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
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.
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.
Mostly useful for testing purposes. Setting Age to 1s works just as
well, but it is surprising that using 0s (or just 0) does not work.
Also clarify this in the documentation.
If you enter unit_add_exec_dependencies with m->where = NULL, you'll
very likely end up aborting somewhere under socket_needs_mount.
(When systemd goes to check to see if the journald socket requires your
mount, it'll do path_startswith(path, m->where)... *kaboom*)
This patch should ensure that:
a) both branches in mount_add_one() set m->where, and
b) mount_add_extras() calls unit_add_exec_dependencies() *after*
setting m->where.
If 'systemctl enable' (and friends) is run inside chroot it always
exits with a bad return code. unit_file_enable() returns the number of
symlink rules that were supposed to be created. So resetting r to 0 and
exiting gracefully should be the correct way.
'systemd-coredumpctl' will list available coredumps:
PID UID GID sig exe
32452 500 500 11 /home/zbyszek/systemd/build/journalctl
32666 500 500 11 /usr/lib64/valgrind/memcheck-amd64-linux
...
'systemd-coredumpctl dump PID' will write the coredump
to specified file or stdout.
Some keymaps apply to a large range of computer models, not all of which have
all of the scan codes in the maps. If a single scan code is invalid, do not
abort but continue with the next entry in the map. Instead just show the error
message for that particular scan code, to help with debugging.
Commit b1f87c76b1 changed sscanf from %i to %u, as scan codes are unsigned
numbers which can be > 0x7FFFFFFF. However, sscanf doesn't accept hexadecimal
numbers for %u. It works fine with %i, so revert this back.
Given that "journalctl -u" exists now there's no need to duplicate this
functionality in systemctl, so let's drop this, especially given that it
always felt a bit awkward to overload "-f" to both --force and --follow,
and to have continues output with a status header for this.
systemctl status -f avahi-daemon
now becomes:
journalctl -fu avahi-daemon
Which is shorter and a lot less redundant.
The new 'unique' API allows listing all unique field values that a field
specified by a field name can take in all entries of the journal. This
allows answering queries such as "What units logged to the journal?",
"What hosts have logged into the journal?", "Which boot IDs have logged
into the journal?".
Ultimately this allows implementation of tools similar to lastlog based
on journal data.
Note that listing these field values will not work for journal files
created with older journald, as the field values are not indexed in
older files.
Much like logind has a client in loginctl, and journald in journalctl
introduce timedatectl, to change the system time (incl. RTC), timezones
and related settings.
If a device unit has aliases defined in udev rules, and there are
other units that depend on that alias, as in
BindTo=sys-subsystem-net-devices-eth0.device
then systemd will fail the start the alias, and any dependent units
will time out. See
https://bugs.freedesktop.org/show_bug.cgi?id=52580
This is because unit_add_name() in device_add_escaped_name() will
return EEXIST.
The solution taken here is to call device_update_unit() on the alias
name. Thus if a unit with the alias name already exists, we reuse it;
otherwise a new unit is created. Creating multiple units for a single
device is perhaps suboptimal, but it's consistent with the treatment
of udev symlinks in device_process_new_device().
Under some circumstances this could lead to a segfault since we we
half-initialized a mount unit, then tried to hook it into the network of
things and while doing that recursively ended up looking at our
half-initialized mount unit again assuming it was fully initialized.
This commit checks for a usage line which contains [{|]reload[|}"] (to
not errnously match force-reload).
Heuristics like this suck, but it solves a real problem and there
appears to be no better way...
The fstab generator adds Before=swap.target by default, and when creating
a custom .swap unit, you can also add Before=swap.target to the unit.
However, it is impossible to not have this ordering dependency right now.
Virtually all existing setups likely use the fstab generator, so this
change is unlikely to break anything.
Use cases:
* iptables.service – atomically reload rules without having to flush
them beforehand (which may leave the system insecure if reload fails)
* rpc-nfsd.service – reexport filesystems after /etc/exports update
without completely stopping and restarting nfsd
(In both cases, the actual service is provided by a kernel module and
does not have any associated user-space processes, thus Type=oneshot.)
Valgrind says:
==29176== Conditional jump or move depends on uninitialised value(s)
==29176== at 0x412A85: cunescape_length_with_prefix (util.c:1565)
==29176== by 0x40B351: dev_kmsg_record (journald-kmsg.c:301)
==29176== by 0x40B653: server_read_dev_kmsg (journald-kmsg.c:347)
==29176== by 0x40B701: server_flush_dev_kmsg (journald-kmsg.c:365)
==29176== by 0x409DE7: main (journald.c:1535)
The MESSAGE_ID=... stanza will appear in countless number of places.
It is just too long to write it out in full each time.
Incidentally, this also fixes a typo of MESSSAGE is three places.
The removal of the TIMEOUT= handling in udevd put firmware requests into the
devpath parent/child dependency tracking. Drivers which block in module_init()
asking userspace for firmware ran into a 30 sec device timeout.
The whole firmware loading willl hopefully move into the kernel and
the fragile-since-day-one fake async driver-core device dance involving
udev can be retired:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=abb139e75c2cdbb955e840d6331cb5863e409d0e
If the final key in any sysctl.d file is a duplicate, systemd-sysctl
will exit with an error (and no explaination why). Ignore this, as
duplicate keys are to be expected when overriding settings in the
directory hierarchy.
This allows unprivileged clients to check for the used virtualization
even when lacking the privileges that some of the virtualization tests
require.
https://bugzilla.gnome.org/show_bug.cgi?id=684801
No longer override the default kernel keymap if nothing is specified in
vconsole.conf.
The default should be to do nothing (i.e., use what is already in the
kernel) unless the distro/admin has explicitly requested it.
No longer override the default kernel font if nothing is specified in
vconsole.conf.
The default kernel font[0] provides ISO-8859-1 and box characters. Users
of Arabic, Cyrilic or Hebrew must set a different font manually as these
character sets were provided by the old default font [1], but are not
any longer.
Rationale:
* it is counter-intuitive that an empty vconsole.conf file is different
from adding FONT="";
* the version of the default font shipped with Arch (which is the
upstream one) behaves very badly during early boot[2] (which should
admittedly be fixed in the font itself);
* the kernel already supplies a default font, it seems reasonable to
use that unless anything else is specified;
* This also avoids a needless slow call to setfont; and
* We don't want to work around problems in the kernel (in case the
compiled-in font is not acceptable for whatever reason).
[0]: <https://dev.archlinux.org/~tomegun/kernel.bdf>
[1]: <https://dev.archlinux.org/~tomegun/latarcyrheb.bdf>
[2]: <http://i.imgur.com/J2tM4.jpg>
As audit is pretty much just a special kind of logging we should treat
it similar, and manage the audit fd in a static variable.
This simplifies the audit fd sharing with the SELinux access checking
code quite a bit.
a) Instead of parsing the bus messages inside of selinux-access.c
simply pass everything pre-parsed in the functions
b) implement the access checking with a macro that resolves to nothing
on non-selinux builds
c) split out the selinux checks into their own sources
selinux-util.[ch]
d) this unifies the job creation code behind the D-Bus calls
Manager.StartUnit() and Unit.Start().
d4e9eb91ea changed the behavior for the F and f actions, wrongly sending
them to glob_item(). Restore the old behavior and shortcut straight to
write_one_file().
This minimal HTTP server can serve journal data via HTTP. Its primary
purpose is synchronization of journal data across the network. It serves
journal data in three formats:
text/plain: the text format known from /var/log/messages
application/json: the journal entries formatted as JSON
application/vnd.fdo.journal: the binary export format of the journal
The HTTP server also serves a small HTML5 app that makes use of the JSON
serialization to present the journal data to the user.
Examples:
This downloads the journal in text format:
# systemctl start systemd-journal-gatewayd.service
# wget http://localhost:19531/entries
Same for JSON:
# curl -H"Accept: application/json" http://localhost:19531/entries
Access via web browser:
$ firefox http://localhost:19531/
Among other cleanups this introduces a threshold for the size of binary
blobs we serialize as integer arrays in the JSON output. THis can be
disabled via --all.
The old code used a timestamp to print a timespan for unsealed journalfiles,
incorrectly showing things like 2230 days of unsealed entries. Print the timespan
between the first and last entry instead.
This requires a little bit of tip-toeing around to explicitly avoid
touching the environment from a sig handler. Instead, simply create a
function to reset the var to its "unset" state, allowing the next call
to columns() to recalculate and cache the new value.
Turns out cpuset needs explicit initialization before we could make use
of it. Thus mounting cpuset with cpu/cpuacct would make it impossible to
just create a group in "cpu" and start it.
Instead of doing hand optimized fd bisect arrays just use plain old
hashmaps. Now I can understand my own code again. Yay!
As a side effect this should fix some bad memory accesses caused by
accesses after mmap(), introduced in 189.
The mmap cache doesn't guarantee that we can look at two files at the
same time. Hence make sure to look at the entries to compare one
after the other, instead of at the same time when comparing them, and
reposition the window in between.
Systemctl would always return 1, because it treated uninteresting dbus
messages ("job added") as errors. Just ignore everything apart from
interesting ("job removed") messages.
Semantics are slightly different, because before unit_name_mangle
returning NULL was ignored, and now it is reported as oom. But
unit_name_mangle only returns NULL on oom.
Note: I did s/MANAGER/SYSTEMD/ everywhere, even though it makes the
patch quite verbose. Nevertheless, keeping MANAGER prefix in some
places, and SYSTEMD prefix in others would just lead to confusion down
the road. Better to rip off the band-aid now.
Lennart Poettering
