047 more than once. I.e. in addition to transitions from the
3048 initrd to the host OS it is now possible to transition to
3049 further OS images from the host. This is useful to implement
3050 offline updating tools.
3052 * Various other additions have been made to the RPM macros
3053 shipped with systemd. Use %udev_rules_update() after
3054 installing new udev rules files. %_udevhwdbdir,
3055 %_udevrulesdir, %_journalcatalogdir, %_tmpfilesdir,
3056 %_sysctldir are now available which resolve to the right
3057 directories for packages to place various data files in.
3059 * journalctl gained the new --full switch (in addition to
3060 --all, to disable ellipsation for long messages.
3062 Contributions from: Anders Olofsson, Auke Kok, Ben Boeckel,
3063 Colin Walters, Cosimo Cecchi, Daniel Wallace, Dave Reisner,
3064 Eelco Dolstra, Holger Hans Peter Freyther, Kay Sievers,
3065 Chun-Yi Lee, Lekensteyn, Lennart Poettering, Mantas Mikulėnas,
3066 Marti Raudsepp, Martin Pitt, Mauro Dreissig, Michael Biebl,
3067 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nis Martensen,
3068 Oleksii Shevchuk, Olivier Brunel, Ramkumar Ramachandra, Thomas
3069 Bächler, Thomas Hindoe Paaboel Andersen, Tom Gundersen, Tony
3070 Camuso, Umut Tezduyar, Zbigniew Jędrzejewski-Szmek
3074 * journalctl gained new --since= and --until= switches to
3075 filter by time. It also now supports nice filtering for
3076 units via --unit=/-u.
3078 * Type=oneshot services may use ExecReload= and do the
3081 * The journal daemon now supports time-based rotation and
3082 vacuuming, in addition to the usual disk-space based
3085 * The journal will now index the available field values for
3086 each field name. This enables clients to show pretty drop
3087 downs of available match values when filtering. The bash
3088 completion of journalctl has been updated
3089 accordingly. journalctl gained a new switch -F to list all
3090 values a certain field takes in the journal database.
3092 * More service events are now written as structured messages
3093 to the journal, and made recognizable via message IDs.
3095 * The timedated, localed and hostnamed mini-services which
3096 previously only provided support for changing time, locale
3097 and hostname settings from graphical DEs such as GNOME now
3098 also have a minimal (but very useful) text-based client
3099 utility each. This is probably the nicest way to changing
3100 these settings from the command line now, especially since
3101 it lists available options and is fully integrated with bash
3104 * There's now a new tool "systemd-coredumpctl" to list and
3105 extract coredumps from the journal.
3107 * We now install a README each in /var/log/ and
3108 /etc/rc.d/init.d explaining where the system logs and init
3109 scripts went. This hopefully should help folks who go to
3110 that dirs and look into the otherwise now empty void and
3111 scratch their heads.
3113 * When user-services are invoked (by systemd --user) the
3114 $MANAGERPID env var is set to the PID of systemd.
3116 * SIGRTMIN+24 when sent to a --user instance will now result
3117 in immediate termination of systemd.
3119 * gatewayd received numerous feature additions such as a
3120 "follow" mode, for live syncing and filtering.
3122 * browse.html now allows filtering and showing detailed
3123 information on specific entries. Keyboard navigation and
3124 mouse screen support has been added.
3126 * gatewayd/journalctl now supports HTML5/JSON
3127 Server-Sent-Events as output.
3129 * The SysV init script compatibility logic will now
3130 heuristically determine whether a script supports the
3131 "reload" verb, and only then make this available as
3134 * "systemctl status --follow" has been removed, use "journalctl
3137 * journald.conf's RuntimeMinSize=, PersistentMinSize= settings
3138 have been removed since they are hardly useful to be
3141 * And I'd like to take the opportunity to specifically mention
3142 Zbigniew for his great contributions. Zbigniew, you rock!
3144 Contributions from: Andrew Eikum, Christian Hesse, Colin
3145 Guthrie, Daniel J Walsh, Dave Reisner, Eelco Dolstra, Ferenc
3146 Wágner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Mantas
3147 Mikulėnas, Martin Mikkelsen, Martin Pitt, Michael Olbrich,
3148 Michael Stapelberg, Michal Schmidt, Sebastian Ott, Thomas
3149 Bächler, Umut Tezduyar, Will Woods, Wulf C. Krueger, Zbigniew
3150 Jędrzejewski-Szmek, Сковорода Никита Андреевич
3154 * If /etc/vconsole.conf is non-existent or empty we will no
3155 longer load any console font or key map at boot by
3156 default. Instead the kernel defaults will be left
3157 intact. This is definitely the right thing to do, as no
3158 configuration should mean no configuration, and hard-coding
3159 font names that are different on all archs is probably a bad
3160 idea. Also, the kernel default key map and font should be
3161 good enough for most cases anyway, and mostly identical to
3162 the userspace fonts/key maps we previously overloaded them
3163 with. If distributions want to continue to default to a
3164 non-kernel font or key map they should ship a default
3165 /etc/vconsole.conf with the appropriate contents.
3167 Contributions from: Colin Walters, Daniel J Walsh, Dave
3168 Reisner, Kay Sievers, Lennart Poettering, Lukas Nykryn, Tollef
3169 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
3173 * journalctl gained a new --cursor= switch to show entries
3174 starting from the specified location in the journal.
3176 * We now enforce a size limit on journal entry fields exported
3177 with "-o json" in journalctl. Fields larger than 4K will be
3178 assigned null. This can be turned off with --all.
3180 * An (optional) journal gateway daemon is now available as
3181 "systemd-journal-gatewayd.service". This service provides
3182 access to the journal via HTTP and JSON. This functionality
3183 will be used to implement live log synchronization in both
3184 pull and push modes, but has various other users too, such
3185 as easy log access for debugging of embedded devices. Right
3186 now it is already useful to retrieve the journal via HTTP:
3188 # systemctl start systemd-journal-gatewayd.service
3189 # wget http://localhost:19531/entries
3191 This will download the journal contents in a
3192 /var/log/messages compatible format. The same as JSON:
3194 # curl -H"Accept: application/json" http://localhost:19531/entries
3196 This service is also accessible via a web browser where a
3197 single static HTML5 app is served that uses the JSON logic
3198 to enable the user to do some basic browsing of the
3199 journal. This will be extended later on. Here's an example
3200 screenshot of this app in its current state:
3202 http://0pointer.de/public/journal-gatewayd
3204 Contributions from: Kay Sievers, Lennart Poettering, Robert
3205 Milasan, Tom Gundersen
3209 * The bash completion logic is now available for journalctl
3212 * We do not mount the "cpuset" controller anymore together with
3213 "cpu" and "cpuacct", as "cpuset" groups generally cannot be
3214 started if no parameters are assigned to it. "cpuset" hence
3215 broke code that assumed it it could create "cpu" groups and
3218 * journalctl -f will now subscribe to terminal size changes,
3219 and line break accordingly.
3221 Contributions from: Dave Reisner, Kay Sievers, Lennart
3222 Poettering, Lukas Nykrynm, Mirco Tischler, Václav Pavlín
3226 * nspawn will now create a symlink /etc/localtime in the
3227 container environment, copying the host's timezone
3228 setting. Previously this has been done via a bind mount, but
3229 since symlinks cannot be bind mounted this has now been
3230 changed to create/update the appropriate symlink.
3232 * journalctl -n's line number argument is now optional, and
3233 will default to 10 if omitted.
3235 * journald will now log the maximum size the journal files may
3236 take up on disk. This is particularly useful if the default
3237 built-in logic of determining this parameter from the file
3238 system size is used. Use "systemctl status
3239 systemd-journald.service" to see this information.
3241 * The multi-seat X wrapper tool has been stripped down. As X
3242 is now capable of enumerating graphics devices via udev in a
3243 seat-aware way the wrapper is not strictly necessary
3244 anymore. A stripped down temporary stop-gap is still shipped
3245 until the upstream display managers have been updated to
3246 fully support the new X logic. Expect this wrapper to be
3247 removed entirely in one of the next releases.
3249 * HandleSleepKey= in logind.conf has been split up into
3250 HandleSuspendKey= and HandleHibernateKey=. The old setting
3251 is not available anymore. X11 and the kernel are
3252 distuingishing between these keys and we should too. This
3253 also means the inhibition lock for these keys has been split
3256 Contributions from: Dave Airlie, Eelco Dolstra, Lennart
3257 Poettering, Lukas Nykryn, Václav Pavlín
3261 * Whenever a unit changes state we will now log this to the
3262 journal and show along the unit's own log output in
3265 * ConditionPathIsMountPoint= can now properly detect bind
3266 mount points too. (Previously, a bind mount of one file
3267 system to another place in the same file system could not be
3268 detected as mount, since they shared struct stat's st_dev
3271 * We will now mount the cgroup controllers cpu, cpuacct,
3272 cpuset and the controllers net_cls, net_prio together by
3275 * nspawn containers will now have a virtualized boot
3276 ID. (i.e. /proc/sys/kernel/random/boot_id is now mounted
3277 over with a randomized ID at container initialization). This
3278 has the effect of making "journalctl -b" do the right thing
3281 * The JSON output journal serialization has been updated not
3282 to generate "endless" list objects anymore, but rather one
3283 JSON object per line. This is more in line how most JSON
3284 parsers expect JSON objects. The new output mode
3285 "json-pretty" has been added to provide similar output, but
3286 neatly aligned for readability by humans.
3288 * We dropped all explicit sync() invocations in the shutdown
3289 code. The kernel does this implicitly anyway in the kernel
3290 reboot() syscall. halt(8)'s -n option is now a compatibility
3293 * We now support virtualized reboot() in containers, as
3294 supported by newer kernels. We will fall back to exit() if
3295 CAP_SYS_REBOOT is not available to the container. Also,
3296 nspawn makes use of this now and will actually reboot the
3297 container if the containerized OS asks for that.
3299 * journalctl will only show local log output by default
3300 now. Use --merge (-m) to show remote log output, too.
3302 * libsystemd-journal gained the new sd_journal_get_usage()
3303 call to determine the current disk usage of all journal
3304 files. This is exposed in the new "journalctl --disk-usage"
3307 * journald gained a new configuration setting SplitMode= in
3308 journald.conf which may be used to control how user journals
3309 are split off. See journald.conf(5) for details.
3311 * A new condition type ConditionFileNotEmpty= has been added.
3313 * tmpfiles' "w" lines now support file globbing, to write
3314 multiple files at once.
3316 * We added Python bindings for the journal submission
3317 APIs. More Python APIs for a number of selected APIs will
3318 likely follow. Note that we intend to add native bindings
3319 only for the Python language, as we consider it common
3320 enough to deserve bindings shipped within systemd. There are
3321 various projects outside of systemd that provide bindings
3322 for languages such as PHP or Lua.
3324 * Many conditions will now resolve specifiers such as %i. In
3325 addition, PathChanged= and related directives of .path units
3326 now support specifiers as well.
3328 * There's now a new RPM macro definition for the system preset
3331 * journald will now warn if it ca not forward a message to the
3332 syslog daemon because its socket is full.
3334 * timedated will no longer write or process /etc/timezone,
3335 except on Debian. As we do not support late mounted /usr
3336 anymore /etc/localtime always being a symlink is now safe,
3337 and hence the information in /etc/timezone is not necessary
3340 * logind will now always reserve one VT for a text getty (VT6
3341 by default). Previously if more than 6 X sessions where
3342 started they took up all the VTs with auto-spawned gettys,
3343 so that no text gettys were available anymore.
3345 * udev will now automatically inform the btrfs kernel logic
3346 about btrfs RAID components showing up. This should make
3347 simple hotplug based btrfs RAID assembly work.
3349 * PID 1 will now increase its RLIMIT_NOFILE to 64K by default
3350 (but not for its children which will stay at the kernel
3351 default). This should allow setups with a lot more listening
3354 * systemd will now always pass the configured timezone to the
3355 kernel at boot. timedated will do the same when the timezone
3358 * logind's inhibition logic has been updated. By default,
3359 logind will now handle the lid switch, the power and sleep
3360 keys all the time, even in graphical sessions. If DEs want
3361 to handle these events on their own they should take the new
3362 handle-power-key, handle-sleep-key and handle-lid-switch
3363 inhibitors during their runtime. A simple way to achiveve
3364 that is to invoke the DE wrapped in an invocation of:
3366 systemd-inhibit --what=handle-power-key:handle-sleep-key:handle-lid-switch ...
3368 * Access to unit operations is now checked via SELinux taking
3369 the unit file label and client process label into account.
3371 * systemd will now notify the administrator in the journal
3372 when he over-mounts a non-empty directory.
3374 * There are new specifiers that are resolved in unit files,
3375 for the host name (%H), the machine ID (%m) and the boot ID
3378 Contributions from: Allin Cottrell, Auke Kok, Brandon Philips,
3379 Colin Guthrie, Colin Walters, Daniel J Walsh, Dave Reisner,
3380 Eelco Dolstra, Jan Engelhardt, Kay Sievers, Lennart
3381 Poettering, Lucas De Marchi, Lukas Nykryn, Mantas Mikulėnas,
3382 Martin Pitt, Matthias Clasen, Michael Olbrich, Pierre Schmitz,
3383 Shawn Landden, Thomas Hindoe Paaboel Andersen, Tom Gundersen,
3384 Václav Pavlín, Yin Kangkai, Zbigniew Jędrzejewski-Szmek
3388 * Support for reading structured kernel messages from
3389 /dev/kmsg has now been added and is enabled by default.
3391 * Support for reading kernel messages from /proc/kmsg has now
3392 been removed. If you want kernel messages in the journal
3393 make sure to run a recent kernel (>= 3.5) that supports
3394 reading structured messages from /dev/kmsg (see
3395 above). /proc/kmsg is now exclusive property of classic
3396 syslog daemons again.
3398 * The libudev API gained the new
3399 udev_device_new_from_device_id() call.
3401 * The logic for file system namespace (ReadOnlyDirectory=,
3402 ReadWriteDirectoy=, PrivateTmp=) has been reworked not to
3403 require pivot_root() anymore. This means fewer temporary
3404 directories are created below /tmp for this feature.
3406 * nspawn containers will now see and receive all submounts
3407 made on the host OS below the root file system of the
3410 * Forward Secure Sealing is now supported for Journal files,
3411 which provide cryptographical sealing of journal files so
3412 that attackers cannot alter log history anymore without this
3413 being detectable. Lennart will soon post a blog story about
3414 this explaining it in more detail.
3416 * There are two new service settings RestartPreventExitStatus=
3417 and SuccessExitStatus= which allow configuration of exit
3418 status (exit code or signal) which will be excepted from the
3419 restart logic, resp. consider successful.
3421 * journalctl gained the new --verify switch that can be used
3422 to check the integrity of the structure of journal files and
3423 (if Forward Secure Sealing is enabled) the contents of
3426 * nspawn containers will now be run with /dev/stdin, /dev/fd/
3427 and similar symlinks pre-created. This makes running shells
3428 as container init process a lot more fun.
3430 * The fstab support can now handle PARTUUID= and PARTLABEL=
3433 * A new ConditionHost= condition has been added to match
3434 against the hostname (with globs) and machine ID. This is
3435 useful for clusters where a single OS image is used to
3436 provision a large number of hosts which shall run slightly
3437 different sets of services.
3439 * Services which hit the restart limit will now be placed in a
3442 Contributions from: Bertram Poettering, Dave Reisner, Huang
3443 Hang, Kay Sievers, Lennart Poettering, Lukas Nykryn, Martin
3444 Pitt, Simon Peeters, Zbigniew Jędrzejewski-Szmek
3448 * When running in --user mode systemd will now become a
3449 subreaper (PR_SET_CHILD_SUBREAPER). This should make the ps
3450 tree a lot more organized.
3452 * A new PartOf= unit dependency type has been introduced that
3453 may be used to group services in a natural way.
3455 * "systemctl enable" may now be used to enable instances of
3458 * journalctl now prints error log levels in red, and
3459 warning/notice log levels in bright white. It also supports
3460 filtering by log level now.
3462 * cgtop gained a new -n switch (similar to top), to configure
3463 the maximum number of iterations to run for. It also gained
3464 -b, to run in batch mode (accepting no input).
3466 * The suffix ".service" may now be omitted on most systemctl
3467 command lines involving service unit names.
3469 * There's a new bus call in logind to lock all sessions, as
3470 well as a loginctl verb for it "lock-sessions".
3472 * libsystemd-logind.so gained a new call sd_journal_perror()
3473 that works similar to libc perror() but logs to the journal
3474 and encodes structured information about the error number.
3476 * /etc/crypttab entries now understand the new keyfile-size=
3479 * shutdown(8) now can send a (configurable) wall message when
3480 a shutdown is cancelled.
3482 * The mount propagation mode for the root file system will now
3483 default to "shared", which is useful to make containers work
3484 nicely out-of-the-box so that they receive new mounts from
3485 the host. This can be undone locally by running "mount
3486 --make-rprivate /" if needed.
3488 * The prefdm.service file has been removed. Distributions
3489 should maintain this unit downstream if they intend to keep
3490 it around. However, we recommend writing normal unit files
3491 for display managers instead.
3493 * Since systemd is a crucial part of the OS we will now
3494 default to a number of compiler switches that improve
3495 security (hardening) such as read-only relocations, stack
3496 protection, and suchlike.
3498 * The TimeoutSec= setting for services is now split into
3499 TimeoutStartSec= and TimeoutStopSec= to allow configuration
3500 of individual time outs for the start and the stop phase of
3503 Contributions from: Artur Zaprzala, Arvydas Sidorenko, Auke
3504 Kok, Bryan Kadzban, Dave Reisner, David Strauss, Harald Hoyer,
3505 Jim Meyering, Kay Sievers, Lennart Poettering, Mantas
3506 Mikulėnas, Martin Pitt, Michal Schmidt, Michal Sekletar, Peter
3507 Alfredsen, Shawn Landden, Simon Peeters, Terence Honles, Tom
3508 Gundersen, Zbigniew Jędrzejewski-Szmek
3512 * The journal and id128 C APIs are now fully documented as man
3515 * Extra safety checks have been added when transitioning from
3516 the initial RAM disk to the main system to avoid accidental
3519 * /etc/crypttab entries now understand the new keyfile-offset=
3522 * systemctl -t can now be used to filter by unit load state.
3524 * The journal C API gained the new sd_journal_wait() call to
3525 make writing synchronous journal clients easier.
3527 * journalctl gained the new -D switch to show journals from a
3530 * journalctl now displays a special marker between log
3531 messages of two different boots.
3533 * The journal is now explicitly flushed to /var via a service
3534 systemd-journal-flush.service, rather than implicitly simply
3535 by seeing /var/log/journal to be writable.
3537 * journalctl (and the journal C APIs) can now match for much
3538 more complex expressions, with alternatives and
3541 * When transitioning from the initial RAM disk to the main
3542 system we will now kill all processes in a killing spree to
3543 ensure no processes stay around by accident.
3545 * Three new specifiers may be used in unit files: %u, %h, %s
3546 resolve to the user name, user home directory resp. user
3547 shell. This is useful for running systemd user instances.
3549 * We now automatically rotate journal files if their data
3550 object hash table gets a fill level > 75%. We also size the
3551 hash table based on the configured maximum file size. This
3552 together should lower hash collisions drastically and thus
3553 speed things up a bit.
3555 * journalctl gained the new "--header" switch to introspect
3556 header data of journal files.
3558 * A new setting SystemCallFilters= has been added to services
3559 which may be used to apply blacklists or whitelists to
3560 system calls. This is based on SECCOMP Mode 2 of Linux 3.5.
3562 * nspawn gained a new --link-journal= switch (and quicker: -j)
3563 to link the container journal with the host. This makes it
3564 very easy to centralize log viewing on the host for all
3565 guests while still keeping the journal files separated.
3567 * Many bugfixes and optimizations
3569 Contributions from: Auke Kok, Eelco Dolstra, Harald Hoyer, Kay
3570 Sievers, Lennart Poettering, Malte Starostik, Paul Menzel, Rex
3571 Tsai, Shawn Landden, Tom Gundersen, Ville Skyttä, Zbigniew
3576 * Several tools now understand kernel command line arguments,
3577 which are only read when run in an initial RAM disk. They
3578 usually follow closely their normal counterparts, but are
3581 * There's a new tool to analyze the readahead files that are
3582 automatically generated at boot. Use:
3584 /usr/lib/systemd/systemd-readahead analyze /.readahead
3586 * We now provide an early debug shell on tty9 if this enabled. Use:
3588 systemctl enable debug-shell.service
3590 * All plymouth related units have been moved into the Plymouth
3591 package. Please make sure to upgrade your Plymouth version
3594 * systemd-tmpfiles now supports getting passed the basename of
3595 a configuration file only, in which case it will look for it
3596 in all appropriate directories automatically.
3598 * udevadm info now takes a /dev or /sys path as argument, and
3599 does the right thing. Example:
3601 udevadm info /dev/sda
3602 udevadm info /sys/class/block/sda
3604 * systemctl now prints a warning if a unit is stopped but a
3605 unit that might trigger it continues to run. Example: a
3606 service is stopped but the socket that activates it is left
3609 * "systemctl status" will now mention if the log output was
3610 shortened due to rotation since a service has been started.
3612 * The journal API now exposes functions to determine the
3613 "cutoff" times due to rotation.
3615 * journald now understands SIGUSR1 and SIGUSR2 for triggering
3616 immediately flushing of runtime logs to /var if possible,
3617 resp. for triggering immediate rotation of the journal
3620 * It is now considered an error if a service is attempted to
3621 be stopped that is not loaded.
3623 * XDG_RUNTIME_DIR now uses numeric UIDs instead of usernames.
3625 * systemd-analyze now supports Python 3
3627 * tmpfiles now supports cleaning up directories via aging
3628 where the first level dirs are always kept around but
3629 directories beneath it automatically aged. This is enabled
3630 by prefixing the age field with '~'.
3632 * Seat objects now expose CanGraphical, CanTTY properties
3633 which is required to deal with very fast bootups where the
3634 display manager might be running before the graphics drivers
3635 completed initialization.
3637 * Seat objects now expose a State property.
3639 * We now include RPM macros for service enabling/disabling
3640 based on the preset logic. We recommend RPM based
3641 distributions to make use of these macros if possible. This
3642 makes it simpler to reuse RPM spec files across
3645 * We now make sure that the collected systemd unit name is
3646 always valid when services log to the journal via
3649 * There's a new man page kernel-command-line(7) detailing all
3650 command line options we understand.
3652 * The fstab generator may now be disabled at boot by passing
3653 fstab=0 on the kernel command line.
3655 * A new kernel command line option modules-load= is now understood
3656 to load a specific kernel module statically, early at boot.
3658 * Unit names specified on the systemctl command line are now
3659 automatically escaped as needed. Also, if file system or
3660 device paths are specified they are automatically turned
3661 into the appropriate mount or device unit names. Example:
3663 systemctl status /home
3664 systemctl status /dev/sda
3666 * The SysVConsole= configuration option has been removed from
3667 system.conf parsing.
3669 * The SysV search path is no longer exported on the D-Bus
3672 * The Names= option is been removed from unit file parsing.
3674 * There's a new man page bootup(7) detailing the boot process.
3676 * Every unit and every generator we ship with systemd now
3677 comes with full documentation. The self-explanatory boot is
3680 * A couple of services gained "systemd-" prefixes in their
3681 name if they wrap systemd code, rather than only external
3682 code. Among them fsck@.service which is now
3683 systemd-fsck@.service.
3685 * The HaveWatchdog property has been removed from the D-Bus
3688 * systemd.confirm_spawn= on the kernel command line should now
3691 * There's a new man page crypttab(5) which details all options
3692 we actually understand.
3694 * systemd-nspawn gained a new --capability= switch to pass
3695 additional capabilities to the container.
3697 * timedated will now read known NTP implementation unit names
3698 from /usr/lib/systemd/ntp-units.d/*.list,
3699 systemd-timedated-ntp.target has been removed.
3701 * journalctl gained a new switch "-b" that lists log data of
3702 the current boot only.
3704 * The notify socket is in the abstract namespace again, in
3705 order to support daemons which chroot() at start-up.
3707 * There is a new Storage= configuration option for journald
3708 which allows configuration of where log data should go. This
3709 also provides a way to disable journal logging entirely, so
3710 that data collected is only forwarded to the console, the
3711 kernel log buffer or another syslog implementation.
3713 * Many bugfixes and optimizations
3715 Contributions from: Auke Kok, Colin Guthrie, Dave Reisner,
3716 David Strauss, Eelco Dolstra, Kay Sievers, Lennart Poettering,
3717 Lukas Nykryn, Michal Schmidt, Michal Sekletar, Paul Menzel,
3718 Shawn Landden, Tom Gundersen
3722 * "systemctl help <unit>" now shows the man page if one is
3725 * Several new man pages have been added.
3727 * MaxLevelStore=, MaxLevelSyslog=, MaxLevelKMsg=,
3728 MaxLevelConsole= can now be specified in
3729 journald.conf. These options allow reducing the amount of
3730 data stored on disk or forwarded by the log level.
3732 * TimerSlackNSec= can now be specified in system.conf for
3733 PID1. This allows system-wide power savings.
3735 Contributions from: Dave Reisner, Kay Sievers, Lauri Kasanen,
3736 Lennart Poettering, Malte Starostik, Marc-Antoine Perennou,
3741 * logind is now capable of (optionally) handling power and
3742 sleep keys as well as the lid switch.
3744 * journalctl now understands the syntax "journalctl
3745 /usr/bin/avahi-daemon" to get all log output of a specific
3748 * CapabilityBoundingSet= in system.conf now also influences
3749 the capability bound set of usermode helpers of the kernel.
3751 Contributions from: Daniel Drake, Daniel J. Walsh, Gert
3752 Michael Kulyk, Harald Hoyer, Jean Delvare, Kay Sievers,
3753 Lennart Poettering, Matthew Garrett, Matthias Clasen, Paul
3754 Menzel, Shawn Landden, Tero Roponen, Tom Gundersen
3758 * Note that we skipped 139 releases here in order to set the
3759 new version to something that is greater than both udev's
3760 and systemd's most recent version number.
3762 * udev: all udev sources are merged into the systemd source tree now.
3763 All future udev development will happen in the systemd tree. It
3764 is still fully supported to use the udev daemon and tools without
3765 systemd running, like in initramfs or other init systems. Building
3766 udev though, will require the *build* of the systemd tree, but
3767 udev can be properly *run* without systemd.
3769 * udev: /lib/udev/devices/ are not read anymore; systemd-tmpfiles
3770 should be used to create dead device nodes as workarounds for broken
3773 * udev: RUN+="socket:..." and udev_monitor_new_from_socket() is
3774 no longer supported. udev_monitor_new_from_netlink() needs to be
3775 used to subscribe to events.
3777 * udev: when udevd is started by systemd, processes which are left
3778 behind by forking them off of udev rules, are unconditionally cleaned
3779 up and killed now after the event handling has finished. Services or
3780 daemons must be started as systemd services. Services can be
3781 pulled-in by udev to get started, but they can no longer be directly
3782 forked by udev rules.
3784 * udev: the daemon binary is called systemd-udevd now and installed
3785 in /usr/lib/systemd/. Standalone builds or non-systemd systems need
3786 to adapt to that, create symlink, or rename the binary after building
3789 * libudev no longer provides these symbols:
3790 udev_monitor_from_socket()
3791 udev_queue_get_failed_list_entry()
3792 udev_get_{dev,sys,run}_path()
3793 The versions number was bumped and symbol versioning introduced.
3795 * systemd-loginctl and systemd-journalctl have been renamed
3796 to loginctl and journalctl to match systemctl.
3798 * The config files: /etc/systemd/systemd-logind.conf and
3799 /etc/systemd/systemd-journald.conf have been renamed to
3800 logind.conf and journald.conf. Package updates should rename
3801 the files to the new names on upgrade.
3803 * For almost all files the license is now LGPL2.1+, changed
3804 from the previous GPL2.0+. Exceptions are some minor stuff
3805 of udev (which will be changed to LGPL2.1 eventually, too),
3806 and the MIT licensed sd-daemon.[ch] library that is suitable
3807 to be used as drop-in files.
3809 * systemd and logind now handle system sleep states, in
3810 particular suspending and hibernating.
3812 * logind now implements a sleep/shutdown/idle inhibiting logic
3813 suitable for a variety of uses. Soonishly Lennart will blog
3814 about this in more detail.
3816 * var-run.mount and var-lock.mount are no longer provided
3817 (which prevously bind mounted these directories to their new
3818 places). Distributions which have not converted these
3819 directories to symlinks should consider stealing these files
3820 from git history and add them downstream.
3822 * We introduced the Documentation= field for units and added
3823 this to all our shipped units. This is useful to make it
3824 easier to explore the boot and the purpose of the various
3827 * All smaller setup units (such as
3828 systemd-vconsole-setup.service) now detect properly if they
3829 are run in a container and are skipped when
3830 appropriate. This guarantees an entirely noise-free boot in
3831 Linux container environments such as systemd-nspawn.
3833 * A framework for implementing offline system updates is now
3834 integrated, for details see:
3835 http://freedesktop.org/wiki/Software/systemd/SystemUpdates
3837 * A new service type Type=idle is available now which helps us
3838 avoiding ugly interleaving of getty output and boot status
3841 * There's now a system-wide CapabilityBoundingSet= option to
3842 globally reduce the set of capabilities for the
3843 system. This is useful to drop CAP_SYS_MKNOD, CAP_SYS_RAWIO,
3844 CAP_NET_RAW, CAP_SYS_MODULE, CAP_SYS_TIME, CAP_SYS_PTRACE or
3845 even CAP_NET_ADMIN system-wide for secure systems.
3847 * There are now system-wide DefaultLimitXXX= options to
3848 globally change the defaults of the various resource limits
3849 for all units started by PID 1.
3851 * Harald Hoyer's systemd test suite has been integrated into
3852 systemd which allows easy testing of systemd builds in qemu
3853 and nspawn. (This is really awesome! Ask us for details!)
3855 * The fstab parser is now implemented as generator, not inside
3858 * systemctl will now warn you if .mount units generated from
3859 /etc/fstab are out of date due to changes in fstab that
3860 have not been read by systemd yet.
3862 * systemd is now suitable for usage in initrds. Dracut has
3863 already been updated to make use of this. With this in place
3864 initrds get a slight bit faster but primarily are much
3865 easier to introspect and debug since "systemctl status" in
3866 the host system can be used to introspect initrd services,
3867 and the journal from the initrd is kept around too.
3869 * systemd-delta has been added, a tool to explore differences
3870 between user/admin configuration and vendor defaults.
3872 * PrivateTmp= now affects both /tmp and /var/tmp.
3874 * Boot time status messages are now much prettier and feature
3875 proper english language. Booting up systemd has never been
3878 * Read-ahead pack files now include the inode number of all
3879 files to pre-cache. When the inode changes the pre-caching
3880 is not attempted. This should be nicer to deal with updated
3881 packages which might result in changes of read-ahead
3884 * We now temporaritly lower the kernel's read_ahead_kb variable
3885 when collecting read-ahead data to ensure the kernel's
3886 built-in read-ahead does not add noise to our measurements
3887 of necessary blocks to pre-cache.
3889 * There's now RequiresMountsFor= to add automatic dependencies
3890 for all mounts necessary for a specific file system path.
3892 * MountAuto= and SwapAuto= have been removed from
3893 system.conf. Mounting file systems at boot has to take place
3896 * nspawn now learned a new switch --uuid= to set the machine
3897 ID on the command line.
3899 * nspawn now learned the -b switch to automatically search
3902 * vt102 is now the default TERM for serial TTYs, upgraded from
3905 * systemd-logind now works on VT-less systems.
3907 * The build tree has been reorganized. The individual
3908 components now have directories of their own.
3910 * A new condition type ConditionPathIsReadWrite= is now available.
3912 * nspawn learned the new -C switch to create cgroups for the
3913 container in other hierarchies.
3915 * We now have support for hardware watchdogs, configurable in
3918 * The scheduled shutdown logic now has a public API.
3920 * We now mount /tmp as tmpfs by default, but this can be
3921 masked and /etc/fstab can override it.
3923 * Since udisks does not make use of /media anymore we are not
3924 mounting a tmpfs on it anymore.
3926 * journalctl gained a new --local switch to only interleave
3927 locally generated journal files.
3929 * We can now load the IMA policy at boot automatically.
3931 * The GTK tools have been split off into a systemd-ui.
3933 Contributions from: Andreas Schwab, Auke Kok, Ayan George,
3934 Colin Guthrie, Daniel Mack, Dave Reisner, David Ward, Elan
3935 Ruusamäe, Frederic Crozat, Gergely Nagy, Guillermo Vidal,
3936 Hannes Reinecke, Harald Hoyer, Javier Jardón, Kay Sievers,
3937 Lennart Poettering, Lucas De Marchi, Léo Gillot-Lamure,
3938 Marc-Antoine Perennou, Martin Pitt, Matthew Monaco, Maxim
3939 A. Mikityanskiy, Michael Biebl, Michael Olbrich, Michal
3940 Schmidt, Nis Martensen, Patrick McCarty, Roberto Sassu, Shawn
3941 Landden, Sjoerd Simons, Sven Anders, Tollef Fog Heen, Tom
3946 * This is mostly a bugfix release
3948 * Support optional initialization of the machine ID from the
3949 KVM or container configured UUID.
3951 * Support immediate reboots with "systemctl reboot -ff"
3953 * Show /etc/os-release data in systemd-analyze output
3955 * Many bugfixes for the journal, including endianness fixes and
3956 ensuring that disk space enforcement works
3958 * sd-login.h is C++ comptaible again
3960 * Extend the /etc/os-release format on request of the Debian
3963 * We now refuse non-UTF8 strings used in various configuration
3964 and unit files. This is done to ensure we do not pass invalid
3965 data over D-Bus or expose it elsewhere.
3967 * Register Mimo USB Screens as suitable for automatic seat
3970 * Read SELinux client context from journal clients in a race
3973 * Reorder configuration file lookup order. /etc now always
3974 overrides /run in order to allow the administrator to always
3975 and unconditionally override vendor supplied or
3976 automatically generated data.
3978 * The various user visible bits of the journal now have man
3979 pages. We still lack man pages for the journal API calls
3982 * We now ship all man pages in HTML format again in the
3985 Contributions from: Dave Reisner, Dirk Eibach, Frederic
3986 Crozat, Harald Hoyer, Kay Sievers, Lennart Poettering, Marti
3987 Raudsepp, Michal Schmidt, Shawn Landden, Tero Roponen, Thierry
3992 * This is mostly a bugfix release
3994 * systems lacking /etc/os-release are no longer supported.
3996 * Various functionality updates to libsystemd-login.so
3998 * Track class of PAM logins to distuingish greeters from
4001 Contributions from: Kay Sievers, Lennart Poettering, Michael
4006 * This is an important bugfix release for v41.
4008 * Building man pages is now optional which should be useful
4009 for those building systemd from git but unwilling to install
4012 * Watchdog support for supervising services is now usable. In
4013 a future release support for hardware watchdogs
4014 (i.e. /dev/watchdog) will be added building on this.
4016 * Service start rate limiting is now configurable and can be
4017 turned off per service. When a start rate limit is hit a
4018 reboot can automatically be triggered.
4020 * New CanReboot(), CanPowerOff() bus calls in systemd-logind.
4022 Contributions from: Benjamin Franzke, Bill Nottingham,
4023 Frederic Crozat, Lennart Poettering, Michael Olbrich, Michal
4024 Schmidt, Michał Górny, Piotr Drąg
4028 * The systemd binary is installed /usr/lib/systemd/systemd now;
4029 An existing /sbin/init symlink needs to be adapted with the
4032 * The code that loads kernel modules has been ported to invoke
4033 libkmod directly, instead of modprobe. This means we do not
4034 support systems with module-init-tools anymore.
4036 * Watchdog support is now already useful, but still not
4039 * A new kernel command line option systemd.setenv= is
4040 understood to set system wide environment variables
4041 dynamically at boot.
4043 * We now limit the set of capabilities of systemd-journald.
4045 * We now set SIGPIPE to ignore by default, since it only is
4046 useful in shell pipelines, and has little use in general
4047 code. This can be disabled with IgnoreSIPIPE=no in unit
4050 Contributions from: Benjamin Franzke, Kay Sievers, Lennart
4051 Poettering, Michael Olbrich, Michal Schmidt, Tom Gundersen,
4056 * This is mostly a bugfix release
4058 * We now expose the reason why a service failed in the
4059 "Result" D-Bus property.
4061 * Rudimentary service watchdog support (will be completed over
4062 the next few releases.)
4064 * When systemd forks off in order execute some service we will
4065 now immediately changes its argv[0] to reflect which process
4066 it will execute. This is useful to minimize the time window
4067 with a generic argv[0], which makes bootcharts more useful
4069 Contributions from: Alvaro Soliverez, Chris Paulson-Ellis, Kay
4070 Sievers, Lennart Poettering, Michael Olbrich, Michal Schmidt,
4071 Mike Kazantsev, Ray Strode
4075 * This is mostly a test release, but incorporates many
4078 * New systemd-cgtop tool to show control groups by their
4081 * Linking against libacl for ACLs is optional again. If
4082 disabled, support tracking device access for active logins
4083 goes becomes unavailable, and so does access to the user
4084 journals by the respective users.
4086 * If a group "adm" exists, journal files are automatically
4087 owned by them, thus allow members of this group full access
4088 to the system journal as well as all user journals.
4090 * The journal now stores the SELinux context of the logging
4091 client for all entries.
4093 * Add C++ inclusion guards to all public headers
4095 * New output mode "cat" in the journal to print only text
4096 messages, without any meta data like date or time.
4098 * Include tiny X server wrapper as a temporary stop-gap to
4099 teach XOrg udev display enumeration. This is used by display
4100 managers such as gdm, and will go away as soon as XOrg
4101 learned native udev hotplugging for display devices.
4103 * Add new systemd-cat tool for executing arbitrary programs
4104 with STDERR/STDOUT connected to the journal. Can also act as
4105 BSD logger replacement, and does so by default.
4107 * Optionally store all locally generated coredumps in the
4108 journal along with meta data.
4110 * systemd-tmpfiles learnt four new commands: n, L, c, b, for
4111 writing short strings to files (for usage for /sys), and for
4112 creating symlinks, character and block device nodes.
4114 * New unit file option ControlGroupPersistent= to make cgroups
4115 persistent, following the mechanisms outlined in
4116 http://www.freedesktop.org/wiki/Software/systemd/PaxControlGroups
4118 * Support multiple local RTCs in a sane way
4120 * No longer monopolize IO when replaying readahead data on
4121 rotating disks, since we might starve non-file-system IO to
4122 death, since fanotify() will not see accesses done by blkid,
4125 * Do not show kernel threads in systemd-cgls anymore, unless
4126 requested with new -k switch.
4128 Contributions from: Dan Horák, Kay Sievers, Lennart
4129 Poettering, Michal Schmidt
4133 * This is mostly a test release, but incorporates many
4136 * The git repository moved to:
4137 git://anongit.freedesktop.org/systemd/systemd
4138 ssh://git.freedesktop.org/git/systemd/systemd
4140 * First release with the journal
4141 http://0pointer.de/blog/projects/the-journal.html
4143 * The journal replaces both systemd-kmsg-syslogd and
4144 systemd-stdout-bridge.
4146 * New sd_pid_get_unit() API call in libsystemd-logind
4148 * Many systemadm clean-ups
4150 * Introduce remote-fs-pre.target which is ordered before all
4151 remote mounts and may be used to start services before all
4154 * Added Mageia support
4156 * Add bash completion for systemd-loginctl
4158 * Actively monitor PID file creation for daemons which exit in
4159 the parent process before having finished writing the PID
4160 file in the daemon process. Daemons which do this need to be
4161 fixed (i.e. PID file creation must have finished before the
4162 parent exits), but we now react a bit more gracefully to them.
4164 * Add colourful boot output, mimicking the well-known output
4165 of existing distributions.
4167 * New option PassCredentials= for socket units, for
4168 compatibility with a recent kernel ABI breakage.
4170 * /etc/rc.local is now hooked in via a generator binary, and
4171 thus will no longer act as synchronization point during
4174 * systemctl list-unit-files now supports --root=.
4176 * systemd-tmpfiles now understands two new commands: z, Z for
4177 relabelling files according to the SELinux database. This is
4178 useful to apply SELinux labels to specific files in /sys,
4181 * Output of SysV services is now forwarded to both the console
4182 and the journal by default, not only just the console.
4184 * New man pages for all APIs from libsystemd-login.
4186 * The build tree got reorganized and a the build system is a
4187 lot more modular allowing embedded setups to specifically
4188 select the components of systemd they are interested in.
4190 * Support for Linux systems lacking the kernel VT subsystem is
4193 * configure's --with-rootdir= got renamed to
4194 --with-rootprefix= to follow the naming used by udev and
4197 * Unless specified otherwise we will now install to /usr instead
4198 of /usr/local by default.
4200 * Processes with '@' in argv[0][0] are now excluded from the
4201 final shut-down killing spree, following the logic explained
4203 http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons
4205 * All processes remaining in a service cgroup when we enter
4206 the START or START_PRE states are now killed with
4207 SIGKILL. That means it is no longer possible to spawn
4208 background processes from ExecStart= lines (which was never
4209 supported anyway, and bad style).
4211 * New PropagateReloadTo=/PropagateReloadFrom= options to bind
4212 reloading of units together.
4214 Contributions from: Bill Nottingham, Daniel J. Walsh, Dave
4215 Reisner, Dexter Morgan, Gregs Gregs, Jonathan Nieder, Kay
4216 Sievers, Lennart Poettering, Michael Biebl, Michal Schmidt,
4217 Michał Górny, Ran Benita, Thomas Jarosch, Tim Waugh, Tollef
4218 Fog Heen, Tom Gundersen, Zbigniew Jędrzejewski-Szmek