platform/upstream/systemd.git
9 years agocore: various small fixes to unit-printf
Evgeny Vereshchagin [Sat, 24 Oct 2015 20:55:56 +0000 (20:55 +0000)]
core: various small fixes to unit-printf

* check memory allocation errors in specifier_cgroup_slice
* %I doesn't fail for non-instantiated units (%i doesn't fail too)
* EOPNOTSUPP for consistency

9 years agoMerge pull request #1664 from again4you/devel/tmp-smack_#5
Lennart Poettering [Sat, 24 Oct 2015 12:58:51 +0000 (14:58 +0200)]
Merge pull request #1664 from again4you/devel/tmp-smack_#5

mount: add new SmackFileSystemRoot= setting for mount unit (v4)

9 years agounits: add 'SmackFileSystemRoot=*' option into tmp.mount
Sangjung Woo [Wed, 21 Oct 2015 12:48:13 +0000 (21:48 +0900)]
units: add 'SmackFileSystemRoot=*' option into tmp.mount

If SMACK is enabled, 'smackfsroot=*' option should be specified when
/tmp is mounted since many non-root processes use /tmp for temporary
usage. If not, /tmp is labeled as '_' and smack denial occurs when
writing.

In order to do that, 'SmackFileSystemRoot=*' is newly added into
tmp.mount.

9 years agomount: add new SmackFileSystemRoot= setting for mount unit
Sangjung Woo [Wed, 21 Oct 2015 08:42:34 +0000 (17:42 +0900)]
mount: add new SmackFileSystemRoot= setting for mount unit

This option specifies the label to assign the root of the file system if
it lacks the Smack extended attribute. Note that this option will be
ignored if kernel does not support the Smack feature by runtime
checking.

9 years agoMerge pull request #1657 from phomes/tmpfiles-simplification
Lennart Poettering [Sat, 24 Oct 2015 10:17:41 +0000 (12:17 +0200)]
Merge pull request #1657 from phomes/tmpfiles-simplification

tmpfiles: simplify mountpoint detection

9 years agoMerge pull request #1659 from vcaputo/journal_verify_envalid
Lennart Poettering [Sat, 24 Oct 2015 09:52:52 +0000 (11:52 +0200)]
Merge pull request #1659 from vcaputo/journal_verify_envalid

journal: s/Envalid/Invalid/

9 years agojournal: s/Envalid/Invalid/
Vito Caputo [Fri, 23 Oct 2015 23:12:31 +0000 (16:12 -0700)]
journal: s/Envalid/Invalid/

9 years agotmpfiles: simplify mountpoint detection
Thomas Hindoe Paaboel Andersen [Fri, 23 Oct 2015 18:15:17 +0000 (20:15 +0200)]
tmpfiles: simplify mountpoint detection

No functional change. Just a simplification. A || (!A && B) is
the same as A || B

Introduced in 78a92a5a2306709e4587e332728a76901323ade9

9 years agoMerge pull request #1653 from keszybz/lz4-compress-time
Lennart Poettering [Fri, 23 Oct 2015 15:42:24 +0000 (17:42 +0200)]
Merge pull request #1653 from keszybz/lz4-compress-time

Limit test-compress-benchmark to approx. 12 s of runtime

9 years agotest-compress-benchmark: properly initialize buffer
Zbigniew Jędrzejewski-Szmek [Sun, 18 Oct 2015 21:59:40 +0000 (17:59 -0400)]
test-compress-benchmark: properly initialize buffer

We were compressing unitialized memory, which should not result in
any problems, but is inelegant.

9 years agotest-compress-benchmark: limit default runtime to 2 seconds per subtest
Zbigniew Jędrzejewski-Szmek [Sun, 18 Oct 2015 21:10:51 +0000 (17:10 -0400)]
test-compress-benchmark: limit default runtime to 2 seconds per subtest

If both lz4 and xz are enabled, this results in a limit of
2×3×2 s ~= 12 s runtime.

Previous implementation started with really small buffer sizes. When
combined with a short time limit this resulteded in abysmal results for xz.
It seems that the initialization overead is really significant for small
buffers. Since xz will not be used by default anymore, this does not
seem worth fixing. Instead buffer sizes are changed to run a
pseudo-random non-repeating pattern. This should allow reasonable testing
for all buffer sizes. For testing, both runtime and the buffer size seed
can be specified on the command line. Sufficiently large runtime allows
all buffer sizes up to 1MB to be tested.

9 years agoMerge pull request #1652 from filbranden/extract1
Lennart Poettering [Fri, 23 Oct 2015 11:46:03 +0000 (13:46 +0200)]
Merge pull request #1652 from filbranden/extract1

util: Replace state with separate booleans in extract_first_word

9 years agoMerge pull request #1647 from evverx/use-extract-in-namespace-parsing
Lennart Poettering [Fri, 23 Oct 2015 11:18:55 +0000 (13:18 +0200)]
Merge pull request #1647 from evverx/use-extract-in-namespace-parsing

core: use extract_first_word for namespace parsing

9 years agoMerge pull request #1648 from evverx/clarify-journalctl-quiet-option
Lennart Poettering [Fri, 23 Oct 2015 10:47:07 +0000 (12:47 +0200)]
Merge pull request #1648 from evverx/clarify-journalctl-quiet-option

journalctl: clarify -q option

9 years agoMerge pull request #1641 from poettering/btrfs-quota
Daniel Mack [Fri, 23 Oct 2015 08:50:53 +0000 (10:50 +0200)]
Merge pull request #1641 from poettering/btrfs-quota

btrfs quota beef up and various other unrelated changes

9 years agoutil: Replace state with separate booleans in extract_first_word
Filipe Brandenburger [Fri, 23 Oct 2015 06:12:31 +0000 (23:12 -0700)]
util: Replace state with separate booleans in extract_first_word

This simplifies the logic and uniformizes the way single and double
quotes are handled. In the end, the code is about 40 lines shorter.

Tested by running the excellent test cases from test-util. Also
installed the systemd binaries including this patch and booted a
system with it, everything looked normal.

9 years agojournalctl: clarify -q option
Evgeny Vereshchagin [Thu, 22 Oct 2015 23:18:17 +0000 (23:18 +0000)]
journalctl: clarify -q option

-q suppresses info messages too

9 years agocore: use extract_first_word for namespace parsing
Evgeny Vereshchagin [Thu, 22 Oct 2015 20:28:28 +0000 (20:28 +0000)]
core: use extract_first_word for namespace parsing

see https://github.com/systemd/systemd/pull/1632#issuecomment-149903791
We should port this loop over to extract_first_word(), too.

9 years agoMerge pull request #1646 from keszybz/man-dnf-install
Lennart Poettering [Thu, 22 Oct 2015 18:28:50 +0000 (20:28 +0200)]
Merge pull request #1646 from keszybz/man-dnf-install

dnf install command update

9 years agoman: also add --enablerepo=updates to dnf invocation
Zbigniew Jędrzejewski-Szmek [Thu, 22 Oct 2015 18:11:22 +0000 (14:11 -0400)]
man: also add --enablerepo=updates to dnf invocation

Without the updates repo, we are installing packages from the time
that that version of Fedora was released. Normally, during the
lifetime of the release most packages are updated, so most of the
packages installed would be outdated, and the first update after
installation would update a massive set of packages. Avoid all this
by installing from the updates repo from the start.

9 years agoMerge pull request #1625 from teg/ndisc
Patrik Flykt [Thu, 22 Oct 2015 17:41:50 +0000 (20:41 +0300)]
Merge pull request #1625 from teg/ndisc

sd-ndisc: Rename (from sd-icmp6-nd) and prepare for handling SLAAC and router discovery

9 years agoman: remove --nogpg from dnf install command line
Zbigniew Jędrzejewski-Szmek [Thu, 22 Oct 2015 17:34:46 +0000 (13:34 -0400)]
man: remove --nogpg from dnf install command line

Keys for previous and future Fedora distributions were added
for the fedora-repos package recently:
https://bugzilla.redhat.com/show_bug.cgi?id=1246701.
There is no need to skip signature checking.

Also, update to the latest and greatest and remove unnecessary quotes.

9 years agosd-ndisc: reduce callbacks
Tom Gundersen [Thu, 22 Oct 2015 15:34:58 +0000 (17:34 +0200)]
sd-ndisc: reduce callbacks

Only callback on error when the statemachine is in a truly broken state. This
is now only the case when we fail to rearm a timer.

9 years agosd-ndisc: drop sd_ndisc_get_prefixlen()
Tom Gundersen [Wed, 21 Oct 2015 00:14:24 +0000 (02:14 +0200)]
sd-ndisc: drop sd_ndisc_get_prefixlen()

This is unused, and in the future we will pass prefixes and prefixlengths directly
to the callbacks when needed rather than having to search for them.

9 years agosd-ndisc: drop sd_ndisc_prefix_match()
Tom Gundersen [Mon, 19 Oct 2015 13:04:22 +0000 (15:04 +0200)]
sd-ndisc: drop sd_ndisc_prefix_match()

This is no longer used.

9 years agosd-ndisc: don't trigger timeout on prefix expiry
Tom Gundersen [Sun, 18 Oct 2015 15:45:22 +0000 (17:45 +0200)]
sd-ndisc: don't trigger timeout on prefix expiry

The caller of the library is no longer notified, so triggering a timer
just to clean up is not necessary. Instead check for and clean up
invalid prefixes lazily.

9 years agosd-ndisc: remove a prefix from the ndisc object when freeing it
Tom Gundersen [Sun, 18 Oct 2015 15:27:10 +0000 (17:27 +0200)]
sd-ndisc: remove a prefix from the ndisc object when freeing it

This follows the coding style, and allows some simplification to the rest of the code.

9 years agosd-ndisc: don't inform the caller of expired prefixes
Tom Gundersen [Sun, 18 Oct 2015 15:19:29 +0000 (17:19 +0200)]
sd-ndisc: don't inform the caller of expired prefixes

The caller should push any lifetime information into the kernel and let the kernel handle
prefix expiration.

9 years agosd-ndisc: clean up state enum
Tom Gundersen [Sun, 18 Oct 2015 15:13:03 +0000 (17:13 +0200)]
sd-ndisc: clean up state enum

There is no need to assign valuse to the states. Also add _INVALID and _MAX,
even though these are not used, it keeps it consistent.

9 years agonetworkd: dhcp6 - do not handle prefix expiration
Tom Gundersen [Sun, 18 Oct 2015 15:06:31 +0000 (17:06 +0200)]
networkd: dhcp6 - do not handle prefix expiration

This ressurects 47d45d3cde45d6545367570264e4e3636bc9e345. We now always use /128 prefixes,
so there is no need for the DHCPv6 code to know about prefixes expiring.

9 years agonetworkd: dhcp6 - DHCPv6 addresses should always be /128
Tom Gundersen [Sun, 18 Oct 2015 14:59:21 +0000 (16:59 +0200)]
networkd: dhcp6 - DHCPv6 addresses should always be /128

The routing information should be configured separately by ND, there is no need to
indicate the prefix again in the DHCPv6 addresses.

See discussion and related links at issue #1520.

9 years agonetworkd: ndisc - split out from dhcp6 code
Tom Gundersen [Fri, 16 Oct 2015 19:09:15 +0000 (21:09 +0200)]
networkd: ndisc - split out from dhcp6 code

9 years agonetworkd: ndisc - make logging messages uniform
Tom Gundersen [Fri, 16 Oct 2015 17:28:27 +0000 (19:28 +0200)]
networkd: ndisc - make logging messages uniform

Refer to Router Discovery rather than ICMPv6.

9 years agosd-ndisc: rename icmp6 to ndisc throughout the code
Tom Gundersen [Fri, 16 Oct 2015 17:17:50 +0000 (19:17 +0200)]
sd-ndisc: rename icmp6 to ndisc throughout the code

9 years agolibsystemd-network: split icm6-util.[ch] out of dhcp6 code
Tom Gundersen [Fri, 16 Oct 2015 16:09:10 +0000 (18:09 +0200)]
libsystemd-network: split icm6-util.[ch] out of dhcp6 code

9 years agonetworkd: rename icmp6 to ndisc
Tom Gundersen [Fri, 16 Oct 2015 15:34:58 +0000 (17:34 +0200)]
networkd: rename icmp6 to ndisc

9 years agosd-ndisc: rename API from sd-icmp6-nd
Tom Gundersen [Fri, 16 Oct 2015 15:07:47 +0000 (17:07 +0200)]
sd-ndisc: rename API from sd-icmp6-nd

9 years agosd-icmp6-nd: rename files to sd-ndisc
Tom Gundersen [Thu, 15 Oct 2015 17:25:31 +0000 (19:25 +0200)]
sd-icmp6-nd: rename files to sd-ndisc

The actual code rename will follow. The reason for the change of name is to make it
simpler and more uniform with how we name other libraries (we don't include the
underlying protocol). The new name also matches the naming in the kernel (which
is particularly relevent here as we expect to let the kernel do some parts of
the protocol and we do others).

9 years agoMerge pull request #1644 from reverendhomer/patch-1
Tom Gundersen [Thu, 22 Oct 2015 15:02:54 +0000 (17:02 +0200)]
Merge pull request #1644 from reverendhomer/patch-1

networkd-manager: fix swapped arguments

9 years agoMerge pull request #1642 from evverx/allow-unbalanced-double-quote-in-relax-mode
Lennart Poettering [Thu, 22 Oct 2015 10:20:12 +0000 (12:20 +0200)]
Merge pull request #1642 from evverx/allow-unbalanced-double-quote-in-relax-mode

util: allow unbalanced double quote in EXTRACT_QUOTES|EXTRACT_RELAX mode

9 years agonetworkd-manager: fix swapped arguments
reverendhomer [Thu, 22 Oct 2015 07:36:07 +0000 (10:36 +0300)]
networkd-manager: fix swapped arguments

fixes Coverity #1328493

9 years agoutil: allow unbalanced double quote in EXTRACT_QUOTES|EXTRACT_RELAX mode
Evgeny Vereshchagin [Thu, 22 Oct 2015 00:37:32 +0000 (00:37 +0000)]
util: allow unbalanced double quote in EXTRACT_QUOTES|EXTRACT_RELAX mode

extract_first_word understands "\'string" but doesn't understand "\"string"
fixed this inconsistency.

9 years agocore: check parsed bus msg in full before applying it
Lennart Poettering [Thu, 22 Oct 2015 00:02:17 +0000 (02:02 +0200)]
core: check parsed bus msg in full before applying it

9 years agonspawn: don't try to resolve passed binary before entering namespace
Lennart Poettering [Wed, 21 Oct 2015 23:33:06 +0000 (01:33 +0200)]
nspawn: don't try to resolve passed binary before entering namespace

Othewise we might follow the symlinks on the host, instead of the
container.

Fixes #1400

9 years agonspawn: rework how we determine private networking settings
Lennart Poettering [Wed, 21 Oct 2015 22:59:18 +0000 (00:59 +0200)]
nspawn: rework how we determine private networking settings

Make sure we acquire CAP_NET_ADMIN if we require virtual networking.

Make sure we imply virtual ethernet correctly when bridge is request.

Fixes: #1511
Fixes: #1554
Fixes: #1590

9 years agounits: also whitelist "blkext" block devices for nspawn service
Lennart Poettering [Wed, 21 Oct 2015 22:39:57 +0000 (00:39 +0200)]
units: also whitelist "blkext" block devices for nspawn service

/dev/loop*p* block devices are of the "blkext" subsystem, not of loop,
hence whitelist this too.

Fixes #1446

9 years agoimport: don't claim we moved .nspawn file into place when in fact we did not
Lennart Poettering [Wed, 21 Oct 2015 22:39:27 +0000 (00:39 +0200)]
import: don't claim we moved .nspawn file into place when in fact we did not

9 years agoimport: correct handling if .nspawn file could not be downloaded
Lennart Poettering [Wed, 21 Oct 2015 22:39:09 +0000 (00:39 +0200)]
import: correct handling if .nspawn file could not be downloaded

9 years agomachinectl: accept "none" and "infinity" as specifier when dropping quotas using...
Lennart Poettering [Wed, 21 Oct 2015 20:41:31 +0000 (22:41 +0200)]
machinectl: accept "none" and "infinity" as specifier when dropping quotas using "machinectl set-limit"

Previously, we already accepted "-" as special value for dropping
limits. Add "infinity", as that's what we support for RLIMITs and hence
should support here to. Also add "none" as that's what the btrfs tools
use.

9 years agoupdate TODO
Lennart Poettering [Wed, 21 Oct 2015 17:48:07 +0000 (19:48 +0200)]
update TODO

9 years agotmpfiles.d: change all subvolumes to use quota
Lennart Poettering [Wed, 21 Oct 2015 17:47:28 +0000 (19:47 +0200)]
tmpfiles.d: change all subvolumes to use quota

Let's make sure the subvolumes we create fit into a sensible definition
of a quota tree.

9 years agotmpfiles: introduce "q" and "Q" for creating quota-enabled btrfs subvolumes
Lennart Poettering [Wed, 21 Oct 2015 17:46:23 +0000 (19:46 +0200)]
tmpfiles: introduce "q" and "Q" for creating quota-enabled btrfs subvolumes

This allows us to set up the quota group hierarchy in a reasonable way
on btrfs file systems.

9 years agoimport: when downloading images, create a subtree quota group for them
Lennart Poettering [Wed, 21 Oct 2015 17:45:44 +0000 (19:45 +0200)]
import: when downloading images, create a subtree quota group for them

9 years agobtrfs: beef-up btrfs support with a limited understanding of quota
Lennart Poettering [Wed, 21 Oct 2015 17:38:21 +0000 (19:38 +0200)]
btrfs: beef-up btrfs support with a limited understanding of quota

With this change we understand more than just leaf quota groups for
btrfs file systems. Specifically:

- When we create a subvolume we can now optionally add the new subvolume
  to all qgroups its parent subvolume was member of too. Alternatively
  it is also possible to insert an intermediary quota group between the
  parent's qgroups and the subvolume's leaf qgroup, which is useful for
  a concept of "subtree" qgroups, that contain a subvolume and all its
  children.

- The remove logic for subvolumes has been updated to optionally remove
  any leaf qgroups or "subtree" qgroups, following the logic above.

- The snapshot logic for subvolumes has been updated to replicate the
  original qgroup setup of the source, if it follows the "subtree"
  design described above. It will not cover qgroup setups that introduce
  arbitrary qgroups, especially those orthogonal to the subvolume
  hierarchy.

This also tries to be more graceful when setting up /var/lib/machines as
btrfs. For example, if mkfs.btrfs is missing we don't even try to set it
up as loopback device.

Fixes #1559
Fixes #1129

9 years agologind: minor clean-ups
Lennart Poettering [Wed, 21 Oct 2015 17:36:22 +0000 (19:36 +0200)]
logind: minor clean-ups

9 years agosystemctl: the various list commands actually can take any number of arguments
Lennart Poettering [Wed, 21 Oct 2015 17:34:52 +0000 (19:34 +0200)]
systemctl: the various list commands actually can take any number of arguments

I accidentally broke this a while back when I ported systemctl to the
verbs logic.

Add support for this back.

9 years agoutil: improve dir_is_empty() call
Lennart Poettering [Mon, 19 Oct 2015 21:58:17 +0000 (23:58 +0200)]
util: improve dir_is_empty() call

Simplify the call, and add dir_is_populated() as inverse call, in order
to make some checks easier to read.

9 years agoMerge pull request #1639 from alkino/master
Lennart Poettering [Wed, 21 Oct 2015 19:01:21 +0000 (21:01 +0200)]
Merge pull request #1639 from alkino/master

core dbus: Check that flush works with memstream

9 years agocore dbus: Check that flush works with memstream
Nicolas Cornu [Wed, 21 Oct 2015 16:17:12 +0000 (18:17 +0200)]
core dbus: Check that flush works with memstream

9 years agoMerge pull request #1637 from alkino/master
Lennart Poettering [Wed, 21 Oct 2015 14:38:35 +0000 (16:38 +0200)]
Merge pull request #1637 from alkino/master

shell-completion: systemd-run: add new property EnvironmentFile

9 years agoshell-completion: systemd-run: add new property EnvironmentFile
Nicolas Cornu [Wed, 21 Oct 2015 14:18:59 +0000 (16:18 +0200)]
shell-completion: systemd-run: add new property EnvironmentFile

9 years agoMerge pull request #1632 from evverx/fix-namespace-parsing
Lennart Poettering [Wed, 21 Oct 2015 13:56:48 +0000 (15:56 +0200)]
Merge pull request #1632 from evverx/fix-namespace-parsing

core: fix namespace parsing

9 years agoMerge pull request #1636 from alkino/master
Lennart Poettering [Wed, 21 Oct 2015 13:55:31 +0000 (15:55 +0200)]
Merge pull request #1636 from alkino/master

Add zsh completion on systemd-run's properties

9 years agoAdd zsh completion on systemd-run's properties
Nicolas Cornu [Wed, 21 Oct 2015 13:51:09 +0000 (15:51 +0200)]
Add zsh completion on systemd-run's properties

9 years agocore: fix namespace parsing
Evgeny Vereshchagin [Wed, 21 Oct 2015 11:49:36 +0000 (11:49 +0000)]
core: fix namespace parsing

ReadOnlyDirectories=-/ works fine

9 years agoMerge pull request #1623 from evverx/run-rw-ro-ia-dirs
Lennart Poettering [Wed, 21 Oct 2015 10:08:44 +0000 (12:08 +0200)]
Merge pull request #1623 from evverx/run-rw-ro-ia-dirs

systemd-run can launch units with ReadWriteDirectories, ReadOnlyDirectories, InaccessibleDirectories

9 years agoMerge pull request #1626 from teg/networkd
Lennart Poettering [Wed, 21 Oct 2015 10:07:08 +0000 (12:07 +0200)]
Merge pull request #1626 from teg/networkd

networkd: assorted fixes

9 years agonetworkd: manager/link - only serialize once per event-loop iteration
Tom Gundersen [Wed, 30 Sep 2015 16:17:43 +0000 (18:17 +0200)]
networkd: manager/link - only serialize once per event-loop iteration

Every time the state is written out we may trigger third-party apps, so
let's be a bit more careful about writing this out unnecessarily.

9 years agonetworkd: link - serialize addresses
Tom Gundersen [Wed, 30 Sep 2015 13:32:16 +0000 (15:32 +0200)]
networkd: link - serialize addresses

9 years agonetworkd: route - add hash_ops
Tom Gundersen [Sat, 3 Oct 2015 16:40:28 +0000 (18:40 +0200)]
networkd: route - add hash_ops

9 years agonetworkd: route - simplify route_new()
Tom Gundersen [Fri, 9 Oct 2015 21:43:52 +0000 (23:43 +0200)]
networkd: route - simplify route_new()

9 years agonetworkd: address - distinguish between addresses added by us and by others
Tom Gundersen [Wed, 30 Sep 2015 12:01:44 +0000 (14:01 +0200)]
networkd: address - distinguish between addresses added by us and by others

We only keep the addresses that we added ourselves in link->addresses, and
introduce a new set link->addresses_foreign to keep addresses of unknown
origin.

Only functional change is that "foreign" addresses no longer prevent a link
from entering "configured" state.

9 years agonetworkd: address - rework firewall rules lifetime
Tom Gundersen [Wed, 30 Sep 2015 12:07:12 +0000 (14:07 +0200)]
networkd: address - rework firewall rules lifetime

Establish the firewall rule before creating the address, and do not create the address
if the firewall rule could not be created. Also, only drop the firewall rule once
the address has been removed from the kernel.

9 years agonetworkd: address - merge _change() into _configure()
Tom Gundersen [Thu, 1 Oct 2015 15:31:14 +0000 (17:31 +0200)]
networkd: address - merge _change() into _configure()

These functions are almost entirely the same, so avoid duplication.

9 years agonetworkd: address - factor out address_update()
Tom Gundersen [Mon, 28 Sep 2015 15:16:12 +0000 (17:16 +0200)]
networkd: address - factor out address_update()

Call back into link_check_ready() whenever an address state change may have
made a link ready.

9 years agonetworkd: link - only consider configured when all addresses are ready
Tom Gundersen [Mon, 28 Sep 2015 11:38:43 +0000 (13:38 +0200)]
networkd: link - only consider configured when all addresses are ready

We were considering a link configured whilst its IPv6 addresses were still
tentative.

Fixes issue #650.

9 years agoshell-completion: systemd-run: add new properties
Evgeny Vereshchagin [Tue, 20 Oct 2015 19:59:43 +0000 (19:59 +0000)]
shell-completion: systemd-run: add new properties

"ReadWriteDirectories", "ReadOnlyDirectories", "InaccessibleDirectories"

9 years agorun: can launch units with ReadWriteDirectories, ReadOnlyDirectories, InaccessibleDir...
Evgeny Vereshchagin [Tue, 20 Oct 2015 19:57:04 +0000 (19:57 +0000)]
run: can launch units with ReadWriteDirectories, ReadOnlyDirectories, InaccessibleDirectories

9 years agoMerge pull request #1622 from phomes/unused-variable
Ronny Chevalier [Tue, 20 Oct 2015 17:48:27 +0000 (19:48 +0200)]
Merge pull request #1622 from phomes/unused-variable

dbus-execute: remove unused variable

9 years agodbus-execute: remove unused variable
Thomas Hindoe Paaboel Andersen [Tue, 20 Oct 2015 17:36:54 +0000 (19:36 +0200)]
dbus-execute: remove unused variable

from ceb728cf

9 years agoMerge pull request #1616 from evverx/run-fix-environment-parsing
Lennart Poettering [Tue, 20 Oct 2015 13:26:19 +0000 (15:26 +0200)]
Merge pull request #1616 from evverx/run-fix-environment-parsing

run: fix Environment parsing

9 years agoMerge pull request #1538 from ssahani/ipv62
Tom Gundersen [Tue, 20 Oct 2015 12:59:27 +0000 (14:59 +0200)]
Merge pull request #1538 from ssahani/ipv62

networkd: add support to configure IPv6 DAD

9 years agorun: fix Environment parsing
Evgeny Vereshchagin [Tue, 20 Oct 2015 02:04:10 +0000 (02:04 +0000)]
run: fix Environment parsing

* `Environment=` resets previous assignments
* `Environment='a=1 b=2'` sets `a` to `1` and `b` to `2`
* `Environment='"a=1 2" b=2"'` sets `a` to `1 2` and `b` to `2`

9 years agoMerge pull request #1619 from iaguis/nspawn-sysfs-netns-3
Lennart Poettering [Tue, 20 Oct 2015 10:22:32 +0000 (12:22 +0200)]
Merge pull request #1619 from iaguis/nspawn-sysfs-netns-3

nspawn: skip /sys-as-tmpfs if we don't use private-network

9 years agoMerge pull request #1568 from poettering/netclass
Daniel Mack [Tue, 20 Oct 2015 08:31:38 +0000 (10:31 +0200)]
Merge pull request #1568 from poettering/netclass

various fixes, for various things

9 years agonspawn: skip /sys-as-tmpfs if we don't use private-network
Iago López Galeiras [Thu, 15 Oct 2015 10:13:13 +0000 (12:13 +0200)]
nspawn: skip /sys-as-tmpfs if we don't use private-network

Since v3.11/7dc5dbc ("sysfs: Restrict mounting sysfs"), the kernel
doesn't allow mounting sysfs if you don't have CAP_SYS_ADMIN rights over
the network namespace.

So the mounting /sys as a tmpfs code introduced in
d8fc6a000fe21b0c1ba27fbfed8b42d00b349a4b doesn't work with user
namespaces if we don't use private-net. The reason is that we mount
sysfs inside the container and we're in the network namespace of the host
but we don't have CAP_SYS_ADMIN over that namespace.

To fix that, we mount /sys as a sysfs (instead of tmpfs) if we don't use
private network and ignore the /sys-as-a-tmpfs code if we find that /sys
is already mounted as sysfs.

Fixes #1555

9 years agodbus-execute: some cleanups when parsing EnvironmentFiles= for transient units
Lennart Poettering [Mon, 19 Oct 2015 21:43:19 +0000 (23:43 +0200)]
dbus-execute: some cleanups when parsing EnvironmentFiles= for transient units

9 years agostrv: Add _cleanup_strv_free_erase_ and _cleanup_string_free_erase_
Zbigniew Jędrzejewski-Szmek [Thu, 15 Oct 2015 14:02:35 +0000 (10:02 -0400)]
strv: Add _cleanup_strv_free_erase_ and _cleanup_string_free_erase_

9 years agotree-wide: whenever we deal with passwords, erase them from memory after use
Lennart Poettering [Wed, 14 Oct 2015 20:40:23 +0000 (22:40 +0200)]
tree-wide: whenever we deal with passwords, erase them from memory after use

A bit snake-oilish, but can't hurt.

9 years agocore: actually set pids_max field after parsing it
Lennart Poettering [Wed, 14 Oct 2015 20:41:25 +0000 (22:41 +0200)]
core: actually set pids_max field after parsing it

Fixes one of the issues pointed out in #1522.

9 years agojournald: make sure r is always initialized
Lennart Poettering [Wed, 14 Oct 2015 17:33:20 +0000 (19:33 +0200)]
journald: make sure r is always initialized

9 years agojournal: fix bad memory freeing
Lennart Poettering [Wed, 14 Oct 2015 17:33:01 +0000 (19:33 +0200)]
journal: fix bad memory freeing

9 years agotree-wide: add more void casts for various syscall invocations
Lennart Poettering [Wed, 14 Oct 2015 17:32:46 +0000 (19:32 +0200)]
tree-wide: add more void casts for various syscall invocations

9 years agocore: fix type of NetClass dbus property
Lennart Poettering [Wed, 14 Oct 2015 17:31:42 +0000 (19:31 +0200)]
core: fix type of NetClass dbus property

it's exposed as uint32_t and it is internally an uint32_t, hence no need
to treat it as an unsigned.

9 years agoman: move documentation about NetClass from systemd.unit(5) to systemd.resource-contr...
Lennart Poettering [Wed, 14 Oct 2015 17:30:40 +0000 (19:30 +0200)]
man: move documentation about NetClass from systemd.unit(5) to systemd.resource-control(5)

This is after all where we expose all the other cgroup props, especially
those that can be adjusted dynamically.

9 years agoupdate CODING_STYLE
Lennart Poettering [Mon, 19 Oct 2015 20:31:58 +0000 (22:31 +0200)]
update CODING_STYLE

9 years agoupdate TODO
Lennart Poettering [Mon, 19 Oct 2015 20:30:11 +0000 (22:30 +0200)]
update TODO

9 years agoMerge pull request #1614 from alkino/master
Lennart Poettering [Mon, 19 Oct 2015 20:09:10 +0000 (22:09 +0200)]
Merge pull request #1614 from alkino/master

Hook more properties for transient units

9 years agoMerge pull request #1613 from phomes/unused-functions-2
Lennart Poettering [Mon, 19 Oct 2015 20:06:06 +0000 (22:06 +0200)]
Merge pull request #1613 from phomes/unused-functions-2

tree-wide: remove unused functions

9 years agoHook more properties for transient units
Nicolas Cornu [Mon, 19 Oct 2015 19:57:01 +0000 (21:57 +0200)]
Hook more properties for transient units

systemd-run can now launch units with EnvironmentFile set.