chiark / gitweb /
Kay Sievers [Mon, 1 Jul 2013 21:55:34 +0000 (23:55 +0200)]
keymap: remove non-existing driver string matches
There are no such strings for input devices in the kernel.
Kay Sievers [Mon, 1 Jul 2013 20:34:43 +0000 (22:34 +0200)]
hwdb: add --device=<device-id> and --filter=<key name glob>
Kay Sievers [Mon, 1 Jul 2013 17:41:19 +0000 (19:41 +0200)]
hwdb: import data
Kay Sievers [Mon, 1 Jul 2013 17:39:50 +0000 (19:39 +0200)]
hwdb: remove support for (not fully implemented) conditional properties
Lennart Poettering [Mon, 1 Jul 2013 01:02:42 +0000 (03:02 +0200)]
core: add support to run transient units in arbitrary slices
Lennart Poettering [Mon, 1 Jul 2013 00:52:17 +0000 (02:52 +0200)]
core: parse Slice= from the unit type specific unit file section
Since not all unit types know Slice= it belongs in the unit type
specific unit file section.
Lennart Poettering [Mon, 1 Jul 2013 00:47:11 +0000 (02:47 +0200)]
scope: properly implement passive validity checking of PIDs field when creating transient scopes
Lennart Poettering [Mon, 1 Jul 2013 00:45:02 +0000 (02:45 +0200)]
core: move ControlGroup and Slice properties out of the dbus "Unit" interface
Slice/ControlGroup only really makes sense for unit types which actually
have cgroups attached to them, hence move them out of the generic Unit
interface and into the specific unit type interfaces.
These fields will continue to be part of Unit though, simply because
things are a log easier that way. However, regardless how this looks
internally we should keep things clean and independent of the specific
implementation of the inside.
Lennart Poettering [Sun, 30 Jun 2013 22:40:56 +0000 (00:40 +0200)]
core: allow setting of the description string for transient units
Lennart Poettering [Sun, 30 Jun 2013 22:03:57 +0000 (00:03 +0200)]
core: add new "scope" unit type for making a unit of pre-existing processes
"Scope" units are very much like service units, however with the
difference that they are created from pre-existing processes, rather
than processes that systemd itself forks off. This means they are
generated programmatically via the bus API as transient units rather
than from static configuration read from disk. Also, they do not provide
execution-time parameters, as at the time systemd adds the processes to
the scope unit they already exist and the parameters cannot be applied
anymore.
The primary benefit of this new unit type is to create arbitrary cgroups
for worker-processes forked off an existing service.
This commit also adds a a new mode to "systemd-run" to run the specified
processes in a scope rather then a transient service.
Lennart Poettering [Sun, 30 Jun 2013 22:02:41 +0000 (00:02 +0200)]
systemctl: show per-unit cgroup tree correctly following the new property
Lennart Poettering [Sun, 30 Jun 2013 22:02:03 +0000 (00:02 +0200)]
snapshot: snapshots are just a special kind of transient units now
Lennart Poettering [Sun, 30 Jun 2013 22:01:29 +0000 (00:01 +0200)]
dbus: expose cgroup properties in introspection everywhere
Lennart Poettering [Sun, 30 Jun 2013 21:56:59 +0000 (23:56 +0200)]
service: correct service bus introspection for timeouts
Lennart Poettering [Sun, 30 Jun 2013 21:56:11 +0000 (23:56 +0200)]
cgroup: implicitly add units to GC queue when their cgroups run empty
Lennart Poettering [Sun, 30 Jun 2013 21:55:36 +0000 (23:55 +0200)]
cgroup: readd proper cgroup empty tracking
Jason St. John [Thu, 27 Jun 2013 16:38:15 +0000 (18:38 +0200)]
man: fix spacing issue in various man pages
Before: libsystemd-daemonpkg-config(1)
After: libsystemd-daemon pkg-config(1)
This fix is more complicated than it should be due to the consecutive
XML elements separated by collapsible whitespace.
Merging the lines and separating the XML elements with an en space or a
non-breaking space is the only solution that results in one, and only
one, space being inserted between them when testing. An em space results
in two spaces being inserted.
Zbigniew Jędrzejewski-Szmek [Sat, 29 Jun 2013 02:29:21 +0000 (22:29 -0400)]
build-sys: "link" libsystemd-id128.so with libsystemd-label
Fixed build on debian wheezy:
./.libs/libudev.so: undefined reference to `cg_create'
Appears to have no influence on the resulting binaries and
libraries. Cf.
b5fafdf63f.
Lennart Poettering [Fri, 28 Jun 2013 16:37:15 +0000 (18:37 +0200)]
main: fix loading of default target
장동준 [Fri, 28 Jun 2013 08:51:28 +0000 (10:51 +0200)]
keymap: Add Samsung Series 3
Signed-off-by: Martin Pitt <martinpitt@gnome.org>
Lennart Poettering [Fri, 28 Jun 2013 02:12:58 +0000 (04:12 +0200)]
core: add transient units
Transient units can be created via the bus API. They are configured via
the method call parameters rather than on-disk files. They are subject
to normal GC. Transient units currently may only be created for
services (however, we will extend this), and currently only ExecStart=
and the cgroup parameters can be configured (also to be extended).
Transient units require a unique name, that previously had no
configuration file on disk.
A tool systemd-run is added that makes use of this functionality to run
arbitrary command lines as transient services:
$ systemd-run /bin/ping www.heise.de
Will cause systemd to create a new transient service and run ping in it.
Lennart Poettering [Fri, 28 Jun 2013 01:47:40 +0000 (03:47 +0200)]
dbus: fix minor memory leak when sending job change signals
Lennart Poettering [Thu, 27 Jun 2013 22:41:24 +0000 (00:41 +0200)]
unit: make sure the dropins we write are high-priority
Lennart Poettering [Thu, 27 Jun 2013 21:21:21 +0000 (23:21 +0200)]
dbus: hookup runtime property changes for mouns, services, sockets, swaps too
Lennart Poettering [Thu, 27 Jun 2013 21:13:17 +0000 (23:13 +0200)]
dbus: make DeviceAllow=/DevicePolicy= writable
Lennart Poettering [Thu, 27 Jun 2013 19:50:35 +0000 (21:50 +0200)]
dbus: make more cgroup attributes runtime settable
Lennart Poettering [Thu, 27 Jun 2013 19:14:56 +0000 (21:14 +0200)]
dbus: add infrastructure for changing multiple properties at once on units and hook some cgroup attributes up to it
This introduces two bus calls to make runtime changes to selected bus
properties, optionally with persistence.
This currently hooks this up only for three cgroup atributes, but this
brings the infrastructure to add more changable attributes.
This allows setting multiple attributes at once, and takes an array
rather than a dictionary of properties, in order to implement simple
resetting of lists using the same approach as when they are sourced from
unit files. This means, that list properties are appended to by this
call, unless they are first reset via assigning the empty list.
Jan Janssen [Thu, 27 Jun 2013 14:24:02 +0000 (16:24 +0200)]
journal-verify: Use proper printf placeholder
Jan Janssen [Wed, 26 Jun 2013 11:43:16 +0000 (13:43 +0200)]
test: Add list testcase
Lennart Poettering [Thu, 27 Jun 2013 02:14:27 +0000 (04:14 +0200)]
core: general cgroup rework
Replace the very generic cgroup hookup with a much simpler one. With
this change only the high-level cgroup settings remain, the ability to
set arbitrary cgroup attributes is removed, so is support for adding
units to arbitrary cgroup controllers or setting arbitrary paths for
them (especially paths that are different for the various controllers).
This also introduces a new -.slice root slice, that is the parent of
system.slice and friends. This enables easy admin configuration of
root-level cgrouo properties.
This replaces DeviceDeny= by DevicePolicy=, and implicitly adds in
/dev/null, /dev/zero and friends if DeviceAllow= is used (unless this is
turned off by DevicePolicy=).
Lennart Poettering [Thu, 27 Jun 2013 00:28:12 +0000 (02:28 +0200)]
core: don't do runaway fork()s if we hit a segfault from our segfault handler
Zbigniew Jędrzejewski-Szmek [Wed, 26 Jun 2013 23:48:32 +0000 (19:48 -0400)]
killall: do not use alloca() in argument list
It is not allowed.
Zbigniew Jędrzejewski-Szmek [Wed, 26 Jun 2013 23:48:24 +0000 (19:48 -0400)]
journalctl: highlight MESSAGE= in verbose output
When looking at verbose output, additional "work" is required to
pick out the interesting MESSAGE= lines from all the fields.
Also, show long fields in full in verbose output mode when
OUTPUT_FULL_WIDTH is specified.
Zbigniew Jędrzejewski-Szmek [Wed, 26 Jun 2013 23:48:19 +0000 (19:48 -0400)]
man: add CONSTANTS section to systemd.directives(7)
Zbigniew Jędrzejewski-Szmek [Wed, 26 Jun 2013 23:47:34 +0000 (19:47 -0400)]
man: use <constant> for various constants which look ugly with quotes
Zbigniew Jędrzejewski-Szmek [Wed, 26 Jun 2013 12:03:53 +0000 (08:03 -0400)]
man: always supply quotes around literals
When manpages are displayed on a terminal, <literal>s are indistinguishable
from surrounding text. Add quotes everywhere, remove duplicate quotes,
and tweak a few lists for consistent formatting.
https://bugzilla.redhat.com/show_bug.cgi?id=874631
Harald Hoyer [Tue, 25 Jun 2013 14:25:38 +0000 (16:25 +0200)]
systemctl: conform to LSB with the "status" return code
check for pid file existance before returning 1
David King [Tue, 25 Jun 2013 10:48:17 +0000 (11:48 +0100)]
man: Small language improvements to sd_journal_open
Harald Hoyer [Tue, 25 Jun 2013 11:06:51 +0000 (13:06 +0200)]
systemctl: conform to LSB with the "status" return code
http://fedoraproject.org/wiki/Packaging:SysVInitScript#Exit_Codes_for_the_Status_Action
https://bugzilla.redhat.com/show_bug.cgi?id=975016
Zbigniew Jędrzejewski-Szmek [Tue, 25 Jun 2013 01:00:28 +0000 (21:00 -0400)]
journald: always vacuum empty offline files
Corrupted empty files are relatively common. I think they are created
when a coredump for a user who never logged anything before is
attempted to be written, but the write does not succeed because the
coredump is too big, but there are probably other ways to create
those, especially if the machine crashes at the right time.
Non-corrupted empty files can also happen, e.g. if a journal file is
opened, but nothing is ever successfully written to it and it is
rotated because of MaxFileSec=. Either way, each "empty" journal file
costs around 3 MB, and there's little point in keeping them around.
Zbigniew Jędrzejewski-Szmek [Tue, 25 Jun 2013 01:02:16 +0000 (21:02 -0400)]
journald: fix space limits reporting
Reporting of the free space was bogus, since the remaining space
was compared with the maximum allowed, instead of the current
use being compared with the maximum allowed. Simplify and fix
by reporting limits directly at the point where they are calculated.
Also, assign a UUID to the message.
Zbigniew Jędrzejewski-Szmek [Mon, 24 Jun 2013 11:59:41 +0000 (07:59 -0400)]
journal/vacuum: cleanup
Jan Janssen [Sat, 22 Jun 2013 13:34:01 +0000 (15:34 +0200)]
libudev: Use correct type for sizeof
Zbigniew Jędrzejewski-Szmek [Sun, 23 Jun 2013 00:04:08 +0000 (20:04 -0400)]
journal-verify: allow unlinked data entries
Sometimes an entry is not successfully written, and we end up with
data items which are "unlinked", not connected to, and not used by any
entry. This will usually happen when we write to write a core dump,
and the initial small data fields are written successfully, but
the huge COREDUMP= field is not written. This situation is hard
to avoid, but the results are mostly harmless. Thus only warn about
unused data items.
Also, be more verbose about why journal files failed verification.
This should help diagnose journal failure modes without resorting
to a hexadecimal editor.
https://bugs.freedesktop.org/show_bug.cgi?id=65235 (esp. see
system.journal attached to the bug report).
Michael Biebl [Fri, 21 Jun 2013 22:55:18 +0000 (00:55 +0200)]
man: Fix small typo
Lennart Poettering [Fri, 21 Jun 2013 13:57:47 +0000 (15:57 +0200)]
update TODO
Lennart Poettering [Fri, 21 Jun 2013 13:57:42 +0000 (15:57 +0200)]
tmpfiles: fix error check
Lennart Poettering [Fri, 21 Jun 2013 13:56:45 +0000 (15:56 +0200)]
journald: bump the journal per-unit ratelimit defaults
Too many people kept hitting them, so let's increase the limits a bit.
https://bugzilla.redhat.com/show_bug.cgi?id=965803
Martin Pitt [Fri, 21 Jun 2013 04:34:38 +0000 (06:34 +0200)]
keymap: Apply to all Latitude and Precision models
https://launchpad.net/bugs/
1193147
Zbigniew Jędrzejewski-Szmek [Fri, 21 Jun 2013 02:25:49 +0000 (22:25 -0400)]
journalctl,systemctl: show messages tagged with OBJECT_SYSTEMD_[USER_]_UNIT
Replace mallocs with alloca while at it.
Zbigniew Jędrzejewski-Szmek [Fri, 21 Jun 2013 01:48:26 +0000 (21:48 -0400)]
Make sure we only show authentic coredump messages
Before we only checked the MESSAGE_ID and COREDUMP_UNIT.
Those are both user-controlled fields.
For COREDUMP_USER_UNIT, relax the rules a bit, and also
allow messages from _UID=0.
Zbigniew Jędrzejewski-Szmek [Wed, 12 Jun 2013 04:24:34 +0000 (00:24 -0400)]
journal: allow callers to specify OBJECT_PID=
When journald encounters a message with OBJECT_PID= set
coming from a priviledged process (UID==0), additional fields
will be added to the message:
OBJECT_UID=,
OBJECT_GID=,
OBJECT_COMM=,
OBJECT_EXE=,
OBJECT_CMDLINE=,
OBJECT_AUDIT_SESSION=,
OBJECT_AUDIT_LOGINUID=,
OBJECT_SYSTEMD_CGROUP=,
OBJECT_SYSTEMD_SESSION=,
OBJECT_SYSTEMD_OWNER_UID=,
OBJECT_SYSTEMD_UNIT= or OBJECT_SYSTEMD_USER_UNIT=.
This is for other logging daemons, like setroubleshoot, to be able to
augment their logs with data about the process.
https://bugzilla.redhat.com/show_bug.cgi?id=951627
Zbigniew Jędrzejewski-Szmek [Fri, 21 Jun 2013 02:40:10 +0000 (22:40 -0400)]
Add hasprefix macro to check prefixes of fixed length
Lennart Poettering [Thu, 20 Jun 2013 23:46:27 +0000 (01:46 +0200)]
login: add an api to determine the slice a PID is located in to libsystemd-login
Auke Kok [Thu, 20 Jun 2013 20:36:33 +0000 (13:36 -0700)]
core: only attempt to connect to a session bus if one likely exists.
Lennart Poettering [Thu, 20 Jun 2013 21:32:13 +0000 (23:32 +0200)]
update TODO
Lennart Poettering [Thu, 20 Jun 2013 21:31:21 +0000 (23:31 +0200)]
man: fix example for DefaultEnvironment=
The example mentioned Environment= rather than DefaultEnvironment=.
Also made some other clean-ups.
Lennart Poettering [Thu, 20 Jun 2013 21:25:23 +0000 (23:25 +0200)]
man: document that shutdown() is only sometimes OK on sockets passed in via socket activation
Umut Tezduyar [Sun, 9 Jun 2013 05:08:46 +0000 (07:08 +0200)]
manager: add DefaultEnvironment option
This complements existing functionality of setting variables
through 'systemctl set-environment', the kernel command line,
and through normal environment variables for systemd in session
mode.
Zbigniew Jędrzejewski-Szmek [Thu, 20 Jun 2013 20:15:04 +0000 (16:15 -0400)]
journalctl: show lines in full with --all
In
31f7bf1 "logs-show: print multiline messages", I forgot
to take into account the fact that --all implies --full for
journalctl.
Harald Hoyer [Thu, 20 Jun 2013 13:46:37 +0000 (15:46 +0200)]
kernel-install: filter out "initrd=" from /proc/cmdline
Zbigniew Jędrzejewski-Szmek [Thu, 20 Jun 2013 04:51:10 +0000 (00:51 -0400)]
logind: uninitialized variable
Zbigniew Jędrzejewski-Szmek [Thu, 20 Jun 2013 04:20:55 +0000 (00:20 -0400)]
build-sys: make intltool and distcheck happy
Zbigniew Jędrzejewski-Szmek [Thu, 20 Jun 2013 04:01:15 +0000 (00:01 -0400)]
nspawn: '-C' option has been removed
Fixup for
9444b1f "logind: add infrastructure to keep track of
machines, and move to slices."
Lennart Poettering [Thu, 20 Jun 2013 02:03:38 +0000 (04:03 +0200)]
sd-login: update machine enumeration/notifications APIs for new /run/systemd/machines/
Lennart Poettering [Thu, 20 Jun 2013 02:00:28 +0000 (04:00 +0200)]
logind: make ListMachines bus call public
Lennart Poettering [Thu, 20 Jun 2013 01:45:08 +0000 (03:45 +0200)]
logind: add infrastructure to keep track of machines, and move to slices
- This changes all logind cgroup objects to use slice objects rather
than fixed croup locations.
- logind can now collect minimal information about running
VMs/containers. As fixed cgroup locations can no longer be used we
need an entity that keeps track of machine cgroups in whatever slice
they might be located. Since logind already keeps track of users,
sessions and seats this is a trivial addition.
- nspawn will now register with logind and pass various bits of metadata
along. A new option "--slice=" has been added to place the container
in a specific slice.
- loginctl gained commands to list, introspect and terminate machines.
- user.slice and machine.slice will now be pulled in by logind.service,
since only logind.service requires this slice.
Łukasz Stelmach [Wed, 19 Jun 2013 12:59:02 +0000 (14:59 +0200)]
Describe handling of an AF_UNIX socket
Describe how to handle an AF_UNIX socket, with Accept set to false,
received from systemd, upon exit.
Signed-off-by: Łukasz Stelmach <l.stelmach@samsung.com>
Zbigniew Jędrzejewski-Szmek [Wed, 19 Jun 2013 22:09:54 +0000 (18:09 -0400)]
core/dbus: properly export cgroup properties on sockets
https://bugs.freedesktop.org/show_bug.cgi?id=65945
Zbigniew Jędrzejewski-Szmek [Wed, 19 Jun 2013 22:08:14 +0000 (18:08 -0400)]
cgls,loginctl,systemctl: fix -l
Fixup for
98a6e13 "journalctl,loginctl,systemctl,systemd-cgls: add -l
as alias for --full".
Lennart Poettering [Tue, 18 Jun 2013 14:25:11 +0000 (16:25 +0200)]
journal: add references to SSKG paper FSS is based on
Lennart Poettering [Tue, 18 Jun 2013 14:06:05 +0000 (16:06 +0200)]
update TODO
Zbigniew Jędrzejewski-Szmek [Tue, 18 Jun 2013 12:48:14 +0000 (08:48 -0400)]
journalctl: properly print headers of empty journals
Sean McGovern [Wed, 12 Jun 2013 13:32:47 +0000 (09:32 -0400)]
udev: handle network controllers in nonstandard domains
Onboard network controllers are not always on PCI domain 0.
[Kay: use int instead of long, add [P] to slot naming, remove sysname var]
Lennart Poettering [Tue, 18 Jun 2013 00:07:35 +0000 (02:07 +0200)]
core: unref slice ref after use
Lennart Poettering [Mon, 17 Jun 2013 23:00:13 +0000 (01:00 +0200)]
core: expose a "Slice" property on Unit objects on the bus
Lennart Poettering [Mon, 17 Jun 2013 22:59:54 +0000 (00:59 +0200)]
units: add default units for system.slice, user.slice, machine.slice
Lennart Poettering [Mon, 17 Jun 2013 19:33:26 +0000 (21:33 +0200)]
core: add new .slice unit type for partitioning systems
In order to prepare for the kernel cgroup rework, let's introduce a new
unit type to systemd, the "slice". Slices can be arranged in a tree and
are useful to partition resources freely and hierarchally by the user.
Each service unit can now be assigned to one of these slices, and later
on login users and machines may too.
Slices translate pretty directly to the cgroup hierarchy, and the
various objects can be assigned to any of the slices in the tree.
Ross Lagerwall [Thu, 13 Jun 2013 09:45:12 +0000 (10:45 +0100)]
rules: only run systemd-sysctl when a network device is added
Otherwise, when a network device is renamed, systemd-sysctl is run twice
with the same network device name: once for ACTION="add" and once for
ACTION="move".
Lennart Poettering [Mon, 17 Jun 2013 19:12:53 +0000 (21:12 +0200)]
mount: when learning about the root mount from mountinfo, don't add conflicting dep for umount.target
That way systemd won't try to umount it at shutdown.
Lennart Poettering [Mon, 17 Jun 2013 16:37:07 +0000 (18:37 +0200)]
rpm: define a %_userunitdir macro
Daniel Albers [Mon, 17 Jun 2013 09:36:35 +0000 (11:36 +0200)]
journalctl,loginctl,systemctl,systemd-cgls: add -l as alias for --full
https://bugs.freedesktop.org/show_bug.cgi?id=65850
Zbigniew Jędrzejewski-Szmek [Fri, 14 Jun 2013 12:53:15 +0000 (08:53 -0400)]
man: update systemd-analyze invocation
Gabriel de Perthuis [Tue, 11 Jun 2013 20:29:32 +0000 (22:29 +0200)]
systemd-analyze: Show critical chains for listed units
Jason St. John [Wed, 12 Jun 2013 17:45:14 +0000 (19:45 +0200)]
man: improve readability of "_TRANSPORT=" section in systemd.journal-fields(7)
The list and descriptions of valid transports was difficult to read, so
break the long sentence up into discrete man page list items to improve
readability.
Zbigniew Jędrzejewski-Szmek [Fri, 14 Jun 2013 03:32:15 +0000 (23:32 -0400)]
man: add sd_j_open_files to return values section
Zbigniew Jędrzejewski-Szmek [Fri, 14 Jun 2013 03:32:14 +0000 (23:32 -0400)]
journald: do not calculate free space too early
Since the system journal wasn't open yet, available_space() returned 0.
Before:
systemd-journal[22170]: Allowing system journal files to grow to 4.0G.
systemd-journal[22170]: Journal size currently limited to 0B due to SystemKeepFree.
After:
systemd-journal[22178]: Allowing system journal files to grow to 4.0G.
systemd-journal[22178]: Journal size currently limited to 3.0G due to SystemKeepFree.
Also, when failing to write a message, show how much space was needed:
"Failed to write entry (26 items,
260123456 bytes) despite vacuuming, ignoring: ...".
Zbigniew Jędrzejewski-Szmek [Fri, 14 Jun 2013 03:32:14 +0000 (23:32 -0400)]
journal: use initialization instead of zeroing
Zbigniew Jędrzejewski-Szmek [Wed, 12 Jun 2013 17:40:54 +0000 (13:40 -0400)]
build-sys: add 'man' target
Useful when working just on the documentation.
Jan Janssen [Wed, 12 Jun 2013 11:05:15 +0000 (13:05 +0200)]
gitignore: Add test-journal-interleaving
Ross Burton [Tue, 11 Jun 2013 16:16:37 +0000 (17:16 +0100)]
build-sys: don't install quotaon.service twice
quotaon.service is already installed through dist_systemunit_DATA, so it doesn't
need to be added to nodist_systemunit_DATA. Installing the same file twice
results in a race condition where the install process can fail.
https://bugs.freedesktop.org/show_bug.cgi?id=65659
[zj: actually remove quotaon.service from the other list.]
Jason St. John [Tue, 11 Jun 2013 16:26:03 +0000 (18:26 +0200)]
bootchart: fix typos in copyright notices
"Corporation" was misspelled as "Coproration"
Lennart Poettering [Tue, 11 Jun 2013 06:55:42 +0000 (08:55 +0200)]
update TODO
Zbigniew Jędrzejewski-Szmek [Mon, 10 Jun 2013 22:10:12 +0000 (18:10 -0400)]
journald: do not overwrite syslog facility when parsing priority
https://bugs.freedesktop.org/show_bug.cgi?id=65610
Zbigniew Jędrzejewski-Szmek [Mon, 10 Jun 2013 19:08:02 +0000 (15:08 -0400)]
build-sys: remove SD_JOURNAL_SYSTEM_ONLY(3) from Makefile
Zbigniew Jędrzejewski-Szmek [Mon, 10 Jun 2013 14:22:08 +0000 (10:22 -0400)]
man: add note that sd_journal_open_files is racy
Zbigniew Jędrzejewski-Szmek [Sat, 8 Jun 2013 02:01:03 +0000 (22:01 -0400)]
journal: letting (interleaved) seqnums go
In the following scenario:
server creates system.journal
server creates user-1000.journal
both journals share the same seqnum_id.
Then
server writes to user-1000.journal first,
and server writes to system.journal a bit later,
and everything is fine.
The server then terminates (crash, reboot, rsyslog testing,
whatever), and user-1000.journal has entries which end with
a lower seqnum than system.journal. Now
server is restarted
server opens user-1000.journal and writes entries to it...
BAM! duplicate seqnums for the same seqnum_id.
Now, we usually don't see that happen, because system.journal
is closed last, and opened first. Since usually at least one
message is written during boot and lands in the system.journal,
the seqnum is initialized from it, and is set to a number higher
than than anything found in user journals. Nevertheless, if
system.journal is corrupted and is rotated, it can happen that
an entry is written to the user journal with a seqnum that is
a duplicate with an entry found in the corrupted system.journal~.
When browsing the journal, journalctl can fall into a loop
where it tries to follow the seqnums, and tries to go the
next location by seqnum, and is transported back in time to
to the older duplicate seqnum. There is not way to find
out the maximum seqnum used in a multiple files, without
actually looking at all of them. But we don't want to do
that because it would be slow, and actually it isn't really
possible, because a file might e.g. be temporarily unaccessible.
Fix the problem by using different seqnum series for user
journals. Using the same seqnum series for rotated journals
is still fine, because we know that nothing will write
to the rotated journal anymore.
Likely related:
https://bugs.freedesktop.org/show_bug.cgi?id=64566
https://bugs.freedesktop.org/show_bug.cgi?id=59856
https://bugs.freedesktop.org/show_bug.cgi?id=64296
https://bugs.archlinux.org/task/35581
https://bugzilla.novell.com/show_bug.cgi?id=817778
Possibly related:
https://bugs.freedesktop.org/show_bug.cgi?id=64293
Zbigniew Jędrzejewski-Szmek [Sat, 8 Jun 2013 01:54:20 +0000 (21:54 -0400)]
tests: add testcase for duplicate seqnums
Marius Vollmer [Fri, 7 Jun 2013 04:50:21 +0000 (00:50 -0400)]
tests: add testcase for skipping-entries-on-direction-change-bug
This test case failed until
a3e6f050de8.
Taken from https://bugs.freedesktop.org/show_bug.cgi?id=65255.
Zbigniew Jędrzejewski-Szmek [Fri, 7 Jun 2013 03:30:46 +0000 (23:30 -0400)]
journal: change direction tests to use the same convention (cp </> np)
The order was different in various places, which makes it harder to
read to code. Also consistently use ternany for all direction checks.
Remove one free(NULL).