chiark / gitweb /
elogind.git
9 years agocore: downgrade unit type not supported message
Umut Tezduyar Lindskog [Fri, 20 Feb 2015 09:53:28 +0000 (10:53 +0100)]
core: downgrade unit type not supported message

Otherwise every daemon reload prints out warnings like:

systemd[1]: Unit type .busname is not supported on this system.
systemd[1]: Unit type .swap is not supported on this system.

9 years agoshared: AFS is also a network filesystem
Cristian Rodríguez [Fri, 20 Feb 2015 18:35:11 +0000 (15:35 -0300)]
shared: AFS is also a network filesystem

9 years agomachined: use x-machine-unix prefix for the container bus on dbus1
Benjamin Franzke [Thu, 19 Feb 2015 19:47:28 +0000 (20:47 +0100)]
machined: use x-machine-unix prefix for the container bus on dbus1

This fixes "machinectl login" on systems configured with --disable-kdbus.

The error was:
machinectl login foo
Failed to get machine PTY: Input/output error

9 years agocore: fix return value on OOM
Thomas Hindoe Paaboel Andersen [Sat, 28 Feb 2015 22:39:55 +0000 (23:39 +0100)]
core: fix return value on OOM

9 years agoupdate TODO
Lennart Poettering [Sat, 28 Feb 2015 16:38:23 +0000 (17:38 +0100)]
update TODO

9 years agocore: rework device state logic
Lennart Poettering [Fri, 27 Feb 2015 20:55:08 +0000 (21:55 +0100)]
core: rework device state logic

This change introduces a new state "tentative" for device units. Device
units are considered "plugged" when udev announced them, "dead" when
they are not available in the kernel, and "tentative" when they are
referenced in /proc/self/mountinfo or /proc/swaps but not (yet)
announced via udev.

This should fix a race when device nodes (like loop devices) are created
and immediately mounted. Previously, systemd might end up seeing the
mount unit before the device, and would thus pull down the mount because
its BindTo dependency on the device would not be fulfilled.

9 years agobootctl: avoid NULL pointer dereference
Zbigniew Jędrzejewski-Szmek [Fri, 27 Feb 2015 22:31:43 +0000 (17:31 -0500)]
bootctl: avoid NULL pointer dereference

9 years agobuild-sys: use -Wno-missing-field-initializers for efi stuff
Zbigniew Jędrzejewski-Szmek [Fri, 27 Feb 2015 22:26:48 +0000 (17:26 -0500)]
build-sys: use -Wno-missing-field-initializers for efi stuff

9 years agoman: correct description of systemd-user-sessions
Zbigniew Jędrzejewski-Szmek [Fri, 27 Feb 2015 22:26:42 +0000 (17:26 -0500)]
man: correct description of systemd-user-sessions

That part of functionality was removed in 7fb3ee51c1b377.

9 years agobuild-sys: make systemd-user-sessions conditional on PAM again
Zbigniew Jędrzejewski-Szmek [Fri, 27 Feb 2015 22:26:26 +0000 (17:26 -0500)]
build-sys: make systemd-user-sessions conditional on PAM again

This undoes a part of previous commit: s-u-s is dependent on HAVE_PAM
again, but not on HAVE_LOGIND.

9 years agouser-sessions: move into own subdir and build independently of logind
Ivan Shapovalov [Wed, 25 Feb 2015 18:47:26 +0000 (21:47 +0300)]
user-sessions: move into own subdir and build independently of logind

Suggested by Zbyszek on IRC.

[zj: /run/nologin is used with PAM. systemd-user-session is independent
     of logind.]

9 years agoshared/condition: fix gcc5 warning
Daniel Mack [Fri, 27 Feb 2015 19:05:26 +0000 (20:05 +0100)]
shared/condition: fix gcc5 warning

Fixes the warning below.

src/shared/condition.c: In function ‘condition_new’:
src/shared/condition.c:47:27: warning: logical not is only applied to the left hand side of comparison [-Wlogical-not-parentheses]
         assert(!parameter == (type == CONDITION_NULL));
                           ^
src/shared/macro.h:42:44: note: in definition of macro ‘_unlikely_’
 #define _unlikely_(x) (__builtin_expect(!!(x),0))
                                            ^
src/shared/macro.h:226:22: note: in expansion of macro ‘assert_se’
 #define assert(expr) assert_se(expr)
                      ^
src/shared/condition.c:47:9: note: in expansion of macro ‘assert’
         assert(!parameter == (type == CONDITION_NULL));
         ^

9 years agonetworkd: add support for Uplink Failure Detection
Alin Rauta [Tue, 17 Feb 2015 12:06:57 +0000 (04:06 -0800)]
networkd: add support for Uplink Failure Detection

Introduce BindCarrier= to indicate the set of links that determine if
the current link should be brought UP or DOWN.

[tomegun: add a bit to commit message]

9 years agoboot: efi - fix missing include
Marcel Holtmann [Fri, 27 Feb 2015 00:19:33 +0000 (01:19 +0100)]
boot: efi - fix missing include

src/boot/efi/splash.c:312:9: warning: implicit declaration of function ‘graphics_mode’ [-Wimplicit-function-declaration]
         err = graphics_mode(TRUE);
         ^

9 years agohwdb: add sdio identifiers for Broadcom WLAN cards
Arend van Spriel [Wed, 25 Feb 2015 10:02:39 +0000 (11:02 +0100)]
hwdb: add sdio identifiers for Broadcom WLAN cards

This patch adds the sdio identifiers known to be supported by
the brcmfmac open-source driver.

9 years agosysctl: add some hints how to override settings
Zbigniew Jędrzejewski-Szmek [Fri, 27 Feb 2015 00:05:51 +0000 (19:05 -0500)]
sysctl: add some hints how to override settings

Also a link to decent documentation for sysrq keys. It is surprising
hard to find.

https://lists.fedoraproject.org/pipermail/devel/2015-February/208412.html

9 years agosysctl: downgrade message about sysctl overrides to debug
Zbigniew Jędrzejewski-Szmek [Fri, 27 Feb 2015 00:00:11 +0000 (19:00 -0500)]
sysctl: downgrade message about sysctl overrides to debug

Printing it at info level was tedious. We don't do that for any other
overrides.

9 years agosd-rtnl: types - fix check for unsupported type
Tom Gundersen [Wed, 25 Feb 2015 20:29:37 +0000 (15:29 -0500)]
sd-rtnl: types - fix check for unsupported type

Spotted by Thomas H.P. Andersen <phomes@gmail.com>

9 years agosd-rtnl: types - don't assert_return in private API
Tom Gundersen [Wed, 25 Feb 2015 20:28:48 +0000 (15:28 -0500)]
sd-rtnl: types - don't assert_return in private API

9 years agoboot: efi - remove custom splash support
Kay Sievers [Thu, 26 Feb 2015 18:51:08 +0000 (19:51 +0100)]
boot: efi - remove custom splash support

9 years agoboot: efi - split graphics and splash handling
Kay Sievers [Thu, 26 Feb 2015 18:45:08 +0000 (19:45 +0100)]
boot: efi - split graphics and splash handling

9 years agoboot: efi - support embedded splash image
Kay Sievers [Thu, 26 Feb 2015 17:33:37 +0000 (18:33 +0100)]
boot: efi - support embedded splash image

9 years agocore: emit changes for NFailedUnits property
Lucas De Marchi [Wed, 18 Feb 2015 16:22:37 +0000 (14:22 -0200)]
core: emit changes for NFailedUnits property

By notifying the clients when this property is changed it's possible to
allow "system health monitor" tools to get transitions like
running<->degraded. This is an alternative to send changes on the
SystemState property since the latter is more difficult to derive.

9 years agoUpdate TODO
Zbigniew Jędrzejewski-Szmek [Thu, 26 Feb 2015 14:27:12 +0000 (09:27 -0500)]
Update TODO

9 years agounit: When stopping due to BindsTo=, log which unit caused it
Colin Walters [Tue, 17 Feb 2015 18:47:34 +0000 (13:47 -0500)]
unit: When stopping due to BindsTo=, log which unit caused it

I'm trying to track down a relatively recent change in systemd
which broke OSTree; see https://bugzilla.gnome.org/show_bug.cgi?id=743891

Systemd started to stop sysroot.mount, and this patch should help
me debug why at least.

While we're here, "break" on the first unit we find that will
deactivate, as there's no point in further iteration.

9 years agotimedated: when performing "SetTime" compensate for program lag
Shawn Landden [Mon, 16 Feb 2015 19:47:08 +0000 (11:47 -0800)]
timedated: when performing "SetTime" compensate for program lag

(David: fix up compile-failure and simplify code a bit)

9 years agobuild-sys: efi - move cflag from x86_64 to common section
Kay Sievers [Thu, 26 Feb 2015 00:55:32 +0000 (01:55 +0100)]
build-sys: efi - move cflag from x86_64 to common section

9 years agoboot: efi - print generic error message if we fail to execute an image
Kay Sievers [Thu, 26 Feb 2015 00:15:18 +0000 (01:15 +0100)]
boot: efi - print generic error message if we fail to execute an image

9 years agoboot: efi - remove default boot splash handling
Kay Sievers [Wed, 25 Feb 2015 14:56:54 +0000 (15:56 +0100)]
boot: efi - remove default boot splash handling

9 years agoupdate TODO
Lennart Poettering [Wed, 25 Feb 2015 23:43:08 +0000 (00:43 +0100)]
update TODO

9 years agomachined: if /var/lib/machines is backed by a loop file, resize it on "machinectl...
Lennart Poettering [Wed, 25 Feb 2015 23:40:43 +0000 (00:40 +0100)]
machined: if /var/lib/machines is backed by a loop file, resize it on "machinectl set-limit"

When the pool size limit is altered with "machinectl set-limit", then
not only set the subvolume quota of the /var/lib/machine subvolume, but
also resize the backing loop file and the btrfs file system on it
dynamically.

9 years agobootctl: check that partition uuid is valid
Thomas Hindoe Paaboel Andersen [Wed, 25 Feb 2015 22:10:35 +0000 (23:10 +0100)]
bootctl: check that partition uuid is valid

It probably is but check anyway to make coverity happy.

CID#1271354

9 years agounit: use weaker dependencies between mount and device units in --user mode
Lennart Poettering [Wed, 25 Feb 2015 21:05:14 +0000 (22:05 +0100)]
unit: use weaker dependencies between mount and device units in --user mode

When running in user mode unmounting of mount units when a device
vanishes is unlikely to work, and even if it would work is already done
by PID 1 anyway. HEnce, when creating implicit dependencies between
mount units and their backing devices, created a Wants= type dependency
in --user mode, but leave a BindsTo= dependency in --system mode.

9 years agonspawn: make kill signal to use for PID 1 configurable
Lennart Poettering [Wed, 25 Feb 2015 21:04:48 +0000 (22:04 +0100)]
nspawn: make kill signal to use for PID 1 configurable

9 years agomachined,machinectl: add calls for changing container/VM quotas
Lennart Poettering [Tue, 24 Feb 2015 22:50:37 +0000 (23:50 +0100)]
machined,machinectl: add calls for changing container/VM quotas

9 years agojournal: make skipping of exhausted journal files effective again
Michal Schmidt [Tue, 24 Feb 2015 18:45:17 +0000 (19:45 +0100)]
journal: make skipping of exhausted journal files effective again

Commit 668c965af "journal: skipping of exhausted journal files is bad if
direction changed" fixed a correctness issue, but it also significantly
limited the cases where the optimization that skips exhausted journal
files could apply.
As a result, some journalctl queries are much slower in v219 than in v218.
(e.g. queries where a "--since" cutoff should have quickly eliminated
older journal files from consideration, but didn't.)

If already in the initial iteration find_location_with_matches() finds
no entry, the journal file's location is not updated. This is fine,
except that:
 - We must update at least f->last_direction. The optimization relies on
   it. Let's separate that from journal_file_save_location() and update
   it immediately after the direction checks.
 - The optimization was conditional on "f->current_offset > 0", but it
   would always be 0 in this scenario. This check is unnecessary for the
   optimization.

9 years agohwdb: Update database of Bluetooth company identifiers
Marcel Holtmann [Wed, 25 Feb 2015 06:27:49 +0000 (07:27 +0100)]
hwdb: Update database of Bluetooth company identifiers

9 years agopo: update Polish translation
Piotr Drąg [Thu, 19 Feb 2015 15:18:10 +0000 (16:18 +0100)]
po: update Polish translation

https://bugs.freedesktop.org/show_bug.cgi?id=89230

9 years agopo: update Italian translation
Daniele Medri [Thu, 19 Feb 2015 16:03:51 +0000 (17:03 +0100)]
po: update Italian translation

9 years agopo: simplify one sentence
Zbigniew Jędrzejewski-Szmek [Wed, 25 Feb 2015 04:55:35 +0000 (23:55 -0500)]
po: simplify one sentence

9 years agopo: fix typo in French translation
Zbigniew Jędrzejewski-Szmek [Wed, 25 Feb 2015 04:53:59 +0000 (23:53 -0500)]
po: fix typo in French translation

Was already fixed elsewhere in 6be9b3606872037d1d48a81fbe045af8b64cea85.

9 years agosystemd: add getrandom syscall numbers for MIPS
Aaro Koskinen [Mon, 23 Feb 2015 14:01:31 +0000 (16:01 +0200)]
systemd: add getrandom syscall numbers for MIPS

Add getrandom syscall numbers for MIPS. Based on Linux 3.17 kernel
(commit 42944521af97a3b25516f15f3149aec3779656dc, "MIPS: Wire up new
syscalls getrandom and memfd_create").

9 years agobootchart: svg: fix checking of list end
Aaro Koskinen [Tue, 24 Feb 2015 16:32:31 +0000 (18:32 +0200)]
bootchart: svg: fix checking of list end

If we have less samples than expected, systemd-bootchart will crash.

9 years agobuild-sys: fail if gnuefi files are not found
Zbigniew Jędrzejewski-Szmek [Tue, 24 Feb 2015 22:35:49 +0000 (17:35 -0500)]
build-sys: fail if gnuefi files are not found

The build would fail later anyway, so it is better to bail
out early.

Also check for the second bios file only if the first one was not
found. I'm not sure which one is preferred. If the other one, the
order should be flipped.

9 years agobootctl: fix mem leaks
Thomas Hindoe Paaboel Andersen [Tue, 24 Feb 2015 22:57:47 +0000 (23:57 +0100)]
bootctl: fix mem leaks

CID#1271347/1271348

9 years agomachined: fix check if host directory could be opened
Thomas Hindoe Paaboel Andersen [Tue, 24 Feb 2015 19:47:53 +0000 (20:47 +0100)]
machined: fix check if host directory could be opened

CID#1271351

9 years agoacl-util: avoid freeing uninitialized pointer
Thomas Hindoe Paaboel Andersen [Tue, 24 Feb 2015 19:40:07 +0000 (20:40 +0100)]
acl-util: avoid freeing uninitialized pointer

CID#1271344/1271345

9 years agounits: add missing unit file
Lennart Poettering [Tue, 24 Feb 2015 17:46:33 +0000 (18:46 +0100)]
units: add missing unit file

9 years agoimportd: enable btrfs quota in /var/lib/machines, if necessary
Lennart Poettering [Tue, 24 Feb 2015 17:43:37 +0000 (18:43 +0100)]
importd: enable btrfs quota in /var/lib/machines, if necessary

9 years agomachined/machinectl: when "machinectl image-status" is used without arguments show...
Lennart Poettering [Tue, 24 Feb 2015 17:23:40 +0000 (18:23 +0100)]
machined/machinectl: when "machinectl image-status" is used without arguments show statistics about pool

9 years agoshared: fix wrong assertion in barrier_set_role()
Cristian Rodríguez [Fri, 20 Feb 2015 18:14:56 +0000 (15:14 -0300)]
shared: fix wrong assertion in barrier_set_role()

 assert(b->pipe[0] >= 0 && b->pipe[0] >= 0);

Test the same condition twice, pretty sure we mean

 assert(b->pipe[0] >= 0 && b->pipe[1] >= 0);

9 years agobuild: make sure to test for -Wflag instead of -Wno-flag
David Herrmann [Tue, 24 Feb 2015 16:49:48 +0000 (17:49 +0100)]
build: make sure to test for -Wflag instead of -Wno-flag

gcc does not support testing for negated warnings. See here for details:
    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63499

This patch changes CC_CHECK_FLAG_APPEND to always test for the non-negated
warnings.

9 years agomachined: actually open up methods to unprivileged clients on dbus1
Lennart Poettering [Tue, 24 Feb 2015 16:26:40 +0000 (17:26 +0100)]
machined: actually open up methods to unprivileged clients on dbus1

Many of machined's operations are now opened up to unprivileged clients
via PolicyKit. Open up the dbus policy so that we can actually make
these calls.

kdbus doesn't reuqire this, hence this wasn't noticed before.

9 years agoimportd: create a loopback btrfs file system for /var/lib/machines, if necessary
Lennart Poettering [Tue, 24 Feb 2015 16:23:53 +0000 (17:23 +0100)]
importd: create a loopback btrfs file system for /var/lib/machines, if necessary

When manipulating container and VM images we need efficient and atomic
directory snapshots and file copies, as well as disk quota. btrfs
provides this, legacy file systems do not. Hence, implicitly create a
loopback file system in /var/lib/machines.raw and mount it to
/var/lib/machines, if that directory is not on btrfs anyway.

This is done implicitly and transparently the first time the user
invokes "machinectl import-xyz".

This allows us to take benefit of btrfs features for container
management without actually having the rest of the system use btrfs.

The loopback is sized 500M initially. Patches to grow it dynamically are
to follow.

9 years agoid128: add new sd_id128_is_null() call
Lennart Poettering [Mon, 23 Feb 2015 23:10:35 +0000 (00:10 +0100)]
id128: add new sd_id128_is_null() call

9 years agoimport: print nice warning if we need btrfs but /var/lib/machines is not btrfs
Lennart Poettering [Thu, 19 Feb 2015 15:19:15 +0000 (16:19 +0100)]
import: print nice warning if we need btrfs but /var/lib/machines is not btrfs

9 years agotest-hashmap: fix gcc5 warning
Daniel Mack [Tue, 24 Feb 2015 15:24:14 +0000 (16:24 +0100)]
test-hashmap: fix gcc5 warning

gcc5 spits out a warning about test-hashmap.c:

  CC       src/test/test-hashmap.o
src/test/test-hashmap.c: In function ‘test_string_compare_func’:
src/test/test-hashmap.c:76:79: warning: logical not is only applied to the left hand side of comparison [-Wlogical-not-parentheses]

9 years agoshared/unit-name: fix gcc5 warning
Daniel Mack [Tue, 24 Feb 2015 15:18:43 +0000 (16:18 +0100)]
shared/unit-name: fix gcc5 warning

Fix the following gcc5 warning:

  CC       src/shared/libsystemd_shared_la-unit-name.lo
src/shared/unit-name.c: In function 'unit_name_is_valid':
src/shared/unit-name.c:102:34: warning: logical not is only applied to the left hand side of comparison [-Wlogical-not-parentheses]
                 if (!template_ok == TEMPLATE_VALID && at+1 == e)
                                  ^

9 years agosystemctl: bump NOFILE only for systemctl_main
Zbigniew Jędrzejewski-Szmek [Tue, 24 Feb 2015 15:10:04 +0000 (10:10 -0500)]
systemctl: bump NOFILE only for systemctl_main

It is not necessary when running as telinit, etc.

https://bugzilla.redhat.com/show_bug.cgi?id=1184712

9 years agobuild: add -Wno-format-signedness
David Herrmann [Tue, 24 Feb 2015 14:59:06 +0000 (15:59 +0100)]
build: add -Wno-format-signedness

gcc5 introduced this option (gcc4 silently ignores it, which is fine).
Given that gcc5 thinks 'unsigned char'/'unsigned short' is promoted to
'int' for var-args, stuff like this spits out warnings:
    uint8_t x;
    printf("%" PRIu8", x);

gcc5 promots 'x' to 'int', instead of 'unsigned int' and thus gets a
signedness-warnings as it expects an 'unsigned int'.

glibc states otherwise: unsigneds are always promoted to 'unsigned int'.
Until gcc and glibc figure this out, lets just ignore that warning (which
is totally useless in its current form).

9 years agobootchart: fix default init path
Martin Pitt [Tue, 24 Feb 2015 13:30:10 +0000 (14:30 +0100)]
bootchart: fix default init path

Commit 6e1bf7ab99 used the wrong directory; we need rootlibexecdir, not
rootlibdir, as the latter is something like /lib/x86_64-linux-gnu/ on
multi-arch systems.

https://launchpad.net/bugs/1423867

9 years agosd-bus: test-bus-kernel-benchmark: don't rely on fixed unique names
Daniel Mack [Tue, 24 Feb 2015 13:01:36 +0000 (14:01 +0100)]
sd-bus: test-bus-kernel-benchmark: don't rely on fixed unique names

The kernel part of kdbus does not allow userspace to make any assumption on
which unique address the first user on the bus will get.

Instead, use sd_bus_get_unique_name() to get the server's address.

9 years agotest-time: test "infinity" parsing in nanoseconds
Daniel Mack [Tue, 24 Feb 2015 12:27:10 +0000 (13:27 +0100)]
test-time: test "infinity" parsing in nanoseconds

9 years agoshared/time-util: fix gcc5 warning
Daniel Mack [Tue, 24 Feb 2015 12:26:09 +0000 (13:26 +0100)]
shared/time-util: fix gcc5 warning

  CC       src/shared/libsystemd_shared_la-time-util.lo
src/shared/time-util.c: In function 'parse_nsec':
src/shared/time-util.c:789:25: warning: logical not is only applied to the left hand side of comparison [-Wlogical-not-parentheses]
                 if (!*s != 0)
                         ^

9 years agosd-bus: sync kdbus.h
Daniel Mack [Tue, 24 Feb 2015 11:10:13 +0000 (12:10 +0100)]
sd-bus: sync kdbus.h

Follow two small changes in the kdbus API:

 * Flags are now returned in cmd->return_flags by KDBUS_CMD_NAME_ACQUIRE

 * struct kdbus_item_list has been dropped. The information stored in
   this struct was redundant since awhile already, as all commands
   report their returned slice size anyway.

9 years agosystemctl: support auditd.service better
Zbigniew Jędrzejewski-Szmek [Sat, 7 Feb 2015 16:35:37 +0000 (11:35 -0500)]
systemctl: support auditd.service better

We would print the filename header before trying to open the file. But since
the header was printed to stdout, and the error to stderr, the error would appear
on the terminal before the header. It is cleaner to open the file first, then
and only then print the header.

Also exit on first error. We shouldn't report success if we were unable to open
a file.

9 years agosystemctl: check validity of PID we received
Zbigniew Jędrzejewski-Szmek [Sat, 7 Feb 2015 16:16:04 +0000 (11:16 -0500)]
systemctl: check validity of PID we received

9 years agoresolved: use == for comparing unsigned against zero
Zbigniew Jędrzejewski-Szmek [Fri, 6 Feb 2015 16:30:19 +0000 (11:30 -0500)]
resolved: use == for comparing unsigned against zero

9 years agotmpfiles: avoid creating duplicate acl entries
Zbigniew Jędrzejewski-Szmek [Tue, 24 Feb 2015 04:19:54 +0000 (23:19 -0500)]
tmpfiles: avoid creating duplicate acl entries

https://bugs.freedesktop.org/show_bug.cgi?id=89202
https://bugs.debian.org/778656

Status quo ante can be restored with:
  getfacl -p /var/log/journal/`cat /etc/machine-id`|grep -v '^#'|sort -u|sudo setfacl --set-file=- /var/log/journal/`cat /etc/machine-id`

9 years agolibsystemd: revert one hunk of duplicate ifdef removal
Zbigniew Jędrzejewski-Szmek [Tue, 24 Feb 2015 01:03:55 +0000 (20:03 -0500)]
libsystemd: revert one hunk of duplicate ifdef removal

This change exposed a bug in kernel headers:

/usr/include/linux/if_bridge.h:184:20: error: field 'ip6' has incomplete type
    struct in6_addr ip6;
                    ^
/usr/include/linux/if_tunnel.h:76:19: error: field 'prefix' has incomplete type
  struct in6_addr  prefix;
                   ^

9 years agoman: explain time units in tmpfiles
Zbigniew Jędrzejewski-Szmek [Mon, 23 Feb 2015 23:59:17 +0000 (18:59 -0500)]
man: explain time units in tmpfiles

https://bugzilla.redhat.com/show_bug.cgi?id=1195294

9 years agosd-bus, shared: fix includes
Daniel Mack [Mon, 23 Feb 2015 23:42:13 +0000 (00:42 +0100)]
sd-bus, shared: fix includes

Include <sys/socket.h> from util.h and bus-message.h in order to
build errors like the ones below on kdbus enabled systems.

./src/shared/util.h:976:32: warning: its scope is only this definition or declaration, which is probably not what you want
In file included from src/libsystemd/sd-bus/bus-kernel.c:37:0:
./src/shared/util.h:1081:28: warning: 'struct msghdr' declared inside parameter list
 void cmsg_close_all(struct msghdr *mh);
                            ^
  CC       src/libsystemd/sd-bus/libsystemd_la-bus-creds.lo
In file included from src/libsystemd/sd-bus/bus-creds.c:25:0:
./src/shared/util.h:976:32: warning: 'struct ucred' declared inside parameter list
 int getpeercred(int fd, struct ucred *ucred);
                                ^

9 years agoremove unused includes
Thomas Hindoe Paaboel Andersen [Sat, 14 Feb 2015 08:44:31 +0000 (09:44 +0100)]
remove unused includes

This patch removes includes that are not used. The removals were found with
include-what-you-use which checks if any of the symbols from a header is
in use.

9 years agonspawn: fix whitespace and typo in partition table blurb
Jan Synacek [Mon, 23 Feb 2015 14:22:40 +0000 (15:22 +0100)]
nspawn: fix whitespace and typo in partition table blurb

9 years agopo: update French translation
Sylvain Plantefève [Thu, 19 Feb 2015 22:31:51 +0000 (23:31 +0100)]
po: update French translation

9 years agobuild-sys: bootctl depends on HAVE_BLKID
Marcel Holtmann [Sun, 22 Feb 2015 09:28:34 +0000 (10:28 +0100)]
build-sys: bootctl depends on HAVE_BLKID

9 years agologind: fix a typo in a polkit description
Piotr Drąg [Thu, 19 Feb 2015 15:12:38 +0000 (16:12 +0100)]
logind: fix a typo in a polkit description

9 years agoREADME: add pkg-config to required deps for autogen
Peter Waller [Mon, 22 Dec 2014 12:33:19 +0000 (12:33 +0000)]
README: add pkg-config to required deps for autogen

Otherwise, several macros are undefined.

9 years agoman: make bootup graph consistent
Chris Morin [Wed, 4 Feb 2015 19:54:34 +0000 (14:54 -0500)]
man: make bootup graph consistent

9 years agobuild-sys: update gitignore
Kay Sievers [Fri, 20 Feb 2015 14:42:32 +0000 (15:42 +0100)]
build-sys: update gitignore

9 years agoshared: handle unnamed sockets in socket_address_equal()
Michal Schmidt [Fri, 20 Feb 2015 01:25:16 +0000 (02:25 +0100)]
shared: handle unnamed sockets in socket_address_equal()

Make sure we don't inspect sun_path of unnamed sockets.
Since we cannot know if two unnamed sockets' adresses refer to the same
socket, just return false.

9 years agoshared: avoid semi-duplicating socket_address_equal()
Michal Schmidt [Fri, 20 Feb 2015 01:13:03 +0000 (02:13 +0100)]
shared: avoid semi-duplicating socket_address_equal()

Just call socket_address_equal() from socket_address_matches_fd()
instead of implementing similar comparing of addresses.

9 years agoshared: use SocketAddress in socket_address_matches_fd()
Michal Schmidt [Fri, 20 Feb 2015 01:04:05 +0000 (02:04 +0100)]
shared: use SocketAddress in socket_address_matches_fd()

Cleanup. No behavior change.

9 years agocore, shared: in deserializing, match same files reached via different paths
Michal Schmidt [Thu, 19 Feb 2015 22:12:38 +0000 (23:12 +0100)]
core, shared: in deserializing, match same files reached via different paths

When dbus.socket is updated like this:
-ListenStream=/var/run/dbus/system_bus_socket
+ListenStream=/run/dbus/system_bus_socket
... and daemon-reload is performed, bad things happen.
During deserialization systemd does not recognize that the two paths
refer to the same named socket and replaces the socket file with a new
one. As a result, applications hang when they try talking to dbus.

Fix this by finding a match not only when the path names are equal, but
also when they point to the same inode.
In socket_address_equal() it is necessary to move the address size
comparison into the abstract sockets branch. For path name sockets the
comparison must not be done and for other families it is redundant
(their sizes are constant and checked by socket_address_verify()).

FIFOs and special files can also have multiple pathnames, so compare the
inodes for them as well. Note that previously the pathname checks used
streq_ptr(), but the paths cannot be NULL.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1186018
9 years agobus: sync with kdbus.git (ABI break)
David Herrmann [Thu, 19 Feb 2015 19:02:40 +0000 (20:02 +0100)]
bus: sync with kdbus.git (ABI break)

This syncs up the new KDBUS_CMD_CONN_INFO behavior:
 - attach-flags are passed in .attach_flags, instead of directly merged
   with the command flags.

9 years agoman: add newlines to the pull-raw example in machinectl(1)
Benjamin Franzke [Thu, 19 Feb 2015 12:10:18 +0000 (13:10 +0100)]
man: add newlines to the pull-raw example in machinectl(1)

They were removed in commit 798d3a52 ("Reindent man pages to 2ch").

9 years agonspawn: chown basic device nodes to userns root
Lennart Poettering [Thu, 19 Feb 2015 11:03:39 +0000 (12:03 +0100)]
nspawn: chown basic device nodes to userns root

9 years agonspawn: fix build on non-selinux systems
Lennart Poettering [Thu, 19 Feb 2015 11:03:12 +0000 (12:03 +0100)]
nspawn: fix build on non-selinux systems

9 years agonspawn: add basic user namespacing support
Lennart Poettering [Thu, 19 Feb 2015 10:30:18 +0000 (11:30 +0100)]
nspawn: add basic user namespacing support

(This is incomplete, /proc and /sys are still owned by root from outside
the container, not inside)

9 years agosysv-generator: fix wrong "Overwriting existing symlink" warnings
Martin Pitt [Thu, 19 Feb 2015 10:06:24 +0000 (11:06 +0100)]
sysv-generator: fix wrong "Overwriting existing symlink" warnings

Fix result testing of is_symlink() to ignore negative results, which happen if
the file name does not exist at all. In this case we do not want a warning and
unlink the non-existing link.

https://bugs.debian.org/778700

9 years agosystemd-sysv-generator test: Adjust to dropped runlevelN.target mapping
Martin Pitt [Thu, 19 Feb 2015 08:09:57 +0000 (09:09 +0100)]
systemd-sysv-generator test: Adjust to dropped runlevelN.target mapping

Commit d5d8429a dropped the explicit runlevelN.target mapping. Adjust the tests
accordingly to explicitly state the expected targets instead of runlevels.

9 years agohwdb: fix ThinkPad X* Tablet special keys
Lubomir Rintel [Wed, 18 Feb 2015 20:02:01 +0000 (21:02 +0100)]
hwdb: fix ThinkPad X* Tablet special keys

ThinkPad tablet firmware has DMI product name and version reversed:

Handle 0x0001, DMI type 1, 27 bytes
System Information
        Manufacturer: LENOVO
        Product Name: 7762AS1
        Version: ThinkPad X61 Tablet
        Serial Number: LKZCDH2
        UUID: 6ADBC681-4FC9-11CB-844F-B47CB9210BE2
        Wake-up Type: Power Switch
        SKU Number: Not Specified
        Family: ThinkPad X61 Tablet

9 years agopo: update Russian translation
Sergey Ptashnick [Thu, 19 Feb 2015 00:32:42 +0000 (03:32 +0300)]
po: update Russian translation

Add strings for fsckd.

9 years agoupdate TODO
Lennart Poettering [Wed, 18 Feb 2015 22:35:51 +0000 (23:35 +0100)]
update TODO

9 years agonspawn: when connected to pipes for stdin/stdout, pass them as-is to PID 1
Lennart Poettering [Wed, 18 Feb 2015 22:32:55 +0000 (23:32 +0100)]
nspawn: when connected to pipes for stdin/stdout, pass them as-is to PID 1

Previously we always invoked the container PID 1 on /dev/console of the
container. With this change we do so only if nspawn was invoked
interactively (i.e. its stdin/stdout was connected to a TTY). In all other
cases we directly pass through the fds unmodified.

This has the benefit that nspawn can be added into shell pipelines.

https://bugs.freedesktop.org/show_bug.cgi?id=87732

9 years agosystemd-boot: fix alignment of version in table
Thomas Hindoe Paaboel Andersen [Wed, 18 Feb 2015 21:18:12 +0000 (22:18 +0100)]
systemd-boot: fix alignment of version in table

9 years agoeverywhere: remove configurability of sysv runlevel to target mapping
Lennart Poettering [Wed, 18 Feb 2015 19:17:49 +0000 (20:17 +0100)]
everywhere: remove configurability of sysv runlevel to target mapping

With this change runlevel 2, 3, 4 are mapped to multi-user.target for
good, and 5 to graphical.target. This was already the previous mapping
but is now no longer reconfigurable, but hard-coded into the core.

This should generally simplify things, but also fix one bug: the
sysv-generator previously generated symlinks to runlevel[2-5].target
units, which possibly weren't picked up if these aliases were otherwise
only referenced by the real names "multi-user.target" and
"graphical.target".

We keep compat aliases "runlevel[2345].target" arround for cases where
this target name is explicitly requested.

9 years agoupdate TODO
Lennart Poettering [Wed, 18 Feb 2015 18:39:54 +0000 (19:39 +0100)]
update TODO

9 years agorun: if we fail to set a property assignment then really fail
Lennart Poettering [Wed, 18 Feb 2015 18:39:40 +0000 (19:39 +0100)]
run: if we fail to set a property assignment then really fail

9 years agonspawn: add support for --property= to set scope properties
Lennart Poettering [Wed, 18 Feb 2015 18:38:55 +0000 (19:38 +0100)]
nspawn: add support for --property= to set scope properties

This is similar to systemd-run's --property= setting.