core: reorder where we add units to queues in unit_notify()
This moves all calls that shall do deferred work on detecting whether to
start/stop the unit or dependent units after a unit state change to the
end of the function, to make things easier to read.
So far, these calls were spread all over the function, and
conditionalized needlessly on MANAGER_RELOADING(). This is unnecessary,
since the queues are not dispatched while reloading anyway, and
immediately before acting on a queued unit we'll check if the suggested
operation really makes sense.
The only conditionalizaiton we leave in is on checking the new unit
state itself, since we have that in a local variable anyway.
core: change BoundBy= dependency handling to be processed by a deferred work queue
So far StopWhenUnneeded= handling and UpheldBy= handling was already
processed by a queue that is dispatched in a deferred mode of operation
instead of instantly. This changes BoundBy= handling to be processed the
same way.
This should ensure that all *event*-to-job propagation is done directly
from unit_notify(), while all *state*-to-job propagation is done from a
deferred work queue, quite systematically. The work queue is submitted
to by unit_notify() too.
Key really is the difference between event and state: some jobs shall be
queued one-time on events (think: OnFailure= + OnSuccess= and similar),
others shall be queued continuously when a specific state is in effect
(think: UpheldBy=). The latter cases are usually effect of the
combination of states of a few units (e.g. StopWhenUnneeded= checks
wether any of the Wants=/Requires=/… deps are still up before acting),
and hence it makes sense to trigger them to be run after an individual
unit's state changed, but process them on a queue that runs whenever
there's nothing else to do that ensures the decision on them is only
taken after all jobs/queued IO events are dispatched, and things
settled, so that it makes sense to come to a combined conclusion. If
we'd dispatch this work immediately inside of unit_notify() we'd always
act instantly, even though another event from another unit that is
already queued might make the work unnecessary or invalid.
This is mostly a commit to make things philosophically clean. It does
not add features, but it should make corner cases more robust.
This is like a really strong version of Wants=, that keeps starting the
specified unit if it is ever found inactive.
This is an alternative to Restart= inside a unit, acknowledging the fact
that whether to keep restarting the unit is sometimes not a property of
the unit itself but the state of the system.
This implements a part of what #4263 requests. i.e. there's no
distinction between "always" and "opportunistic". We just dumbly
implement "always" and become active whenever we see no job queued for
an inactive unit that is supposed to be upheld.
This is similar to OnFailure= but is activated whenever a unit returns
into inactive state successfully.
I was always afraid of adding this, since it effectively allows building
loops and makes our engine Turing complete, but it pretty much already
was it was just hidden.
Given that we have per-unit ratelimits as well as an event loop global
ratelimit I feel safe to add this finally, given it actually is useful.
core: add new PropagateStopTo= dependency (and inverse)
This takes inspiration from PropagatesReloadTo=, but propagates
stop jobs instead of restart jobs.
This is defined based on exactly two atoms: UNIT_ATOM_PROPAGATE_STOP +
UNIT_ATOM_RETROACTIVE_STOP_ON_STOP. The former ensures that when the
unit the dependency is originating from is stopped based on user
request, we'll propagate the stop job to the target unit, too. In
addition, when the originating unit suddenly stops from external causes
the stopping is propagated too. Note that this does *not* include the
UNIT_ATOM_CANNOT_BE_ACTIVE_WITHOUT atom (which is used by BoundBy=),
i.e. this dependency is purely about propagating "edges" and not
"levels", i.e. it's about propagating specific events, instead of
continious states.
This is supposed to be useful for dependencies between .mount units and
their backing .device units. So far we either placed a BindsTo= or
Requires= dependency between them. The former gave a very clear binding
of the to units together, however was problematic if users establish
mounnts manually with different block device sources than our
configuration defines, as we there might come to the conclusion that the
backing device was absent and thus we need to umount again what the user
mounted. By combining Requires= with the new StopPropagatedFrom= (i.e.
the inverse PropagateStopTo=) we can get behaviour that matches BindsTo=
in every single atom but one: UNIT_ATOM_CANNOT_BE_ACTIVE_WITHOUT is
absent, and hence the level-triggered logic doesn't apply.
Let's add an implicit reverse dep OnFailureOf=. This is exposed via the
bus to make things more debuggable: you can now ask systemd for which
units a specific unit is the failure handler.
OnFailure= was the only dependency type that had no inverse, this fixes
that.
Now that deps are a bit cheaper, it should be OK to add deps that only
serve debug purposes.
core: convert Slice= into a proper dependency (and add a back dependency)
The slice a unit is assigned to is currently a UnitRef reference. Let's
turn it into a proper dependency, to simplify and clean up code a bit.
Now that new dep types are cheaper, deps should generally be preferable
over everything else, if the concept applies.
This brings one major benefit: we often have to iterate through all unit
a slice contains. So far we iterated through all Before= dependencies of
the slice unit to achieve that, filtering out unrelated units, and
taking benefit of the fact that slice units are implicitly ordered
Before= the units they contain. By making Slice= a proper dependency,
and having an accompanying SliceOf= dependency type, this is much
simpler and nicer as we can directly enumerate the units a slice
contains.
The forward dependency is actually called InSlice internally, since we
already used the UNIT_SLICE name as UnitType field. However, since we
don't intend to expose the dependency to users as dep anyway (we already
have the regular Slice D-Bus property for this) this shouldn't matter.
The SliceOf= implicit dependency type (the erverse of Slice=/InSlice=)
is exported over the bus, to make things a bit nicer to debug and
discoverable.
In a later commit we intend to move the slice logic to use proper
dependencies instead of a "UnitRef" object. This preparatory commit
drops direct use of the slice UnitRef object for a static inline
function UNIT_GET_SLICE() that is both easier to grok, and allows us to
easily replace its internal implementation later on.
Luca Boccassi [Sun, 23 May 2021 12:43:38 +0000 (13:43 +0100)]
meson: look for bpftool in /usr/sbin too
On Debian, bpftool is installed in /usr/sbin, which is not in $PATH for
non-root users by default, so finding it fails.
Add a secondary, hard-coded '/usr/sbin/bpftool' after 'bpftool' so that
meson can find it.
наб [Tue, 16 Mar 2021 15:59:07 +0000 (16:59 +0100)]
kernel-install: don't erase previous errors if a hook returns 77
The manpage says that exiting 77 is the same as exiting 0,
then skipping all other hooks, but the behaviour heretofor
was to exit 0, skip all, and behave as if all hooks exited 0
sysusers: make sysusers work with dnf --installroot
This is not very pretty, but the code in fs-util.c already provisions for
missing /proc. We ourselves are careful to set up /proc, but not everybody
is and it is important for sysusers to also work where shadow-utils would:
I would like to replace calls to useradd and groupadd in Fedora systemd rpm
scriptlets with a call to sysusers. It has a number of advantages:
- dogfooding
- we don't need to manually duplicate the information from our sysusers
files to scriptlets
- a dependency on shadow-utils is dropped, which transitively drops dependencies
on setup and fedora-repos and bunch of other stuff.
We could try to get 'dnf' and 'rpm --root' and such to be reworked,
but not in any reasonable timeframe. And even if this was done, we'd still
want to support older rpm/dnf versions.
sysusers: add debug lines to failing getent/putent operations
I'm trying to use systemd-sysusers for systemd.rpm itself, and the invocation
in dnf chroot is failing like this:
...
Creating group input with gid 999.
Creating group kvm with gid 36.
Creating group render with gid 998.
Creating group sgx with gid 997.
Creating group systemd-journal with gid 190.
Creating group systemd-network with gid 192.
Creating user systemd-network (systemd Network Management) with uid 192 and gid 192.
Creating group systemd-oom with gid 996.
Creating user systemd-oom (systemd Userspace OOM Killer) with uid 996 and gid 996.
Creating group systemd-resolve with gid 193.
Creating user systemd-resolve (systemd Resolver) with uid 193 and gid 193.
Creating group systemd-timesync with gid 995.
Creating user systemd-timesync (systemd Time Synchronization) with uid 995 and gid 995.
Creating group systemd-coredump with gid 994.
Creating user systemd-coredump (systemd Core Dumper) with uid 994 and gid 994.
Failed to write files: Function not implemented
Let's add more info to make such failures easier to debug.
Roshan Shariff [Mon, 24 May 2021 07:30:38 +0000 (01:30 -0600)]
rules.d: Properly quote $env{MODALIAS}
Add quotes around use of $env{MODALIAS} in rules.d/80-drivers.rules. The
modalias can contain whitespace, for example when it is dynamically generated
using device or vendor IDs.
Yu Watanabe [Sun, 23 May 2021 07:59:40 +0000 (16:59 +0900)]
udev/net: do not manage loopback interfaces
There are nothing we can configure in udevd for loopback interfaces;
no ethertool configs can be applied, MAC address, interface name should
not be touched.
Yu Watanabe [Thu, 20 May 2021 21:22:37 +0000 (06:22 +0900)]
udev/net: drop fallback logic to set speed
ethtool_set_glinksettings() already fallback to use ETHTOOL_GSET/ETHTOOL_SSET
commands when ETHTOOL_GLINKSETTINGS/ETHTOOL_SLINKSETTINGS are not
supported.
Hans de Goede [Sat, 22 May 2021 17:42:21 +0000 (19:42 +0200)]
hwdb: keyboard: Ignore duplicate atkbd keypresses on touchpad on/off toggle
The atkbd device on the Lenovo Yoga 300-11IBR 2-in-1 sends unknown
keycodes when the touchpad is toggled on/off:
[ 1918.995562] atkbd serio0: Unknown key pressed (translated set 2, code 0x63 on isa0060/serio0).
[ 1918.995610] atkbd serio0: Use 'setkeycodes 63 <keycode>' to make it known.
[ 1919.032121] atkbd serio0: Unknown key released (translated set 2, code 0x63 on isa0060/serio0).
[ 1919.032135] atkbd serio0: Use 'setkeycodes 63 <keycode>' to make it known.
[ 1926.098414] atkbd serio0: Unknown key pressed (translated set 2, code 0x62 on isa0060/serio0).
[ 1926.098461] atkbd serio0: Use 'setkeycodes 62 <keycode>' to make it known.
[ 1926.146537] atkbd serio0: Unknown key released (translated set 2, code 0x62 on isa0060/serio0).
[ 1926.146583] atkbd serio0: Use 'setkeycodes 62 <keycode>' to make it known.
The "Ideapad extra buttons" driver alreadys sends f22 / f23 key-events
when the touchpad is toggles off, so map the keycodes for the duplicate
atkbd events to unknown to silence these kernel warnings.
sleep: use SleepOperation enum everywhere and drop sleep_settings()
Instead of comparing strings everywhere, let's use the new enum. This
allows us to drop sleep_settings(), since the operation enum can be
directly used as index into the config settings.
Some minor other refactoring is done, but mostly just shifting thing
around a bit, no actual change in behaviour.
units: make sure importd has CAP_LINUX_IMMUTABLE flag
Since d8f9686c0f1f276c0a687d9bd69f3adf33f15a95 we use the chattr +i flag
for marking containers in directories as reead-only. But to do so we
need the cap for it, hence grant it.
I'm working on building initramfs images directly from normal packages, and it
doesn't make sense for those units to be started. Pristine system rpms need to
behave correctly as much as possible also in the initrd, and those units are
enabled by the rpms. There usually isn't enough time for the timer to actually
fire, but starting it gives a line on the console and generally looks confusing
and sloppy. Flushing the journal means that its actually lost, since the real
/var is not available yet.
Another approach would be not enable those units, but right now they are
statically enabled, and changing that would be more work, and doesn't really
seem necessary, since the condition checks are very quick.
Checking for /etc/initrd-release is the standard condition that the initrd
units use, so let's do the same here.
Previously we'd pass all return values of read_virtual_file() to
log_info_errno() as error, but that makes no sense, given that we
sometimes return positive one with means "not truncated" but we'd show
as "Permission denied. Let's fix this, and log differently for sucess
and error.
Using format strings for concatenating strings is pretty unefficient,
and using PATH_MAX buffers unpretty as well. Let's revert to using
strjoina() as before.
However, to fix the fuzz issue at hand, let's explicitly verify the two
input strings ensuring they are valid path names. This includes a length
check (to 2K each), thus making things prettier, faster and using less
memory again.