chiark / gitweb /
elogind.git
7 years agodocs: gtk-doc warnings are annoying
Kay Sievers [Sat, 3 Nov 2012 21:04:06 +0000 (22:04 +0100)]
docs: gtk-doc warnings are annoying

7 years agovconsole: remove Frugalware legacy file support
Miklos Vajna [Sat, 3 Nov 2012 15:23:00 +0000 (16:23 +0100)]
vconsole: remove Frugalware legacy file support

7 years agoutil : fallback to plain ASCII drawing if locale is not UTF-8
Michal Schmidt [Fri, 2 Nov 2012 16:35:30 +0000 (17:35 +0100)]
util : fallback to plain ASCII drawing if locale is not UTF-8

When printing cgroup and sysfs hierarchies, avoid using UTF-8 box drawing
characters if the locale is not UTF-8.

https://bugzilla.redhat.com/show_bug.cgi?id=871153

7 years agoutil: add is_locale_utf8()
Michal Schmidt [Fri, 2 Nov 2012 16:27:15 +0000 (17:27 +0100)]
util: add is_locale_utf8()

journalctl and vconsole-setup both implement utf8 locale detection.
Let's have a common function for it.
The next patch will add another use.

7 years agoanalyze: use GDBus instead of dbus-python
Cosimo Cecchi [Fri, 2 Nov 2012 15:29:02 +0000 (11:29 -0400)]
analyze: use GDBus instead of dbus-python

7 years agotimedatectl: show "DST active: n/a" if no DST data is available
Kay Sievers [Fri, 2 Nov 2012 15:45:02 +0000 (05:45 -1000)]
timedatectl: show "DST active: n/a" if no DST data is available

7 years agoTODO: update
Kay Sievers [Fri, 2 Nov 2012 14:18:29 +0000 (15:18 +0100)]
TODO: update

7 years agotimedatectl: explain everything nobody wants to know about DST
Kay Sievers [Fri, 2 Nov 2012 02:45:07 +0000 (03:45 +0100)]
timedatectl: explain everything nobody wants to know about DST

7 years agoshared: add timer_get_dst()
Kay Sievers [Thu, 1 Nov 2012 23:40:02 +0000 (00:40 +0100)]
shared: add timer_get_dst()

7 years agoTODO: update
Kay Sievers [Thu, 1 Nov 2012 23:34:07 +0000 (00:34 +0100)]
TODO: update

7 years agoudev: break over-long lines
Kay Sievers [Wed, 31 Oct 2012 00:55:51 +0000 (01:55 +0100)]
udev: break over-long lines

7 years agoman: typo fixes
Thomas Hindoe Paaboel Andersen [Thu, 1 Nov 2012 22:35:34 +0000 (23:35 +0100)]
man: typo fixes

Fixes a few more typos. Also changes a "Accept=no" to
"Accept=false" to be consistent with the previous examples
in the same man page.

7 years agoupdate TODO
Lennart Poettering [Wed, 31 Oct 2012 01:57:26 +0000 (02:57 +0100)]
update TODO

7 years agoservice: drop special HTTP server target, as it is a bad idea anf Fedora specific
Lennart Poettering [Wed, 31 Oct 2012 01:55:04 +0000 (02:55 +0100)]
service: drop special HTTP server target, as it is a bad idea anf Fedora specific

This was premarily intended to support the LSB facility $httpd which is
only known by Fedora, and a bad idea since it lacks any real-life
usecase.

Similar, drop support for some other old Fedora-specific facilities.

Also, document the rules for introduction of new facilities, to clarify
the situation for the future.

7 years agoman: explain a bit more detailed what happens on suspend
Lennart Poettering [Tue, 30 Oct 2012 23:19:34 +0000 (00:19 +0100)]
man: explain a bit more detailed what happens on suspend

7 years agoman: document that Documentation= follows an order
Lennart Poettering [Tue, 30 Oct 2012 23:06:08 +0000 (00:06 +0100)]
man: document that Documentation= follows an order

7 years agoupdate TODO
Lennart Poettering [Tue, 30 Oct 2012 23:04:59 +0000 (00:04 +0100)]
update TODO

7 years agoupdate TODO: insserv belongs on the chopping block
Lennart Poettering [Tue, 30 Oct 2012 22:57:50 +0000 (23:57 +0100)]
update TODO: insserv belongs on the chopping block

7 years agostrv: cleanup error path loops
Lennart Poettering [Tue, 30 Oct 2012 17:29:45 +0000 (18:29 +0100)]
strv: cleanup error path loops

https://bugzilla.redhat.com/show_bug.cgi?id=858799

7 years agoupdate TODO
Lennart Poettering [Tue, 30 Oct 2012 17:29:35 +0000 (18:29 +0100)]
update TODO

7 years agoshared: "max" in the string->number conversion is meant to be inclusive
Michal Schmidt [Tue, 30 Oct 2012 14:45:50 +0000 (15:45 +0100)]
shared: "max" in the string->number conversion is meant to be inclusive

7 years agoshared, core: do not always accept numbers in string lookups
Michal Schmidt [Tue, 30 Oct 2012 13:29:38 +0000 (14:29 +0100)]
shared, core: do not always accept numbers in string lookups

The behaviour of the common name##_from_string conversion is surprising.
It accepts not only the strings from name##_table but also any number
that falls within the range of the table. The order of items in most of
our tables is an internal affair. It should not be visible to the user.

I know of a case where the surprising numeric conversion leads to a crash.

We will allow the direct numeric conversion only for the tables where the
mapping of strings to numeric values has an external meaning. This holds
for the following lookup tables:
 - netlink_family, ioprio_class, ip_tos, sched_policy - their numeric
   values are stable as they are defined by the Linux kernel interface.
 - log_level, log_facility_unshifted - the well-known syslog interface.

We allow the user to use numeric values whose string names systemd does
not know. For instance, the user may want to test a new kernel featuring
a scheduling policy that did not exist when his systemd version was
released. A slightly unpleasant effect of this is that the
name##_to_string conversion cannot return pointers to constant strings
anymore. The strings have to be allocated on demand and freed by the
caller.

7 years agobash-completion: add completion for coredumpctl
Zbigniew Jędrzejewski-Szmek [Tue, 30 Oct 2012 09:18:17 +0000 (10:18 +0100)]
bash-completion: add completion for coredumpctl

7 years agocoredumpctl: add --field/-F option
Zbigniew Jędrzejewski-Szmek [Tue, 30 Oct 2012 10:15:24 +0000 (10:15 +0000)]
coredumpctl: add --field/-F option

Useful for completion generation.

7 years agocoredumpctl: add --no-legend option
Zbigniew Jędrzejewski-Szmek [Tue, 30 Oct 2012 08:45:19 +0000 (09:45 +0100)]
coredumpctl: add --no-legend option

Useful for completion generation.

7 years agocoredumpctl: fix program return code
Zbigniew Jędrzejewski-Szmek [Tue, 30 Oct 2012 08:44:32 +0000 (09:44 +0100)]
coredumpctl: fix program return code

7 years agocoredumpctl: add guard to options table
Zbigniew Jędrzejewski-Szmek [Tue, 30 Oct 2012 08:35:53 +0000 (09:35 +0100)]
coredumpctl: add guard to options table

It is not nice to segfault on unknown options :(

7 years agolibsystemd-daemon: fix style
Michal Schmidt [Tue, 30 Oct 2012 09:30:44 +0000 (10:30 +0100)]
libsystemd-daemon: fix style

7 years agoshared, libsystemd-daemon: check for empty strings in strto*l conversions
Michal Schmidt [Tue, 30 Oct 2012 09:29:40 +0000 (10:29 +0100)]
shared, libsystemd-daemon: check for empty strings in strto*l conversions

strtol() and friends may set EINVAL if no conversion was performed, but
they are not required to do so. In practice they don't. We need to check
for it.

https://bugzilla.redhat.com/show_bug.cgi?id=870577

7 years agoSMACK: Add configuration options. (v3)
Auke Kok [Mon, 29 Oct 2012 22:30:05 +0000 (15:30 -0700)]
SMACK: Add configuration options. (v3)

This adds SMACK label configuration options to socket units.

SMACK labels should be applied to most objects on disk well before
execution time, but two items remain that are generated dynamically
at run time that require SMACK labels to be set in order to enforce
MAC on all objects.

Files on disk can be labelled using package management.

For device nodes, simple udev rules are sufficient to add SMACK labels
at boot/insertion time.

Sockets can be created at run time and systemd does just that for
several services. In order to protect FIFO's and UNIX domain sockets,
we must instruct systemd to apply SMACK labels at runtime.

This patch adds the following options:

Smack - applicable to FIFO's.
SmackIpIn/SmackIpOut - applicable to sockets.

No external dependencies are required to support SMACK, as setting
the labels is done using fsetxattr(). The labels can be set on a
kernel that does not have SMACK enabled either, so there is no need
to #ifdef any of this code out.

For more information about SMACK, please see Documentation/Smack.txt
in the kernel source code.

v3 of this patch changes the config options to be CamelCased.

7 years agologind: it's OK if a process on an pty requests a session for seat0
Lennart Poettering [Sun, 28 Oct 2012 16:37:16 +0000 (17:37 +0100)]
logind: it's OK if a process on an pty requests a session for seat0

After all, if a sudo/su inside an X terminal should get added to the
same session as the X session itself.

7 years agologind: unify all session lock loop
Lennart Poettering [Sun, 28 Oct 2012 16:25:23 +0000 (17:25 +0100)]
logind: unify all session lock loop

7 years agoupdate TODO
Lennart Poettering [Sun, 28 Oct 2012 12:55:47 +0000 (13:55 +0100)]
update TODO

7 years agosystemd: mount the EFI variable filesystem
Lee, Chun-Yi [Sat, 27 Oct 2012 03:23:22 +0000 (11:23 +0800)]
systemd: mount the EFI variable filesystem

Add efivarfs to the mount_table in mount-setup.c, so the EFI variable
filesystem will be mounted when systemd executed.

The EFI variable filesystem will merge in v3.7 or v3.8 linux kernel.

Cc: Kay Sievers <kay@vrfy.org>
Cc: Lennart Poettering <lennart@poettering.net>
Cc: Mantas Mikulėnas <grawity@gmail.com>
Cc: Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl>
Cc: Matt Fleming <matt.fleming@intel.com>
Cc: Jeremy Kerr <jeremy.kerr@canonical.com>
Cc: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Lee, Chun-Yi <jlee@suse.com>
7 years agounits: agetty overrides TERM
Michal Schmidt [Mon, 29 Oct 2012 20:59:34 +0000 (21:59 +0100)]
units: agetty overrides TERM

Environment=TERM=... has no effect on agetty who sets it by itself. To
really set TERM to a specified value, it has to be given on the command
line.

https://bugzilla.redhat.com/show_bug.cgi?id=870622

7 years agoutil: improve overflow checks
Michal Schmidt [Mon, 29 Oct 2012 20:04:47 +0000 (21:04 +0100)]
util: improve overflow checks

commit 49371bb fixed the observed division by zero, but missed another
occurrence of the same bug. It was also not the optimal fix. We can
simply make the divisor a constant by swapping it with the compared
value.

7 years agohostnamectl: do not choke on set-hostname with no argument
Kay Sievers [Mon, 29 Oct 2012 19:54:26 +0000 (20:54 +0100)]
hostnamectl: do not choke on set-hostname with no argument

https://bugzilla.redhat.com/show_bug.cgi?id=871172

7 years agoupdate TODO
Kay Sievers [Mon, 29 Oct 2012 19:54:19 +0000 (20:54 +0100)]
update TODO

7 years agoutil: avoid divide by zero FPE
Dave Reisner [Mon, 29 Oct 2012 19:49:34 +0000 (15:49 -0400)]
util: avoid divide by zero FPE

In early userspace, if kernel initialization happens extremely quickly,
a call to systemd-timestamp can potentially result in division by zero.
Ensure that the check in timespec_load, which only makes sense if tv_sec
is greater than zero, is guarded by this condition.

7 years agoNEWS: fix typo
Zbigniew Jędrzejewski-Szmek [Mon, 29 Oct 2012 09:52:31 +0000 (09:52 +0000)]
NEWS: fix typo

7 years agoswap: fix swap behaviour with symlinks
Olivier Brunel [Sat, 13 Oct 2012 12:24:15 +0000 (14:24 +0200)]
swap: fix swap behaviour with symlinks

Starting a swap unit pointing to (What) a symlink (e.g. /dev/mapper/swap
or /dev/disk/by-uuid/...) would have said unit marked active, following
the one using the "actual" device (/dev/{dm-1,sda3}), but that new unit
would be seen as inactive.
Since all requests to stop swap units would follow/redirect to it,
and it is seen inactive, nothing would be done (swapoff never called).

This is because this unit would be treated twice in
swap_process_new_swap, the second call to swap_add_one causing it to
eventually be marked inactive.

7 years agoswap: modernize style
Zbigniew Jędrzejewski-Szmek [Wed, 19 Sep 2012 15:38:00 +0000 (17:38 +0200)]
swap: modernize style

7 years agoswap: use automatic cleanup
Zbigniew Jędrzejewski-Szmek [Wed, 19 Sep 2012 15:06:28 +0000 (17:06 +0200)]
swap: use automatic cleanup

7 years agoswap: introduce helper variable
Zbigniew Jędrzejewski-Szmek [Wed, 19 Sep 2012 08:52:11 +0000 (10:52 +0200)]
swap: introduce helper variable

Just for readability, no funcational change.

7 years agologin: trivial grammar fix
Zbigniew Jędrzejewski-Szmek [Sun, 28 Oct 2012 12:30:00 +0000 (13:30 +0100)]
login: trivial grammar fix

7 years agosystemctl: skip JOBS column if no jobs
Zbigniew Jędrzejewski-Szmek [Sun, 28 Oct 2012 12:22:37 +0000 (13:22 +0100)]
systemctl: skip JOBS column if no jobs

Output is very constrained. This change saves 4 columns in the common
case.

7 years agobash-completion: avoid usage of ls for listing devices
Dave Reisner [Sun, 28 Oct 2012 16:27:27 +0000 (12:27 -0400)]
bash-completion: avoid usage of ls for listing devices

7 years agobash-completion: add minimal 'udevadm' support
Kay Sievers [Sun, 28 Oct 2012 16:18:55 +0000 (17:18 +0100)]
bash-completion: add minimal 'udevadm' support

7 years agoNEWS: fix a typo
Ben Boeckel [Sat, 27 Oct 2012 07:08:56 +0000 (03:08 -0400)]
NEWS: fix a typo

7 years agologind: add 'lock' as possible choice for handling hw keys
Lennart Poettering [Sun, 28 Oct 2012 11:29:27 +0000 (12:29 +0100)]
logind: add 'lock' as possible choice for handling hw keys

7 years agolibudev: hwdb - cleanup list before getting new properties
Kay Sievers [Sun, 28 Oct 2012 03:59:38 +0000 (04:59 +0100)]
libudev: hwdb - cleanup list before getting new properties

7 years agoudev: add "udevadm hwdb --test=<modalias>"
Kay Sievers [Sun, 28 Oct 2012 03:41:15 +0000 (04:41 +0100)]
udev: add "udevadm hwdb --test=<modalias>"

7 years agoudev: get rid of SYSCONFDIR
Kay Sievers [Sun, 28 Oct 2012 03:04:22 +0000 (04:04 +0100)]
udev: get rid of SYSCONFDIR

7 years agocompletion: fix typo in accessing array index
Ramkumar Ramachandra [Fri, 2 Dec 2011 16:44:32 +0000 (22:14 +0530)]
completion: fix typo in accessing array index

Remove spurious '}'. This error went unnoticed so far because Bash
doesn't complain.

7 years agounits: teach m4 scripts in units/ about Debian's rc.local
Ramkumar Ramachandra [Fri, 2 Dec 2011 16:44:31 +0000 (22:14 +0530)]
units: teach m4 scripts in units/ about Debian's rc.local

This makes the behaviour wrt. to rc[-.]local consistent between
various distributions supporting it.

7 years agoTweak TODO
Zbigniew Jędrzejewski-Szmek [Sat, 27 Oct 2012 22:43:44 +0000 (00:43 +0200)]
Tweak TODO

7 years agoman/logind.conf.xml: Add missing 'suspend' value for Handle*Key options.
Thomas Bächler [Wed, 24 Oct 2012 19:47:06 +0000 (21:47 +0200)]
man/logind.conf.xml: Add missing 'suspend' value for Handle*Key options.

7 years agoutil: fix possible integer overflows
Michal Sekletar [Thu, 25 Oct 2012 14:16:17 +0000 (16:16 +0200)]
util: fix possible integer overflows

7 years agolocalectl: fix memleak, use _cleanup_strv_free_
Michal Sekletar [Fri, 26 Oct 2012 12:16:18 +0000 (14:16 +0200)]
localectl: fix memleak, use _cleanup_strv_free_

l might contain zero strings, however there is still memory
allocated for NULL terminator, use _cleanup_strv_free_ instead to
prevent tiny leak in such case.

7 years agolocalectl: fix memleak, jump to finish before returning
Michal Sekletar [Fri, 26 Oct 2012 12:15:00 +0000 (14:15 +0200)]
localectl: fix memleak, jump to finish before returning

7 years agojournal: fix memleak, call set_free before return
Michal Sekletar [Fri, 26 Oct 2012 12:11:37 +0000 (14:11 +0200)]
journal: fix memleak, call set_free before return

7 years agounits: reword rescue mode hints
Zbigniew Jędrzejewski-Szmek [Tue, 23 Oct 2012 14:32:12 +0000 (16:32 +0200)]
units: reword rescue mode hints

Do not suggest to the user that commands can be issued before
logging in.

sulogin prints it own message, which mentions ^D, so there's no need
to repeat it here.

7 years agoupdate TODO
Lennart Poettering [Sat, 27 Oct 2012 22:49:49 +0000 (00:49 +0200)]
update TODO

7 years agologind: support for hybrid sleep (i.e. suspend+hibernate at the same time)
Lennart Poettering [Sat, 27 Oct 2012 22:49:04 +0000 (00:49 +0200)]
logind: support for hybrid sleep (i.e. suspend+hibernate at the same time)

7 years agolibudev: import hwdb and export lookup interface
Kay Sievers [Sat, 27 Oct 2012 18:59:01 +0000 (20:59 +0200)]
libudev: import hwdb and export lookup interface

7 years agohwclock: do not seal the kernel's time-warp call from inside the initrd
Kay Sievers [Sat, 27 Oct 2012 14:23:32 +0000 (16:23 +0200)]
hwclock: do not seal the kernel's time-warp call from inside the initrd

7 years agobash-compleiton: add missing --disk-usage option for journalctl
Dave Reisner [Fri, 26 Oct 2012 15:25:17 +0000 (11:25 -0400)]
bash-compleiton: add missing --disk-usage option for journalctl

7 years agoutil: return the remaining string in startswith()
Lennart Poettering [Fri, 26 Oct 2012 23:20:01 +0000 (01:20 +0200)]
util: return the remaining string in startswith()

7 years agocoredumpctl: add 'gdb' verb to start gdb right-away on a collected coredump
Lennart Poettering [Fri, 26 Oct 2012 23:19:47 +0000 (01:19 +0200)]
coredumpctl: add 'gdb' verb to start gdb right-away on a collected coredump

7 years agocoredumpctl: show timestamps in list
Lennart Poettering [Fri, 26 Oct 2012 18:34:39 +0000 (20:34 +0200)]
coredumpctl: show timestamps in list

7 years agojournal: special case the trivial cache chain cache entry
Lennart Poettering [Fri, 26 Oct 2012 18:25:36 +0000 (20:25 +0200)]
journal: special case the trivial cache chain cache entry

7 years agocoredumpctl: optimize journal entry parsing a bit by enumerating only once
Lennart Poettering [Fri, 26 Oct 2012 18:25:10 +0000 (20:25 +0200)]
coredumpctl: optimize journal entry parsing a bit by enumerating only once

7 years agocoredumpctl: initialize global vars
Lennart Poettering [Fri, 26 Oct 2012 18:23:28 +0000 (20:23 +0200)]
coredumpctl: initialize global vars

7 years agojournal: provide an API that allows client to figure out whether they need to recheck...
Lennart Poettering [Fri, 26 Oct 2012 18:05:19 +0000 (20:05 +0200)]
journal: provide an API that allows client to figure out whether they need to recheck the journal manually for changes in regular intervals

Network file systems generally do not offer inotify() that would work
across the network. We hence cannot rely on inotify() exclusiely in
those case. Provide an API to determine these cases, and suggest doing
manual regular rechecks.

Note that this is not complete yet, as we need to rescan journal dirs on
network file systems explicitly to find new/removed files

7 years agojournal: fix parsing of monotonic kernel timestamps
Lennart Poettering [Fri, 26 Oct 2012 12:55:03 +0000 (14:55 +0200)]
journal: fix parsing of monotonic kernel timestamps

7 years agokeymap: Add HP EliteBook 8440p
Martin Pitt [Fri, 26 Oct 2012 04:31:14 +0000 (06:31 +0200)]
keymap: Add HP EliteBook 8440p

Thanks to Glen Ditchfield <gjditchfield@acm.org>!

https://launchpad.net/bugs/1071579

7 years agoupdate TODO
Lennart Poettering [Fri, 26 Oct 2012 01:51:10 +0000 (03:51 +0200)]
update TODO

7 years agoudev: builtin - do not fail builtin initialization if one of them returns an error
Kay Sievers [Fri, 26 Oct 2012 01:43:56 +0000 (03:43 +0200)]
udev: builtin - do not fail builtin initialization if one of them returns an error

7 years agojournal: introduce entry array chain cache
Lennart Poettering [Fri, 26 Oct 2012 01:24:03 +0000 (03:24 +0200)]
journal: introduce entry array chain cache

When traversing entry array chains for a bisection or for retrieving an
item by index we previously always started at the beginning of the
chain. Since we tend to look at the same chains repeatedly, let's cache
where we have been the last time, and maybe we can skip ahead with this
the next time.

This turns most bisections and index lookups from O(log(n)*log(n)) into
O(log(n)). More importantly however, we seek around on disk much less,
which is good to reduce buffer cache and seek times on rotational disks.

7 years agotest: extend test-send to send some weirder data
Lennart Poettering [Thu, 25 Oct 2012 23:08:08 +0000 (01:08 +0200)]
test: extend test-send to send some weirder data

7 years agojournal: properly determine cutoff max date
Lennart Poettering [Thu, 25 Oct 2012 23:07:41 +0000 (01:07 +0200)]
journal: properly determine cutoff max date

7 years agoman: minor updates
Lennart Poettering [Thu, 25 Oct 2012 23:07:07 +0000 (01:07 +0200)]
man: minor updates

7 years agoupdate TODO
Lennart Poettering [Thu, 25 Oct 2012 23:06:50 +0000 (01:06 +0200)]
update TODO

7 years agoman: fix example
Lennart Poettering [Thu, 25 Oct 2012 23:06:43 +0000 (01:06 +0200)]
man: fix example

7 years agosysctl: parse all keys in a config file
Michal Sekletar [Thu, 25 Oct 2012 14:16:19 +0000 (16:16 +0200)]
sysctl: parse all keys in a config file

https://bugzilla.redhat.com/show_bug.cgi?id=869779

7 years agoudev: kmod - fix typo
Kay Sievers [Thu, 25 Oct 2012 22:28:23 +0000 (00:28 +0200)]
udev: kmod - fix typo

7 years agoman: typo fixes
Thomas Hindoe Paaboel Andersen [Thu, 25 Oct 2012 22:16:47 +0000 (00:16 +0200)]
man: typo fixes

https://bugs.freedesktop.org/show_bug.cgi?id=55890

Fixed typos, serial comma, and removed "either" as there were more
than two options. Also did an extra rename of "system-shutdown"
to "systemd-shutdown" that was forgotten in commit
8bd3b8620c80d0f2383f2fb04315411fc8077ca1

7 years agoudev: kmod, hwdb - do not fail if databases are not available
Kay Sievers [Thu, 25 Oct 2012 20:58:02 +0000 (22:58 +0200)]
udev: kmod, hwdb - do not fail if databases are not available

7 years agoudev: hwdb - exit if no database is available
Kay Sievers [Thu, 25 Oct 2012 20:22:29 +0000 (22:22 +0200)]
udev: hwdb - exit if no database is available

7 years agorules: fix usb_id/hwdb mixup in tty-description.rules
Kay Sievers [Thu, 25 Oct 2012 20:17:05 +0000 (22:17 +0200)]
rules: fix usb_id/hwdb mixup in tty-description.rules

<grawity> btw, in 15ce372b75a "call 'hwdb' with --subsystem=" you
  actually added "usb_id --subsystem=usb" (75-tty-description.rules)
<kay> ouch :)

7 years agoudev: hwdb validate() return when the database is not opened
Kay Sievers [Thu, 25 Oct 2012 20:08:33 +0000 (22:08 +0200)]
udev: hwdb validate() return when the database is not opened

7 years agoudev: hwdb - remove run_once
Kay Sievers [Thu, 25 Oct 2012 19:40:01 +0000 (21:40 +0200)]
udev: hwdb - remove run_once

7 years agoudev: hwdb - properly initialize search structure
Kay Sievers [Thu, 25 Oct 2012 19:39:01 +0000 (21:39 +0200)]
udev: hwdb - properly initialize search structure

7 years agorules: call hwdb for all devices with a 'modalias'
Kay Sievers [Thu, 25 Oct 2012 19:34:46 +0000 (21:34 +0200)]
rules: call hwdb for all devices with a 'modalias'

7 years agorules: call 'hwdb' with --subsystem= where appropriate
Kay Sievers [Thu, 25 Oct 2012 19:33:22 +0000 (21:33 +0200)]
rules: call 'hwdb' with --subsystem= where appropriate

7 years agoudev: hwdb: add acpi/pnp vendor description file
Kay Sievers [Thu, 25 Oct 2012 19:32:22 +0000 (21:32 +0200)]
udev: hwdb: add acpi/pnp vendor description file

7 years agoudev: set optind = 0, not the usual 1, to reset getopt_long()s internal state
Kay Sievers [Thu, 25 Oct 2012 19:31:38 +0000 (21:31 +0200)]
udev: set optind = 0, not the usual 1, to reset getopt_long()s internal state

7 years agojob: avoid recursion into transaction code from job cancelation
Michal Schmidt [Thu, 25 Oct 2012 00:31:49 +0000 (02:31 +0200)]
job: avoid recursion into transaction code from job cancelation

I hit an "assert(j->installed)" failure in transaction_apply(). Looking
into the backtrace I saw what happened:
1. The system was booting. var.mount/start was an installed job.
2. I pressed Ctrl+Alt+Del.
3. reboot.target was going to be isolated.
4. transaction_apply() proceeded to install a var.mount/stop job.
5. job_install() canceled the conflicting start job.
6. Depending jobs ended recursively with JOB_DEPENDENCY, among them was
   local-fs.target/start.
7. Its OnFailure action triggered - emergency.target was now going to be
   isolated.
8. We recursed back into transaction_apply() where the half-installed
   var.mount/stop job confused us.

Recursing from job installation back into the transaction code cannot be
a good idea. Avoid the problem by canceling the conflicting job
non-recursively in job_install(). I don't think we'll miss anything by
not recursing here. After all, we are called from transaction_apply().
We will not be installing just this one job, but all jobs from a
transaction. All requirement dependencies will be included in it and
will be installed separately. Every transaction job will get a chance
to cancel its own conflicting installed job.

7 years agojob: add comments to JobResult values
Michal Schmidt [Wed, 24 Oct 2012 23:22:22 +0000 (01:22 +0200)]
job: add comments to JobResult values

7 years agoupdate TODO
Lennart Poettering [Wed, 24 Oct 2012 23:20:18 +0000 (01:20 +0200)]
update TODO