chiark / gitweb /
9 years agounits: stop when it has done its job
Lennart Poettering [Mon, 25 Jun 2012 10:01:09 +0000 (12:01 +0200)]
units: stop when it has done its job

9 years agoman: document systemd-vconsole-setup.service
Lennart Poettering [Mon, 25 Jun 2012 10:00:43 +0000 (12:00 +0200)]
man: document systemd-vconsole-setup.service

9 years agoman: introduce bootup(7)
Lennart Poettering [Fri, 22 Jun 2012 21:14:19 +0000 (23:14 +0200)]
man: introduce bootup(7)

9 years agoman: add more references to k-c-l(7)
Lennart Poettering [Fri, 22 Jun 2012 14:59:32 +0000 (16:59 +0200)]
man: add more references to k-c-l(7)

9 years agoman: fix references to udev man page
Lennart Poettering [Fri, 22 Jun 2012 14:57:13 +0000 (16:57 +0200)]
man: fix references to udev man page

9 years agoman: add reference to bootparam(7) even though it is terribly out of date
Lennart Poettering [Fri, 22 Jun 2012 14:57:01 +0000 (16:57 +0200)]
man: add reference to bootparam(7) even though it is terribly out of date

9 years agoman: mention other initrd implementations in kernel-command-line(7)
Lennart Poettering [Fri, 22 Jun 2012 14:29:26 +0000 (16:29 +0200)]
man: mention other initrd implementations in kernel-command-line(7)

9 years agounit: drop the Names= option
Lennart Poettering [Fri, 22 Jun 2012 14:24:57 +0000 (16:24 +0200)]
unit: drop the Names= option

Names= is a source of errors, simply because alias names specified like
this only become relevant after a unit has been loaded but cannot be
used to load a unit.

Let's get rid of the confusion and drop this field. To establish alias
names peope should use symlinks, which have the the benefit of being
useful as key to load a unit, even though they are not taken into
account if unit names are listed but they haven't been explicitly
referenced before.

9 years agodbus: don't expose SysVInitPath and SysVRcndPath on the bus
Lennart Poettering [Fri, 22 Jun 2012 14:18:47 +0000 (16:18 +0200)]
dbus: don't expose SysVInitPath and SysVRcndPath on the bus

People should use systemd.pc if anything at all to determine these
directories, and people should not assume that the bus fields are part
of the supported API, so let's just drop this.

9 years agocore: remove sysv_console option
Lennart Poettering [Fri, 22 Jun 2012 11:16:28 +0000 (13:16 +0200)]
core: remove sysv_console option

This option never made much sense. It was originally intended to make
sure that the usual startup output of sysv scripts goes to the terminal.
However, since SysV scripts started from a terminal would not output to
that terminal, but rather /dev/console this effect was more often than
not actually taking place. Nowadays systemd has much nicer boot time
status output than SysV which makes the sysv output redundant. Finally,
all output of services goes to the journal anyway, and is not lost.

Hence, let's drop this option, and simplify things a bit.

9 years agosystemctl: automatically turn paths and unescaped unit names into proper unit names
Lennart Poettering [Fri, 22 Jun 2012 11:08:48 +0000 (13:08 +0200)]
systemctl: automatically turn paths and unescaped unit names into proper unit names

This makes sure that

  systemctl status /home

is implicitly translated to:

  systemctl status /home.mount

Similar, /dev/foobar becomes dev-foobar.device.

Also, all characters that cannot be part of a unit name are implicitly

9 years agomodules-load: parse driver=/rd.driver= kernel command line option
Lennart Poettering [Fri, 22 Jun 2012 09:39:04 +0000 (11:39 +0200)]
modules-load: parse driver=/rd.driver= kernel command line option

This generalizes logic that already has been available in dracut before.

9 years agoman: refer to udevd by its service name not binary name
Lennart Poettering [Fri, 22 Jun 2012 08:38:40 +0000 (10:38 +0200)]
man: refer to udevd by its service name not binary name

Since the binary name is now hidden away in /usr/lib/ the primary user
handle for the udev service is the unit name, hence change the man page
to be available under the unit name, and make the binary name an alias
for it.

9 years agofstab-generator: support fstab=/rd.fstab= kernel cmdline
Lennart Poettering [Fri, 22 Jun 2012 08:27:05 +0000 (10:27 +0200)]
fstab-generator: support fstab=/rd.fstab= kernel cmdline

This generalizes functionality already available in dracut.

9 years agoman: add more links to kernel-command-line(7)
Lennart Poettering [Fri, 22 Jun 2012 08:13:50 +0000 (10:13 +0200)]
man: add more links to kernel-command-line(7)

9 years agocryptsetup: allow configuration of LUKS disks via the kernel cmdline
Lennart Poettering [Fri, 22 Jun 2012 08:11:06 +0000 (10:11 +0200)]
cryptsetup: allow configuration of LUKS disks via the kernel cmdline

This generalizes a bit of the functionality already available in dracut.

9 years agoman: prettify logind man page
Lennart Poettering [Thu, 21 Jun 2012 23:47:43 +0000 (01:47 +0200)]
man: prettify logind man page

9 years agoman: minor typo fixes
Lennart Poettering [Thu, 21 Jun 2012 23:38:42 +0000 (01:38 +0200)]
man: minor typo fixes

9 years agoman: document all kernel command line options we understand
Lennart Poettering [Thu, 21 Jun 2012 23:35:52 +0000 (01:35 +0200)]
man: document all kernel command line options we understand

9 years agojournal: fix sd_journal_stream_fd()
Lennart Poettering [Thu, 21 Jun 2012 22:19:13 +0000 (00:19 +0200)]
journal: fix sd_journal_stream_fd()

9 years agojournal: set the _SYSTEMD_UNIT field for messages from terminated processes
Eelco Dolstra [Thu, 21 Jun 2012 20:40:47 +0000 (16:40 -0400)]
journal: set the _SYSTEMD_UNIT field for messages from terminated processes

As described in

the journal currently doesn't set fields such as _SYSTEMD_UNIT
properly for messages coming from processes that have already
terminated.  This means among other things that "systemctl status" may
not show some of the output of services that wrote messages just
before they exited.

This patch fixes this by having processes that log to the journal
write their unit identifier to journald when the connection to
/run/systemd/journal/stdout is opened.  Journald stores the unit ID
and uses it to fill in _SYSTEMD_UNIT when it cannot be obtained
normally (i.e. from the cgroup).  To prevent impersonating another
unit, this information is only used when the caller is root.

This doesn't fix the general problem of getting metadata about
messages from terminated processes (which requires some kernel
support), but it allows "systemctl status" and similar queries to do
the Right Thing for units that log via stdout/stderr.

9 years agoreadahead: minor code style fixes
Lennart Poettering [Thu, 21 Jun 2012 22:09:23 +0000 (00:09 +0200)]
readahead: minor code style fixes

9 years agoreadahead: make sure to close pack file before exiting, to be valgrind clean
Lennart Poettering [Thu, 21 Jun 2012 22:03:25 +0000 (00:03 +0200)]
readahead: make sure to close pack file before exiting, to be valgrind clean

9 years agoreadahead: use log_error() for logging errors
Lennart Poettering [Thu, 21 Jun 2012 21:58:31 +0000 (23:58 +0200)]
readahead: use log_error() for logging errors

9 years agoreadahead: make use of util.h's page_size() call
Lennart Poettering [Thu, 21 Jun 2012 21:56:03 +0000 (23:56 +0200)]
readahead: make use of util.h's page_size() call

9 years agoreadahead: merge three binaries into one
Lennart Poettering [Thu, 21 Jun 2012 21:53:20 +0000 (23:53 +0200)]
readahead: merge three binaries into one

since the binaries share much of the same code and we better load only
one binary instead of two from disk at early boot let's merge the three
readahead binaries into one. This also allows us to drop a lot of
duplicated code.

9 years agopreset: don't look for preset files in /lib unless /usr is split off
Lennart Poettering [Thu, 21 Jun 2012 21:22:53 +0000 (23:22 +0200)]
preset: don't look for preset files in /lib unless /usr is split off

9 years agoman: document the new RPM macros in daemon(7)
Lennart Poettering [Thu, 21 Jun 2012 20:36:37 +0000 (22:36 +0200)]
man: document the new RPM macros in daemon(7)

9 years agorpm: include RPM macro definition for .spec snippets
Lennart Poettering [Thu, 21 Jun 2012 20:14:17 +0000 (22:14 +0200)]
rpm: include RPM macro definition for .spec snippets

Let's try to standardize a bit the RPM macros used for
installing/uninstalling services.

This only covers the non-SysV compat bits, since that tends to vary
widely between the various distros.


Add %{?systemd_requires} to the header of the spec file. And then:

%systemd_post foobar.service

%systemd_preun foobar.service

%systemd_postun foobar.service

And, instead of the latter, in case the service shall be restarted on updates:

%systemd_postun_restart foobar.service

9 years agologin: wrap CanTTY and CanGraphical seat attributes in libsystemd-login
Lennart Poettering [Thu, 21 Jun 2012 14:52:54 +0000 (16:52 +0200)]
login: wrap CanTTY and CanGraphical seat attributes in libsystemd-login

9 years agoman: document new sd_session_get_state() call
Lennart Poettering [Thu, 21 Jun 2012 14:31:06 +0000 (16:31 +0200)]
man: document new sd_session_get_state() call

9 years agologind: introduce a state for session, being one of online, active, closing
Lennart Poettering [Thu, 21 Jun 2012 14:14:53 +0000 (16:14 +0200)]
logind: introduce a state for session, being one of online, active, closing

online = logged in
active = logged in and session is in the fg
closing = nominally logged out but some left-over processes still around

Related to:

9 years agologind: expose CanGraphical and CanTTY properties on seat objects
Lennart Poettering [Thu, 21 Jun 2012 11:48:01 +0000 (13:48 +0200)]
logind: expose CanGraphical and CanTTY properties on seat objects

Since we boot so fast now that gdm might get started before the
graphics drivers are properly loaded and probed we might end up
announcing seat0 to gdm before it has graphics capabilities. Which will
cause gdm/X11 cause to fail later on.

To fix this race, let's expose CanGraphical and CanTTY fields on all
seats, which clarify whether a seat is suitable for gdm resp, suitable
for text logins. gdm then needs to watch CanGraphical and spawn X11 on
it only if it is true.

This way:

USB graphics seats will expose CanGraphical=yes, CanTTY=no

Machines with no graphics drivers at all, but a text console:
CanGraphical=no, CanTTY=yes

Machines with CONFIG_VT turned off: CanGraphical=yes, CanTTY=no

And the most important case: seat0 where the graphics driver has not
been probed yet boot up with CanGraphical=no, CanTTY=yes first, which
then changes to CanGraphical=yes as soon as the probing is complete.

9 years agoupdate TODO
Lennart Poettering [Wed, 20 Jun 2012 14:03:57 +0000 (16:03 +0200)]
update TODO

9 years agoupdate TODO
Lennart Poettering [Wed, 20 Jun 2012 12:31:00 +0000 (14:31 +0200)]
update TODO

9 years agotmpfiles: exclude the first level directories in /run/user from automatic clean up
Lennart Poettering [Wed, 20 Jun 2012 07:05:50 +0000 (09:05 +0200)]
tmpfiles: exclude the first level directories in /run/user from automatic clean up

It's logind's job to maintain those user dirs, so avoid automatic clean
up for them. However, we do cover everything within them.

9 years agoanalyze: Python 3 compatibility
Shawn Landden [Thu, 14 Jun 2012 22:08:19 +0000 (15:08 -0700)]
analyze: Python 3 compatibility

Plot verb doesn't currently work b/c Cairo bug sending to stdout in python 3.

9 years agologin: initialize XDG_RUNTIME_DIR as /run/user/$UID instead of /run/user/$USER
Lennart Poettering [Tue, 19 Jun 2012 15:25:28 +0000 (17:25 +0200)]
login: initialize XDG_RUNTIME_DIR as /run/user/$UID instead of /run/user/$USER

The sssd folks would like to place the kerberos credential cache in
XDG_RUNTIME_DIR, but need to do that in the PAM auth hooks, before
pam_systemd is run as part of the PAM session setup. Hence, in order to
make this easy for them: avoid usage of usernames, and use user IDs
instead thus making an additional NSS lookup unnecessary in the kerberos
bits, but still have the directory well-defined so that the kerberos
bits can determine it before pam_systemd is run.

9 years agovconsole-setup: enable utf-8 mode explicitly
Tom Gundersen [Sun, 10 Jun 2012 18:37:14 +0000 (20:37 +0200)]
vconsole-setup: enable utf-8 mode explicitly

Rather than assuming the console is in utf-8 mode if nothing else is
specified, be a bit more robust and enable it explicitly.

This fixes a regression compared with Arch's initscripts when not
using a framebuffer as the old VGA console would not be in utf-8
mode by default.

Furthermore, this would allow vconsole-setup to be used after boot
to change the vconsole into utf-8 mode in case it has been set to
non-utf-8 mode for whatever reason. I.e, the following would leave
the console in utf-8 mode as expected:

 # export LANG=en_US.ISO-8859-1
 # /usr/lib/systemd/systemd-vconsole-setup
 # export LANG=en_US.UTF-8
 # /usr/lib/systemd/systemd-vconsole-setup

Reported-by: Xyne <>
Reported-by: Thomas Bächler <>
Cc: Dave Reisner <>
9 years agosystemd: return error when asked to stop unknown unit
Michal Sekletar [Tue, 19 Jun 2012 08:00:54 +0000 (10:00 +0200)]
systemd: return error when asked to stop unknown unit

Command systemctl stop foo.service, will print error message, when
foo.service is unknown to systemd, i.e. there is no unit file loaded for
this service.

9 years agolibudev: clarify that udev_monitor_receive_device() is non-blocking
Kay Sievers [Mon, 18 Jun 2012 13:45:01 +0000 (15:45 +0200)]
libudev: clarify that udev_monitor_receive_device() is non-blocking

Based on a patch from: Sam Thursfield <>

9 years agojournal: fix monotonic seeking
Lennart Poettering [Sat, 9 Jun 2012 13:20:42 +0000 (15:20 +0200)]
journal: fix monotonic seeking

9 years agojournal: rotate on SIGUSR2
Lennart Poettering [Sat, 9 Jun 2012 13:04:14 +0000 (15:04 +0200)]
journal: rotate on SIGUSR2

9 years agojournal: fix SD_JOURNAL_SYSTEM_ONLY flag
Lennart Poettering [Sat, 9 Jun 2012 13:03:22 +0000 (15:03 +0200)]
journal: fix SD_JOURNAL_SYSTEM_ONLY flag

9 years agojournal: expose and make use of cutoff times of journal
Lennart Poettering [Sat, 9 Jun 2012 08:32:38 +0000 (10:32 +0200)]
journal: expose and make use of cutoff times of journal

This helps explaining when the log output of "systemctl status" is
incomplete because the logs got rotated since the service was started.

9 years agojournal: correct list link up on hash collisions
Lennart Poettering [Sat, 9 Jun 2012 08:30:44 +0000 (10:30 +0200)]
journal: correct list link up on hash collisions

9 years agojournal: add basic object definition for signatures
Lennart Poettering [Tue, 5 Jun 2012 18:45:54 +0000 (20:45 +0200)]
journal: add basic object definition for signatures

9 years agojournal: replace arena offset by header size
Lennart Poettering [Tue, 5 Jun 2012 18:32:55 +0000 (20:32 +0200)]
journal: replace arena offset by header size

9 years agologind: more robust handling of VT-less systems
Michal Schmidt [Fri, 15 Jun 2012 20:22:24 +0000 (22:22 +0200)]
logind: more robust handling of VT-less systems

Restore the check that was removed in commit 74afee9c. Its removal
caused a regression on some s390x systems where for whatever reason the
device node /dev/tty0 exists and makes the preceding access() check

9 years agoservice: timeout for oneshot services
Lukas Nykryn [Thu, 14 Jun 2012 15:07:07 +0000 (17:07 +0200)]
service: timeout for oneshot services

Add possibility to specify timeout for oneshot services.

  Added minor fixups. -- michich ]

9 years agosystemctl: clearer error message for missing install information
Michal Schmidt [Thu, 14 Jun 2012 21:16:07 +0000 (23:16 +0200)]
systemctl: clearer error message for missing install information

Some users found it difficult to understand what systemctl was telling
Instead of "install information" talk about "[Install] section", which
is more likely to ring a bell. And suggest that it is intentional, so
that users do not attempt to "correct" the unit files.

9 years agotmpfiles: create char devices with correct SELinux context
Michal Schmidt [Thu, 14 Jun 2012 14:01:19 +0000 (16:01 +0200)]
tmpfiles: create char devices with correct SELinux context

9 years agounit-name: introduce unit_dbus_path_from_name()
Michal Schmidt [Wed, 13 Jun 2012 16:22:08 +0000 (18:22 +0200)]
unit-name: introduce unit_dbus_path_from_name()

Use the same function in core and in systemctl.
get_unit_path() in systemctl becomes unnecessary.

9 years agosystemctl: warn about all active triggers, not just sockets
Michal Schmidt [Wed, 13 Jun 2012 16:27:41 +0000 (18:27 +0200)]
systemctl: warn about all active triggers, not just sockets

9 years agosystemctl: fix iteration in check_listening_sockets()
Michal Schmidt [Wed, 13 Jun 2012 15:47:51 +0000 (17:47 +0200)]
systemctl: fix iteration in check_listening_sockets()

9 years agosystemctl: remove is_socket_listening
Michal Schmidt [Wed, 13 Jun 2012 15:29:11 +0000 (17:29 +0200)]
systemctl: remove is_socket_listening

We can use the functionality of check_unit(). Factor out

9 years agosystemctl: style fixes for the previous patch
Michal Schmidt [Wed, 13 Jun 2012 13:52:27 +0000 (15:52 +0200)]
systemctl: style fixes for the previous patch

Use the usual indentation, bracketing style, and no assignments in ifs.
Since check_listening_sockets provides just optional hints for the user,
don't pass its DBusErrors to the caller.

9 years agosystemctl will print warning when stopping unit
Michal Sekletar [Wed, 13 Jun 2012 12:14:13 +0000 (14:14 +0200)]
systemctl will print warning when stopping unit

systemctl now prints warning and list of sockets in listenning state which can
trigger start of service which is about to be stopped

9 years agoman: fix 'sysytemd' typos
Michal Schmidt [Tue, 12 Jun 2012 22:08:46 +0000 (00:08 +0200)]
man: fix 'sysytemd' typos

9 years agojournal-file: fix mmap leak
Michal Schmidt [Tue, 12 Jun 2012 14:45:09 +0000 (16:45 +0200)]
journal-file: fix mmap leak

9 years agologind: fix check for multiple sessions
Michal Schmidt [Tue, 12 Jun 2012 07:31:43 +0000 (09:31 +0200)]
logind: fix check for multiple sessions

The "$action-multiple-sessions" polkit actions are defined as
"$action while other users are logged in". To me this implies that the
following sessions should not count:
 - greeter sessions
 - user sessions belonging to the same user as the one who's asking

Not sure how to treat class SESSION_LOCK_SCREEN. I never have these.
I just ignore every class that's not SESSION_USER.

9 years agoudev: support "udevadm info /dev/sda; udevadm info /sys/class/block/sda"
Kay Sievers [Sun, 10 Jun 2012 20:53:07 +0000 (22:53 +0200)]
udev: support "udevadm info /dev/sda; udevadm info /sys/class/block/sda"

9 years agotmpfiles: fix error message
Kay Sievers [Sun, 10 Jun 2012 17:31:39 +0000 (19:31 +0200)]
tmpfiles: fix error message

9 years agotmpfiles: print error if basename lookup fails; document it in manpage
Kay Sievers [Sun, 10 Jun 2012 17:21:50 +0000 (19:21 +0200)]
tmpfiles: print error if basename lookup fails; document it in manpage

9 years agotmpfiles: allow to specify basename only: systemd-tmpfiles <program.conf>
Dave Reisner [Sat, 9 Jun 2012 02:31:19 +0000 (22:31 -0400)]
tmpfiles: allow to specify basename only: systemd-tmpfiles <program.conf>

Allow passing of basename only, instead of the absolute path; letting
systemd-tmpfiles perform a path lookup for the proper fragment path in
the config directories.

This allows distributions to call: systemd-tmpfiles <program.conf> on
upgrade of a package, with respecting the possibly overriden (or even
masked) tmpfile.

9 years agoman: replace tabs with spaces
Kay Sievers [Sun, 10 Jun 2012 16:32:11 +0000 (18:32 +0200)]
man: replace tabs with spaces

9 years agoupdate .gitignore
Kay Sievers [Sun, 10 Jun 2012 16:28:53 +0000 (18:28 +0200)]
update .gitignore

9 years agoman: systemd-tmpfiles - document proper config-file-stack order
Kay Sievers [Sun, 10 Jun 2012 16:26:11 +0000 (18:26 +0200)]
man: systemd-tmpfiles - document proper config-file-stack order

9 years agoarch: drop sysvcompat support
Tom Gundersen [Fri, 8 Jun 2012 23:08:37 +0000 (01:08 +0200)]
arch: drop sysvcompat support

Our downstream generator takes care of all the sysvcompat support we need.

9 years agoudev: always use $(rootprefix)/lib/udev for libexecdir
Kay Sievers [Sun, 10 Jun 2012 15:23:43 +0000 (17:23 +0200)]
udev: always use $(rootprefix)/lib/udev for libexecdir

On Sat, Jun 9, 2012 at 12:46 AM, Malte Starostik <> wrote:
> From: Malte Starostik <>
> Rules get installed in $(libexecdir)/udev/, so are keymaps.  Helper
> binaries go to $(rootprefix)/lib/udev though.  Problem is, in the code,
> both are referenced via UDEVLIBEXECDIR which is defined to the former
> location.  Result: systemd-udev can't find e.g. the keymap binary to
> apply keymaps.

9 years agosystemd-debug-shell: add to POTFILES.skip
Kay Sievers [Thu, 7 Jun 2012 17:34:13 +0000 (19:34 +0200)]
systemd-debug-shell: add to POTFILES.skip

9 years agodelete plymouth units; moved to plymouth
Kay Sievers [Thu, 7 Jun 2012 17:27:17 +0000 (19:27 +0200)]
delete plymouth units; moved to plymouth

9 years agounits: add systemd-debug-shell.service
Michal Schmidt [Thu, 7 Jun 2012 14:03:33 +0000 (16:03 +0200)]
units: add systemd-debug-shell.service

If enabled, the service provides a root shell from an early boot. It also
remains active late into shutdown. It is useful for debugging startup and
shutdown problems.

This is the unit referenced in

9 years agoudev: systemd-udev-settle.service fix After=
Kay Sievers [Wed, 6 Jun 2012 12:30:16 +0000 (14:30 +0200)]
udev: systemd-udev-settle.service fix After=

9 years agounits: avoid redundant VT clearing by agetty
Michal Schmidt [Wed, 6 Jun 2012 09:26:30 +0000 (11:26 +0200)]
units: avoid redundant VT clearing by agetty

TTYVTDisallocate=yes already clears the VT. agetty does not need to do
it again. Run it with --noclear.

Felix Miata found the double clearing confusing in this bugreport:
Add a comment explaining what clears the VT.

9 years agoreadahead: Add tool to analyze the contents of the pack file. v3.
Auke Kok [Tue, 5 Jun 2012 20:32:20 +0000 (13:32 -0700)]
readahead: Add tool to analyze the contents of the pack file. v3.

This patch adds code to compile 'systemd-readahead-analyze' and install
it into $bindir.

Use this program to parse the contents of the readahead pack file, or
an arbitrary pack file and display which files are listed in it, and
how much of the files are requested to be readahead.

This code is not new - it's partially taken from sreadahead (formerly
maintained by Arjan van der Ven and me, and was originally written
by me), and adapted with the right bits to parse the systemd
readahead pack files, which are slightly different in format.

v2 adds a common READAHEAD_PACK_FILE_VERSION used in all the code
to provide a quick way to assure all these programs are always
synchronized. v3 fixes the integer math.

9 years agoudev: remove remaining selinux labeling for file in /run
Kay Sievers [Tue, 5 Jun 2012 15:31:32 +0000 (17:31 +0200)]
udev: remove remaining selinux labeling for file in /run

9 years agoudev: kernel cmdline - accept rd.* parameters
Kay Sievers [Mon, 4 Jun 2012 20:46:32 +0000 (22:46 +0200)]
udev: kernel cmdline - accept rd.* parameters

9 years agosilence gcc warning on 32 bit
Kay Sievers [Mon, 4 Jun 2012 20:27:39 +0000 (22:27 +0200)]
silence gcc warning on 32 bit

9 years agobuild-sys: release 185 v185
Kay Sievers [Mon, 4 Jun 2012 18:15:16 +0000 (20:15 +0200)]
build-sys: release 185

9 years agosystemctl: rename "man" to "help"
Kay Sievers [Mon, 4 Jun 2012 17:48:32 +0000 (19:48 +0200)]
systemctl: rename "man" to "help"

9 years agoudev: reset /proc/self/oom_score_adj in worker processes
Kay Sievers [Mon, 4 Jun 2012 16:10:50 +0000 (18:10 +0200)]
udev: reset /proc/self/oom_score_adj in worker processes

Only the main daemon process should be excluded from OOM handling,
not the worker processes or their child processes.

9 years agoremove support for deprecated /proc/self/oom_adj
Kay Sievers [Mon, 4 Jun 2012 14:41:10 +0000 (16:41 +0200)]
remove support for deprecated /proc/self/oom_adj

9 years agoudev: limit minimum worker count to 16
Kay Sievers [Mon, 4 Jun 2012 14:40:31 +0000 (16:40 +0200)]
udev: limit minimum worker count to 16

9 years agoupdate TODO
Kay Sievers [Mon, 4 Jun 2012 14:40:08 +0000 (16:40 +0200)]
update TODO

9 years agoudev: lower the starting children_max -> drop the minimum required RAM from 23mb...
Lauri Kasanen [Sun, 3 Jun 2012 15:11:58 +0000 (18:11 +0300)]
udev: lower the starting children_max -> drop the minimum required RAM from 23mb to 5mb

Udev was the limiting factor for us on low-RAM systems.
Given an average RSS of 180kb, 128 workers would require ~23mb of RAM.

Now, please consider what happens when there is only, say, 15mb free.
Udev protects itself from OOM, and the kernel can do nothing but panic.

28 workers * 0.18mb = ~5mb. This change should not affect more powerful
systems much, given that they still get the addition from the amount of RAM.

9 years agofstab-generator: avoid mangling of mount source string
Kay Sievers [Mon, 4 Jun 2012 13:21:05 +0000 (15:21 +0200)]
fstab-generator: avoid mangling of mount source string

This is a valid source entry in /etc/fstab:
  // /data cifs noperm,auto

On Mon, Jun 4, 2012 at 2:04 PM, Dave Reisner <> wrote:
> On Mon, Jun 04, 2012 at 12:57:47PM +0200, Kay Sievers wrote:
>> Changed it to use:
>>   path_is_absolute()
>> instead of:
>>   is_path(),
>> so that we still sanitize the input we might match against.
>> Let me know, if you think that could still cause any problems?

> Yes, this will still break CIFS shares.

9 years agounit-name: never create a unit name with a leading '.'
Kay Sievers [Mon, 4 Jun 2012 12:57:24 +0000 (14:57 +0200)]
unit-name: never create a unit name with a leading '.'

Supposed to prevent creating unit files like:
  ├── dev-sda1.device.wants
  │   └── .dot.mount -> /run/systemd/generator/.dot.mount
  ├── .dot.mount
  # cat /etc/fstab
  /dev/sda1    /.dot           vfat ro           1 3

which we later skip reading because of the leading '.'.

9 years agologind: punt duplicate definition of InhibitWhat
Malte Starostik [Sun, 3 Jun 2012 19:39:21 +0000 (21:39 +0200)]
logind: punt duplicate definition of InhibitWhat

Trivial fix for:
  src/login/logind-inhibit.h:37:3: error: redefinition of typedef 'InhibitWhat'
  src/login/logind-inhibit.h:26:26: note: previous declaration of 'InhibitWhat' was here

9 years agofstab-generator: avoid mangling of non-path mount source and dest
Kay Sievers [Mon, 4 Jun 2012 10:52:14 +0000 (12:52 +0200)]
fstab-generator: avoid mangling of non-path mount source and dest

This can invalidate otherwise valid source paths with trailing slashes,
such as "host:/" in the case of a network mount.

Based on a patch from Dave Reisner <>, which
removed the slash mangling entirely.

Changed it to match on the leading slash to exclude non-path values.

9 years agokeymaps: keyboard-force-release - fix shebang
Dave Reisner [Sat, 2 Jun 2012 22:27:14 +0000 (18:27 -0400)]
keymaps: keyboard-force-release - fix shebang

This script will still run without the shebang, but we won't get the
intended effect of the errexit flag in the interpreter line.

9 years agosystemd-detect-virt: fix "option '--quiet' requires an argument"
Kay Sievers [Mon, 4 Jun 2012 10:39:22 +0000 (12:39 +0200)]
systemd-detect-virt: fix "option '--quiet' requires an argument"

9 years agokeymaps: replace dir tokens in keyboard-force-release
Dave Reisner [Sat, 2 Jun 2012 18:46:05 +0000 (14:46 -0400)]
keymaps: replace dir tokens in keyboard-force-release

Use the correct udev libexec dir as well, not systemd's.

9 years agomissing: define MS_STRICTATIME if not defined already
Lennart Poettering [Fri, 1 Jun 2012 19:25:29 +0000 (21:25 +0200)]
missing: define MS_STRICTATIME if not defined already

9 years agofix typo
Kay Sievers [Fri, 1 Jun 2012 15:56:11 +0000 (17:56 +0200)]
fix typo

9 years agounits: fix man section
Lennart Poettering [Fri, 1 Jun 2012 15:26:28 +0000 (17:26 +0200)]
units: fix man section

9 years agojournal: allow setting of a cutoff log level for disk storage, syslog, kmsg, console...
Lennart Poettering [Fri, 1 Jun 2012 15:26:16 +0000 (17:26 +0200)]
journal: allow setting of a cutoff log level for disk storage, syslog, kmsg, console forwarding

9 years agojournal: support changing the console tty to forward to
Lennart Poettering [Thu, 31 May 2012 23:06:10 +0000 (01:06 +0200)]
journal: support changing the console tty to forward to

9 years agoman: document systemd-journal
Lennart Poettering [Thu, 31 May 2012 23:05:38 +0000 (01:05 +0200)]
man: document systemd-journal

9 years agoman: write man page for systemd-logind
Lennart Poettering [Thu, 31 May 2012 22:14:07 +0000 (00:14 +0200)]
man: write man page for systemd-logind