chiark / gitweb /
8 years agoresolved: support for TCP DNS queries
Lennart Poettering [Wed, 16 Jul 2014 18:15:47 +0000 (20:15 +0200)]
resolved: support for TCP DNS queries

8 years agodns-packet: allow dynamic resizing of DNS packets
Lennart Poettering [Wed, 16 Jul 2014 16:04:14 +0000 (18:04 +0200)]
dns-packet: allow dynamic resizing of DNS packets

8 years agodns-domain: introduce macros for accessing all DNS header fields
Lennart Poettering [Wed, 16 Jul 2014 16:03:46 +0000 (18:03 +0200)]
dns-domain: introduce macros for accessing all DNS header fields

8 years agoupdate TODO
Lennart Poettering [Wed, 16 Jul 2014 15:24:43 +0000 (17:24 +0200)]
update TODO

8 years agojournal: add systemd-journal-remote to sysusers
Kay Sievers [Wed, 16 Jul 2014 13:14:46 +0000 (15:14 +0200)]
journal: add systemd-journal-remote to sysusers

8 years agojournal-remote: remove obsolete variable
Zbigniew Jędrzejewski-Szmek [Wed, 16 Jul 2014 11:52:41 +0000 (07:52 -0400)]
journal-remote: remove obsolete variable

8 years agojournal-remote: fix double typedef and add missing header
Zbigniew Jędrzejewski-Szmek [Wed, 16 Jul 2014 11:49:21 +0000 (07:49 -0400)]
journal-remote: fix double typedef and add missing header

8 years agoshared: include stdbool.h in mkdir.h
Sjoerd Simons [Wed, 16 Jul 2014 10:09:56 +0000 (12:09 +0200)]
shared: include stdbool.h in mkdir.h

8 years agobuild-sys: don't move libgudev to /lib
Michael Biebl [Wed, 16 Jul 2014 10:09:47 +0000 (12:09 +0200)]
build-sys: don't move libgudev to /lib

It depends on libgobject and libgmodule which are installed in /usr/lib.

8 years agotest-compress-benchmark: add missing % before PRIu64 format
Thomas Hindoe Paaboel Andersen [Wed, 16 Jul 2014 06:59:13 +0000 (08:59 +0200)]
test-compress-benchmark: add missing % before PRIu64 format

8 years agotest-socket-util: silence warnings
Thomas Hindoe Paaboel Andersen [Wed, 16 Jul 2014 06:52:11 +0000 (08:52 +0200)]
test-socket-util: silence warnings

8 years agounits/serial-getty@.service: use the default RestartSec
Michael Olbrich [Tue, 15 Jul 2014 16:28:10 +0000 (18:28 +0200)]
units/serial-getty@.service: use the default RestartSec

For pluggable ttys such as USB serial devices, the getty is restarted
and exits in a loop until the remove event reaches systemd. Under
certain circumstances the restart loop can overload the system in a
way that prevents the remove event from reaching systemd for a long
time (e.g. at least several minutes on a small embedded system).

Use the default RestartSec to prevent the restart loop from
overloading the system. Serial gettys are interactive units, so
waiting an extra 100ms really doesn't make a difference anyways
compared to the time it takes the user to log in.

8 years agoresolve: avoid use of uninitalized variable
Zbigniew Jędrzejewski-Szmek [Wed, 16 Jul 2014 02:47:03 +0000 (22:47 -0400)]
resolve: avoid use of uninitalized variable

8 years agotest-compress-benchmark: silence warnings
Marc-Antoine Perennou [Wed, 16 Jul 2014 01:13:06 +0000 (10:13 +0900)]
test-compress-benchmark: silence warnings

and btw make it pass for 32bits where size_t != uint64_t

8 years agojournal-remote: avoid copying input data
Zbigniew Jędrzejewski-Szmek [Sat, 12 Jul 2014 03:20:08 +0000 (23:20 -0400)]
journal-remote: avoid copying input data

Instead of copying fields into new memory allocations, simply keep pointers
into the receive buffer. Data in this buffer is only copied when there is not
enough space for new data and a large chunk of the buffer contains old data.

8 years agoClear up confusion wrt. ENTRY_SIZE_MAX and DATA_SIZE_MAX
Zbigniew Jędrzejewski-Szmek [Mon, 14 Jul 2014 20:53:23 +0000 (16:53 -0400)]
Clear up confusion wrt. ENTRY_SIZE_MAX and DATA_SIZE_MAX

Define DATA_SIZE_MAX to mean the maximum size of a single
field, and ENTRY_SIZE_MAX to mean the size of the whole
entry, with some rough calculation of overhead over the payload.

Check if entries are not too big when processing native journal

8 years agoµhttp-util: fix compilation without gnutls
Zbigniew Jędrzejewski-Szmek [Sat, 12 Jul 2014 03:17:57 +0000 (23:17 -0400)]
µhttp-util: fix compilation without gnutls

8 years agoman: describe new filename rules for journal-remote
Zbigniew Jędrzejewski-Szmek [Thu, 10 Jul 2014 19:03:28 +0000 (15:03 -0400)]
man: describe new filename rules for journal-remote

8 years agoman: document systemd-journal-upload
Zbigniew Jędrzejewski-Szmek [Thu, 10 Jul 2014 18:50:50 +0000 (14:50 -0400)]
man: document systemd-journal-upload

8 years agojournal-upload: add config file
Zbigniew Jędrzejewski-Szmek [Wed, 16 Jul 2014 02:22:05 +0000 (22:22 -0400)]
journal-upload: add config file

8 years agojournal-remote: let user specify just the main part of the url
Zbigniew Jędrzejewski-Szmek [Thu, 10 Jul 2014 05:39:49 +0000 (01:39 -0400)]
journal-remote: let user specify just the main part of the url

We can append /upload ourselves.

8 years agojournal-remote: rework fd and writer reference handling
Zbigniew Jędrzejewski-Szmek [Wed, 2 Jul 2014 04:15:37 +0000 (00:15 -0400)]
journal-remote: rework fd and writer reference handling

8 years agojournal-remote: improve some messages
Zbigniew Jędrzejewski-Szmek [Wed, 2 Jul 2014 03:18:44 +0000 (23:18 -0400)]
journal-remote: improve some messages

8 years agoFix problem with allocating large buffers and log leftovers
Zbigniew Jędrzejewski-Szmek [Wed, 2 Jul 2014 03:07:45 +0000 (23:07 -0400)]
Fix problem with allocating large buffers and log leftovers

8 years agoConstify ConfigTableItem tables
Zbigniew Jędrzejewski-Szmek [Wed, 16 Jul 2014 01:03:11 +0000 (21:03 -0400)]
Constify ConfigTableItem tables

8 years agojournal-remote: allow splitting incoming logs by source host
Zbigniew Jędrzejewski-Szmek [Sun, 22 Jun 2014 17:36:31 +0000 (13:36 -0400)]
journal-remote: allow splitting incoming logs by source host

Previously existing scheme where the file name would be based on
the source was just too ugly and unpredicatable. Now there are
only two options:
  1. just one file (until rotation),
  2. one file per source host, using the hostname as filename part.
For the cases where the source is specified by the user, only
option one is allowed, and the full of the file must be specified.

8 years agoAllow addresses to be specified for --listen-... args
Zbigniew Jędrzejewski-Szmek [Mon, 30 Jun 2014 05:11:32 +0000 (01:11 -0400)]
Allow addresses to be specified for --listen-... args

Hostnames still aren't accepted.

8 years agoAdd simple generator of fake journal export stream
Zbigniew Jędrzejewski-Szmek [Mon, 30 Jun 2014 04:57:15 +0000 (00:57 -0400)]
Add simple generator of fake journal export stream

8 years agoshared/socket-label: fix error message
Zbigniew Jędrzejewski-Szmek [Mon, 30 Jun 2014 02:06:48 +0000 (22:06 -0400)]
shared/socket-label: fix error message

Was: Failed to listen on [::]:2000: Success

8 years agoshared/socket-util: add function to query remote address
Zbigniew Jędrzejewski-Szmek [Sun, 29 Jun 2014 23:24:56 +0000 (19:24 -0400)]
shared/socket-util: add function to query remote address

8 years agojournal-remote: add units and read certs from default locations
Zbigniew Jędrzejewski-Szmek [Mon, 31 Mar 2014 03:08:02 +0000 (23:08 -0400)]
journal-remote: add units and read certs from default locations

8 years agojournal-upload: add watchdog support
Zbigniew Jędrzejewski-Szmek [Wed, 2 Apr 2014 00:30:13 +0000 (20:30 -0400)]
journal-upload: add watchdog support

8 years agojournal-upload: make state persistent
Zbigniew Jędrzejewski-Szmek [Tue, 1 Apr 2014 13:09:35 +0000 (09:09 -0400)]
journal-upload: make state persistent

8 years agojournal-upload: use journal as the source
Zbigniew Jędrzejewski-Szmek [Sat, 29 Mar 2014 04:37:25 +0000 (00:37 -0400)]
journal-upload: use journal as the source

8 years agojournal-upload: HTTPS support
Zbigniew Jędrzejewski-Szmek [Sat, 29 Mar 2014 04:44:48 +0000 (00:44 -0400)]
journal-upload: HTTPS support

8 years agojournal-upload: a tool to push messages to systemd-journal-remote
Zbigniew Jędrzejewski-Szmek [Tue, 18 Mar 2014 02:54:28 +0000 (22:54 -0400)]
journal-upload: a tool to push messages to systemd-journal-remote

8 years agojournal-remote: reject fields above maximum size
Zbigniew Jędrzejewski-Szmek [Mon, 31 Mar 2014 02:35:37 +0000 (22:35 -0400)]
journal-remote: reject fields above maximum size

Also fix an infinite loop on E2BIG.

Remember what range we already scanned for '\n', to avoid
quadratic behaviour on long "text" fields.

8 years agojournal-remote: small fixes
Zbigniew Jędrzejewski-Szmek [Sun, 23 Mar 2014 16:36:05 +0000 (12:36 -0400)]
journal-remote: small fixes

8 years agobuild-sys: add check for libcurl
Zbigniew Jędrzejewski-Szmek [Tue, 18 Mar 2014 02:52:53 +0000 (22:52 -0400)]
build-sys: add check for libcurl

8 years agoMove network-related journal programs to src/journal-remote/
Zbigniew Jędrzejewski-Szmek [Wed, 26 Mar 2014 02:30:24 +0000 (22:30 -0400)]
Move network-related journal programs to src/journal-remote/

Directory src/journal has become one of the largest directories,
and since systemd-journal-gatewayd, systemd-journal-remote, and
forthcoming systemd-journal-upload are all closely related, create
a separate directory for them.

8 years agomicrohttp-util: rework gnutls logging
Zbigniew Jędrzejewski-Szmek [Sat, 29 Mar 2014 15:58:11 +0000 (11:58 -0400)]
microhttp-util: rework gnutls logging

8 years agojournal: allow files with no data whatsoever
Zbigniew Jędrzejewski-Szmek [Sun, 30 Mar 2014 18:20:34 +0000 (14:20 -0400)]
journal: allow files with no data whatsoever

If a file was opened for writing, and then closed immediately without
actually writing any entries, on subsequent opening, it would be
considered "corrupted". This should be totally fine, and even in
read mode, an empty file can become non-empty later on.

8 years agoresolved: add busname unit file
Kay Sievers [Wed, 16 Jul 2014 01:48:10 +0000 (03:48 +0200)]
resolved: add busname unit file

8 years agoresolved: add legacy dbus service and policy files
Kay Sievers [Wed, 16 Jul 2014 01:39:09 +0000 (03:39 +0200)]
resolved: add legacy dbus service and policy files

8 years agoupdate TODO
Lennart Poettering [Wed, 16 Jul 2014 01:32:29 +0000 (03:32 +0200)]
update TODO

8 years agoresolved: add small NSS module that uses resolved to resolve DNS names
Lennart Poettering [Wed, 16 Jul 2014 01:31:30 +0000 (03:31 +0200)]
resolved: add small NSS module that uses resolved to resolve DNS names

8 years agonss: various minor fixes to nss-myhostname + nss-mymachines
Lennart Poettering [Wed, 16 Jul 2014 01:30:40 +0000 (03:30 +0200)]
nss: various minor fixes to nss-myhostname + nss-mymachines

8 years agodns-domain: never allow labels that are larger than 63 chars
Lennart Poettering [Wed, 16 Jul 2014 01:29:20 +0000 (03:29 +0200)]
dns-domain: never allow labels that are larger than 63 chars

8 years agodns-domain: fix generation of reverse IP address lookup name
Lennart Poettering [Wed, 16 Jul 2014 01:28:52 +0000 (03:28 +0200)]
dns-domain: fix generation of reverse IP address lookup name

8 years agoresolve: add distinct bus error code for hosts that exist but lack A or AAAA records
Lennart Poettering [Wed, 16 Jul 2014 01:28:18 +0000 (03:28 +0200)]
resolve: add distinct bus error code for hosts that exist but lack A or AAAA records

8 years agoupdate TODO
Lennart Poettering [Tue, 15 Jul 2014 22:31:27 +0000 (00:31 +0200)]
update TODO

8 years agoresolved: add a DNS client stub resolver
Lennart Poettering [Tue, 15 Jul 2014 22:26:02 +0000 (00:26 +0200)]
resolved: add a DNS client stub resolver

Let's turn resolved into a something truly useful: a fully asynchronous
DNS stub resolver that subscribes to network changes.

(More to come: caching, LLMNR, mDNS/DNS-SD, DNSSEC, IDN, NSS module)

8 years agonetworkd: remove unused variable
Thomas Hindoe Paaboel Andersen [Tue, 15 Jul 2014 18:38:49 +0000 (20:38 +0200)]
networkd: remove unused variable

8 years agonetworkd: netdev - set mac for bond/bridge devicse
Tom Gundersen [Tue, 15 Jul 2014 17:07:35 +0000 (19:07 +0200)]
networkd: netdev - set mac for bond/bridge devicse

Suggested by poma.

8 years agosd-dhcp-client: make request broadcasts opt-in
Tom Gundersen [Tue, 15 Jul 2014 16:55:31 +0000 (18:55 +0200)]
sd-dhcp-client: make request broadcasts opt-in

It appears there is no good way to decide whether or not broadcasts should be enabled,
there is hardware that must have broadcast, and there are networks that only allow
unicast. So we give up and make this configurable.

By default, unicast is used, but if the kernel were to inform us abotu certain
interfaces requiring broadcast, we could change this to opt-in by default in
those cases.

8 years agorules: uaccess - add ID_SOFTWARE_RADIO
Kay Sievers [Tue, 15 Jul 2014 15:35:53 +0000 (17:35 +0200)]
rules: uaccess - add ID_SOFTWARE_RADIO

On Tue, Jul 15, 2014 at 1:52 PM, Alick Zhao <> wrote:
>>> So maybe ID_SOFTWARE_RADIO ?
>> Hmm, SDR is more a term for a generic technology than for a device
>> class. To me it does not really sound like an administrator would know
>> what this is.
>> What exactly is the device or subsystem you want to make accessible to
>> locally logged-in users only?
> Initially it is bladeRF, but many more are of interest: USRP, rtl-sdr,
> HackRF, ... [1]
> I agree an administrator might not know what SDR is, since it is
> currently still not widely known, and makes sense only for amateurs
> and researchers. But as a SDR fan, I see many new SDR peripherals
> are created recently, and expect to see more. So a generic ID seems
> reasonable to me.
> [1]

8 years agocore: fix oneshot service resource control
Umut Tezduyar Lindskog [Tue, 15 Jul 2014 06:36:29 +0000 (08:36 +0200)]
core: fix oneshot service resource control

Oneshot services's cgroup is removed when the service
exits. An assert is hit otherwise.

8 years agoshell-completion: restore completion for -p
Zbigniew Jędrzejewski-Szmek [Mon, 14 Jul 2014 23:24:46 +0000 (19:24 -0400)]
shell-completion: restore completion for -p

It was broken since systemd was moved out of /bin.

For zsh it was never there.

8 years agotimesyncd: add sockaddr_pretty wrapper
Zbigniew Jędrzejewski-Szmek [Mon, 14 Jul 2014 22:29:27 +0000 (18:29 -0400)]
timesyncd: add sockaddr_pretty wrapper

8 years agotimesyncd: only listen to clock changes when connected
Zbigniew Jędrzejewski-Szmek [Tue, 15 Jul 2014 13:52:17 +0000 (09:52 -0400)]
timesyncd: only listen to clock changes when connected

This reverts previous commit and applies a different fix.

manager_clock_watch() callback calls manager_send_request() to kick
off a resync. We can only do that when we're actually connected to
something. It is not useful to setup the callback from manager_new().

Now the callback will be dropped in manager_connect() and requested
in manager_begin().

8 years agotimesyncd: suppress resync at system time change when not connected
Kay Sievers [Tue, 15 Jul 2014 00:24:35 +0000 (02:24 +0200)]
timesyncd: suppress resync at system time change when not connected

Jul 04 17:46:03 orchid systemd[1]: Starting Network Time Synchronization...
Jul 04 17:46:03 orchid systemd[1]: Started Network Time Synchronization.
Jul 04 17:46:22 orchid systemd-timesyncd[301]: System time changed. Resyncing.
Jul 04 17:46:22 orchid systemd-timesyncd[301]: Assertion 'm->current_server_name'

8 years agorules: consistently use "?*" instead of "*?"
Kay Sievers [Tue, 15 Jul 2014 00:04:47 +0000 (02:04 +0200)]
rules: consistently use "?*" instead of "*?"

8 years agoTODO
Tom Gundersen [Mon, 14 Jul 2014 20:08:10 +0000 (22:08 +0200)]

8 years agonetworkd: link - fix memory leak
Tom Gundersen [Sat, 12 Jul 2014 23:11:52 +0000 (01:11 +0200)]
networkd: link - fix memory leak

Make link_initialized() idempotent to avoid taking refs on several udev_device

8 years agotest: network - skip if we don't have permissions to create netdevs
Tom Gundersen [Sat, 12 Jul 2014 21:15:17 +0000 (23:15 +0200)]
test: network - skip if we don't have permissions to create netdevs

8 years agosd-dhcp-server: make gcc happy
Tom Gundersen [Sat, 12 Jul 2014 21:07:33 +0000 (23:07 +0200)]
sd-dhcp-server: make gcc happy

It complains about optoffset possibly being uninitialized. It is wrong,
but let's just initialize it.

8 years agosd-dhcp6-client: make gcc happy
Tom Gundersen [Sat, 12 Jul 2014 21:07:33 +0000 (23:07 +0200)]
sd-dhcp6-client: make gcc happy

It complains about {max,init}_retransmit_time possibly being
uninitialized. It is wrong, but let's just initialize it.

8 years agonetworkd: netdev - introduce vtable for netdev kinds
Tom Gundersen [Sun, 6 Jul 2014 12:07:34 +0000 (14:07 +0200)]
networkd: netdev - introduce vtable for netdev kinds

Split each netdev kind into its own .h/.c.

8 years agonetworkd: netdev - rework load_one
Tom Gundersen [Fri, 4 Jul 2014 15:03:37 +0000 (17:03 +0200)]
networkd: netdev - rework load_one

We now:
 - parse config
 - match on environment
 - verify and complement config
 - create netdev

8 years agonetworkd: netdev - split out bridge creation
Tom Gundersen [Sat, 5 Jul 2014 13:00:59 +0000 (15:00 +0200)]
networkd: netdev - split out bridge creation

8 years agonetworkd: netdev - rename 'enslave' to 'join'
Tom Gundersen [Sat, 5 Jul 2014 12:53:54 +0000 (14:53 +0200)]
networkd: netdev - rename 'enslave' to 'join'

Enslave only really makes sense when referring to bridges and bonds, so try to be
a bit more neutral.

8 years agoman: systemd.netdev - make it clear that we do not touch preexisting netdevs
Tom Gundersen [Mon, 14 Jul 2014 10:25:42 +0000 (12:25 +0200)]
man: systemd.netdev - make it clear that we do not touch preexisting netdevs

We will happily use bridges/bonds as master devices, but we will not change their settings if they were created by
someone else.

8 years agonetworkd: add back route destination support
Tom Gundersen [Mon, 14 Jul 2014 10:14:23 +0000 (12:14 +0200)]
networkd: add back route destination support

This was accidentally dropped when adding metric support.

8 years agonetworkd: bond - explicitly map to kernel mode values
Tom Gundersen [Mon, 14 Jul 2014 10:13:45 +0000 (12:13 +0200)]
networkd: bond - explicitly map to kernel mode values

8 years agonetworkd: dhcp add vendor class indentifier option 60
Susant Sahani [Mon, 14 Jul 2014 08:04:18 +0000 (13:34 +0530)]
networkd: dhcp add vendor class indentifier option 60

Vendor Class Identifier be used by DHCP clients to identify
their vendor type and configuration. When using this option,
vendors can define their own specific identifier values, such
as to convey a particular hardware or operating system
configuration or other identifying information.

Vendor-specified DHCP options—features that let administrators assign
separate options to clients with similar configuration requirements.
For example, if DHCP-aware clients for example we want to separate
different gateway and option for different set of people
(dev/test/hr/finance) in a org or devices for example web/database
servers or let's say in a embedded device etc and require a different
default gateway or DNS server than the rest of clients.

8 years agonetworkd: make metric of routes configurable
Susant Sahani [Thu, 10 Jul 2014 17:39:58 +0000 (23:09 +0530)]
networkd: make metric of routes configurable

Now route metric can be configuted via conf file:

example conf:



ip route output
default via dev em1 metric 10

[tomegun: squash TODO update and reword man page a bit]

8 years agonetworkd: return 1 from successful event handlers
Tom Gundersen [Mon, 14 Jul 2014 09:04:13 +0000 (11:04 +0200)]
networkd: return 1 from successful event handlers

8 years agoTODO: update networkd
Tom Gundersen [Mon, 14 Jul 2014 08:50:37 +0000 (10:50 +0200)]
TODO: update networkd

8 years agofileio: quote more shell characters in envfiles
Mantas Mikulėnas [Sun, 13 Jul 2014 15:49:00 +0000 (18:49 +0300)]
fileio: quote more shell characters in envfiles

Turns out, making strings shell-proof is harder than expected:

    # machinectl set-hostname "foo|poweroff" && . /etc/machine-info

(This could be simplified by quoting *and* escaping all characters,
which is harmless in shell but unnecessary.)

8 years agopath-lookup: don't make ~/.local/share/systemd/user a symlink
Tanu Kaskinen [Sat, 12 Jul 2014 15:59:19 +0000 (18:59 +0300)]
path-lookup: don't make ~/.local/share/systemd/user a symlink

We already encourage upstreams to keep the default configuration
separate from user customizations for software that is installed in
the system location. Let's allow that separation also for software
that is installed in the home directory.

Some discussion:

8 years agoman: mention XDG_DATA_HOME in systemd.unit
Tanu Kaskinen [Sat, 12 Jul 2014 16:36:16 +0000 (19:36 +0300)]
man: mention XDG_DATA_HOME in systemd.unit

8 years agoman: sysusers.d correct default user shell
Sjoerd Simons [Sun, 13 Jul 2014 14:56:16 +0000 (16:56 +0200)]
man: sysusers.d correct default user shell

For the non-root user sysusers uses nologin as the default shell, not
login. Correct the documentation to match the code.

8 years agoman: add systemd-coredump(8) and a bunch of links
Zbigniew Jędrzejewski-Szmek [Mon, 14 Jul 2014 00:32:46 +0000 (20:32 -0400)]
man: add systemd-coredump(8) and a bunch of links

8 years agoAdd function to open temp files in selinux mode
Zbigniew Jędrzejewski-Szmek [Mon, 14 Jul 2014 01:10:38 +0000 (21:10 -0400)]
Add function to open temp files in selinux mode

8 years agosysusers: preserve label of /etc/{passwd, group}
Colin Walters [Sun, 13 Jul 2014 20:35:33 +0000 (13:35 -0700)]
sysusers: preserve label of /etc/{passwd, group}

These files are specially labeled on SELinux systems, and we need to
preserve that label.

8 years agobuild-sys: Do not distribute generated emergency.service
Jon Severinsson [Fri, 11 Jul 2014 12:37:36 +0000 (14:37 +0200)]
build-sys: Do not distribute generated emergency.service

It is already in nodist_systemunit_DATA and if it is
shipped, it contains the hardcoded path to systemctl
which will cause it to fail to start when
rootprefix != prefix and rootbindir != bindir.

8 years agoRevert "build-sys: include PolicyKit files as part of distribution"
Mike Gilbert [Fri, 4 Jul 2014 18:43:14 +0000 (14:43 -0400)]
Revert "build-sys: include PolicyKit files as part of distribution"

This reverts commit 0c26bfc3d21fdb3963f1248c237e2f1a33b5566d.

src/core/ depends on values which
are specified at configure time, so we cannot ship the corresponding
policy file in the tarball.

Since we need to regenerate one policy file, we might as well generate
them all.

8 years agosd-event: don't require a signal event source to be enabled for the child event sourc...
Lennart Poettering [Fri, 11 Jul 2014 14:48:35 +0000 (16:48 +0200)]
sd-event: don't require a signal event source to be enabled for the child event source to work

8 years agogitignore: ignore .swp files
David Herrmann [Fri, 11 Jul 2014 14:35:37 +0000 (16:35 +0200)]
gitignore: ignore .swp files

vim places them in the source-tree while editing files. Ignore them.

8 years agosd-event: always call epoll_ctl() on mask-updates if edge-triggered
David Herrmann [Wed, 9 Jul 2014 22:47:23 +0000 (00:47 +0200)]
sd-event: always call epoll_ctl() on mask-updates if edge-triggered

A call to sd_event_source_set_io_events() skipps calling into the kernel
if the new event-mask matches the old one. This is safe for
level-triggered sources as the kernel moves them onto the ready-list
automatically if events change. However, edge-triggered sources might not
be on the ready-list even though events are present.

A call to sd_event_source_set_io_events() with EPOLLET set might thus be
used to just move the io-source onto the ready-list so the next poll
will return it again. This is very useful to avoid starvation in
priority-based event queues.

Imagine a read() loop on an edge-triggered fd. If we cannot read data fast
enough to drain the receive queue, we might decide to skip reading for now
and schedule it for later. On edge-triggered io-sources we have to make
sure it's put on the ready-list so the next dispatch-round will return it
again if it's still the highest priority task. We could make sd-event
handle edge-triggered sources directly and allow marking them ready again.
However, it's much simpler to let the kernel do that for now via

8 years agoshared: fix coding-style for ring-buffer implementation
David Herrmann [Wed, 9 Jul 2014 18:20:46 +0000 (20:20 +0200)]
shared: fix coding-style for ring-buffer implementation

We use "typedef struct Ring Ring" with camel-case for internal objects.
So rename "struct ring" to "Ring".

8 years agoshared: add MIN3 macro
David Herrmann [Mon, 30 Jun 2014 13:43:40 +0000 (15:43 +0200)]
shared: add MIN3 macro

This is like MIN but evaluates 3 arguments. We already have MAX3, so add
the equivalent for MIN.

8 years agoFix build without any compression enabled
Zbigniew Jędrzejewski-Szmek [Fri, 11 Jul 2014 14:42:06 +0000 (10:42 -0400)]
Fix build without any compression enabled

8 years agoendian: explicitly include endian.h wherever we want to use __BYTE_ORDER
Lennart Poettering [Fri, 11 Jul 2014 14:13:13 +0000 (16:13 +0200)]
endian: explicitly include endian.h wherever we want to use __BYTE_ORDER

8 years agoalways check for __BYTE_ORDER == __BIG_ENDIAN when checking for endianess
Lennart Poettering [Fri, 11 Jul 2014 13:56:16 +0000 (15:56 +0200)]
always check for __BYTE_ORDER == __BIG_ENDIAN when checking for endianess

Let's always stick to glibc's way to determine byte order, and not mix
autoconf-specific checks with gcc checks.

8 years agohostnamed: introduce new location machin-info field, too
Lennart Poettering [Fri, 11 Jul 2014 13:50:32 +0000 (15:50 +0200)]
hostnamed: introduce new location machin-info field, too

8 years agohostnamed: drop nss-myhostname check
Lennart Poettering [Fri, 11 Jul 2014 13:38:17 +0000 (15:38 +0200)]
hostnamed: drop nss-myhostname check

The check only cares about whether the module is installed, not enabled.
But installation we should know anyway, after all we ship the module
with systemd these days...

8 years agohostnamed: make use of in_charset() to verify charset
Lennart Poettering [Fri, 11 Jul 2014 13:37:11 +0000 (15:37 +0200)]
hostnamed: make use of in_charset() to verify charset

8 years agonss-myhostname: simplify array building a bit
Lennart Poettering [Fri, 11 Jul 2014 13:34:18 +0000 (15:34 +0200)]
nss-myhostname: simplify array building a bit

8 years agohostnamed: minor modernization
Lennart Poettering [Fri, 11 Jul 2014 13:31:49 +0000 (15:31 +0200)]
hostnamed: minor modernization