chiark / gitweb /
elogind.git
5 years agoshell-completion: add full support for completing busctl command lines, with services...
Lennart Poettering [Tue, 23 Dec 2014 21:43:16 +0000 (22:43 +0100)]
shell-completion: add full support for completing busctl command lines, with services, objects, interfaces, members, and signatures

5 years agobusctl: when introspecting objects, optionally limit output by interface name
Lennart Poettering [Tue, 23 Dec 2014 21:42:55 +0000 (22:42 +0100)]
busctl: when introspecting objects, optionally limit output by interface name

5 years agogitignore: hide test-lldp files
Lennart Poettering [Tue, 23 Dec 2014 20:34:55 +0000 (21:34 +0100)]
gitignore: hide test-lldp files

5 years agomachined: introduce polkit for OpenLogin() call
Lennart Poettering [Tue, 23 Dec 2014 20:28:22 +0000 (21:28 +0100)]
machined: introduce polkit for OpenLogin() call

This way "machinectl login" can be opened up to run without privileges.

5 years agobus: add missing bus-policy.[ch]
Lennart Poettering [Tue, 23 Dec 2014 20:06:01 +0000 (21:06 +0100)]
bus: add missing bus-policy.[ch]

Accidentally forgot to commit this. Sorry!

5 years agobuild-sys: move core/build.h → shared/build.h
Lennart Poettering [Tue, 23 Dec 2014 18:19:11 +0000 (19:19 +0100)]
build-sys: move core/build.h → shared/build.h

After all, pretty much all our tools include it, and it should hence be
shared.

Also move sysfs-show.h from core/ to login/, since it has no point to
exist in core.

5 years agoupdate TODO
Lennart Poettering [Tue, 23 Dec 2014 18:14:26 +0000 (19:14 +0100)]
update TODO

5 years agomachinectl: rework 'machinectl login' to use OpenMachineLogin()
Lennart Poettering [Tue, 23 Dec 2014 18:14:04 +0000 (19:14 +0100)]
machinectl: rework 'machinectl login' to use OpenMachineLogin()

5 years agomachined: add new call OpenMachineLogin() that starts a getty in a container on a...
Lennart Poettering [Tue, 23 Dec 2014 18:12:49 +0000 (19:12 +0100)]
machined: add new call OpenMachineLogin() that starts a getty in a container on a pty and returns the pty master fd to the client

This is a one-stop solution for "machinectl login", and should simplify
getting logins in containers.

5 years agosd-bus: teach x-container-unix: bus protoocol to connect to the namespace of a PID...
Lennart Poettering [Tue, 23 Dec 2014 18:11:48 +0000 (19:11 +0100)]
sd-bus: teach x-container-unix: bus protoocol to connect to the namespace of a PID instead of a container name

5 years agocondition: don't include files from src/core
Lennart Poettering [Tue, 23 Dec 2014 18:07:01 +0000 (19:07 +0100)]
condition: don't include files from src/core

5 years agoimport: fix compiler warning
Lennart Poettering [Tue, 23 Dec 2014 18:05:23 +0000 (19:05 +0100)]
import: fix compiler warning

5 years agoenv-util: don't include files from src/core/
Lennart Poettering [Tue, 23 Dec 2014 18:04:56 +0000 (19:04 +0100)]
env-util: don't include files from src/core/

5 years agocore: rearrange code so that libsystemd/sd-bus/ does not include header files from...
Lennart Poettering [Tue, 23 Dec 2014 17:36:04 +0000 (18:36 +0100)]
core: rearrange code so that libsystemd/sd-bus/ does not include header files from core

Stuff in src/shared or src/libsystemd should *never* include code from
src/core or any of the tools, so don't do that here either. It's not OK!

5 years agobus-proxyd: ignore errors from sd_bus_creds_get_well_known_names()
Daniel Mack [Tue, 23 Dec 2014 17:36:29 +0000 (18:36 +0100)]
bus-proxyd: ignore errors from sd_bus_creds_get_well_known_names()

sd_bus_creds_get_well_known_names() fails with -ENODATA in case the
message has no names attached, which is intended behavior if the
remote connection didn't own any names at the time of sending.

The function already deals with 'sender_names' being an empty strv,
so we can just continue in such cases.

5 years agobus-proxyd: handle -ESRCH and -ENXIO gracefully
Daniel Mack [Tue, 23 Dec 2014 12:37:52 +0000 (13:37 +0100)]
bus-proxyd: handle -ESRCH and -ENXIO gracefully

Messages to destinations that are not currently owned by any bus connection
will cause kdbus related function to return with either -ENXIO or -ESRCH.

Such conditions should not make the proxyd terminate but send a sane
SD_BUS_ERROR_NAME_HAS_NO_OWNER error reply to the proxied connection.

5 years agoupdate TODO
Lennart Poettering [Tue, 23 Dec 2014 01:11:48 +0000 (02:11 +0100)]
update TODO

5 years agoutil: add allocation loop to gettyname_malloc()
Lennart Poettering [Tue, 23 Dec 2014 01:10:08 +0000 (02:10 +0100)]
util: add allocation loop to gettyname_malloc()

5 years agonspawn,pty: port over to new ptsname_malloc() helper
Lennart Poettering [Tue, 23 Dec 2014 01:02:08 +0000 (02:02 +0100)]
nspawn,pty: port over to new ptsname_malloc() helper

5 years agosystemd-run: support -t mode when combined with -M
Lennart Poettering [Tue, 23 Dec 2014 00:58:49 +0000 (01:58 +0100)]
systemd-run: support -t mode when combined with -M

For that, ask machined for a container PTY and use that.

5 years agomachined: add OpenMachinePTY() bus call for allocating a PTY device within a container
Lennart Poettering [Mon, 22 Dec 2014 20:17:29 +0000 (21:17 +0100)]
machined: add OpenMachinePTY() bus call for allocating a PTY device within a container

Then, port "machinectl" over to make use of it.

5 years agosystemd-run: add --quiet mode to suppress informational message on TTY usage
Lennart Poettering [Mon, 22 Dec 2014 19:39:10 +0000 (20:39 +0100)]
systemd-run: add --quiet mode to suppress informational message on TTY usage

5 years agomachinectl,nspawn: don't print extra final newline if pty terminal output was newline...
Lennart Poettering [Mon, 22 Dec 2014 19:33:45 +0000 (20:33 +0100)]
machinectl,nspawn: don't print extra final newline if pty terminal output was newline-terinated anyway

5 years agorun: add a new "-t" mode for invoking a binary on an allocated TTY
Lennart Poettering [Mon, 22 Dec 2014 18:45:32 +0000 (19:45 +0100)]
run: add a new "-t" mode for invoking a binary on an allocated TTY

5 years agoimport: rename 'poll-dck' to 'pull-dkr'
Lennart Poettering [Mon, 22 Dec 2014 18:42:27 +0000 (19:42 +0100)]
import: rename 'poll-dck' to 'pull-dkr'

I figure "pull-dck" is not a good name, given that one could certainly
read the verb in a way that might be funny for 16year-olds. ;-)

Also, don't hardcode the index URL to use, make it runtime and configure
time configurable instead.

5 years agoshared: json - support escaping utf16 surrogate pairs
Tom Gundersen [Mon, 22 Dec 2014 13:53:40 +0000 (14:53 +0100)]
shared: json - support escaping utf16 surrogate pairs

We originally only supported escaping ucs2 encoded characters (as \uxxxx). This
only covers the BMP. Support escaping also utf16 surrogate pairs (on the form
\uxxxx\uyyyy) to cover all of unicode.

5 years agoshared: utf8 - support decoding the full utf16
Tom Gundersen [Mon, 22 Dec 2014 11:57:05 +0000 (12:57 +0100)]
shared: utf8 - support decoding the full utf16

We originally only supported the BMP (i.e., we treated UTF-16 as UCS-2).

5 years agoshared: utf8 - support ucs4 -> utf8
Tom Gundersen [Sun, 21 Dec 2014 23:58:26 +0000 (00:58 +0100)]
shared: utf8 - support ucs4 -> utf8

Originally we only supported ucs2, so move the ucs4 version from libsystemd-terminal to shared
and use that everywhere.

5 years agosd-bus: ignore KDBUS_ITEM_TIMESTAMP in kernel messages
Daniel Mack [Mon, 22 Dec 2014 18:59:53 +0000 (19:59 +0100)]
sd-bus: ignore KDBUS_ITEM_TIMESTAMP in kernel messages

Kernel notifications carry a timestamp now, so make sure
bus_kernel_translate_message() doesn't complain when it stumbles across
them.

5 years agobusname: fix CMD_FREE ioctl
Daniel Mack [Mon, 22 Dec 2014 15:51:15 +0000 (16:51 +0100)]
busname: fix CMD_FREE ioctl

The KDBUS_CMD_FREE ioctl struct has a size field now, which needs to be set.

5 years agoverbs: fix typo in error message
Dave Reisner [Mon, 22 Dec 2014 02:28:49 +0000 (21:28 -0500)]
verbs: fix typo in error message

5 years agohwdb: Lenovo USB mouse MO28UOL DPI data
Frank Theile [Sun, 21 Dec 2014 22:18:55 +0000 (08:18 +1000)]
hwdb: Lenovo USB mouse MO28UOL DPI data

https://bugs.freedesktop.org/show_bug.cgi?id=87504

5 years agosd-bus: sync kdbus.h (ABI break)
Daniel Mack [Sat, 20 Dec 2014 18:23:49 +0000 (19:23 +0100)]
sd-bus: sync kdbus.h (ABI break)

Sync kdbus.h with upstream changes:

  * Two optional cancellation points where added for synchronously
    blocking KDBUS_CMD_SEND commands: A sigmask to change the mask
    of accepted signals before the task is put to sleep, and a
    generic file descriptor that can be written to, in order to cancel
    the command. Both methods are currently unused.

  * The KDBUS_CMD_CANCEL ioctl was removed. sd-bus was never using
    that command, so there's no change needed.

  * Some kerneldoc fixes

5 years agobuild: add option to disable hwdb
Umut Tezduyar Lindskog [Fri, 19 Dec 2014 10:47:30 +0000 (11:47 +0100)]
build: add option to disable hwdb

[tomegun: fix Makefile-man.am, based on fix from Michael Biebl]

5 years agosd-lldp: minor header cleanup
Tom Gundersen [Fri, 19 Dec 2014 21:42:03 +0000 (22:42 +0100)]
sd-lldp: minor header cleanup

 * (potentially) public headers must reside in src/systemd/ (not in
   src/libsystemd*)
 * some private (not prefixed with sd_) functions moved from sd-lldp.h to
   lldp-internal.h
 * introduce lldp-util.h for the cleanup macro, as these should not be public
 * rename the cleanup macro, we always name them _cleanup_foo_, never
   _cleanup_sd_foo_
 * mark some function arguments as 'const'

5 years agoupdate TODO
Lennart Poettering [Fri, 19 Dec 2014 19:44:49 +0000 (20:44 +0100)]
update TODO

5 years agomachined: add a full bus object for images
Lennart Poettering [Fri, 19 Dec 2014 19:43:18 +0000 (20:43 +0100)]
machined: add a full bus object for images

5 years agomachined: add new GetImage() bus call for retrieving the bus path for an image
Lennart Poettering [Fri, 19 Dec 2014 19:07:23 +0000 (20:07 +0100)]
machined: add new GetImage() bus call for retrieving the bus path for an image

5 years agosysv-generator: properly add Makefile symlink
Lennart Poettering [Fri, 19 Dec 2014 19:04:55 +0000 (20:04 +0100)]
sysv-generator: properly add Makefile symlink

5 years agoutil: rename ignore_file() to hidden_file()
Lennart Poettering [Fri, 19 Dec 2014 19:03:36 +0000 (20:03 +0100)]
util: rename ignore_file() to hidden_file()

hidden_file() is a bit more precise, since dot files usually shouldn't
be ignored, but certainly be considered hidden.

5 years agoupdate TODO
Lennart Poettering [Fri, 19 Dec 2014 18:19:20 +0000 (19:19 +0100)]
update TODO

5 years agomachinectl: port machinectl to new verbs logic
Lennart Poettering [Fri, 19 Dec 2014 18:19:05 +0000 (19:19 +0100)]
machinectl: port machinectl to new verbs logic

5 years agoimport: Verb[] array can be static, too
Lennart Poettering [Fri, 19 Dec 2014 18:18:48 +0000 (19:18 +0100)]
import: Verb[] array can be static, too

5 years agoverbs: when invoking the default verb, pass a faked argv array, with just the verb...
Lennart Poettering [Fri, 19 Dec 2014 18:17:52 +0000 (19:17 +0100)]
verbs: when invoking the default verb, pass a faked argv array, with just the verb in it

That way the dispatcher calls know how they got called.

5 years agomachined/machinectl: add logic to show list of available images
Lennart Poettering [Fri, 19 Dec 2014 17:42:50 +0000 (18:42 +0100)]
machined/machinectl: add logic to show list of available images

This adds a new bus call to machined that enumerates /var/lib/container
and returns all trees stored in it, distuingishing three types:

        - GPT disk images, which are files suffixed with ".gpt"
        - directory trees
        - btrfs subvolumes

5 years agoutil: when creating temporary filename for atomic creation of files, add an extra...
Lennart Poettering [Fri, 19 Dec 2014 17:40:23 +0000 (18:40 +0100)]
util: when creating temporary filename for atomic creation of files, add an extra "#" to the name

That way, we have a simple, somewhat reliable way to detect such
temporary files, by simply checking if they start with ".#".

5 years agojournal: skipping of exhausted journal files is bad if direction changed
Michal Schmidt [Fri, 19 Dec 2014 16:11:22 +0000 (17:11 +0100)]
journal: skipping of exhausted journal files is bad if direction changed

EOF is meaningless if the direction of iteration changes.
Move the EOF optimization under the direction check.

This fixes test-journal-interleaving for me.

Thanks to Filipe Brandenburger for telling me about the failure.

5 years agojournal: make next_with_matches() always use f->current_offset
Michal Schmidt [Fri, 19 Dec 2014 13:42:55 +0000 (14:42 +0100)]
journal: make next_with_matches() always use f->current_offset

next_with_matches() is odd in that its "unit64_t *offset" parameter is
both input and output. In other it's purely for output.

The function is called from two places in next_beyond_location(). In
both of them "&cp" is used as the argument and in both cases cp is
guaranteed to equal f->current_offset.

Let's just have next_with_matches() ignore "*offset" on input and
operate with f->current_offset.

I did not investigate why it is, but it makes my usual benchmark run
reproducibly faster:

$ time ./journalctl --since=2014-06-01 --until=2014-07-01 > /dev/null

real    0m4.032s
user    0m3.896s
sys     0m0.135s

(Compare to preceding commit, where real was 4.4s.)

5 years agojournal: fix skipping of duplicate entries in iteration
Michal Schmidt [Fri, 19 Dec 2014 14:05:30 +0000 (15:05 +0100)]
journal: fix skipping of duplicate entries in iteration

I accidentally broke the detection of duplicate entries in 7943f42275
"journal: optimize iteration by returning previously found candidate
entry".

When we have a known location of a candidate entry, we must not return
from next_beyond_location() immediately. We must go through the
duplicates detection to make sure the candidate differs from the
already iterated entry.

This fix slows down iteration a bit, but it's still faster than it
was before the rework.

$ time ./journalctl --since=2014-06-01 --until=2014-07-01 > /dev/null

real    0m4.448s
user    0m4.298s
sys     0m0.149s

(Compare with results from commit 7943f42275, where real was 5.3s before
the rework.)

5 years agonetworkd: link - plug leak
Tom Gundersen [Fri, 19 Dec 2014 07:34:04 +0000 (08:34 +0100)]
networkd: link - plug leak

5 years agonetworkctl: lldp - respect arg_legend
Tom Gundersen [Fri, 19 Dec 2014 07:33:46 +0000 (08:33 +0100)]
networkctl: lldp - respect arg_legend

5 years agoLLDP: Add support for networkctl
Susant Sahani [Thu, 11 Dec 2014 04:29:55 +0000 (09:59 +0530)]
LLDP: Add support for networkctl

5 years agonetworkd: integrate LLDP
Susant Sahani [Sun, 23 Nov 2014 04:26:14 +0000 (09:56 +0530)]
networkd: integrate LLDP

This patch integrates LLDP with networkd.

Example conf:
file : lldp.network

[Match]
Name=em1

[Network]
LLDP=yes

5 years agonetworkd: Introduce Link Layer Discovery Protocol (LLDP)
Susant Sahani [Sun, 23 Nov 2014 04:16:36 +0000 (09:46 +0530)]
networkd: Introduce Link Layer Discovery Protocol (LLDP)

This patch introduces LLDP support to networkd. it implements the
receiver side of the protocol.

The Link Layer Discovery Protocol (LLDP) is an industry-standard,
vendor-neutral method to allow networked devices to advertise
capabilities, identity, and other information onto a LAN. The Layer 2
protocol, detailed in IEEE 802.1AB-2005.LLDP allows network devices
that operate at the lower layers of a protocol stack (such as
Layer 2 bridges and switches) to learn some of the capabilities
and characteristics of LAN devices available to higher
layer protocols.

5 years agonetworkctl: port to verbs helper
Tom Gundersen [Fri, 19 Dec 2014 02:16:45 +0000 (03:16 +0100)]
networkctl: port to verbs helper

5 years agosystemd-hwdb: port to new verbs helper
Tom Gundersen [Fri, 19 Dec 2014 02:01:27 +0000 (03:01 +0100)]
systemd-hwdb: port to new verbs helper

5 years agoimport: dck - fix curl error handling
Tom Gundersen [Fri, 19 Dec 2014 02:00:56 +0000 (03:00 +0100)]
import: dck - fix curl error handling

5 years agoupdate TODO
Lennart Poettering [Fri, 19 Dec 2014 01:07:35 +0000 (02:07 +0100)]
update TODO

5 years agoimport: add new minimal tool "systemd-import" for pulling down foreign containers...
Lennart Poettering [Fri, 19 Dec 2014 00:59:52 +0000 (01:59 +0100)]
import: add new minimal tool "systemd-import" for pulling down foreign containers and install them locally

This adds a simply but powerful tool for downloading container images
from the most popular container solution used today. Use it like
this:

       # systemd-import pull-dck mattdm/fedora
       # systemd-nspawn -M fedora

This will donwload the layers for "mattdm/fedora", and make them
available locally as /var/lib/container/fedora.

The tool is pretty complete, as long as it's only about pulling down
images, or updating them. Pushing or searching is not supported yet.

5 years agoutil: add generalization of verb parsing for command line tools
Lennart Poettering [Fri, 19 Dec 2014 00:48:05 +0000 (01:48 +0100)]
util: add generalization of verb parsing for command line tools

We should move loginctl, timedatectl, machinectl over to use this new
API instead of a manual one.

5 years agoutil: make sure rm_rf() can be called on symlinks (with the effect of deleting it)
Lennart Poettering [Fri, 19 Dec 2014 00:47:19 +0000 (01:47 +0100)]
util: make sure rm_rf() can be called on symlinks (with the effect of deleting it)

5 years agostrv: ass new strv_is_uniq() and strv_reverse() calls
Lennart Poettering [Fri, 19 Dec 2014 00:31:59 +0000 (01:31 +0100)]
strv: ass new strv_is_uniq() and strv_reverse() calls

5 years agoexecute: the runtime directory can only be on tmpfs, hence don't use rm_rf_dangerous...
Lennart Poettering [Fri, 19 Dec 2014 00:31:27 +0000 (01:31 +0100)]
execute: the runtime directory can only be on tmpfs, hence don't use rm_rf_dangerous() needlessly

5 years agonss-myhostname: introduce is_gateway() similar to the existing is_hostname()
Lennart Poettering [Thu, 18 Dec 2014 00:44:34 +0000 (01:44 +0100)]
nss-myhostname: introduce is_gateway() similar to the existing is_hostname()

5 years agoload-fragment: allow quoting in command name and document allowed escapes
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 23:08:13 +0000 (18:08 -0500)]
load-fragment: allow quoting in command name and document allowed escapes

The handling of the command name and other arguments is unified. This
simplifies things and should make them more predictable for users.
Incidentally, this makes ExecStart handling match the .desktop file
specification, apart for the requirment for an absolute path.

https://bugs.freedesktop.org/show_bug.cgi?id=86171

5 years agoTreat a trailing backslash as an error
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 22:51:38 +0000 (17:51 -0500)]
Treat a trailing backslash as an error

Commit a2a5291b3f5 changed the parser to reject unfinished quoted
strings. Unfortunately it introduced an error where a trailing
backslash would case an infinite loop. Of course this must fixed, but
the question is what to to instead. Allowing trailing backslashes and
treating them as normal characters would be one option, but this seems
suboptimal. First, there would be inconsistency between handling of
quoting and of backslashes. Second, a trailing backslash is most
likely an error, at it seems better to point it out to the user than
to try to continue.

Updated rules:
ExecStart=/bin/echo \\ → OK, prints a backslash
ExecStart=/bin/echo \ → error
ExecStart=/bin/echo "x → error
ExecStart=/bin/echo "x"y → error

5 years agotest-strv: use STRV_MAKE
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 22:01:42 +0000 (17:01 -0500)]
test-strv: use STRV_MAKE

5 years agotree-wide: make condition_free_list return NULL
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 17:33:05 +0000 (12:33 -0500)]
tree-wide: make condition_free_list return NULL

5 years agocore: make exec_command_free_list return NULL
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 17:29:24 +0000 (12:29 -0500)]
core: make exec_command_free_list return NULL

5 years agohwdb: move Logitech mouse to the right spot
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 23:42:04 +0000 (18:42 -0500)]
hwdb: move Logitech mouse to the right spot

5 years agohwdb: Logitech, Inc. RX 250 Optical Mouse DPI data
Erik Auerswald [Thu, 18 Dec 2014 23:46:29 +0000 (18:46 -0500)]
hwdb: Logitech, Inc. RX 250 Optical Mouse DPI data

https://bugs.freedesktop.org/show_bug.cgi?id=87435

5 years agohwdb: Logitech, Inc. RX 300 Optical Mouse DPI data
Tim JP [Thu, 18 Dec 2014 23:40:44 +0000 (18:40 -0500)]
hwdb: Logitech, Inc. RX 300 Optical Mouse DPI data

https://bugs.freedesktop.org/show_bug.cgi?id=87456

5 years agotest-cap-list: allow mismatch in case
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 22:40:32 +0000 (17:40 -0500)]
test-cap-list: allow mismatch in case

5 years agopath-util: fix breakage in path_is_mount_point
Dave Reisner [Thu, 18 Dec 2014 23:10:46 +0000 (18:10 -0500)]
path-util: fix breakage in path_is_mount_point

This fixes 2 problems introduced by 6feeeab0bc:

1) If name_to_handle_at returns ENOSYS for the child, we'll wrongly
return -ENOSYS when it returns the same for the parent. Immediately
jump to the fallback logic when we get ENOSYS.

2) If name_to_handle_at returns EOPNOTSUPP for the child but suceeds
for the parent, we'll be comparing an uninitialized value (mount_id) to
an initialized value (mount_id_parent). Initialize the mount_id
variables to invalid mount_ids to avoid this.

5 years agosystemd-hwdb: introduce new tool
Tom Gundersen [Mon, 15 Dec 2014 23:48:24 +0000 (00:48 +0100)]
systemd-hwdb: introduce new tool

This pulls out the hwdb managment from udevadm into an independent tool.

The old code is left in place for backwards compatibility, and easy of
testing, but all documentation is dropped to encourage use of the new
tool instead.

5 years agonetworkd: add FDB support
Alin Rauta [Wed, 17 Dec 2014 15:35:36 +0000 (07:35 -0800)]
networkd: add FDB support

5 years agojournal: next_with_matches() now does not need a mapped object as input
Michal Schmidt [Wed, 17 Dec 2014 14:46:30 +0000 (15:46 +0100)]
journal: next_with_matches() now does not need a mapped object as input

Now that journal_file_next_entry() does not need a pointer to the
current object, next_with_matches() does not need it either.

5 years agojournal: journal_file_next_entry() does not need pointer to current Object
Michal Schmidt [Wed, 17 Dec 2014 14:45:10 +0000 (15:45 +0100)]
journal: journal_file_next_entry() does not need pointer to current Object

The current offset is sufficient information.

5 years agojournal: optimize iteration by returning previously found candidate entry
Michal Schmidt [Wed, 17 Dec 2014 14:07:25 +0000 (15:07 +0100)]
journal: optimize iteration by returning previously found candidate entry

In next_beyond_location() when the JournalFile's location type is
LOCATION_SEEK, it means there's nothing to do, because we already have
the location of the candidate entry. Do an early return. Note that now
next_beyond_location() does not anymore guarantee on return that the
entry is mapped, but previous patches made sure the caller does not
care.

This optimization is at least as good as "journal: optimize iteration:
skip files that cannot improve current candidate entry" was.

Timing results on my workstation, using:
$ time ./journalctl -q --since=2014-06-01 --until=2014-07-01 > /dev/null

Before "Revert "journal: optimize iteration: skip files that cannot
improve current candidate entry":

real    0m5.349s
user    0m5.166s
sys     0m0.181s

Now:

real    0m3.901s
user    0m3.724s
sys     0m0.176s

5 years agojournal: optimize iteration by skipping exhausted files
Michal Schmidt [Thu, 18 Dec 2014 13:21:55 +0000 (14:21 +0100)]
journal: optimize iteration by skipping exhausted files

If from a previous iteration we know we are at the end of a journal
file, don't bother looking into the file again. This is complicated by
the fact that the EOF does not have to be permanent (think of
"journalctl -f"). So we also check if the number of entries in the
journal file changed.

This optimization has a similar effect as "journal: optimize iteration:
skip whole files behind current location" had.

5 years agojournal: drop unnecessary parameters of next_beyond_location()
Michal Schmidt [Wed, 17 Dec 2014 14:37:52 +0000 (15:37 +0100)]
journal: drop unnecessary parameters of next_beyond_location()

offset is redundant, because the caller can rely on f->current_offset.
The object pointer the function saves in *ret is thrown away by the caller.

5 years agojournal: remove redundant variable new_offset
Michal Schmidt [Thu, 18 Dec 2014 08:56:04 +0000 (09:56 +0100)]
journal: remove redundant variable new_offset

The file's current_offset is already updated at this point, so let's use
it.

5 years agojournal: compare candidate entries using JournalFiles' locations
Michal Schmidt [Wed, 17 Dec 2014 13:06:28 +0000 (14:06 +0100)]
journal: compare candidate entries using JournalFiles' locations

When comparing the locations of candidate entries, we can rely on the
location information stored in struct JournalFile.

5 years agojournal: simplify set_location()
Michal Schmidt [Tue, 16 Dec 2014 21:38:09 +0000 (22:38 +0100)]
journal: simplify set_location()

set_location() is called from real_journal_next() when a winning entry
has been picked from among the candidates in journal files.

The location type is always set to LOCATION_DISCRETE. No need to pass
it as a parameter.
The per-JournalFile location information is already updated at this
point. No need for having the direction and offset here.

5 years agojournal: keep per-JournalFile location info during iteration
Michal Schmidt [Tue, 16 Dec 2014 20:03:36 +0000 (21:03 +0100)]
journal: keep per-JournalFile location info during iteration

In next_beyond_location() when we find a candidate entry in a journal
file, save its location information in struct JournalFile.

The purpose of remembering the locations of candidate entries is to be
able to save work in the next iteration. This patch does only the
remembering part.

LOCATION_SEEK means the location identifies a candidate entry.
When a winner is picked from among candidates, it becomes
LOCATION_DISCRETE.
LOCATION_TAIL here signifies we've iterated the file to the end (or the
beginning in the case of reversed direction).

5 years agojournal: abstract the resetting of JournalFile's location
Michal Schmidt [Tue, 16 Dec 2014 19:54:56 +0000 (20:54 +0100)]
journal: abstract the resetting of JournalFile's location

5 years agojournal: move definition of LocationType to journal-file.h
Michal Schmidt [Tue, 16 Dec 2014 19:51:58 +0000 (20:51 +0100)]
journal: move definition of LocationType to journal-file.h

In preparation for individual JournalFiles maintaining a location
of their own.

5 years agoRevert "journal: optimize iteration: skip whole files behind current location"
Michal Schmidt [Tue, 16 Dec 2014 19:32:41 +0000 (20:32 +0100)]
Revert "journal: optimize iteration: skip whole files behind current location"

This reverts commit b7c88ab8cc7d55a43450bf3dea750f95f2e910d6.

This optimization will be made redundant by the following patches.

5 years agoRevert "journal: optimize iteration: skip files that cannot improve current candidate...
Michal Schmidt [Tue, 16 Dec 2014 19:32:34 +0000 (20:32 +0100)]
Revert "journal: optimize iteration: skip files that cannot improve current candidate entry"

This reverts commit f8b5a3b75fb55f0acb85c21424b3893c822742e9.

This optimization will be made redundant by the following patches.

5 years agojournal: delete unused function journal_file_skip_entry()
Michal Schmidt [Tue, 16 Dec 2014 14:47:01 +0000 (15:47 +0100)]
journal: delete unused function journal_file_skip_entry()

Its only caller is a test.

5 years agojournal: delete unused function journal_file_move_to_entry_by_offset()
Michal Schmidt [Tue, 16 Dec 2014 18:06:30 +0000 (19:06 +0100)]
journal: delete unused function journal_file_move_to_entry_by_offset()

5 years agocore: use raw_clone instead of fork in signal handler
Zbigniew Jędrzejewski-Szmek [Wed, 17 Dec 2014 04:53:23 +0000 (23:53 -0500)]
core: use raw_clone instead of fork in signal handler

fork() is not async-signal-safe and calling it from the signal handler
could result in a deadlock when at_fork() handlers are called. Using
the raw clone() syscall sidesteps that problem.

The tricky part is that raise() does not work, since getpid() does not
work. Add raw_getpid() to get the real pid, and use kill() instead of
raise().

https://bugs.freedesktop.org/show_bug.cgi?id=86604

5 years agotest-unit-file: add test for semicolon escaping
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 05:52:28 +0000 (00:52 -0500)]
test-unit-file: add test for semicolon escaping

https://bugs.freedesktop.org/show_bug.cgi?id=87393

5 years agoload-fragment: properly unescape \;
tomsod-m ya ru [Thu, 18 Dec 2014 04:01:06 +0000 (23:01 -0500)]
load-fragment: properly unescape \;

https://bugs.freedesktop.org/show_bug.cgi?id=87393

5 years agopath: make the check for unsupported name_to_handle_at symmetric
Zbigniew Jędrzejewski-Szmek [Thu, 18 Dec 2014 02:05:58 +0000 (21:05 -0500)]
path: make the check for unsupported name_to_handle_at symmetric

If child supports, but the parent does not, or when the child does
not support, but the parent does, assume the child is a mount point.

Only if neither supports use the fallback.

5 years agopath: follow symbolic link for parent path (2)
Umut Tezduyar Lindskog [Thu, 18 Dec 2014 00:43:58 +0000 (19:43 -0500)]
path: follow symbolic link for parent path (2)

c0e57ba9e22ee937722958d8b912ade2a37f206d fixed the fallback path.
We should do the same for name_to_handle_at().

5 years agomachinectl: add new commands for copying files from/to containers
Lennart Poettering [Thu, 18 Dec 2014 00:35:58 +0000 (01:35 +0100)]
machinectl: add new commands for copying files from/to containers

5 years agoutil: in make_stdio() use dup2() rather than dup3()
Lennart Poettering [Wed, 17 Dec 2014 20:54:00 +0000 (21:54 +0100)]
util: in make_stdio() use dup2() rather than dup3()

dup3() allows setting O_CLOEXEC which we are not interested in. However,
it also fails if called with the same fd as input and output, which is
something we don't want. Hence use dup2().

Also, we need to explicitly turn off O_CLOEXEC for the fds, in case the
input fd was O_CLOEXEC and < 3.

5 years agoupdate TODO
Lennart Poettering [Wed, 17 Dec 2014 20:53:54 +0000 (21:53 +0100)]
update TODO

5 years agomachinectl: implement "bind" command to create additional bind mounts from host to...
Lennart Poettering [Wed, 17 Dec 2014 20:51:45 +0000 (21:51 +0100)]
machinectl: implement "bind" command to create additional bind mounts from host to container during runtime