1 systemd System and Service Manager
5 (This changes list is very incomplete, and the release is
8 * Introduce a new API "sd-hwdb.h" for querying the hardware
9 metadata database. With this minimal interface one can query
10 and enumerate the udev hwdb, decoupled from the old libudev
11 library. libudev's interface for this is now only a wrapper
12 around sd-hwdb. A new tool systemd-hwdb has been added to
13 interface with and update the database.
15 * When any of systemd's tools copies files (for example due to
16 tmpfiles' C lines) a btrfs reflink will attempted first,
17 before bytewise copying is done.
19 * systemd-nspawn gained a new --ephemeral switch. When
20 specified a btrfs snapshot is taken of the container's root
21 directory, and immediately removed when the container
22 terminates again. Thus, a container can be started whose
23 changes never alter the container's root directory, and are
24 lost on container termination. This switch can also be used
25 for starting a container off the root file system of the
26 host without affecting the host OS. This switch is only
27 available on btrfs file systems.
29 * systemd-nspawn gained a new --template= switch. It takes the
30 path to a container tree to use as template for the tree
31 specified via --directory=, should that director be
32 missing. This allows instantiating containers dynamically,
33 on first run. This switch is only available on btrfs file
36 * When a .mount unit refers to a mount point on which multiple
37 mounts are stacked, and the .mount unit is stopped all of
38 the stacked mount points will now be unmounted until no
41 * systemd now has an explicit notion of supported and
42 unsupported unit types. Jobs enqueued for unsupported unit
43 types will now fail with an "unsupported" error code. More
44 specifically .swap, .automount and .device units are not
45 supported in containers, .busname units are not supported on
46 non-kdbus systems. .swap and .automount are also not
47 supported if their respective kernel compile time options
50 * machinectl gained support for two new "copy-from" and
51 "copy-to" commands for copying files from a running
52 container to the host or vice versa.
54 * machinectl gained support for a new "bind" command to bind
55 mount host directories into local containers. This is
56 currently only supported for nspawn containers.
58 * networkd gained support for configuring bridge forwarding
59 database entries (fdb) from .network files.
61 * A new tiny daemon "systemd-importd" has been added that can
62 download container images in tar, raw, qcow2 or dkr formats,
63 and make them available locally in /var/lib/machines, so
64 that they can run as nspawn containers. The daemon can GPG
65 verify the downloads (not supported for dkr, since it has no
66 provisions for verifying downloads). It will transparently
67 decompress bz2, xz, gzip compressed downloads if necessary,
68 and restore sparse files on disk. The daemon uses privilege
69 separation to ensure the actual download logic runs with
70 fewer privileges than the deamon itself. machinectl has
71 gained new commands "pull-tar", "pull-raw" and "pull-dkr" to
72 make the functionality of importd available to the
73 user. With this in place the Fedora and Ubuntu "Cloud"
74 images can be downloaded and booted as containers unmodified
75 (the Fedora images lack the appropriate GPG signature files
76 currently, so they cannot be verified, but this will change
77 soon, hopefully). Note that downloading images is currently
78 only fully supported on btrfs.
80 * machinectl is now able to list container images found in
81 /var/lib/machines, along with some metadata about sizes of
82 disk and similar. If the directory is located on btrfs and
83 quota is enabled, this includes quota display. A new command
84 "image-status" has been added that shows additional
85 information about images.
87 * machinectl is now able to clone container images
88 efficiently, if the underlying file system (btrfs) supports
89 it, with the new "machinectl list-images" command. It also
90 gained commands for renaming and removing images, as well as
91 marking them read-only or read-write (supported also on
94 * networkd gained support for collecting LLDP network
95 announcements, from hardware that supports this. This is
96 shown in networkctl output.
98 * systemd-run gained support for a new -t (--pty) switch for
99 invoking a binary on a pty whose input and output is
100 connected to the invoking terminal. This allows executing
101 processes as system services while interactively
102 communicating with them via the terminal. Most interestingly
103 this is supported across container boundaries. Invoking
104 "systemd-run -t /bin/bash" is an alternative to running a
105 full login session, the difference being that the former
106 will not register a session, nor go through the PAM session
109 * tmpfiles gained support for a new "v" line type for creating
110 btrfs subvolumes. If the underlying file system is a legacy
111 file system, this automatically degrades to creating a
112 normal directory. Among others /var/lib/machines is now
113 created like this at boot, should it be missing.
115 * The directory /var/lib/containers/ has been deprecated and
116 been replaced by /var/lib/machines. The term "machines" has
117 been used in the systemd context as generic term for both
118 VMs and containers, and hence appears more appropriate for
119 this, as the directory can also contain raw images bootable
122 * systemd-nspawn when invoked with -M but without --directory=
123 or --image= is now capable of searching for the container
124 root directory, subvolume or disk image automatically, in
125 /var/lib/machines. systemd-nspawn@.service has been updated
126 to make use of this, thus allowing it to be used for raw
129 * A new machines.target unit has been introduced that is
130 supposed to group all containers/VMs invoked as services on
131 the system. systemd-nspawn@.service has been updated to
134 * machinectl gained a new "start" command, for invoking a
135 container as a service. "machinectl start foo" is mostly
136 equivalent to "systemctl start systemd-nspawn@foo.service",
137 but handles escaping in a nicer way.
139 * systemd-nspawn will now mount most of the cgroupfs tree
140 read-only into each container, with the exception of the
141 container's own subtree in the name=systemd hierarchy.
143 * journald now sets the special FS_NOCOW file flag for its
144 journal files. This should improve performance on btrfs, by
145 avoiding heavy fragmentation when journald's write-pattern
146 is used on COW file systems. It degrades btrfs' data
147 integrity guarantees for the files to the same levels as for
148 ext3/ext4 however. This should be OK though as journald does
149 its own data integrity checks and all its objects are
150 checksummed on disk. Also, journald should handle btrfs disk
151 full events a lot more gracefully now, by processing SIGBUS
152 errors, and not relying on fallocate() anymore.
154 * When journald detects that journal files it is writing to
155 have been deleted it will immediately start new journal
158 * systemd now provides a way to store file descriptors
159 per-service in PID 1.This is useful for daemons to ensure
160 that fds they require are not lost during a daemon
161 restart. The fds are passed to the deamon on the next
162 invocation in the same way socket activation fds are
163 passed. This is now used by journald to ensure that the
164 various sockets connected to all the system's stdout/stderr
165 are not lost when journald is restarted. File descriptors
166 may be stored in PID 1 via the sd_pid_notify_with_fds() API,
167 an extension to sd_notify(). Note that a limit is enforced
168 on the number of fds a service can store in PID 1, and it
169 defaults to 0, so that no fds may be stored, unless this is
170 explicitly turned on.
172 * The default TERM variable to use for units connected to a
173 terminal, when no other value is explicitly is set is now
174 vt220 rather than vt102. This should be fairly safe still,
175 but allows PgUp/PgDn work.
177 * The /etc/crypttab option header= as known from Debian is now
180 * "loginctl user-status" and "loginctl session-status" will
181 now show the last 10 lines of log messages of the
182 user/session following the status output. Similar,
183 "machinectl status" will show the last 10 log lines
184 associated with a virtual machine or container
185 service. (Note that this is usually not the log messages
186 done in the VM/container itself, but simply what the
187 container manager logs. For nspawn this includes all console
190 * "loginctl session-status" without further argument will now
191 show the status of the session of the caller. Similar,
192 "lock-session", "unlock-session", "activate",
193 "enable-linger", "disable-linger" may now be called without
194 session/user parameter in which case they apply to the
195 caller's session/user.
197 * An X11 session scriptlet is now shipped that uploads
198 $DISPLAY and $XAUTHORITY into the environment of the systemd
199 --user daemon if a session begins. This should improve
200 compatibility with X11 enabled applications run as systemd
203 * Generators are now subject to masking via /etc and /run, the
204 same way as unit files.
206 * networkd .network files gained support for configuring
207 per-link IPv4/IPv6 packet forwarding as well as IPv4
208 masquerading. This is by default turned on for veth links to
209 containers, as registered by systemd-nspawn. This means that
210 nspawn containers run with --network-veth will now get
211 automatic routed access to the host's networks without any
212 further configuration or setup, as long as networkd runs on
215 * systemd-nspawn gained the --port= (-p) switch to expose TCP
216 or UDP posts of a container on the host. With this in place
217 it is possible to run containers with private veth links
218 (--network-veth), and have their functionality exposed on
219 the host as if their services were running directly on the
222 * systemd-nspawn's --network-beth switch now gained a short
223 version "-n", since with the changes above it is now truly
224 useful out-of-the-box. The systemd-nspawn@.service has been
225 updated to make use of it too by default.
227 * systemd-nspawn will now maintain a per-image R/W lock, to
228 ensure that the same image is not started more than once
229 writable. (It's OK to run an image multiple times
230 simultaneously in read-only mode.)
232 * systemd-nspawn's --image= option is now capable of
233 dissecting and booting MBR and GPT disk images that contain
234 only a single active Linux partition. Previously it
235 supported only GPT disk images with proper GPT type
236 IDs. This allows running cloud images from major
237 distributions directly with systemd-nspawn, without
240 * In addition to collecting mouse dpi data in the udev
241 hardware database, there's now support for collecting angle
242 information for mouse scroll wheels. The database is
243 supposed to guarantee similar scrolling behaviour on mice
244 that it knows about. There's also support for collecting
245 information about Touchpad types.
247 * udev's input_id built-in will now also collect touch screen
248 dimension data and attach it to probed devices.
250 * /etc/os-release gained support for a Distribution Privacy
253 * networkd gained support for creating "ipvlan", "gretap",
254 "ip6gre", "ip6gretap" and "ip6tnl" network devices.
256 * systemd-tmpfiles gained support for "a" lines for setting
259 * systemd-nspawn will now mount /tmp in the container to
260 tmpfs, automatically.
262 * systemd now exposes the memory.usage_in_bytes cgroup
263 attribute and shows it for each service in the "systemctl
264 status" output, if available.
266 * When the user presses Ctrl-Alt-Del more than 7x within 2s an
267 immediate reboot is triggered. This useful if shutdown is
268 hung and is unable to complete, to expedite the
269 operation. Note that this kind of reboot will still unmount
270 all file systems, and hence should not result in fsck being
273 * A .device unit for an optical block device will now be
274 considered active only when a medium is in the drive. Also,
275 mount units are now bound to their backing devices thus
276 triggering automatic unmounting when devices become
277 unavailable. With this in place systemd will now
278 automatically unmount left-over mounts when a CD-ROM is
279 ejected or an USB stick is yanked from the system.
281 * networkd-wait-online now has support for waiting for
282 specific interfaces only (with globbing), and for giving up
283 after a configurable timeout.
285 * networkd now exits when idle. It will be automatically
286 restarted as soon as interfaces show up, are removed or
287 change state. networkd will stay around as long as there is
288 at least one DHCP state machine or similar around, that keep
291 * networkd may now configure IPv6 link-local addressing in
292 addition to IPv4 link-local addressing.
294 * The IPv6 "token" for use in SLAAC may now be configured for
295 each .network interface in networkd.
297 * Routes configured with networkd may now be assigned a scope
300 * networkd's [Match] sections now support globbing and lists
301 of multiple space-separated matches per item.
305 * When querying unit file enablement status (for example via
306 "systemctl is-enabled"), a new state "indirect" is now known
307 which indicates that a unit might not be enabled itself, but
308 another unit listed in its Alias= setting might be.
310 * Similar to the various existing ConditionXYZ= settings for
311 units there are now matching AssertXYZ= settings. While
312 failing conditions cause a unit to be skipped, but its job
313 to succeed, failing assertions declared like this will cause
314 a unit start operation and its job to fail.
316 * hostnamed now knows a new chassis type "embedded".
318 * systemctl gained a new "edit" command. When used on a unit
319 file this allows extending unit files with .d/ drop-in
320 configuration snippets or editing the full file (after
321 copying it from /usr/lib to /etc). This will invoke the
322 user's editor (as configured with $EDITOR), and reload the
323 modified configuration after editing.
325 * "systemctl status" now shows the suggested enablement state
326 for a unit, as declared in the (usually vendor-supplied)
329 * nss-myhostname will now resolve the single-label host name
330 "gateway" to the locally configured default IP routing
331 gateways, ordered by their metrics. This assigns a stable
332 name to the used gateways, regardless which ones are
333 currently configured. Note that the name will only be
334 resolved after all other name sources (if nss-myhostname is
335 configured properly) and should hence not negatively impact
336 systems that use the single-label host name "gateway" in
339 * systemd-inhibit now allows filtering by mode when listing
342 * Scope and service units gained a new "Delegate" boolean
343 property, which when set allows processes running inside the
344 unit to further partition resources. This is primarily
345 useful for systemd user instances as well as container
348 * journald will now pick up audit messages directly from
349 the kernel, and log them like any other log message. The
350 audit fields are split up and fully indexed. This means that
351 journalctl in many ways is now a (nicer!) alternative to
352 ausearch, the traditional audit client. Note that this
353 implements only a minimal audit client, if you want the
354 special audit modes like reboot-on-log-overflow, please use
355 the traditional auditd instead, which can be used in
356 parallel to journald.
358 * The ConditionSecurity= unit file option now understands the
359 special string "audit" to check whether auditing is
362 * journalctl gained two new commands --vacuum-size= and
363 --vacuum-time= to delete old journal files until the
364 remaining ones take up no more the specified size on disk,
365 or are not older than the specified time.
367 * A new, native PPPoE library has been added to sd-network,
368 systemd's library of light-weight networking protocols. This
369 library will be used in a future version of networkd to
370 enable PPPoE communication without an external pppd daemon.
372 * The busctl tool now understands a new "capture" verb that
373 works similar to "monitor", but writes a packet capture
374 trace to STDOUT that can be redirected to a file which is
375 compatible with libcap's capture file format. This can then
376 be loaded in Wireshark and similar tools to inspect bus
379 * The busctl tool now understands a new "tree" verb that shows
380 the object trees of a specific service on the bus, or of all
383 * The busctl tool now understands a new "introspect" verb that
384 shows all interfaces and members of objects on the bus,
385 including their signature and values. This is particularly
386 useful to get more information about bus objects shown by
387 the new "busctl tree" command.
389 * The busctl tool now understands new verbs "call",
390 "set-property" and "get-property" for invoking bus method
391 calls, setting and getting bus object properties in a
394 * busctl gained a new --augment-creds= argument that controls
395 whether the tool shall augment credential information it
396 gets from the bus with data from /proc, in a possibly
399 * nspawn's --link-journal= switch gained two new values
400 "try-guest" and "try-host" that work like "guest" and
401 "host", but do not fail if the host has no persistent
402 journalling enabled. -j is now equivalent to
403 --link-journal=try-guest.
405 * macvlan network devices created by nspawn will now have
406 stable MAC addresses.
408 * A new SmackProcessLabel= unit setting has been added, which
409 controls the SMACK security label processes forked off by
410 the respective unit shall use.
412 * If compiled with --enable-xkbcommon, systemd-localed will
413 verify x11 keymap settings by compiling the given keymap. It
414 will spew out warnings if the compilation fails. This
415 requires libxkbcommon to be installed.
417 * When a coredump is collected a larger number of metadata
418 fields is now collected and included in the journal records
419 created for it. More specifically control group membership,
420 environment variables, memory maps, working directory,
421 chroot directory, /proc/$PID/status, and a list of open file
422 descriptors is now stored in the log entry.
424 * The udev hwdb now contains DPI information for mice. For
427 http://who-t.blogspot.de/2014/12/building-a-dpi-database-for-mice.html
429 * All systemd programs that read standalone configuration
430 files in /etc now also support a corresponding series of
431 .conf.d configuration directories in /etc/, /run/,
432 /usr/local/lib/, /usr/lib/, and (if configured with
433 --enable-split-usr) /lib/. In particular, the following
434 configuration files now have corresponding configuration
435 directories: system.conf user.conf, logind.conf,
436 journald.conf, sleep.conf, bootchart.conf, coredump.conf,
437 resolved.conf, timesyncd.conf, journal-remote.conf, and
438 journal-upload.conf. Note that distributions should use the
439 configuration directories in /usr/lib/; the directories in
440 /etc/ are reserved for the system administrator.
442 * systemd-rfkill will no longer take the rfkill device name
443 into account when storing rfkill state on disk, as the name
444 might be dynamically assigned and not stable. Instead, the
445 ID_PATH udev variable combined with the rfkill type (wlan,
446 bluetooth, ...) is used.
448 * A new service systemd-machine-id-commit.service has been
449 added. When used on systems where /etc is read-only during
450 boot, and /etc/machine-id is not initialized (but an empty
451 file), this service will copy the temporary machine ID
452 created as replacement into /etc after the system is fully
453 booted up. This is useful for systems that are freshly
454 installed with a non-initialized machine ID, but should get
455 a fixed machine ID for subsequent boots.
457 * networkd's .netdev files now provide a large set of
458 configuration parameters for VXLAN devices. Similar, the
459 bridge port cost parameter is now configurable in .network
460 files. There's also new support for configuring IP source
461 routing. networkd .link files gained support for a new
462 OriginalName= match that is useful to match against the
463 original interface name the kernel assigned. .network files
464 may include MTU= and MACAddress= fields for altering the MTU
465 and MAC address while being connected to a specific network
468 * The LUKS logic gained supported for configuring
469 UUID-specific key files. There's also new support for naming
470 LUKS device from the kernel command line, using the new
473 * Timer units may now be transiently created via the bus API
474 (this was previously already available for scope and service
475 units). In addition it is now possible to create multiple
476 transient units at the same time with a single bus call. The
477 "systemd-run" tool has been updated to make use of this for
478 running commands on a specified time, in at(1)-style.
480 * tmpfiles gained support for "t" lines, for assigning
481 extended attributes to files. Among other uses this may be
482 used to assign SMACK labels to files.
484 Contributions from: Alin Rauta, Alison Chaiken, Andrej
485 Manduch, Bastien Nocera, Chris Atkinson, Chris Leech, Chris
486 Mayo, Colin Guthrie, Colin Walters, Cristian Rodríguez,
487 Daniele Medri, Daniel Mack, Dan Williams, Dan Winship, Dave
488 Reisner, David Herrmann, Didier Roche, Felipe Sateler, Gavin
489 Li, Hans de Goede, Harald Hoyer, Iago López Galeiras, Ivan
490 Shapovalov, Jakub Filak, Jan Janssen, Jan Synacek, Joe
491 Lawrence, Josh Triplett, Kay Sievers, Lennart Poettering,
492 Lukas Nykryn, Łukasz Stelmach, Maciej Wereski, Mantas
493 Mikulėnas, Marcel Holtmann, Martin Pitt, Maurizio Lombardi,
494 Michael Biebl, Michael Chapman, Michael Marineau, Michal
495 Schmidt, Michal Sekletar, Olivier Brunel, Patrik Flykt, Peter
496 Hutterer, Przemyslaw Kedzierski, Rami Rosen, Ray Strode,
497 Richard Schütz, Richard W.M. Jones, Ronny Chevalier, Ross
498 Lagerwall, Sean Young, Stanisław Pitucha, Susant Sahani,
499 Thomas Haller, Thomas Hindoe Paaboel Andersen, Tom Gundersen,
500 Torstein Husebø, Umut Tezduyar Lindskog, Vicente Olivert
501 Riera, WaLyong Cho, Wesley Dawson, Zbigniew Jędrzejewski-Szmek
503 -- Berlin, 2014-12-10
507 * journalctl gained the new options -t/--identifier= to match
508 on the syslog identifier (aka "tag"), as well as --utc to
509 show log timestamps in the UTC timezone. journalctl now also
510 accepts -n/--lines=all to disable line capping in a pager.
512 * journalctl gained a new switch, --flush, that synchronously
513 flushes logs from /run/log/journal to /var/log/journal if
514 persistent storage is enabled. systemd-journal-flush.service
515 now waits until the operation is complete.
517 * Services can notify the manager before they start a reload
518 (by sending RELOADING=1) or shutdown (by sending
519 STOPPING=1). This allows the manager to track and show the
520 internal state of daemons and closes a race condition when
521 the process is still running but has closed its D-Bus
524 * Services with Type=oneshot do not have to have any ExecStart
527 * User units are now loaded also from
528 $XDG_RUNTIME_DIR/systemd/user/. This is similar to the
529 /run/systemd/user directory that was already previously
530 supported, but is under the control of the user.
532 * Job timeouts (i.e. time-outs on the time a job that is
533 queued stays in the run queue) can now optionally result in
534 immediate reboot or power-off actions (JobTimeoutAction= and
535 JobTimeoutRebootArgument=). This is useful on ".target"
536 units, to limit the maximum time a target remains
537 undispatched in the run queue, and to trigger an emergency
538 operation in such a case. This is now used by default to
539 turn off the system if boot-up (as defined by everything in
540 basic.target) hangs and does not complete for at least
541 15min. Also, if power-off or reboot hang for at least 30min
542 an immediate power-off/reboot operation is triggered. This
543 functionality is particularly useful to increase reliability
544 on embedded devices, but also on laptops which might
545 accidentally get powered on when carried in a backpack and
546 whose boot stays stuck in a hard disk encryption passphrase
549 * systemd-logind can be configured to also handle lid switch
550 events even when the machine is docked or multiple displays
551 are attached (HandleLidSwitchDocked= option).
553 * A helper binary and a service have been added which can be
554 used to resume from hibernation in the initramfs. A
555 generator will parse the resume= option on the kernel
556 command line to trigger resume.
558 * A user console daemon systemd-consoled has been
559 added. Currently, it is a preview, and will so far open a
560 single terminal on each session of the user marked as
561 Desktop=systemd-console.
563 * Route metrics can be specified for DHCP routes added by
566 * The SELinux context of socket-activated services can be set
567 from the information provided by the networking stack
568 (SELinuxContextFromNet= option).
570 * Userspace firmware loading support has been removed and
571 the minimum supported kernel version is thus bumped to 3.7.
573 * Timeout for udev workers has been increased from 1 to 3
574 minutes, but a warning will be printed after 1 minute to
575 help diagnose kernel modules that take a long time to load.
577 * Udev rules can now remove tags on devices with TAG-="foobar".
579 * systemd's readahead implementation has been removed. In many
580 circumstances it didn't give expected benefits even for
581 rotational disk drives and was becoming less relevant in the
582 age of SSDs. As none of the developers has been using
583 rotating media anymore, and nobody stepped up to actively
584 maintain this component of systemd it has now been removed.
586 * Swap units can use Options= to specify discard options.
587 Discard options specified for swaps in /etc/fstab are now
590 * Docker containers are now detected as a separate type of
593 * The Password Agent protocol gained support for queries where
594 the user input is shown, useful e.g. for user names.
595 systemd-ask-password gained a new --echo option to turn that
598 * The default sysctl.d/ snippets will now set:
600 net.core.default_qdisc = fq_codel
602 This selects Fair Queuing Controlled Delay as the default
603 queuing discipline for network interfaces. fq_codel helps
604 fight the network bufferbloat problem. It is believed to be
605 a good default with no tuning required for most workloads.
606 Downstream distributions may override this choice. On 10Gbit
607 servers that do not do forwarding, "fq" may perform better.
608 Systems without a good clocksource should use "pfifo_fast".
610 * If kdbus is enabled during build a new option BusPolicy= is
611 available for service units, that allows locking all service
612 processes into a stricter bus policy, in order to limit
613 access to various bus services, or even hide most of them
614 from the service's view entirely.
616 * networkctl will now show the .network and .link file
617 networkd has applied to a specific interface.
619 * sd-login gained a new API call sd_session_get_desktop() to
620 query which desktop environment has been selected for a
623 * UNIX utmp support is now compile-time optional to support
626 * systemctl gained two new commands "add-wants" and
627 "add-requires" for pulling in units from specific targets
630 * If the word "rescue" is specified on the kernel command line
631 the system will now boot into rescue mode (aka
632 rescue.target), which was previously available only by
633 specifying "1" or "systemd.unit=rescue.target" on the kernel
634 command line. This new kernel command line option nicely
635 mirrors the already existing "emergency" kernel command line
638 * New kernel command line options mount.usr=, mount.usrflags=,
639 mount.usrfstype= have been added that match root=, rootflags=,
640 rootfstype= but allow mounting a specific file system to
643 * The $NOTIFY_SOCKET is now also passed to control processes of
644 services, not only the main process.
646 * This version reenables support for fsck's -l switch. This
647 means at least version v2.25 of util-linux is required for
648 operation, otherwise dead-locks on device nodes may
649 occur. Again: you need to update util-linux to at least
650 v2.25 when updating systemd to v217.
652 * The "multi-seat-x" tool has been removed from systemd, as
653 its functionality has been integrated into X servers 1.16,
654 and the tool is hence redundant. It is recommended to update
655 display managers invoking this tool to simply invoke X
656 directly from now on, again.
658 * Support for the new ALLOW_INTERACTIVE_AUTHORIZATION D-Bus
659 message flag has been added for all of systemd's PolicyKit
660 authenticated method calls has been added. In particular
661 this now allows optional interactive authorization via
662 PolicyKit for many of PID1's privileged operations such as
663 unit file enabling and disabling.
665 * "udevadm hwdb --update" learnt a new switch "--usr" for
666 placing the rebuilt hardware database in /usr instead of
667 /etc. When used only hardware database entries stored in
668 /usr will be used, and any user database entries in /etc are
669 ignored. This functionality is useful for vendors to ship a
670 pre-built database on systems where local configuration is
671 unnecessary or unlikely.
673 * Calendar time specifications in .timer units now also
674 understand the strings "semi-annually", "quarterly" and
675 "minutely" as shortcuts (in addition to the preexisting
676 "anually", "hourly", ...).
678 * systemd-tmpfiles will now correctly create files in /dev
679 at boot which are marked for creation only at boot. It is
680 recommended to always create static device nodes with 'c!'
681 and 'b!', so that they are created only at boot and not
682 overwritten at runtime.
684 * When the watchdog logic is used for a service (WatchdogSec=)
685 and the watchdog timeout is hit the service will now be
686 terminated with SIGABRT (instead of just SIGTERM), in order
687 to make sure a proper coredump and backtrace is
688 generated. This ensures that hanging services will result in
689 similar coredump/backtrace behaviour as services that hit a
692 Contributions from: Andreas Henriksson, Andrei Borzenkov,
693 Angus Gibson, Ansgar Burchardt, Ben Wolsieffer, Brandon L.
694 Black, Christian Hesse, Cristian Rodríguez, Daniel Buch,
695 Daniele Medri, Daniel Mack, Dan Williams, Dave Reisner, David
696 Herrmann, David Sommerseth, David Strauss, Emil Renner
697 Berthing, Eric Cook, Evangelos Foutras, Filipe Brandenburger,
698 Gustavo Sverzut Barbieri, Hans de Goede, Harald Hoyer, Hristo
699 Venev, Hugo Grostabussiat, Ivan Shapovalov, Jan Janssen, Jan
700 Synacek, Jonathan Liu, Juho Son, Karel Zak, Kay Sievers, Klaus
701 Purer, Koen Kooi, Lennart Poettering, Lukas Nykryn, Lukasz
702 Skalski, Łukasz Stelmach, Mantas Mikulėnas, Marcel Holtmann,
703 Marius Tessmann, Marko Myllynen, Martin Pitt, Michael Biebl,
704 Michael Marineau, Michael Olbrich, Michael Scherer, Michal
705 Schmidt, Michal Sekletar, Miroslav Lichvar, Patrik Flykt,
706 Philippe De Swert, Piotr Drąg, Rahul Sundaram, Richard
707 Weinberger, Robert Milasan, Ronny Chevalier, Ruben Kerkhof,
708 Santiago Vila, Sergey Ptashnick, Simon McVittie, Sjoerd
709 Simons, Stefan Brüns, Steven Allen, Steven Noonan, Susant
710 Sahani, Sylvain Plantefève, Thomas Hindoe Paaboel Andersen,
711 Timofey Titovets, Tobias Hunger, Tom Gundersen, Torstein
712 Husebø, Umut Tezduyar Lindskog, WaLyong Cho, Zbigniew
715 -- Berlin, 2014-10-28
719 * timedated no longer reads NTP implementation unit names from
720 /usr/lib/systemd/ntp-units.d/*.list. Alternative NTP
721 implementations should add a
723 Conflicts=systemd-timesyncd.service
725 to their unit files to take over and replace systemd's NTP
726 default functionality.
728 * systemd-sysusers gained a new line type "r" for configuring
729 which UID/GID ranges to allocate system users/groups
730 from. Lines of type "u" may now add an additional column
731 that specifies the home directory for the system user to be
732 created. Also, systemd-sysusers may now optionally read user
733 information from STDIN instead of a file. This is useful for
734 invoking it from RPM preinst scriptlets that need to create
735 users before the first RPM file is installed since these
736 files might need to be owned by them. A new
737 %sysusers_create_inline RPM macro has been introduced to do
738 just that. systemd-sysusers now updates the shadow files as
739 well as the user/group databases, which should enhance
740 compatibility with certain tools like grpck.
742 * A number of bus APIs of PID 1 now optionally consult
743 PolicyKit to permit access for otherwise unprivileged
744 clients under certain conditions. Note that this currently
745 doesn't support interactive authentication yet, but this is
746 expected to be added eventually, too.
748 * /etc/machine-info now has new fields for configuring the
749 deployment environment of the machine, as well as the
750 location of the machine. hostnamectl has been updated with
751 new command to update these fields.
753 * systemd-timesyncd has been updated to automatically acquire
754 NTP server information from systemd-networkd, which might
755 have been discovered via DHCP.
757 * systemd-resolved now includes a caching DNS stub resolver
758 and a complete LLMNR name resolution implementation. A new
759 NSS module "nss-resolve" has been added which can be used
760 instead of glibc's own "nss-dns" to resolve hostnames via
761 systemd-resolved. Hostnames, addresses and arbitrary RRs may
762 be resolved via systemd-resolved D-Bus APIs. In contrast to
763 the glibc internal resolver systemd-resolved is aware of
764 multi-homed system, and keeps DNS server and caches separate
765 and per-interface. Queries are sent simultaneously on all
766 interfaces that have DNS servers configured, in order to
767 properly handle VPNs and local LANs which might resolve
768 separate sets of domain names. systemd-resolved may acquire
769 DNS server information from systemd-networkd automatically,
770 which in turn might have discovered them via DHCP. A tool
771 "systemd-resolve-host" has been added that may be used to
772 query the DNS logic in resolved. systemd-resolved implements
773 IDNA and automatically uses IDNA or UTF-8 encoding depending
774 on whether classic DNS or LLMNR is used as transport. In the
775 next releases we intend to add a DNSSEC and mDNS/DNS-SD
776 implementation to systemd-resolved.
778 * A new NSS module nss-mymachines has been added, that
779 automatically resolves the names of all local registered
780 containers to their respective IP addresses.
782 * A new client tool "networkctl" for systemd-networkd has been
783 added. It currently is entirely passive and will query
784 networking configuration from udev, rtnetlink and networkd,
785 and present it to the user in a very friendly
786 way. Eventually, we hope to extend it to become a full
787 control utility for networkd.
789 * .socket units gained a new DeferAcceptSec= setting that
790 controls the kernels' TCP_DEFER_ACCEPT sockopt for
791 TCP. Similar, support for controlling TCP keep-alive
792 settings has been added (KeepAliveTimeSec=,
793 KeepAliveIntervalSec=, KeepAliveProbes=). Also, support for
794 turning off Nagle's algorithm on TCP has been added
797 * logind learned a new session type "web", for use in projects
798 like Cockpit which register web clients as PAM sessions.
800 * timer units with at least one OnCalendar= setting will now
801 be started only after timer-sync.target has been
802 reached. This way they will not elapse before the system
803 clock has been corrected by a local NTP client or
804 similar. This is particular useful on RTC-less embedded
805 machines, that come up with an invalid system clock.
807 * systemd-nspawn's --network-veth= switch should now result in
808 stable MAC addresses for both the outer and the inner side
811 * systemd-nspawn gained a new --volatile= switch for running
812 container instances with /etc or /var unpopulated.
814 * The kdbus client code has been updated to use the new Linux
815 3.17 memfd subsystem instead of the old kdbus-specific one.
817 * systemd-networkd's DHCP client and server now support
818 FORCERENEW. There are also new configuration options to
819 configure the vendor client identifier and broadcast mode
822 * systemd will no longer inform the kernel about the current
823 timezone, as this is necessarily incorrect and racy as the
824 kernel has no understanding of DST and similar
825 concepts. This hence means FAT timestamps will be always
826 considered UTC, similar to what Android is already
827 doing. Also, when the RTC is configured to the local time
828 (rather than UTC) systemd will never synchronize back to it,
829 as this might confuse Windows at a later boot.
831 * systemd-analyze gained a new command "verify" for offline
832 validation of unit files.
834 * systemd-networkd gained support for a couple of additional
835 settings for bonding networking setups. Also, the metric for
836 statically configured routes may now be configured. For
837 network interfaces where this is appropriate the peer IP
838 address may now be configured.
840 * systemd-networkd's DHCP client will no longer request
841 broadcasting by default, as this tripped up some networks.
842 For hardware where broadcast is required the feature should
843 be switched back on using RequestBroadcast=yes.
845 * systemd-networkd will now set up IPv4LL addresses (when
846 enabled) even if DHCP is configured successfully.
848 * udev will now default to respect network device names given
849 by the kernel when the kernel indicates that these are
850 predictable. This behavior can be tweaked by changing
851 NamePolicy= in the relevant .link file.
853 * A new library systemd-terminal has been added that
854 implements full TTY stream parsing and rendering. This
855 library is supposed to be used later on for implementing a
856 full userspace VT subsystem, replacing the current kernel
859 * A new tool systemd-journal-upload has been added to push
860 journal data to a remote system running
861 systemd-journal-remote.
863 * journald will no longer forward all local data to another
864 running syslog daemon. This change has been made because
865 rsyslog (which appears to be the most commonly used syslog
866 implementation these days) no longer makes use of this, and
867 instead pulls the data out of the journal on its own. Since
868 forwarding the messages to a non-existent syslog server is
869 more expensive than we assumed we have now turned this
870 off. If you run a syslog server that is not a recent rsyslog
871 version, you have to turn this option on again
872 (ForwardToSyslog= in journald.conf).
874 * journald now optionally supports the LZ4 compressor for
875 larger journal fields. This compressor should perform much
876 better than XZ which was the previous default.
878 * machinectl now shows the IP addresses of local containers,
879 if it knows them, plus the interface name of the container.
881 * A new tool "systemd-escape" has been added that makes it
882 easy to escape strings to build unit names and similar.
884 * sd_notify() messages may now include a new ERRNO= field
885 which is parsed and collected by systemd and shown among the
886 "systemctl status" output for a service.
888 * A new component "systemd-firstboot" has been added that
889 queries the most basic systemd information (timezone,
890 hostname, root password) interactively on first
891 boot. Alternatively it may also be used to provision these
892 things offline on OS images installed into directories.
894 * The default sysctl.d/ snippets will now set
896 net.ipv4.conf.default.promote_secondaries=1
898 This has the benefit of no flushing secondary IP addresses
899 when primary addresses are removed.
901 Contributions from: Ansgar Burchardt, Bastien Nocera, Colin
902 Walters, Dan Dedrick, Daniel Buch, Daniel Korostil, Daniel
903 Mack, Dan Williams, Dave Reisner, David Herrmann, Denis
904 Kenzior, Eelco Dolstra, Eric Cook, Hannes Reinecke, Harald
905 Hoyer, Hong Shick Pak, Hui Wang, Jean-André Santoni, Jóhann
906 B. Guðmundsson, Jon Severinsson, Karel Zak, Kay Sievers, Kevin
907 Wells, Lennart Poettering, Lukas Nykryn, Mantas Mikulėnas,
908 Marc-Antoine Perennou, Martin Pitt, Michael Biebl, Michael
909 Marineau, Michael Olbrich, Michal Schmidt, Michal Sekletar,
910 Miguel Angel Ajo, Mike Gilbert, Olivier Brunel, Robert
911 Schiele, Ronny Chevalier, Simon McVittie, Sjoerd Simons, Stef
912 Walter, Steven Noonan, Susant Sahani, Tanu Kaskinen, Thomas
913 Blume, Thomas Hindoe Paaboel Andersen, Timofey Titovets,
914 Tobias Geerinckx-Rice, Tomasz Torcz, Tom Gundersen, Umut
915 Tezduyar Lindskog, Zbigniew Jędrzejewski-Szmek
917 -- Berlin, 2014-08-19
921 * A new tool systemd-sysusers has been added. This tool
922 creates system users and groups in /etc/passwd and
923 /etc/group, based on static declarative system user/group
924 definitions in /usr/lib/sysusers.d/. This is useful to
925 enable factory resets and volatile systems that boot up with
926 an empty /etc directory, and thus need system users and
927 groups created during early boot. systemd now also ships
928 with two default sysusers.d/ files for the most basic
929 users and groups systemd and the core operating system
932 * A new tmpfiles snippet has been added that rebuilds the
933 essential files in /etc on boot, should they be missing.
935 * A directive for ensuring automatic clean-up of
936 /var/cache/man/ has been removed from the default
937 configuration. This line should now be shipped by the man
938 implementation. The necessary change has been made to the
939 man-db implementation. Note that you need to update your man
940 implementation to one that ships this line, otherwise no
941 automatic clean-up of /var/cache/man will take place.
943 * A new condition ConditionNeedsUpdate= has been added that
944 may conditionalize services to only run when /etc or /var
945 are "older" than the vendor operating system resources in
946 /usr. This is useful for reconstructing or updating /etc
947 after an offline update of /usr or a factory reset, on the
948 next reboot. Services that want to run once after such an
949 update or reset should use this condition and order
950 themselves before the new systemd-update-done.service, which
951 will mark the two directories as fully updated. A number of
952 service files have been added making use of this, to rebuild
953 the udev hardware database, the journald message catalog and
954 dynamic loader cache (ldconfig). The systemd-sysusers tool
955 described above also makes use of this now. With this in
956 place it is now possible to start up a minimal operating
957 system with /etc empty cleanly. For more information on the
958 concepts involved see this recent blog story:
960 http://0pointer.de/blog/projects/stateless.html
962 * A new system group "input" has been introduced, and all
963 input device nodes get this group assigned. This is useful
964 for system-level software to get access to input devices. It
965 complements what is already done for "audio" and "video".
967 * systemd-networkd learnt minimal DHCPv4 server support in
968 addition to the existing DHCPv4 client support. It also
969 learnt DHCPv6 client and IPv6 Router Solicitation client
970 support. The DHCPv4 client gained support for static routes
971 passed in from the server. Note that the [DHCPv4] section
972 known in older systemd-networkd versions has been renamed to
973 [DHCP] and is now also used by the DHCPv6 client. Existing
974 .network files using settings of this section should be
975 updated, though compatibility is maintained. Optionally, the
976 client hostname may now be sent to the DHCP server.
978 * networkd gained support for vxlan virtual networks as well
979 as tun/tap and dummy devices.
981 * networkd gained support for automatic allocation of address
982 ranges for interfaces from a system-wide pool of
983 addresses. This is useful for dynamically managing a large
984 number of interfaces with a single network configuration
985 file. In particular this is useful to easily assign
986 appropriate IP addresses to the veth links of a large number
989 * RPM macros for processing sysusers, sysctl and binfmt
990 drop-in snippets at package installation time have been
993 * The /etc/os-release file should now be placed in
994 /usr/lib/os-release. The old location is automatically
995 created as symlink. /usr/lib is the more appropriate
996 location of this file, since it shall actually describe the
997 vendor operating system shipped in /usr, and not the
998 configuration stored in /etc.
1000 * .mount units gained a new boolean SloppyOptions= setting
1001 that maps to mount(8)'s -s option which enables permissive
1002 parsing of unknown mount options.
1004 * tmpfiles learnt a new "L+" directive which creates a symlink
1005 but (unlike "L") deletes a pre-existing file first, should
1006 it already exist and not already be the correct
1007 symlink. Similar, "b+", "c+" and "p+" directives have been
1008 added as well, which create block and character devices, as
1009 well as fifos in the filesystem, possibly removing any
1010 pre-existing files of different types.
1012 * For tmpfiles' "L", "L+", "C" and "C+" directives the final
1013 'argument' field (which so far specified the source to
1014 symlink/copy the files from) is now optional. If omitted the
1015 same file os copied from /usr/share/factory/ suffixed by the
1016 full destination path. This is useful for populating /etc
1017 with essential files, by copying them from vendor defaults
1018 shipped in /usr/share/factory/etc.
1020 * A new command "systemctl preset-all" has been added that
1021 applies the service preset settings to all installed unit
1022 files. A new switch --preset-mode= has been added that
1023 controls whether only enable or only disable operations
1026 * A new command "systemctl is-system-running" has been added
1027 that allows checking the overall state of the system, for
1028 example whether it is fully up and running.
1030 * When the system boots up with an empty /etc, the equivalent
1031 to "systemctl preset-all" is executed during early boot, to
1032 make sure all default services are enabled after a factory
1035 * systemd now contains a minimal preset file that enables the
1036 most basic services systemd ships by default.
1038 * Unit files' [Install] section gained a new DefaultInstance=
1039 field for defining the default instance to create if a
1040 template unit is enabled with no instance specified.
1042 * A new passive target cryptsetup-pre.target has been added
1043 that may be used by services that need to make they run and
1044 finish before the first LUKS cryptographic device is set up.
1046 * The /dev/loop-control and /dev/btrfs-control device nodes
1047 are now owned by the "disk" group by default, opening up
1048 access to this group.
1050 * systemd-coredump will now automatically generate a
1051 stack trace of all core dumps taking place on the system,
1052 based on elfutils' libdw library. This stack trace is logged
1055 * systemd-coredump may now optionally store coredumps directly
1056 on disk (in /var/lib/systemd/coredump, possibly compressed),
1057 instead of storing them unconditionally in the journal. This
1058 mode is the new default. A new configuration file
1059 /etc/systemd/coredump.conf has been added to configure this
1060 and other parameters of systemd-coredump.
1062 * coredumpctl gained a new "info" verb to show details about a
1063 specific coredump. A new switch "-1" has also been added
1064 that makes sure to only show information about the most
1065 recent entry instead of all entries. Also, as the tool is
1066 generally useful now the "systemd-" prefix of the binary
1067 name has been removed. Distributions that want to maintain
1068 compatibility with the old name should add a symlink from
1069 the old name to the new name.
1071 * journald's SplitMode= now defaults to "uid". This makes sure
1072 that unprivileged users can access their own coredumps with
1073 coredumpctl without restrictions.
1075 * New kernel command line options "systemd.wants=" (for
1076 pulling an additional unit during boot), "systemd.mask="
1077 (for masking a specific unit for the boot), and
1078 "systemd.debug-shell" (for enabling the debug shell on tty9)
1079 have been added. This is implemented in the new generator
1080 "systemd-debug-generator".
1082 * systemd-nspawn will now by default filter a couple of
1083 syscalls for containers, among them those required for
1084 kernel module loading, direct x86 IO port access, swap
1085 management, and kexec. Most importantly though
1086 open_by_handle_at() is now prohibited for containers,
1087 closing a hole similar to a recently discussed vulnerability
1088 in docker regarding access to files on file hierarchies the
1089 container should normally not have access to. Note that for
1090 nspawn we generally make no security claims anyway (and
1091 this is explicitly documented in the man page), so this is
1092 just a fix for one of the most obvious problems.
1094 * A new man page file-hierarchy(7) has been added that
1095 contains a minimized, modernized version of the file system
1096 layout systemd expects, similar in style to the FHS
1097 specification or hier(5). A new tool systemd-path(1) has
1098 been added to query many of these paths for the local
1101 * Automatic time-based clean-up of $XDG_RUNTIME_DIR is no
1102 longer done. Since the directory now has a per-user size
1103 limit, and is cleaned on logout this appears unnecessary,
1104 in particular since this now brings the lifecycle of this
1105 directory closer in line with how IPC objects are handled.
1107 * systemd.pc now exports a number of additional directories,
1108 including $libdir (which is useful to identify the library
1109 path for the primary architecture of the system), and a
1110 couple of drop-in directories.
1112 * udev's predictable network interface names now use the dev_port
1113 sysfs attribute, introduced in linux 3.15 instead of dev_id to
1114 distinguish between ports of the same PCI function. dev_id should
1115 only be used for ports using the same HW address, hence the need
1118 * machined has been updated to export the OS version of a
1119 container (read from /etc/os-release and
1120 /usr/lib/os-release) on the bus. This is now shown in
1121 "machinectl status" for a machine.
1123 * A new service setting RestartForceExitStatus= has been
1124 added. If configured to a set of exit signals or process
1125 return values, the service will be restarted when the main
1126 daemon process exits with any of them, regardless of the
1129 * systemctl's -H switch for connecting to remote systemd
1130 machines has been extended so that it may be used to
1131 directly connect to a specific container on the
1132 host. "systemctl -H root@foobar:waldi" will now connect as
1133 user "root" to host "foobar", and then proceed directly to
1134 the container named "waldi". Note that currently you have to
1135 authenticate as user "root" for this to work, as entering
1136 containers is a privileged operation.
1138 Contributions from: Andreas Henriksson, Benjamin Steinwender,
1139 Carl Schaefer, Christian Hesse, Colin Ian King, Cristian
1140 Rodríguez, Daniel Mack, Dave Reisner, David Herrmann, Eugene
1141 Yakubovich, Filipe Brandenburger, Frederic Crozat, Hristo
1142 Venev, Jan Engelhardt, Jonathan Boulle, Kay Sievers, Lennart
1143 Poettering, Luke Shumaker, Mantas Mikulėnas, Marc-Antoine
1144 Perennou, Marcel Holtmann, Michael Marineau, Michael Olbrich,
1145 Michał Bartoszkiewicz, Michal Sekletar, Patrik Flykt, Ronan Le
1146 Martret, Ronny Chevalier, Ruediger Oertel, Steven Noonan,
1147 Susant Sahani, Thadeu Lima de Souza Cascardo, Thomas Hindoe
1148 Paaboel Andersen, Tom Gundersen, Tom Hirst, Umut Tezduyar
1149 Lindskog, Uoti Urpala, Zbigniew Jędrzejewski-Szmek
1151 -- Berlin, 2014-07-03
1155 * As an experimental feature, udev now tries to lock the
1156 disk device node (flock(LOCK_SH|LOCK_NB)) while it
1157 executes events for the disk or any of its partitions.
1158 Applications like partitioning programs can lock the
1159 disk device node (flock(LOCK_EX)) and claim temporary
1160 device ownership that way; udev will entirely skip all event
1161 handling for this disk and its partitions. If the disk
1162 was opened for writing, the close will trigger a partition
1163 table rescan in udev's "watch" facility, and if needed
1164 synthesize "change" events for the disk and all its partitions.
1165 This is now unconditionally enabled, and if it turns out to
1166 cause major problems, we might turn it on only for specific
1167 devices, or might need to disable it entirely. Device Mapper
1168 devices are excluded from this logic.
1170 * We temporarily dropped the "-l" switch for fsck invocations,
1171 since they collide with the flock() logic above. util-linux
1172 upstream has been changed already to avoid this conflict,
1173 and we will readd "-l" as soon as util-linux with this
1174 change has been released.
1176 * The dependency on libattr has been removed. Since a long
1177 time, the extended attribute calls have moved to glibc, and
1178 libattr is thus unnecessary.
1180 * Virtualization detection works without priviliges now. This
1181 means the systemd-detect-virt binary no longer requires
1182 CAP_SYS_PTRACE file capabilities, and our daemons can run
1183 with fewer privileges.
1185 * systemd-networkd now runs under its own "systemd-network"
1186 user. It retains the CAP_NET_ADMIN, CAP_NET_BIND_SERVICE,
1187 CAP_NET_BROADCAST, CAP_NET_RAW capabilities though, but
1188 loses the ability to write to files owned by root this way.
1190 * Similar, systemd-resolved now runs under its own
1191 "systemd-resolve" user with no capabilities remaining.
1193 * Similar, systemd-bus-proxyd now runs under its own
1194 "systemd-bus-proxy" user with only CAP_IPC_OWNER remaining.
1196 * systemd-networkd gained support for setting up "veth"
1197 virtual ethernet devices for container connectivity, as well
1198 as GRE and VTI tunnels.
1200 * systemd-networkd will no longer automatically attempt to
1201 manually load kernel modules necessary for certain tunnel
1202 transports. Instead, it is assumed the kernel loads them
1203 automatically when required. This only works correctly on
1204 very new kernels. On older kernels, please consider adding
1205 the kernel modules to /etc/modules-load.d/ as a work-around.
1207 * The resolv.conf file systemd-resolved generates has been
1208 moved to /run/systemd/resolve/. If you have a symlink from
1209 /etc/resolv.conf, it might be necessary to correct it.
1211 * Two new service settings, ProtectHome= and ProtectSystem=,
1212 have been added. When enabled, they will make the user data
1213 (such as /home) inaccessible or read-only and the system
1214 (such as /usr) read-only, for specific services. This allows
1215 very light-weight per-service sandboxing to avoid
1216 modifications of user data or system files from
1217 services. These two new switches have been enabled for all
1218 of systemd's long-running services, where appropriate.
1220 * Socket units gained new SocketUser= and SocketGroup=
1221 settings to set the owner user and group of AF_UNIX sockets
1222 and FIFOs in the file system.
1224 * Socket units gained a new RemoveOnStop= setting. If enabled,
1225 all FIFOS and sockets in the file system will be removed
1226 when the specific socket unit is stopped.
1228 * Socket units gained a new Symlinks= setting. It takes a list
1229 of symlinks to create to file system sockets or FIFOs
1230 created by the specific Unix sockets. This is useful to
1231 manage symlinks to socket nodes with the same life-cycle as
1234 * The /dev/log socket and /dev/initctl FIFO have been moved to
1235 /run, and have been replaced by symlinks. This allows
1236 connecting to these facilities even if PrivateDevices=yes is
1237 used for a service (which makes /dev/log itself unavailable,
1238 but /run is left). This also has the benefit of ensuring
1239 that /dev only contains device nodes, directories and
1240 symlinks, and nothing else.
1242 * sd-daemon gained two new calls sd_pid_notify() and
1243 sd_pid_notifyf(). They are similar to sd_notify() and
1244 sd_notifyf(), but allow overriding of the source PID of
1245 notification messages if permissions permit this. This is
1246 useful to send notify messages on behalf of a different
1247 process (for example, the parent process). The
1248 systemd-notify tool has been updated to make use of this
1249 when sending messages (so that notification messages now
1250 originate from the shell script invoking systemd-notify and
1251 not the systemd-notify process itself. This should minimize
1252 a race where systemd fails to associate notification
1253 messages to services when the originating process already
1256 * A new "on-abnormal" setting for Restart= has been added. If
1257 set, it will result in automatic restarts on all "abnormal"
1258 reasons for a process to exit, which includes unclean
1259 signals, core dumps, timeouts and watchdog timeouts, but
1260 does not include clean and unclean exit codes or clean
1261 signals. Restart=on-abnormal is an alternative for
1262 Restart=on-failure for services that shall be able to
1263 terminate and avoid restarts on certain errors, by
1264 indicating so with an unclean exit code. Restart=on-failure
1265 or Restart=on-abnormal is now the recommended setting for
1266 all long-running services.
1268 * If the InaccessibleDirectories= service setting points to a
1269 mount point (or if there are any submounts contained within
1270 it), it is now attempted to completely unmount it, to make
1271 the file systems truly unavailable for the respective
1274 * The ReadOnlyDirectories= service setting and
1275 systemd-nspawn's --read-only parameter are now recursively
1276 applied to all submounts, too.
1278 * Mount units may now be created transiently via the bus APIs.
1280 * The support for SysV and LSB init scripts has been removed
1281 from the systemd daemon itself. Instead, it is now
1282 implemented as a generator that creates native systemd units
1283 from these scripts when needed. This enables us to remove a
1284 substantial amount of legacy code from PID 1, following the
1285 fact that many distributions only ship a very small number
1286 of LSB/SysV init scripts nowadays.
1288 * Privileged Xen (dom0) domains are not considered
1289 virtualization anymore by the virtualization detection
1290 logic. After all, they generally have unrestricted access to
1291 the hardware and usually are used to manage the unprivileged
1294 * systemd-tmpfiles gained a new "C" line type, for copying
1295 files or entire directories.
1297 * systemd-tmpfiles "m" lines are now fully equivalent to "z"
1298 lines. So far, they have been non-globbing versions of the
1299 latter, and have thus been redundant. In future, it is
1300 recommended to only use "z". "m" has hence been removed
1301 from the documentation, even though it stays supported.
1303 * A tmpfiles snippet to recreate the most basic structure in
1304 /var has been added. This is enough to create the /var/run →
1305 /run symlink and create a couple of structural
1306 directories. This allows systems to boot up with an empty or
1307 volatile /var. Of course, while with this change, the core OS
1308 now is capable with dealing with a volatile /var, not all
1309 user services are ready for it. However, we hope that sooner
1310 or later, many service daemons will be changed upstream so
1311 that they are able to automatically create their necessary
1312 directories in /var at boot, should they be missing. This is
1313 the first step to allow state-less systems that only require
1314 the vendor image for /usr to boot.
1316 * systemd-nspawn has gained a new --tmpfs= switch to mount an
1317 empty tmpfs instance to a specific directory. This is
1318 particularly useful for making use of the automatic
1319 reconstruction of /var (see above), by passing --tmpfs=/var.
1321 * Access modes specified in tmpfiles snippets may now be
1322 prefixed with "~", which indicates that they shall be masked
1323 by whether the existing file or directory is currently
1324 writable, readable or executable at all. Also, if specified,
1325 the sgid/suid/sticky bits will be masked for all
1328 * A new passive target unit "network-pre.target" has been
1329 added which is useful for services that shall run before any
1330 network is configured, for example firewall scripts.
1332 * The "floppy" group that previously owned the /dev/fd*
1333 devices is no longer used. The "disk" group is now used
1334 instead. Distributions should probably deprecate usage of
1337 Contributions from: Camilo Aguilar, Christian Hesse, Colin Ian
1338 King, Cristian Rodríguez, Daniel Buch, Dave Reisner, David
1339 Strauss, Denis Tikhomirov, John, Jonathan Liu, Kay Sievers,
1340 Lennart Poettering, Mantas Mikulėnas, Mark Eichin, Ronny
1341 Chevalier, Susant Sahani, Thomas Blume, Thomas Hindoe Paaboel
1342 Andersen, Tom Gundersen, Umut Tezduyar Lindskog, Zbigniew
1345 -- Berlin, 2014-06-11
1349 * A new "systemd-timesyncd" daemon has been added for
1350 synchronizing the system clock across the network. It
1351 implements an SNTP client. In contrast to NTP
1352 implementations such as chrony or the NTP reference server,
1353 this only implements a client side, and does not bother with
1354 the full NTP complexity, focusing only on querying time from
1355 one remote server and synchronizing the local clock to
1356 it. Unless you intend to serve NTP to networked clients or
1357 want to connect to local hardware clocks, this simple NTP
1358 client should be more than appropriate for most
1359 installations. The daemon runs with minimal privileges, and
1360 has been hooked up with networkd to only operate when
1361 network connectivity is available. The daemon saves the
1362 current clock to disk every time a new NTP sync has been
1363 acquired, and uses this to possibly correct the system clock
1364 early at bootup, in order to accommodate for systems that
1365 lack an RTC such as the Raspberry Pi and embedded devices,
1366 and to make sure that time monotonically progresses on these
1367 systems, even if it is not always correct. To make use of
1368 this daemon, a new system user and group "systemd-timesync"
1369 needs to be created on installation of systemd.
1371 * The queue "seqnum" interface of libudev has been disabled, as
1372 it was generally incompatible with device namespacing as
1373 sequence numbers of devices go "missing" if the devices are
1374 part of a different namespace.
1376 * "systemctl list-timers" and "systemctl list-sockets" gained
1377 a --recursive switch for showing units of these types also
1378 for all local containers, similar in style to the already
1379 supported --recursive switch for "systemctl list-units".
1381 * A new RebootArgument= setting has been added for service
1382 units, which may be used to specify a kernel reboot argument
1383 to use when triggering reboots with StartLimitAction=.
1385 * A new FailureAction= setting has been added for service
1386 units which may be used to specify an operation to trigger
1387 when a service fails. This works similarly to
1388 StartLimitAction=, but unlike it, controls what is done
1389 immediately rather than only after several attempts to
1390 restart the service in question.
1392 * hostnamed got updated to also expose the kernel name,
1393 release, and version on the bus. This is useful for
1394 executing commands like hostnamectl with the -H switch.
1395 systemd-analyze makes use of this to properly display
1396 details when running non-locally.
1398 * The bootchart tool can now show cgroup information in the
1399 graphs it generates.
1401 * The CFS CPU quota cgroup attribute is now exposed for
1402 services. The new CPUQuota= switch has been added for this
1403 which takes a percentage value. Setting this will have the
1404 result that a service may never get more CPU time than the
1405 specified percentage, even if the machine is otherwise idle.
1407 * systemd-networkd learned IPIP and SIT tunnel support.
1409 * LSB init scripts exposing a dependency on $network will now
1410 get a dependency on network-online.target rather than simply
1411 network.target. This should bring LSB handling closer to
1412 what it was on SysV systems.
1414 * A new fsck.repair= kernel option has been added to control
1415 how fsck shall deal with unclean file systems at boot.
1417 * The (.ini) configuration file parser will now silently
1418 ignore sections whose name begins with "X-". This may be
1419 used to maintain application-specific extension sections in unit
1422 * machined gained a new API to query the IP addresses of
1423 registered containers. "machinectl status" has been updated
1424 to show these addresses in its output.
1426 * A new call sd_uid_get_display() has been added to the
1427 sd-login APIs for querying the "primary" session of a
1428 user. The "primary" session of the user is elected from the
1429 user's sessions and generally a graphical session is
1430 preferred over a text one.
1432 * A minimal systemd-resolved daemon has been added. It
1433 currently simply acts as a companion to systemd-networkd and
1434 manages resolv.conf based on per-interface DNS
1435 configuration, possibly supplied via DHCP. In the long run
1436 we hope to extend this into a local DNSSEC enabled DNS and
1439 * The systemd-networkd-wait-online tool is now enabled by
1440 default. It will delay network-online.target until a network
1441 connection has been configured. The tool primarily integrates
1442 with networkd, but will also make a best effort to make sense
1443 of network configuration performed in some other way.
1445 * Two new service options StartupCPUShares= and
1446 StartupBlockIOWeight= have been added that work similarly to
1447 CPUShares= and BlockIOWeight= however only apply during
1448 system startup. This is useful to prioritize certain services
1449 differently during bootup than during normal runtime.
1451 * hostnamed has been changed to prefer the statically
1452 configured hostname in /etc/hostname (unless set to
1453 'localhost' or empty) over any dynamic one supplied by
1454 dhcp. With this change, the rules for picking the hostname
1455 match more closely the rules of other configuration settings
1456 where the local administrator's configuration in /etc always
1457 overrides any other settings.
1459 Contributions fron: Ali H. Caliskan, Alison Chaiken, Bas van
1460 den Berg, Brandon Philips, Cristian Rodríguez, Daniel Buch,
1461 Dan Kilman, Dave Reisner, David Härdeman, David Herrmann,
1462 David Strauss, Dimitris Spingos, Djalal Harouni, Eelco
1463 Dolstra, Evan Nemerson, Florian Albrechtskirchinger, Greg
1464 Kroah-Hartman, Harald Hoyer, Holger Hans Peter Freyther, Jan
1465 Engelhardt, Jani Nikula, Jason St. John, Jeffrey Clark,
1466 Jonathan Boulle, Kay Sievers, Lennart Poettering, Lukas
1467 Nykryn, Lukasz Skalski, Łukasz Stelmach, Mantas Mikulėnas,
1468 Marcel Holtmann, Martin Pitt, Matthew Monaco, Michael
1469 Marineau, Michael Olbrich, Michal Sekletar, Mike Gilbert, Nis
1470 Martensen, Patrik Flykt, Philip Lorenz, poma, Ray Strode,
1471 Reyad Attiyat, Robert Milasan, Scott Thrasher, Stef Walter,
1472 Steven Siloti, Susant Sahani, Tanu Kaskinen, Thomas Bächler,
1473 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar
1474 Lindskog, WaLyong Cho, Will Woods, Zbigniew
1477 -- Beijing, 2014-05-28
1481 * When restoring the screen brightness at boot, stay away from
1482 the darkest setting or from the lowest 5% of the available
1483 range, depending on which is the larger value of both. This
1484 should effectively protect the user from rebooting into a
1485 black screen, should the brightness have been set to minimum
1488 * sd-login gained a new sd_machine_get_class() call to
1489 determine the class ("vm" or "container") of a machine
1490 registered with machined.
1492 * sd-login gained new calls
1493 sd_peer_get_{session,owner_uid,unit,user_unit,slice,machine_name}(),
1494 to query the identity of the peer of a local AF_UNIX
1495 connection. They operate similarly to their sd_pid_get_xyz()
1498 * PID 1 will now maintain a system-wide system state engine
1499 with the states "starting", "running", "degraded",
1500 "maintenance", "stopping". These states are bound to system
1501 startup, normal runtime, runtime with at least one failed
1502 service, rescue/emergency mode and system shutdown. This
1503 state is shown in the "systemctl status" output when no unit
1504 name is passed. It is useful to determine system state, in
1505 particularly when doing so for many systems or containers at
1508 * A new command "list-machines" has been added to "systemctl"
1509 that lists all local OS containers and shows their system
1510 state (see above), if systemd runs inside of them.
1512 * systemctl gained a new "-r" switch to recursively enumerate
1513 units on all local containers, when used with the
1514 "list-unit" command (which is the default one that is
1515 executed when no parameters are specified).
1517 * The GPT automatic partition discovery logic will now honour
1518 two GPT partition flags: one may be set on a partition to
1519 cause it to be mounted read-only, and the other may be set
1520 on a partition to ignore it during automatic discovery.
1522 * Two new GPT type UUIDs have been added for automatic root
1523 partition discovery, for 32-bit and 64-bit ARM. This is not
1524 particularly useful for discovering the root directory on
1525 these architectures during bare-metal boots (since UEFI is
1526 not common there), but still very useful to allow booting of
1527 ARM disk images in nspawn with the -i option.
1529 * MAC addresses of interfaces created with nspawn's
1530 --network-interface= switch will now be generated from the
1531 machine name, and thus be stable between multiple invocations
1534 * logind will now automatically remove all IPC objects owned
1535 by a user if she or he fully logs out. This makes sure that
1536 users who are logged out cannot continue to consume IPC
1537 resources. This covers SysV memory, semaphores and message
1538 queues as well as POSIX shared memory and message
1539 queues. Traditionally, SysV and POSIX IPC had no life-cycle
1540 limits. With this functionality, that is corrected. This may
1541 be turned off by using the RemoveIPC= switch of logind.conf.
1543 * The systemd-machine-id-setup and tmpfiles tools gained a
1544 --root= switch to operate on a specific root directory,
1547 * journald can now forward logged messages to the TTYs of all
1548 logged in users ("wall"). This is the default for all
1549 emergency messages now.
1551 * A new tool systemd-journal-remote has been added to stream
1552 journal log messages across the network.
1554 * /sys/fs/cgroup/ is now mounted read-only after all cgroup
1555 controller trees are mounted into it. Note that the
1556 directories mounted beneath it are not read-only. This is a
1557 security measure and is particularly useful because glibc
1558 actually includes a search logic to pick any tmpfs it can
1559 find to implement shm_open() if /dev/shm is not available
1560 (which it might very well be in namespaced setups).
1562 * machinectl gained a new "poweroff" command to cleanly power
1563 down a local OS container.
1565 * The PrivateDevices= unit file setting will now also drop the
1566 CAP_MKNOD capability from the capability bound set, and
1567 imply DevicePolicy=closed.
1569 * PrivateDevices=, PrivateNetwork= and PrivateTmp= is now used
1570 comprehensively on all long-running systemd services where
1571 this is appropriate.
1573 * systemd-udevd will now run in a disassociated mount
1574 namespace. To mount directories from udev rules, make sure to
1575 pull in mount units via SYSTEMD_WANTS properties.
1577 * The kdbus support gained support for uploading policy into
1578 the kernel. sd-bus gained support for creating "monitoring"
1579 connections that can eavesdrop into all bus communication
1580 for debugging purposes.
1582 * Timestamps may now be specified in seconds since the UNIX
1583 epoch Jan 1st, 1970 by specifying "@" followed by the value
1586 * Native tcpwrap support in systemd has been removed. tcpwrap
1587 is old code, not really maintained anymore and has serious
1588 shortcomings, and better options such as firewalls
1589 exist. For setups that require tcpwrap usage, please
1590 consider invoking your socket-activated service via tcpd,
1591 like on traditional inetd.
1593 * A new system.conf configuration option
1594 DefaultTimerAccuracySec= has been added that controls the
1595 default AccuracySec= setting of .timer units.
1597 * Timer units gained a new WakeSystem= switch. If enabled,
1598 timers configured this way will cause the system to resume
1599 from system suspend (if the system supports that, which most
1602 * Timer units gained a new Persistent= switch. If enabled,
1603 timers configured this way will save to disk when they have
1604 been last triggered. This information is then used on next
1605 reboot to possible execute overdue timer events, that
1606 could not take place because the system was powered off.
1607 This enables simple anacron-like behaviour for timer units.
1609 * systemctl's "list-timers" will now also list the time a
1610 timer unit was last triggered in addition to the next time
1611 it will be triggered.
1613 * systemd-networkd will now assign predictable IPv4LL
1614 addresses to its local interfaces.
1616 Contributions from: Brandon Philips, Daniel Buch, Daniel Mack,
1617 Dave Reisner, David Herrmann, Gerd Hoffmann, Greg
1618 Kroah-Hartman, Hendrik Brueckner, Jason St. John, Josh
1619 Triplett, Kay Sievers, Lennart Poettering, Marc-Antoine
1620 Perennou, Michael Marineau, Michael Olbrich, Miklos Vajna,
1621 Patrik Flykt, poma, Sebastian Thorarensen, Thomas Bächler,
1622 Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom Gundersen,
1623 Umut Tezduyar Lindskog, Wieland Hoffmann, Zbigniew
1626 -- Berlin, 2014-03-25
1630 * A new unit file setting RestrictAddressFamilies= has been
1631 added to restrict which socket address families unit
1632 processes gain access to. This takes address family names
1633 like "AF_INET" or "AF_UNIX", and is useful to minimize the
1634 attack surface of services via exotic protocol stacks. This
1635 is built on seccomp system call filters.
1637 * Two new unit file settings RuntimeDirectory= and
1638 RuntimeDirectoryMode= have been added that may be used to
1639 manage a per-daemon runtime directories below /run. This is
1640 an alternative for setting up directory permissions with
1641 tmpfiles snippets, and has the advantage that the runtime
1642 directory's lifetime is bound to the daemon runtime and that
1643 the daemon starts up with an empty directory each time. This
1644 is particularly useful when writing services that drop
1645 privileges using the User= or Group= setting.
1647 * The DeviceAllow= unit setting now supports globbing for
1648 matching against device group names.
1650 * The systemd configuration file system.conf gained new
1651 settings DefaultCPUAccounting=, DefaultBlockIOAccounting=,
1652 DefaultMemoryAccounting= to globally turn on/off accounting
1653 for specific resources (cgroups) for all units. These
1654 settings may still be overridden individually in each unit
1657 * systemd-gpt-auto-generator is now able to discover /srv and
1658 root partitions in addition to /home and swap partitions. It
1659 also supports LUKS-encrypted partitions now. With this in
1660 place, automatic discovery of partitions to mount following
1661 the Discoverable Partitions Specification
1662 (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec)
1663 is now a lot more complete. This allows booting without
1664 /etc/fstab and without root= on the kernel command line on
1665 systems prepared appropriately.
1667 * systemd-nspawn gained a new --image= switch which allows
1668 booting up disk images and Linux installations on any block
1669 device that follow the Discoverable Partitions Specification
1670 (see above). This means that installations made with
1671 appropriately updated installers may now be started and
1672 deployed using container managers, completely
1673 unmodified. (We hope that libvirt-lxc will add support for
1674 this feature soon, too.)
1676 * systemd-nspawn gained a new --network-macvlan= setting to
1677 set up a private macvlan interface for the
1678 container. Similarly, systemd-networkd gained a new
1679 Kind=macvlan setting in .netdev files.
1681 * systemd-networkd now supports configuring local addresses
1684 * A new tool systemd-network-wait-online has been added to
1685 synchronously wait for network connectivity using
1688 * The sd-bus.h bus API gained a new sd_bus_track object for
1689 tracking the life-cycle of bus peers. Note that sd-bus.h is
1690 still not a public API though (unless you specify
1691 --enable-kdbus on the configure command line, which however
1692 voids your warranty and you get no API stability guarantee).
1694 * The $XDG_RUNTIME_DIR runtime directories for each user are
1695 now individual tmpfs instances, which has the benefit of
1696 introducing separate pools for each user, with individual
1697 size limits, and thus making sure that unprivileged clients
1698 can no longer negatively impact the system or other users by
1699 filling up their $XDG_RUNTIME_DIR. A new logind.conf setting
1700 RuntimeDirectorySize= has been introduced that allows
1701 controlling the default size limit for all users. It
1702 defaults to 10% of the available physical memory. This is no
1703 replacement for quotas on tmpfs though (which the kernel
1704 still does not support), as /dev/shm and /tmp are still
1705 shared resources used by both the system and unprivileged
1708 * logind will now automatically turn off automatic suspending
1709 on laptop lid close when more than one display is
1710 connected. This was previously expected to be implemented
1711 individually in desktop environments (such as GNOME),
1712 however has been added to logind now, in order to fix a
1713 boot-time race where a desktop environment might not have
1714 been started yet and thus not been able to take an inhibitor
1715 lock at the time where logind already suspends the system
1716 due to a closed lid.
1718 * logind will now wait at least 30s after each system
1719 suspend/resume cycle, and 3min after system boot before
1720 suspending the system due to a closed laptop lid. This
1721 should give USB docking stations and similar enough time to
1722 be probed and configured after system resume and boot in
1723 order to then act as suspend blocker.
1725 * systemd-run gained a new --property= setting which allows
1726 initialization of resource control properties (and others)
1727 for the created scope or service unit. Example: "systemd-run
1728 --property=BlockIOWeight=10 updatedb" may be used to run
1729 updatedb at a low block IO scheduling weight.
1731 * systemd-run's --uid=, --gid=, --setenv=, --setenv= switches
1732 now also work in --scope mode.
1734 * When systemd is compiled with kdbus support, basic support
1735 for enforced policies is now in place. (Note that enabling
1736 kdbus still voids your warranty and no API compatibility
1739 Contributions from: Andrey Borzenkov, Ansgar Burchardt, Armin
1740 K., Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
1741 Harald Hoyer, Henrik Grindal Bakken, Jasper St. Pierre, Kay
1742 Sievers, Kieran Clancy, Lennart Poettering, Lukas Nykryn,
1743 Mantas Mikulėnas, Marcel Holtmann, Mark Oteiza, Martin Pitt,
1744 Mike Gilbert, Peter Rajnoha, poma, Samuli Suominen, Stef
1745 Walter, Susant Sahani, Tero Roponen, Thomas Andersen, Thomas
1746 Bächler, Thomas Hindoe Paaboel Andersen, Tomasz Torcz, Tom
1747 Gundersen, Umut Tezduyar Lindskog, Uoti Urpala, Zachary Cook,
1748 Zbigniew Jędrzejewski-Szmek
1750 -- Berlin, 2014-03-12
1754 * systemd will now relabel /dev after loading the SMACK policy
1755 according to SMACK rules.
1757 * A new unit file option AppArmorProfile= has been added to
1758 set the AppArmor profile for the processes of a unit.
1760 * A new condition check ConditionArchitecture= has been added
1761 to conditionalize units based on the system architecture, as
1762 reported by uname()'s "machine" field.
1764 * systemd-networkd now supports matching on the system
1765 virtualization, architecture, kernel command line, host name
1768 * logind is now a lot more aggressive when suspending the
1769 machine due to a closed laptop lid. Instead of acting only
1770 on the lid close action, it will continuously watch the lid
1771 status and act on it. This is useful for laptops where the
1772 power button is on the outside of the chassis so that it can
1773 be reached without opening the lid (such as the Lenovo
1774 Yoga). On those machines, logind will now immediately
1775 re-suspend the machine if the power button has been
1776 accidentally pressed while the laptop was suspended and in a
1777 backpack or similar.
1779 * logind will now watch SW_DOCK switches and inhibit reaction
1780 to the lid switch if it is pressed. This means that logind
1781 will not suspend the machine anymore if the lid is closed
1782 and the system is docked, if the laptop supports SW_DOCK
1783 notifications via the input layer. Note that ACPI docking
1784 stations do not generate this currently. Also note that this
1785 logic is usually not fully sufficient and Desktop
1786 Environments should take a lid switch inhibitor lock when an
1787 external display is connected, as systemd will not watch
1790 * nspawn will now make use of the devices cgroup controller by
1791 default, and only permit creation of and access to the usual
1792 API device nodes like /dev/null or /dev/random, as well as
1793 access to (but not creation of) the pty devices.
1795 * We will now ship a default .network file for
1796 systemd-networkd that automatically configures DHCP for
1797 network interfaces created by nspawn's --network-veth or
1798 --network-bridge= switches.
1800 * systemd will now understand the usual M, K, G, T suffixes
1801 according to SI conventions (i.e. to the base 1000) when
1802 referring to throughput and hardware metrics. It will stay
1803 with IEC conventions (i.e. to the base 1024) for software
1804 metrics, according to what is customary according to
1805 Wikipedia. We explicitly document which base applies for
1806 each configuration option.
1808 * The DeviceAllow= setting in unit files now supports a syntax
1809 to whitelist an entire group of devices node majors at once,
1810 based on the /proc/devices listing. For example, with the
1811 string "char-pts", it is now possible to whitelist all
1812 current and future pseudo-TTYs at once.
1814 * sd-event learned a new "post" event source. Event sources of
1815 this type are triggered by the dispatching of any event
1816 source of a type that is not "post". This is useful for
1817 implementing clean-up and check event sources that are
1818 triggered by other work being done in the program.
1820 * systemd-networkd is no longer statically enabled, but uses
1821 the usual [Install] sections so that it can be
1822 enabled/disabled using systemctl. It still is enabled by
1825 * When creating a veth interface pair with systemd-nspawn, the
1826 host side will now be prefixed with "vb-" if
1827 --network-bridge= is used, and with "ve-" if --network-veth
1828 is used. This way, it is easy to distinguish these cases on
1829 the host, for example to apply different configuration to
1830 them with systemd-networkd.
1832 * The compatibility libraries for libsystemd-journal.so,
1833 libsystem-id128.so, libsystemd-login.so and
1834 libsystemd-daemon.so do not make use of IFUNC
1835 anymore. Instead, we now build libsystemd.so multiple times
1836 under these alternative names. This means that the footprint
1837 is drastically increased, but given that these are
1838 transitional compatibility libraries, this should not matter
1839 much. This change has been made necessary to support the ARM
1840 platform for these compatibility libraries, as the ARM
1841 toolchain is not really at the same level as the toolchain
1842 for other architectures like x86 and does not support
1843 IFUNC. Please make sure to use --enable-compat-libs only
1844 during a transitional period!
1846 Contributions from: Andreas Fuchs, Armin K., Colin Walters,
1847 Daniel Mack, Dave Reisner, David Herrmann, Djalal Harouni,
1848 Holger Schurig, Jason A. Donenfeld, Jason St. John, Jasper
1849 St. Pierre, Kay Sievers, Lennart Poettering, Łukasz Stelmach,
1850 Marcel Holtmann, Michael Scherer, Michal Sekletar, Mike
1851 Gilbert, Samuli Suominen, Thomas Bächler, Thomas Hindoe
1852 Paaboel Andersen, Tom Gundersen, Umut Tezduyar Lindskog,
1853 Zbigniew Jędrzejewski-Szmek
1855 -- Berlin, 2014-02-24
1859 * A new component "systemd-networkd" has been added that can
1860 be used to configure local network interfaces statically or
1861 via DHCP. It is capable of bringing up bridges, VLANs, and
1862 bonding. Currently, no hook-ups for interactive network
1863 configuration are provided. Use this for your initrd,
1864 container, embedded, or server setup if you need a simple,
1865 yet powerful, network configuration solution. This
1866 configuration subsystem is quite nifty, as it allows wildcard
1867 hotplug matching in interfaces. For example, with a single
1868 configuration snippet, you can configure that all Ethernet
1869 interfaces showing up are automatically added to a bridge,
1870 or similar. It supports link-sensing and more.
1872 * A new tool "systemd-socket-proxyd" has been added which can
1873 act as a bidirectional proxy for TCP sockets. This is
1874 useful for adding socket activation support to services that
1875 do not actually support socket activation, including virtual
1876 machines and the like.
1878 * Add a new tool to save/restore rfkill state on
1881 * Save/restore state of keyboard backlights in addition to
1882 display backlights on shutdown/boot.
1884 * udev learned a new SECLABEL{} construct to label device
1885 nodes with a specific security label when they appear. For
1886 now, only SECLABEL{selinux} is supported, but the syntax is
1887 prepared for additional security frameworks.
1889 * udev gained a new scheme to configure link-level attributes
1890 from files in /etc/systemd/network/*.link. These files can
1891 match against MAC address, device path, driver name and type,
1892 and will apply attributes like the naming policy, link speed,
1893 MTU, duplex settings, Wake-on-LAN settings, MAC address, MAC
1894 address assignment policy (randomized, ...).
1896 * The configuration of network interface naming rules for
1897 "permanent interface names" has changed: a new NamePolicy=
1898 setting in the [Link] section of .link files determines the
1899 priority of possible naming schemes (onboard, slot, mac,
1900 path). The default value of this setting is determined by
1901 /usr/lib/net/links/99-default.link. Old
1902 80-net-name-slot.rules udev configuration file has been
1903 removed, so local configuration overriding this file should
1904 be adapated to override 99-default.link instead.
1906 * When the User= switch is used in a unit file, also
1907 initialize $SHELL= based on the user database entry.
1909 * systemd no longer depends on libdbus. All communication is
1910 now done with sd-bus, systemd's low-level bus library
1913 * kdbus support has been added to PID 1 itself. When kdbus is
1914 enabled, this causes PID 1 to set up the system bus and
1915 enable support for a new ".busname" unit type that
1916 encapsulates bus name activation on kdbus. It works a little
1917 bit like ".socket" units, except for bus names. A new
1918 generator has been added that converts classic dbus1 service
1919 activation files automatically into native systemd .busname
1922 * sd-bus: add a light-weight vtable implementation that allows
1923 defining objects on the bus with a simple static const
1924 vtable array of its methods, signals and properties.
1926 * systemd will not generate or install static dbus
1927 introspection data anymore to /usr/share/dbus-1/interfaces,
1928 as the precise format of these files is unclear, and
1929 nothing makes use of it.
1931 * A proxy daemon is now provided to proxy clients connecting
1932 via classic D-Bus AF_UNIX sockets to kdbus, to provide full
1933 compatibility with classic D-Bus.
1935 * A bus driver implementation has been added that supports the
1936 classic D-Bus bus driver calls on kdbus, also for
1937 compatibility purposes.
1939 * A new API "sd-event.h" has been added that implements a
1940 minimal event loop API built around epoll. It provides a
1941 couple of features that direct epoll usage is lacking:
1942 prioritization of events, scales to large numbers of timer
1943 events, per-event timer slack (accuracy), system-wide
1944 coalescing of timer events, exit handlers, watchdog
1945 supervision support using systemd's sd_notify() API, child
1948 * A new API "sd-rntl.h" has been added that provides an API
1949 around the route netlink interface of the kernel, similar in
1950 style to "sd-bus.h".
1952 * A new API "sd-dhcp-client.h" has been added that provides a
1953 small DHCPv4 client-side implementation. This is used by
1956 * There is a new kernel command line option
1957 "systemd.restore_state=0|1". When set to "0", none of the
1958 systemd tools will restore saved runtime state to hardware
1959 devices. More specifically, the rfkill and backlight states
1962 * The FsckPassNo= compatibility option in mount/service units
1963 has been removed. The fstab generator will now add the
1964 necessary dependencies automatically, and does not require
1965 PID1's support for that anymore.
1967 * journalctl gained a new switch, --list-boots, that lists
1968 recent boots with their times and boot IDs.
1970 * The various tools like systemctl, loginctl, timedatectl,
1971 busctl, systemd-run, ... have gained a new switch "-M" to
1972 connect to a specific, local OS container (as direct
1973 connection, without requiring SSH). This works on any
1974 container that is registered with machined, such as those
1975 created by libvirt-lxc or nspawn.
1977 * systemd-run and systemd-analyze also gained support for "-H"
1978 to connect to remote hosts via SSH. This is particularly
1979 useful for systemd-run because it enables queuing of jobs
1980 onto remote systems.
1982 * machinectl gained a new command "login" to open a getty
1983 login in any local container. This works with any container
1984 that is registered with machined (such as those created by
1985 libvirt-lxc or nspawn), and which runs systemd inside.
1987 * machinectl gained a new "reboot" command that may be used to
1988 trigger a reboot on a specific container that is registered
1989 with machined. This works on any container that runs an init
1990 system of some kind.
1992 * systemctl gained a new "list-timers" command to print a nice
1993 listing of installed timer units with the times they elapse
1996 * Alternative reboot() parameters may now be specified on the
1997 "systemctl reboot" command line and are passed to the
1998 reboot() system call.
2000 * systemctl gained a new --job-mode= switch to configure the
2001 mode to queue a job with. This is a more generic version of
2002 --fail, --irreversible, and --ignore-dependencies, which are
2003 still available but not advertised anymore.
2005 * /etc/systemd/system.conf gained new settings to configure
2006 various default timeouts of units, as well as the default
2007 start limit interval and burst. These may still be overridden
2010 * PID1 will now export on the bus profile data of the security
2011 policy upload process (such as the SELinux policy upload to
2014 * journald: when forwarding logs to the console, include
2015 timestamps (following the setting in
2016 /sys/module/printk/parameters/time).
2018 * OnCalendar= in timer units now understands the special
2019 strings "yearly" and "annually". (Both are equivalent)
2021 * The accuracy of timer units is now configurable with the new
2022 AccuracySec= setting. It defaults to 1min.
2024 * A new dependency type JoinsNamespaceOf= has been added that
2025 allows running two services within the same /tmp and network
2026 namespace, if PrivateNetwork= or PrivateTmp= are used.
2028 * A new command "cat" has been added to systemctl. It outputs
2029 the original unit file of a unit, and concatenates the
2030 contents of additional "drop-in" unit file snippets, so that
2031 the full configuration is shown.
2033 * systemctl now supports globbing on the various "list-xyz"
2034 commands, like "list-units" or "list-sockets", as well as on
2035 those commands which take multiple unit names.
2037 * journalctl's --unit= switch gained support for globbing.
2039 * All systemd daemons now make use of the watchdog logic so
2040 that systemd automatically notices when they hang.
2042 * If the $container_ttys environment variable is set,
2043 getty-generator will automatically spawn a getty for each
2044 listed tty. This is useful for container managers to request
2045 login gettys to be spawned on as many ttys as needed.
2047 * %h, %s, %U specifier support is not available anymore when
2048 used in unit files for PID 1. This is because NSS calls are
2049 not safe from PID 1. They stay available for --user
2050 instances of systemd, and as special case for the root user.
2052 * loginctl gained a new "--no-legend" switch to turn off output
2055 * The "sd-login.h" API gained three new calls:
2056 sd_session_is_remote(), sd_session_get_remote_user(),
2057 sd_session_get_remote_host() to query information about
2060 * The udev hardware database now also carries vendor/product
2061 information of SDIO devices.
2063 * The "sd-daemon.h" API gained a new sd_watchdog_enabled() to
2064 determine whether watchdog notifications are requested by
2067 * Socket-activated per-connection services now include a
2068 short description of the connection parameters in the
2071 * tmpfiles gained a new "--boot" option. When this is not used,
2072 only lines where the command character is not suffixed with
2073 "!" are executed. When this option is specified, those
2074 options are executed too. This partitions tmpfiles
2075 directives into those that can be safely executed at any
2076 time, and those which should be run only at boot (for
2077 example, a line that creates /run/nologin).
2079 * A new API "sd-resolve.h" has been added which provides a simple
2080 asynchronous wrapper around glibc NSS host name resolution
2081 calls, such as getaddrinfo(). In contrast to glibc's
2082 getaddrinfo_a(), it does not use signals. In contrast to most
2083 other asynchronous name resolution libraries, this one does
2084 not reimplement DNS, but reuses NSS, so that alternate
2085 host name resolution systems continue to work, such as mDNS,
2086 LDAP, etc. This API is based on libasyncns, but it has been
2087 cleaned up for inclusion in systemd.
2089 * The APIs "sd-journal.h", "sd-login.h", "sd-id128.h",
2090 "sd-daemon.h" are no longer found in individual libraries
2091 libsystemd-journal.so, libsystemd-login.so,
2092 libsystemd-id128.so, libsystemd-daemon.so. Instead, we have
2093 merged them into a single library, libsystemd.so, which
2094 provides all symbols. The reason for this is cyclic
2095 dependencies, as these libraries tend to use each other's
2096 symbols. So far, we have managed to workaround that by linking
2097 a copy of a good part of our code into each of these
2098 libraries again and again, which, however, makes certain
2099 things hard to do, like sharing static variables. Also, it
2100 substantially increases footprint. With this change, there
2101 is only one library for the basic APIs systemd
2102 provides. Also, "sd-bus.h", "sd-memfd.h", "sd-event.h",
2103 "sd-rtnl.h", "sd-resolve.h", "sd-utf8.h" are found in this
2104 library as well, however are subject to the --enable-kdbus
2105 switch (see below). Note that "sd-dhcp-client.h" is not part
2106 of this library (this is because it only consumes, never
2107 provides, services of/to other APIs). To make the transition
2108 easy from the separate libraries to the unified one, we
2109 provide the --enable-compat-libs compile-time switch which
2110 will generate stub libraries that are compatible with the
2111 old ones but redirect all calls to the new one.
2113 * All of the kdbus logic and the new APIs "sd-bus.h",
2114 "sd-memfd.h", "sd-event.h", "sd-rtnl.h", "sd-resolve.h",
2115 and "sd-utf8.h" are compile-time optional via the
2116 "--enable-kdbus" switch, and they are not compiled in by
2117 default. To make use of kdbus, you have to explicitly enable
2118 the switch. Note however, that neither the kernel nor the
2119 userspace API for all of this is considered stable yet. We
2120 want to maintain the freedom to still change the APIs for
2121 now. By specifying this build-time switch, you acknowledge
2122 that you are aware of the instability of the current
2125 * Also, note that while kdbus is pretty much complete,
2126 it lacks one thing: proper policy support. This means you
2127 can build a fully working system with all features; however,
2128 it will be highly insecure. Policy support will be added in
2129 one of the next releases, at the same time that we will
2130 declare the APIs stable.
2132 * When the kernel command line argument "kdbus" is specified,
2133 systemd will automatically load the kdbus.ko kernel module. At
2134 this stage of development, it is only useful for testing kdbus
2135 and should not be used in production. Note: if "--enable-kdbus"
2136 is specified, and the kdbus.ko kernel module is available, and
2137 "kdbus" is added to the kernel command line, the entire system
2138 runs with kdbus instead of dbus-daemon, with the above mentioned
2139 problem of missing the system policy enforcement. Also a future
2140 version of kdbus.ko or a newer systemd will not be compatible with
2141 each other, and will unlikely be able to boot the machine if only
2142 one of them is updated.
2144 * systemctl gained a new "import-environment" command which
2145 uploads the caller's environment (or parts thereof) into the
2146 service manager so that it is inherited by services started
2147 by the manager. This is useful to upload variables like
2148 $DISPLAY into the user service manager.
2150 * A new PrivateDevices= switch has been added to service units
2151 which allows running a service with a namespaced /dev
2152 directory that does not contain any device nodes for
2153 physical devices. More specifically, it only includes devices
2154 such as /dev/null, /dev/urandom, and /dev/zero which are API
2157 * logind has been extended to support behaviour like VT
2158 switching on seats that do not support a VT. This makes
2159 multi-session available on seats that are not the first seat
2160 (seat0), and on systems where kernel support for VTs has
2161 been disabled at compile-time.
2163 * If a process holds a delay lock for system sleep or shutdown
2164 and fails to release it in time, we will now log its
2165 identity. This makes it easier to identify processes that
2166 cause slow suspends or power-offs.
2168 * When parsing /etc/crypttab, support for a new key-slot=
2169 option as supported by Debian is added. It allows indicating
2170 which LUKS slot to use on disk, speeding up key loading.
2172 * The sd_journald_sendv() API call has been checked and
2173 officially declared to be async-signal-safe so that it may
2174 be invoked from signal handlers for logging purposes.
2176 * Boot-time status output is now enabled automatically after a
2177 short timeout if boot does not progress, in order to give
2178 the user an indication what she or he is waiting for.
2180 * The boot-time output has been improved to show how much time
2181 remains until jobs expire.
2183 * The KillMode= switch in service units gained a new possible
2184 value "mixed". If set, and the unit is shut down, then the
2185 initial SIGTERM signal is sent only to the main daemon
2186 process, while the following SIGKILL signal is sent to
2187 all remaining processes of the service.
2189 * When a scope unit is registered, a new property "Controller"
2190 may be set. If set to a valid bus name, systemd will send a
2191 RequestStop() signal to this name when it would like to shut
2192 down the scope. This may be used to hook manager logic into
2193 the shutdown logic of scope units. Also, scope units may now
2194 be put in a special "abandoned" state, in which case the
2195 manager process which created them takes no further
2196 responsibilities for it.
2198 * When reading unit files, systemd will now verify
2199 the access mode of these files, and warn about certain
2200 suspicious combinations. This has been added to make it
2201 easier to track down packaging bugs where unit files are
2202 marked executable or world-writable.
2204 * systemd-nspawn gained a new "--setenv=" switch to set
2205 container-wide environment variables. The similar option in
2206 systemd-activate was renamed from "--environment=" to
2207 "--setenv=" for consistency.
2209 * systemd-nspawn has been updated to create a new kdbus domain
2210 for each container that is invoked, thus allowing each
2211 container to have its own set of system and user buses,
2212 independent of the host.
2214 * systemd-nspawn gained a new --drop-capability= switch to run
2215 the container with less capabilities than the default. Both
2216 --drop-capability= and --capability= now take the special
2217 string "all" for dropping or keeping all capabilities.
2219 * systemd-nspawn gained new switches for executing containers
2220 with specific SELinux labels set.
2222 * systemd-nspawn gained a new --quiet switch to not generate
2223 any additional output but the container's own console
2226 * systemd-nspawn gained a new --share-system switch to run a
2227 container without PID namespacing enabled.
2229 * systemd-nspawn gained a new --register= switch to control
2230 whether the container is registered with systemd-machined or
2231 not. This is useful for containers that do not run full
2232 OS images, but only specific apps.
2234 * systemd-nspawn gained a new --keep-unit which may be used
2235 when invoked as the only program from a service unit, and
2236 results in registration of the unit service itself in
2237 systemd-machined, instead of a newly opened scope unit.
2239 * systemd-nspawn gained a new --network-interface= switch for
2240 moving arbitrary interfaces to the container. The new
2241 --network-veth switch creates a virtual Ethernet connection
2242 between host and container. The new --network-bridge=
2243 switch then allows assigning the host side of this virtual
2244 Ethernet connection to a bridge device.
2246 * systemd-nspawn gained a new --personality= switch for
2247 setting the kernel personality for the container. This is
2248 useful when running a 32-bit container on a 64-bit host. A
2249 similar option Personality= is now also available for service
2252 * logind will now also track a "Desktop" identifier for each
2253 session which encodes the desktop environment of it. This is
2254 useful for desktop environments that want to identify
2255 multiple running sessions of itself easily.
2257 * A new SELinuxContext= setting for service units has been
2258 added that allows setting a specific SELinux execution
2259 context for a service.
2261 * Most systemd client tools will now honour $SYSTEMD_LESS for
2262 settings of the "less" pager. By default, these tools will
2263 override $LESS to allow certain operations to work, such as
2264 jump-to-the-end. With $SYSTEMD_LESS, it is possible to
2265 influence this logic.
2267 * systemd's "seccomp" hook-up has been changed to make use of
2268 the libseccomp library instead of using its own
2269 implementation. This has benefits for portability among
2272 * For usage together with SystemCallFilter=, a new
2273 SystemCallErrorNumber= setting has been introduced that
2274 allows configuration of a system error number to be returned
2275 on filtered system calls, instead of immediately killing the
2276 process. Also, SystemCallArchitectures= has been added to
2277 limit access to system calls of a particular architecture
2278 (in order to turn off support for unused secondary
2279 architectures). There is also a global
2280 SystemCallArchitectures= setting in system.conf now to turn
2281 off support for non-native system calls system-wide.
2283 * systemd requires a kernel with a working name_to_handle_at(),
2284 please see the kernel config requirements in the README file.
2286 Contributions from: Adam Williamson, Alex Jia, Anatol Pomozov,
2287 Ansgar Burchardt, AppleBloom, Auke Kok, Bastien Nocera,
2288 Chengwei Yang, Christian Seiler, Colin Guthrie, Colin Walters,
2289 Cristian Rodríguez, Daniel Buch, Daniele Medri, Daniel J
2290 Walsh, Daniel Mack, Dan McGee, Dave Reisner, David Coppa,
2291 David Herrmann, David Strauss, Djalal Harouni, Dmitry Pisklov,
2292 Elia Pinto, Florian Weimer, George McCollister, Goffredo
2293 Baroncelli, Greg Kroah-Hartman, Hendrik Brueckner, Igor
2294 Zhbanov, Jan Engelhardt, Jan Janssen, Jason A. Donenfeld,
2295 Jason St. John, Jasper St. Pierre, Jóhann B. Guðmundsson, Jose
2296 Ignacio Naranjo, Karel Zak, Kay Sievers, Kristian Høgsberg,
2297 Lennart Poettering, Lubomir Rintel, Lukas Nykryn, Lukasz
2298 Skalski, Łukasz Stelmach, Luke Shumaker, Mantas Mikulėnas,
2299 Marc-Antoine Perennou, Marcel Holtmann, Marcos Felipe Rasia de
2300 Mello, Marko Myllynen, Martin Pitt, Matthew Monaco, Michael
2301 Marineau, Michael Scherer, Michał Górny, Michal Sekletar,
2302 Michele Curti, Oleksii Shevchuk, Olivier Brunel, Patrik Flykt,
2303 Pavel Holica, Raudi, Richard Marko, Ronny Chevalier, Sébastien
2304 Luttringer, Sergey Ptashnick, Shawn Landden, Simon Peeters,
2305 Stefan Beller, Susant Sahani, Sylvain Plantefeve, Sylvia Else,
2306 Tero Roponen, Thomas Bächler, Thomas Hindoe Paaboel Andersen,
2307 Tom Gundersen, Umut Tezduyar Lindskog, Unai Uribarri, Václav
2308 Pavlín, Vincent Batts, WaLyong Cho, William Giokas, Yang
2309 Zhiyong, Yin Kangkai, Yuxuan Shui, Zbigniew Jędrzejewski-Szmek
2311 -- Berlin, 2014-02-20
2315 * logind has gained support for facilitating privileged input
2316 and drm device access for unprivileged clients. This work is
2317 useful to allow Wayland display servers (and similar
2318 programs, such as kmscon) to run under the user's ID and
2319 access input and drm devices which are normally
2320 protected. When this is used (and the kernel is new enough)
2321 logind will "mute" IO on the file descriptors passed to
2322 Wayland as long as it is in the background and "unmute" it
2323 if it returns into the foreground. This allows secure
2324 session switching without allowing background sessions to
2325 eavesdrop on input and display data. This also introduces
2326 session switching support if VT support is turned off in the
2327 kernel, and on seats that are not seat0.
2329 * A new kernel command line option luks.options= is understood
2330 now which allows specifying LUKS options for usage for LUKS
2331 encrypted partitions specified with luks.uuid=.
2333 * tmpfiles.d(5) snippets may now use specifier expansion in
2334 path names. More specifically %m, %b, %H, %v, are now
2335 replaced by the local machine id, boot id, hostname, and
2336 kernel version number.
2338 * A new tmpfiles.d(5) command "m" has been introduced which
2339 may be used to change the owner/group/access mode of a file
2340 or directory if it exists, but do nothing if it does not.
2342 * This release removes high-level support for the
2343 MemorySoftLimit= cgroup setting. The underlying kernel
2344 cgroup attribute memory.soft_limit= is currently badly
2345 designed and likely to be removed from the kernel API in its
2346 current form, hence we should not expose it for now.
2348 * The memory.use_hierarchy cgroup attribute is now enabled for
2349 all cgroups systemd creates in the memory cgroup
2350 hierarchy. This option is likely to be come the built-in
2351 default in the kernel anyway, and the non-hierarchical mode
2352 never made much sense in the intrinsically hierarchical
2355 * A new field _SYSTEMD_SLICE= is logged along with all journal
2356 messages containing the slice a message was generated
2357 from. This is useful to allow easy per-customer filtering of
2358 logs among other things.
2360 * systemd-journald will no longer adjust the group of journal
2361 files it creates to the "systemd-journal" group. Instead we
2362 rely on the journal directory to be owned by the
2363 "systemd-journal" group, and its setgid bit set, so that the
2364 kernel file system layer will automatically enforce that
2365 journal files inherit this group assignment. The reason for
2366 this change is that we cannot allow NSS look-ups from
2367 journald which would be necessary to resolve
2368 "systemd-journal" to a numeric GID, because this might
2369 create deadlocks if NSS involves synchronous queries to
2370 other daemons (such as nscd, or sssd) which in turn are
2371 logging clients of journald and might block on it, which
2372 would then dead lock. A tmpfiles.d(5) snippet included in
2373 systemd will make sure the setgid bit and group are
2374 properly set on the journal directory if it exists on every
2375 boot. However, we recommend adjusting it manually after
2376 upgrades too (or from RPM scriptlets), so that the change is
2377 not delayed until next reboot.
2379 * Backlight and random seed files in /var/lib/ have moved into
2380 the /var/lib/systemd/ directory, in order to centralize all
2381 systemd generated files in one directory.
2383 * Boot time performance measurements (as displayed by
2384 "systemd-analyze" for example) will now read ACPI 5.0 FPDT
2385 performance information if that's available to determine how
2386 much time BIOS and boot loader initialization required. With
2387 a sufficiently new BIOS you hence no longer need to boot
2388 with Gummiboot to get access to such information.
2390 Contributions from: Andrey Borzenkov, Chen Jie, Colin Walters,
2391 Cristian Rodríguez, Dave Reisner, David Herrmann, David
2392 Mackey, David Strauss, Eelco Dolstra, Evan Callicoat, Gao
2393 feng, Harald Hoyer, Jimmie Tauriainen, Kay Sievers, Lennart
2394 Poettering, Lukas Nykryn, Mantas Mikulėnas, Martin Pitt,
2395 Michael Scherer, Michał Górny, Mike Gilbert, Patrick McCarty,
2396 Sebastian Ott, Tom Gundersen, Zbigniew Jędrzejewski-Szmek
2398 -- Berlin, 2013-10-02
2402 * The Restart= option for services now understands a new
2403 on-watchdog setting, which will restart the service
2404 automatically if the service stops sending out watchdog keep
2405 alive messages (as configured with WatchdogSec=).
2407 * The getty generator (which is responsible for bringing up a
2408 getty on configured serial consoles) will no longer only
2409 start a getty on the primary kernel console but on all
2410 others, too. This makes the order in which console= is
2411 specified on the kernel command line less important.
2413 * libsystemd-logind gained a new sd_session_get_vt() call to
2414 retrieve the VT number of a session.
2416 * If the option "tries=0" is set for an entry of /etc/crypttab
2417 its passphrase is queried indefinitely instead of any
2418 maximum number of tries.
2420 * If a service with a configure PID file terminates its PID
2421 file will now be removed automatically if it still exists
2422 afterwards. This should put an end to stale PID files.
2424 * systemd-run will now also take relative binary path names
2425 for execution and no longer insists on absolute paths.
2427 * InaccessibleDirectories= and ReadOnlyDirectories= now take
2428 paths that are optionally prefixed with "-" to indicate that
2429 it should not be considered a failure if they do not exist.
2431 * journalctl -o (and similar commands) now understands a new
2432 output mode "short-precise", it is similar to "short" but
2433 shows timestamps with usec accuracy.
2435 * The option "discard" (as known from Debian) is now
2436 synonymous to "allow-discards" in /etc/crypttab. In fact,
2437 "discard" is preferred now (since it is easier to remember
2440 * Some licensing clean-ups were made, so that more code is now
2441 LGPL-2.1 licensed than before.
2443 * A minimal tool to save/restore the display backlight
2444 brightness across reboots has been added. It will store the
2445 backlight setting as late as possible at shutdown, and
2446 restore it as early as possible during reboot.
2448 * A logic to automatically discover and enable home and swap
2449 partitions on GPT disks has been added. With this in place
2450 /etc/fstab becomes optional for many setups as systemd can
2451 discover certain partitions located on the root disk
2452 automatically. Home partitions are recognized under their
2453 GPT type ID 933ac7e12eb44f13b8440e14e2aef915. Swap
2454 partitions are recognized under their GPT type ID
2455 0657fd6da4ab43c484e50933c84b4f4f.
2457 * systemd will no longer pass any environment from the kernel
2458 or initrd to system services. If you want to set an
2459 environment for all services, do so via the kernel command
2460 line systemd.setenv= assignment.
2462 * The systemd-sysctl tool no longer natively reads the file
2463 /etc/sysctl.conf. If desired, the file should be symlinked
2464 from /etc/sysctl.d/99-sysctl.conf. Apart from providing
2465 legacy support by a symlink rather than built-in code, it
2466 also makes the otherwise hidden order of application of the
2467 different files visible. (Note that this partly reverts to a
2468 pre-198 application order of sysctl knobs!)
2470 * The "systemctl set-log-level" and "systemctl dump" commands
2471 have been moved to systemd-analyze.
2473 * systemd-run learned the new --remain-after-exit switch,
2474 which causes the scope unit not to be cleaned up
2475 automatically after the process terminated.
2477 * tmpfiles learned a new --exclude-prefix= switch to exclude
2478 certain paths from operation.
2480 * journald will now automatically flush all messages to disk
2481 as soon as a message at the log level CRIT, ALERT or EMERG
2484 Contributions from: Andrew Cook, Brandon Philips, Christian
2485 Hesse, Christoph Junghans, Colin Walters, Daniel Schaal,
2486 Daniel Wallace, Dave Reisner, David Herrmann, Gao feng, George
2487 McCollister, Giovanni Campagna, Hannes Reinecke, Harald Hoyer,
2488 Herczeg Zsolt, Holger Hans Peter Freyther, Jan Engelhardt,
2489 Jesper Larsen, Kay Sievers, Khem Raj, Lennart Poettering,
2490 Lukas Nykryn, Maciej Wereski, Mantas Mikulėnas, Marcel
2491 Holtmann, Martin Pitt, Michael Biebl, Michael Marineau,
2492 Michael Scherer, Michael Stapelberg, Michal Sekletar, Michał
2493 Górny, Olivier Brunel, Ondrej Balaz, Ronny Chevalier, Shawn
2494 Landden, Steven Hiscocks, Thomas Bächler, Thomas Hindoe
2495 Paaboel Andersen, Tom Gundersen, Umut Tezduyar, WANG Chao,
2496 William Giokas, Zbigniew Jędrzejewski-Szmek
2498 -- Berlin, 2013-09-13
2502 * The documentation has been updated to cover the various new
2503 concepts introduced with 205.
2505 * Unit files now understand the new %v specifier which
2506 resolves to the kernel version string as returned by "uname
2509 * systemctl now supports filtering the unit list output by
2510 load state, active state and sub state, using the new
2513 * "systemctl status" will now show the results of the
2514 condition checks (like ConditionPathExists= and similar) of
2515 the last start attempts of the unit. They are also logged to
2518 * "journalctl -b" may now be used to look for boot output of a
2519 specific boot. Try "journalctl -b -1" for the previous boot,
2520 but the syntax is substantially more powerful.
2522 * "journalctl --show-cursor" has been added which prints the
2523 cursor string the last shown log line. This may then be used
2524 with the new "journalctl --after-cursor=" switch to continue
2525 browsing logs from that point on.
2527 * "journalctl --force" may now be used to force regeneration
2530 * Creation of "dead" device nodes has been moved from udev
2531 into kmod and tmpfiles. Previously, udev would read the kmod
2532 databases to pre-generate dead device nodes based on meta
2533 information contained in kernel modules, so that these would
2534 be auto-loaded on access rather then at boot. As this
2535 does not really have much to do with the exposing actual
2536 kernel devices to userspace this has always been slightly
2537 alien in the udev codebase. Following the new scheme kmod
2538 will now generate a runtime snippet for tmpfiles from the
2539 module meta information and it now is tmpfiles' job to the
2540 create the nodes. This also allows overriding access and
2541 other parameters for the nodes using the usual tmpfiles
2542 facilities. As side effect this allows us to remove the
2543 CAP_SYS_MKNOD capability bit from udevd entirely.
2545 * logind's device ACLs may now be applied to these "dead"
2546 devices nodes too, thus finally allowing managed access to
2547 devices such as /dev/snd/sequencer whithout loading the
2548 backing module right-away.
2550 * A new RPM macro has been added that may be used to apply
2551 tmpfiles configuration during package installation.
2553 * systemd-detect-virt and ConditionVirtualization= now can
2554 detect User-Mode-Linux machines (UML).
2556 * journald will now implicitly log the effective capabilities
2557 set of processes in the message metadata.
2559 * systemd-cryptsetup has gained support for TrueCrypt volumes.
2561 * The initrd interface has been simplified (more specifically,
2562 support for passing performance data via environment
2563 variables and fsck results via files in /run has been
2564 removed). These features were non-essential, and are
2565 nowadays available in a much nicer way by having systemd in
2566 the initrd serialize its state and have the hosts systemd
2567 deserialize it again.
2569 * The udev "keymap" data files and tools to apply keyboard
2570 specific mappings of scan to key codes, and force-release
2571 scan code lists have been entirely replaced by a udev
2572 "keyboard" builtin and a hwdb data file.
2574 * systemd will now honour the kernel's "quiet" command line
2575 argument also during late shutdown, resulting in a
2576 completely silent shutdown when used.
2578 * There's now an option to control the SO_REUSEPORT socket
2579 option in .socket units.
2581 * Instance units will now automatically get a per-template
2582 subslice of system.slice unless something else is explicitly
2583 configured. For example, instances of sshd@.service will now
2584 implicitly be placed in system-sshd.slice rather than
2585 system.slice as before.
2587 * Test coverage support may now be enabled at build time.
2589 Contributions from: Dave Reisner, Frederic Crozat, Harald
2590 Hoyer, Holger Hans Peter Freyther, Jan Engelhardt, Jan
2591 Janssen, Jason St. John, Jesper Larsen, Kay Sievers, Lennart
2592 Poettering, Lukas Nykryn, Maciej Wereski, Martin Pitt, Michael
2593 Olbrich, Ramkumar Ramachandra, Ross Lagerwall, Shawn Landden,
2594 Thomas H.P. Andersen, Tom Gundersen, Tomasz Torcz, William
2595 Giokas, Zbigniew Jędrzejewski-Szmek
2597 -- Berlin, 2013-07-23
2601 * Two new unit types have been introduced:
2603 Scope units are very similar to service units, however, are
2604 created out of pre-existing processes -- instead of PID 1
2605 forking off the processes. By using scope units it is
2606 possible for system services and applications to group their
2607 own child processes (worker processes) in a powerful way
2608 which then maybe used to organize them, or kill them
2609 together, or apply resource limits on them.
2611 Slice units may be used to partition system resources in an
2612 hierarchical fashion and then assign other units to them. By
2613 default there are now three slices: system.slice (for all
2614 system services), user.slice (for all user sessions),
2615 machine.slice (for VMs and containers).
2617 Slices and scopes have been introduced primarily in
2618 context of the work to move cgroup handling to a
2619 single-writer scheme, where only PID 1
2620 creates/removes/manages cgroups.
2622 * There's a new concept of "transient" units. In contrast to
2623 normal units these units are created via an API at runtime,
2624 not from configuration from disk. More specifically this
2625 means it is now possible to run arbitrary programs as
2626 independent services, with all execution parameters passed
2627 in via bus APIs rather than read from disk. Transient units
2628 make systemd substantially more dynamic then it ever was,
2629 and useful as a general batch manager.
2631 * logind has been updated to make use of scope and slice units
2632 for managing user sessions. As a user logs in he will get
2633 his own private slice unit, to which all sessions are added
2634 as scope units. We also added support for automatically
2635 adding an instance of user@.service for the user into the
2636 slice. Effectively logind will no longer create cgroup
2637 hierarchies on its own now, it will defer entirely to PID 1
2638 for this by means of scope, service and slice units. Since
2639 user sessions this way become entities managed by PID 1
2640 the output of "systemctl" is now a lot more comprehensive.
2642 * A new mini-daemon "systemd-machined" has been added which
2643 may be used by virtualization managers to register local
2644 VMs/containers. nspawn has been updated accordingly, and
2645 libvirt will be updated shortly. machined will collect a bit
2646 of meta information about the VMs/containers, and assign
2647 them their own scope unit (see above). The collected
2648 meta-data is then made available via the "machinectl" tool,
2649 and exposed in "ps" and similar tools. machined/machinectl
2650 is compile-time optional.
2652 * As discussed earlier, the low-level cgroup configuration
2653 options ControlGroup=, ControlGroupModify=,
2654 ControlGroupPersistent=, ControlGroupAttribute= have been
2655 removed. Please use high-level attribute settings instead as
2656 well as slice units.
2658 * A new bus call SetUnitProperties() has been added to alter
2659 various runtime parameters of a unit. This is primarily
2660 useful to alter cgroup parameters dynamically in a nice way,
2661 but will be extended later on to make more properties
2662 modifiable at runtime. systemctl gained a new set-properties
2663 command that wraps this call.
2665 * A new tool "systemd-run" has been added which can be used to
2666 run arbitrary command lines as transient services or scopes,
2667 while configuring a number of settings via the command
2668 line. This tool is currently very basic, however already
2669 very useful. We plan to extend this tool to even allow
2670 queuing of execution jobs with time triggers from the
2671 command line, similar in fashion to "at".
2673 * nspawn will now inform the user explicitly that kernels with
2674 audit enabled break containers, and suggest the user to turn
2677 * Support for detecting the IMA and AppArmor security
2678 frameworks with ConditionSecurity= has been added.
2680 * journalctl gained a new "-k" switch for showing only kernel
2681 messages, mimicking dmesg output; in addition to "--user"
2682 and "--system" switches for showing only user's own logs
2685 * systemd-delta can now show information about drop-in
2686 snippets extending unit files.
2688 * libsystemd-bus has been substantially updated but is still
2689 not available as public API.
2691 * systemd will now look for the "debug" argument on the kernel
2692 command line and enable debug logging, similar to what
2693 "systemd.log_level=debug" already did before.
2695 * "systemctl set-default", "systemctl get-default" has been
2696 added to configure the default.target symlink, which
2697 controls what to boot into by default.
2699 * "systemctl set-log-level" has been added as a convenient
2700 way to raise and lower systemd logging threshold.
2702 * "systemd-analyze plot" will now show the time the various
2703 generators needed for execution, as well as information
2704 about the unit file loading.
2706 * libsystemd-journal gained a new sd_journal_open_files() call
2707 for opening specific journal files. journactl also gained a
2708 new switch to expose this new functionality. Previously we
2709 only supported opening all files from a directory, or all
2710 files from the system, as opening individual files only is
2711 racy due to journal file rotation.
2713 * systemd gained the new DefaultEnvironment= setting in
2714 /etc/systemd/system.conf to set environment variables for
2717 * If a privileged process logs a journal message with the
2718 OBJECT_PID= field set, then journald will automatically
2719 augment this with additional OBJECT_UID=, OBJECT_GID=,
2720 OBJECT_COMM=, OBJECT_EXE=, ... fields. This is useful if
2721 system services want to log events about specific client
2722 processes. journactl/systemctl has been updated to make use
2723 of this information if all log messages regarding a specific
2726 Contributions from: Auke Kok, Chengwei Yang, Colin Walters,
2727 Cristian Rodríguez, Daniel Albers, Daniel Wallace, Dave
2728 Reisner, David Coppa, David King, David Strauss, Eelco
2729 Dolstra, Gabriel de Perthuis, Harald Hoyer, Jan Alexander
2730 Steffens, Jan Engelhardt, Jan Janssen, Jason St. John, Johan
2731 Heikkilä, Karel Zak, Karol Lewandowski, Kay Sievers, Lennart
2732 Poettering, Lukas Nykryn, Mantas Mikulėnas, Marius Vollmer,
2733 Martin Pitt, Michael Biebl, Michael Olbrich, Michael Tremer,
2734 Michal Schmidt, Michał Bartoszkiewicz, Nirbheek Chauhan,
2735 Pierre Neidhardt, Ross Burton, Ross Lagerwall, Sean McGovern,
2736 Thomas Hindoe Paaboel Andersen, Tom Gundersen, Umut Tezduyar,
2737 Václav Pavlín, Zachary Cook, Zbigniew Jędrzejewski-Szmek,
2738 Łukasz Stelmach, 장동준
2742 * The Python bindings gained some minimal support for the APIs
2743 exposed by libsystemd-logind.
2745 * ConditionSecurity= gained support for detecting SMACK. Since
2746 this condition already supports SELinux and AppArmor we only
2747 miss IMA for this. Patches welcome!
2749 Contributions from: Karol Lewandowski, Lennart Poettering,
2750 Zbigniew Jędrzejewski-Szmek
2754 * systemd-nspawn will now create /etc/resolv.conf if
2755 necessary, before bind-mounting the host's file onto it.
2757 * systemd-nspawn will now store meta information about a
2758 container on the container's cgroup as extended attribute
2759 fields, including the root directory.
2761 * The cgroup hierarchy has been reworked in many ways. All
2762 objects any of the components systemd creates in the cgroup
2763 tree are now suffixed. More specifically, user sessions are
2764 now placed in cgroups suffixed with ".session", users in
2765 cgroups suffixed with ".user", and nspawn containers in
2766 cgroups suffixed with ".nspawn". Furthermore, all cgroup
2767 names are now escaped in a simple scheme to avoid collision
2768 of userspace object names with kernel filenames. This work
2769 is preparation for making these objects relocatable in the
2770 cgroup tree, in order to allow easy resource partitioning of
2771 these objects without causing naming conflicts.
2773 * systemctl list-dependencies gained the new switches
2774 --plain, --reverse, --after and --before.
2776 * systemd-inhibit now shows the process name of processes that
2777 have taken an inhibitor lock.
2779 * nss-myhostname will now also resolve "localhost"
2780 implicitly. This makes /etc/hosts an optional file and
2781 nicely handles that on IPv6 ::1 maps to both "localhost" and
2784 * libsystemd-logind.so gained a new call
2785 sd_get_machine_names() to enumerate running containers and
2786 VMs (currently only supported by very new libvirt and
2787 nspawn). sd_login_monitor can now be used to watch
2788 VMs/containers coming and going.
2790 * .include is not allowed recursively anymore, and only in
2791 unit files. Usually it is better to use drop-in snippets in
2792 .d/*.conf anyway, as introduced with systemd 198.
2794 * systemd-analyze gained a new "critical-chain" command that
2795 determines the slowest chain of units run during system
2796 boot-up. It is very useful for tracking down where
2797 optimizing boot time is the most beneficial.
2799 * systemd will no longer allow manipulating service paths in
2800 the name=systemd:/system cgroup tree using ControlGroup= in
2801 units. (But is still fine with it in all other dirs.)
2803 * There's a new systemd-nspawn@.service service file that may
2804 be used to easily run nspawn containers as system
2805 services. With the container's root directory in
2806 /var/lib/container/foobar it is now sufficient to run
2807 "systemctl start systemd-nspawn@foobar.service" to boot it.
2809 * systemd-cgls gained a new parameter "--machine" to list only
2810 the processes within a certain container.
2812 * ConditionSecurity= now can check for "apparmor". We still
2813 are lacking checks for SMACK and IMA for this condition
2814 check though. Patches welcome!
2816 * A new configuration file /etc/systemd/sleep.conf has been
2817 added that may be used to configure which kernel operation
2818 systemd is supposed to execute when "suspend", "hibernate"
2819 or "hybrid-sleep" is requested. This makes the new kernel
2820 "freeze" state accessible to the user.
2822 * ENV{SYSTEMD_WANTS} in udev rules will now implicitly escape
2823 the passed argument if applicable.
2825 Contributions from: Auke Kok, Colin Guthrie, Colin Walters,
2826 Cristian Rodríguez, Daniel Buch, Daniel Wallace, Dave Reisner,
2827 Evangelos Foutras, Greg Kroah-Hartman, Harald Hoyer, Josh
2828 Triplett, Kay Sievers, Lennart Poettering, Lukas Nykryn,
2829 MUNEDA Takahiro, Mantas Mikulėnas, Mirco Tischler, Nathaniel
2830 Chen, Nirbheek Chauhan, Ronny Chevalier, Ross Lagerwall, Tom
2831 Gundersen, Umut Tezduyar, Ville Skyttä, Zbigniew
2836 * The output of 'systemctl list-jobs' got some polishing. The
2837 '--type=' argument may now be passed more than once. A new
2838 command 'systemctl list-sockets' has been added which shows
2839 a list of kernel sockets systemd is listening on with the
2840 socket units they belong to, plus the units these socket
2843 * The experimental libsystemd-bus library got substantial
2844 updates to work in conjunction with the (also experimental)
2845 kdbus kernel project. It works well enough to exchange
2846 messages with some sophistication. Note that kdbus is not
2847 ready yet, and the library is mostly an elaborate test case
2848 for now, and not installable.
2850 * systemd gained a new unit 'systemd-static-nodes.service'
2851 that generates static device nodes earlier during boot, and
2852 can run in conjunction with udev.
2854 * libsystemd-login gained a new call sd_pid_get_user_unit()
2855 to retrieve the user systemd unit a process is running
2856 in. This is useful for systems where systemd is used as
2859 * systemd-nspawn now places all containers in the new /machine
2860 top-level cgroup directory in the name=systemd
2861 hierarchy. libvirt will soon do the same, so that we get a
2862 uniform separation of /system, /user and /machine for system
2863 services, user processes and containers/virtual
2864 machines. This new cgroup hierarchy is also useful to stick
2865 stable names to specific container instances, which can be
2866 recognized later this way (this name may be controlled
2867 via systemd-nspawn's new -M switch). libsystemd-login also
2868 gained a new call sd_pid_get_machine_name() to retrieve the
2869 name of the container/VM a specific process belongs to.
2871 * bootchart can now store its data in the journal.
2873 * libsystemd-journal gained a new call
2874 sd_journal_add_conjunction() for AND expressions to the
2875 matching logic. This can be used to express more complex
2876 logical expressions.
2878 * journactl can now take multiple --unit= and --user-unit=
2881 * The cryptsetup logic now understands the "luks.key=" kernel
2882 command line switch for specifying a file to read the
2883 decryption key from. Also, if a configured key file is not
2884 found the tool will now automatically fall back to prompting
2887 * Python systemd.journal module was updated to wrap recently
2888 added functions from libsystemd-journal. The interface was
2889 changed to bring the low level interface in s.j._Reader
2890 closer to the C API, and the high level interface in
2891 s.j.Reader was updated to wrap and convert all data about
2894 Contributions from: Anatol Pomozov, Auke Kok, Harald Hoyer,
2895 Henrik Grindal Bakken, Josh Triplett, Kay Sievers, Lennart
2896 Poettering, Lukas Nykryn, Mantas Mikulėnas Marius Vollmer,
2897 Martin Jansa, Martin Pitt, Michael Biebl, Michal Schmidt,
2898 Mirco Tischler, Pali Rohar, Simon Peeters, Steven Hiscocks,
2899 Tom Gundersen, Zbigniew Jędrzejewski-Szmek
2903 * journalctl --update-catalog now understands a new --root=
2904 option to operate on catalogs found in a different root
2907 * During shutdown after systemd has terminated all running
2908 services a final killing loop kills all remaining left-over
2909 processes. We will now print the name of these processes
2910 when we send SIGKILL to them, since this usually indicates a
2913 * If /etc/crypttab refers to password files stored on
2914 configured mount points automatic dependencies will now be
2915 generated to ensure the specific mount is established first
2916 before the key file is attempted to be read.
2918 * 'systemctl status' will now show information about the
2919 network sockets a socket unit is listening on.
2921 * 'systemctl status' will also shown information about any
2922 drop-in configuration file for units. (Drop-In configuration
2923 files in this context are files such as
2924 /etc/systemd/systemd/foobar.service.d/*.conf)
2926 * systemd-cgtop now optionally shows summed up CPU times of
2927 cgroups. Press '%' while running cgtop to switch between
2928 percentage and absolute mode. This is useful to determine
2929 which cgroups use up the most CPU time over the entire
2930 runtime of the system. systemd-cgtop has also been updated
2931 to be 'pipeable' for processing with further shell tools.
2933 * 'hostnamectl set-hostname' will now allow setting of FQDN
2936 * The formatting and parsing of time span values has been
2937 changed. The parser now understands fractional expressions
2938 such as "5.5h". The formatter will now output fractional
2939 expressions for all time spans under 1min, i.e. "5.123456s"
2940 rather than "5s 123ms 456us". For time spans under 1s
2941 millisecond values are shown, for those under 1ms
2942 microsecond values are shown. This should greatly improve
2943 all time-related output of systemd.
2945 * libsystemd-login and libsystemd-journal gained new
2946 functions for querying the poll() events mask and poll()
2947 timeout value for integration into arbitrary event
2950 * localectl gained the ability to list available X11 keymaps
2951 (models, layouts, variants, options).
2953 * 'systemd-analyze dot' gained the ability to filter for
2954 specific units via shell-style globs, to create smaller,
2955 more useful graphs. I.e. it is now possible to create simple
2956 graphs of all the dependencies between only target units, or
2957 of all units that Avahi has dependencies with.
2959 Contributions from: Cristian Rodríguez, Dr. Tilmann Bubeck,
2960 Harald Hoyer, Holger Hans Peter Freyther, Kay Sievers, Kelly
2961 Anderson, Koen Kooi, Lennart Poettering, Maksim Melnikau,
2962 Marc-Antoine Perennou, Marius Vollmer, Martin Pitt, Michal
2963 Schmidt, Oleksii Shevchuk, Ronny Chevalier, Simon McVittie,
2964 Steven Hiscocks, Thomas Weißschuh, Umut Tezduyar, Václav
2965 Pavlín, Zbigniew Jędrzejewski-Szmek, Łukasz Stelmach
2969 * The boot-time readahead implementation for rotating media
2970 will now read the read-ahead data in multiple passes which
2971 consist of all read requests made in equidistant time
2972 intervals. This means instead of strictly reading read-ahead
2973 data in its physical order on disk we now try to find a
2974 middle ground between physical and access time order.
2976 * /etc/os-release files gained a new BUILD_ID= field for usage
2977 on operating systems that provide continuous builds of OS
2980 Contributions from: Auke Kok, Eelco Dolstra, Kay Sievers,
2981 Lennart Poettering, Lukas Nykryn, Martin Pitt, Václav Pavlín
2982 William Douglas, Zbigniew Jędrzejewski-Szmek
2986 * systemd-python gained an API exposing libsystemd-daemon.
2988 * The SMACK setup logic gained support for uploading CIPSO
2991 * Behaviour of PrivateTmp=, ReadWriteDirectories=,
2992 ReadOnlyDirectories= and InaccessibleDirectories= has
2993 changed. The private /tmp and /var/tmp directories are now
2994 shared by all processes of a service (which means
2995 ExecStartPre= may now leave data in /tmp that ExecStart= of
2996 the same service can still access). When a service is
2997 stopped its temporary directories are immediately deleted
2998 (normal clean-up with tmpfiles is still done in addition to
3001 * By default, systemd will now set a couple of sysctl
3002 variables in the kernel: the safe sysrq options are turned
3003 on, IP route verification is turned on, and source routing
3004 disabled. The recently added hardlink and softlink
3005 protection of the kernel is turned on. These settings should
3006 be reasonably safe, and good defaults for all new systems.
3008 * The predictable network naming logic may now be turned off
3009 with a new kernel command line switch: net.ifnames=0.
3011 * A new libsystemd-bus module has been added that implements a
3012 pretty complete D-Bus client library. For details see:
3014 http://lists.freedesktop.org/archives/systemd-devel/2013-March/009797.html
3016 * journald will now explicitly flush the journal files to disk
3017 at the latest 5min after each write. The file will then also
3018 be marked offline until the next write. This should increase
3019 reliability in case of a crash. The synchronization delay
3020 can be configured via SyncIntervalSec= in journald.conf.
3022 * There's a new remote-fs-setup.target unit that can be used
3023 to pull in specific services when at least one remote file
3024 system is to be mounted.
3026 * There are new targets timers.target and paths.target as
3027 canonical targets to pull user timer and path units in
3028 from. This complements sockets.target with a similar
3029 purpose for socket units.
3031 * libudev gained a new call udev_device_set_attribute_value()
3032 to set sysfs attributes of a device.
3034 * The udev daemon now sets the default number of worker
3035 processes executed in parallel based on the number of available
3036 CPUs instead of the amount of available RAM. This is supposed
3037 to provide a more reliable default and limit a too aggressive
3038 paralellism for setups with 1000s of devices connected.
3040 Contributions from: Auke Kok, Colin Walters, Cristian
3041 Rodríguez, Daniel Buch, Dave Reisner, Frederic Crozat, Hannes
3042 Reinecke, Harald Hoyer, Jan Alexander Steffens, Jan
3043 Engelhardt, Josh Triplett, Kay Sievers, Lennart Poettering,
3044 Mantas Mikulėnas, Martin Pitt, Mathieu Bridon, Michael Biebl,
3045 Michal Schmidt, Michal Sekletar, Miklos Vajna, Nathaniel Chen,
3046 Oleksii Shevchuk, Ozan Çağlayan, Thomas Hindoe Paaboel
3047 Andersen, Tollef Fog Heen, Tom Gundersen, Umut Tezduyar,
3048 Zbigniew Jędrzejewski-Szmek
3052 * Configuration of unit files may now be extended via drop-in
3053 files without having to edit/override the unit files
3054 themselves. More specifically, if the administrator wants to
3055 change one value for a service file foobar.service he can
3056 now do so by dropping in a configuration snippet into
3057 /etc/systemd/system/foobar.service.d/*.conf. The unit logic
3058 will load all these snippets and apply them on top of the
3059 main unit configuration file, possibly extending or
3060 overriding its settings. Using these drop-in snippets is
3061 generally nicer than the two earlier options for changing
3062 unit files locally: copying the files from
3063 /usr/lib/systemd/system/ to /etc/systemd/system/ and editing
3064 them there; or creating a new file in /etc/systemd/system/
3065 that incorporates the original one via ".include". Drop-in
3066 snippets into these .d/ directories can be placed in any
3067 directory systemd looks for units in, and the usual
3068 overriding semantics between /usr/lib, /etc and /run apply
3071 * Most unit file settings which take lists of items can now be
3072 reset by assigning the empty string to them. For example,
3073 normally, settings such as Environment=FOO=BAR append a new
3074 environment variable assignment to the environment block,
3075 each time they are used. By assigning Environment= the empty
3076 string the environment block can be reset to empty. This is
3077 particularly useful with the .d/*.conf drop-in snippets
3078 mentioned above, since this adds the ability to reset list
3079 settings from vendor unit files via these drop-ins.
3081 * systemctl gained a new "list-dependencies" command for
3082 listing the dependencies of a unit recursively.
3084 * Inhibitors are now honored and listed by "systemctl
3085 suspend", "systemctl poweroff" (and similar) too, not only
3086 GNOME. These commands will also list active sessions by
3089 * Resource limits (as exposed by the various control group
3090 controllers) can now be controlled dynamically at runtime
3091 for all units. More specifically, you can now use a command
3092 like "systemctl set-cgroup-attr foobar.service cpu.shares
3093 2000" to alter the CPU shares a specific service gets. These
3094 settings are stored persistently on disk, and thus allow the
3095 administrator to easily adjust the resource usage of
3096 services with a few simple commands. This dynamic resource
3097 management logic is also available to other programs via the
3098 bus. Almost any kernel cgroup attribute and controller is
3101 * systemd-vconsole-setup will now copy all font settings to
3102 all allocated VTs, where it previously applied them only to
3105 * libsystemd-login gained the new sd_session_get_tty() API
3108 * This release drops support for a few legacy or
3109 distribution-specific LSB facility names when parsing init
3110 scripts: $x-display-manager, $mail-transfer-agent,
3111 $mail-transport-agent, $mail-transfer-agent, $smtp,
3112 $null. Also, the mail-transfer-agent.target unit backing
3113 this has been removed. Distributions which want to retain
3114 compatibility with this should carry the burden for
3115 supporting this themselves and patch support for these back
3116 in, if they really need to. Also, the facilities $syslog and
3117 $local_fs are now ignored, since systemd does not support
3118 early-boot LSB init scripts anymore, and these facilities
3119 are implied anyway for normal services. syslog.target has
3122 * There are new bus calls on PID1's Manager object for
3123 cancelling jobs, and removing snapshot units. Previously,
3124 both calls were only available on the Job and Snapshot
3127 * systemd-journal-gatewayd gained SSL support.
3129 * The various "environment" files, such as /etc/locale.conf
3130 now support continuation lines with a backslash ("\") as
3131 last character in the line, similarly in style (but different)
3132 to how this is supported in shells.
3134 * For normal user processes the _SYSTEMD_USER_UNIT= field is
3135 now implicitly appended to every log entry logged. systemctl
3136 has been updated to filter by this field when operating on a
3137 user systemd instance.
3139 * nspawn will now implicitly add the CAP_AUDIT_WRITE and
3140 CAP_AUDIT_CONTROL capabilities to the capabilities set for
3141 the container. This makes it easier to boot unmodified
3142 Fedora systems in a container, which however still requires
3143 audit=0 to be passed on the kernel command line. Auditing in
3144 kernel and userspace is unfortunately still too broken in
3145 context of containers, hence we recommend compiling it out
3146 of the kernel or using audit=0. Hopefully this will be fixed
3147 one day for good in the kernel.
3149 * nspawn gained the new --bind= and --bind-ro= parameters to
3150 bind mount specific directories from the host into the
3153 * nspawn will now mount its own devpts file system instance
3154 into the container, in order not to leak pty devices from
3155 the host into the container.
3157 * systemd will now read the firmware boot time performance
3158 information from the EFI variables, if the used boot loader
3159 supports this, and takes it into account for boot performance
3160 analysis via "systemd-analyze". This is currently supported
3161 only in conjunction with Gummiboot, but could be supported
3162 by other boot loaders too. For details see:
3164 http://www.freedesktop.org/wiki/Software/systemd/BootLoaderInterface
3166 * A new generator has been added that automatically mounts the
3167 EFI System Partition (ESP) to /boot, if that directory
3168 exists, is empty, and no other file system has been
3169 configured to be mounted there.
3171 * logind will now send out PrepareForSleep(false) out
3172 unconditionally, after coming back from suspend. This may be
3173 used by applications as asynchronous notification for
3174 system resume events.
3176 * "systemctl unlock-sessions" has been added, that allows
3177 unlocking the screens of all user sessions at once, similar
3178 to how "systemctl lock-sessions" already locked all users
3179 sessions. This is backed by a new D-Bus call UnlockSessions().
3181 * "loginctl seat-status" will now show the master device of a
3182 seat. (i.e. the device of a seat that needs&nb
~ianmdlvl / elogind.git / blob