chiark / gitweb /
9 years agoudev: keymap - remove weird 'c2 a0' character sequences which break the check
Kay Sievers [Wed, 11 Apr 2012 22:42:43 +0000 (00:42 +0200)]
udev: keymap - remove weird 'c2 a0' character sequences which break the check

$ hexdump -C  src/udev/keymap/keymaps/samsung-90x3a
00000000  30 78 39 36 20 6b 62 64  69 6c 6c 75 6d 75 70 c2  |0x96 kbdillumup.|
00000010  a0 c2 a0 c2 a0 c2 a0 c2  a0 c2 a0 c2 a0 c2 a0 20  |............... |
00000020  23 20 46 6e 2b 46 38 20  6b 65 79 62 6f 61 72 64  |# Fn+F8 keyboard|

9 years agoOne can specify in which cgroup hierarchies a systemd-nspawn container will appear
Léo Gillot-Lamure [Wed, 11 Apr 2012 11:27:19 +0000 (12:27 +0100)]
One can specify in which cgroup hierarchies a systemd-nspawn container will appear

9 years agorelicense to LGPLv2.1 (with exceptions)
Lennart Poettering [Wed, 11 Apr 2012 22:20:58 +0000 (00:20 +0200)]
relicense to LGPLv2.1 (with exceptions)

We finally got the OK from all contributors with non-trivial commits to
relicense systemd from GPL2+ to LGPL2.1+.

Some udev bits continue to be GPL2+ for now, but we are looking into
relicensing them too, to allow free copy/paste of all code within

The bits that used to be MIT continue to be MIT.

The big benefit of the relicensing is that closed source code may now
link against and friends.

9 years agoudev: - fix source directory
Kay Sievers [Wed, 11 Apr 2012 22:08:26 +0000 (00:08 +0200)]
udev: - fix source directory

9 years agoupdate TODO
Lennart Poettering [Wed, 11 Apr 2012 21:56:58 +0000 (23:56 +0200)]
update TODO

9 years agopolkit: when spawning off agent, wait until the agent is fully initialized
Lennart Poettering [Wed, 11 Apr 2012 20:37:13 +0000 (22:37 +0200)]
polkit: when spawning off agent, wait until the agent is fully initialized

9 years agounits: exclude gettys from isolate requests
Lennart Poettering [Wed, 11 Apr 2012 19:58:33 +0000 (21:58 +0200)]
units: exclude gettys from isolate requests

gettys are nowadays mostly autospawned and hence usually subject to
being shut down on isolate requests, since they are no dependency of any
other unit. This is a bad idea if the user isolates between
multi-user.graphical and, hence exclude them from the

This has the effect that gettys no longer cleaned up when is isolated, which might actualy be considered a
feature, even though it is a change from previous behaviour...

Note that the one getty that really matters (the one on tty1) is still
removed when isolating to since it conflicts with

9 years agoudev: fix RUN+= logic regarding builtin vs. program
Kay Sievers [Wed, 11 Apr 2012 20:34:25 +0000 (22:34 +0200)]
udev: fix RUN+= logic regarding builtin vs. program

9 years agoudev: fix keymap mkdir
Marc-Antoine Perennou [Wed, 11 Apr 2012 20:24:53 +0000 (22:24 +0200)]
udev: fix keymap mkdir

9 years agoenable proper access timestamps on all tmpfs mounts
Kay Sievers [Wed, 11 Apr 2012 20:18:55 +0000 (22:18 +0200)]
enable proper access timestamps on all tmpfs mounts

9 years agotmpfiles: open directories with O_NOATIME to preserve timestamp
Kay Sievers [Wed, 11 Apr 2012 19:33:12 +0000 (21:33 +0200)]
tmpfiles: open directories with O_NOATIME to preserve timestamp

  # stat /tmp/pulse-Du5ectm60QYM | grep 'Access: 20'
  Access: 2012-04-11 21:32:48.444920237 +0200
  # systemd-tmpfiles --clean
  # stat /tmp/pulse-Du5ectm60QYM | grep 'Access: 20'
  Access: 2012-04-11 21:36:27.628925459 +0200

  # stat /tmp/pulse-Du5ectm60QYM | grep 'Access: 20'
  Access: 2012-04-11 21:32:48.444920237 +0200
  # ./systemd-tmpfiles --clean
  # stat /tmp/pulse-Du5ectm60QYM | grep 'Access: 20'
  Access: 2012-04-11 21:32:48.444920237 +0200


9 years agopolkit: temporarily spawn of a polkit agent in terminals for possibly authenticated...
Lennart Poettering [Wed, 11 Apr 2012 16:50:16 +0000 (18:50 +0200)]
polkit: temporarily spawn of a polkit agent in terminals for possibly authenticated operations

9 years agomain: disarm watchdog when preparing for reexecution
Lennart Poettering [Wed, 11 Apr 2012 16:40:22 +0000 (18:40 +0200)]
main: disarm watchdog when preparing for reexecution

9 years agoput in 'if HAVE_ACL' and rename acl.[ch] to acl-util.[ch]
Kay Sievers [Wed, 11 Apr 2012 14:39:31 +0000 (16:39 +0200)]
put in 'if HAVE_ACL' and rename acl.[ch] to acl-util.[ch]

9 years agomove sources into core/
Kay Sievers [Wed, 11 Apr 2012 10:59:52 +0000 (12:59 +0200)]
move sources into core/

9 years agomain: unset some bash specific environment variables that might get leaked to us
Lennart Poettering [Wed, 11 Apr 2012 11:19:26 +0000 (13:19 +0200)]
main: unset some bash specific environment variables that might get leaked to us

9 years agomain: drop container/initrd env vars from inherited set
Lennart Poettering [Wed, 11 Apr 2012 10:56:51 +0000 (12:56 +0200)]
main: drop container/initrd env vars from inherited set

Leave the env vars used in the container/initrd logic set for PID1, but
don't inherit them to any children.

9 years agomachine-id: don't delete runtime machine-id, and place it in /run
Lennart Poettering [Wed, 11 Apr 2012 10:49:52 +0000 (12:49 +0200)]
machine-id: don't delete runtime machine-id, and place it in /run

9 years agomain: add URL to cgroups check message
Lennart Poettering [Wed, 11 Apr 2012 10:48:27 +0000 (12:48 +0200)]
main: add URL to cgroups check message

9 years agosplit selinux label operations out of cgroup-util, socket-util
Kay Sievers [Wed, 11 Apr 2012 10:30:53 +0000 (12:30 +0200)]
split selinux label operations out of cgroup-util, socket-util

This prevents linking of selinux and libdl for another 15 binaries.

9 years agobuild-sys: move remaining headers out EXTRA_DIST, install sd-readahead.h
Kay Sievers [Wed, 11 Apr 2012 01:54:23 +0000 (03:54 +0200)]
build-sys: move remaining headers out EXTRA_DIST, install sd-readahead.h

9 years agobuild-sys: add $(AM_LDFLAGS) where needed
Kay Sievers [Wed, 11 Apr 2012 01:29:31 +0000 (03:29 +0200)]
build-sys: add $(AM_LDFLAGS) where needed

9 years agopam_systemd: add missing libsystemd_audit.l
Kay Sievers [Wed, 11 Apr 2012 01:15:34 +0000 (03:15 +0200)]
pam_systemd: add missing libsystemd_audit.l

9 years agoshutdownd: rework interface, allow subscribing to scheduled shutdowns
Lennart Poettering [Wed, 11 Apr 2012 00:04:46 +0000 (02:04 +0200)]
shutdownd: rework interface, allow subscribing to scheduled shutdowns

This extends the shutdownd interface to expose schedule shutdown
information in /run/systemd/shutdown/schedule.

This also cleans up the shutdownd protocol and documents it in a header
file sd-shutdown.h.

This is supposed to be used by client code that wants to control and
monitor scheduled shutdown.

9 years agosystemctl: don't forward poweroff/reboot requests to logind if time is set or immedia...
Lennart Poettering [Tue, 10 Apr 2012 22:36:44 +0000 (00:36 +0200)]
systemctl: don't forward poweroff/reboot requests to logind if time is set or immediate or dry-run execution is requested

logind can't handle scheduled poweroff/reboot requests, nor
immediate/dry-run requests, hence don't attempt forwarding to logind if
these options are used.

9 years agologs-show: fix output of log lines lacking comm
Lennart Poettering [Tue, 10 Apr 2012 22:34:06 +0000 (00:34 +0200)]
logs-show: fix output of log lines lacking comm

9 years agounits: introduce
Lennart Poettering [Tue, 10 Apr 2012 22:11:18 +0000 (00:11 +0200)]
units: introduce

This separates user/group NSS lookups from host/network NSS lookups.

By default order all network mounts after host/network NSS lookups now,
and logind execution after user/group NSS lookups.

9 years agoupdate TODO
Kay Sievers [Tue, 10 Apr 2012 20:47:48 +0000 (22:47 +0200)]
update TODO

9 years agorename to and put selinux deps in
Kay Sievers [Tue, 10 Apr 2012 19:54:31 +0000 (21:54 +0200)]
rename to and put selinux deps in

Only 34 of 74 tools need libselinux linked, and libselinux is a pain
with its unconditional library constructor.

9 years agoudev: move man pages to udev section
Kay Sievers [Tue, 10 Apr 2012 17:57:23 +0000 (19:57 +0200)]
udev: move man pages to udev section

9 years agodbus: make the service property StartLimitAction writeable
Michael Olbrich [Fri, 6 Apr 2012 19:38:02 +0000 (21:38 +0200)]
dbus: make the service property StartLimitAction writeable

9 years agodbus: add generic DEFINE_BUS_PROPERTY_SET_ENUM() macro for enum properties
Michael Olbrich [Fri, 6 Apr 2012 19:38:01 +0000 (21:38 +0200)]
dbus: add generic DEFINE_BUS_PROPERTY_SET_ENUM() macro for enum properties

The defined function can be used as BusPropertySetCallback.

9 years agodbus: add data argument to BusPropertySetCallback
Michael Olbrich [Fri, 6 Apr 2012 19:38:00 +0000 (21:38 +0200)]
dbus: add data argument to BusPropertySetCallback

BusPropertyCallback already has the argument. It is necesary for the
callback to know what data to access.

9 years agowatchdog: really return the actual watchdog timeout
Michael Olbrich [Fri, 6 Apr 2012 19:37:26 +0000 (21:37 +0200)]
watchdog: really return the actual watchdog timeout

In the current code setting the return argument is never reached.

9 years agomove list.h, macro.h, ioprio.h to shared/
Kay Sievers [Tue, 10 Apr 2012 17:42:07 +0000 (19:42 +0200)]
move list.h, macro.h, ioprio.h to shared/

9 years agomove pager.[ch] to shared/
Kay Sievers [Tue, 10 Apr 2012 17:32:48 +0000 (19:32 +0200)]
move pager.[ch] to shared/

9 years agouse instead of source file
Kay Sievers [Tue, 10 Apr 2012 17:23:08 +0000 (19:23 +0200)]
use instead of source file

9 years agouse instead of source file
Kay Sievers [Tue, 10 Apr 2012 17:14:42 +0000 (19:14 +0200)]
use instead of source file

9 years agomove dep from basic to core
Kay Sievers [Tue, 10 Apr 2012 16:59:47 +0000 (18:59 +0200)]
move dep from basic to core

9 years agomove cgroup-util.[ch] to shared/
Kay Sievers [Tue, 10 Apr 2012 16:46:51 +0000 (18:46 +0200)]
move cgroup-util.[ch] to shared/

9 years agoudev: remove deps already pulled in by LIBADD
Kay Sievers [Tue, 10 Apr 2012 15:39:43 +0000 (17:39 +0200)]
udev: remove deps already pulled in by LIBADD

9 years agoudev: fix gcc warnings showing up after adding $(AM_CFLAGS)
Kay Sievers [Tue, 10 Apr 2012 14:41:52 +0000 (16:41 +0200)]
udev: fix gcc warnings showing up after adding $(AM_CFLAGS)

9 years agobuild-sys: add $(AM_CFLAGS) where needed
Kay Sievers [Tue, 10 Apr 2012 14:02:09 +0000 (16:02 +0200)]
build-sys: add $(AM_CFLAGS) where needed

9 years agoutil: move ACL code into internal library
Kay Sievers [Tue, 10 Apr 2012 12:53:38 +0000 (14:53 +0200)]
util: move ACL code into internal library

9 years agoutil: move all to shared/ and split external dependencies in separate internal libraries
Kay Sievers [Tue, 10 Apr 2012 11:39:02 +0000 (13:39 +0200)]
util: move all to shared/ and split external dependencies in separate internal libraries

  $ ldd /lib/systemd/systemd-timestamp =>  (0x00007fffb05ff000) => /lib64/ (0x00007f90aac57000) => /lib64/ (0x00007f90aaa53000) => /lib64/ (0x00007f90aa84a000) => /lib64/ (0x00007f90aa494000)
  /lib64/ (0x00007f90aae90000) => /lib64/ (0x00007f90aa290000) => /lib64/ (0x00007f90aa08a000) => /lib64/ (0x00007f90a9e6e000)

  $ ldd systemd-timestamp =>  (0x00007fff3cbff000) => /lib64/ (0x00007f5eaa1c3000) => /lib64/ (0x00007f5ea9fbb000) => /lib64/ (0x00007f5ea9c04000)
  /lib64/ (0x00007f5eaa3fc000) => /lib64/ (0x00007f5ea9a00000) => /lib64/ (0x00007f5ea97e4000)

9 years agoudev: set errno = ENOSYS for removed interfaces
Kay Sievers [Tue, 10 Apr 2012 11:33:43 +0000 (13:33 +0200)]
udev: set errno = ENOSYS for removed interfaces

9 years agoupdate TODO
Kay Sievers [Mon, 9 Apr 2012 20:07:18 +0000 (22:07 +0200)]
update TODO

9 years agoudev: remove RUN+="socket:.." and udev_monitor_new_from_netlink()
Kay Sievers [Mon, 9 Apr 2012 18:45:45 +0000 (20:45 +0200)]
udev: remove RUN+="socket:.."  and udev_monitor_new_from_netlink()

9 years agoudev: convert 'uaccess' to a builtin
Kay Sievers [Mon, 9 Apr 2012 14:37:54 +0000 (16:37 +0200)]
udev: convert 'uaccess' to a builtin

9 years agoudev: keymap: remove rootprefix mangling from script
Kay Sievers [Sun, 8 Apr 2012 15:08:09 +0000 (17:08 +0200)]
udev: keymap: remove rootprefix mangling from script

9 years agoudev: fix test-udev binary
Kay Sievers [Sun, 8 Apr 2012 15:03:17 +0000 (17:03 +0200)]
udev: fix test-udev binary

9 years agoudev: remove support for /lib/udev/devices/; tmpfiles should be used
Kay Sievers [Sun, 8 Apr 2012 14:50:16 +0000 (16:50 +0200)]
udev: remove support for /lib/udev/devices/; tmpfiles should be used

9 years agoudev: fix rules sort order
Tom Gundersen [Sat, 7 Apr 2012 11:41:13 +0000 (13:41 +0200)]
udev: fix rules sort order

Commit 91418155ae9034f466d436c314cd136309bc557d moved around the code,
but did not chang ethe array index.

9 years agoudev: switch to systemd logging functions
Kay Sievers [Sun, 8 Apr 2012 14:06:20 +0000 (16:06 +0200)]
udev: switch to systemd logging functions

9 years agobuild-sys: bump systemd version to 'udev version 182'
Kay Sievers [Fri, 6 Apr 2012 17:52:49 +0000 (19:52 +0200)]
build-sys: bump systemd version to 'udev version 182'

9 years agosystemd: add hardware watchdog support
Lennart Poettering [Thu, 5 Apr 2012 20:08:10 +0000 (22:08 +0200)]
systemd: add hardware watchdog support

This adds minimal hardware watchdog support to PID 1. The idea is that
PID 1 supervises and watchdogs system services, while the hardware
watchdog is used to supervise PID 1.

This adds two hardware watchdog configuration options, for the runtime
watchdog and for a shutdown watchdog. The former is active during normal
operation, the latter only at reboots to ensure that if a clean reboot
times out we reboot nonetheless.

If the runtime watchdog is enabled PID 1 will automatically wake up at
half the configured interval and write to the watchdog daemon.

By default we enable the shutdown watchdog, but leave the runtime
watchdog disabled in order not to break independent hardware watchdog
daemons people might be using.

This is only the most basic hookup. If necessary we can later on hook
up the watchdog ping more closely with services deemed crucial.

9 years agojob: use a lookup table for merging of job types
Michal Schmidt [Thu, 5 Apr 2012 06:34:05 +0000 (08:34 +0200)]
job: use a lookup table for merging of job types

It is easier to see what job_type_merge() is doing when the merging
rules are written in the form of a table.

job_type_is_superset() contained redundant information. It can be
simplified to a simple rule: Type A is a superset of B iff merging A
with B gives A.

Two job types are conflicting iff they are not mergeable.

Make job_type_lookup_merge() the core function to decide the type
merging. All other job_type_*() are just short wrappers around it.
They can be inline.

test-job-type gives the same results as before.
btw, the systemd binary is smaller by almost 1 KB.

9 years agoupdate TODO
Kay Sievers [Wed, 4 Apr 2012 12:57:36 +0000 (14:57 +0200)]
update TODO

9 years agoman: update udev man pages
Kay Sievers [Wed, 4 Apr 2012 12:42:32 +0000 (14:42 +0200)]
man: update udev man pages

9 years agobuild-sys: add a few missing headers
Kay Sievers [Wed, 4 Apr 2012 12:08:37 +0000 (14:08 +0200)]
build-sys: add a few missing headers

9 years agounits: direct users to the journal for logs when entering rescue mode
Lennart Poettering [Wed, 4 Apr 2012 11:52:02 +0000 (13:52 +0200)]
units: direct users to the journal for logs when entering rescue mode

9 years agoudev: replace UDEV_EXPORT with _public_
Kay Sievers [Wed, 4 Apr 2012 11:30:09 +0000 (13:30 +0200)]
udev: replace UDEV_EXPORT with _public_

9 years agoudev: ata_id - remove assert() until we switch over to systemd logging
Kay Sievers [Wed, 4 Apr 2012 11:29:37 +0000 (13:29 +0200)]
udev: ata_id - remove assert() until we switch over to systemd logging

9 years agoudev: enable logging
Kay Sievers [Wed, 4 Apr 2012 11:14:29 +0000 (13:14 +0200)]
udev: enable logging

9 years agobuild-sys: remove vala hack, which did not allow to list headers in sources
Kay Sievers [Wed, 4 Apr 2012 10:48:07 +0000 (12:48 +0200)]
build-sys: remove vala hack, which did not allow to list headers in sources

9 years agokeymap: Add support for Lenovo v480 touchpad toggle hotkey.
Ayan George [Wed, 4 Apr 2012 06:38:38 +0000 (08:38 +0200)]
keymap: Add support for Lenovo v480 touchpad toggle hotkey.

Signed-off-by: Martin Pitt <>
9 years agounits/: use @SYSTEMCTL@ instead of hardcoded paths
Dave Reisner [Wed, 4 Apr 2012 04:22:21 +0000 (00:22 -0400)]
units/: use @SYSTEMCTL@ instead of hardcoded paths

Especially in the case of --enable-split-usr, several units will point
to the wrong location for systemctl. Use @SYSTEMCTL@ which will always
contain the proper path.

9 years agoudev: avoid building selinux parts without have_selinux
Dave Reisner [Wed, 4 Apr 2012 04:22:20 +0000 (00:22 -0400)]
udev: avoid building selinux parts without have_selinux

9 years agoudev: fix path in udev.service
Kay Sievers [Wed, 4 Apr 2012 04:21:23 +0000 (06:21 +0200)]
udev: fix path in udev.service

9 years agoudev: fix gcc warnings
Kay Sievers [Wed, 4 Apr 2012 03:31:21 +0000 (05:31 +0200)]
udev: fix gcc warnings

9 years agoMerge branch 'master' of ssh://
Kay Sievers [Wed, 4 Apr 2012 03:23:51 +0000 (05:23 +0200)]
Merge branch 'master' of ssh://

9 years agoudev: fix gcc warnings
Kay Sievers [Wed, 4 Apr 2012 03:18:14 +0000 (05:18 +0200)]
udev: fix gcc warnings

9 years agomove imported udev into place
Kay Sievers [Tue, 3 Apr 2012 19:24:46 +0000 (21:24 +0200)]
move imported udev into place

9 years agojournal: don't export the boot id twice per entry
Lennart Poettering [Tue, 3 Apr 2012 23:00:09 +0000 (01:00 +0200)]
journal: don't export the boot id twice per entry

9 years agoman: document the _TRANSPORT journal field
Lennart Poettering [Tue, 3 Apr 2012 22:43:40 +0000 (00:43 +0200)]
man: document the _TRANSPORT journal field

9 years agoman: clarify the formatting of timestamps
Lennart Poettering [Tue, 3 Apr 2012 21:08:04 +0000 (23:08 +0200)]
man: clarify the formatting of timestamps

9 years agoman: update documentation of special units
Lennart Poettering [Tue, 3 Apr 2012 20:32:05 +0000 (22:32 +0200)]
man: update documentation of special units

9 years agoman: document special journal fields
Lennart Poettering [Tue, 3 Apr 2012 20:31:48 +0000 (22:31 +0200)]
man: document special journal fields

9 years agoimport udev repository
Kay Sievers [Tue, 3 Apr 2012 19:08:04 +0000 (21:08 +0200)]
import udev repository

9 years agologind: log with AUTH facility
Lennart Poettering [Tue, 3 Apr 2012 17:25:29 +0000 (19:25 +0200)]
logind: log with AUTH facility

9 years agofix a couple of AF_UNIX connect() calls
Lennart Poettering [Tue, 3 Apr 2012 17:24:12 +0000 (19:24 +0200)]
fix a couple of AF_UNIX connect() calls

9 years agojournal: in json and export mode use double underscores to prefix location fields
Lennart Poettering [Tue, 3 Apr 2012 16:15:28 +0000 (18:15 +0200)]
journal: in json and export mode use double underscores to prefix location fields

Many programming languages don't allow variable names beginning in dots,
hence let's use double underscores for the location fields instead. This
gets us the simple rule:

__ is the prefix for location fields (i.e. fields that are used to
identify entries, rather than part of the entries)

_ is the prefix for trusted fields (i.e. those fields journald itself
adds to all entries)

no prefix for unrusted fields (i.e. all fields normal client code sends

9 years agosystemctl: make -f short for both --follow and --force
Lennart Poettering [Tue, 3 Apr 2012 12:43:48 +0000 (14:43 +0200)]
systemctl: make -f short for both --follow and --force

9 years agoupdate TODO
Lennart Poettering [Tue, 3 Apr 2012 12:25:51 +0000 (14:25 +0200)]
update TODO

9 years agoservice: schedule JOB_RESTART from SERVICE_AUTO_RESTART state
David Ward [Tue, 3 Apr 2012 03:45:35 +0000 (23:45 -0400)]
service: schedule JOB_RESTART from SERVICE_AUTO_RESTART state


9 years agoanalyze: Cosmetic exit when the bootup is not yet complete when plotting.
Colin Guthrie [Mon, 2 Apr 2012 10:08:20 +0000 (11:08 +0100)]
analyze: Cosmetic exit when the bootup is not yet complete when plotting.

This is just a nicer message than a python traceback.

9 years agoRevert "socket: if we fail to create an instantiated service for a socket, don't...
Michal Schmidt [Mon, 2 Apr 2012 19:49:00 +0000 (21:49 +0200)]
Revert "socket: if we fail to create an instantiated service for a socket, don't put the socket in failure mode"

This reverts commit 9586cdfab6a2638078702b7fea7e16b3a71899e2.
(but not the TODO hunk).

The bug was already fixed by 1a710b43. And if other errors occur, we
don't want to leave the socket active in order to avoid having socket

9 years agoinstall: check for proper return from dirent_ensure_type
Dave Reisner [Mon, 2 Apr 2012 13:08:35 +0000 (09:08 -0400)]
install: check for proper return from dirent_ensure_type

Fixes 'systemctl list-unit-files', which previously returned only:

  Failed to issue method call: No such file or directory

9 years agoupdate TODO
Lennart Poettering [Mon, 2 Apr 2012 18:54:15 +0000 (20:54 +0200)]
update TODO

9 years agojournal: implicitly add code location to all messages logged with the native interface
Lennart Poettering [Mon, 2 Apr 2012 17:24:30 +0000 (19:24 +0200)]
journal: implicitly add code location to all messages logged with the native interface

This logic can be turned off by defining SD_JOURNAL_SUPPRESS_LOCATION
before including sd-journal.h.

This also saves/restores errno in all logging functions, in order to be
useful as logging calls without side-effects.

This also adds a couple of __unlikely__ around the early checks in the
logging calls, in order to minimize the runtime impact.

9 years agojournal: decrease default mmap window size to allow a bigger number of journals to...
Lennart Poettering [Mon, 2 Apr 2012 17:04:56 +0000 (19:04 +0200)]
journal: decrease default mmap window size to allow a bigger number of journals to be traversed in parallel

9 years agobash-completion: update naming of loginctl
Dave Reisner [Sun, 1 Apr 2012 23:04:23 +0000 (19:04 -0400)]
bash-completion: update naming of loginctl

18b754d3 changed the name of systemd-loginctl to loginctl, but didn't
update the bash-completion to match.

9 years agorename /etc/systemd/systemd-{login,journal}d.conf to {login,journal}d.conf
Kay Sievers [Fri, 30 Mar 2012 21:18:33 +0000 (23:18 +0200)]
rename /etc/systemd/systemd-{login,journal}d.conf to {login,journal}d.conf

9 years agojob: add debug prints where job type gets changed
Michal Schmidt [Tue, 27 Mar 2012 23:26:04 +0000 (01:26 +0200)]
job: add debug prints where job type gets changed

9 years agojob: fix loss of ordering with restart jobs
Michal Schmidt [Tue, 27 Mar 2012 22:42:27 +0000 (00:42 +0200)]
job: fix loss of ordering with restart jobs

Suppose that foo.service/start is a job waiting on other job bar.service/start
to finish. And then foo.service/restart is enqueued (not using

Currently this makes foo.service start immediately, forgetting about the
ordering to bar.service.

The runnability check for JOB_RESTART jobs looks only at dependencies for
stopping. That's actually correct, because restart jobs should be treated the
same as stop jobs at first. The bug is that job_run_and_invalidate() does not
treat them exactly the same as stop jobs. unit_start() gets called without
checking for the runnability of the converted JOB_START job.

The fix is to simplify the switch in job_run_and_invalidate(). Handle
JOB_RESTART identically to JOB_STOP.
Also simplify the handling of JOB_TRY_RESTART - just convert it to JOB_RESTART
if the unit is active and let it fall through to the JOB_RESTART case.
Similarly for JOB_RELOAD_OR_START - have a fall through to JOB_START.

In job_finish_and_invalidate() it's not necessary to check for JOB_TRY_RESTART
with JOB_DONE, because JOB_TRY_RESTART jobs will have been converted to
JOB_RESTART already.

Speeding up the restart of services in "auto-restart" state still works as

but it's still not perfect. With this fix the try-restart action will wait for
the restart to complete in the right order, but the optimal behaviour would be
to finish quickly (without disturbing the start job).

9 years agojournal: properly handle if we interleave files with different boot ids
Lennart Poettering [Tue, 27 Mar 2012 16:50:34 +0000 (18:50 +0200)]
journal: properly handle if we interleave files with different boot ids

If we try to locate a monotonic time in a file that doesn't have any
entries with the matching boot id, then don't fail on it, simply
fall back to calendar time.

9 years agounits: mount /tmp as tmpfs
Kay Sievers [Tue, 27 Mar 2012 15:13:05 +0000 (17:13 +0200)]
units: mount /tmp as tmpfs

The default setups should be a stateless as possible. /tmp as tmpfs is
the intended default for general purpose systems.

Small temporary files should not be stored on disk; lager files, or
files which should potentially survive a reboot, belong into /var/tmp.

Also catch up with some good old UNIX history.

More details are here:

9 years agounits: get rid of var-run.mount and var-lock.mount
Lennart Poettering [Tue, 27 Mar 2012 15:11:00 +0000 (17:11 +0200)]
units: get rid of var-run.mount and var-lock.mount

Since a number of distribitions don't need this compat glue anymore drop
it from systemd upstream. Distributions which still haven't converted
to /run can steal these unit files from the git history if they need to.

9 years agounits: don't mount tmpfs on /media anymore
Lennart Poettering [Tue, 27 Mar 2012 15:04:22 +0000 (17:04 +0200)]
units: don't mount tmpfs on /media anymore

udisks2 doesn't use /media anymore, instead mounts removable media in a
user-private directory beneath /run. /media is hence mostly obsolete and
hence it makes little sense to continue to mount a tmpfs to it.

Distributions should consider dropping the mount point entirely since
nothing uses it anymore.

9 years agocat: fix priority type
Lennart Poettering [Mon, 26 Mar 2012 22:20:48 +0000 (00:20 +0200)]
cat: fix priority type

Needs to be "int", not "char". Spotted by Frederic Crozat.