Finit carried out a reboot, halt, or poweroff requested over the
initctl socket, but replied with a NACK. The command printed an
empty line and fell back to signalling PID 1.
The result was never set from the action taken. Return what
shutdown_now() returns.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
shutdown ignored everything after its options, so a delayed shutdown
went down right away:
shutdown -r +10 "Upgrading, back in five"
Take a time, now, +MIN, or the next HH:MM, and an optional message.
Finit keeps the countdown, so it survives logging out:
- warns logged-in users at 60, 30, 15 and 10 minutes left, then
every minute from five, and on the console
- writes the message to /etc/nologin from five minutes before,
leaving one already in place alone
- shutdown -c cancels, and a new shutdown TIME replaces it
Without a time, or with now and no message, shutdown acts right away
as before. An immediate reboot, halt, or power-off, also over D-Bus,
cancels a scheduled one. The warnings are written by a child, reading
utmp and closing a serial tty can block.
Fixes#149
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When the main process of a service exits, Finit sends SIGKILL to its
process group. That killed the action script of avahi-autoipd, which
it runs on SIGTERM to deconfigure the interface, before it was done.
Add stop-scope, in the block format and as the stop-scope: token:
service avahi-autoipd {
stop-scope = "process"
command = "avahi-autoipd eth0"
}
- pgroup: the stop signal to the main process, SIGKILL to its
process group when it exits, as before and the default
- process: only the main process
- cgroup: the stop signal to all of the cgroup, SIGKILL to the rest
- mixed: the stop signal to the main process, SIGKILL to the rest of
the cgroup, like systemd's KillMode=mixed
SIGKILL to a cgroup uses cgroup.kill when the kernel has it. Without
a cgroup, cgroup and mixed fall back to the process group. Services
in the same .conf file share a cgroup, Finit warns when one of them
uses a cgroup scope.
Fixes#413
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Stopping a forking service after its start script had exited, but
before the daemon had written its PID file, did not hold. When the
PID file showed up, the service was reported running again and the
daemon was left alone:
initctl stop ext
service_forked() cleared any block on the service, the user's stop
included.
Keep a user stop, and let the state machine stop the late daemon of
a service no longer enabled. The service is briefly RUNNING on its
way to STOPPING, so service/NAME/running blinks on and off.
Follow-up to #452
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A stopped service with a daemon-owned PID file came back as running
when another process wrote that file, e.g., a container started by
hand:
sudo podman start dash
turned container:dash from stopped to running in initctl, with the
podman PID adopted, moved to the service's cgroup, and its conditions
set. The same happened to a service waiting for its conditions.
The plugin acted on any new or changed PID file, only skipping
services stopping or running scripts, and service_forked() then
cleared the stop.
Only let a service that is starting or running claim its PID file.
Fixes#452
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A notify:systemd service that calls sd_notify() early could lose its
READY=1 and never become ready. Seen with udevd in an initramfs.
The socket address contained the PID of the service, so it was bound
in the parent after fork(), racing the service.
Let the kernel pick a unique abstract address instead, bound before
fork(), and pass it in NOTIFY_SOCKET.
Fixes#505
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
After a SIGHUP, from initctl reload or a changed service, Finit waits
for a pid-notify daemon to recreate or touch its PID file to reassert
readiness. keventd only reloaded its rules, so with a daemon-owned
PID file it never did.
Touch it with pidfile() after reloading the rules, once keventd has
created it. With -c that is after the coldplug queue has drained.
Fixes#503
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Changing a service from @alice:wheel to @alice, or removing group from
a block, and reloading kept the service running as group wheel, also
after a restart.
The group is only set when the configuration has one, and on reload
the service is updated in place, so the old one stayed.
Clear it before parsing, like a new service gets it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service listing more than four supplementary groups, in extra-groups
or as @user:group,sup1,sup2,..., got only the first four:
too many supplementary groups, max 4
The names were stored in a fixed array of four. It was also only
reset when the new configuration listed groups, so groups dropped from
a service lingered until reboot.
Store the list as one string, split when the service starts, and set
it on every (re)registration. A list longer than 511 characters drops
the name it cuts off, which may otherwise match another group.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A run/task/service whose command starts like an option without a ':'
lost its command, e.g. the example in finit.conf(5):
run [2345] @joe:users logger "Hello world"
was registered as the log option, running "Hello world". Same for
pidof, pidproxy, cgroupfs-mount, and in block format too.
The option parser matches with MATCH_CMD(), a prefix match, also for
the options with no trailing ':' to end the prefix: log, pid, cgroup,
respawn, nowarn, and the rest.
Match these against the whole token, allowing only the separator that
starts a value: "log:", "pid:", "cgroup." and "cgroup:".
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A command that takes -- as an argument lost it, and everything after
it, to the description:
service name {
command = "cmd -- arg"
description = "a cmd"
}
ran `cmd` with description `arg`.
The block translator builds a legacy one-liner, appending the
description as "-- text", and service_register() splits that line at
the first "-- ", here the one in the command.
Split the description off only for legacy one-liners. Blocks now hand
it to the new service_register_desc() as an argument.
Fixes#512
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service whose condition goes into flux during a reload is paused
with its reload still pending. If another reload was requested in
the meantime, re-parsing its unchanged .conf file cleared the pending
mark, so the service was resumed without ever being reloaded. Seen
with sshd <pid/syslogd> on Infix, where a configuration change that
touched both landed as two reloads in a row and sshd kept its old
listen addresses.
The mark is only ever cleared once the change has been applied, so a
mark that is still set when the file is parsed again means exactly
that: not applied yet. Leave it alone.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Covers the longer user and group names, the tmpfiles mode and owner
work, the shutdown remount order, and the stop/reload script timeout.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Same treatment as d/D got: e went through fisdir() and a path based
chmod/chown, f/F did the chmod/chown by path after closing the file.
Both now work on the open fd, so the mode and owner end up on the
directory or file that was just checked or written. f/F use open(2)
directly, which lets a plain f rely on O_EXCL for create-if-missing
instead of the earlier stat().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
(cherry picked from commit 75739fbeef)
Apply a PAM session to run/task/sysv/services Finit starts, pam_limits
above all, so a service running as a given user picks up that user's
limits the way a login does.
Add a new `pam` setting for the new block format (only), like the
per-service directories, naming a file in /etc/pam.d:
service weston {
user = "weston"
pam = "weston-autologin"
command = "/usr/bin/weston --continue-without-input"
}
pam_close_session() has to be called by a process still holding the
handle, and the handle does not survive exec(). Hence the keeper: it
holds the handle, drops to the service's credentials, and waits for a
parent-death signal before closing the session. Same shape as
systemd's (sd-pam), for the same reason, and one per fork, so the
script hooks open and close their own.
The keeper closes the descriptors it inherited from Finit and only
those. Closing everything would also take out what pam_open_session()
opened for itself, a keyring fd or a lock file, and leave the modules
to close a session with those pulled out from under them. Closing
nothing, as (sd-pam) does, would leave it holding the write end of the
notify pipe for the service's whole lifetime and starve notify = "s6"
services of their ready signal. So the fds open before pam_start()
are snapshotted and exactly those are closed, while the ones PAM opens
after are marked close-on-exec so the daemon does not inherit them
either.
A refused value, a denied account stack, an uninstalled pam.d file,
and a build without PAM support all keep the service from starting
rather than running it with the stacks skipped: one that quietly loses
pam_limits and its private /tmp, with nothing said. Capabilities a
module like pam_cap.so granted are merged into the IAB Finit applies
instead of being replaced by it, which only helps a service that also
sets capabilities, the other arm being a plain setuid() with nothing
left to restore once permitted is empty.
The test sysroot gains pam_permit.so, pam_deny.so and pam_limits.so,
which ldd cannot see, libpam dlopen()s them, and the test skips when
the host has none to stage. The negative cases pin the exit status
rather than only asserting crashed, which serv reports for any early
exit, so a bad command or an unwritable pidfile cannot pass for a
rejected session.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The ChangeLog is hard-wrapped at 72 columns, which the GitHub releases
page renders verbatim as a ragged right edge. Join the continuation
lines of each paragraph and list item, as already done in Infix, and
let the browser do the wrapping.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
GitHub now shows a sha256 digest for every uploaded release asset, so
the sidecar hash files only clutter the asset list.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Building it by default installs our libsystemd.so.0 in $libdir. Where
the real one is already present, and contrib/debian/build.sh configures
--prefix=/usr --exec-prefix= on a Debian host, ours outranks it in the
loader cache, and every program linking libsystemd loses
sd_journal_stream_fd and the LIBSYSTEMD_209 symbol versions. The test
sysroot ran into exactly that, dbus-daemon exited 127 on every restart.
Restore the default to no. sulogin and watchdogd stay on, they only add
binaries. distcheck asks for the library so it stays covered.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
dbus-broker fails, the bus never comes up:
Service dbus[19] died (exit status: 127), restarting
The sysroot stages dbus-daemon from the host, which links the real
libsystemd, and since v5.0 builds ours by default the install puts a
second libsystemd.so.0 in the sysroot. It wins the ld.so cache, and
dbus-daemon finds neither sd_journal_stream_fd nor the LIBSYSTEMD_209
symbol versions in it.
Uninstall the library from the sysroot after the install. The test serv
compiles sd-daemon.c in, so nothing there needs it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
v5.0 ships keventd and the D-Bus support enabled out of the box, but the
remaining bundled pieces stayed opt-in, and the help text for two of them
already claimed otherwise.
Default all three to yes; --without-sulogin, --without-watchdog, and
--without-libsystemd opt out. The distcheck and CI configure lines drop
the flags they no longer need, so CI exercises the defaults.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
make distcheck fails in install:
/usr/bin/install: cannot create regular file '/etc/dbus-1/system.d/org.finit.conf': Permission denied
The policy was installed to $sysconfdir/dbus-1/system.d. distcheck only
overrides the prefix, so the file escaped its sandbox and aimed for the
real /etc.
Install it where dbus looks for package owned policy, leaving
/etc/dbus-1/system.d to the admin. The test bus config reads it
relative to itself.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
fuzz-msg-parse.sh looked for its driver next to the script, which only
holds when srcdir and builddir are the same. Every VPATH build skipped
the sweep, claiming D-Bus was off:
SKIP: fuzz-msg-parse.sh
distcheck, and with it the release job, is a VPATH build. Use
top_builddir, which the test environment already exports.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
make distcheck fails in dbus-device:
sh: /libexec/finit/keventd: not found
distcheck configures with --prefix=$(distdir)/_inst, so the bundled
helpers install below that prefix, not in /libexec/finit, which the test
had hard coded.
Export the configured pkglibexecdir as FINIT_EXECPATH in test.env and
run keventd from there, the same path finit itself is built with.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The engine covers most of the udev grammar but not all of it, and the
gaps are invisible until a rule silently does nothing. Write down what
is implemented and where it parts ways with udev(7), rather than
leaving people to infer it from a ruleset that happens to work. The
man page gets the directory precedence and a pointer to udev(7) and
the User's Guide, which hold the details.
Rename the menu entry to Device Manager while here, matching how the
watchdog daemon is listed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The rules directories were read one after the other, each sorted on
its own. udev instead sorts all files from all directories together
and lets a file mask one of the same name in a lower priority
directory. The difference broke both standard override idioms
without a sound: copying a rules file to /etc/udev/rules.d/ to edit
it left the original running too, and symlinking a name to /dev/null
disabled nothing.
Collect everything with glob(), drop entries shadowed by a later
directory, and sort the survivors by filename, same as udev and the
tmpfiles.d handling in Finit itself. The -r directory sits above
/etc/ so a test ruleset can override anything shipped.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Two races, both around the post:script Finit runs when it gives up on a
service.
Finit marks the service crashed before it forks post:script, so the file
the test greps for lands a moment later. Waiting for the state is not
enough.
slay then takes the PID from initctl status, and service_post_script()
sets svc->pid to the script's PID, so once Finit has given up the PID
reported for the service is the post:script. Killing that takes out the
script instead of the service and the file never arrives at all. The
guard for this was already there, with a comment describing it, but
inside the loop that waits for a PID to appear, so it only covered the
case where there was none. A PID that was already there went straight
to kill -9. Hence the gcc leg killing once more at lap 13, after Finit
had stopped restarting, where clang stopped at 12.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Stanzas in /lib/finit/system cannot use if:keventd, or any other
built-in service: if: is evaluated at parse time, and the generated
.conf files for built-ins were read after the system directory, so the
referenced service did not exist yet and the stanza was dropped.
The late position was deliberate: it kept bundled services from starting
before udev (5d703fd3). The motivating case was dbus, which has since
moved to a build-time 20-dbus.conf (2e0b1d6f), and the built-ins that
remain want the early slot: watchdogd should start as soon as possible,
keventd is the device manager, and runparts is gated on int/bootstrap so
its position never mattered.
Reading /run/finit/system first also puts the override chain in its
natural order: built-in/generated defaults, then bundled (read-only)
system files in /lib, then the administrator (override) .conf file in
/etc/finit.d, always read last.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Half the daemon was missing: -p, -r, -S, -t from SYNOPSIS and
OPTIONS, class/ and driver/ conditions, SIGHUP rules reload, the
rules engine and its directories, /run/udev/data, and the D-Bus
socket in FILES.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Three copies of read-first-line-from-a-file (sysfs_read_file, a
static sysfs_read, fgetline), two condition emitters symlinking the
same reconf marker (cond_emit, sys_cond), two seqnum-stable loops,
and five hand-rolled logit() prototypes, none with a printf
attribute, so no format string was ever checked.
Keep one of each: cond_emit() is exported and mkpaths its parents,
which also retires init_dev_condition_dir() and the sys/ directory
priming loop; logit() moves to keventd.h with the format attribute;
the stability test becomes seqnum_stable(), shared by the coldplug
gate and settle; kev_now_ms() serves dbus.c too. rule_ctx_free()
and kev_seq_baseline() had one caller each in their own files, now
static.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
initctl reload freezes conditions to the old generation and each
owner re-asserts. Finit's own providers do this in-process; an
external provider whose conditions are generation files, rather than
the oneshot symlinks keventd uses, has no way to know the moment.
Emit Manager1.ConfigReloaded when reconfiguration completes.
keventd needs no subscriber: its conditions are symlinks to the
reconf marker itself, so they read the current generation by
construction and never flux, which the device bus test now pins
down.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Introspection, queue-state properties, immediate and bus-first
settle, Info by devpath, Trigger(add, net) observed via the
DeviceProcessed signal, and RulesReload -- driven by dummy interface
hotplug like keventd.sh. New call-ss and call-u client modes, and
getprop learns the b and t variants.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The udev parity gaps that were blocked on IPC -- settle, trigger,
info, queue introspection, runtime rule reload -- become bus methods.
keventd serves its own socket the way Finit serves /run/finit/bus:
brokerless, libink, one socket per daemon, no forwarding between the
two.
Settle(u) -> b parked until the queue drains or the timeout
passes; true when settled
Trigger(s, s) replay events, action + subsystem glob
Info(s) -> a{ss} /run/udev/data properties for a devpath
RulesReload() -> u re-read rules dirs, returns rule count
QueueEmpty (b), SeqnumProcessed (t) properties
DeviceProcessed (ss) signal after each fully handled event
The queue state is the highest kernel seqnum keventd has handled,
baselined at startup, against /sys/kernel/uevent_seqnum. keventd -S
now asks the running daemon first and falls back to seqnum polling.
In passive mode Trigger and RulesReload refuse. Adds
link_call_connection() for the park bookkeeping.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
udevdb_read_parent() contained the general capability, keyed from
sysfs when no uevent is in hand. Split it out as
udevdb_read_devpath() for the D-Bus Info method; parent lookup
becomes a wrapper.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
coldplug_trigger(action, subsystem-glob) replays events for a subset
of devices, the D-Bus Trigger method needs both knobs; coldplug() is
now the ("add", NULL) case. nftw() has no user cookie, so the
parameters ride in file statics.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The park machinery was welded to broker uid resolution; a handler
that cannot answer yet, like a device-settle call waiting for the
event queue to drain, had no way to defer. link_call_park() holds
the request, link_call_resume() re-runs the handler with
link_call_resumed() reading true, and the expire sweep remains the
backstop for a resume that never comes.
Resume also no longer drops a local caller's kernel group set in the
privileged re-check: group source now keys on broker-ness.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Device seqnums are 64-bit; the writer and reader stopped at u32.
Adds t/x/d to the skip path so a{sv} consumers tolerate them.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
dbus.md was orphaned: not in dist_docs_DATA, not linked from the
user guide. Wire it into the doc dist and link it from the index
features list, features.md, initctl.md, and plugins.md, where the
dbus.so plugin entry now disambiguates the external system bus from
the built-in org.finit API.
Document the 64-peer cap, the supported AddMatch keys, the busconfig
policy file, the legacy-parity edge semantics with the deliberate
SetRunlevel InvalidArgs divergence, the reload-signal behavior of
Service1.Reload, and the reboot family timeout. Refresh the stale
initctl.md usage paste, add monitor and the D-Bus transport to
initctl(8), add /run/finit/bus to the filesystem layout, and flatten
the ChangeLog D-Bus entry to house style.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The D-Bus support is default-enabled, so the HAVE_DBUS paths only
bit-rot silently without this: the leg caught initctl failing to
build with --disable-dbus on its first local run. Also asserts the
binaries carry no bus references and smoke-runs one non-dbus test.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
initctl_CPPFLAGS is only assigned under the DBUS conditional, but an
automake per-target variable exists even when its conditional is
false, so a --disable-dbus build dropped AM_CPPFLAGS entirely:
util.c:319:16: error: invalid use of undefined type 'struct FTW'
Assign the base flags unconditionally and append under DBUS.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The path encoding in libink/path.c was untested for identities with
separators; dbus-service.sh only used the bare keventd identity.
Declare a dhcp-client:eth1 service and verify the escaped path, that
the object introspects, and that Identity round-trips.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
All three org.finit signals were invisible to generated proxies, and
org.freedesktop.DBus was missing from the standard interfaces even
though Hello, AddMatch, and RemoveMatch are answered.
Add a link_signal_t table to the vtable, emitted like methods and
properties, declare the Manager1 and Cond1 signals, and complete the
static XML with PropertiesChanged and org.freedesktop.DBus.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The D-Bus methods carried byte-for-byte copies of api.c's static
start/stop/restart helpers. Promote them to service.c alongside
service_reload(), which already serves both callers, and reduce both
sides to svc_parse_jobstr-style adapters.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Regression tests for the recent handler fixes: bogus SetRunlevel is
InvalidArgs, Signal on a stopped service is Failed, and the reboot
family declares the timeout argument. Reboot cannot be invoked
without taking down the sandbox, so the latter is asserted via
introspection. New call-u and call-su modes in dbus-auth-client.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
initctl -t N reboot arms an emergency bypass timer over the legacy
socket, but the bus methods took no argument, so the timeout was
silently dropped whenever D-Bus was up.
Reboot, Halt, and Poweroff now take a timeout in seconds, 0 for
none, armed via the same shutdown_bypass() the legacy path uses.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Service1 Start/Stop/Restart discarded the action return value, their
Manager1 twins map it to org.finit.Error.Failed. Cond1 Set/Clear
replied success even when the condition symlink operation failed,
where legacy initctl exits 73. Verify the resulting condition state
with cond_get() rather than the noupdate return values, which report
no-change, not failure, and would reject an idempotent re-set.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Manager1.Signal silently skipped stopped services, so the same
command gave different exit codes depending on transport: the legacy
INIT_CMD_SIGNAL path fails when the service is not running. Mirror
the legacy behavior.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The bus method called sm_runlevel() unconditionally. In runlevel 0
and 6 that aborts an in-flight shutdown, which INIT_CMD_RUNLVL
refuses with a warning, and during bootstrap it switches immediately
where the legacy path defers via cfglevel to the end of runlevel S.
Port both. A bad runlevel argument still returns InvalidArgs, where
the legacy protocol acks silently: a typed interface rejects garbage.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The legacy INIT_CMD_SUSPEND is refused in runlevel S, 0, and 6, the
bus method suspended unconditionally, even mid-shutdown. Add the
same guard, replying WrongRunlevel like the reboot family.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Once it gives up it forks the post:script and reports that PID as the
service's, so a slay still waiting for the service to come back killed
the script instead, and crashing.sh lost the /tmp/post it checks for.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The unified keventd has no automated coverage, only the devmon
fallback is exercised by the test suite. Network interfaces are
the one device class an unprivileged test can hotplug: the sandbox
has its own network namespace, so 'ip link add' makes the kernel
emit genuine uevents.
Verify keventd readiness, <class/net/IFNAME> driving a service --
and <dev/IFNAME> NOT asserted, interfaces are not device nodes --
libudev-compatible n<ifindex> keying in /run/udev/data, conditions
surviving initctl reload, and cleanup on interface remove.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service with a <class/net/eth0>, or any other keventd-provided,
condition is never started when the device appears. keventd asserts
the condition file, but devmon only watches /dev, so no cond_update()
ever reaches affected services. Reload made it worse:
devmon_reconf() clears any registered dev/ condition without a /dev
node behind it.
Watch the dev/, class/, and driver/ condition directories, like the
sys and usr plugins do for their namespaces, and treat an existing
condition file as device presence in devmon_reconf().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
keventd links libblkid, which finit itself does not, but sysroot.mk
only copied the libraries finit links. Inside the sysroot keventd
then fails to start:
Service keventd[18] died (exit status: 127)
Collect libraries from finit and everything installed under
libexec/finit/ instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When started with -c (the default when keventd is the device manager),
gate the pidfile on the kernel's uevent_seqnum having been stable for
200ms. Up to now ready signaling with the pidfile was done right after
coldplug() triggered the kernel to re-emit events, but before uev_run()
had drained any of them, so <pid/keventd> really only meant "listening
on netlink".
With the gate, services that depend on <pid/keventd> can now assume /dev
is populated and persistent symlinks are live.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Stop-gap "settle" equivalent of udevadm settle for migration scenarios.
Polls /sys/kernel/uevent_seqnum every 50ms and exits zero when the
sequence number has been stable for 200ms (or non-zero after -t SECONDS
timeout, default 30s).
This is racy by design -- a slow probe firing after we return still
races -- so the doc steers users toward dev/, class/, and bind/
conditions for any service they control. Settle is for legacy boot
scripts and init transitions where condition wiring isn't feasible.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
dev/<X> only fires when a device gets a /dev node, which leaves a lot of
embedded-relevant hardware uncoverable: DSA switch cores, IIO sensors,
LEDs, backlight, PHYs, regulators -- all live purely under sysfs.
Two new condition namespaces:
class/<subsystem>/<sysname> asserted on every sysfs class device add
(e.g. <class/leds/blue>)
driver/<name> asserted while the driver is bound to at
least one device (e.g. <driver/mt7530>)
A driver can bind to several devices, so driver/ conditions are
refcounted: asserted on first bind, cleared when the last device is
unbound.
dev_cond() is generalized into a static cond_emit(prefix, rel, set) so
class_cond() and driver_cond() share the same mkpath + symlink/erase
code. The name avoids colliding with src/cond.h's public cond_path()
helper (unrelated function that returns a condition's filesystem path).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a rule references an absolute helper path (e.g. /lib/udev/fido_id,
/lib/udev/scsi_id) that the system does not ship and we have no matching
builtin either, return 1 from try_builtin_fallback() so the caller does
not fork /bin/sh on a binary that's known to be missing. Previously
each such uevent left a zombie 127 child for keventd's sigchld_cb to
reap -- harmless but noisy and wasteful at coldplug.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
modprobe_load() and run_program() used to temporarily restore the
default SIGCHLD handler around fork+waitpid because the main loop set
SIGCHLD=SIG_IGN. With the libuev conversion, sigchld_cb in keventd.c
handles reaping via signalfd -- and signalfd does not interfere with
synchronous waitpid(pid, ...) -- so the dance is dead code.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Replace the raw poll() + global running/reload_rules flag dance with
libuev. All four signals (SIGUSR1, SIGTERM, SIGHUP, SIGCHLD) are now
handled via uev_signal_init() -- libuev uses signalfd internally so the
handlers run in normal main-loop context, not async signal context.
The SIGCHLD handler waitpid(-1, WNOHANG) reaps any subprocess (modprobe
loads, rule RUN+= helpers) instead of the old SIG_IGN auto-reap trick.
Synchronous waitpid() in modprobe_load() and run_program() still works
because signalfd queues the signal -- it only reaches sigchld_cb on the
next main-loop iteration.
The netlink receive loop becomes uevent_cb() registered via uev_io_init(),
with the receive buffer passed via the watcher's arg slot (avoids
file-static state). ENOBUFS and EINTR/EAGAIN are still tolerated, any
other recv() error still panic()s.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
keventd is the only thing in the tree that links libblkid, so a tree
that used to build now stops in configure with no hint of which package
to install. Say so where people look for dependencies, and give CI the
package it now needs.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add 27 stock rules installed to /lib/udev/rules.d/. Rules invoking
/lib/udev/<helper> fall back to keventd builtins (path_id, usb_id,
blkid, hwdb, kmod, net_id, input_id) when the helper binary is absent;
user-supplied helpers in /lib/udev/ still take precedence.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Transform keventd from a power-supply monitor + basic hotplug handler into
a full udev-compatible device manager.
Rules engine (rules.c):
- Full .rules file parser covering all udev key types: ACTION, KERNEL,
SUBSYSTEM, DEVPATH, ENV, ATTR, SYSCTL, TAG, RESULT, PROGRAM, TEST,
parent-chain KERNELS/SUBSYSTEMS/ATTRS/DRIVERS, and more
- Pattern matching: plain string, fnmatch glob, and pipe-separated alternatives
- Operators: ==, !=, =, +=, -=, :=
- Assignments: NAME=, MODE=, OWNER=, GROUP=, SYMLINK+=, ENV{k}=, TAG+=, RUN+=
- IMPORT{program|file|builtin|parent|cmdline|db}=
- PROGRAM= with stdout capture for subsequent RESULT== matching
- GOTO=/LABEL= flow control
- Loads *.rules from /lib/udev/rules.d, /run/udev/rules.d, /etc/udev/rules.d
and an optional extra directory (-r DIR); reloads on SIGHUP
Builtin framework (builtin.c):
- kmod: load module by MODALIAS or explicit alias
- hwdb: match device against *.hwdb text files in udev hwdb dirs; builds
correct lookup key per subsystem — evdev:input:b*v*p*e* for input,
usb:v*p* for USB, raw modalias for PCI/platform
- path_id: build stable ID_PATH / ID_PATH_TAG from sysfs topology (PCI, USB,
ATA, NVMe, platform, ACPI, virtio)
- usb_id: read idVendor/idProduct/bcdDevice/serial from sysfs; look up
ID_VENDOR_FROM_DATABASE and ID_MODEL_FROM_DATABASE from usb.ids
(hwdata package) when available; silent fallback when absent
- input_id: classify input devices (keyboard, mouse, joystick, touchscreen,
touchpad) from evdev capability bitmasks in sysfs
- net_id: generate predictable names — MAC-based enx<mac> and PCI-slot-based
enp<bus>s<dev>[f<func>]
- blkid: probe filesystem type, UUID, and label via libblkid; sets ID_FS_*
and ID_PART_TABLE_* properties
Network interface renaming (uevent.c):
- netdev_add() renames interfaces via SIOCSIFNAME when a NAME= rule matched,
then sets the Finit dev/ condition on the final name; and any setup using
persistent interface naming via udev rules
Device node and symlink improvements (uevent.c):
- NAME=, MODE=, OWNER=, GROUP= overrides from matched rules applied at
mknod/chown time, falling back to the built-in permission table
- SYMLINK+= links from rules applied alongside built-in by-id/by-path links
Device property database (udevdb.c):
- Persist per-device E:/S:/I: records to /run/udev/data/ on ADD/CHANGE,
delete on REMOVE; IMPORT{db}= restores saved properties into event env
Build:
- libblkid (util-linux) is now required for keventd
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is a backwards compatible mode for users upgrading and not noticing
that keventd is now build by default.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
After keventd processes a uevent (creating device nodes, loading
modules, etc.), rebroadcast the original event to netlink group
0x4 so that libudev-zero consumers -- graphical applications,
Wayland/X11 compositors, libinput, and anything else using libudev
to monitor device hotplug -- can receive device events.
Rebroadcast is enabled by default. Use -g to override the target
netlink group mask, or -G to disable rebroadcast entirely. Bit 0
(kernel group) is always masked out to prevent feedback loops.
Ref: https://github.com/finit-project/finit/issues/451#issuecomment-3817233886
See: https://github.com/illiliti/libudev-zero
Suggested-by: Aaron Andersen <aaron@fosslib.net>
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Rewrite doc/keventd.md from a 14-line stub into comprehensive
documentation covering all features of the new unified keventd:
device node creation, persistent symlinks, firmware loading,
module loading, coldplug, conditions, and command-line usage.
Update doc/conditions.md to list keventd as the primary provider
of dev/* and sys/pwr/* conditions, with devmon as fallback when
an external device manager is used instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Evolve keventd from a power_supply-only monitor into a full device
manager capable of replacing mdev/mdevd on embedded systems. This
is the first step towards Finit v5.0 where keventd absorbs devmon.
New capabilities:
- Parse all uevent actions (add, remove, change, bind, unbind)
- Create and remove /dev nodes with subsystem-aware permissions
- Create persistent symlinks in /dev/disk/by-{id,path} and
/dev/input/by-{id,path}, tracked for cleanup on device removal
- Load firmware from /lib/firmware/ via the sysfs loading protocol
- Spawn modprobe for MODALIAS events (async, non-blocking)
- Coldplug support via -c flag (walks /sys/devices to replay events)
- Set dev/* conditions for Finit's service dependency system
The original power_supply monitoring and sys/pwr/ac condition are
preserved.
New files: keventd.h (structures/API), uevent.c (all device logic).
The receive buffer is increased to 8K with a 1MB socket buffer to
reduce event loss during coldplug bursts.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A pass over the whole branch before merge, mostly in libink since
that is the new code and the part exposed to the wire. Grouped here
rather than scattered so the review is easy to read in one place.
libink parser and dispatch:
- Bound reader lengths so a 32-bit size_t can't wrap a wire length
past the guard and read out of bounds. Reachable pre-auth on any
bus, so it matters on the 32-bit targets Finit runs on.
- Drop a peer when a reply send fails instead of limping on with a
half-written frame; a built-in whose send failed used to fall
through and put a second frame on the wire.
initctl:
- Copy a D-Bus error name out of the reply before closing the client;
the reply points into memory the close frees. Both error paths now
share one helper so this can't creep back.
Authorization:
- Take the caller's groups from the kernel (SO_PEERCRED plus
SO_PEERGROUPS) rather than getpwuid()/getgrouplist(), which go
through NSS and can block PID 1 on a slow LDAP or SSSD backend.
The check is now a lookup against the group resolved once at init,
with no NSS and no 256 KiB array on the stack. A caller reaching
us through a broker carries no group set, so system-bus privileged
methods are root-only; the local bus keeps group support. See
libink/README.md for the note on lifting that.
Shutdown:
- Call dbus_exit() from the shutdown path so the server, its peers,
and the socket are let go cleanly. The teardown existed but nobody
called it.
Tests, CI, docs:
- A fuzz target for the message parser, run as a quick sweep in the
suite and properly under libFuzzer in CI, with the corpus carried
between runs. The -as-uid tests drop groups the way a login does
so SO_PEERGROUPS sees the right set, and widen the test socket to
reach the per-method check behind the 0660 gate. Bring the GitHub
actions up to versions that run on Node 24, and tidy a few small
things a /simplify pass turned up.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
__msg_parse() turns bytes off a socket into pointers, before anything
has vouched for the peer, and it is the only place in libink that
does. It had no test of its own beyond whatever the other tests
happened to send it, all of it well-formed.
The target checks the parser's contract, not merely that it survived.
A header field must point into the header field array, and terminate
inside it, and the parse must never claim more bytes than it was
handed. Crash-only would pass a parser that walked into the body and
returned fields from there, since those bytes were handed over too.
The expected bounds are derived from the raw header rather than from
the parser, so the two have to agree independently.
Every input is copied into an allocation sized to it first. Reading
past the end of a roomy buffer stays inside the allocation and the
sanitizer never sees it; against an exact one the same read is a
fault, which is where the sharpest findings come from.
Under libFuzzer it is an ordinary fuzz target and named files replay,
which is how a find gets reproduced. With no arguments it runs a
fixed sweep -- every truncation, every single-byte corruption, every
value of the length that decides where the header ends, and seeded
garbage -- so the suite covers the same contract on every build,
without clang or a corpus in the tree. It takes 40 ms.
CI fuzzes it properly on every pull request, keeps the crashers, and
carries the corpus between runs so it reaches deeper over time than
any single run can. Note that clang links the fuzzer runtime against
the newest GCC tree it finds, so the libstdc++ headers have to match
that one and not the default compiler, which is worth saying since
installing the obvious package leaves you exactly where you started.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A rule carrying sender, destination, or eavesdrop was refused whole,
and a peer whose AddMatch fails gets no signals at all. That is a
poor trade for keys clients attach as a matter of course: better a
filter wider than asked for than a subscription that never happened.
They are accepted and ignored rather than honoured. Widening costs
nothing here since Finit is the only sender on this bus, and what it
emits through the match table is state any peer that got this far may
already read.
argN and argNpath still take the whole rule down. Honouring them
means parsing message bodies, and nothing asks for them yet.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
"Set not yet implemented" reads as a promise. Finit exposes no
writable property and has no use for one: everything a caller might
want to change is a Manager1 method, where the authorization lives.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Only the file header said we refuse 'B' messages, and nobody reads a
header comment when they are looking at why a parse failed. Put it at
the check, with an XXX so it turns up in a grep for known gaps.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A call is parked until the resolver says who sent it, and an outbound
call sits in a pending slot until its reply lands. Neither had a way
to give up. A broker that answers GetConnectionUnixUser slowly, or
not at all, leaves the caller waiting forever and keeps the slot; four
of those and every later privileged call is refused with
LimitsExceeded until Finit restarts.
libink cannot time itself out, it has no event loop, so the deadline
is the embedder's to keep. One sweep per connection covers both, and
the ordering between them stays in the library rather than in each
embedder: calls first, because one timing out usually resolves the
park it was made for, and AccessDenied tells that caller more than a
bare timeout.
The sweep is armed when a resolve is deferred and stops rearming as
soon as nothing is outstanding, so a system that never meets a broker
never wakes up for it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Parked calls and outbound calls awaiting a reply only ever happen on a
connection talking to a broker, but the parked array sat on the server
and the pending array on every connection. A server with no broker
carried 4 KiB of slots it could never fill, and both were reachable
from code paths that have no business in them.
Move both behind one struct, allocated on the first park or call and
freed with the connection. link_server_t goes from 4400 to 168 bytes;
link_connection_t barely moves, its buffers dominate, but an ordinary
peer no longer carries reply-tracking it never uses.
Tokens are now per bus rather than per server, so link_uid_resolved()
takes the connection the answer is about. Every resolver already has
it: it is the first argument to both the resolver and the reply
callback.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
It still described treating every system-bus caller as unprivileged as
the state of things, which stopped being true when Finit learned to ask
the broker who sent a call.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>